diff --git a/D03_PlanAnalysis/D03_PlanAnalysis_Analyze.py b/D03_PlanAnalysis/D03_PlanAnalysis_Analyze.py
index ad1dce02b..337223f36 100644
--- a/D03_PlanAnalysis/D03_PlanAnalysis_Analyze.py
+++ b/D03_PlanAnalysis/D03_PlanAnalysis_Analyze.py
@@ -18,21 +18,31 @@ from __future__ import annotations
 import hashlib
 import json
 import logging
+import math
+import multiprocessing
+from concurrent.futures import ProcessPoolExecutor, as_completed
 from pathlib import Path
 from typing import Any, Callable
 
+import numpy as np
+import shapely
+from shapely.geometry import LineString, MultiLineString
+
 from B05_Profile.B05_Profile_Engine_Grade import (
     legal_grade_limit_pct,
     legal_plan_radius_min_m,
     resolve_design_speed,
 )
 from B05_Profile.B05_Profile_Engine_RouteInitial import (
+    _POOLS,
+    _TERRAINS,
     generate_initial_route,
     load_terrain_and_streams,
 )
 from B05_Profile.B05_Profile_Engine_RouteInitial_Store import request_fingerprint
 from common_util.common_util_json import atomic_write_json
 from common_util.common_util_surface_confirmation import SHEET_ANALYSIS_FILTER
+from config.config_system_terrain import PLAN_ANALYZE_WORKERS
 from D03_PlanAnalysis.D03_PlanAnalysis_Terrain import PLAN_CORRIDOR_M, models_dir_of
 
 logger = logging.getLogger(__name__)
@@ -43,6 +53,9 @@ SELECTION = {"source_filter": SHEET_ANALYSIS_FILTER, "method": "dtm", "smooth":
 # D 등록에 지형 구분 칸 없음 — B `_road_settings` 못 읽을 때 폴백 그대로(가장 완화)
 TERRAIN_TYPE = "special"
 ALGO = 1
+# 계류 거르기 칸 · 지형 상자 여유(m) — `near_streams`
+STREAM_TILE_M = 200.0
+STREAM_PAD_M = 100.0
 
 Progress = Callable[[int, str], None]
 Loader = Callable[..., tuple[Any, list]]
@@ -90,6 +103,38 @@ def _route_view(result: dict[str, Any]) -> dict[str, Any]:
     return {key: result.get(key) for key in keep}
 
 
+def _engine_inputs(criteria: dict[str, Any]) -> dict[str, Any]:
+    return {
+        "weights": None,
+        "max_grade": criteria["max_grade"],
+        "min_radius_m": criteria["min_radius_m"],
+        "stream_offset_m": None,
+        "grade_smooth_m": None,
+        "target_grade": None,
+        "with_profile": False,
+        "design_speed": criteria["design_speed"],
+    }
+
+
+def segment_fingerprint(
+    seg: dict[str, Any],
+    terrain: Any,
+    streams: list,
+    criteria: dict[str, Any],
+    corridor_m: float = PLAN_CORRIDOR_M,
+) -> str:
+    """구간 입력 지문 — 갈래 넷 B 저장 키 · 복도 · 판."""
+    expected = [_point(seg["start"]), _point(seg["end"])]
+    inputs = _engine_inputs(criteria)
+    keys = [
+        request_fingerprint(expected, terrain, streams, mode=mode, **inputs)[0]
+        for mode in PLAN_MODES
+    ]
+    return hashlib.blake2b(
+        repr((keys, float(corridor_m), ALGO)).encode(), digest_size=16
+    ).hexdigest()
+
+
 def analyze_segment(
     project_root: Path,
     seg: dict[str, Any],
@@ -100,23 +145,8 @@ def analyze_segment(
 ) -> dict[str, Any]:
     """연도 구간 하나 · 갈래 넷 — 입력 지문이 저장본과 같으면 그대로(`stored` True)."""
     expected = [_point(seg["start"]), _point(seg["end"])]
-    inputs = {
-        "weights": None,
-        "max_grade": criteria["max_grade"],
-        "min_radius_m": criteria["min_radius_m"],
-        "stream_offset_m": None,
-        "grade_smooth_m": None,
-        "target_grade": None,
-        "with_profile": False,
-        "design_speed": criteria["design_speed"],
-    }
-    keys = [
-        request_fingerprint(expected, terrain, streams, mode=mode, **inputs)[0]
-        for mode in PLAN_MODES
-    ]
-    fingerprint = hashlib.blake2b(
-        repr((keys, float(corridor_m), ALGO)).encode(), digest_size=16
-    ).hexdigest()
+    inputs = _engine_inputs(criteria)
+    fingerprint = segment_fingerprint(seg, terrain, streams, criteria, corridor_m)
     stored = read_segment(project_root, seg)
     if stored is not None and stored.get("fingerprint") == fingerprint:
         return {**stored, "stored": True}
@@ -148,6 +178,60 @@ def analyze_segment(
     return {**result, "stored": False}
 
 
+def near_streams(terrain: Any, streams: list) -> list:
+    """지형 상자 안 어느 점에서든 가장 가까운 계류일 수 있는 선만 — 나머지는 엔진 값에 안 닿음.
+
+    칸마다 닿는 거리 = 칸 가운데 계류 거리 + 칸 반대각(거리는 1-립시츠) — 그 안에 드는 선만 남기면
+    상자 안 점의 계류 거리 · 건너기 · 격자 거리가 통째와 같음(통째 거리 = 추천 시간의 큰 몫).
+    """
+    if not streams:
+        return streams
+    x0, x1 = float(np.min(terrain.x)) - STREAM_PAD_M, float(np.max(terrain.x)) + STREAM_PAD_M
+    y0, y1 = float(np.min(terrain.y)) - STREAM_PAD_M, float(np.max(terrain.y)) + STREAM_PAD_M
+    xs = np.linspace(x0, x1, max(1, math.ceil((x1 - x0) / STREAM_TILE_M)) + 1)
+    ys = np.linspace(y0, y1, max(1, math.ceil((y1 - y0) / STREAM_TILE_M)) + 1)
+    lo_x, lo_y = np.meshgrid(xs[:-1], ys[:-1])
+    hi_x, hi_y = np.meshgrid(xs[1:], ys[1:])
+    tiles = shapely.box(lo_x.ravel(), lo_y.ravel(), hi_x.ravel(), hi_y.ravel())
+    reach = shapely.distance(shapely.centroid(tiles), MultiLineString(streams)) + 0.5 * math.hypot(
+        xs[1] - xs[0], ys[1] - ys[0]
+    )
+    lines = np.array([LineString(line) for line in streams], dtype=object)
+    keep = (shapely.distance(lines[:, None], tiles[None, :]) <= reach[None, :]).any(axis=1)
+    return [line for line, kept in zip(streams, keep) if kept]
+
+
+def _load(project_root: Path, year: int, loader: Loader) -> tuple[Any, list]:
+    terrain, streams = loader(
+        project_root,
+        SELECTION,
+        models_dir_of(project_root, year),
+        derived_dir(project_root) / f"cost_{year}",
+    )
+    return terrain, near_streams(terrain, streams)
+
+
+def _segment_job(
+    project_root: Path,
+    seg: dict[str, Any],
+    criteria: dict[str, Any],
+    corridor_m: float,
+    loader: Loader,
+) -> dict[str, Any]:
+    """연도 구간 하나 — 지형 읽기부터(제 프로세스면 끝날 때 비용면 · 묶음 메모리를 놓음)."""
+    terrain, streams = _load(project_root, int(seg["year"]), loader)
+    return analyze_segment(project_root, seg, terrain, streams, criteria, corridor_m)
+
+
+def _worker_job(*args: Any) -> dict[str, Any]:
+    """나란히 셈 프로세스 — 구간 하나 뒤 B 기억(비용면 · 묶음)을 비워 다음 구간 메모리를 놓음."""
+    try:
+        return _segment_job(*args)
+    finally:
+        _POOLS.clear()
+        _TERRAINS.clear()
+
+
 def analyze_plan(
     project_root: Path,
     segments: list[dict[str, Any]],
@@ -156,23 +240,52 @@ def analyze_plan(
     loader: Loader = load_terrain_and_streams,
     corridor_m: float = PLAN_CORRIDOR_M,
     on_progress: Progress | None = None,
+    workers: int = PLAN_ANALYZE_WORKERS,
 ) -> list[dict[str, Any]]:
-    """연도 구간마다 지형(연도 DEM · 연도 비용면 캐시)을 읽어 `analyze_segment`."""
-    results = []
-    for i, seg in enumerate(segments):
-        year = int(seg["year"])
+    """연도 구간마다(연도 DEM · 연도 비용면 캐시) `analyze_segment`.
+
+    다시 셀 구간이 둘 이상이면 프로세스 `workers` 개로 나란히 — 구간끼리 입력 · 저장본이 따로라
+    결과 같음. `workers` 1 = 이 프로세스(가짜 지형 시험)."""
+
+    def report(percent: int, message: str) -> None:
         if on_progress is not None:
-            on_progress(100 * i // len(segments), f"{year}년 구간 노선 추천 중")
-        terrain, streams = loader(
-            project_root,
-            SELECTION,
-            models_dir_of(project_root, year),
-            derived_dir(project_root) / f"cost_{year}",
-        )
-        results.append(analyze_segment(project_root, seg, terrain, streams, criteria, corridor_m))
-    if on_progress is not None:
-        on_progress(100, "분석 끝")
-    return results
+            on_progress(percent, message)
+
+    results: list[dict[str, Any] | None] = [None] * len(segments)
+    todo = []
+    for i, seg in enumerate(segments):
+        stored = read_segment(project_root, seg)
+        if stored is not None:  # 저장본이 있을 때만 이 프로세스에서 지형을 읽어 지문을 봄
+            terrain, streams = _load(project_root, int(seg["year"]), loader)
+            if stored.get("fingerprint") == segment_fingerprint(
+                seg, terrain, streams, criteria, corridor_m
+            ):
+                results[i] = {**stored, "stored": True}
+                continue
+        todo.append(i)
+
+    count = min(workers, len(todo))
+    if count <= 1:
+        for n, i in enumerate(todo):
+            report(100 * n // len(segments), f"{int(segments[i]['year'])}년 구간 노선 추천 중")
+            results[i] = _segment_job(project_root, segments[i], criteria, corridor_m, loader)
+    else:
+        years = "·".join(str(int(segments[i]["year"])) for i in todo)
+        report(0, f"{years}년 구간 노선 추천 중(나란히 {count})")
+        context = multiprocessing.get_context("spawn")
+        with ProcessPoolExecutor(count, mp_context=context) as pool:
+            futures = {
+                pool.submit(_worker_job, project_root, segments[i], criteria, corridor_m, loader): i
+                for i in todo
+            }
+            for n, future in enumerate(as_completed(futures), 1):
+                i = futures[future]
+                results[i] = future.result()
+                report(
+                    100 * n // len(todo), f"{int(segments[i]['year'])}년 구간 끝 ({n}/{len(todo)})"
+                )
+    report(100, "분석 끝")
+    return results  # type: ignore[return-value]
 
