科研 Skill

疲劳断裂寿命分析

fatigue-fracture-analysis

Estimates fatigue life and crack-growth life for metal parts from S-N data, load histories, and fracture-mechanics inputs. Use for Miner’s-rule damage, rainflow counting, or Paris-law crack growth.

发布者 Codar安全审核通过v0.1.0更新于 2026-11-18

SKILL.md

Fatigue and Fracture Analysis

Estimates how long a metal part will survive under cyclic loading, by two complementary routes: S-N (stress-life) fatigue with cumulative-damage summation, and linear elastic fracture mechanics (LEFM) crack-growth life. Use this when a part sees repeated loading and the question is "how many cycles" or "how many crack-growth cycles to failure", not a single static overload (use machine-element-sizing or quick-fea for that).

Purpose

  • Turn a measured or assumed load history into a fatigue-damage estimate using Miner's rule, with proper cycle counting (rainflow) for variable-amplitude signals.
  • Fit or apply an S-N curve (Basquin equation) for a material, including mean-stress correction (Goodman, Gerber, or Soderberg) and a stress-concentration (notch) factor.
  • Estimate remaining life of a known or postulated crack with the Paris law, and compare the critical crack size against fracture toughness.
  • Report a safety factor and the dominant source of uncertainty (material scatter, load measurement, or SCF/K estimate), not just a cycle count.

Prerequisites

numpy, scipy, and matplotlib (pip install numpy scipy matplotlib). A material S-N dataset or standard parameters (references/material_fatigue_data.md has indicative values for common steels and aluminum alloys — always prefer a dataset specific to the actual material and surface condition when one is available). A load history (time series of stress or strain) for variable-amplitude work, or a single stress amplitude and mean stress for constant-amplitude work.

Workflow

  1. Classify the problem.

    • Constant-amplitude, high-cycle (elastic, >~10^4 cycles): use the S-N route directly.
    • Variable-amplitude: rainflow-count the load history first, then apply S-N + Miner's rule to each counted cycle bin.
    • Known or suspected crack (weld defect, inclusion, scratch, prior fatigue damage): use the Paris-law route, and check LEFM applies (small-scale yielding — see Limitations).
    • Low-cycle fatigue (<~10^4 cycles, visible plastic strain each cycle): the Basquin S-N fit is not valid; use a strain-life (Coffin-Manson) approach instead — flag this to the user rather than silently applying S-N.
  2. Get or fit the S-N curve. python scripts/fatigue.py sn-fit --data cycles_stress.csv fits S = S_f' * (2N)^b (Basquin) by log-log linear regression and reports the fatigue strength coefficient, exponent, and R². Without data, use literature values from references/material_fatigue_data.md, state the source, and flag it as an estimate.

  3. Correct for mean stress if the cycle has a nonzero mean (most real loads do): python scripts/fatigue.py mean-stress --amp 120 --mean 40 --ultimate 550 --method goodman Goodman is conservative for brittle/typical cases; Gerber fits ductile metal data better but is non-conservative near zero mean stress — default to Goodman unless the user asks.

  4. Count cycles in a variable-amplitude history: python scripts/fatigue.py rainflow --signal load.csv --col stress outputs cycle ranges, means, and counts (ASTM E1049 rainflow algorithm). Never sum peak-to-peak ranges directly without rainflow counting — it over- or under-counts damage.

  5. Sum damage (Miner's rule): python scripts/fatigue.py miner --cycles rainflow_output.csv --sn-b -0.1 --sn-sf 900 reports total damage D = sum(n_i / N_i) and life in repeats-of-the-history if D < 1 (failure predicted when D ≥ 1). Report D itself, not just pass/fail — a D of 0.05 and a D of 0.9 both "pass" today but have very different margins.

