An application of hybrid Taguchi-ANN to predict tool wear for turning EN24 material

Abstract

This work is an attempt to predict tool wear for turning EN24 material by the hybrid Taguchi-ANN (Taguchi-Artificial Neural Network) method. The objective is to minimize the tool wear. The independent factors are cutting environment, feed rate, depth of cut, nose radius, and tool type. A Spinner numerical control lathe is used to assess performance. As per the Taguchi orthogonal array, 27 experiments are conducted for each value of the uncontrollable factor (spindle vibration). Optimal setting is structured by Taguchi analysis and the response table. The additive model is used to predict the response. Conformity test is carried out to check whether the predicted and experimental values of response are within the range given by the confidence interval. Furthermore, the ANN is used to predict and analyze the tool wear. The result showed that the supremely important parameter is depth of cut and the least important parameter is tool type. The ideal set found is A3, B3, C3, D1, and E3. Through ANN analysis, it is observed that the experimental values are very close to the predicted values of tool wear. The predicted value at optimal setting is 0.0401 mm. The experimental values at optimal setting is 0.0422 mm. In addition, the study showed that when the feed rate and nose radius are both set to high levels and the depth of cut is medium, using an uncoated tungsten carbide tool with minimal lubrication results in the least amount of tool wear.