Optimal brain surgeon pruning of neural network models of manufacturing processes

Abstract EN

In this paper ,Optimal Brain Surgeon (OBS) pruning algorithm is proposed to optimize network architecture with respect to testing patterns error and overcoming the over fitting problem .

Turning process is used as case study to improve the performance of the neural network – surface roughness model .Using the proposed algorithm reduced the prediction error on testing patterns from 0.6237 to 0.2854 based on the absolute percent error estimate .Also.

noticeable improvement is made in correlation coefficient from 0.8656 to 0.9807 making the network more reliable for new operating conditions.