关键词: 神经网络, 麦夸特算法, 预测模型, 力学性能

Abstract:

The paper proposes a method of applying the feedforward artificial neural network with the LevenbergMarquardt training algorithm for the problem of composite mechanical properties prediction. By using the second derivative information, the network convergence speed is promoted and the generalization performance is enhanced. Taking the wheat strawreinforced composite for instance, the nonlinear mapping is set up from four influence factors (mold temperature, mold pressure, fibre content and time ) to its tensile strength and toughness. The simulation results show the founded network model has preferable learning and generalization capabilities, which performs effectively in predicting composite mechanical properties. Besides, the model is used to optimize the process parameters of compression molding and find the range of the best parameters.

Key words: neural network;LevenbergMarquardt algorithm;predicting model;mechanical property