Abstract—Machined surface roughness is an important parameter used in the evaluation of the surface integrity of machined parts and components. This paper proposes a new computational intelligence approach to predicting the machined surface roughness in metal machining. In this approach, wavelet packet transform (WPT) is incorporated into artificial neural networks (ANN) to develop two ANN models for predicting average roughness Ra and root-mean-square roughness Rq, respectively. Each model has eight inputs, including the cutting speed, the feed rate, energy of wavelet packets for three cutting force components, and energy of wavelet packets for three cutting vibration components. Forty-five machining experiments were performed to collect relevant data to train and test the ANN models. Based on the test data, the average mean square errors (MSE) were 1.23% for predicting average roughness Ra and 2.85% for predicting root-mean-square roughness Rq. These results show that the ANN models developed from the present study have high prediction accuracy.
Index Terms—Artificial neural networks (ANN), machined surface roughness, predictive modeling, wavelet packet transform (WPT).
Ning Fang and P. Srinivasa Pai are with the College of Engineering, Utah State University, Logan, UT 84322, USA (e-mail: firstname.lastname@example.org, email@example.com).
Cite: Ning Fang and P. Srinivasa Pai, "A New Computational Intelligence Approach to Predicting the Machined Surface Roughness in Metal Machining," International Journal of Machine Learning and Computing vol. 8, no. 6, pp. 524-529, 2018.