Carbon fiber composite lightning strike damage residual strength prediction method based on physical information neural network

Abstract

The present application relates to the technical field of composite lightning damage assessment and intelligent prediction, aiming at the problems of long test evaluation period, high cost, poor generalization ability under small sample condition and lack of physical consistency of traditional test, a lightning damage residual strength prediction method for carbon fiber composite material based on physical information neural network is disclosed. The method firstly acquires multi-source data after lightning and extracts cross-scale characteristic parameters; a physical information neural network model is constructed, and the non-steady heat conduction equation, the current continuity equation and the residual strength degradation constraint are explicitly introduced into the loss function as physical constraint terms; a phased training strategy driven by test and multi-physical field simulation data is adopted to obtain the lightning damage residual strength prediction model, and finally the residual strength ratio is output and the damage grade is divided. The present application significantly improves the precision and generalization ability of residual strength prediction under small sample condition, and has the advantages of strong physical interpretability and convenient engineering implementation.

Images