Study On The System Reliability Of Structure Based On The Artificial Neural Network And The Elastic Modulus Reduction Method

Identification of failure modes of engineering structures and calculation of system reliability are important foundations to ensure the safety and reliability of engineering structures.The traditional structural failure mode identification method has the problems of path dependence and low calculation efficiency,while the traditional system reliability calculation method relies on failure mode identification and joint failure probability calculation.At the same time,in practical engineering,the relationship between structural random variables and structural function is difficult to express explicitly,which makes the reliability analysis based on function more difficult to realize.BP neural network method can approximate any nonlinear mapping relationship,and solve the problem of implicit function solution by fitting the function.The elastic modulus reduction method(EMRM)puts forward a new elastic modulus adjustment strategy based on the principle of conservation of energy and the load-bearing ratio of elements considering the combined internal force effect,that is,by reducing the elastic modulus of high-load elements,the stiffness damage is simulated,and then the failure evolution of the whole structure is simulated until the structure reaches the limit state,and the failure elements are not identified and marked in the process,so it has the characteristics of no path dependence.In view of this,this paper first proposes a reliability analysis method to solve implicit function structure based on BP neural network method,then combines BP neural network method with EMRM to propose a new method for structural failure mode identification and system reliability analysis(BP-EMRM),and then optimizes this method to improve its calculation efficiency.Finally,the system reliability of complex engineering structures is calculated by using BP-EMRM proposed in this paper.The research content of this paper mainly includes the following parts:(1)In order to solve the difficulty in calculating the reliability of structures with implicit function,in the second chapter of this paper,based on BP neural network method,a reliability analysis method for structures with implicit function is proposed.Firstly,the sample points are randomly sampled according to the probability distribution of random variables,which are substituted into the finite element software for analysis,and the actual response values of the structure under each group of sample points are obtained,and the difference between the allowable response values and the actual response values is taken as the functional value of the structure;Then,the function values of sample points and each group of sample points are used as the training data of BP neural network,and the BP neural network is established and trained to fit the functional function of the structure.Finally,the trained neural network is taken as the functional function of the structure,and the reliability of the structure is calculated by Monte Carlo method.(2)Aiming at the limitations of traditional failure mode identification and system reliability calculation methods,in chapter 3,based on BP neural network method and EMRM method,a new method of structural failure mode identification and system reliability calculation-BP-EMRM method is proposed,which avoids the problems of path dependence of traditional failure modes and traditional system reliability analysis methods relying on failure mode identification.Firstly,the sample points are randomly sampled according to the probability distribution characteristics of random variables,and the sample points are substituted into EMRM method group by group to analyze the ultimate bearing capacity of the structure,and the homogeneous generalized yield function value of each unit section and the overall ultimate load value of the structure in the plastic limit state are obtained.Secondly,the functional functions of each unit section and structural system are established respectively,and their function values under each sample point are calculated.Then,taking the sample points and the functional function values under each sample point as training data,the BP neural network of each unit section and structural system is established and trained respectively.After that,based on the trained neural network and Monte Carlo method,the failure probability and reliability index of each unit section and structural system are calculated respectively,so as to obtain the system reliability of the structure.Finally,the failure probability of each unit section is sorted from large to small,and it is divided into batches according to the order of magnitude,and the section of each batch of units is changed into plastic hinge in turn until the structure becomes a mechanism,thus the main failure modes of the structure are obtained.(3)Through analysis,it is found that there are two main reasons that affect the calculation efficiency of BP-EMRM method in calculating structural system reliability.On the one hand,more sample points are needed when the reliability index of structural system converges,and on the other hand,the finite element calculation time of sample points is longer.In order to improve the calculation efficiency of BP-EMRM method,the fourth chapter of this paper optimizes BP-EMRM method based on Latin hypercube sampling and BP neural network parameters.Firstly,Latin hypercube sampling is used instead of random sampling to study the influence of Latin hypercube sampling on computational efficiency.Then,on this basis,the parameter values of BP neural network are changed,and the influence of BP neural network parameters on calculation efficiency is studied,and the recommended value range of BP neural network parameters when calculating system reliability index by BP-EMRM method is given.The results show that the optimization of BP-EMRM method by Latin hypercube sampling and BP neural network parameters can reduce the number of sample points required for the convergence of structural system reliability index,thus improving the calculation efficiency of BP-EMRM method.(4)In order to further verify the applicability of BP-EMRM method,the optimized BP-EMRM method is used to calculate the system reliability of penstock structure of hydropower station in chapter 5 of this paper.The results show that the optimized BP-EMRM method can correctly and efficiently calculate the system reliability of the penstock structure of hydropower station,and is suitable for the reliability calculation of complex structures.