 
 def pick_mode(result: dict[str, Any], wanted: str | None) -> str | None:
diff --git a/E03_StabSection/E03_StabSection_Engine_Prob.py b/E03_StabSection/E03_StabSection_Engine_Prob.py
new file mode 100644
index 000000000..fc6e913c0
--- /dev/null
+++ b/E03_StabSection/E03_StabSection_Engine_Prob.py
@@ -0,0 +1,441 @@
+"""E03 2D 확률 해석 · 민감도 (요구 351 · PLAN 537-1) — Slide2 · SLOPE/W Probabilistic 과 같은 생각.
+
+확률 변수 = 물성(c · φ · γ) · 지하수위(수위선 위아래 이동 m) · r_u — 정규 · 대수정규 · 절단(평균 아래 · 위
+거리 = Slide2 Relative Min · Max · 물리 한계 c ≥ 0 · 0 ≤ φ < 89 · γ > 0 · 0 ≤ r_u < 1 도 절단) · 변수 사이 상관
+(Slide2 c–φ Correlation · 표준정규에서 촐레스키) → 몬테카를로 · 라틴 하이퍼큐브 N 회
+→ 안전율 표본 → 평균 · 표준편차 · Pf = P(F < 1) · β 정규 (μ − 1)/σ · β 대수정규
+ln(μ/√(1+V²))/√ln(1+V²) · 히스토그램 · 누적분포 · P(F < 기준안전율).
+갈래: fixed = 결정론 임계면 고정(절편 기하 한 번 · 물성만 바꿔 묶음 평가 · 빠름 · Slide2 Global Minimum) ·
+search = 표본마다 임계면 다시 찾음(`Engine_Search.analyze` · 느림 · 프로세스 병렬 · Slide2 Overall Slope).
+민감도 = 변수 하나씩 범위(절단 끝 · 없으면 ±3σ) 50 등분 훑기(나머지 평균 · 임계면 고정 · Slide2 Sensitivity 와 같음)
+· 토네이도 = 끝값 F 폭 순.
+γ 변수 = γt 표본 · γsat 는 같은 차만큼 옮김(γsat − γt 그대로).
+search 일꾼 = `python -m E03_StabSection.E03_StabSection_Engine_Prob <job.json>`(서버 프로세스에서 풀을 열지 않음 ·
+`Worker` 머리글) · 한 줄마다 {"done": n} · 끝에 job["out"] 파일.
+NumPy · SciPy(stats) 만 · 계산 자리 = ③ 서버 단독.
+"""
+
+from __future__ import annotations
+
+import copy
+import json
+import math
+import sys
+import time
+from pathlib import Path
+
+import numpy as np
+from scipy.stats import norm
+
+from E03_StabSection import E03_StabSection_Engine_LEM as lem
+from E03_StabSection.E03_StabSection_Engine_Search import analyze
+from E03_StabSection.E03_StabSection_Engine_Station import MIN_DEPTH_M
+from E03_StabSection.E03_StabSection_Engine_Surface import (
+    circle_slices,
+    evaluate,
+    poly_slices,
+    three_point_center,
+)
+
+PROPS = ("c", "phi", "gamma", "water", "ru")
+#: 물리 한계(절단) — 지하수 이동은 한계 없음
+BOUNDS = {"c": (0.0, math.inf), "phi": (0.0, 89.0), "gamma": (1e-3, math.inf), "ru": (0.0, 0.99)}
+#: 절단 없을 때 민감도 훑기 범위(±σ 배)
+SWEEP_SIGMA = 3.0
+SWEEP_POINTS = 51
+HIST_BINS = 30
+CDF_POINTS = 101
+N_SLICES = 50
+
+
+# ---- 분포 · 표본 --------------------------------------------------------------
+def _bounds(v: dict) -> tuple[float, float]:
+    """절단 구간 [a, b] — 평균 ∓ 아래 · 위 거리 ∩ 물리 한계."""
+    lo_phys, hi_phys = BOUNDS.get(v["prop"], (-math.inf, math.inf))
+    mean = float(v["mean"])
+    a = mean - float(v["lo"]) if v.get("lo") is not None else -math.inf
+    b = mean + float(v["hi"]) if v.get("hi") is not None else math.inf
+    return max(a, lo_phys), min(b, hi_phys)
+
+
+def _lognormal(v: dict) -> tuple[float, float]:
+    """대수정규 (μ_ln, σ_ln) — 평균 · 표준편차 그대로 맞춤."""
+    mean, std = float(v["mean"]), float(v["std"])
+    s2 = math.log1p((std / mean) ** 2)
+    return math.log(mean) - 0.5 * s2, math.sqrt(s2)
+
+
+def ppf(v: dict, u: np.ndarray) -> np.ndarray:
+    """균등 u(0~1) → 변수 값(절단 분포 역함수)."""
+    u = np.asarray(u, float)
+    mean, std = float(v["mean"]), float(v.get("std") or 0.0)
+    if std <= 0:
+        return np.full(u.shape, mean)
+    a, b = _bounds(v)
+    if v.get("dist") == "lognormal" and v["prop"] != "water":
+        if mean <= 0:
+            raise ValueError(f"대수정규는 평균 > 0: {v}")
+        mu, s = _lognormal(v)
+        za = (math.log(a) - mu) / s if a > 0 else -math.inf
+        zb = (math.log(b) - mu) / s if b < math.inf else math.inf
+        pa, pb = norm.cdf(za), norm.cdf(zb)
+        return np.exp(mu + s * norm.ppf(pa + u * (pb - pa)))
+    pa, pb = norm.cdf((a - mean) / std), norm.cdf((b - mean) / std)
+    return mean + std * norm.ppf(pa + u * (pb - pa))
+
+
+def sample(
+    variables: list[dict],
+    n: int,
+    *,
+    sampling: str = "lhs",
+    seed: int = 0,
+    correlations: list | None = None,
+) -> np.ndarray:
+    """표본 (n, 변수 수) — lhs = 라틴 하이퍼큐브(변수마다 n 칸 하나씩 · 칸 차례 섞음) · mc = 단순 무작위.
+
+    correlations = [(i, j, ρ), …] — 표준정규 칸에서 촐레스키로 섞은 뒤 분포로(LHS 층은 섞인 변수만 흐려짐)."""
+    rng = np.random.default_rng(int(seed))
+    k = len(variables)
+    if sampling == "lhs":
+        order = np.argsort(rng.random((n, k)), axis=0)
+        u = (order + rng.random((n, k))) / n
+    else:
+        u = rng.random((n, k))
+    u = np.clip(u, 1e-12, 1 - 1e-12)
+    if correlations:
+        corr = np.eye(k)
+        for i, j, rho in correlations:
+            corr[i, j] = corr[j, i] = float(rho)
+        u = norm.cdf(norm.ppf(u) @ np.linalg.cholesky(corr).T)
+        u = np.clip(u, 1e-12, 1 - 1e-12)
+    return np.column_stack([ppf(v, u[:, j]) for j, v in enumerate(variables)]) if k else u
+
+
+def sweep_range(v: dict) -> tuple[float, float]:
+    """민감도 훑기 범위 — 절단 끝 · 없으면 평균 ±3σ(물리 한계 안)."""
+    mean, std = float(v["mean"]), float(v.get("std") or 0.0)
+    a, b = _bounds(v)
+    lo = a if math.isfinite(a) and v.get("lo") is not None else mean - SWEEP_SIGMA * std
+    hi = b if math.isfinite(b) and v.get("hi") is not None else mean + SWEEP_SIGMA * std
+    lo_phys, hi_phys = BOUNDS.get(v["prop"], (-math.inf, math.inf))
+    return max(lo, lo_phys), min(hi, hi_phys)
+
+
+# ---- 통계 -------------------------------------------------------------------
+def statistics(F: np.ndarray, required: float | None = None, bins: int = HIST_BINS) -> dict:
+    """안전율 표본 → 평균 · 표준편차 · Pf · β 정규 · 대수정규 · 히스토그램 · 누적분포(해 없는 표본 = 셈 밖)."""
+    F = np.asarray(F, float)
+    ok = F[np.isfinite(F)]
+    out = {"n": int(len(F)), "n_valid": int(len(ok))}
+    if len(ok) < 2:
+        return {**out, "mean": None, "std": None, "pf": None, "beta_normal": None}
+    mean, std = float(ok.mean()), float(ok.std(ddof=1))
+    V = std / mean if mean > 0 else math.nan
+    beta_ln = (
+        math.log(mean / math.sqrt(1 + V * V)) / math.sqrt(math.log1p(V * V))
+        if mean > 0 and V > 0
+        else None
+    )
+    counts, edges = np.histogram(ok, bins=bins)
+    p = np.linspace(0.0, 1.0, CDF_POINTS)
+    out.update(
+        mean=mean,
+        std=std,
+        cov=V if math.isfinite(V) else None,
+        min=float(ok.min()),
+        max=float(ok.max()),
+        pf=float(np.mean(ok < 1.0)),
+        n_fail=int(np.sum(ok < 1.0)),
+        beta_normal=(mean - 1.0) / std if std > 0 else None,
+        beta_lognormal=beta_ln,
+        required=required,
+        p_below_required=float(np.mean(ok < required)) if required else None,
+        histogram={"edges": edges.tolist(), "counts": counts.tolist()},
+        cdf={"p": p.tolist(), "F": np.quantile(ok, p).tolist()},
+    )
+    return out
+
+
+# ---- 변수 → 모형 ---------------------------------------------------------------
+def perturb(model: dict, variables: list[dict], row) -> dict:
+    """모형 사본 — 표본 한 줄을 물성 · 지하수위에 넣음(search 갈래)."""
+    out = dict(model)
+    mats = {m["id"]: dict(m) for m in model["materials"]}
+    dz = 0.0
+    for v, x in zip(variables, row):
+        if v["prop"] == "water":
+            dz += float(x)
+            continue
+        if v["prop"] == "ru":
+            gw = model.get("groundwater") or {}
+            if gw.get("mode") == "ru":
+                out["groundwater"] = {**gw, "ru": float(x)}
+            continue
+        m = mats[v["material"]]
+        if v["prop"] == "gamma":
+            gt = float(m["gamma_t"])
+            m["gamma_sat"] = float(m.get("gamma_sat", gt)) + float(x) - gt
+            m["gamma_t"] = float(x)
+        else:
+            m[v["prop"]] = float(x)
+    out["materials"] = list(mats.values())
+    gw = out.get("groundwater") or {}
+    if dz and gw.get("mode") == "line":
+        out["groundwater"] = {**gw, "line": [[x, z + dz] for x, z in gw["line"]]}
+    return out
+
+
+class Fixed:
+    """임계면 고정 — 모형 · 면 하나를 한 번 준비해 표본마다 물성 칸만 갈아 끼우고 절편 묶음으로 평가."""
+
+    def __init__(self, model: dict, surface: dict, direction: int, n_slices: int = N_SLICES):
+        self.prep = lem.Prepared(model, direction)
+        self.n = n_slices
+        s = self.prep.direction
+        self.index = {m["id"]: i for i, m in enumerate(self.prep.materials)}
+        if surface["kind"] == "circle":
+            xs = sorted((s * surface["entry"][0], s * surface["exit"][0]))
+            self.geom = (
+                "circle",
+                s * surface["center"][0],
+                surface["center"][1],
+                surface["radius"],
+            )
+            self.geom += tuple(xs)
+        else:
+            pts = sorted((s * x, z) for x, z in surface["points"])
+            self.geom = ("polyline", np.array([p[0] for p in pts]), np.array([p[1] for p in pts]))
+
+    def _prep(self, variables: list[dict], row) -> lem.Prepared:
+        base = self.prep
+        p = copy.copy(base)
+        gt, gs = base.gamma_t.copy(), base.gamma_sat.copy()
+        c, tp = base.c.copy(), base.tanphi.copy()
+        dz = 0.0
+        for v, x in zip(variables, row):
+            if v["prop"] == "water":
+                dz += float(x)
+                continue
+            if v["prop"] == "ru":
+                if base.gw_mode == "ru":
+                    p.ru = float(x)
+                continue
+            i = self.index[v["material"]]
+            if v["prop"] == "c":
+                c[i] = x
+            elif v["prop"] == "phi":
+                tp[i] = math.tan(math.radians(x))
+            else:
+                gs[i] += x - gt[i]
+                gt[i] = x
+        p.gamma_t, p.gamma_sat, p.c, p.tanphi = gt, gs, c, tp
+        p.table = base.table.copy()
+        p.table[:, 4, :] = gt[base.mat]
+        p.table[:, 5, :] = gs[base.mat] - gt[base.mat]
+        if dz and base.gw_mode == "line":
+            p.water_z = base.water_z + dz
+        return p
+
+    def _slices(self, prep: lem.Prepared) -> tuple[dict, np.ndarray]:
+        g = self.geom
+        if g[0] == "circle":
+            _, xc, zc, R, xE, xX = g
+            sl = circle_slices(prep, [xc], [zc], [R], [xE], [xX], self.n)
+            return sl, np.array([[xc, zc]])
+        vx, vz = g[1][None, :], g[2][None, :]
+        return poly_slices(prep, vx, vz, self.n), three_point_center(vx, vz)
+
+    def F(self, variables: list[dict], X: np.ndarray, method: str) -> np.ndarray:
+        """표본 줄마다 안전율(해 없음 = nan)."""
+        X = np.atleast_2d(np.asarray(X, float))
+        parts = [self._slices(self._prep(variables, row)) for row in X]
+        sl = _stack([p[0] for p in parts])
+        center = np.vstack([p[1] for p in parts])
+        mp_seed = None
+        if method == "morgenstern_price":  # describe 와 같은 출발 참고값(같은 면 Spencer)
+            mp_seed = evaluate(sl, center, ("spencer",))["spencer"]["F"]
+            sl.pop("_pre", None)
+        F = evaluate(sl, center, (method,), mp_seed=mp_seed)[method]["F"]
+        return np.where(sl["valid"], F, np.nan)
+
+
+def _stack(parts: list[dict]) -> dict:
+    """한 줄짜리 절편 묶음들 → (S, …) 하나(배열 칸만 잇고 스칼라는 그대로)."""
+    out = {}
+    for key, value in parts[0].items():
+        if key == "_pre":
+            continue
+        if isinstance(value, np.ndarray) and value.ndim >= 1 and value.shape[0] == 1:
+            out[key] = np.concatenate([p[key] for p in parts])
+        else:
+            out[key] = value
+    return out
+
+
+def _chunks(n: int, size: int):
+    for i in range(0, n, size):
+        yield slice(i, min(i + size, n))
+
+
+def sensitivity(
+    fixed: Fixed, variables: list[dict], method: str, points: int = SWEEP_POINTS
+) -> dict:
+    """변수 하나씩 범위 훑기(나머지 평균) · 토네이도 = 끝값 F 폭 큰 순."""
+    means = np.array([float(v["mean"]) for v in variables])
+    rows, spans = [], []
+    for j, v in enumerate(variables):
+        lo, hi = sweep_range(v)
+        xs = np.linspace(lo, hi, points) if hi > lo else np.full(points, means[j])
+        block = np.tile(means, (points, 1))
+        block[:, j] = xs
+        rows.append(block)
+        spans.append(xs)
+    F0 = float(fixed.F(variables, means[None, :], method)[0])
+    if not rows:
+        return {"F_mean": F0, "curves": [], "tornado": []}
+    F = fixed.F(variables, np.vstack(rows), method).reshape(len(variables), points)
+    num = lambda a: [float(x) if np.isfinite(x) else None for x in a]  # noqa: E731
+    curves = []
+    for v, xs, f in zip(variables, spans, F):
+        ends = [f[0], f[-1]]
+        swing = abs(ends[1] - ends[0]) if all(np.isfinite(ends)) else None
+        curves.append(
+            {
+                "name": v["name"],
+                "x": num(xs),
+                "F": num(f),
+                "F_lo": num([ends[0]])[0],
+                "F_hi": num([ends[1]])[0],
+                "swing": swing,
+            }
+        )
+    tornado = sorted(
+        (c for c in curves if c["swing"] is not None), key=lambda c: c["swing"], reverse=True
+    )
+    return {"F_mean": F0 if math.isfinite(F0) else None, "curves": curves, "tornado": [
+        {"name": c["name"], "F_lo": c["F_lo"], "F_hi": c["F_hi"], "swing": c["swing"]}
+        for c in tornado
+    ]}  # fmt: skip
+
+
+# ---- 매회 탐색(search) ------------------------------------------------------------
+_CTX: dict = {}
+
+
+def _init(ctx: dict) -> None:
+    _CTX.clear()
+    _CTX.update(ctx)
+
+
+def _search_one(row) -> float:
+    c = _CTX
+    m = c["method"]
+    methods = tuple(dict.fromkeys(("bishop", m)))
+    res = analyze(
+        perturb(c["model"], c["variables"], row),
+        direction=c["direction"],
+        limits=c["limits"],
+        methods=methods,
+        noncircular=m not in ("ordinary", "bishop"),  # 비원호 Fellenius · Bishop = 참고 칸
+        min_depth=MIN_DEPTH_M,
+    )
+    best = res["methods"][m]["min"]
+    return float(best["F"]) if best and best.get("F") is not None else math.nan
+
+
+def search_F(ctx: dict, X: np.ndarray, workers: int = 1, emit=None) -> np.ndarray:
+    """표본마다 임계면 다시 찾음 — workers > 1 = 프로세스 풀(서버 밖 일꾼에서만)."""
+    out = np.full(len(X), np.nan)
+    if workers <= 1:
+        _init(ctx)
+        for i, row in enumerate(X):
+            out[i] = _search_one(row)
+            emit and emit(i + 1)
+        return out
+    from concurrent.futures import ProcessPoolExecutor, as_completed
+
+    with ProcessPoolExecutor(max_workers=workers, initializer=_init, initargs=(ctx,)) as pool:
+        futures = {pool.submit(_search_one, row): i for i, row in enumerate(X)}
+        for done, future in enumerate(as_completed(futures), 1):
+            out[futures[future]] = future.result()
+            emit and emit(done)
+    return out
+
+
+# ---- 한 번 해석 ---------------------------------------------------------------
+def run(
+    model: dict,
+    *,
+    surface: dict,
+    direction: int,
+    limits=None,
+    variables: list[dict],
+    correlations: list | None = None,
+    n: int = 1000,
+    sampling: str = "lhs",
+    seed: int = 0,
+    mode: str = "fixed",
+    method: str = "bishop",
+    required: float | None = None,
+    workers: int = 1,
+    emit=None,
+) -> dict:
+    """확률 해석 + 민감도 한 벌. emit(끝난 표본 수) = 진행률."""
+    t0 = time.perf_counter()
+    X = sample(variables, n, sampling=sampling, seed=seed, correlations=correlations)
+    fixed = Fixed(model, surface, direction)
+    t1 = time.perf_counter()
+    if mode == "search":
+        ctx = {
+            "model": model,
+            "variables": variables,
+            "direction": direction,
+            "limits": limits,
+            "method": method,
+        }
+        F = search_F(ctx, X, workers, emit)
+    else:
+        F = np.full(n, np.nan)
+        for part in _chunks(n, 250):
+            F[part] = fixed.F(variables, X[part], method)
+            emit and emit(part.stop)
+    t2 = time.perf_counter()
+    sens = sensitivity(fixed, variables, method)
+    t3 = time.perf_counter()
+    realized = [
+        {"name": v["name"], "mean": float(X[:, j].mean()), "std": float(X[:, j].std(ddof=1))}
+        for j, v in enumerate(variables)
+    ]
+    return {
+        "mode": mode,
+        "method": method,
+        "n": n,
+        "sampling": sampling,
+        "seed": seed,
+        "variables": variables,
+        "correlations": correlations or [],
+        "realized": realized,
+        "stats": statistics(F, required),
+        "F": [round(float(f), 4) if np.isfinite(f) else None for f in F],
+        "sensitivity": sens,
+        "elapsed_s": {
+            "prepare": round(t1 - t0, 3),
+            "samples": round(t2 - t1, 3),
+            "sensitivity": round(t3 - t2, 3),
+            "total": round(t3 - t0, 3),
+        },
+    }
+
+
+def main() -> None:
+    """일꾼 — job = run() 인자 + out(결과 파일) · 한 줄마다 진행."""
+    path = Path(sys.argv[1])
+    job = json.loads(path.read_text(encoding="utf-8"))
+    out = Path(job.pop("out"))
+    result = run(**job, emit=lambda k: print(json.dumps({"done": k}), flush=True))
+    out.write_text(json.dumps(result, ensure_ascii=False), encoding="utf-8")
+
+
+if __name__ == "__main__":
+    main()
diff --git a/E03_StabSection/E03_StabSection_Router_Prob.py b/E03_StabSection/E03_StabSection_Router_Prob.py
new file mode 100644
index 000000000..bc8e4fcfe
--- /dev/null
+++ b/E03_StabSection/E03_StabSection_Router_Prob.py
@@ -0,0 +1,400 @@
+"""E03 확률 해석 · 민감도 API (요구 351 · PLAN 537-1) — 대표 단면만.
+
+- `GET /api/projects/{id}/stab/prob/setup/{station_key}?side=&condition=` — 그 단면 지층 물성(E02 값 = 평균) ·
+  기본값 후보(`prob_defaults.json` · 근거 · 사용자 확정 전) · 지하수 갈래 · 기준안전율 · 마지막 결과
+- `POST …/stab/prob/{station_key}` — 분포 · 표본 수 · 갈래 → 뒤에서(진행 `GET …/stab/prob/progress`) ·
+  같은 지문 = 저장본(fresh)
+- `GET …/stab/prob` — 지금 2D 결과에 붙은 기록 목록(측점 · 비탈 · 조건마다 가장 최근 · E05 「확률해석」 시트)
+
+평균 = E02 지반 조건(작업본) 값 · 바꾸려면 E02 에서. 임계면 = 지금 2D 결과의 그 방법 최소 면.
+결과 = 파생 저장본 `_derived/stab2d_prob/<지문>` · 지문 = 원본 측점 지문 + 요청 + 코드.
+엔진 = `Engine_Prob` — 임계면 고정 = 서버 스레드(빠름) · 매회 탐색 = 일꾼 프로세스(풀 · 느림).
+계산 자리 = ③ 서버 단독.
+"""
+
+from __future__ import annotations
+
+import asyncio
+import hashlib
+import json
+import logging
+import os
+import subprocess
+import sys
+import tempfile
+import time
+from functools import lru_cache
+from pathlib import Path
+from typing import Any, Literal
+from uuid import UUID
+
+from fastapi import APIRouter, HTTPException
+from pydantic import BaseModel, Field
+
+from common_util.common_util_derived_store import (
+    derived_dir,
+    derived_key,
+    read_derived,
+    write_derived,
+)
+from E02_StabGround import E02_StabGround_Router as e02
+from E03_StabSection import E03_StabSection_Engine_Prob as prob
+from E03_StabSection import E03_StabSection_Router as e03
+from E03_StabSection import E03_StabSection_Router_WhatIf as whatif
+from E03_StabSection import E03_StabSection_View as view
+from E03_StabSection.E03_StabSection_Engine_LEM import WALL_ID
+from E03_StabSection.E03_StabSection_Engine_Model import section_model_from_b06
+from E03_StabSection.E03_StabSection_Types import groundwater as norm_groundwater
+
+logger = logging.getLogger(__name__)
+router = APIRouter(prefix="/api/projects", tags=["E03 Stab Probabilistic"])
+KIND = "stab2d_prob"
+KEEP = 200
+DEFAULTS_PATH = (
+    Path(__file__).resolve().parents[1] / "resources/master_template/stability/prob_defaults.json"
+)
+#: 매회 탐색 표본 수 상한(측점 하나 탐색 수 초 × N)
+SEARCH_MAX_N = 2000
+#: 일꾼 프로세스 수 — None = 코어 − 2 · 1 = 서버 안에서 차례로(시험)
+WORKERS: int | None = None
+
+_JOBS: dict[str, dict[str, Any]] = {}
+_TASKS: set[asyncio.Task] = set()
+
+
+@lru_cache(maxsize=1)
+def code_hash() -> str:
+    digest = hashlib.sha1(e03.engine_hash().encode())
+    for name in ("E03_StabSection_Engine_Prob.py", Path(__file__).name):
+        digest.update((Path(__file__).parent / name).read_bytes())
+    return digest.hexdigest()[:12]
+
+
+def load_defaults() -> dict[str, Any]:
+    return json.loads(DEFAULTS_PATH.read_text(encoding="utf-8"))
+
+
+class VarIn(BaseModel):
+    """확률 변수 하나 — c · φ · γ = 변동계수 · 지하수위 이동(m) · r_u = 표준편차 · 절단 = 평균 ± k·σ."""
+
+    material: str | None = None
+    prop: Literal["c", "phi", "gamma", "water", "ru"]
+    dist: Literal["normal", "lognormal"] = "normal"
+    cov: float | None = Field(None, ge=0.0, le=2.0)
+    std: float | None = Field(None, ge=0.0, le=20.0)
+    trunc_sigma: float | None = Field(3.0, gt=0.0, le=10.0)
+
+
+class ProbRequest(BaseModel):
+    side: Literal["left", "right"]
+    condition: Literal["dry", "wet", "seismic"]
+    method: Literal["ordinary", "bishop", "janbu", "spencer", "morgenstern_price"] = "bishop"
+    mode: Literal["fixed", "search"] = "fixed"
+    n: int = Field(1000, ge=50, le=20000)
+    sampling: Literal["lhs", "mc"] = "lhs"
+    seed: int = Field(0, ge=0, le=2**31 - 1)
+    variables: list[VarIn] = Field(default_factory=list, max_length=40)
+    rho_c_phi: float = Field(0.0, ge=-0.95, le=0.95)
+
+
+def _model(ctx: dict[str, Any], condition: str) -> dict[str, Any]:
+    """그 조건 모형 — `analyze_station` 과 같은 입력(지층 · 지반선 · 벽 · 하중 · E02 지하수 갈래)."""
+    kw = ctx["kw"]
+    model = section_model_from_b06(
+        ctx["cross"],
+        ctx["design"],
+        materials=kw["materials"],
+        layers=kw["layers"],
+        ground=kw["ground"],
+        wet=condition == "wet",
+        seismic_kh=kw["seismic_kh"] if condition == "seismic" else 0.0,
+        road_surcharge_kpa=kw["road_surcharge_kpa"],
+        walls=kw.get("walls"),
+    )
+    model["groundwater"] = norm_groundwater(kw["groundwater_for"](condition, model))
+    return model
+
+
+def _slot(root: Path, station_key: str, side: str, condition: str, method: str) -> dict:
+    """측점 · 비탈 · 조건 — 대표 단면만 · 모형 · 임계면 · 기준안전율."""
+    ctx = whatif._context(root, station_key)
+    rows = [dict(r) for r in e03._latest_summary(root, ctx["plan"]["fp"])[0]["stations"]]
+    view.mark_representative(rows, e03._overrides(root))
+    mark = next((r.get("representative") or {} for r in rows if r["key"] == station_key), {})
+    if not mark.get("on"):
+        raise HTTPException(
+            status_code=422, detail="대표 단면만 확률 해석합니다 — 먼저 대표 단면으로."
+        )
+    if condition not in ctx["kw"]["conditions"]:
+        raise HTTPException(status_code=422, detail="그 조건은 해석하지 않았습니다.")
+    case = next(
+        (
+            c
+            for c in ctx["value"]["out"]["cases"]
+            if c["condition"] == condition and c["side"] == side
+        ),
+        None,
+    )
+    if case is None:
+        raise HTTPException(status_code=422, detail="그 쪽 비탈이 없습니다.")
+    methods = case["result"]["methods"]
+    surface = (methods.get(method) or {}).get("min") or None
+    model = _model(ctx, condition)
+    sd = next(s for s in model["sides"] if s["side"] == side)
+    return {
+        "ctx": ctx,
+        "model": model,
+        "side": sd,
+        "role": case["role"],
+        "surface": surface,
+        "methods": methods,
+        "required": view.required_factor(ctx["plan"]["factors"], condition, case["role"]),
+    }
+
+
+def _materials(model: dict[str, Any]) -> list[dict[str, Any]]:
+    """단면에 쓰인 물성(벽 뺌 · 지층 차례)."""
+    used = list(dict.fromkeys(r["material_id"] for r in model["regions"]))
+    by = {m["id"]: m for m in model["materials"]}
+    return [by[i] for i in used if i != WALL_ID and i in by]
+
+
+def _variables(req: ProbRequest, model: dict[str, Any]) -> tuple[list[dict], list]:
+    """요청 → 엔진 변수(평균 = 모형 값 · σ 0 은 뺌) · c–φ 상관 짝."""
+    mats = {m["id"]: m for m in _materials(model)}
+    gw = model.get("groundwater") or {}
+    out = []
+    for v in req.variables:
+        if v.prop in ("water", "ru"):
+            need = "line" if v.prop == "water" else "ru"
+            if gw.get("mode") != need:
+                raise HTTPException(
+                    status_code=422,
+                    detail=f"{'지하수위' if v.prop == 'water' else 'r_u'} 변수는 그 조건 지하수 갈래가 "
+                    f"{'수위선' if need == 'line' else 'r_u'} 일 때만.",
+                )
+            mean, std = (0.0, v.std) if v.prop == "water" else (float(gw["ru"]), v.std)
+            name = "지하수위" if v.prop == "water" else "r_u"
+        else:
+            m = mats.get(v.material or "")
+            if m is None:
+                raise HTTPException(status_code=422, detail=f"단면에 없는 물성: {v.material}")
+            mean = float(m["gamma_t"] if v.prop == "gamma" else m[v.prop])
+            std = mean * float(v.cov) if v.cov is not None else v.std
+            name = f"{m.get('name') or m['id']} {dict(c='c', phi='φ', gamma='γ')[v.prop]}"
+        if not std:
+            continue
+        if v.dist == "lognormal" and (mean <= 0 or v.prop in ("water",)):
+            raise HTTPException(status_code=422, detail=f"대수정규는 평균 > 0 만: {name}")
+        k = v.trunc_sigma
+        out.append(
+            {
+                "name": name,
+                "material": v.material if v.prop not in ("water", "ru") else None,
+                "prop": v.prop,
+                "dist": v.dist,
+                "mean": mean,
+                "std": float(std),
+                "lo": None if k is None else k * float(std),
+                "hi": None if k is None else k * float(std),
+            }
+        )
+    pairs = []
+    if req.rho_c_phi:
+        for i, a in enumerate(out):
+            for j, b in enumerate(out):
+                if a["prop"] == "c" and b["prop"] == "phi" and a["material"] == b["material"]:
+                    pairs.append((i, j, req.rho_c_phi))
+    return out, pairs
+
+
+def _run_search(job: dict[str, Any], emit: Any) -> dict[str, Any]:
+    """매회 탐색 일꾼(`python -m …Engine_Prob job.json`) — 한 줄마다 emit · 실패면 RuntimeError."""
+    with tempfile.TemporaryDirectory(prefix="aislo_e03p_") as folder:
+        path, out = Path(folder) / "job.json", Path(folder) / "out.json"
+        path.write_text(json.dumps({**job, "out": str(out)}, ensure_ascii=False), encoding="utf-8")
+        log = Path(folder) / "stderr.txt"
+        with log.open("w", encoding="utf-8") as errors:
+            process = subprocess.Popen(
+                [sys.executable, "-m", "E03_StabSection.E03_StabSection_Engine_Prob", str(path)],
+                cwd=e03.PROJECT_ROOT,
+                stdout=subprocess.PIPE,
+                stderr=errors,
+                text=True,
+                encoding="utf-8",
+            )
+            for line in process.stdout:
+                if line.startswith("{"):
+                    emit(json.loads(line).get("done"))
+            code = process.wait()
+        if code != 0:
+            tail = log.read_text(encoding="utf-8", errors="replace").strip().splitlines()[-1:]
+            raise RuntimeError(f"확률 해석 일꾼 실패: {tail}")
+        return json.loads(out.read_text(encoding="utf-8"))
+
+
+def run_prob(root: Path, station_key: str, req: ProbRequest, report: Any) -> dict[str, Any]:
+    slot = _slot(root, station_key, req.side, req.condition, req.method)
+    if slot["surface"] is None:
+        raise HTTPException(status_code=422, detail="그 방법 임계면이 없습니다(해 없음).")
+    if req.mode == "search" and req.n > SEARCH_MAX_N:
+        raise HTTPException(status_code=422, detail=f"매회 탐색은 표본 {SEARCH_MAX_N} 회까지.")
+    variables, pairs = _variables(req, slot["model"])
+    if not variables:
+        raise HTTPException(status_code=422, detail="확률 변수가 없습니다(변동계수 · 표준편차 0).")
+    row = slot["ctx"]["row"]
+    key = derived_key(code_hash(), row["fp"], req.model_dump())
+    cached = read_derived(root, KIND, key)
+    if cached:
+        return cached["value"]
+    job = {
+        "model": slot["model"],
+        "surface": slot["surface"],
+        "direction": slot["side"]["direction"],
+        "limits": list(slot["side"]["limits"]),
+        "variables": variables,
+        "correlations": pairs,
+        "n": req.n,
+        "sampling": req.sampling,
+        "seed": req.seed,
+        "mode": req.mode,
+        "method": req.method,
+        "required": slot["required"],
+    }
+
+    def emit(done: int | None) -> None:
+        if done:
+            report(2 + 93 * int(done) // req.n, f"표본 {done} / {req.n}")
+
+    report(1, "시작")
+    workers = WORKERS or max(1, (os.cpu_count() or 2) - 2)
+    if req.mode == "search" and workers > 1:
+        result = _run_search({**job, "workers": workers}, emit)
+    else:
+        result = prob.run(**job, emit=emit)
+    record = {
+        "station_key": station_key,
+        "chainage_m": row["chainage_m"],
+        "label": row.get("label"),
+        "base_fp": row["fp"],
+        "side": req.side,
+        "role": slot["role"],
+        "condition": req.condition,
+        "required": slot["required"],
+        "fs_deterministic": slot["surface"]["F"],
+        "surface_kind": slot["surface"]["kind"],
+        "request": req.model_dump(),
+        "result": result,
+        "computed_at": time.strftime("%Y-%m-%dT%H:%M:%S"),
+    }
+    write_derived(root, KIND, key, record, keep=KEEP)
+    return record
+
+
+def records(root: Path) -> list[dict[str, Any]]:
+    """가장 최근 2D 요약 측점에 붙은 기록 — (측점 · 비탈 · 조건)마다 가장 최근 하나 · 누가거리 순."""
+    summary, _ = e03._latest_summary(root, "")
+    fps = {r["fp"] for r in (summary or {}).get("stations") or []}
+    latest: dict[tuple, dict[str, Any]] = {}
+    for path in derived_dir(root, KIND).glob("*.json"):
+        value = (read_derived(root, KIND, path.stem) or {}).get("value")
+        if not value or value.get("base_fp") not in fps:
+            continue
+        slot = (value["station_key"], value["side"], value["condition"])
+        if slot not in latest or value["computed_at"] > latest[slot]["computed_at"]:
+            latest[slot] = value
+    return sorted(
+        latest.values(), key=lambda v: (float(v["chainage_m"]), v["side"], v["condition"])
+    )
+
+
+def setup(root: Path, station_key: str, side: str, condition: str) -> dict[str, Any]:
+    slot = _slot(root, station_key, side, condition, "bishop")
+    gw = slot["model"].get("groundwater") or {}
+    last = next(
+        (
+            r
+            for r in records(root)
+            if (r["station_key"], r["side"], r["condition"]) == (station_key, side, condition)
+        ),
+        None,
+    )
+    return {
+        "station_key": station_key,
+        "side": side,
+        "condition": condition,
+        "role": slot["role"],
+        "required": slot["required"],
+        "fs_deterministic": {m: (v.get("min") or {}).get("F") for m, v in slot["methods"].items()},
+        "materials": [
+            {k: m.get(k) for k in ("id", "name", "c", "phi", "gamma_t")}
+            for m in _materials(slot["model"])
+        ],
+        "groundwater": {"mode": gw.get("mode", "none"), "ru": gw.get("ru")},
+        "defaults": load_defaults(),
+        "search_max_n": SEARCH_MAX_N,
+        "last": last,
+    }
+
+
+@router.get("/{project_id}/stab/prob/setup/{station_key}", response_model=None)
+async def get_setup(project_id: UUID, station_key: str, side: str, condition: str) -> dict:
+    root, _ = await e02._project(project_id)
+    return await asyncio.to_thread(setup, root, station_key, side, condition)
+
+
+@router.post("/{project_id}/stab/prob/{station_key}", response_model=None)
+async def post_prob(project_id: UUID, station_key: str, payload: ProbRequest) -> dict:
+    """뒤에서 — 같은 지문이면 바로 저장본(fresh) · 진행 = `GET …/stab/prob/progress`."""
+    root, _ = await e02._project(project_id)
+    key = str(project_id)
+    if (_JOBS.get(key) or {}).get("running"):
+        raise HTTPException(status_code=409, detail="이미 확률 해석 중입니다.")
+    # 입력 검사 · 저장본 확인은 바로(422 를 그 자리에서 돌려줌)
+    slot = await asyncio.to_thread(
+        _slot, root, station_key, payload.side, payload.condition, payload.method
+    )
+    _variables(payload, slot["model"])
+    fp = derived_key(code_hash(), slot["ctx"]["row"]["fp"], payload.model_dump())
+    saved = read_derived(root, KIND, fp)
+    if saved:
+        return {"status": "fresh", "record": saved["value"]}
+    job = {
+        "running": True,
+        "percent": 0,
+        "message": "시작",
+        "error": None,
+        "station_key": station_key,
+    }
+    _JOBS[key] = job
+
+    def report(percent: int, message: str) -> None:
+        job.update(percent=int(percent), message=message)
+
+    async def work() -> None:
+        try:
+            await asyncio.to_thread(run_prob, root, station_key, payload, report)
+            job.update(percent=100, message="끝")
+        except Exception as error:  # noqa: BLE001 — 진행률에 까닭을 그대로 올림
+            logger.exception("E03 확률 해석 실패: project_id=%s", project_id)
+            job["error"] = str(getattr(error, "detail", None) or error)
+        finally:
+            job["running"] = False
+
+    task = asyncio.create_task(work(), name=f"E03 prob {project_id}")
+    _TASKS.add(task)
+    task.add_done_callback(_TASKS.discard)
+    return {"status": "accepted", **job}
+
+
+@router.get("/{project_id}/stab/prob/progress", response_model=None)
+async def get_progress(project_id: UUID) -> dict:
+    await e02._project(project_id)
+    job = _JOBS.get(str(project_id))
+    return dict(job) if job else {"running": False, "percent": 0, "message": "", "error": None}
+
+
+@router.get("/{project_id}/stab/prob", response_model=None)
+async def get_records(project_id: UUID) -> dict:
+    root, _ = await e02._project(project_id)
+    return {"records": await asyncio.to_thread(records, root)}
diff --git a/E03_StabSection/E03_StabSection_UI_Page.ts b/E03_StabSection/E03_StabSection_UI_Page.ts
index e9ad0a2df..e458e70f4 100644
--- a/E03_StabSection/E03_StabSection_UI_Page.ts
+++ b/E03_StabSection/E03_StabSection_UI_Page.ts
@@ -6,6 +6,7 @@
  *   계산 = 서버 단독([해석] → POST …/stab/analyze2d · 잠금판 진행률 = GET …/stab/progress) ·
  *   대표 단면 켬/끔 = sessionStorage 캐시 → [대표 단면 저장] = PUT(자동저장 없음).
  *   「보강 검토」 탭 = 우측을 가정 재해석 · 벽 외적 안정으로 바꿈(`UI_Reinforce` · 서버 따로 저장본).
+ *   「확률 · 민감도」 탭 = 대표 단면 확률 해석(`UI_Prob` · 서버 따로 저장본 · 카드 안 진행률).
  * ========================================================================== */
 
