Add Monte Carlo (MCM) acceptance evaluation

Add a 10000-iteration Monte Carlo simulation on top of the existing GUM uncertainty path. For IRa and Ir it reports the simulated mean, standard deviation, P2.5/P97.5 (95% coverage interval), and pass/fail probability against configurable standard values (default IRa/Ir <= 1.0). - src/mcm.rs: dependency-free SplitMix64 PRNG + Box-Muller normals, deterministic seed for reproducible results - domain: AcceptanceLimits, McmResult, McmIndexStats; wire into SampleInput/CalculationResult - UI: compact calibration table, standard-value inputs, side-by-side GUM/MCM result cards, and a placeholder before calculation - run.md: run and packaging instructions Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
2026-06-08 15:13:05 +08:00 · 2026-06-08 15:13:05 +08:00 · 0c06d51c4a
parent c1080c1b03
commit 0c06d51c4a
8 changed files with 611 additions and 75 deletions
--- a/run.md
+++ b/run.md
@ -0,0 +1,105 @@
+# 运行与打包说明
+
+建筑材料放射性判定系统（Rust + Tauri + React + TypeScript）。
+
+## 1. 环境要求
+
+| 工具 | 版本 | 说明 |
+| --- | --- | --- |
+| Node.js | ≥ 18 | 前端构建（Vite） |
+| npm | 随 Node 安装 | 依赖管理 |
+| Rust | ≥ 1.78（已验证 1.93） | 计算引擎与 Tauri 后端，`rustup` 安装 |
+| Tauri 前置依赖 | Windows: WebView2 Runtime（Win11 已内置） | 见 <https://tauri.app/start/prerequisites/> |
+
+首次准备：
+
+```powershell
+npm install
+```
+
+> Rust 依赖在首次 `cargo`/`tauri` 命令时自动下载编译。
+
+## 2. 开发运行
+
+### 2.1 仅前端（浏览器预览，不含 Tauri 原生命令）
+
+```powershell
+npm run dev
+```
+
+打开 <http://localhost:1420>。注意：此模式下 `invoke("calculate")` 不可用，仅用于调样式。
+
+### 2.2 完整桌面应用（推荐）
+
+```powershell
+npm run tauri:dev
+```
+
+该命令会自动启动 Vite 开发服务器并编译 Rust 后端，弹出桌面窗口，支持热重载。
+
+## 3. 测试
+
+计算引擎（含蒙特卡洛仿真）的单元/集成测试：
+
+```powershell
+cargo test -p ceramic-radioactivity
+```
+
+前端类型检查与生产构建：
+
+```powershell
+npm run build
+```
+
+## 4. 打包（生成 Windows 安装包）
+
+1. 在 `src-tauri/tauri.conf.json` 中将 `bundle.active` 设为 `true`：
+
+   ```json
+   "bundle": {
+     "active": true,
+     "targets": ["nsis", "msi"],
+     "icon": ["icons/icon.ico"]
+   }
+   ```
+
+   - `nsis` 生成 `.exe` 安装包，`msi` 生成 `.msi`（需安装 WiX Toolset，可只保留 `nsis`）。
+   - `icon` 需指向有效图标；可用 `npx tauri icon path\to\logo.png` 生成整套图标。
+
+2. 执行构建：
+
+   ```powershell
+   npm run tauri:build
+   ```
+
+3. 产物输出目录：
+
+   ```
+   src-tauri/target/release/                         # 可执行文件 建筑材料放射性判定.exe
+   src-tauri/target/release/bundle/nsis/             # NSIS 安装包
+   src-tauri/target/release/bundle/msi/              # MSI 安装包（如启用）
+   ```
+
+> 首次 release 构建较慢（需编译全部 Rust 依赖），后续增量构建会显著加快。
+
+## 5. 项目结构
+
+```
+src/                  计算引擎（库 crate：ceramic-radioactivity）
+  ├─ domain.rs        输入/输出/校准/判定/MCM 数据结构
+  ├─ calculator.rs    GUM 法不确定度计算与三态判定
+  └─ mcm.rs           蒙特卡洛法仿真（10000 次，含 PRNG）
+tests/                计算引擎测试
+src-tauri/            Tauri 桌面外壳（command: calculate）
+ui/                   React + Ant Design 前端
+run.md                本说明
+```
+
+## 6. 功能说明（计算与判定）
+
+- **GUM 法**：A 类（标准差 / 极差法）、B 类不确定度合成，输出 IRa、Ir 及相对不确定度。
+  - `ur ≤ 20%` → `OK`；否则次数 < 6 → `请增加试验次数至 6 次`；否则 → `校准仪器后重新测量`。
+- **蒙特卡洛法（MCM）**：对各核素校准比活度按正态分布抽样，10000 次仿真传播至 IRa、Ir，
+  输出**平均值、标准偏差、P2.5、P97.5**（95% 包含区间），并与**标准值**比较给出**合格 / 不合格概率**。
+  - 标准值（IRa、Ir 限值）可在界面“校准参数 / 判定标准值”中修改，默认均为 `1.0`（GB 6566 主体材料）。
+  - 仿真使用固定随机种子，结果可复现。
--- a/src/calculator.rs
+++ b/src/calculator.rs
@ -2,6 +2,7 @@ use crate::domain::{
    CalculationError, CalculationResult, Conclusion, IndexResult, NuclideMeasurements,
    NuclideResult, SampleInput,
 };
+use crate::mcm::run_monte_carlo;

