Generalized fuzzy model for metal cutting data selection (original) (raw)
Related papers
This paper outlines a comparative study of different Artificial Neural Network models and 'adaptive-networks-based fuzzy inference systems' approach for predicting the surface roughness (Ra) using cutting speed (v), feed rate (fr) and cutting depth (d) when machining Nickel-based super alloys with uncoated carbide tool under milling conditions. Nickel-based super alloys are generally known to be one of the most difficult materials to machine because of hardness, high strength at high temperature, affinity to react with the tool materials and low thermal diffusivity. In this work, it was develop an accurate and reliable model for prediction of surface roughness for Nickel-based super alloys in end milling operations. In this study, to ensure the effectiveness of connectionist techniques, different models such as multi-layered perceptron, Elman recurrent neural network, radial basis function network and Adaptive-Networks-based Fuzzy Inference Systems were designed and optimal networks parameters were found. All the designed networks modeling and prediction ability were found using root mean square errors and correlations analysis. Finally, it was concluded that Elman recurrent neural network and adaptive-networks-based fuzzy inference systems could be used to predict the surface roughness for nickel-based super alloys in end milling process compared to the other connectionist models.