On transfer learning for chatter detection in turning using wavelet packet transform and ensemble empirical mode decomposition

Abstract

The increasing availability of sensor data at machine tools makes automatic chatter detection algorithms a trending topic in metal cutting. Two prominent and advanced methods for feature extraction via signal decomposition are Wavelet Packet Transform (WPT) and Ensemble Empirical Mode Decomposition (EEMD). We apply these two methods to time series acquired from an acceleration sensor at the tool holder of a lathe. Different turning experiments with varying dynamic behavior of the machine tool structure were performed. We compare the performance of these two methods with Support Vector Machine (SVM) classifier combined with Recursive Feature Elimination (RFE). We also show that the common WPT-based approach of choosing wavelet packets with the highest energy ratios as representative features for chatter does not always result in packets that enclose the chatter frequency, thus reducing the classification accuracy. Further, we test the transfer learning capability of each of these methods by training the classifier on one of the cutting configurations and then testing it on the other cases. It is found that when training and testing on data from the same cutting configuration both methods yield high accuracies reaching in one of the cases as high as 94% and 91%, respectively, for WPT and EEMD. However, EEMD is shown to outperform WPT in transfer learning applications with accuracy of up to 84%. Therefore, for systems where the movement of the cutting center leads to significant variations in the stiffness of the machine-tool system, we recommend using EEMD over WPT for training a classifier. This is because EEMD retains higher accuracy rates in comparison to WPT when the input data stream deviates from the data that was used to train the classifier.

Chatter detection for milling using novel p-leader multifractal features

“…However, a solid knowledge background about machining dynamics is needed for stability diagrams, and is not practical for industrial users to operate. Moreover, these models often do not account for the effect of the changing dynamics or highly complex machining operations (Yesilli et al, 2019).…”

Intelligent Fault Diagnosis of Machines Based on Adaptive Transfer Density Peaks Search Clustering

“…In the study of diagnosing for a small amount of unlabelled mixed data, transfer learning recently has been applied in the fields of mechanical fault diagnosis by researchers [6]. For example, transfer learning was utilized to detect machine tool chatter based on the detected features of a vibration signal via wavelet packet transform (WPT) and ensemble empirical modal decomposition (EEMD) [7]. Shao et al proposed a method based on deep transfer autoencoder for intelligent diagnosis of bearing faults among different mechanical equipment [8].…”

Chatter Diagnosis in Milling Using Supervised Learning and Topological Features Vector

“…The standard approach for chatter recognition from cutting signals has mostly focused on extracting features by decomposing the time series and combining them with supervised learning algorithms-most commonly Support Vector Machine (SVM). The two most widely used decompositions are Wavelet Packet Transform (WPT) [6][7][8][9][10][11][12][13][14] and Ensemble Empirical Mode Decomposition (EEMD) [14][15][16][17][18]. Both WPT and EEMD require manually preprocessing the signal to identify the most informative parts of the signal which carry chatter signatures, which is characterized by the part of the decomposition whose spectrum contain the chatter frequency.…”

“…Once the informative decompositions are obtained, they are used to compute several time and frequency domain features for chatter classification. Many times the resulting features are too many and they overfit the model; therefore, the traditional tools are often equipped with a feature ranking process to prune the features' vector [7,14,16].…”

“…This requires a dedicated skilled user to examine the signal for each configuration, tag chatter and chatter free cases, and extract the informative decompositions. In other words, classifiers trained using WPT or EEMD features can be difficult to generalize to different cutting conditions even for the same machine [14]. Recently, topological features were also explored for chatter detection in turning with the advantage of bypassing the manual feature extraction phase inherent to WPT and EEMD [19].…”