Statistics Analysis of the Structural Stress Process Based on the Wavelet Neural Network

Abstract

Aiming to the structure reliability analysis, a large amount of structural loading distribution data are required. In general the simulation or real test are used to get these data. But in fact large machine structure with low fault rate is often unable to collect necessary statistic data. The new developing machine also has not large amount of statistical data due to no precedent of use. And moreover the numerical simulation is very enormous for human and time consumption, which its results need be tested and verified. Considering the advantages and disadvantages of the real test and simulation, statistical analysis of loading process is developed based on the wavelet neural network (WNN), to gain the stress mean and standard deviation, and as well as their respective coefficient. Through the application to the stress statistical analysis of the four-bar arm frame crane, the results show that the measured stress data as samples to train WNN can further ensure actual prediction. It is very high efficiency to predict the loading distribution data using the trained WNN. The stress mean and standard deviation required by the reliability analysis can be obtained. Its results can also meet the requirements of the project.