 import "./E03_StabSection_UI_Style.css";
@@ -40,6 +41,7 @@ import {
   type WhatIfRecord,
 } from "./E03_StabSection_UI_Reinforce";
 import { buildSupport, type SupportAsk, type SupportRecord } from "./E03_StabSection_UI_Support";
+import { createProbPanel } from "./E03_StabSection_UI_Prob";
 import type {
   Analysis,
   Condition,
@@ -100,6 +102,7 @@ export function renderE03StabSection(root: HTMLElement): void {
     return void main.append(el("p", { className: "e03-empty", text: t("E03_NoProject") }));
   const pid: string = projectId;
   const base = `/projects/${pid}/stab`;
+  const prob = createProbPanel(pid, base, () => paintSection());
 
   let result: Result2d | null = null;
   const sections = new Map<string, SectionView>();
@@ -110,7 +113,7 @@ export function renderE03StabSection(root: HTMLElement): void {
   let side: string | null = null;
   let picked: number | null = null;
   const toggles: SectionToggles = { map: true, centers: false, candidates: true };
-  let mode: "result" | "reinforce" = "result";
+  let mode: "result" | "reinforce" | "prob" = "result";
   let busy = false;
   const slopeRecs = new Map<string, WhatIfRecord>(); // 측점 열쇠 → 마지막 가정
   const wallRecs = new Map<string, WallRecord>();
@@ -310,6 +313,7 @@ export function renderE03StabSection(root: HTMLElement): void {
       options: [
         { value: "result", text: t("E03_Tab_Result") },
         { value: "reinforce", text: t("E03_Tab_Reinforce") },
+        { value: "prob", text: t("E03_Tab_Prob") },
       ],
       value: mode,
       compact: true,
@@ -336,6 +340,12 @@ export function renderE03StabSection(root: HTMLElement): void {
     if (view.regions.some((r) => r.material_id === WALL_ID))
       legendHost.append(el("p", { className: "e03-muted", text: t("E03_Walls_Hint") }));
     const row = currentRow();
+    if (mode === "prob" && row) {
+      resultHost.append(
+        prob.render({ row, condition, side, editable: !!result?.editable && !busy }),
+      );
+      return;
+    }
     if (mode === "reinforce" && row) {
       if (row.walls && !wallRecs.has(row.key) && !busy && result?.editable)
         queueMicrotask(() => void runWall(row.key, { method: "coulomb", water_ratio: 0 }));
diff --git a/E03_StabSection/E03_StabSection_UI_Prob.css b/E03_StabSection/E03_StabSection_UI_Prob.css
new file mode 100644
index 000000000..4ea757351
--- /dev/null
+++ b/E03_StabSection/E03_StabSection_UI_Prob.css
@@ -0,0 +1,94 @@
+/* E03 확률 · 민감도 탭(537-1) — 입력 표 · 히스토그램 · 민감도 선 · 토네이도 막대 */
+.e03-prob-vars input[type="number"] {
+  width: 5.5em;
+}
+
+.e03-prob-vars select {
+  width: auto;
+}
+
+.e03-prob-run {
+  flex-wrap: wrap;
+  align-items: flex-end;
+}
+
+.e03-prob-src summary {
+  cursor: pointer;
+  color: var(--color-text-secondary);
+}
+
+.e03-prob-bar {
+  fill: var(--e03-mat-0);
+  fill-opacity: 0.75;
+}
+
+.e03-prob-bar--fail {
+  fill: var(--e03-ng);
+}
+
+.e03-prob-mark {
+  stroke-width: 1.5;
+}
+
+.e03-prob-mark--one {
+  stroke: var(--e03-ng);
+}
+
+.e03-prob-mark--req {
+  stroke: var(--e03-mat-3);
+  stroke-dasharray: 5 3;
+}
+
+.e03-prob-mark--det {
+  stroke: var(--e03-critical);
+  stroke-dasharray: 2 2;
+}
+
+.e03-prob-mark__text {
+  font-size: 10px;
+  stroke: none;
+  fill: var(--color-text-secondary);
+}
+
+.e03-prob-line {
+  fill: none;
+  stroke-width: 2;
+}
+
+.e03-prob-line--0 {
+  stroke: var(--e03-mat-0);
+}
+.e03-prob-line--1 {
+  stroke: var(--e03-mat-1);
+}
+.e03-prob-line--2 {
+  stroke: var(--e03-mat-2);
+}
+.e03-prob-line--3 {
+  stroke: var(--e03-mat-3);
+}
+.e03-prob-line--4 {
+  stroke: var(--e03-mat-4);
+}
+.e03-prob-line--5 {
+  stroke: var(--e03-mat-5);
+}
+.e03-prob-line--6 {
+  stroke: var(--e03-mat-6);
+}
+.e03-prob-line--7 {
+  stroke: var(--e03-mat-7);
+}
+
+.e03-prob-tbar--lo {
+  fill: var(--e03-mat-1);
+}
+
+.e03-prob-tbar--hi {
+  fill: var(--e03-mat-0);
+}
+
+.e03-prob-tlabel {
+  font-size: 10px;
+  fill: var(--color-text-secondary);
+}
diff --git a/E03_StabSection/E03_StabSection_UI_Prob.ts b/E03_StabSection/E03_StabSection_UI_Prob.ts
new file mode 100644
index 000000000..3cd5b777e
--- /dev/null
+++ b/E03_StabSection/E03_StabSection_UI_Prob.ts
@@ -0,0 +1,384 @@
+/* =============================================================================
+ * E03_StabSection_UI_Prob.ts
+ * E03 「확률 · 민감도」 탭(요구 351 · PLAN 537-1) — 대표 단면 · 고른 비탈 · 조건.
+ *   분포 입력(단면 물성 × c · φ · γ · 지하수위 · r_u — 쓰기 · 분포 · 변동계수/표준편차 · 절단 ±kσ) ·
+ *   표본(LHS · MC · N · 시드) · 갈래(임계면 고정 · 매회 탐색) · 방법 · c–φ 상관 →
+ *   서버 계산(POST …/stab/prob/{측점} · 뒤에서 · 진행 GET …/stab/prob/progress · 카드 안 진행률).
+ *   기본값 후보 = 서버 `prob_defaults.json`(근거 · 사용자 확정 전) · 입력 = sessionStorage 캐시(자동저장 없음).
+ *   결과 = 요약 표 · 히스토그램 · 누적분포 · 민감도 · 토네이도(`UI_ProbChart`).
+ * ========================================================================== */
+
+import "./E03_StabSection_UI_Prob.css";
+import { API_BASE_URL } from "@config/config_frontend";
+import {
+  createButton,
+  createInputField,
+  createSelectField,
+  el,
+  showToast,
+} from "@ui/ui_template_elements";
+import { t, type LocaleKey } from "@ui/ui_template_locale";
+import {
+  changedCount,
+  fillCache,
+  probRequest,
+  probRows,
+  type ProbCache,
+  type ProbRecord,
+  type ProbRow,
+  type ProbSetup,
+} from "./logic/E03_StabSection_Prob";
+import type { Condition, StationRow } from "./logic/E03_StabSection_Logic";
+import {
+  cdfChart,
+  histogramChart,
+  sensitivityChart,
+  tornadoChart,
+} from "./E03_StabSection_UI_ProbChart";
+
+export interface ProbView {
+  row: StationRow;
+  condition: Condition;
+  side: string | null;
+  editable: boolean;
+}
+
+interface Progress {
+  running: boolean;
+  percent: number;
+  message: string;
+  error: string | null;
+}
+
+const fmt = (v: number | null | undefined, d = 3): string =>
+  v === null || v === undefined || !Number.isFinite(v) ? "-" : v.toFixed(d);
+const sleep = (ms: number): Promise<void> => new Promise((r) => setTimeout(r, ms));
+const METHODS = ["bishop", "janbu", "spencer", "morgenstern_price", "ordinary"] as const;
+const methodName = (m: string): string =>
+  t(m === "morgenstern_price" ? "E03_Pb_Method_mp" : (`E03_Method_${m}` as LocaleKey));
+
+async function api<T>(path: string, init?: RequestInit): Promise<T> {
+  const response = await fetch(`${API_BASE_URL}${path}`, {
+    credentials: "include",
+    headers: { "Content-Type": "application/json" },
+    ...init,
+  });
+  const body = (await response.json().catch(() => ({}))) as T & { detail?: unknown };
+  if (!response.ok)
+    throw new Error(typeof body.detail === "string" ? body.detail : `HTTP ${response.status}`);
+  return body;
+}
+
+/** 탭 하나 — 준비(setup) · 결과 · 진행을 들고 다시 그릴 때 `repaint` 를 부름 */
+export function createProbPanel(projectId: string, base: string, repaint: () => void) {
+  const cacheKey = `e03_prob_cache:${projectId}`;
+  const setups = new Map<string, ProbSetup | "loading" | string>();
+  const records = new Map<string, ProbRecord>();
+  let cache: ProbCache | null = JSON.parse(sessionStorage.getItem(cacheKey) ?? "null");
+  let busy: Progress | null = null;
+
+  const save = (): void => sessionStorage.setItem(cacheKey, JSON.stringify(cache));
+  const slotOf = (v: ProbView, side: string): string => `${v.row.key}|${side}|${v.condition}`;
+
+  async function load(slot: string, key: string, side: string, condition: string): Promise<void> {
+    setups.set(slot, "loading");
+    try {
+      const q = `side=${side}&condition=${condition}`;
+      const s = await api<ProbSetup>(`${base}/prob/setup/${encodeURIComponent(key)}?${q}`);
+      setups.set(slot, s);
+      if (s.last) records.set(slot, s.last);
+    } catch (error) {
+      setups.set(slot, error instanceof Error ? error.message : t("Common_Status_Error"));
+    }
+    repaint();
+  }
+
+  async function run(slot: string, setup: ProbSetup): Promise<void> {
+    if (!cache) return;
+    busy = { running: true, percent: 0, message: "", error: null };
+    repaint();
+    try {
+      const body = probRequest(setup, cache);
+      const path = `${base}/prob/${encodeURIComponent(setup.station_key)}`;
+      const first = await api<{ status: string; record?: ProbRecord }>(path, {
+        method: "POST",
+        body: JSON.stringify(body),
+      });
+      if (first.status !== "fresh") {
+        for (;;) {
+          await sleep(700);
+          busy = await api<Progress>(`${base}/prob/progress`);
+          repaint();
+          if (!busy.running) break;
+        }
+        if (busy.error) throw new Error(busy.error);
+      }
+      busy = null;
+      await load(slot, setup.station_key, setup.side, setup.condition);
+      showToast(t("E03_Pb_Done"), "success");
+    } catch (error) {
+      showToast(error instanceof Error ? error.message : t("Common_Status_Error"), "error");
+    } finally {
+      busy = null;
+      repaint();
+    }
+  }
+
+  function inputTable(setup: ProbSetup): HTMLElement {
+    const table = el("table", {
+      className: "e03-table e03-prob-vars",
+      attrs: { "data-role": "prob-vars" },
+    });
+    const heads = [
+      "E03_Pb_Use",
+      "E03_Pb_Var",
+      "E03_Pb_Mean",
+      "E03_Pb_Dist",
+      "E03_Pb_Spread",
+      "E03_Pb_Trunc",
+    ];
+    table.append(
+      el("thead", {
+        children: [el("tr", { children: heads.map((k) => el("th", { text: t(k as LocaleKey) })) })],
+      }),
+    );
+    const body = el("tbody");
+    for (const row of probRows(setup)) body.append(inputRow(row));
+    table.append(body);
+    return table;
+  }
+
+  function inputRow(row: ProbRow): HTMLElement {
+    const v = cache!.vars[row.key];
+    const tr = el("tr", { attrs: { "data-key": row.key } });
+    const use = el("input", { attrs: { type: "checkbox", "data-field": "use" } });
+    use.checked = v.use;
+    use.addEventListener("change", () => ((v.use = use.checked), save()));
+    const dist = el("select", {
+      className: "ui-input ui-input--sm",
+      attrs: { "data-field": "dist" },
+    });
+    for (const d of row.prop === "water" ? ["normal"] : ["normal", "lognormal"])
+      dist.append(el("option", { text: t(`E03_Pb_Dist_${d}` as LocaleKey), attrs: { value: d } }));
+    dist.value = v.dist;
+    dist.addEventListener("change", () => ((v.dist = dist.value as typeof v.dist), save()));
+    const num = (value: number | null, field: string, apply: (n: number | null) => void) => {
+      const input = el("input", {
+        className: "ui-input ui-input--sm",
+        attrs: { type: "number", step: "0.01", min: "0", "data-field": field },
+      });
+      input.value = value === null ? "" : String(value);
+      input.addEventListener("input", () => {
+        const n = Number(input.value);
+        if (input.value.trim() === "") apply(null);
+        else if (Number.isFinite(n) && n >= 0) apply(n);
+        save();
+      });
+      return input;
+    };
+    const unit = row.kind === "cov" ? t("E03_Pb_Cov") : row.prop === "water" ? "σ m" : "σ";
+    const label =
+      row.prop === "water" || row.prop === "ru"
+        ? t(`E03_Pb_Prop_${row.prop}` as LocaleKey)
+        : `${row.name} ${t(`E03_Pb_Prop_${row.prop}` as LocaleKey)}`;
+    tr.append(
+      el("td", { children: [use] }),
+      el("td", { text: label }),
+      el("td", { className: "is-num", text: row.prop === "water" ? "±0" : fmt(row.mean, 2) }),
+      el("td", { children: [dist] }),
+      el("td", {
+        children: [
+          num(v.value, "value", (n) => (v.value = n ?? 0)),
+          el("span", { className: "e03-muted", text: ` ${unit}` }),
+        ],
+      }),
+      el("td", { children: [num(v.trunc, "trunc", (n) => (v.trunc = n))] }),
+    );
+    return tr;
+  }
+
+  function runFields(setup: ProbSetup, slot: string, editable: boolean): HTMLElement {
+    const r = cache!.run;
+    const wrap = el("div", { className: "e03-panel__row e03-prob-run" });
+    const set = <K extends keyof typeof r>(k: K, v: (typeof r)[K]): void => {
+      r[k] = v;
+      save();
+    };
+    const sel = (
+      label: LocaleKey,
+      value: string,
+      opts: [string, string][],
+      on: (v: string) => void,
+    ) => {
+      const f = createSelectField({
+        label: t(label),
+        options: opts.map(([value, text]) => ({ value, text })),
+        value,
+        compact: true,
+        onChange: on,
+      });
+      f.select.dataset.field = label;
+      return f.root;
+    };
+    const numField = (label: LocaleKey, value: number, step: string, on: (n: number) => void) => {
+      const f = createInputField({
+        label: t(label),
+        type: "number",
+        compact: true,
+        value: String(value),
+      });
+      f.input.step = step;
+      f.input.dataset.field = label;
+      f.input.addEventListener("input", () => {
+        const n = Number(f.input.value);
+        if (f.input.value.trim() !== "" && Number.isFinite(n)) on(n);
+      });
+      return f.root;
+    };
+    wrap.append(
+      sel(
+        "E03_Pb_Mode",
+        r.mode,
+        [
+          ["fixed", t("E03_Pb_Mode_fixed")],
+          ["search", t("E03_Pb_Mode_search")],
+        ],
+        (v) => set("mode", v as typeof r.mode),
+      ),
+      sel(
+        "E03_Pb_Method",
+        r.method,
+        METHODS.map((m) => [m, methodName(m)]),
+        (v) => set("method", v as typeof r.method),
+      ),
+      sel(
+        "E03_Pb_Sampling",
+        r.sampling,
+        [
+          ["lhs", "LHS"],
+          ["mc", "Monte Carlo"],
+        ],
+        (v) => set("sampling", v as typeof r.sampling),
+      ),
+      numField("E03_Pb_N", r.n, "100", (n) => set("n", n)),
+      numField("E03_Pb_Seed", r.seed, "1", (n) => set("seed", n)),
+      numField("E03_Pb_Rho", r.rho, "0.1", (n) => set("rho", n)),
+    );
+    const go = createButton({
+      label: busy ? `${t("E03_Rf_Busy")} ${busy.percent}%` : t("E03_Pb_Run"),
+      variant: "filled",
+      onClick: () => void run(slot, setup),
+    });
+    go.dataset.role = "prob-run";
+    go.disabled = !editable || !!busy;
+    const reset = createButton({
+      label: t("E03_Pb_Defaults"),
+      variant: "ghost",
+      onClick: () => {
+        cache = fillCache(setup, null);
+        save();
+        repaint();
+      },
+    });
+    reset.disabled = !!busy;
+    wrap.append(reset, go);
+    return wrap;
+  }
+
+  function summary(rec: ProbRecord): HTMLElement {
+    const s = rec.result.stats;
+    const rows: [LocaleKey, string][] = [
+      ["E03_Pb_Det", fmt(rec.fs_deterministic)],
+      ["E03_Pb_MeanF", fmt(s.mean)],
+      ["E03_Pb_Std", fmt(s.std)],
+      ["E03_Pb_Pf", `${fmt((s.pf ?? 0) * 100, 2)} %`],
+      ["E03_Pb_BetaN", fmt(s.beta_normal)],
+      ["E03_Pb_BetaLn", fmt(s.beta_lognormal)],
+      [
+        "E03_Pb_PReq",
+        s.p_below_required === null || s.p_below_required === undefined
+          ? "-"
+          : `${fmt(s.p_below_required * 100, 2)} %`,
+      ],
+      ["E03_Pb_Valid", `${s.n_valid} / ${s.n}`],
+      ["E03_Pb_Time", `${fmt(rec.result.elapsed_s.total, 1)} s`],
+    ];
+    const dl = el("dl", { className: "e03-kpi", attrs: { "data-role": "prob-summary" } });
+    for (const [k, v] of rows)
+      dl.append(el("dt", { text: t(k) }), el("dd", { text: v, attrs: { "data-k": k } }));
+    const r = rec.result;
+    const note = `${t(`E03_Pb_Mode_${r.mode}` as LocaleKey)} · ${methodName(r.method)} · ${r.sampling.toUpperCase()} ${r.n} · seed ${r.seed} · ${rec.computed_at}`;
+    return el("div", { children: [dl, el("p", { className: "e03-muted", text: note })] });
+  }
+
+  function sources(setup: ProbSetup, changed: number): HTMLElement {
+    const d = el("details", { className: "e03-prob-src" });
+    const tag = changed
+      ? t("E03_Pb_UserInput").replace("{n}", String(changed))
+      : t("E03_Pb_Unconfirmed");
+    d.append(el("summary", { text: tag }));
+    for (const [k, v] of Object.entries(setup.defaults.sources))
+      d.append(el("p", { className: "e03-muted", text: `${k} — ${v}` }));
+    return d;
+  }
+
+  function render(v: ProbView): HTMLElement {
+    const host = el("div", { className: "e03-card e03-prob", attrs: { "data-role": "prob" } });
+    const head = (key: LocaleKey): HTMLElement =>
+      el("h3", { className: "e03-card__title", text: t(key) });
+    host.append(head("E03_Tab_Prob"));
+    if (!v.row.representative?.on) {
+      host.append(el("p", { className: "e03-empty", text: t("E03_Pb_RepOnly") }));
+      return host;
+    }
+    const side = v.side ?? v.row.sides[0]?.side;
+    if (!side) return host;
+    const slot = slotOf(v, side);
+    const got = setups.get(slot);
+    if (got === undefined) queueMicrotask(() => void load(slot, v.row.key, side, v.condition));
+    if (typeof got !== "object") {
+      const text = typeof got === "string" && got !== "loading" ? got : t("E03_Loading");
+      host.append(el("p", { className: "e03-empty", text }));
+      return host;
+    }
+    cache = fillCache(got, cache);
+    const sideText = `${t(side === "left" ? "E03_Left" : "E03_Right")} · ${t(got.role === "cut" ? "E03_Cut" : "E03_Fill")}`;
+    host.append(
+      el("p", {
+        className: "e03-muted",
+        text: `${sideText} · ${t(`E03_Cond_${v.condition}` as LocaleKey)} · ${t("E03_Pb_MeanNote")}`,
+      }),
+    );
+    host.append(
+      inputTable(got),
