Development of artificial neural network for surface roughness and machine prediction

Abstract

An ANN is a system triggered by the process of biological neurons with the aim of learning a certain system. This study focuses on development of ArtificialNeuralNetwork (ANN) model for surface roughness and machining prediction. It is easy, precise and based on linear relationship between the neural network output when all of the input parameters are constant at their mean values other than the input parameter which is given to crucial testing and target values of the network. This can be achieved by providing stimulus to the neuronal model, estimating the output, and regulating the weights until the preffered output is attained. The composition of artificial neural network present data where surface roughness (Ra) is taken as output parameter to produce ANN's response.Furthermore, selected sigmoid transfer function has its activation function in determining the actual value of a node in the ANNmodel ANN model.Right selection of machining parameter has been discovered to be a crucial in building a link between quality and productivity in an economic way.In conclusion, the neural network with the most favourable composition gives a productive approach to suggest an objective for surface roughness of the raw material under diverse cutting situations.The highest absolute percentage error in ANN model prediction was found to be 2.31% with average percentage error of 0.31%