Contents lists available at ScienceDirect Journal of Materials Processing Tech. journal homepage: www.elsevier.com/locate/jmatprotec A novel sound-based belt condition monitoring method for robotic grinding using optimally pruned extreme learning machine Xiaoqiang Zhang a , Huabin Chen a, ⁎ , Jijin Xu a , Xuefeng Song a , Junwei Wang a , Xiaoqi Chen a,b, ⁎ a Shanghai Key Laboratory of Materials Laser Processing and Modification, School of Materials Science and Engineering, Shanghai Jiao Tong University, Shanghai 200240, PR China b Department of Mechanical Engineering, University of Canterbury, Christchurch, New Zealand ARTICLE INFO Keywords: Robotic belt grinding Sound-based belt condition monitoring Grinding ability factor Discrete wavelet deposition Optimally pruned extreme learning machine ABSTRACT A novel method based on sound signals is proposed to realize realtime quantitative monitoring of abrasive belt conditions in robotic grinding system. Robotic belt grinding experiments are carried out on Inconel 718 alloy workpieces. The belt wear styles are observed under scanning electron microscope (SEM). The belt wear con- ditions are quantified by the grinding ability factor. During grinding process, sound signals are obtained con- currently by an omnidirectional capacitance microphone. Fast Fourier Transform (FFT) and Discrete Wavelet Decomposition (DWD) are conducted to separate the belt-condition-related sound signals from the raw signals. Then sound features are extracted to establish a novel data-driven model using Optimally Pruned Extreme Learning Machine (OP-ELM). The model is developed to predict the belt grinding ability factor. Experimental datasets are used to train and validate the established model. The results show that the grinding sound in high frequency region from 10 to 15 kHz is sensitive to the belt wear conditions, and that the OP-ELM model speedily and accurately predicts the belt grinding ability factor after optimizing the hidden layer numbers and kernel function type. Compared with experimental results, the mean absolute percentage errors (MAPE) of the pre- dicted grinding ability factor are less than 0.4% and the maximum absolute percentage errors (MAXE) are within 14%. It is thus concluded that the proposed sound-based monitoring approach using the OP-ELM model is robust in accurately predicting the grinding belt condition despite grinding dynamics involving contact pressures. 1. Introduction At present, robotic grinding finds many applications in polishing, spruing and deburring. However, the absence of accurate material removal model has severely limited its performance. The material removal is closely related to material, grinding parameters and tool conditions. Among them, tool condition is a complicated factor because it is time-varying and multi- factor coupled. As the material is removed, the topography of grinding tool gradually changes, which decreases the material removal ability and in- creases the heat production. These phenomena are especially evident in grinding and polishing difficult-to-machine materials such as superalloys and titanium alloys (Wegener et al., 2011). Therefore, grinding tool mon- itoring is important to achieving accurate grinding. Much attention has been focused on the research issue in recent years. Arunachalam and Vijayaraghavan (2015) used machine vision sensors to assess grinding wheel conditions. Lipiński et al. (2014) applied laser scan- ning microscopy to evaluate the geometric features of the abrasive tool’s topography. However, these direct methods have to stop the rotation of grinding tools, and observe a very small number of abrasives. As such, the direct monitoring methods are inefficient and random. Therefore, many more researchers have devoted their efforts to indirect monitoring meth- odologies based on the process signals, including force, vibration, power and acoustic emission (AE) signals. Feng et al. (2009) integrated grinding force signals with system vibration signals to monitor a micro-grinding wheel conditions. Lezanski (2001) fused multiple sensors including vibra- tion, forces and acoustic emission by a neural network and fuzzy logic al- gorithm to evaluate the grinding wheel conditions. Tian et al. (2017) de- veloped a portable power monitoring system to track process changes and abrasive wheel conditions. Mokbel and Maksoud (2000) studied the spectral amplitude features of AE signals emitted in grinding processes with different grinding wheels and concluded that AE technique was capable to monitor the diamond wheel conditions. Warren Liao et al. (2007) and (Warren Liao, 2010) applied AE features to distinguish the “dull” wheel and “sharp” wheels. Although AE monitoring has been frequently reported, there are still some issues for practical applications, such as the instability of raw https://doi.org/10.1016/j.jmatprotec.2018.05.013 Received 5 February 2018; Received in revised form 18 April 2018; Accepted 7 May 2018 ⁎ Corresponding authors at: Shanghai Key Laboratory of Materials Laser Processing and Modification, School of Materials Science and Engineering, Shanghai Jiao Tong University, Shanghai 200240, PR China. E-mail addresses: hbchen@sjtu.edu.cn (H. Chen), xiaoqi.chen@canterbury.ac.nz (X. Chen). Journal of Materials Processing Tech. 260 (2018) 9–19 0924-0136/ © 2018 Elsevier B.V. All rights reserved. T