 const ACCEPTANCE_LIMIT_PERCENT: f64 = 20.0;

@ -25,6 +26,8 @@ pub fn calculate_sample(input: SampleInput) -> Result<CalculationResult, Calcula
        Conclusion::RecalibrateInstrument
    };

+    let mcm = run_monte_carlo(&ra, &th, &k, &input.limits);
+
    Ok(CalculationResult {
        measurement_count: n,
        ra,
@ -33,6 +36,7 @@ pub fn calculate_sample(input: SampleInput) -> Result<CalculationResult, Calcula
        ira,
        ir,
        conclusion,
+        mcm,
    })
 }

--- a/src/domain.rs
+++ b/src/domain.rs
@ -8,6 +8,25 @@ pub struct SampleInput {
    pub ra: NuclideMeasurements,
    pub th: NuclideMeasurements,
    pub k: NuclideMeasurements,
+    /// 合格判定标准值（限值）。前端可省略，默认 IRa ≤ 1.0、Ir ≤ 1.0（GB 6566 主体材料）。
+    #[serde(default)]
+    pub limits: AcceptanceLimits,
+}
+
+/// 内照射指数 IRa 与外照射指数 Ir 的合格判定标准值（限值）。
+#[derive(Debug, Clone, Copy, PartialEq, Serialize, Deserialize)]
+pub struct AcceptanceLimits {
+    pub ira_limit: f64,
+    pub ir_limit: f64,
+}
+
+impl Default for AcceptanceLimits {
+    fn default() -> Self {
+        Self {
+            ira_limit: 1.0,
+            ir_limit: 1.0,
+        }
+    }
 }

 #[derive(Debug, Clone, PartialEq, Serialize, Deserialize)]
@ -32,6 +51,42 @@ pub struct CalculationResult {
    pub ira: IndexResult,
    pub ir: IndexResult,
    pub conclusion: Conclusion,
+    /// 蒙特卡洛法（MCM）仿真结果。
+    pub mcm: McmResult,
+}
+
+/// 蒙特卡洛法（MCM）整体仿真结果。
+#[derive(Debug, Clone, PartialEq, Serialize, Deserialize)]
+pub struct McmResult {
+    /// 仿真次数（默认 10000）。
+    pub iterations: usize,
+    /// 内照射指数 IRa 的仿真统计。
+    pub ira: McmIndexStats,
+    /// 外照射指数 Ir 的仿真统计。
+    pub ir: McmIndexStats,
+    /// 综合合格概率：每次仿真中 IRa 与 Ir 同时不超过各自标准值的比例。
+    pub overall_pass_probability: f64,
+    /// 综合不合格概率 = 1 - overall_pass_probability。
+    pub overall_fail_probability: f64,
+}
+
+/// 单个指数的 MCM 仿真统计量。
+#[derive(Debug, Clone, Copy, PartialEq, Serialize, Deserialize)]
+pub struct McmIndexStats {
+    /// 仿真平均值。
+    pub mean: f64,
+    /// 仿真标准偏差。
+    pub std_dev: f64,
+    /// 2.5 百分位（95% 包含区间下限）。
+    pub p2_5: f64,
+    /// 97.5 百分位（95% 包含区间上限）。
+    pub p97_5: f64,
+    /// 用于比较的标准值（限值）。
+    pub standard_value: f64,
+    /// 合格概率：仿真值不超过标准值的比例。
+    pub pass_probability: f64,
+    /// 不合格概率 = 1 - pass_probability。
+    pub fail_probability: f64,
 }

