Development of a Neural Network Simulator for Studying the Constitutive Behavior of Structural Composite Materials

Table 1

Optimized forward and inverse ANN training conditions for the case of 7 material parameters and 70 strains (fiber and composite strains).

Epochs	Training data	Network architecture	Learning rate	Momentum rate	Normalization method for strain data
10,000	4,000	7-80-80-70 (Forward); 70-80-80-1 (Inverse)	0.4	0.6	2-group (Forward); Individual (Inverse)