A Neural Network Based Real Time Controller for Turning Process

In this paper, the design and implementation of an effective neural network model for turning process identification as well as a neural network controller to track a desired vibration level of the turning machine is as an example of using the neural network for manufacturing process control. Multi – Layer Perceptron (MLP) neural network architecture with Levenberg Marquardt (LM) algorithm has been utilized to train the turning process identifier. Two different strategies have been used for training turning process identifier, and for training the controller model, where there is no mathematical model till now could relate the vibration level to the input turning process parameters “feed, speed, and depth of cut”. The vibration signal obtained by the experimental work has been used to train a neural network for identification and control of the turning process. The developed Neuro – controller has been checked by applying different reference vibration signals where it isfound that the controller has good ability to track the reference within maximum settling time that does not exceed (4 sec for 95% of the signal); maximum overshot not exceed (30%) of the reference signal used for checking.