 #[derive(Debug, Clone, Copy, PartialEq, Serialize, Deserialize)]
--- a/src/lib.rs
+++ b/src/lib.rs
@ -1,8 +1,9 @@
 mod calculator;
 mod domain;
+mod mcm;

 pub use calculator::calculate_sample;
 pub use domain::{
-    CalculationError, CalculationResult, CalibrationParams, Conclusion, IndexResult,
-    NuclideMeasurements, NuclideResult, SampleInput,
+    AcceptanceLimits, CalculationError, CalculationResult, CalibrationParams, Conclusion,
+    IndexResult, McmIndexStats, McmResult, NuclideMeasurements, NuclideResult, SampleInput,
 };
--- a/src/mcm.rs
+++ b/src/mcm.rs
@ -0,0 +1,146 @@
+//! 蒙特卡洛法（MCM）不确定度传播与合格概率评估。
+//!
+//! 按 JJF 1059.2 的思路，对内照射指数 IRa 与外照射指数 Ir 进行 10000 次仿真：
+//! 每次仿真从各核素校准比活度的分布中抽样，经测量模型传播后得到 IRa、Ir 的样本，
+//! 进而给出仿真平均值、标准偏差、95% 包含区间（P2.5、P97.5），并与标准值比较给出合格概率。
+//!
+//! 每个核素的校准比活度按正态分布 `N(mean_calibrated, combined_uncertainty)` 抽样，
+//! 其中均值与标准（合成）不确定度直接复用 GUM 法已算出的 `NuclideResult`，
+//! 因此 MCM 的均值/标准偏差与 GUM 结果保持一致，而包含区间与合格概率由仿真分布直接得到。
+
+use crate::domain::{AcceptanceLimits, McmIndexStats, McmResult, NuclideResult};
+
+/// 默认仿真次数。
+pub const DEFAULT_ITERATIONS: usize = 10_000;
+
+/// 固定随机种子，保证同一输入得到可复现的仿真结果（便于出具一致的报告与测试）。
+const SEED: u64 = 0x853C_49E6_748F_EA9B;
+
+/// 对给定的核素结果与标准值执行 MCM 仿真。
+pub fn run_monte_carlo(
+    ra: &NuclideResult,
+    th: &NuclideResult,
+    k: &NuclideResult,
+    limits: &AcceptanceLimits,
+) -> McmResult {
+    let iterations = DEFAULT_ITERATIONS;
+    let mut rng = SplitMix64::new(SEED);
+
+    let mut ira_samples = Vec::with_capacity(iterations);
+    let mut ir_samples = Vec::with_capacity(iterations);
+    let mut overall_pass = 0usize;
+
+    for _ in 0..iterations {
+        let c_ra = sample_normal(&mut rng, ra.mean_calibrated, ra.combined_uncertainty);
+        let c_th = sample_normal(&mut rng, th.mean_calibrated, th.combined_uncertainty);
+        let c_k = sample_normal(&mut rng, k.mean_calibrated, k.combined_uncertainty);
+
+        let ira = c_ra / 200.0;
+        let ir = c_ra / 370.0 + c_th / 260.0 + c_k / 4200.0;
+
+        if ira <= limits.ira_limit && ir <= limits.ir_limit {
+            overall_pass += 1;
+        }
+
+        ira_samples.push(ira);
+        ir_samples.push(ir);
+    }
+
+    let ira_stats = summarize(&mut ira_samples, limits.ira_limit);
+    let ir_stats = summarize(&mut ir_samples, limits.ir_limit);
+    let overall_pass_probability = overall_pass as f64 / iterations as f64;
+
+    McmResult {
+        iterations,
+        ira: ira_stats,
+        ir: ir_stats,
+        overall_pass_probability,
+        overall_fail_probability: 1.0 - overall_pass_probability,
+    }
+}
+
+/// 由样本计算统计量。会就地排序 `samples` 以取百分位。
+fn summarize(samples: &mut [f64], standard_value: f64) -> McmIndexStats {
