Study On Method For Fatigue Reliability Design And Analysis Of Mechanical Structure

Modern mechanical equipment(such as shield tunneling machine,high-speed rail,aero-engine etc.)moving in a large,high performance,complex service environment direction.Once a key component in the equipment is damaged,it will cause major property loss or even catastrophic accidents.Therefore,more stringent requirements are put forward for the reliability of modern mechanical equipment.Fatigue failure is still one of the main threats to mechanical equipment.The classical fatigue failure theory can solve the structural fatigue life prediction problems when the main influence factors of fatigue life(such as mechanical properties of the material,shape and size of parts,surface roughness and periodic alternating loads)are deterministic.However,these factors in fact are randomly distributed around the determined value,some of them have great discreteness.It will make the fatigue life of the same type of mechanical products under the same working condition sometimes vary greatly.The mechanical equipment still has a great potential safety hazard though it has been checked for fatigue safety.The classical fatigue failure theory cannot solve the problem of mechanical structure fatigue failure reliability.The effective way for solving this problem is combining the reliability design method with the anti-fatigue design method and carrying out the reliability-based structural fatigue optimization design.However,the research on this topic is limit.This paper aims to propose a method that can be used to structural fatigue reliability analysis and design.The method employs damage coupled stress-strain field FEA analysis and uncertain modeling method of material's fatigue properties.The main contributions and innovations of the dissertation are as follows:(1)A FEM analysis method that coupled damage is proposed to predict the structural fatigue life.Considering the immediate effect of fatigue damage on FEA calculation,the effective stress is updated during the FEA analysis process,which makes structural fatigue simulation conforms to practical fatigue failure mechanisms.In this paper,Lemaitre's low cycle fatigue damage evolution law is used to couple the cyclic plastic constitutive equation.Then the cyclic stress-strain relationship of AISI H11 is simulated in ABAQUS by defining its low cycle fatigue constitutive relationship in UMAT.(2)Reliability analysis based on probability and interval analysis is carried out for discrete life of strain fatigue.Strain fatigue life reliability analysis methods are proposed based on Manson-Coffin equation.This paper employs the two-parameter Weibull distribution to establish the probability-strain-life curve of AISI H11 by combing maximum likelihood estimation and genetic algorithm;the samples of Coffin-Manson parameters were added by random combination at first.Then the ellipsoid model of AISI H11's Coffin-Manson parameters is established by transforming four-dimensional problems into that of two-dimension.Finally,the fatigue life interval is predicted according to the definition of non-probabilistic reliability index.(3)A method that combing support vector machine and Monte Carlo simulation is proposed to improve the computational efficiency in practical structural fatigue reliability analysis.In order to reduce the number of training samples,the method makes full use of support vector classifier and support vector regression and employs the strategy that an initial training sample pool is added by step by step;for the problems of small failure probability,subset simulation is used to generate the efficient samples for training sample pool,which improves calculation efficiency when estimation accuracy is guaranteed.(4)It's significant to develop a stable and efficient reliability optimization design method for structural fatigue design in practical engineering applications.A hybrid directional step method is proposed for reliability-based design optimization in this paper.The method makes the performance function value smaller than that of the previous point by adding a step,thus keeping the performance function values of iterations presenting a decreasing sequence.The results of numerical examples show that the proposed method presents good robustness and efficient computation for highly nonlinear and concave function problems...