Prediction Of Stress Concentration Based On Gray Theory And Neural Network

In this article, stress concentration problem existed universally in engineering structure has been studied .On the basis of studies already done, gray system theory has been applied to process experimental data in terms of stresses (or strains) nearly a region of stress concentration and the maximum stress can also be obtained easily .As far as certain structure that has larger error between the result of prediction and approximate calculation, composite model is given to modify gray model .As a try, neural network is proposed to determine the maximum stress nearby stress concentration .combining experiment and numerical calculation .The network structure with 3-6-1 is established based on the fundamental theory and technology of electric measurement of strain gauge .The relationship between the stresses (or strains) at the three points and the maximum stress of stress concentration point is gotten by training network .Through the analysis of feature and applicability of the two models ,the result indicates that the two methods proposed have higher precision and reliability on prediction . Meanwhile, it has provided a sound basis to obtain the maximum stress nearby stress concentration in engineering structure using the method of electric measurement of strain gauge .In addition the preprocessing method of data adopted in neural network takes model to a more extent.