Fatigue state assessment method for metal components based on probability flow learning
Through a probabilistic flow learning method, a small amount of residual stress data is used to construct a data-driven model and a digital-analog fusion model, which solves the challenge of estimating fatigue damage of metal components under variable amplitude load, and achieves efficient and low-cost fatigue state evaluation.
CN115753049BActive Publication Date: 2025-03-28CHONGQING UNIV
2 Cites 0 Cited by
Patent Information
- Application Number
- CN202211416638.4
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2022-11-13
- Publication Date
- 2025-03-28
- Estimated Expiration
- 2042-11-13
AI Technical Summary
Technical Problem
Under variable amplitude load conditions, fatigue damage estimation of metal components is challenging and requires a large amount of degradation data, resulting in high costs.
Method used
Using a probabilistic flow learning method, by constructing a data-driven model and a digital-to-analog fusion model, a small amount of residual stress measurement data is used to achieve fatigue assessment of full life and phased life.
Benefits of technology
This method can effectively predict the fatigue cycle of metal components, improve the confidence of the evaluation results, and greatly reduce the evaluation cost.
✦ Generated by Eureka AI based on patent content.
Abstract
The present invention discloses a method for fatigue state assessment of metal components based on probability flow learning, comprising the following steps: Step 1: Collect data: Step 2: Construct a data-driven model: Map a smooth ordinary differential equation to a vector field to constrain monotonicity; Use a stochastic differential equation to introduce random disturbances into the vector field to obtain a differential vector field; Use induced variables and induced states to encode the mean and variance of Gaussian process regression into drift and diffusion coefficients of the differential vector field; Step 3: Construct a digital-analog fusion model: Use residual stress as a state particle, random differential equations as latent space state equations, and physical models as observation space state equations to construct a digital-analog fusion state space model; Step 4: Fatigue state assessment: Use the fatigue cycle obtained based on the residual and the fatigue cycle obtained by the data-driven model, use the least squares method to output the final predicted fatigue cycle, and realize fatigue state assessment.
Need to check novelty before this filing date? Find Prior Art