  6. Crack-growth life (if applicable): python scripts/fatigue.py paris --a0 0.5 --af 12 --C 6.9e-12 --m 3.0 --dK-func "constant" --dsigma 80 integrates da/dN = C*(ΔK)^m from initial flaw size a0 to critical/final size af (from K_IC or a geometry limit) and reports cycles to failure. ΔK = Y*Δσ*sqrt(pi*a) with a geometry factor Y (1.0 for a simple through-crack is the default; use a handbook Y for the actual geometry — this changes the answer significantly and must not be left at 1.0 silently).

  7. Report: load classification, S-N or Paris-law parameters and their source, mean- stress or geometry correction used, damage sum or crack-growth life, safety factor against the design life, and the single biggest uncertainty in the estimate.

Limitations

  • S-N/Miner's rule assumes no sequence effects (load order doesn't matter) and no retardation after overloads; this is known to be inaccurate for spectra with large occasional overloads (overload-induced crack retardation is not modeled).
  • LEFM (Paris law) requires small-scale yielding — plastic zone small relative to crack length and remaining ligament. For thin sections or high-toughness/low-yield-strength materials this can be violated; this skill does not check it automatically, so state the assumption and flag when the initial crack is a large fraction of the section.
  • Material fatigue data in references/material_fatigue_data.md is indicative only (typical literature values), not certified design data; always prefer a dataset or handbook value specific to the real material, heat treatment, and surface finish.
  • No corrosion-fatigue, fretting-fatigue, thermal-fatigue, or weld-specific (e.g. IIW FAT class) methods are included; flag these cases to the user instead of silently applying a plain S-N curve.
  • references/material_fatigue_data.md and the scripts are illustrative; cross-check any load-bearing or safety-critical conclusion against structural-analysis / quick-fea stress results and, where relevant, a qualified fatigue engineer's sign-off.

Related skills

machine-element-sizing (static sizing and basic shaft fatigue of standard elements), structural-analysis (FEA-based stress for pump casings/impellers, including its own fatigue discussion), quick-fea (general stress field from a STEP part to feed this skill's S-N route), reliability-engineering (Weibull life analysis at the population level, versus this skill's single-part deterministic life estimate).

Quality checks

  • Reported damage sum and crack-growth life always come with the input parameters and their source (measured, fitted, or literature) printed alongside, not just a bare number.
  • Rainflow-counted cycle totals account for the entire input history (verify sum(counts) * 0.5 or 1.0 per half/full cycle matches the signal's peak-valley count).
  • The script's --selftest mode (python scripts/fatigue.py --selftest) checks the Paris integration against a closed-form constant-ΔK solution and the rainflow counter against a hand-countable reference signal; run it once after any edit to the script.

示例提示词

这个轴承座承受交变载荷,帮我用雨流计数和迈纳法则估算疲劳寿命。

所属插件

机械设计与先进制造面向机械设计与制造研究:文本/草图生成 CAD、FreeCAD/SolidWorks 参数化建模、机构综合与零件选型、公差与 GD&T、DFM 与设计评审、快速有限元与应力代理模型、增材制造工艺与缺陷预测、疲劳断裂寿命。

安装该插件会一并获得同领域的 35 个 Skill / MCP。

同插件的其他科研 Skill

常见问题

疲劳断裂寿命分析能做什么?

Estimates fatigue life and crack-growth life for metal parts from S-N data, load histories, and fracture-mechanics inputs. Use for Miner’s-rule damage, rainflow counting, or Paris-law crack growth。

如何安装疲劳断裂寿命分析?

在科搭 Codar 左侧点击「插件」,在「组件」页签搜索「疲劳断裂寿命分析」并安装;在 Claude Code 或 Codex 中可以发送本页的安装提示词,或下载 Skill 包后解压到 ~/.claude/skills 或 ~/.codex/skills。

疲劳断裂寿命分析需要哪些权限?

该科研 Skill没有声明额外权限;执行具体操作时仍由 Agent 按自身规则请求确认。

疲劳断裂寿命分析经过官方审核吗?

是。SpeedSkills 收录的每个版本都会记录内容摘要与权限声明,当前版本已通过安全审核。