Optimising Machine-Learning-Based Fault Prediction in Foundry Production
IWANN '09 Proceedings of the 10th International Work-Conference on Artificial Neural Networks: Part II: Distributed Computing, Artificial Intelligence, Bioinformatics, Soft Computing, and Ambient Assisted Living
Hi-index | 0.00 |
By integrating the predominances of neural network (NN) and genetic algorithm (GA), it takes aid at heavy ingot casting flaw to optimize the foundry technique parameters. Under the matching schemes of foundry technique parameters that gained with uniform design method, the casting process of ingot was simulated by finite element method (FEM). A neural network was set up to reflect the influence of foundry technique parameter on casting flaw; every scheme is taken as the training sample or test sample. A program was made to combine NN with GA. After 300 generations of GA, the solution is stable. With the optimal casting parameters, the casting flaws are reduced and less than any sample results.