+    let n = samples.len();
+    let mean = samples.iter().sum::<f64>() / n as f64;
+    let variance = samples
+        .iter()
+        .map(|value| (value - mean).powi(2))
+        .sum::<f64>()
+        / (n as f64 - 1.0);
+    let std_dev = variance.sqrt();
+
+    let pass = samples.iter().filter(|&&value| value <= standard_value).count();
+    let pass_probability = pass as f64 / n as f64;
+
+    samples.sort_by(|a, b| a.partial_cmp(b).unwrap_or(std::cmp::Ordering::Equal));
+    let p2_5 = percentile(samples, 0.025);
+    let p97_5 = percentile(samples, 0.975);
+
+    McmIndexStats {
+        mean,
+        std_dev,
+        p2_5,
+        p97_5,
+        standard_value,
+        pass_probability,
+        fail_probability: 1.0 - pass_probability,
+    }
+}
+
+/// 已排序样本的线性插值百分位（p 取 0..=1）。
+fn percentile(sorted: &[f64], p: f64) -> f64 {
+    let n = sorted.len();
+    if n == 0 {
+        return f64::NAN;
+    }
+    if n == 1 {
+        return sorted[0];
+    }
+    let rank = p * (n as f64 - 1.0);
+    let lower = rank.floor() as usize;
+    let upper = rank.ceil() as usize;
+    let weight = rank - lower as f64;
+    sorted[lower] * (1.0 - weight) + sorted[upper] * weight
+}
+
+/// 由标准正态抽样得到 `N(mean, std_dev)` 样本（std_dev 为 0 时退化为常数）。
+fn sample_normal(rng: &mut SplitMix64, mean: f64, std_dev: f64) -> f64 {
+    if std_dev == 0.0 {
+        return mean;
+    }
+    mean + std_dev * rng.next_standard_normal()
+}
+
+/// SplitMix64 伪随机数发生器：无外部依赖、确定性、统计性质良好。
+struct SplitMix64 {
+    state: u64,
+}
+
+impl SplitMix64 {
+    fn new(seed: u64) -> Self {
+        Self { state: seed }
+    }
+
+    fn next_u64(&mut self) -> u64 {
+        self.state = self.state.wrapping_add(0x9E37_79B9_7F4A_7C15);
+        let mut z = self.state;
+        z = (z ^ (z >> 30)).wrapping_mul(0xBF58_476D_1CE4_E5B9);
+        z = (z ^ (z >> 27)).wrapping_mul(0x94D0_49BB_1331_11EB);
+        z ^ (z >> 31)
+    }
+
+    /// 返回 [0, 1) 区间内的均匀分布样本。
+    fn next_uniform(&mut self) -> f64 {
+        // 取高 53 位映射到 [0, 1)，与 f64 尾数精度一致。
+        (self.next_u64() >> 11) as f64 / (1u64 << 53) as f64
+    }
+
+    /// Box-Muller 变换生成标准正态样本。
+    fn next_standard_normal(&mut self) -> f64 {
+        // 避免 ln(0)：将 u1 下限钳到一个极小正数。
+        let u1 = self.next_uniform().max(f64::MIN_POSITIVE);
+        let u2 = self.next_uniform();
+        (-2.0 * u1.ln()).sqrt() * (std::f64::consts::TAU * u2).cos()
+    }
+}
--- a/tests/calculator_tests.rs
+++ b/tests/calculator_tests.rs
@ -1,5 +1,6 @@
 use ceramic_radioactivity::{
-    calculate_sample, CalibrationParams, Conclusion, NuclideMeasurements, SampleInput,
+    calculate_sample, AcceptanceLimits, CalibrationParams, Conclusion, NuclideMeasurements,
+    SampleInput,
 };

