Sign in Subscribe
Fatigue (Materials)

Evidence-Enhanced Data-Model-Driven Approach for Predicting Fatigue Crack Propagation

Strategic Insights

— 1 min read

Researchers from Zhejiang University have developed a novel approach for simulating fatigue crack propagation in steel structures, utilizing advanced surrogate modeling techniques and data-driven physical model updating. The study aims to provide reliable decision-making basis for the dynamic life management and maintenance of steel structures.

The researchers proposed an evidence-enhanced data-model-driven approach, integrating advanced surrogate modeling techniques with data-driven physical model updating based on measurement. This approach utilizes a sparrow search algorithm-optimized deep neural network (SSA-DNN) to determine the stress intensity factor (SIF) efficiently, reducing computation time from 300-600 seconds to 0.005-0.01 seconds. Additionally, a dynamic Bayesian network (DBN) is employed to integrate crack propagation data and update physical models effectively combined with importance sampling.

Key Takeaways:

  • The proposed approach integrates advanced surrogate modeling techniques with data-driven physical model updating based on measurement.
  • The SSA-DNN is used to determine the SIF efficiently, reducing computation time from 300-600 seconds to 0.005-0.01 seconds.
  • A DBN is employed to integrate crack propagation data and update physical models effectively combined with importance sampling.
  • The results demonstrate that the prediction error for fatigue crack propagation life is maintained within 5%.
  • The study presents a robust and generalizable framework for modeling dynamic crack propagation.
  • The approach can be extended to engineering digital twins for reliable decision-making basis.
  • The research has been peer-reviewed and published in the journal Engineering Failure Analysis.

Statistics:

  • Computation time reduced from 300-600 seconds to 0.005-0.01 seconds using the SSA-DNN.
  • Prediction error for fatigue crack propagation life maintained within 5%.
  • The study provides a robust and generalizable framework for modeling dynamic crack propagation.
  • The research is supported by the National Natural Science Foundation of China (NSFC), Continuation Project of the Zhejiang Provincial Natural Science Foundation for Distinguished Young Scientists, Fundamental Research Funds for the Central Universities, and Sichuan Provincial Transportation Technology Project.

Sources:

  • An Evidence-enhanced Data-model-driven Approach To Dynamic Prediction of Fatigue Crack Propagation Life. Engineering Failure Analysis, 2025;181.
  • Hangzhou, People's Republic of China, Asia, Data Modeling, Information Technology, Zhejiang University, [Source: NewsRx. Reports from Zhejiang University Provide New Insights into Data Modeling (An Evidence-enhanced Data-model-driven Approach To Dynamic Prediction of Fatigue Crack Propagation Life). Information Technology Newsweekly. November 4, 2025; p 641.

Read more