+      runFields(got, slot, v.editable),
+      sources(got, changedCount(got, cache)),
+    );
+    if (busy)
+      host.append(
+        el("p", {
+          className: "e03-muted",
+          text: `${busy.percent}% · ${busy.message}`,
+          attrs: { "data-role": "prob-progress" },
+        }),
+      );
+    const rec = records.get(slot);
+    if (!rec) return host;
+    const res = rec.result;
+    host.append(
+      head("E03_Pb_Result"),
+      summary(rec),
+      head("E03_Pb_Hist"),
+      histogramChart(res, rec.required, rec.fs_deterministic),
+      head("E03_Pb_Cdf"),
+      cdfChart(res, rec.required),
+      head("E03_Pb_Sens"),
+      sensitivityChart(res),
+      head("E03_Pb_Tornado"),
+      tornadoChart(res),
+    );
+    return host;
+  }
+
+  return { render, isBusy: (): boolean => !!busy };
+}
diff --git a/E03_StabSection/E03_StabSection_UI_ProbChart.ts b/E03_StabSection/E03_StabSection_UI_ProbChart.ts
new file mode 100644
index 000000000..57ba2ce39
--- /dev/null
+++ b/E03_StabSection/E03_StabSection_UI_ProbChart.ts
@@ -0,0 +1,220 @@
+/* =============================================================================
+ * E03_StabSection_UI_ProbChart.ts
+ * E03 「확률 · 민감도」 그래프 SVG(PLAN 537-1) — 안전율 히스토그램(F < 1 빨강 · F = 1 · 기준 선) ·
+ *   누적분포 · 민감도(가로 = 범위 % · 변수마다 선 · Slide2 Sensitivity Plot) · 토네이도(끝값 F 폭 순).
+ * ========================================================================== */
+
+import { el } from "@ui/ui_template_elements";
+import { t } from "@ui/ui_template_locale";
+import { axisTicks, type ProbCurve, type ProbResult } from "./logic/E03_StabSection_Prob";
+import { pointsText, type Point } from "./logic/E03_StabSection_Logic";
+import { svgEl } from "./E03_StabSection_UI_Section";
+
+const W = 380;
+const H = 200;
+const PAD = { l: 44, r: 12, t: 12, b: 28 };
+const fmt = (v: number | null | undefined, d = 2): string =>
+  v === null || v === undefined || !Number.isFinite(v) ? "-" : v.toFixed(d);
+
+interface Frame {
+  svg: SVGSVGElement;
+  sx: (x: number) => number;
+  sy: (y: number) => number;
+}
+
+/** 축 틀 — 가로 · 세로 눈금 · 축 이름 */
+function frame(
+  role: string,
+  xr: [number, number],
+  yr: [number, number],
+  xLabel: string,
+  yLabel: string,
+  pad = PAD,
+  yAxis = true,
+): Frame {
+  const xt = axisTicks(xr[0], xr[1], 5);
+  const yt = axisTicks(yr[0], yr[1], 4);
+  const [x0, x1] = [xt[0], xt[xt.length - 1]];
+  const [y0, y1] = [yt[0], yt[yt.length - 1]];
+  const sx = (x: number): number =>
+    pad.l + ((x - x0) / Math.max(x1 - x0, 1e-9)) * (W - pad.l - pad.r);
+  const sy = (y: number): number =>
+    H - pad.b - ((y - y0) / Math.max(y1 - y0, 1e-9)) * (H - pad.t - pad.b);
+  const svg = svgEl("svg", { viewBox: `0 0 ${W} ${H}`, "data-role": role }) as SVGSVGElement;
+  const axis = svgEl("g", { class: "e03-axis" });
+  for (const v of yAxis ? yt : []) {
+    axis.append(svgEl("line", { x1: pad.l, x2: W - pad.r, y1: sy(v), y2: sy(v) }));
+    axis.append(svgEl("text", { x: pad.l - 4, y: sy(v) + 4, "text-anchor": "end" }, String(v)));
+  }
+  for (const v of xt)
+    axis.append(svgEl("text", { x: sx(v), y: H - pad.b + 14, "text-anchor": "middle" }, String(v)));
+  axis.append(svgEl("text", { x: W - pad.r, y: H - 2, "text-anchor": "end" }, xLabel));
+  axis.append(svgEl("text", { x: 2, y: pad.t - 2 }, yLabel));
+  svg.append(axis);
+  return { svg, sx, sy };
+}
+
+function mark(f: Frame, x: number, cls: string, label: string): void {
+  f.svg.append(
+    svgEl("line", {
+      x1: f.sx(x),
+      x2: f.sx(x),
+      y1: PAD.t,
+      y2: H - PAD.b,
+      class: `e03-prob-mark ${cls}`,
+    }),
+    svgEl("text", { x: f.sx(x) + 3, y: PAD.t + 10, class: `e03-prob-mark__text ${cls}` }, label),
+  );
+}
+
+function box(role: string, svg: SVGSVGElement, legend?: HTMLElement): HTMLElement {
+  const host = el("div", { className: "e03-graph", attrs: { "data-role": `${role}-box` } });
+  host.append(svg as unknown as HTMLElement);
+  if (legend) host.append(legend);
+  return host;
+}
+
+/** 안전율 히스토그램 — 칸 개수 · F < 1 칸 빨강 · F = 1 · 기준 · 결정론 선 */
+export function histogramChart(
+  res: ProbResult,
+  required: number | null,
+  fsDet: number | null,
+): HTMLElement {
+  const h = res.stats.histogram;
+  if (!h) return el("div");
+  const { edges, counts } = h;
+  const xr: [number, number] = [
+    Math.min(edges[0], 1),
+    Math.max(edges[edges.length - 1], required ?? 0),
+  ];
+  const f = frame("prob-hist", xr, [0, Math.max(...counts, 1)], "Fs", t("E03_Pb_Count"));
+  counts.forEach((n, i) => {
+    const [a, b] = [edges[i], edges[i + 1]];
+    const bar = svgEl("rect", {
+      x: f.sx(a),
+      y: f.sy(n),
+      width: Math.max(f.sx(b) - f.sx(a) - 0.5, 0.5),
+      height: f.sy(0) - f.sy(n),
+      class: b <= 1 ? "e03-prob-bar e03-prob-bar--fail" : "e03-prob-bar",
+      "data-count": n,
+    });
+    bar.append(svgEl("title", {}, `${fmt(a, 3)} ~ ${fmt(b, 3)} · ${n}`));
+    f.svg.append(bar);
+  });
+  mark(f, 1, "e03-prob-mark--one", "F = 1");
+  if (required) mark(f, required, "e03-prob-mark--req", `${t("E03_Required")} ${fmt(required)}`);
+  if (fsDet) mark(f, fsDet, "e03-prob-mark--det", t("E03_Pb_Det"));
+  return box("prob-hist", f.svg);
+}
+
+/** 누적분포 — 가로 F · 세로 P(F ≤ x) · F = 1 에서 Pf */
+export function cdfChart(res: ProbResult, required: number | null): HTMLElement {
+  const c = res.stats.cdf;
+  if (!c) return el("div");
+  const xr: [number, number] = [Math.min(c.F[0], 1), Math.max(c.F[c.F.length - 1], required ?? 0)];
+  const f = frame("prob-cdf", xr, [0, 1], "Fs", "P(F ≤ x)");
+  const pts: Point[] = c.F.map((x, i) => [f.sx(x), f.sy(c.p[i])]);
+  f.svg.append(
+    svgEl("polyline", { points: pointsText(pts), class: "e03-prob-line e03-prob-line--0" }),
+  );
+  mark(f, 1, "e03-prob-mark--one", `Pf ${fmt((res.stats.pf ?? 0) * 100, 1)} %`);
+  if (required) mark(f, required, "e03-prob-mark--req", `${t("E03_Required")} ${fmt(required)}`);
+  return box("prob-cdf", f.svg);
+}
+
+function legendOf(curves: ProbCurve[]): HTMLElement {
+  const legend = el("div", { className: "e03-legend" });
+  curves.forEach((c, i) => {
+    const item = el("span", { className: "e03-legend__item" });
+    const sw = svgEl("svg", { width: 18, height: 8 });
+    sw.append(
+      svgEl("line", {
+        x1: 0,
+        x2: 18,
+        y1: 4,
+        y2: 4,
+        class: `e03-prob-line e03-prob-line--${i % 8}`,
+      }),
+    );
+    item.append(sw as unknown as HTMLElement, c.name);
+    legend.append(item);
+  });
+  return legend;
+}
+
+/** 민감도 — 가로 = 범위 %(0 = 아래 끝 · 100 = 위 끝) · 나머지 변수 평균 · 임계면 고정 */
+export function sensitivityChart(res: ProbResult): HTMLElement {
+  const curves = res.sensitivity.curves;
+  const all = curves.flatMap((c) => c.F).filter((v): v is number => v !== null);
+  if (!all.length) return el("div");
+  const f = frame(
+    "prob-sens",
+    [0, 100],
+    [Math.min(...all, 1), Math.max(...all)],
+    t("E03_Pb_Range"),
+    "Fs",
+  );
+  curves.forEach((c, i) => {
+    const n = c.F.length;
+    const pts: Point[] = [];
+    c.F.forEach((v, k) => {
+      if (v !== null) pts.push([f.sx((100 * k) / Math.max(n - 1, 1)), f.sy(v)]);
+    });
+    const line = svgEl("polyline", {
+      points: pointsText(pts),
+      class: `e03-prob-line e03-prob-line--${i % 8}`,
+      "data-name": c.name,
+    });
+    line.append(svgEl("title", {}, `${c.name} · ${fmt(c.x[0])} ~ ${fmt(c.x[n - 1])}`));
+    f.svg.append(line);
+  });
+  return box("prob-sens", f.svg, legendOf(curves));
+}
+
+/** 토네이도 — 변수마다 아래 끝 · 위 끝 F 막대(평균 F 기준) · 폭 큰 순 */
+export function tornadoChart(res: ProbResult): HTMLElement {
+  const rows = res.sensitivity.tornado;
+  const F0 = res.sensitivity.F_mean;
+  if (!rows.length || F0 === null) return el("div");
+  const vals = rows.flatMap((r) => [r.F_lo, r.F_hi]).filter((v): v is number => v !== null);
+  const pad = { ...PAD, l: 96 };
+  const f = frame(
+    "prob-tornado",
+    [Math.min(...vals), Math.max(...vals)],
+    [0, 1],
+    "Fs",
+    "",
+    pad,
+    false,
+  );
+  const step = (H - PAD.t - PAD.b) / rows.length;
+  rows.forEach((r, i) => {
+    const y = PAD.t + i * step + step * 0.15;
+    const hh = step * 0.7;
+    for (const [v, cls] of [
+      [r.F_lo, "e03-prob-tbar--lo"],
+      [r.F_hi, "e03-prob-tbar--hi"],
+    ] as const) {
+      if (v === null) continue;
+      const [a, b] = [Math.min(f.sx(v), f.sx(F0)), Math.max(f.sx(v), f.sx(F0))];
+      const bar = svgEl("rect", {
+        x: a,
+        y,
+        width: Math.max(b - a, 0.5),
+        height: hh,
+        class: `e03-prob-tbar ${cls}`,
+      });
+      bar.append(svgEl("title", {}, `${r.name} · ${fmt(v, 3)}`));
+      f.svg.append(bar);
+    }
+    f.svg.append(
+      svgEl(
+        "text",
+        { x: pad.l - 4, y: y + hh / 2 + 4, "text-anchor": "end", class: "e03-prob-tlabel" },
+        r.name,
+      ),
+    );
+  });
+  mark(f, F0, "e03-prob-mark--det", fmt(F0, 3));
+  return box("prob-tornado", f.svg);
+}
diff --git a/E03_StabSection/logic/E03_StabSection_Prob.ts b/E03_StabSection/logic/E03_StabSection_Prob.ts
new file mode 100644
index 000000000..5d810d405
--- /dev/null
+++ b/E03_StabSection/logic/E03_StabSection_Prob.ts
@@ -0,0 +1,266 @@
+/* =============================================================================
+ * E03_StabSection_Prob.ts
+ * E03 「확률 · 민감도」 탭 계산(PLAN 537-1) — 변수 줄(단면 물성 × c · φ · γ + 지하수위 · r_u) ·
+ *   기본값 후보(서버 `prob_defaults.json`) · 입력 캐시 → 요청 꼴 · 그래프 눈금.
+ *   화면 · fetch · sessionStorage 없음(페이지가 부름).
+ * ========================================================================== */
+
+export type ProbProp = "c" | "phi" | "gamma" | "water" | "ru";
+export type ProbDist = "normal" | "lognormal";
+export type ProbMethod = "ordinary" | "bishop" | "janbu" | "spencer" | "morgenstern_price";
+
+export interface ProbPropDefault {
+  dist: ProbDist;
+  cov?: number;
+  std?: number;
+  trunc_sigma: number | null;
+}
+
+export interface ProbDefaults {
+  run: {
+    n: number;
+    sampling: "lhs" | "mc";
+    seed: number;
+    mode: "fixed" | "search";
+    method: ProbMethod;
+  };
+  properties: Record<ProbProp, ProbPropDefault>;
+  rho_c_phi: number;
+  sources: Record<string, string>;
+  materials?: Record<string, Partial<Record<ProbProp, Partial<ProbPropDefault>>>>;
+}
+
+export interface ProbMaterial {
+  id: string;
+  name: string | null;
+  c: number;
+  phi: number;
+  gamma_t: number;
+}
+
+export interface ProbStats {
+  n: number;
+  n_valid: number;
+  mean: number | null;
+  std: number | null;
+  cov?: number | null;
+  min?: number;
+  max?: number;
+  pf: number | null;
+  n_fail?: number;
+  beta_normal: number | null;
+  beta_lognormal?: number | null;
+  required?: number | null;
+  p_below_required?: number | null;
+  histogram?: { edges: number[]; counts: number[] };
+  cdf?: { p: number[]; F: number[] };
+}
+
+export interface ProbCurve {
+  name: string;
+  x: (number | null)[];
+  F: (number | null)[];
+  F_lo: number | null;
+  F_hi: number | null;
+  swing: number | null;
+}
+
+export interface ProbVariable {
+  name: string;
+  material: string | null;
+  prop: ProbProp;
+  dist: ProbDist;
+  mean: number;
+  std: number;
+  lo: number | null;
+  hi: number | null;
+}
+
+export interface ProbResult {
+  mode: "fixed" | "search";
+  method: ProbMethod;
+  n: number;
+  sampling: string;
+  seed: number;
+  variables: ProbVariable[];
+  stats: ProbStats;
+  sensitivity: {
+    F_mean: number | null;
+    curves: ProbCurve[];
+    tornado: { name: string; F_lo: number | null; F_hi: number | null; swing: number }[];
+  };
+  elapsed_s: { total: number; samples: number };
+}
+
+export interface ProbRecord {
+  station_key: string;
+  side: string;
+  condition: string;
+  required: number | null;
+  fs_deterministic: number | null;
+  request: Record<string, unknown>;
+  result: ProbResult;
+  computed_at: string;
+}
+
+export interface ProbSetup {
+  station_key: string;
+  side: string;
+  condition: string;
+  role: "cut" | "fill";
+  required: number | null;
+  fs_deterministic: Record<string, number | null>;
+  materials: ProbMaterial[];
+  groundwater: { mode: "none" | "line" | "ru"; ru: number | null };
+  defaults: ProbDefaults;
+  search_max_n: number;
+  last: ProbRecord | null;
+}
+
+/** 변수 줄 하나 — kind cov = 변동계수(평균 곱) · std = 표준편차(m · r_u) */
+export interface ProbRow {
+  key: string;
+  material: string | null;
+  prop: ProbProp;
+  name: string;
+  mean: number;
+  kind: "cov" | "std";
+}
+
+export interface ProbInput {
+  use: boolean;
+  dist: ProbDist;
+  value: number;
+  trunc: number | null;
+}
+
+export interface ProbRunInput {
+  n: number;
+  sampling: "lhs" | "mc";
+  seed: number;
+  mode: "fixed" | "search";
+  method: ProbMethod;
+  rho: number;
+}
+
+export interface ProbCache {
+  vars: Record<string, ProbInput>;
+  run: ProbRunInput;
+}
+
+export const varKey = (material: string | null, prop: ProbProp): string =>
+  `${material ?? "*"}|${prop}`;
+
+/** 단면 물성마다 c · φ · γ + 지하수 갈래가 수위선이면 지하수위 · r_u 면 r_u */
+export function probRows(setup: ProbSetup): ProbRow[] {
+  const rows: ProbRow[] = [];
+  for (const m of setup.materials) {
+    const name = m.name ?? m.id;
+    rows.push(
+      { key: varKey(m.id, "c"), material: m.id, prop: "c", name, mean: m.c, kind: "cov" },
+      { key: varKey(m.id, "phi"), material: m.id, prop: "phi", name, mean: m.phi, kind: "cov" },
+      {
+        key: varKey(m.id, "gamma"),
+        material: m.id,
+        prop: "gamma",
+        name,
+        mean: m.gamma_t,
+        kind: "cov",
+      },
+    );
+  }
+  const gw = setup.groundwater;
+  if (gw.mode === "line")
+    rows.push({
+      key: varKey(null, "water"),
+      material: null,
+      prop: "water",
+      name: "",
+      mean: 0,
+      kind: "std",
+    });
+  if (gw.mode === "ru")
+    rows.push({
+      key: varKey(null, "ru"),
+      material: null,
+      prop: "ru",
+      name: "",
+      mean: gw.ru ?? 0,
+      kind: "std",
+    });
+  return rows;
+}
+
+/** 기본값 후보 — 성질 기본 위에 물성별 덮어쓰기 · 0 이면 끔 */
+export function defaultInput(defaults: ProbDefaults, row: ProbRow): ProbInput {
+  const base = defaults.properties[row.prop];
+  const own = (row.material && defaults.materials?.[row.material]?.[row.prop]) || {};
+  const merged = { ...base, ...own };
+  const value = Number((row.kind === "cov" ? merged.cov : merged.std) ?? 0);
+  return {
+    use: value > 0,
+    dist: row.prop === "water" ? "normal" : merged.dist,
+    value,
+    trunc: merged.trunc_sigma ?? null,
+  };
+}
+
+/** 캐시가 없는 줄만 기본값 후보로 채움 */
+export function fillCache(setup: ProbSetup, cache: ProbCache | null): ProbCache {
+  const d = setup.defaults;
+  const vars: Record<string, ProbInput> = { ...(cache?.vars ?? {}) };
+  for (const row of probRows(setup)) vars[row.key] ??= defaultInput(d, row);
+  const run = cache?.run ?? { ...d.run, rho: d.rho_c_phi };
+  return { vars, run };
+}
+
+/** 기본값과 다른 줄 수(「사용자 입력」 표기) */
+export function changedCount(setup: ProbSetup, cache: ProbCache): number {
+  return probRows(setup).filter((row) => {
+    const a = defaultInput(setup.defaults, row);
+    const b = cache.vars[row.key];
+    return (
+      b && (a.use !== b.use || a.dist !== b.dist || a.value !== b.value || a.trunc !== b.trunc)
+    );
+  }).length;
+}
+
+/** POST …/stab/prob/{측점} 몸 — 켠 줄만 · 값 0 은 뺌 */
+export function probRequest(setup: ProbSetup, cache: ProbCache): Record<string, unknown> {
+  const variables = probRows(setup)
+    .map((row) => ({ row, v: cache.vars[row.key] }))
+    .filter(({ v }) => v?.use && v.value > 0)
+    .map(({ row, v }) => ({
+      material: row.material,
+      prop: row.prop,
+      dist: v.dist,
+      ...(row.kind === "cov" ? { cov: v.value } : { std: v.value }),
+      trunc_sigma: v.trunc,
+    }));
+  const r = cache.run;
+  return {
+    side: setup.side,
+    condition: setup.condition,
+    method: r.method,
+    mode: r.mode,
+    n: Math.round(r.n),
+    sampling: r.sampling,
+    seed: Math.round(r.seed),
+    rho_c_phi: r.rho,
+    variables,
+  };
+}
+
+/** 그래프 눈금 — 0 을 억지로 넣지 않음(안전율 축) */
+export function axisTicks(min: number, max: number, count = 5): number[] {
+  const lo = Math.min(min, max);
+  const hi = Math.max(lo + 1e-9, max);
+  const raw = (hi - lo) / Math.max(count, 1);
+  const mag = 10 ** Math.floor(Math.log10(raw));
+  const step = [1, 2, 5, 10].map((m) => m * mag).find((v) => v >= raw) ?? raw;
+  const out: number[] = [];
+  for (let v = Math.floor(lo / step) * step; v <= hi + step * 1e-6; v += step)
+    out.push(Math.round(v * 1e6) / 1e6);
+  if (out[out.length - 1] < hi) out.push(out[out.length - 1] + step);
+  return out;
+}
diff --git a/E05_StabReport/E05_StabReport_Build.py b/E05_StabReport/E05_StabReport_Build.py
index 0105cd5e1..bf0407150 100644
--- a/E05_StabReport/E05_StabReport_Build.py
+++ b/E05_StabReport/E05_StabReport_Build.py
@@ -3,14 +3,14 @@
 입력 = E02 지반 조건(`ground.json`) · E03 2D 파생 저장본(`stab2d_summary` 가장 최근 · 측점별
 `stab2d_station`) · 대표 단면(E03 자동 + `representative.json` 켬/끔) · E04 3D 저장본 요약
 (있으면) · E07 광역 저장본(있으면) · E03 보강 검토 저장본(있으면 · 없으면 s8 뺌) ·
-E06 암반 해석 저장본(있으면 · 없으면 s11 뺌) ·
+E06 암반 해석 저장본(있으면 · 없으면 s11 뺌) · E03 확률 해석 저장본(있으면 · 없으면 s12 뺌) ·
 E01 복사본(측점 · 횡단 · 노선).
 양식 = `resources/master_template/report/stab_report.json`
 칸 자리 = 같은 폴더 `stab_report_slots.md`.
 