 fn default_input() -> SampleInput {
@ -28,6 +29,10 @@ fn default_input() -> SampleInput {
                coverage_factor: 2.0,
            },
        },
+        limits: AcceptanceLimits {
+            ira_limit: 1.0,
+            ir_limit: 1.0,
+        },
    }
 }

@ -80,6 +85,62 @@ fn rejects_mismatched_measurement_counts() {
    assert!(err.to_string().contains("same measurement count"));
 }

+#[test]
+fn monte_carlo_matches_analytical_mean_and_uncertainty() {
+    let result = calculate_sample(default_input()).expect("valid sample should calculate");
+
+    assert_eq!(result.mcm.iterations, 10_000);
+
+    // MCM 仿真均值/标准偏差应与 GUM 解析结果一致（仿真随机误差在 ~1% 量级）。
+    assert_close(result.mcm.ira.mean, result.ira.value, 5e-4);
+    assert_close(result.mcm.ir.mean, result.ir.value, 5e-4);
+    assert_close(
+        result.mcm.ira.std_dev,
+        result.ira.standard_uncertainty,
+        result.ira.standard_uncertainty * 0.05,
+    );
+    assert_close(
+        result.mcm.ir.std_dev,
+        result.ir.standard_uncertainty,
+        result.ir.standard_uncertainty * 0.05,
+    );
+
+    // 95% 包含区间应包住均值。
+    assert!(result.mcm.ira.p2_5 < result.mcm.ira.mean);
+    assert!(result.mcm.ira.mean < result.mcm.ira.p97_5);
+
+    // 默认样本的 IRa≈0.46、Ir≈0.75 都远低于标准值 1.0，合格概率应为 1。
+    assert_close(result.mcm.ira.pass_probability, 1.0, 1e-12);
+    assert_close(result.mcm.ir.pass_probability, 1.0, 1e-12);
+    assert_close(result.mcm.overall_pass_probability, 1.0, 1e-12);
+    assert_close(result.mcm.overall_fail_probability, 0.0, 1e-12);
+}
+
+#[test]
+fn monte_carlo_is_deterministic_for_same_input() {
+    let first = calculate_sample(default_input()).expect("valid sample should calculate");
+    let second = calculate_sample(default_input()).expect("valid sample should calculate");
+
+    assert_eq!(first.mcm, second.mcm);
+}
+
+#[test]
+fn monte_carlo_gives_about_half_pass_probability_when_limit_equals_mean() {
+    let mut input = default_input();
+    let analytical = calculate_sample(input.clone()).expect("valid sample should calculate");
+
+    // 将 IRa 标准值设为其均值，合格概率应接近 0.5。
+    input.limits.ira_limit = analytical.ira.value;
+    let result = calculate_sample(input).expect("valid sample should calculate");
+
+    assert_close(result.mcm.ira.pass_probability, 0.5, 0.03);
+    assert_close(
+        result.mcm.ira.fail_probability,
+        1.0 - result.mcm.ira.pass_probability,
+        1e-12,
+    );
+}
+
 fn assert_close(actual: f64, expected: f64, tolerance: f64) {
    assert!(
        (actual - expected).abs() <= tolerance,
--- a/ui/src/App.tsx
+++ b/ui/src/App.tsx
@ -1,16 +1,6 @@
 import { useMemo, useRef, useState } from "react";
 import { invoke } from "@tauri-apps/api/core";
-import {
-  Alert,
-  Button,
-  Card,
-  ConfigProvider,
-  Descriptions,
-  Form,
-  InputNumber,
-  Table,
-  Tag
-} from "antd";
+import { Alert, Button, Card, ConfigProvider, InputNumber, Table, Tag } from "antd";
 import zhCN from "antd/locale/zh_CN";

 type Conclusion = "Ok" | "IncreaseMeasurementsToSix" | "RecalibrateInstrument";
@ -26,10 +16,34 @@ type NuclideMeasurements = {
  calibration: CalibrationParams;
 };

+type AcceptanceLimits = {
+  ira_limit: number;
+  ir_limit: number;
+};
+
 type SampleInput = {
  ra: NuclideMeasurements;
  th: NuclideMeasurements;
  k: NuclideMeasurements;
+  limits: AcceptanceLimits;
+};
+
+type McmIndexStats = {
+  mean: number;
+  std_dev: number;
+  p2_5: number;
+  p97_5: number;
+  standard_value: number;
+  pass_probability: number;
+  fail_probability: number;
+};
+
+type McmResult = {
+  iterations: number;
+  ira: McmIndexStats;
+  ir: McmIndexStats;
+  overall_pass_probability: number;
+  overall_fail_probability: number;
 };

 type NuclideResult = {
@ -56,6 +70,7 @@ type CalculationResult = {
  ira: IndexResult;
  ir: IndexResult;
  conclusion: Conclusion;
+  mcm: McmResult;
 };

 type MeasurementRow = {
@ -67,6 +82,10 @@ type MeasurementRow = {

 type ResultRow = { name: string } & NuclideResult;

