Damage identification method based on improved similar Bayesian calculation

Abstract

The invention relates to a damage identification method based on improved similar Bayesian calculation. According to the damage identification method, a likelihood function of a parameter is not required to be calculated on solving of parameter posterior probability distribution by using similar Bayesian calculation, and the problem that the likelihood function cannot be solved in actual engineering application is solved; when Markov Chain Monte Carlo sampling is carried out, a probability model which is similar to a solved problem is established, then random sampling on the model is carried out, and a statistical characteristic estimated value can be obtained by using the sample and serves as an approximate solution of the original problem. Required calculated quantity on a complex engineering problem is quite high by using a Markov Chain Monte Carlo sampling method; by the damage identification method, a statistical characteristic value which is responded by a structure corresponding to a parameter sample is quickly calculated by a random response surface in a sampling process, a phenomenon that numerical solution is carried out by a finite element model is avoided, so that the calculation efficiency is greatly improved, and the problem that a Bayesian method cannot be implemented under the conditions of multiple parameters and large sample amount due to over large calculated quantity is solved.

Description

technical field [0001] The invention relates to a damage identification method based on improved approximate Bayesian calculation. Background technique [0002] Traditional actual engineering structures have been subjected to complex operating environments and external loads for a long time, and damages of different degrees and types will inevitably occur. With the growth of time, the existing damage will continue to accumulate, resulting in continuous degradation of structural performance. If the damage is not detected in time and effective reinforcement measures are not taken, catastrophic accidents may occur in the structure under extreme conditions. Damage identification as the core content of structural health monitoring system [1] , is one of the research hotspots in related fields in recent years. The existing damage identification methods can be classified into two categories: deterministic and uncertain methods. [2-4] . In the deterministic method, the paramete...

AI Technical Summary

Problems solved by technology

However, the regularization constant in the parameter posterior probability distribution formula is often unable to be solved, and the Markov chain Monte Carlo method is required [17] to obtain an approximate solution for the posterior distribution

When the structural model is complex and contains many unknown parameters, the solution to the Bayesian correction problem will also become very difficult and the amount of calculation will increase significantly

A more common problem is that the expression of the likelihood function cannot be obtained, which greatly reduces the practicability of the Bayesian method

Images