 몫: 사용자(표지 s1 · 개요 s2) = 처음 만들 때만 채움 ·
 프로그램(목차 s9 · 입력조건 s3 · 측점표 s4 · 대표단면 s5 · 3D요약 s6 · 광역요약 s7 · 보강검토 s8 ·
-암반검토 s11 · 부록_절편 s10) = [보고서 갱신] 때 양식에서 통째로 다시 씀. 대표 단면 · 부록 = `_Section` ·
+암반검토 s11 · 확률해석 s12 · 부록_절편 s10) = [보고서 갱신] 때 양식에서 통째로 다시 씀. 대표 단면 · 부록 = `_Section` ·
 목차 쪽 번호 · 머리말/꼬리말 · 긴 표 쪽 나눔 = `_Print` · 칸 공용 = `_Cells`.
 판정 글 칸 = 엔진 판정 값 + N.G 서식 10 · O.K 서식 11 을 직접 붙임
 (인쇄 HTML · xlsx 는 조건부 서식을 안 먹음).
@@ -38,6 +38,7 @@ from E04_Stab3D import E04_Stab3D_Run as run3d
 from E05_StabReport import E05_StabReport_Area as area
 from E05_StabReport import E05_StabReport_Figure as figure
 from E05_StabReport import E05_StabReport_Print as printer
+from E05_StabReport import E05_StabReport_Prob as prob
 from E05_StabReport import E05_StabReport_Reinforce as reinforce
 from E05_StabReport import E05_StabReport_Rock as rock
 from E05_StabReport import E05_StabReport_Section as section
@@ -62,7 +63,7 @@ FORM_PATH = ROOT / "resources" / "master_template" / "report" / "stab_report.jso
 STAGE = "E05_StabReport"
 REPORT, META = "report.json", "report_meta.json"
 USER_SHEETS = ("s1", "s2")
-PROGRAM_SHEETS = ("s9", "s3", "s4", "s5", "s6", "s7", "s8", "s11", "s10")
+PROGRAM_SHEETS = ("s9", "s3", "s4", "s5", "s6", "s7", "s8", "s11", "s12", "s10")
 REPRESENTATIVE = Path("E03_StabSection") / "representative.json"
 BAND_KEY = "fig/3d_band.png"
 
@@ -170,6 +171,7 @@ def analysis_fingerprint(project_root: Path) -> str:
         "area": area.stamp(project_root),
         "reinforce": reinforce.stamp(project_root),
         "rock": rock.stamp(project_root),
+        "prob": prob.stamp(project_root),
         "form": hashlib.sha1(FORM_PATH.read_bytes()).hexdigest(),
     }
     text = json.dumps(body, ensure_ascii=False, sort_keys=True, default=str)
@@ -221,6 +223,7 @@ def gather(project_root: Path) -> dict[str, Any]:
         "area": area.read(project_root),
         "reinforce": reinforce.read(project_root),
         "rock": rock.read(project_root),
+        "prob": prob.read(project_root),
         "fingerprint": analysis_fingerprint(project_root),
     }
 
@@ -453,6 +456,10 @@ def build_program(book: dict[str, Any], inputs: dict[str, Any]) -> list[tuple[st
         rock.fill_rock(by["s11"], inputs["rock"], _labels(inputs["stations"]))
     else:  # 암반 해석이 없으면 시트를 뺌
         book["시트"] = [s for s in book["시트"] if s["id"] != "s11"]
+    if inputs.get("prob"):
+        prob.fill_prob(by["s12"], inputs["prob"], _labels(inputs["stations"]))
+    else:  # 확률 해석이 없으면 시트를 뺌
+        book["시트"] = [s for s in book["시트"] if s["id"] != "s12"]
     printer.fill_toc(by["s9"], book, None)  # 쪽 번호 = 저장 전 `printer.number_toc`
     return jobs
 