+type McmRow = { name: string } & McmIndexStats;
+
+type CalibrationRow = { key: string; name: string } & CalibrationParams;
+
 type FocusableInput = {
  focus: () => void;
 };
@ -77,6 +96,8 @@ const defaultCalibration = {
  k: { factor: 0.961, expanded_uncertainty_percent: 6.7, coverage_factor: 2 }
 };

+const defaultLimits: AcceptanceLimits = { ira_limit: 1.0, ir_limit: 1.0 };
+
 const initialRows: MeasurementRow[] = [
  { key: 1, ra: 100, th: 110, k: 560 },
  { key: 2, ra: 102, th: 111, k: 565 },
@ -97,8 +118,14 @@ function formatNumber(value: number, digits = 4) {
  return value.toFixed(digits);
 }

+function formatPercent(value: number, digits = 2) {
+  if (!Number.isFinite(value)) return "-";
+  return (value * 100).toFixed(digits) + "%";
+}
+
 function App() {
  const [rows, setRows] = useState<MeasurementRow[]>(initialRows);
+  const [limits, setLimits] = useState<AcceptanceLimits>(defaultLimits);
  const [result, setResult] = useState<CalculationResult | null>(null);
  const [error, setError] = useState<string | null>(null);
  const [loading, setLoading] = useState(false);
@ -129,7 +156,8 @@ function App() {
    return {
      ra: { measured_values: toValues("ra"), calibration: defaultCalibration.ra },
      th: { measured_values: toValues("th"), calibration: defaultCalibration.th },
-      k: { measured_values: toValues("k"), calibration: defaultCalibration.k }
+      k: { measured_values: toValues("k"), calibration: defaultCalibration.k },
+      limits
    };
  };

@ -221,53 +249,137 @@ function App() {
              </div>
            </Card>

-            <Card title="校准参数" className="panel compact-panel">
-              <Form layout="vertical">
-                {Object.entries(defaultCalibration).map(([key, calibration]) => (
-                  <Descriptions key={key} bordered size="small" column={1} className="calibration-block">
-                    <Descriptions.Item label={key.toUpperCase()}>{calibration.factor}</Descriptions.Item>
-                    <Descriptions.Item label="U">{calibration.expanded_uncertainty_percent}%</Descriptions.Item>
-                    <Descriptions.Item label="k">{calibration.coverage_factor}</Descriptions.Item>
-                  </Descriptions>
-                ))}
-              </Form>
+            <Card title="校准参数 / 判定标准值" className="panel compact-panel">
+              <Table<CalibrationRow>
+                className="calibration-table"
+                pagination={false}
+                size="small"
+                rowKey="key"
+                dataSource={[
+                  { key: "ra", name: "Ra", ...defaultCalibration.ra },
+                  { key: "th", name: "Th", ...defaultCalibration.th },
+                  { key: "k", name: "K", ...defaultCalibration.k }
+                ]}
+                columns={[
+                  { title: "核素", dataIndex: "name", align: "center" },
+                  { title: "a", dataIndex: "factor", align: "center" },
+                  {
+                    title: "U",
+                    dataIndex: "expanded_uncertainty_percent",
+                    align: "center",
+                    render: (value: number) => `${value}%`
+                  },
+                  { title: "k", dataIndex: "coverage_factor", align: "center" }
+                ]}
+              />
+              <div className="limits-form">
+                <div className="limit-field">
+                  <span className="limit-label">IRa 标准值</span>
+                  <InputNumber
+                    value={limits.ira_limit}
+                    min={0}
+                    step={0.1}
+                    onChange={(value) => setLimits((current) => ({ ...current, ira_limit: value ?? 0 }))}
+                  />
+                </div>
+                <div className="limit-field">
+                  <span className="limit-label">Ir 标准值</span>
+                  <InputNumber
+                    value={limits.ir_limit}
+                    min={0}
+                    step={0.1}
+                    onChange={(value) => setLimits((current) => ({ ...current, ir_limit: value ?? 0 }))}
+                  />
+                </div>
+              </div>
            </Card>
          </div>

