Development of a methodology for solving problems of predicting the deformed shape of a plate under load using neural networks

Abstract

A methodology for solving problems of predicting the deformed shape of a plate using neural networks has been developed. The application of CAD/CAE systems for research and obtaining empirical data of stress-strain state of structures, as well as for training neural networks is considered. The developed methodology has been tested on a workpiece presented in the form of a plate. As a result of neural networks training, the following data were obtained: data for determination of displacements of the points of deformed surface of the plate by pre-known coordinates of the loads application; data for determination the coordinates of the points of loads application by the pre-known points displacements on the plate surface. The results of forecasts of neural networks with empirical data are analyzed.