diff --git a/E05_StabReport/E05_StabReport_Print.py b/E05_StabReport/E05_StabReport_Print.py
index ff06e3dbf..d2dfe7603 100644
--- a/E05_StabReport/E05_StabReport_Print.py
+++ b/E05_StabReport/E05_StabReport_Print.py
@@ -32,6 +32,7 @@ TITLES = {
     "s7": "광역 자연사면 위험 요약",
     "s8": "보강 검토 — 비탈 완화 · 소단 · 벽 외적 안정",
     "s11": "암반 비탈면 검토 — 평사투영 · 평면 · 쐐기 안전율",
+    "s12": "확률 해석 · 민감도 — 파괴확률 · 신뢰도 지수",
     "s10": "대표 단면 절편 계산표",
 }
 APPENDIX = {"s10"}
diff --git a/E05_StabReport/E05_StabReport_Prob.py b/E05_StabReport/E05_StabReport_Prob.py
new file mode 100644
index 000000000..4388dce08
--- /dev/null
+++ b/E05_StabReport/E05_StabReport_Prob.py
@@ -0,0 +1,145 @@
+"""E05 보고서 s12 확률해석 — E03 확률 해석 저장본(대표 단면 · 비탈 · 조건마다 가장 최근) → 시트(PLAN 537-1).
+
+칸 자리 = `resources/master_template/report/stab_report_slots.md` s12.
+저장본이 없으면 시트를 뺌(있을 때만) — `build_program` 이 책에서 지움.
+"""
+
+from __future__ import annotations
+
+import copy
+import hashlib
+import json
+from pathlib import Path
+from typing import Any
+
+from E03_StabSection.E03_StabSection_Router_Prob import load_defaults, records
+from E05_StabReport.E05_StabReport_Cells import _set, mark
+
+CONDITION_NAMES = {"dry": "건기", "wet": "우기", "seismic": "지진"}
+SIDE_NAMES = {"left": "좌", "right": "우"}
+ROLE_NAMES = {"cut": "깎기", "fill": "쌓기"}
+METHOD_NAMES = {
+    "ordinary": "Fellenius",
+    "bishop": "Bishop",
+    "janbu": "Janbu 보정",
+    "spencer": "Spencer",
+    "morgenstern_price": "M-P",
+}
+MODE_NAMES = {"fixed": "임계면 고정", "search": "매회 탐색"}
+DIST_NAMES = {"normal": "정규", "lognormal": "대수정규"}
+COLS = "ABCDEFGHIJK"
+
+
+def read(project_root: Path) -> list[dict[str, Any]]:
+    return records(Path(project_root))
+
+
+def stamp(project_root: Path) -> str:
+    """해석 지문용 — 저장본 목록 내용(표본 F 뺌)."""
+    body = [
+        {**r, "result": {k: v for k, v in r["result"].items() if k != "F"}}
+        for r in read(project_root)
+    ]
+    text = json.dumps(body, ensure_ascii=False, sort_keys=True, default=str)
+    return hashlib.sha1(text.encode("utf-8")).hexdigest()
+
+
+def _is_default(rec: dict[str, Any], defaults: dict[str, Any]) -> bool:
+    """요청 변수 · 표본 설정이 기본값 후보 그대로인지(「사용자 확정 전」 표기)."""
+    req = rec["request"]
+    props, own = defaults["properties"], defaults.get("materials") or {}
+    for v in req["variables"]:
+        base = {
+            **props[v["prop"]],
+            **((own.get(v.get("material") or "") or {}).get(v["prop"]) or {}),
+        }
+        spread = v.get("cov") if v.get("cov") is not None else v.get("std")
+        want = base.get("cov") if "cov" in base else base.get("std")
+        if (
+            v["dist"] != base["dist"]
+            or spread != want
+            or v.get("trunc_sigma") != base["trunc_sigma"]
+        ):
+            return False
+    run = defaults["run"]
+    same = all(req.get(k) == run[k] for k in ("n", "sampling", "seed", "mode", "method"))
+    return same and req.get("rho_c_phi", 0.0) == defaults["rho_c_phi"]
+
+
+def fill_prob(sheet: dict[str, Any], recs: list[dict[str, Any]], labels: dict) -> None:
+    """s12 — 1. 측점 · 비탈 · 조건마다 한 줄(6 행 서식) · 2. 변수마다 한 줄(10 행 서식 · 토네이도 순)."""
+    tpl = {
+        r: {c: copy.deepcopy(sheet["칸"].get(f"{c}{r}") or {}) for c in COLS} for r in (6, 8, 9, 10)
+    }
+    for a1 in [k for k in sheet["칸"] if int("".join(ch for ch in k if ch.isdigit())) >= 6]:
+        del sheet["칸"][a1]
+    sheet["병합"] = [
+        m for m in sheet["병합"] if int("".join(ch for ch in m.split(":")[0] if ch.isdigit())) < 6
+    ]
+    sheet["행"] = {k: v for k, v in sheet["행"].items() if int(k) < 6}
+
+    def put_row(r: int, styles: dict[str, dict]) -> None:
+        for c, cell in styles.items():
+            sheet["칸"][f"{c}{r}"] = copy.deepcopy(cell)
+
+    def name(rec: dict[str, Any]) -> str:
+        ch = float(rec["chainage_m"])
+        return rec.get("label") or labels.get(round(ch, 3)) or f"{ch:.2f}"
+
+    def where(rec: dict[str, Any]) -> str:
+        return (
+            f"{SIDE_NAMES[rec['side']]} {ROLE_NAMES.get(rec['role'], rec['role'])} · "
+            f"{CONDITION_NAMES.get(rec['condition'], rec['condition'])}"
+        )
+
+    defaults = load_defaults()
+    same = all(_is_default(rec, defaults) for rec in recs)
+    _set(sheet, "A3", mark("변동계수 · 분포 · 표본 = 문헌 기본값 후보(근거 E03 탭)", same))
+    pct = lambda v: None if v is None else 100.0 * float(v)  # noqa: E731
+    r = 6
+    for rec in recs:
+        res, st = rec["result"], rec["result"]["stats"]
+        put_row(r, tpl[6])
+        _set(sheet, f"A{r}", name(rec))
+        _set(sheet, f"B{r}", where(rec))
+        _set(
+            sheet,
+            f"C{r}",
+            f"{METHOD_NAMES.get(res['method'], res['method'])} · {MODE_NAMES[res['mode']]}",
+        )
+        _set(sheet, f"D{r}", int(st["n_valid"]))
+        _set(sheet, f"E{r}", rec.get("fs_deterministic"))
+        for c, k in zip("FG", ("mean", "std")):
+            _set(sheet, f"{c}{r}", st.get(k))
+        _set(sheet, f"H{r}", pct(st.get("pf")))
+        _set(sheet, f"I{r}", st.get("beta_normal"))
+        _set(sheet, f"J{r}", st.get("beta_lognormal"))
+        _set(sheet, f"K{r}", pct(st.get("p_below_required")))
+        r += 1
+    r += 1
+    put_row(r, tpl[8])
+    sheet["병합"].append(f"A{r}:K{r}")
+    sheet["행"][str(r)] = {"높이": 22}
+    sheet["행"][str(r + 1)] = {"높이": 30}
+    put_row(r + 1, tpl[9])
+    r += 2
+    for rec in recs:
+        res = rec["result"]
+        curves = {c["name"]: c for c in res["sensitivity"]["curves"]}
+        rank = {t["name"]: i + 1 for i, t in enumerate(res["sensitivity"]["tornado"])}
+        for v in sorted(res["variables"], key=lambda v: rank.get(v["name"], 99)):
+            c = curves.get(v["name"]) or {}
+            xs = c.get("x") or [None]
+            put_row(r, tpl[10])
+            _set(sheet, f"A{r}", f"{name(rec)} · {where(rec)}")
+            _set(sheet, f"B{r}", v["name"])
+            _set(sheet, f"C{r}", DIST_NAMES.get(v["dist"], v["dist"]))
+            _set(sheet, f"D{r}", float(v["mean"]))
+            _set(sheet, f"E{r}", float(v["std"]))
+            _set(sheet, f"F{r}", v["lo"] / v["std"] if v.get("lo") is not None else "-")
+            _set(sheet, f"G{r}", xs[0])
+            _set(sheet, f"H{r}", c.get("F_lo"))
+            _set(sheet, f"I{r}", xs[-1])
+            _set(sheet, f"J{r}", c.get("F_hi"))
+            _set(sheet, f"K{r}", rank.get(v["name"]))
+            r += 1
diff --git a/config/config_system_terrain.py b/config/config_system_terrain.py
index 113e576e5..62d543a72 100644
--- a/config/config_system_terrain.py
+++ b/config/config_system_terrain.py
@@ -149,6 +149,9 @@ SHEET_SURFACE_METHODS = [
     ).split(",")
     if method.strip()
 ]
+# D 예정노선 [분석] 연도 구간을 나란히 셀 프로세스 수 상한(프로세스 하나 약 0.8 GB) — 여럿이 동시에
+# 돌리면 줄임.
+PLAN_ANALYZE_WORKERS = int(os.getenv("PLAN_ANALYZE_WORKERS", "3"))
 # 확정에 쓸 기본 방식 — 라플라스 (2026-08-30 사용자 확정).
 SHEET_SURFACE_DEFAULT_METHOD = os.getenv("SHEET_SURFACE_DEFAULT_METHOD", "laplace")
 # 자동 전처리가 만드는 도엽 방식 — 기본 하나뿐이다. 여섯 방식을 매번 다 만들면
diff --git a/main.py b/main.py
index a1ce11a71..cd8ab38f4 100644
--- a/main.py
+++ b/main.py
@@ -86,6 +86,7 @@ from E05_StabReport.E05_StabReport_Router import router as e05_stab_report_route
 from E07_StabArea.E07_StabArea_Router import router as e07_stab_area_router
 from E03_StabSection.E03_StabSection_Router import router as e03_stab_section_router
 from E03_StabSection.E03_StabSection_Router_WhatIf import router as e03_stab_whatif_router
+from E03_StabSection.E03_StabSection_Router_Prob import router as e03_stab_prob_router
 from E06_StabRock.E06_StabRock_Router import router as e06_stab_rock_router
 from M01_MasterData.M01_MasterData_Router import router as m01_master_data_router
 from M02_MasterTemplete.M02_Labor_Router import router as m02_labor_router
@@ -499,6 +500,7 @@ app.include_router(e05_stab_report_router, dependencies=protected_with_company)
 app.include_router(e07_stab_area_router, dependencies=protected_with_company)
 app.include_router(e03_stab_section_router, dependencies=protected_with_company)
 app.include_router(e03_stab_whatif_router, dependencies=protected_with_company)
+app.include_router(e03_stab_prob_router, dependencies=protected_with_company)
 app.include_router(e06_stab_rock_router, dependencies=protected_with_company)
 app.include_router(b05_corridor_router, dependencies=protected_with_company)
 app.include_router(b05_route_replan_router, dependencies=protected_with_company)
diff --git a/resources/master_template/report/stab_report.json b/resources/master_template/report/stab_report.json
index 1b9d49ef2..1136f7c7d 100644
--- a/resources/master_template/report/stab_report.json
+++ b/resources/master_template/report/stab_report.json
@@ -4133,6 +4133,452 @@
         }
       }
     },
+    {
+      "id": "s12",
+      "이름": "확률해석",
+      "칸": {
+        "A1": {
+          "서식": 1,
+          "값": "확률 해석 · 민감도 (대표 단면)"
+        },
+        "B1": {
+          "서식": 1
+        },
+        "C1": {
+          "서식": 1
+        },
+        "D1": {
+          "서식": 1
+        },
+        "E1": {
+          "서식": 1
+        },
+        "F1": {
+          "서식": 1
+        },
+        "G1": {
+          "서식": 1
+        },
+        "H1": {
+          "서식": 1
+        },
+        "I1": {
+          "서식": 1
+        },
+        "J1": {
+          "서식": 1
+        },
+        "K1": {
+          "서식": 1
+        },
+        "A2": {
+          "서식": 2,
+          "값": "물성 = 확률 변수(문헌 추정 변동계수) · Pf = P(Fs < 1) · β 정규 = (μ − 1)/σ · β 대수정규 = ln(μ/√(1+V²))/√ln(1+V²) — Slide2 · SLOPE/W Probabilistic 과 같은 정의"
+        },
+        "B2": {
+          "서식": 2
+        },
+        "C2": {
+          "서식": 2
+        },
+        "D2": {
+          "서식": 2
+        },
+        "E2": {
+          "서식": 2
+        },
+        "F2": {
+          "서식": 2
+        },
+        "G2": {
+          "서식": 2
+        },
+        "H2": {
+          "서식": 2
+        },
+        "I2": {
+          "서식": 2
+        },
+        "J2": {
+          "서식": 2
+        },
+        "K2": {
+          "서식": 2
+        },
+        "A3": {
+          "서식": 12
+        },
+        "B3": {
+          "서식": 12
+        },
+        "C3": {
+          "서식": 12
+        },
+        "D3": {
+          "서식": 12
+        },
+        "E3": {
+          "서식": 12
+        },
+        "F3": {
+          "서식": 12
+        },
+        "G3": {
+          "서식": 12
+        },
+        "H3": {
+          "서식": 12
+        },
+        "I3": {
+          "서식": 12
+        },
+        "J3": {
+          "서식": 12
+        },
+        "K3": {
+          "서식": 12
+        },
+        "A4": {
+          "서식": 3,
+          "값": "1. 측점별 결과 — 안전율 분포 · 파괴확률 · 신뢰도 지수"
+        },
+        "B4": {
+          "서식": 3
+        },
+        "C4": {
+          "서식": 3
+        },
+        "D4": {
+          "서식": 3
+        },
+        "E4": {
+          "서식": 3
+        },
+        "F4": {
+          "서식": 3
+        },
+        "G4": {
+          "서식": 3
+        },
+        "H4": {
+          "서식": 3
+        },
+        "I4": {
+          "서식": 3
+        },
+        "J4": {
+          "서식": 3
+        },
+        "K4": {
+          "서식": 3
+        },
+        "A5": {
+          "서식": 6,
+          "값": "측점"
+        },
+        "B5": {
+          "서식": 6,
+          "값": "비탈 · 조건"
+        },
+        "C5": {
+          "서식": 6,
+          "값": "방법 · 임계면"
+        },
+        "D5": {
+          "서식": 6,
+          "값": "표본"
+        },
+        "E5": {
+          "서식": 6,
+          "값": "결정론 Fs"
+        },
+        "F5": {
+          "서식": 6,
+          "값": "Fs 평균"
+        },
+        "G5": {
+          "서식": 6,
+          "값": "표준편차"
+        },
+        "H5": {
+          "서식": 6,
+          "값": "Pf %"
+        },
+        "I5": {
+          "서식": 6,
+          "값": "β 정규"
+        },
+        "J5": {
+          "서식": 6,
+          "값": "β 대수정규"
+        },
+        "K5": {
+          "서식": 6,
+          "값": "P(F<기준) %"
+        },
+        "A6": {
+          "서식": 7
+        },
+        "B6": {
+          "서식": 7
+        },
+        "C6": {
+          "서식": 5
+        },
+        "D6": {
+          "서식": 9
+        },
+        "E6": {
+          "서식": 8
+        },
+        "F6": {
+          "서식": 8
+        },
+        "G6": {
+          "서식": 8
+        },
+        "H6": {
+          "서식": 8
+        },
+        "I6": {
+          "서식": 8
+        },
+        "J6": {
+          "서식": 8
+        },
+        "K6": {
+          "서식": 8
+        },
+        "A8": {
+          "서식": 3,
+          "값": "2. 확률 변수 · 민감도 — 토네이도 순(나머지 평균 · 임계면 고정 · 범위 끝 Fs)"
+        },
+        "B8": {
+          "서식": 3
+        },
+        "C8": {
+          "서식": 3
+        },
+        "D8": {
+          "서식": 3
+        },
+        "E8": {
+          "서식": 3
+        },
+        "F8": {
+          "서식": 3
+        },
+        "G8": {
+          "서식": 3
+        },
+        "H8": {
+          "서식": 3
+        },
+        "I8": {
+          "서식": 3
+        },
+        "J8": {
+          "서식": 3
+        },
+        "K8": {
+          "서식": 3
+        },
+        "A9": {
+          "서식": 6,
+          "값": "측점 · 비탈 · 조건"
+        },
+        "B9": {
+          "서식": 6,
+          "값": "변수"
+        },
+        "C9": {
+          "서식": 6,
+          "값": "분포"
+        },
+        "D9": {
+          "서식": 6,
+          "값": "평균"
+        },
+        "E9": {
+          "서식": 6,
+          "값": "표준편차"
+        },
+        "F9": {
+          "서식": 6,
+          "값": "절단 ±σ"
+        },
+        "G9": {
+          "서식": 6,
+          "값": "아래 끝"
+        },
+        "H9": {
+          "서식": 6,
+          "값": "Fs(아래)"
+        },
+        "I9": {
+          "서식": 6,
+          "값": "위 끝"
+        },
+        "J9": {
+          "서식": 6,
+          "값": "Fs(위)"
+        },
+        "K9": {
+          "서식": 6,
+          "값": "순위"
+        },
+        "A10": {
+          "서식": 7
+        },
+        "B10": {
+          "서식": 5
+        },
+        "C10": {
+          "서식": 7
+        },
+        "D10": {
+          "서식": 8
+        },
+        "E10": {
+          "서식": 8
+        },
+        "F10": {
+          "서식": 16
+        },
+        "G10": {
+          "서식": 8
+        },
+        "H10": {
+          "서식": 8
+        },
+        "I10": {
+          "서식": 8
+        },
+        "J10": {
+          "서식": 8
+        },
+        "K10": {
+          "서식": 9
+        },
+        "A12": {
+          "서식": 5,
+          "값": "확률 해석 없음 — E03 [확률 · 민감도] 뒤 [보고서 갱신]"
+        },
+        "B12": {
+          "서식": 5
+        },
+        "C12": {
+          "서식": 5
+        },
+        "D12": {
+          "서식": 5
+        },
+        "E12": {
+          "서식": 5
+        },
+        "F12": {
+          "서식": 5
+        },
+        "G12": {
+          "서식": 5
+        },
+        "H12": {
+          "서식": 5
+        },
+        "I12": {
+          "서식": 5
+        },
+        "J12": {
+          "서식": 5
+        },
+        "K12": {
+          "서식": 5
+        }
+      },
+      "병합": [
+        "A1:K1",
+        "A2:K2",
+        "A3:K3",
+        "A4:K4",
+        "A8:K8",
+        "A12:K12"
+      ],
+      "열": {
+        "A": {
+          "폭": 13
+        },
+        "B": {
+          "폭": 16
+        },
+        "C": {
+          "폭": 18
+        },
+        "D": {
+          "폭": 7
+        },
+        "E": {
+          "폭": 8
+        },
+        "F": {
+          "폭": 8
+        },
+        "G": {
+          "폭": 8
+        },
+        "H": {
+          "폭": 8
+        },
+        "I": {
+          "폭": 8
+        },
+        "J": {
+          "폭": 8
+        },
+        "K": {
+          "폭": 8
+        }
+      },
+      "행": {
+        "1": {
+          "높이": 36
+        },
+        "2": {
+          "높이": 30
+        },
+        "3": {
+          "높이": 22
+        },
+        "4": {
+          "높이": 22
+        },
+        "5": {
+          "높이": 30
+        },
+        "8": {
+          "높이": 22
+        },
+        "9": {
+          "높이": 30
+        }
+      },
+      "보기": {
+        "눈금선": false
+      },
+      "인쇄": {
+        "용지": "A4",
+        "방향": "landscape",
+        "맞춤": {
+          "폭": 1,
+          "높이": 0
+        },
+        "여백": {
+          "위": 15,
+          "아래": 15,
+          "왼": 10,
+          "오른": 10
+        }
+      }
+    },
     {
       "id": "s10",
       "이름": "부록_절편",
diff --git a/resources/master_template/report/stab_report_slots.md b/resources/master_template/report/stab_report_slots.md
index b566cf62d..47d2cf4d9 100644
--- a/resources/master_template/report/stab_report_slots.md
+++ b/resources/master_template/report/stab_report_slots.md
@@ -4,8 +4,8 @@
 
 ## 공통
 
-- 몫: 사용자 몫(표지 · 개요) = 처음 만들 때만 채움 · 다시 만들 때(보고서 갱신) 안 건드림. 프로그램 몫(목차 · 입력조건 · 측점표 · 대표단면 · 3D요약 · 광역요약 · 보강검토 · 암반검토 · 부록_절편) = 갱신 때 통째 다시 씀 · 작업본에 없는 프로그램 몫 시트는 양식 차례 자리에 끼움 · 새 서식은 작업본 서식 표 끝에 붙여 번호를 바꿈.
-- 시트 차례: 표지 · 목차 · 개요 · 입력조건 · 측점표 · 대표단면 · 3D요약 · 광역요약 · (보강검토) · (암반검토) · 부록_절편.
+- 몫: 사용자 몫(표지 · 개요) = 처음 만들 때만 채움 · 다시 만들 때(보고서 갱신) 안 건드림. 프로그램 몫(목차 · 입력조건 · 측점표 · 대표단면 · 3D요약 · 광역요약 · 보강검토 · 암반검토 · 확률해석 · 부록_절편) = 갱신 때 통째 다시 씀 · 작업본에 없는 프로그램 몫 시트는 양식 차례 자리에 끼움 · 새 서식은 작업본 서식 표 끝에 붙여 번호를 바꿈.
+- 시트 차례: 표지 · 목차 · 개요 · 입력조건 · 측점표 · 대표단면 · 3D요약 · 광역요약 · (보강검토) · (암반검토) · (확률해석) · 부록_절편.
 - PDF: 표지 = 쪽 번호 · 머리말 없음 · 본문 = 머리말(왼 사업명 = 표지 `C8` · 오른 보고서 이름) · 꼬리말 「- n / N -」(목차 = 1 쪽).
 - 사용자 확정 전 표기 한 규칙 — 기본값 그대로 = 「(기본값 · 사용자 확정 전)」 · 바꿈 = 「(사용자 입력)」 · 물성 = 「문헌 추정치 · 사용자 확정 전」.
 - 서식 번호(`서식` 표): 1 title · 2 sub · 3 section · 4 label · 5 text · 6 head · 7 body(가운데 글) · 8 fs(0.00) · 9 int(0) · 10 ng(N.G 빨강 채움) · 11 ok(O.K 초록) · 12 notice · 13 num1(#,##0.0) · 14 num0(#,##0) · 15 para(위 맞춤 긴 글) · 16 prop(0.0) · 17 chain(#,##0.00) · 18 head8 · 19 num8(0.00) · 20 num8(0.0) · 21 int8 · 22 목차 내용 · 23 목차 쪽 · 24 목차 장 · 25 sum8(0.00 · 회색) · 26 표지 제목.
@@ -126,6 +126,16 @@
 | 11 · 12 행 꼴(제목 3 · 머리 6) 다음 절리군 하나 = 한 줄(13 행 서식) | A 이름(7) · B 경사방향 · C 경사 · D φj · E cj(16) · F 지속성(13) · G · H 구간 시작 · 끝(7) · 모르면 「-」 · 없으면 「절리군 없음」 |
 | 절리 줄 다음 한 줄 띄고 15 · 16 행 꼴 · 그 아래 측점 비탈 × 위험 모드 × 조건 = 한 줄(17 행 서식) | A 측점(7) · B 비탈 글(쪽 · 경사 / 방향 · 높이)(5) · C 파괴 형태(평면 · 쐐기 · 전도 + 절리)(7) · D 조건(7) · E Fs · F 기준(8) · G 판정(7 → 10 · 11 · 전도 「검토」) · H 비고(블록 안 생김 · 미끄러짐 면 · 지속성)(5) · 위험 모드 없는 비탈 = 「없음」 한 줄 |
 
+## s12 확률해석 (프로그램 몫 · E03 확률 해석 저장본 · 없으면 시트를 뺌 · 가로 쪽)
+
+| 칸 | 내용 |
+| --- | --- |
+| `A3` (A:K 병합 · 서식 12) | 「변동계수 · 분포 · 표본 = 문헌 기본값 후보(근거 E03 탭)」 + 사용자 확정 전 표기(모든 기록이 `prob_defaults.json` 그대로면 기본값 · 아니면 사용자 입력) |
+| 4 · 5 행 꼴(제목 3 · 머리 6) 다음 (측점 · 비탈 · 조건)마다 한 줄(6 행 서식) | A 측점(7) · B 비탈 · 조건(7) · C 방법 · 임계면 고정 / 매회 탐색(5) · D 유효 표본(9) · E 결정론 Fs · F 평균 · G 표준편차 · H Pf % · I β 정규 · J β 대수정규 · K P(F < 기준) %(8) |
+| 한 줄 띄고 8 · 9 행 꼴 · 그 아래 기록마다 변수 = 한 줄(10 행 서식 · 토네이도 순) | A 측점 · 비탈 · 조건(7) · B 변수(5) · C 분포(7) · D 평균 · E 표준편차(8) · F 절단 ±σ(16 · 없으면 「-」) · G 아래 끝 · H 그때 Fs · I 위 끝 · J 그때 Fs(8) · K 순위(9) |
+
+- 저장본 = (측점 · 비탈 · 조건)마다 가장 최근 하나 · 지금 2D 결과 측점 지문에 붙은 것만.
+
 ## s10 부록_절편 (프로그램 몫 · 대표 단면마다 한 쪽)
 
 양식 1~5 행 = 덩이 꼴(1 제목 · 2 단위 · 3 머리 · 4 절편 줄 서식 · 5 합계 서식). 대표 단면마다 지배 조건(Fs / 기준 가장 낮음) 임계면의 절편표.
diff --git a/resources/master_template/stability/prob_defaults.json b/resources/master_template/stability/prob_defaults.json
new file mode 100644
index 000000000..8d3a1119b
--- /dev/null
+++ b/resources/master_template/stability/prob_defaults.json
@@ -0,0 +1,56 @@
+{
+  "version": 1,
+  "note": "확률 해석 기본값 후보 — 문헌 · 상용 SW 예제 값 · 현장 시험 없음 · 사용자 확정 전(E03 확률 · 민감도 탭에서 바꿈)",
+  "run": {
+    "n": 1000,
+    "sampling": "lhs",
+    "seed": 0,
+    "mode": "fixed",
+    "method": "bishop"
+  },
+  "properties": {
+    "c": {
+      "dist": "lognormal",
+      "cov": 0.3,
+      "trunc_sigma": 3.0
+    },
+    "phi": {
+      "dist": "normal",
+      "cov": 0.1,
+      "trunc_sigma": 3.0
+    },
+    "gamma": {
+      "dist": "normal",
+      "cov": 0.05,
+      "trunc_sigma": 3.0
+    },
+    "water": {
+      "dist": "normal",
+      "std": 0.0,
+      "trunc_sigma": 3.0
+    },
+    "ru": {
+      "dist": "normal",
+      "std": 0.0,
+      "trunc_sigma": 3.0
+    }
+  },
+  "rho_c_phi": 0.0,
+  "sources": {
+    "c": "COV 0.30(범위 0.2~0.5) · 대수정규(음수 막음) — 국내 화강풍화토 실측 정남수 외(2011) 0.34~0.49 · 장정윤 · 박혁진(2017) 0.21 · 해석 가정 Cho(2018 · 2019) 0.20 · Lee 외(2020) 0.25 · ISSMGE TC304(2021) 홍콩 완전풍화화강암 0.51 · Harr(1987) 최대 0.40 · 붕적토 0.40 = 국내 자료 없어 국외 범위 위쪽(협의 필요)",
+    "phi": "COV 0.10(범위 0.05~0.15) · 정규 — 국내 화강풍화토 실측 정남수 외(2011) 0.09~0.12 · 장정윤 · 박혁진(2017) 0.03 · Cho(2018 · 2019) · Lee 외(2020) 0.10 · TC304(2021) 모래 평균 0.079 · 홍콩 CDG 0.096 · Duncan(2000) 2~13 %",
+    "gamma": "COV 0.05(범위 0.03~0.07) · 정규 — Duncan(2000) 3~7 % · 홍콩 CDG 2.6 % · Li & Lumb(1987) 5 % · 국내 논문은 고정값으로 둠",
+    "water": "지하수위 이동 표준편차 기본 0(확률 변수 아님) · 수위선 갈래(E02)일 때만",
+    "ru": "r_u 표준편차 기본 0 · r_u 갈래(E02)일 때만 · Li & Lumb(1987) COV 10 %",
+    "rho_c_phi": "기본 0(상관 무시 = 보수적) · Slide2 · Cho(2019) 기본 −0.5 · 문헌 −0.25~−0.75(Wolff 1985 −0.47 · Lumb 1970 −0.7~−0.37)",
+    "trunc_sigma": "평균 ±3σ 절단 — Slide2 「3x 표준편차」 · SLOPE/W James Bay 예제",
+    "run": "LHS 1000 · 임계면 고정(Slide2 Global Minimum) · Bishop — Slide2 Tutorial 08"
+  },
+  "materials": {
+    "colluvium": {
+      "c": {
+        "cov": 0.4
+      }
+    }
+  }
+}
diff --git a/resources/tester/test_530_2_plan_analyze.py b/resources/tester/test_530_2_plan_analyze.py
index 1e50b9275..cbab24f0f 100644
--- a/resources/tester/test_530_2_plan_analyze.py
+++ b/resources/tester/test_530_2_plan_analyze.py
@@ -17,8 +17,7 @@ SEGMENTS = [
 ]
 