          {error ? <Alert type="error" message={error} showIcon /> : null}

-          {result ? (
-            <Card title="计算结果" className="panel">
-              <div className="result-grid">
-                <ResultTile title="IRa" value={result.ira.value} uncertainty={result.ira.relative_uncertainty_percent} />
-                <ResultTile title="Ir" value={result.ir.value} uncertainty={result.ir.relative_uncertainty_percent} />
-                <div className="result-tile conclusion-tile">
-                  <span>判定</span>
-                  <Tag color={result.conclusion === "Ok" ? "success" : "warning"}>
-                    {conclusionText[result.conclusion]}
-                  </Tag>
+          <div className="results-area">
+            {result ? (
+              <div className="results-row">
+              <Card title="计算结果（GUM 法）" className="panel">
+                <div className="result-grid">
+                  <ResultTile title="IRa" value={result.ira.value} uncertainty={result.ira.relative_uncertainty_percent} />
+                  <ResultTile title="Ir" value={result.ir.value} uncertainty={result.ir.relative_uncertainty_percent} />
+                  <div className="result-tile conclusion-tile">
+                    <span>判定</span>
+                    <Tag color={result.conclusion === "Ok" ? "success" : "warning"}>
+                      {conclusionText[result.conclusion]}
+                    </Tag>
+                  </div>
                </div>
+                <Table<ResultRow>
+                  pagination={false}
+                  rowKey="name"
+                  size="small"
+                  dataSource={[
+                    { name: "Ra-226", ...result.ra },
+                    { name: "Th-232", ...result.th },
+                    { name: "K-40", ...result.k }
+                  ]}
+                  columns={[
+                    { title: "核素", dataIndex: "name" },
+                    { title: "均值", dataIndex: "mean_measured", render: (value: number) => formatNumber(value) },
+                    { title: "校准活度", dataIndex: "mean_calibrated", render: (value: number) => formatNumber(value) },
+                    { title: "A 类", dataIndex: "type_a_uncertainty", render: (value: number) => formatNumber(value) },
+                    { title: "B 类相对", dataIndex: "type_b_relative", render: (value) => formatNumber(value * 100, 3) + "%" },
+                    { title: "合成", dataIndex: "combined_uncertainty", render: (value: number) => formatNumber(value) }
+                  ]}
+                />
+              </Card>
+
+              <Card title={`蒙特卡洛仿真（MCM，${result.mcm.iterations} 次）`} className="panel">
+                <div className="result-grid">
+                  <div className="result-tile">
+                    <span>综合合格概率</span>
+                    <strong>{formatPercent(result.mcm.overall_pass_probability)}</strong>
+                    <small>IRa 与 Ir 同时合格</small>
+                  </div>
+                  <div className="result-tile">
+                    <span>综合不合格概率</span>
+                    <strong>{formatPercent(result.mcm.overall_fail_probability)}</strong>
+                    <small>任一指数超标</small>
+                  </div>
+                  <div className="result-tile conclusion-tile">
+                    <span>仿真判定</span>
+                    <Tag color={result.mcm.overall_pass_probability >= 0.95 ? "success" : "warning"}>
+                      {result.mcm.overall_pass_probability >= 0.95 ? "合格" : "不合格风险"}
+                    </Tag>
+                  </div>
+                </div>
+                <Table<McmRow>
+                  pagination={false}
+                  rowKey="name"
+                  size="small"
+                  dataSource={[
+                    { name: "IRa", ...result.mcm.ira },
+                    { name: "Ir", ...result.mcm.ir }
+                  ]}
+                  columns={[
+                    { title: "指数", dataIndex: "name" },
+                    { title: "平均值", dataIndex: "mean", render: (value: number) => formatNumber(value) },
+                    { title: "标准偏差", dataIndex: "std_dev", render: (value: number) => formatNumber(value) },
+                    { title: "P2.5", dataIndex: "p2_5", render: (value: number) => formatNumber(value) },
+                    { title: "P97.5", dataIndex: "p97_5", render: (value: number) => formatNumber(value) },
+                    { title: "标准值", dataIndex: "standard_value", render: (value: number) => formatNumber(value, 2) },
+                    {
+                      title: "合格概率",
+                      dataIndex: "pass_probability",
+                      render: (value: number) => (
+                        <Tag color={value >= 0.95 ? "success" : "warning"}>{formatPercent(value)}</Tag>
+                      )
+                    }
+                  ]}
+                />
+              </Card>
              </div>
-              <Table<ResultRow>
-                pagination={false}
-                rowKey="name"
-                size="small"
-                dataSource={[
-                  { name: "Ra-226", ...result.ra },
-                  { name: "Th-232", ...result.th },
-                  { name: "K-40", ...result.k }
-                ]}
-                columns={[
-                  { title: "核素", dataIndex: "name" },
-                  { title: "均值", dataIndex: "mean_measured", render: (value: number) => formatNumber(value) },
-                  { title: "校准活度", dataIndex: "mean_calibrated", render: (value: number) => formatNumber(value) },
-                  { title: "A 类不确定度", dataIndex: "type_a_uncertainty", render: (value: number) => formatNumber(value) },
-                  { title: "B 类相对值", dataIndex: "type_b_relative", render: (value) => formatNumber(value * 100, 3) + "%" },
-                  { title: "合成不确定度", dataIndex: "combined_uncertainty", render: (value: number) => formatNumber(value) }
-                ]}
-              />
-            </Card>
-          ) : null}
+            ) : (
+              <div className="results-placeholder">请计算后查看结果</div>
+            )}
+          </div>
        </section>
      </main>
    </ConfigProvider>
--- a/ui/src/styles.css
+++ b/ui/src/styles.css
@ -12,20 +12,47 @@ body {
 .app-shell {
  min-height: 100vh;
  padding: 12px;
+  display: flex;
+  flex-direction: column;
 }

