Study On Related Problems Of Fatigue Life Prediction Of Welded Structures

The welded structures have been widely used in many fields such as mechanical engineering, vehicles ships, marine engineering, petroleum chemical industry and so on. Fatigue failure is one of the main failure modes for some important welds components, because they need to suffer more complex and random loads in practice. Probability characteristics on the fatigue life of the welded structures are studied by using probability theory and method in this paper, which is important to objectively predict the fatigue life and reliability of the welded structures.The prediction on the probability of fatigue life of the welded structures is as research object in this paper. The main research contents are as follows:(1) Fatigue life test and stress analysis of the welded structures of high-speed rail vehicle body are done in the laboratory condition. According to the characteristics of vehicle body, two sets of samples (two kinds of samples applied different loads in each set) are used in fatigue life comparative test. For each experiment, the samples applied by sample axial stretch loading or complex bending loading are test fatigue life. The fatigue experimental results are obtained and used to establish two kinds of static models by the finite element software ANSYS for two samples suffered on different loading. The stress of their danger points is calculated. By comparing the fatigue life data of two different samples, fracture position, and stress distribution simulated by finite element software, the feasibility that complex bending fatigue test is substitute by simple stretch fatigue test is discussed. In addition, the affect and correction of mean stress are analyzed.(2) The information extracting method for and effects different fatigue experimental data are analyzed and compared. Based on the experimental data, by using probability and statistics methods, the life data without censored data are statistically analyzed and the distribution types of the life data and P-S-N curve are fitted The life data with censored data are estimated by the parameters used the maximum likelihood estimation method and K-S test. Compared with two processing results, the effect of censored data should be fully considered and be not ignored for different kinds of experimental data. So the method that censored data are ignored for traditional processing procedure of fatigue life data can underestimate fatigue properties of the samples to some extent (including the life data of sample with overflow and fracturing in non-standard distance segment).(3) Taking the hydraulic supports (large welded structure) as the background, the stress distributions of the hydraulic support in different conditions are calculated by using the finite element software. Simultaneously, the true stress values are obtained through actual structure tests at the same condition and the dangerous position bracket is detected.Real loading process of the hydraulic support is further studied by the mutual mechanism of the hydraulic support and surrounding rock of underground. Through the Monte Carlo numerical simulation method, load spectrum reflecting complexity and random characteristics of the suffered load is programmed. Service life of the hydraulic support is predicted by using local should predict method and Miner cumulative damage theory.(4) The reliability of the structure of the hydraulic support is valuated. The probabilistic characteristics of the load processes of the hydraulic support are described as macro and microaspects. The cyclic variable amplitude load is equivalent to constant amplitude cyclic load with the force amplitude according to damage equivalent principle. During ergodic stationary random process, the Miner linear cumulative damage rule is used to calculate the results. On this basis, the fatigue reliability model of the hydraulic support parts (many parts damage structure as a series system) is established with the help of calculated mechanism of the total probability and statistical average algorithm. The model can objectively and truly reflect the failure correlation between different damage parts.