7-5-1 Neural Network results
7-5-1 Neural Network results
- The fourth NN configuration had structure 7-3-1:
-
input layer with 7 nodes for each of ε, ε,’ and T, the logarithmic functions of ε and ε’, the inverse function of T, and curve peak strain εp
-
hidden layer had 5 nodes
-
output layer had 1 node for flow stress σ prediction
Desired and predicted flow stress vs. strain for test 28
To further improve the information content of the NN input, curve peak strain εp was added to the input features to construct the 7-component input vectors ε, ε', T, ln(ε), ln(ε'), 1/T, εp used for training and testing the 7-5-1 NN. Desired flow stress and predicted flow stress are plotted vs. strain. A generally good fit is verified both in the work hardening and the work softening regions. The 7-5-1 NN gives a much better agreement with experimental data than all the previous NN configurations examined, providing more accurate predictions in the full region of the stress-strain curve, from work hardening to dynamic recrystallisation of the mild steel material. The presence in the input vectors of features accounting for both the analytical relationships existing among the process parameters and the influence of peak strain on the material behaviour modelling allowed for an accurate description of the mild steel material flow stress under hot forging conditions
Who's online
There are currently 0 users and 254 guests online.