 .workspace {
+  width: 100%;
  max-width: 1280px;
  margin: 0 auto;
+  flex: 1;
+  display: flex;
+  flex-direction: column;
+  min-height: 0;
 }

 .content-grid {
  display: grid;
  grid-template-columns: minmax(0, 1fr) 240px;
-  grid-auto-rows: 450px;
+  grid-auto-rows: 440px;
  align-items: stretch;
  gap: 10px;
  margin-bottom: 10px;
+  flex: 0 0 auto;
+}
+
+.results-area {
+  flex: 1;
+  min-height: 0;
+  display: flex;
+  flex-direction: column;
+}
+
+.results-placeholder {
+  flex: 1;
+  display: flex;
+  align-items: center;
+  justify-content: center;
+  color: #c0c4cc;
+  font-size: 26px;
+  font-weight: 600;
+  letter-spacing: 2px;
+  user-select: none;
 }

 .panel {
@ -70,13 +97,36 @@ body {
 }

 .compact-panel .ant-card-body {
-  height: calc(100% - 38px);
-  overflow: hidden;
-  padding: 8px 10px 6px;
+  height: auto;
+  overflow: visible;
+  padding: 10px;
 }

-.calibration-block {
-  margin-bottom: 8px;
+.calibration-table {
+  margin-bottom: 12px;
+}
+
+.limits-form {
+  display: flex;
+  flex-direction: column;
+  gap: 8px;
+}
+
+.limit-field {
+  display: flex;
+  align-items: center;
+  gap: 8px;
+}
+
+.limit-label {
+  flex-shrink: 0;
+  white-space: nowrap;
+  color: #5f6368;
+  font-size: 13px;
+}
+
+.limit-field .ant-input-number {
+  flex: 1;
 }

 .ant-input-number {
@ -91,6 +141,17 @@ body {
  margin-top: 8px;
 }

+.results-row {
+  display: grid;
+  grid-template-columns: minmax(0, 1fr) minmax(0, 1fr);
+  gap: 10px;
+  align-items: stretch;
+}
+
+.results-row > .panel {
+  height: 100%;
+}
+
 .result-grid {
  display: grid;
  grid-template-columns: repeat(3, minmax(0, 1fr));
@ -99,8 +160,8 @@ body {
 }

 .result-tile {
-  min-height: 68px;
-  padding: 8px 10px;
+  min-height: 58px;
+  padding: 7px 10px;
  border: 1px solid #e5e7eb;
  border-radius: 8px;
  background: #ffffff;
@ -116,7 +177,7 @@ body {
 }

 .result-tile strong {
-  font-size: 20px;
+  font-size: 18px;
  line-height: 1.2;
 }

@ -145,23 +206,14 @@ body {
  padding: 0 7px;
 }

-.ant-descriptions.ant-descriptions-small .ant-descriptions-item-label,
-.ant-descriptions.ant-descriptions-small .ant-descriptions-item-content {
-  padding: 5px 8px;
-  line-height: 1.35;
-}
-
-.ant-descriptions.ant-descriptions-small .ant-descriptions-item-label {
-  width: 54px;
-}
-
@media (max-width: 900px) {
  .app-shell {
    padding: 10px;
  }

  .content-grid,
-  .result-grid {
+  .result-grid,
+  .results-row {
    grid-template-columns: 1fr;
  }