Journal of Northeastern University(Natural Science) ›› 2021, Vol. 42 ›› Issue (10): 1444-1450.DOI: 10.12068/j.issn.1005-3026.2021.10.011

• Mechanical Engineering • Previous Articles     Next Articles

Active Learning Algorithm of Structural Reliability Based on Kriging and MCMC

ZHANG Hao-yan, BI Qiu-shi, LI Bo, GUO Guang-yong   

  1. School of Mechanical and Aerospace Engineering, Jilin University, Changchun 130025, China.
  • Revised:2021-01-04 Accepted:2021-01-04 Published:2021-10-22
  • Contact: BI Qiu-shi
  • About author:-
  • Supported by:
    -

Abstract: Because the performance functions of many engineering problems are more complicated and the calculation cost is high in the reliability analysis of mechanical structures, the agent model is often used to fit the implicit performance function to reduce the calculation cost. In order to use less sample information and obtain higher reliability calculation accuracy, the Kriging agent model is combined with the learning function, and an active learning reliability analysis and calculation method is proposed. This method finds the sample points with the best learning effect to update the Kriging model, which improves the fitting accuracy of the model. Markov chain Monte Carlo(MCMC)method is used to evaluate the reliability of the structure, which speeds up the convergence speed of sample points and saves sample space. Analyzing the results of 4 calculation examples shows that compared with the other methods, this method can obtain higher precision calculation results with fewer sample points, and reduce the calculation cost.

Key words: reliability; Kriging; MCMC; active learning; failure probability

CLC Number: