Mechanical Reliability Models With Failure Dependency And Its Application

Failure dependency is an important factor in the determination of the analysis accuracy of reliability.In the mechanical products(including part and mechanical system),lots of factors can bring failure dependency,and make the dependency produced among the random variables of part’s failure mode,between different failure modes of part as well as different parts of a mechanical system.Therefore,with consideration of the types of the factors causing failure dependency,the overal objective of this dissertation is to address the key challenges and critical issues in static and dynamic reliability modelling of mechanical products with failure dependency.The main contents and results of this thesis list as follows:(1)Part failure mode can be expressed by a function which contains many random variables with or without dependency between them.To solve this type of problem,in light of structural reliability theories,this dissertation presents a unified model for analyzing reliability of a part with one failure mode which is subject to random variable dependency.The method involves RF transformation,RT transformation,AFOSM,and so on.Results indicate that random variable dependency has a significant impact on the estimation accuracy of reliability.(2)For a mechanical system with single input and single output,failure dependency between the parts is considered to be caused by a single factor.To deal with this type of issue,this dissertation develops a saddlepoint approximation based mechanical system reliability analysis method.This method involves saddlepoint approximation theory,numerical integration method,conditional reliability concept,and so on.The main merits of this method are:the resultant equation of system reliability is transformed from a complex N-dimensional integration to an one-dimensional integral,which is substantially simpler and more efficient to solve;the failure dependency can be described more accurately.(3)The issues of part reliability with failure modes dependency and mechanical system reliability with part failure dependency have many common features,and the different feature can be deduced as the existence of the different number of shared random variables.Therefore,both of them are considered as the issues of system reliability with multi-shared random variables.To deal with this type of problem,the corrected-partial least squares regression-response surface method based mechanical system reliability model is developed.The method involves partial least squares regression theory,response surface method,latin hypercube sampling method,and so on.Results indicate that the proposed method provides accurate and computationally efficient estimates of system reliability,and is effective particularly when a response evaluation entails costly finite-element,mesh free,or other numerical analysis.(4)Errors are the main reasons that cause the dynamic failure dependency between the parts,and a type of gear transmission system is chosen as the typical case to study the reliability model developing method of this type of problem which is named as the issue of random error based failure dependency.The method involves dynamic differential equation,the maximum extreme value theory,order statistics theory,random variable inverse transformation theory,and so on.Then the proposed method is employed to analysis the dynamic reliability of the system with gear installation errors,and the effects of failure dependency on the reliability are discussed.(5)A tube hydroforming based precision assembly method is presented to assembly the stator can by considering ultra-thin with large length-radius ratio feature of the stator can,complex structure feature of the stator core and complex assembly relationship between stator can and stator core.Then the proposed reliability analysis method was applied to analysis the assembly reliability of the stator can,and the effects of failure dependency on the estimation accuracy of reliability under certain conditions are discussed.Results indicate that:the designed assembly equipment fulfills the requirements of precision assembly;failure correlation brings a negative correlation relationship between the different assembly performance functions.In summary,based on the types of factors causing the failure dependency,four types of mechanical reliability model with consideration of failure dependency are developed:a unified part reliability model with consideration of random variable dependency,a saddlepoint approximation based system static reliability model considering one shared random variable based failure dependency,a corrected-partial least squares regression-response surface method based system static reliability model with consideration of multi-shared random variables based failure dependency,and a system dynamic reliability model with random error based failure dependency.With the corrected-partial least squares regression-response surface method based system static reliability model applied to analysis the assembly reliability of the stator can of the AP1000 reactor coolant pump,the feasibility of the model is verified and some achievements are obtained.The research of this dissertation resolves some issues existing in developing the reliability model of mechanical product with failure dependency.