 
-@pytest.fixture(scope="module")
-def hill() -> Terrain:
+def _hill() -> Terrain:
     x = np.arange(0.0, 1102.0, 2.0)
     y = np.arange(0.0, 402.0, 2.0)
     xx, yy = np.meshgrid(x, y)
@@ -26,6 +25,16 @@ def hill() -> Terrain:
     return Terrain(x, y, z, np.ones(xx.shape, dtype=bool))
 
 
+def hill_loader(root, selection, models_dir, cache_dir):
+    """나란히 셈 프로세스가 이름으로 부르는 지형(모듈 맨 위 함수)."""
+    return _hill(), []
+
+
+@pytest.fixture(scope="module")
+def hill() -> Terrain:
+    return _hill()
+
+
 @pytest.fixture
 def run(tmp_path, hill):
     _POOLS.clear()
@@ -37,7 +46,7 @@ def run(tmp_path, hill):
 
     def go(criteria=CRITERIA, corridor_m=150.0):
         return analyze.analyze_plan(
-            tmp_path, SEGMENTS, criteria, loader=loader, corridor_m=corridor_m
+            tmp_path, SEGMENTS, criteria, loader=loader, corridor_m=corridor_m, workers=1
         )
 
     go.calls = calls
@@ -91,6 +100,54 @@ def test_same_input_reads_store(run, monkeypatch) -> None:
         run(criteria={**CRITERIA, "max_grade": 0.12})
 
 
+def test_parallel_years_match_sequential(tmp_path) -> None:
+    """연도 구간 나란히(프로세스 둘) = 차례로 — 갈래 넷 노선 좌표 · 잰 값 차 0(531-6)."""
+    _POOLS.clear()
+    seq = analyze.analyze_plan(
+        tmp_path / "a", SEGMENTS, CRITERIA, loader=hill_loader, corridor_m=150.0, workers=1
+    )
+    _POOLS.clear()
+    steps = []
+    par = analyze.analyze_plan(
+        tmp_path / "b",
+        SEGMENTS,
+        CRITERIA,
+        loader=hill_loader,
+        corridor_m=150.0,
+        workers=2,
+        on_progress=lambda p, m: steps.append(m),
+    )
+    assert [r["routes"] for r in par] == [r["routes"] for r in seq]
+    assert [r["fingerprint"] for r in par] == [r["fingerprint"] for r in seq]
+    assert "나란히 2" in steps[0] and steps[-1] == "분석 끝"
+    # 저장본은 제 프로세스에서 지문만 보고 그대로
+    again = analyze.analyze_plan(
+        tmp_path / "b", SEGMENTS, CRITERIA, loader=hill_loader, corridor_m=150.0, workers=2
+    )
+    assert [r["stored"] for r in again] == [True, True]
+    _POOLS.clear()
+
+
+def test_near_streams_keeps_every_nearest_line(hill) -> None:
+    """계류 거르기 — 지형 상자 안 점의 계류 거리 · 건너기가 통째와 같고 먼 선은 빠짐(531-6)."""
+    import shapely
+    from shapely.geometry import LineString, MultiLineString
+
+    rng = np.random.default_rng(7)
+    streams = []
+    for _ in range(80):  # 지형(1.1 × 0.4 km) 둘레 5 km 에 흩은 계류
+        start = rng.uniform(-2000.0, 3000.0, 2)
+        streams.append([tuple(start + step) for step in np.cumsum(rng.normal(0, 60, (6, 2)), 0)])
+    kept = analyze.near_streams(hill, streams)
+    assert 0 < len(kept) < len(streams)
+    points = shapely.points(rng.uniform((0.0, 0.0), (1100.0, 400.0), (4000, 2)))
+    whole, part = MultiLineString(streams), MultiLineString(kept)
+    assert np.array_equal(shapely.distance(points, whole), shapely.distance(points, part))
+    route = LineString([(0.0, 0.0), (1100.0, 400.0)])
+    assert route.intersection(whole).equals(route.intersection(part))
+    assert analyze.near_streams(hill, []) == []
+
+
 def test_corridor_argument_keeps_b_default(hill) -> None:
     line = [SEGMENTS[0]["start"], SEGMENTS[0]["end"]]
     base = SearchGrid(line, hill, [], stream_offset_m=10.0)
diff --git a/resources/tester/test_534_3_stab_report_form.py b/resources/tester/test_534_3_stab_report_form.py
index 4a4bb2c51..cc625b2bc 100644
--- a/resources/tester/test_534_3_stab_report_form.py
+++ b/resources/tester/test_534_3_stab_report_form.py
@@ -38,7 +38,7 @@ PRINT_KEYS = {"용지", "방향", "배율", "맞춤", "여백", "머리행", "
 A1 = re.compile(r"^([A-Z]{1,3})(\d+)$")
 NAMES = [
     "표지", "목차", "개요", "입력조건", "측점표", "대표단면", "3D요약", "광역요약", "보강검토", "암반검토",
-    "부록_절편",
+    "확률해석", "부록_절편",
 ]  # fmt: skip
 
 needs_node = pytest.mark.skipif(shutil.which("node") is None, reason="node 가 없음")
@@ -151,7 +151,7 @@ def test_양식_계약():
     book = _form()
     _contract(book)
     assert [s["이름"] for s in book["시트"]] == NAMES
-    ids = ["s1", "s9", "s2", "s3", "s4", "s5", "s6", "s7", "s8", "s11", "s10"]
+    ids = ["s1", "s9", "s2", "s3", "s4", "s5", "s6", "s7", "s8", "s11", "s12", "s10"]
     assert [s["id"] for s in book["시트"]] == ids
 
