Using support vector machines and acoustic noise signal for degradation analysis of rotating machinery

Abstract

An automated approach to degradation analysis is proposed that uses a rotating machine's acoustic signal to determine Remaining Useful Life (RUL). High resolution spectral features are extracted from the acoustic data collected over the entire lifetime of the machine. A novel approach to the computation of Mutual Information based Feature Subset Selection is applied, to remove redundant and irrelevant features, that does not require class label boundaries of the dataset or spectral locations of developing defect to be known or pre-estimated. Using subsets of the feature space, multi-class linear and Radial Basis Function (RBF) Support Vector Machine (SVM) classifiers are developed and a comparison of their performance is provided. Performance of all classifiers is found to be very high, 85 to 98%, with RBF SVMs outperforming linear SVMs when a smaller number of features are used. As larger numbers of features are used for classification, the problem space becomes more linearly separable and the linear SVMs are shown to have comparable performance. A detailed analysis of the misclassifications is provided and an approach to better understand and interpret costly misclassifications is discussed. While defining class label boundaries using an automated k-means clustering algorithm improves performance with an accuracy of approximately 99%, further analysis shows that in 88% of all misclassifications the actual class of failure had the next highest probability of occurring. Thus, a system that incorporates probability distributions as a measure of confidence for the predicted RUL would provide additional valuable information for scheduling preventative maintenance.

Envelope analysis and data-driven approaches to acoustic feature extraction for predicting the remaining useful life of rotating machinery

“…This approach does not require a-priori information regarding the spectral location of potential defects and their resonances and determines the relevant spectral features for monitoring using information obtained from the data acquired over the lifetime of the machine only. This feature extraction technique is described in more detail in [7]. Figure 2 illustrates which spectral components are selected using the MI-FSS criterion for given subset sizes of 0, 16, 64 and 128 features.…”

“…Predicting the RUL allows for improved reliability of machinery, scheduling of maintenance prior to failure to prevent machine downtime and the removal of the cost of unscheduled maintenance. Predicting the RUL of a machine has been explored to a much lesser extent in the literature [4,5,7].…”

“…A Data-Driven approach to feature extraction is therefore proposed to addresses this weakness in the EA implementation. This Data-Driven approach utilizes Information Theory for Feature Subset Selection to select only these salient spectral features across the entire spectrum and hence move noisy features from the feature vector [5,7]. This approach does not require any a-priori information regarding the frequency range of potential defects.…”

A treelike framework combining fault diagnosis and RUL prediction

“…DCNN is commonly used for pattern recognition and machine vision tasks, but a single method based on DCNN may not be suitable for processing datasets used for prediction, as such datasets typically do not have true fault labels in practical applications. Based on this problem, some researchers have combined physical analysis with DCNNbased methods to solve the problem of obtaining unlabeled datasets in actual industrial production [18,19]. However, this method is currently less applied in the field of bearings, and existing methods are not suitable for combining the physical analysis of bearings with DL.…”

Βελτιώσεις Στην Προληπτική Συντήρηση Μηχανολογικού Εξοπλισμού

“…Παράλληλα με τις αριθμητικές μεθόδους γίνεται εφαρμογή και τεχνητής νοημοσύνης (artificial intelligence). Οι πιο διαδεδομένες μέθοδοι αφορούν στην χρήση: Τεχνητών νευρωνικών δικτύων (artificial neural networks -ANN) [109], [110] Support vector machines -SVM [111], [112].…”