Sixteenth National Convention of Mechanical Engineers and

The selection of proper machining-parameters viz. feed, speed and depth of cut is an important task for obtaining the desired performance in the turning process. This work proposes a neuro-fuzzy control scheme for the selection of feed and speed in a single pass turning. The scheme provides the manufacturer a flexibility of specifying the objectives in linguistic terms. The scheme is dependent on the sensory feedback by force and vibration sensors. A well-trained neural network predicts the surface roughness and dimensional deviation, which in combination determine the quality of the work-piece. A fuzzy set model is developed to give quantitative assessment of the total machining performance encompassing surface finish, dimensional accuracy, cost of production and metal removal rate. The first technique models the inverse process dynamics using feed forward neural network with one hidden layer similar to the work of Azouzi and Guillot (1998), and the second technique uses genetic algorithm for obtaining global optimum solution. It has been found that genetic algorithm offers advantage over the former optimization technique. Obtained results have also been compared by traditional optimization method in metal cutting and relative advantages of the present scheme have been discussed.