Analysis of Implicit Functional Function Structure Reliability Based on Custom Neural Network  
Author Juan Du


Co-Author(s) Yun He; Haibin Li


Abstract For complex structural systems, because the structure is highly nonlinear and the functional function has no explicit expression, the traditional method can not fit the functional function well. To solve the problem of implicit functional function, a response surface method based on custom neural network is proposed to analyze its structure reliability. This method makes no additional assumptions about the form of function, the exponential function is used as the hidden layer activation function of the neural network, and the properties of a multi-layer neural network can approximate any nonlinear function with arbitrary precision are used to construct the custom neural network structure. Multiple groups of input and response of the structure are obtained by means of numerical simulation or experiment, which can be used as the training sample data to train the neural network. After training, the neural network not only realizes the display and expression of the structural function, but also improves the fitting accuracy of the functional function. The results of numerical examples show that the proposed method has higher computational accuracy. Compared with the polynomial response surface method, the proposed method has a better fitting effect for the implicit functional function of high-dimensional and highly nonlinear structures, which provides an effective modeling and analysis method for the calculation of complex structural systems reliability.


Keywords implicit function, custom neural network, response surface method, structural reliability, fitting precision
    Article #:  RQD26-235

Proceedings of 26th ISSAT International Conference on Reliability & Quality in Design
Virtual Event

August 5-7, 2021