 
@@ -175,6 +175,7 @@ def test_양식_인쇄_쪽나눔_머리행_그림():
         "광역요약": 0,
         "보강검토": 0,
         "암반검토": 0,
+        "확률해석": 0,
     }
     # N.G 빨강 조건부 서식 — 측점표 판정 열 · 서식 표에 N.G 칸 서식(빌더가 인쇄 · xlsx 용으로 직접 씀)
     assert by["측점표"]["조건부서식"][0]["범위"][0]["c0"] == 9
@@ -239,9 +240,9 @@ def test_견본_pdf_쪽수_머리행(png):
     sections = [i for i, t in enumerate(texts) if "대표 단면" in t and "안전율 Fs" in t]
     assert len(sections) == 2  # 단면마다 한 쪽
     # 표지 1 + 목차 1 + 개요 1 + 입력조건 1 + 측점표 n + 대표단면 2 + 3D 1 + 광역 1 + 보강 1 + 암반 1
-    # + 부록 1
+    # + 확률 1 + 부록 1
     # (양식 빈 판)
-    assert len(pages) == 1 + 1 + 1 + 1 + len(stab) + 2 + 1 + 1 + 1 + 1 + 1
+    assert len(pages) == 1 + 1 + 1 + 1 + len(stab) + 2 + 1 + 1 + 1 + 1 + 1 + 1
     assert all(float(pages[i].mediabox.width) > float(pages[i].mediabox.height) for i in stab)
     html = spreadsheet.print_htmls([{"book": book, "figures": {k: png for k in figs}}])[0]
     assert html.count("<img ") == 3  # 단면 2 + 3D 띠 1
diff --git a/resources/tester/test_537_1_e03_prob.py b/resources/tester/test_537_1_e03_prob.py
new file mode 100644
index 000000000..af806ae21
--- /dev/null
+++ b/resources/tester/test_537_1_e03_prob.py
@@ -0,0 +1,227 @@
+# -*- coding: utf-8 -*-
+"""537-1 E03 2D 확률 해석 · 민감도 — 엔진(`E03_StabSection_Engine_Prob`).
+
+- 무한사면 해석해: F = c/(γ z sinβ cosβ) + (1 − m γw/γ) tanφ/tanβ · c · φ 정규(MIT OCW 1.151 App. 17 ·
+  z 5 · m 0.5 · β 35° · c N(25, 5) · φ N(30°, 7.5°)) — Pf · μ · σ 정확값 = φ 위 수치 적분 · LHS 1만 ±2 %
+  · c · tanφ 정규면 F 도 정규 → Pf = Φ(−β) · 절편 경로(임계면 고정 · 건조 무한사면) = 닫힌 꼴.
+- Slide2 Tutorial 08(Probabilistic Analysis · LHS 1000 · Global Minimum · Bishop): FS 평균 1.145 · σ 0.1206
+  · PF 10.2 % · RI 정규 1.204 · 대수정규 1.238.
+- Slide2 Verification #36(Li & Lumb 1987 · Hassan & Wolff 1999 · r_u): 결정론 1.340(H&W 1.334) · 평균 1.355 ·
+  PF 1.5 % · RI 대수정규 2.482(H&W 2.336).
+- 고정 시드 재현 · LHS 층 · 절단 · 대수정규 평균 · 민감도(평균 = 결정론 F · 토네이도 순) · 매회 탐색 갈래.
+
+    ./venv/Scripts/python.exe -m pytest resources/tester/test_537_1_e03_prob.py -q
+"""
+
+from __future__ import annotations
+
+import math
+import sys
+from pathlib import Path
+
+import numpy as np
+import pytest
+from scipy import integrate
+from scipy.stats import norm
+
+ROOT = Path(__file__).resolve().parents[2]
+sys.path.insert(0, str(ROOT))
+
+from E03_StabSection import E03_StabSection_Engine_Prob as prob  # noqa: E402
+from E03_StabSection.E03_StabSection_Engine_Search import analyze  # noqa: E402
+
+
+def var(name, prop, mean, std, rel=None, mat="s", dist="normal"):
+    return {
+        "name": name, "material": mat, "prop": prop, "dist": dist,
+        "mean": mean, "std": std, "lo": rel, "hi": rel,
+    }  # fmt: skip
+
+
+def near(v, ref, tol):
+    assert v is not None and abs(v - ref) <= tol * abs(ref), (v, ref)
+
+
+# ---- 무한사면 해석해 ---------------------------------------------------------------
+Z, BETA, G, GW = 5.0, math.radians(35.0), 20.0, 9.81
+C_PHI = [var("c", "c", 25.0, 5.0), var("phi", "phi", 30.0, 7.5)]
+
+
+def _ab(m):
+    return 1 / (G * Z * math.sin(BETA) * math.cos(BETA)), (1 - m * GW / G) / math.tan(BETA)
+
+
+def _exact(m):
+    """φ 위 수치 적분 — Pf = ∫ Φ((1 − b tanφ − a μc)/(a σc)) f(φ) dφ · 평균 · 표준편차."""
+    a, b = _ab(m)
+    f = lambda p: norm.pdf(p, 30.0, 7.5)  # noqa: E731
+    F = lambda p: a * 25.0 + b * math.tan(math.radians(p))  # noqa: E731
+    pf = integrate.quad(lambda p: norm.cdf((1 - F(p)) / (a * 5.0)) * f(p), 0, 89, limit=200)[0]
+    mu = integrate.quad(lambda p: F(p) * f(p), 0, 89)[0]
+    m2 = integrate.quad(lambda p: (F(p) ** 2 + (a * 5.0) ** 2) * f(p), 0, 89)[0]
+    return pf, mu, math.sqrt(m2 - mu * mu)
+
+
+def test_infinite_slope_closed_form_pf_beta():
+    """MIT App. 17 무한사면 — LHS 1만 Pf · β ±2 % · 평균 · 표준편차 ±1 %(φ 정규 · 각도)."""
+    a, b = _ab(0.5)
+    X = prob.sample(C_PHI, 10000, seed=0)
+    s = prob.statistics(a * X[:, 0] + b * np.tan(np.radians(X[:, 1])))
+    pf, mu, sd = _exact(0.5)
+    near(pf, 0.2252, 0.01)  # 같은 문헌 MC 2×10⁶
+    near(s["pf"], pf, 0.02)
+    near(s["beta_normal"], (mu - 1) / sd, 0.02)
+    near(s["mean"], mu, 0.01)
+    near(s["std"], sd, 0.01)
+    V = sd / mu
+    near(
+        s["beta_lognormal"],
+        math.log(mu / math.sqrt(1 + V * V)) / math.sqrt(math.log1p(V * V)),
+        0.02,
+    )
+
+
+def test_infinite_slope_normal_tanphi_exact():
+    """c · tanφ 정규 · 독립 → F 정규 → Pf = Φ(−β) 정확(LHS 2만 ±2 %)."""
+    a, b = _ab(0.5)
+    vs = [var("c", "c", 25.0, 5.0), {**var("t", "c", 0.57735, 0.11547), "prop": "x"}]
+    X = prob.sample(vs, 20000, seed=0)
+    s = prob.statistics(a * X[:, 0] + b * X[:, 1])
+    mu = a * 25.0 + b * 0.57735
+    sd = math.hypot(a * 5.0, b * 0.11547)
+    near(mu, 1.154412, 1e-5)
+    near(sd, 0.163756, 1e-4)
+    near(s["beta_normal"], (mu - 1) / sd, 0.02)
+    near(s["pf"], norm.cdf(-(mu - 1) / sd), 0.02)
+
+
+def _infinite_model():
+    """건조 무한사면 β 35° · 지표 아래 5 m 나란한 면 · 모형 양 끝까지(끝 효과 없음)."""
+    t, L = math.tan(BETA), 100.0
+    model = {
+        "materials": [{"id": "s", "name": "s", "gamma_t": G, "c": 25.0, "phi": 30.0}],
+        "regions": [{"material_id": "s", "polygon": [[0, 0], [L, -t * L], [L, -t * L - 20], [0, -20]]}],
+    }  # fmt: skip
+    return model, {"kind": "polyline", "points": [[0, -Z], [L, -t * L - Z]]}
+
+
+@pytest.mark.parametrize("method", ["ordinary", "bishop", "janbu"])
+def test_fixed_surface_slices_equal_closed_form(method):
+    """임계면 고정 절편 경로 = 무한사면 닫힌 꼴(물성 바꾼 줄마다) · LHS 1만 Pf ±2 %."""
+    model, surface = _infinite_model()
+    fixed = prob.Fixed(model, surface, 1)
+    a, b = _ab(0.0)
+    X = np.array([[25.0, 30.0], [10.0, 20.0], [40.0, 35.0]])
+    got = fixed.F(C_PHI, X, method)
+    assert got == pytest.approx(a * X[:, 0] + b * np.tan(np.radians(X[:, 1])), rel=1e-9)
+    out = prob.run(model, surface=surface, direction=1, variables=C_PHI, n=10000, method=method)
+    near(out["stats"]["pf"], _exact(0.0)[0], 0.02)
+
+
+# ---- Slide2 대조 -------------------------------------------------------------------
+T08 = {
+    "materials": [{"id": "s", "name": "s", "gamma_t": 19.0, "c": 5.0, "phi": 30.0}],
+    "regions": [{"material_id": "s", "polygon": [
+        [0, 0], [130, 0], [130, 50], [80, 50], [50, 30], [0, 30]]}],
+}  # fmt: skip
+T08_VARS = [var("c", "c", 5, 1, 3), var("phi", "phi", 30, 3, 9), var("g", "gamma", 19, 0.5, 1.5)]
+
+
+def _surface(model, method="bishop"):
+    return analyze(model, direction=-1, noncircular=False, methods=(method,))["methods"][method][
+        "min"
+    ]
+
+
+def test_slide2_tutorial08_global_minimum():
+    """Tutorial 08 — LHS 1000 · 고정 시드 셋 평균: FS 평균 ±1 % · σ ±3 % · PF ±2 %p · RI ±3 %."""
+    surface = _surface(T08)
+    near(surface["F"], 1.142, 0.01)
+    runs = [
+        prob.run(T08, surface=surface, direction=-1, variables=T08_VARS, n=1000, seed=k)["stats"]
+        for k in range(3)
+    ]
+    avg = {
+        k: float(np.mean([r[k] for r in runs]))
+        for k in ("mean", "std", "pf", "beta_normal", "beta_lognormal")
+    }
+    near(avg["mean"], 1.145, 0.01)
+    near(avg["std"], 0.1206, 0.03)
+    assert abs(avg["pf"] - 0.102) <= 0.02
+    near(avg["beta_normal"], 1.204, 0.03)
+    near(avg["beta_lognormal"], 1.238, 0.03)
+
+
+def test_slide2_36_li_lumb_ru():
+    """#36 r_u 확률 변수 · MC 1만 — 결정론 ±1 % · 평균 ±1.5 % · PF ±1 %p · RI 대수정규 = Hassan & Wolff ±3 %."""
+    model = {
+        "materials": [{"id": "s", "name": "s", "gamma_t": 18.0, "c": 18.0, "phi": 30.0}],
+        "regions": [{"material_id": "s", "polygon": [[0, 0], [20, 0], [20, 15], [15, 15], [5, 5], [0, 5]]}],
+        "groundwater": {"mode": "ru", "ru": 0.2},
+    }  # fmt: skip
+    surface = _surface(model)
+    near(surface["F"], 1.340, 0.01)
+    vs = [
+        var("c", "c", 18, 3.6), var("phi", "phi", 30, 3), var("g", "gamma", 18, 0.9),
+        var("ru", "ru", 0.2, 0.02, mat=None),
+    ]  # fmt: skip
+    s = prob.run(model, surface=surface, direction=-1, variables=vs, n=10000, sampling="mc")[
+        "stats"
+    ]
+    near(s["mean"], 1.355, 0.015)
+    assert abs(s["pf"] - 0.015) <= 0.01
+    near(s["beta_lognormal"], 2.336, 0.03)
+
+
+# ---- 표본 · 재현 · 민감도 · 갈래 ---------------------------------------------------------
+def test_seed_reproducible_and_lhs_strata():
+    """같은 시드 = 같은 표본 · 다른 시드 = 다름 · LHS = 변수마다 n 칸에 하나씩 · 절단 안 · 상관."""
+    a = prob.sample(T08_VARS, 500, seed=7)
+    assert np.array_equal(a, prob.sample(T08_VARS, 500, seed=7))
+    assert not np.array_equal(a, prob.sample(T08_VARS, 500, seed=8))
+    for j, v in enumerate(T08_VARS):
+        lo, hi = v["mean"] - v["lo"], v["mean"] + v["hi"]
+        assert a[:, j].min() >= lo and a[:, j].max() <= hi
+        pa, pb = norm.cdf(-3), norm.cdf(3)
+        u = (norm.cdf((a[:, j] - v["mean"]) / v["std"]) - pa) / (pb - pa)
+        assert sorted(np.floor(u * 500).astype(int)) == list(range(500))
+    rho = prob.sample(C_PHI, 5000, seed=1, correlations=[(0, 1, -0.5)])
+    assert np.corrcoef(rho.T)[0, 1] == pytest.approx(-0.5, abs=0.03)
+
+
+def test_truncation_and_lognormal_moments():
+    """대수정규 = 평균 · σ 그대로 · 음수 없음 · 정규 c 는 물리 한계 0 에서 절단."""
+    ln = prob.sample([var("c", "c", 10.0, 5.0, dist="lognormal")], 20000, seed=2)[:, 0]
+    near(ln.mean(), 10.0, 0.01)
+    near(ln.std(), 5.0, 0.03)
+    assert ln.min() > 0
+    nc = prob.sample([var("c", "c", 5.0, 5.0)], 5000, seed=2)[:, 0]
+    assert nc.min() >= 0
+
+
+def test_sensitivity_and_tornado():
+    """민감도 — 평균 줄 F = 결정론 F · 곡선 51 점 · c ↑ = F ↑ · 토네이도 = 폭 큰 순."""
+    surface = _surface(T08)
+    out = prob.run(T08, surface=surface, direction=-1, variables=T08_VARS, n=200)
+    sens = out["sensitivity"]
+    assert sens["F_mean"] == pytest.approx(surface["F"], rel=1e-9)
+    curves = {c["name"]: c for c in sens["curves"]}
+    assert len(curves["c"]["F"]) == 51 and curves["c"]["x"][0] == pytest.approx(2.0)
+    assert np.all(np.diff(curves["c"]["F"]) > 0) and np.all(np.diff(curves["phi"]["F"]) > 0)
+    swings = [t["swing"] for t in sens["tornado"]]
+    assert swings == sorted(swings, reverse=True) and sens["tornado"][0]["name"] in ("c", "phi")
+    assert out["stats"]["n"] == 200 and len(out["F"]) == 200
+
+
+def test_search_mode_matches_fixed_at_low_variance():
+    """매회 탐색 — 변동이 아주 작으면 고정면과 같은 F(±1 %) · 표본 수 · 시간 칸."""
+    surface = _surface(T08)
+    tiny = [var("c", "c", 5, 0.01), var("phi", "phi", 30, 0.01)]
+    out = prob.run(T08, surface=surface, direction=-1, variables=tiny, n=4, mode="search")
+    for f in out["F"]:
+        near(f, surface["F"], 0.01)
+    assert out["mode"] == "search" and out["elapsed_s"]["samples"] > 0
+
+
+if __name__ == "__main__":
+    sys.exit(pytest.main([__file__, "-q"]))
diff --git a/resources/tester/test_537_1_e03_prob_route.py b/resources/tester/test_537_1_e03_prob_route.py
new file mode 100644
index 000000000..8ac57d404
--- /dev/null
+++ b/resources/tester/test_537_1_e03_prob_route.py
@@ -0,0 +1,295 @@
+# -*- coding: utf-8 -*-
+"""537-1 E03 확률 해석 길(`E03_StabSection_Router_Prob`) — 검증_sub4 동결 복사 · 측점 둘 · Bishop 원호만.
+
+- 준비(setup): 단면 물성 = E02 값 · 기본값 후보 · 기준안전율 · 결정론 F
+- 임계면 고정: 평균 줄 F = 저장된 결정론 F · Pf · β · 표본 수 · 같은 요청 = 저장본(다시 안 셈) · 기록 목록
+- 대표 단면 아님 · 단면에 없는 물성 · 수위선 아닌데 지하수위 변수 · 변수 없음 = 422
+- 매회 탐색 일꾼(프로세스 풀 · 진행 줄 · 결과 파일)
+- E05 확률해석 시트(있을 때만 · 결과 = 저장본 · 기본값 그대로 = 사용자 확정 전 · 바꾸면 사용자 입력 표기)
+- 화면 계산(logic/E03_StabSection_Prob.ts): 기본값 후보 · 물성 덮어쓰기 · 요청 꼴 = 서버 시험 요청
+
+    ./venv/Scripts/python.exe -m pytest resources/tester/test_537_1_e03_prob_route.py -q
+"""
+
+from __future__ import annotations
+
+import asyncio
+import json
+import shutil
+import subprocess
+import sys
+import tempfile
+from pathlib import Path
+from uuid import UUID
+
+import pytest
+from fastapi import HTTPException
+
+ROOT = Path(__file__).resolve().parents[2]
+sys.path.insert(0, str(ROOT))
+sys.path.insert(0, str(Path(__file__).resolve().parent))
+
+from common_util import common_util_project_copy as cp  # noqa: E402
+from E01_StabRegister import E01_StabRegister_Source as src  # noqa: E402
+from E02_StabGround import E02_StabGround_Router as e02  # noqa: E402
+from E03_StabSection import E03_StabSection_Router as router  # noqa: E402
+from E03_StabSection import E03_StabSection_Router_Prob as rp  # noqa: E402
+from E03_StabSection import E03_StabSection_Router_WhatIf as wf  # noqa: E402
+from E03_StabSection import E03_StabSection_View as view  # noqa: E402
+from E03_StabSection import E03_StabSection_Worker as worker  # noqa: E402
+from E03_StabSection.E03_StabSection_Engine_Station import analyze_station  # noqa: E402
+
+SOURCE_ID = "c725d747-50ae-46b0-87ce-4d4e56a53cf3"  # 검증_sub4(읽기만)
+PID = "d" * 8 + "-" + "d" * 4 + "-" + "d" * 4 + "-" + "d" * 4 + "-" + "d" * 12
+
+
+async def _copy(target: Path) -> None:
+    import aiomysql
+
+    from config.config_db import DB_HOST, DB_NAME, DB_PASSWORD, DB_PORT, DB_USER
+
+    connection = await aiomysql.connect(
+        host=DB_HOST, port=DB_PORT, user=DB_USER, password=DB_PASSWORD, db=DB_NAME, autocommit=True
+    )
+    try:
+        await cp.copy_project(
+            connection, SOURCE_ID, scope="stability", new_id=PID, project_root=target
+        )
+    finally:
+        connection.close()
+
+
+def _fast(cross, design, **kw):
+    return analyze_station(cross, design, **{**kw, "methods": ("bishop",), "noncircular": False})
+
+
+@pytest.fixture(scope="module")
+def project():
+    folder = Path(tempfile.mkdtemp(prefix="aislo_e03_prob_"))
+    try:
+        asyncio.run(_copy(folder / "copy"))
+    except Exception as error:  # noqa: BLE001
+        shutil.rmtree(folder, ignore_errors=True)
+        pytest.skip(f"검증 원본 복사 실패: {error}")
+    root = folder / "copy"
+    mp = pytest.MonkeyPatch()
+    original = src.stations
+    mp.setattr(router.src, "stations", lambda r: original(r)[20:22])
+
+    async def fake_project(_pid):
+        return root, {"company_id": 1, "project_kind": "STABILITY"}
+
+    mp.setattr(e02, "_project", fake_project)
+    mp.setattr(worker, "analyze_station", _fast)
+    mp.setattr(wf, "analyze_station", _fast)
+    mp.setattr(router, "WORKERS", 1)
+    mp.setattr(rp, "WORKERS", 1)
+    router._JOBS.clear()
+
+    async def go():
+        await router.analyze2d(UUID(PID))
+        while router._TASKS:
+            await asyncio.gather(*list(router._TASKS))
+
+    asyncio.run(go())
+    yield root
+    mp.undo()
+    shutil.rmtree(folder, ignore_errors=True)
+
+
+def _slot(root: Path) -> tuple[str, str]:
+    """대표 단면 중 깎기 · 쌓기 비탈 하나."""
+    rows = router._latest_summary(root, "")[0]["stations"]
+    view.mark_representative(rows, router._overrides(root))
+    row = next(r for r in rows if (r.get("representative") or {}).get("on"))
+    return row["key"], row["sides"][0]["side"]
+
+
+def _req(side, mats=("weathered_soil",), **kw):
+    """단면 물성마다 c(대수정규 0.3) · φ(0.1) · γ(0.05)."""
+    variables = [
+        v
+        for m in mats
+        for v in (
+            {"material": m, "prop": "c", "dist": "lognormal", "cov": 0.3},
+            {"material": m, "prop": "phi", "cov": 0.1},
+            {"material": m, "prop": "gamma", "cov": 0.05},
+        )
+    ]
+    return rp.ProbRequest(
+        **{"side": side, "condition": "dry", "n": 400, "variables": variables, **kw}
+    )
+
+
+def test_setup_and_fixed_run(project, monkeypatch):
+    key, side = _slot(project)
+    s = rp.setup(project, key, side, "dry")
+    ids = [m["id"] for m in s["materials"]]
+    assert "weathered_soil" in ids and "wall" not in ids
+    assert s["defaults"]["properties"]["c"]["cov"] == 0.3 and s["required"] in (1.5, 1.3)
+    fs = s["fs_deterministic"]["bishop"]
+    seen = []
+    req = _req(side, ids)
+    rec = rp.run_prob(project, key, req, lambda p, m: seen.append(p))
+    res = rec["result"]
+    assert res["sensitivity"]["F_mean"] == pytest.approx(fs, rel=1e-9)
+    assert rec["fs_deterministic"] == fs and res["stats"]["n"] == 400
+    st = res["stats"]
+    assert 0 <= st["pf"] <= 1 and st["std"] > 0 and abs(st["mean"] - fs) < 0.1 * fs
+    assert st["beta_normal"] == pytest.approx((st["mean"] - 1) / st["std"])
+    assert seen and max(seen) <= 95
+    # 같은 요청 = 저장본 · 기록 목록 · setup 의 마지막 결과
+    calls = []
+    monkeypatch.setattr(rp.prob, "run", lambda **kw: calls.append(1))
+    again = rp.run_prob(project, key, req, lambda p, m: None)
+    assert not calls and again["computed_at"] == rec["computed_at"]
+    listed = rp.records(project)
+    assert [(r["station_key"], r["side"], r["condition"]) for r in listed] == [(key, side, "dry")]
+    assert rp.setup(project, key, side, "dry")["last"]["computed_at"] == rec["computed_at"]
+
+
+def test_rejects(project):
+    key, side = _slot(project)
+    with pytest.raises(HTTPException) as e:
+        rp.run_prob(
+            project, key, _req(side, variables=[{"material": "xx", "prop": "c", "cov": 0.2}]), print
+        )
+    assert e.value.status_code == 422
+    with pytest.raises(HTTPException) as e:
+        rp.run_prob(project, key, _req(side, variables=[{"prop": "water", "std": 0.5}]), print)
+    assert e.value.status_code == 422 and "수위선" in e.value.detail
+    with pytest.raises(HTTPException) as e:
+        rp.run_prob(
+            project,
+            key,
+            _req(side, variables=[{"material": "weathered_soil", "prop": "c", "cov": 0}]),
+            print,
+        )
+    assert "변수가 없습니다" in e.value.detail
+    # 대표 단면 끔 → 422
+    path = project / router.REPRESENTATIVE
+    old = path.read_text(encoding="utf-8") if path.is_file() else None
+    path.parent.mkdir(parents=True, exist_ok=True)
+    path.write_text(json.dumps({"overrides": {key: False}}), encoding="utf-8")
+    try:
+        with pytest.raises(HTTPException) as e:
+            rp.run_prob(project, key, _req(side), print)
+        assert "대표 단면" in e.value.detail
+    finally:
+        path.unlink() if old is None else path.write_text(old, encoding="utf-8")
+
+
+def test_search_worker_process():
+    """매회 탐색 일꾼 — 프로세스 풀 둘 · 진행 줄 · 결과 파일(Slide2 Tutorial 08 모형 · 작은 변동)."""
+    from test_537_1_e03_prob import T08, _surface, var
+
+    job = {
+        "model": T08,
+        "surface": _surface(T08),
+        "direction": -1,
+        "limits": None,
+        "variables": [var("c", "c", 5, 0.01), var("phi", "phi", 30, 0.01)],
+        "n": 4,
+        "mode": "search",
+        "workers": 2,
+    }
+    done = []
+    out = rp._run_search(job, done.append)
+    assert sorted(done) == [1, 2, 3, 4] and out["n"] == 4
+    assert all(abs(f - 1.139) < 0.02 for f in out["F"])
+
+
+def test_e05_prob_sheet(project, monkeypatch):
+    """E05 s12 — 측점 줄 = 저장본 값 · 변수 줄 = 토네이도 순 · 기본값 그대로 표기 · 없으면 시트 뺌."""
+    from E05_StabReport import E05_StabReport_Build as build
+
+    recs = rp.records(project)
+    assert len(recs) == 1
+    rec = recs[0]
+    book = build.form()
+    build.build_program(book, build.gather(project))
+    s12 = next(s for s in book["시트"] if s["id"] == "s12")
+    cells = s12["칸"]
+    st = rec["result"]["stats"]
+    assert cells["E6"]["값"] == pytest.approx(rec["fs_deterministic"], abs=1e-6)
+    assert cells["F6"]["값"] == pytest.approx(st["mean"], abs=1e-6)
+    assert cells["H6"]["값"] == pytest.approx(100 * st["pf"], abs=1e-6)
+    assert cells["J6"]["값"] == pytest.approx(st["beta_lognormal"], abs=1e-6)
+    assert "사용자 입력" in cells["A3"]["값"]  # 표본 400 ≠ 기본 1000
+    d = rp.load_defaults()
+    plain = {
+        **rec["request"],
+        **{k: d["run"][k] for k in ("n", "sampling", "seed", "mode", "method")},
+    }
+    assert build.prob._is_default({"request": plain}, d)
+    plain["variables"] = [{**plain["variables"][0], "cov": 0.25}]
+    assert not build.prob._is_default({"request": plain}, d)
+    first = rec["result"]["sensitivity"]["tornado"][0]["name"]
+    assert cells["B10"]["값"] == first and cells["K10"]["값"] == 1
+    monkeypatch.setattr(build.prob, "read", lambda root: [])
+    empty = build.form()
+    build.build_program(empty, build.gather(project))
+    assert "s12" not in {s["id"] for s in empty["시트"]}
+
+
+TYPESCRIPT = ROOT / "config" / "node_modules" / "typescript"
+LOGIC_TS = ROOT / "E03_StabSection" / "logic" / "E03_StabSection_Prob.ts"
+LOGIC_SCRIPT = r"""
+const ts = require(process.argv[1]);
+const fs = require("fs");
+const src = fs.readFileSync(process.argv[2], "utf8");
+const opts = { module: ts.ModuleKind.CommonJS, target: ts.ScriptTarget.ES2020 };
+const js = ts.transpileModule(src, { compilerOptions: opts }).outputText;
+const box = { exports: {} };
+new Function("exports", "module", js)(box.exports, box);
+const m = box.exports;
+const setup = JSON.parse(fs.readFileSync(0, "utf8"));
+const cache = m.fillCache(setup, null);
+const req = m.probRequest(setup, cache);
+cache.vars[m.varKey("colluvium", "c")].value = 0.5;
+process.stdout.write(JSON.stringify({ req, changed: m.changedCount(setup, cache),
+  ticks: m.axisTicks(0.83, 1.47, 5) }));
+"""
+
+
+@pytest.mark.skipif(not TYPESCRIPT.is_dir(), reason="config/node_modules 없음")
+def test_logic_ts_request_matches_server():
+    """기본값 후보(붕적토 c 0.4 덮어씀) → 요청 꼴 = 서버 ProbRequest 로 읽힘 · 바꾼 줄 수 · 눈금."""
+    setup = {
+        "side": "left",
+        "condition": "wet",
+        "materials": [
+            {"id": "weathered_soil", "name": "풍화토", "c": 15, "phi": 30, "gamma_t": 18},
+            {"id": "colluvium", "name": "붕적토", "c": 5, "phi": 27, "gamma_t": 17.5},
+        ],
+        "groundwater": {"mode": "line", "ru": None},
+        "defaults": rp.load_defaults(),
+    }
+    out = subprocess.run(
+        ["node", "-e", LOGIC_SCRIPT, str(TYPESCRIPT), str(LOGIC_TS)],
+        input=json.dumps(setup),
+        capture_output=True,
+        text=True,
+        encoding="utf-8",
+        check=True,
+    )
+    got = json.loads(out.stdout)
+    req = rp.ProbRequest(**got["req"])
+    assert (req.side, req.condition, req.n, req.sampling, req.mode) == (
+        "left",
+        "wet",
+        1000,
+        "lhs",
+        "fixed",
+    )
+    by = {(v.material, v.prop): v for v in req.variables}
+    assert len(by) == 6  # 물 σ 0 = 뺌
+    assert by[("colluvium", "c")].cov == 0.4 and by[("weathered_soil", "c")].cov == 0.3
+    assert by[("weathered_soil", "c")].dist == "lognormal" and by[("colluvium", "phi")].cov == 0.1
+    assert all(v.trunc_sigma == 3.0 for v in req.variables)
+    assert got["changed"] == 1
+    assert got["ticks"][0] <= 0.83 and got["ticks"][-1] >= 1.47 and 0 not in got["ticks"]
+
+
+if __name__ == "__main__":
+    sys.exit(pytest.main([__file__, "-q"]))
diff --git a/ui_template/ui_template_locale_e03.ts b/ui_template/ui_template_locale_e03.ts
index 2b8b49dd7..d336e7733 100644
--- a/ui_template/ui_template_locale_e03.ts
+++ b/ui_template/ui_template_locale_e03.ts
@@ -212,4 +212,64 @@ export const ui_locales_e03 = {
   E03_Sp_Mode_prestress: ["선인장(능동)", "Prestress (active)"],
   E03_Sp_Mode_not_crossing: ["활동면 안 만남", "Not crossing"],
   E03_Sp_Mode_pending: ["결과 없음", "No result"],
+  /* --- 확률 · 민감도 탭(537-1) --- */
+  E03_Tab_Prob: ["확률 · 민감도", "Probabilistic · Sensitivity"],
+  E03_Pb_RepOnly: [
+    "대표 단면만 확률 해석합니다 — 좌측 목록에서 별(대표 단면)을 켜고 [대표 단면 저장] 뒤 보세요.",
+    "Probabilistic analysis runs on representative sections only — star it in the list and save first.",
+  ],
+  E03_Pb_MeanNote: [
+    "평균 = E02 지반 조건 값(바꾸려면 E02) · 분포 · 변동계수는 기본값 후보",
+    "Mean = E02 ground values (change in E02) · distributions · COV are default candidates",
+  ],
+  E03_Pb_Use: ["쓰기", "Use"],
+  E03_Pb_Var: ["변수", "Variable"],
+  E03_Pb_Mean: ["평균", "Mean"],
+  E03_Pb_Dist: ["분포", "Distribution"],
+  E03_Pb_Dist_normal: ["정규", "Normal"],
+  E03_Pb_Dist_lognormal: ["대수정규", "Lognormal"],
+  E03_Pb_Spread: ["변동", "Spread"],
+  E03_Pb_Cov: ["변동계수", "COV"],
+  E03_Pb_Trunc: ["절단 ±σ 배", "Truncate ±kσ"],
+  E03_Pb_Prop_c: ["c kPa", "c kPa"],
+  E03_Pb_Prop_phi: ["φ °", "φ °"],
+  E03_Pb_Prop_gamma: ["γ kN/m³", "γ kN/m³"],
+  E03_Pb_Prop_water: ["지하수위 이동 m", "Water table shift m"],
+  E03_Pb_Prop_ru: ["간극수압비 r_u", "Pore pressure ratio r_u"],
+  E03_Pb_Mode: ["임계면", "Slip surface"],
+  E03_Pb_Mode_fixed: ["임계면 고정(빠름)", "Fixed critical (fast)"],
+  E03_Pb_Mode_search: ["매회 탐색(정밀 · 느림)", "Search each sample (slow)"],
+  E03_Pb_Method: ["방법", "Method"],
+  E03_Pb_Method_mp: ["M-P(반정현)", "M-P (half-sine)"],
+  E03_Pb_Sampling: ["표본 방법", "Sampling"],
+  E03_Pb_N: ["표본 수", "Samples"],
+  E03_Pb_Seed: ["시드", "Seed"],
+  E03_Pb_Rho: ["c–φ 상관 ρ", "c–φ correlation ρ"],
+  E03_Pb_Run: ["확률 해석", "Run Probabilistic"],
+  E03_Pb_Defaults: ["기본값", "Defaults"],
+  E03_Pb_Done: ["확률 해석 끝", "Probabilistic analysis done"],
+  E03_Pb_Unconfirmed: [
+    "기본값 후보 · 사용자 확정 전 — 근거",
+    "Default candidates · not confirmed — sources",
+  ],
+  E03_Pb_UserInput: ["사용자 입력 {n} 줄 · 근거", "User input {n} rows · sources"],
+  E03_Pb_Result: ["확률 해석 결과", "Probabilistic Result"],
+  E03_Pb_Det: ["결정론 Fs", "Deterministic FS"],
+  E03_Pb_MeanF: ["Fs 평균", "Mean FS"],
+  E03_Pb_Std: ["Fs 표준편차", "FS std. dev."],
+  E03_Pb_Pf: ["파괴확률 Pf(F < 1)", "Probability of failure"],
+  E03_Pb_BetaN: ["신뢰도 지수 β 정규", "Reliability index (normal)"],
+  E03_Pb_BetaLn: ["신뢰도 지수 β 대수정규", "Reliability index (lognormal)"],
+  E03_Pb_PReq: ["P(F < 기준안전율)", "P(F < required)"],
+  E03_Pb_Valid: ["유효 표본", "Valid samples"],
+  E03_Pb_Time: ["계산 시간", "Elapsed"],
+  E03_Pb_Hist: ["안전율 히스토그램", "FS Histogram"],
+  E03_Pb_Cdf: ["누적분포", "Cumulative Distribution"],
+  E03_Pb_Sens: [
+    "민감도(나머지 평균 · 임계면 고정)",
+    "Sensitivity (others at mean · fixed surface)",
+  ],
+  E03_Pb_Tornado: ["토네이도(범위 끝 Fs)", "Tornado (FS at range ends)"],
+  E03_Pb_Count: ["표본 수", "Count"],
+  E03_Pb_Range: ["범위 %", "Percent of range"],
 } as const satisfies Record<string, LocaleEntry>;