Fault diagnosis of induction motor based on decision trees and adaptive neuro-fuzzy inference

https://doi.org/10.1016/j.eswa.2007.12.010Get rights and content

Abstract

This paper presents a fault diagnosis method based on adaptive neuro-fuzzy inference system (ANFIS) in combination with decision trees. Classification and regression tree (CART) which is one of the decision tree methods is used as a feature selection procedure to select pertinent features from data set. The crisp rules obtained from the decision tree are then converted to fuzzy if-then rules that are employed to identify the structure of ANFIS classifier. The hybrid of back-propagation and least squares algorithm are utilized to tune the parameters of the membership functions. In order to evaluate the proposed algorithm, the data sets obtained from vibration signals and current signals of the induction motors are used. The results indicate that the CART–ANFIS model has potential for fault diagnosis of induction motors.

Introduction

Induction motors are the workhorse of many different industrial applications due to their ruggedness and versatility. Although the induction motors are well constructed and robust, the possibility of faults is inherent due to stresses involved in the conversion of electrical to mechanical energy and vice versa. The faults of induction motors may not only cause the interruption of product operation but also increase costs, decrease product quality and affect the safety of operators. Early detection of incipient faults can minimize breakdown and reduces maintenance time. Furthermore, the availability and reliability of machines will be also increased. Consequently, fault diagnosis for detection of faults in induction motors has been the subject of considerable research in recent years to avoid stoppage of product operation. For increased productivity and safety reason, there has been an increasing demand for automated predictive maintenance and fault diagnosis system.

The most common faults of induction motors are bearing failures, stator phase winding failures, broken rotor bar or cracked rotor end-rings and air-gap irregularities (Acosta, Verucchi, & Gelso, 2006). Different approaches for motors incipient fault detection and diagnosis have been successfully proposed (Benbouzid and Nejjari, 2001, Casimir et al., 2006, Satish and Samar, 2005, Widodo et al., 2007, Yang et al., 2004, Yang and Kim, 2006). Most of these techniques involve vibration analysis and stator current analysis because they are easy to measure and highly reliable. Most of the current research works in motor incipient fault detection and diagnosis focused on integrating two or more intelligent techniques to obtain a hybrid model to utilize the excellent property and capability of individual classifier.

Artificial neural networks (ANNs) have been proven as a reliable technique to diagnose the condition of a motor and have good learning capability. However, ANNs are not interpretable and understandable, and are incapable of explaining a particular decision to the user in a human-comprehensible form. Fuzzy logic is another method, which has been used for fault detection and diagnosis (Benbouzid & Nejjari, 2001). It has the ability of modeling human knowledge in a form of if-then rules using easily understandable linguistic term. It has the capability of transforming linguistic and heuristic terms into numerical values for use in complex machine computation via fuzzy rules and membership functions. The if-then rules as well as the initial parameters of membership functions are normally prepared by an expert. Thus, fuzzy logic requires fine-tuning in order to obtain acceptable rule base and optimize parameters for available data (Shukri, Khalid, Yusuf, & Shafawi, 2004). The individual problems from fuzzy logic or ANN alone can be solved by the integration of both methods. This approach has been applied for motor fault diagnosis (Goode & Chow, 1995).

The adaptive neuro-fuzzy inference system (ANFIS) (Jang, 1993) is a specific kind of neuro-fuzzy classifier approach which integrates the ANNs adaptive capability and the fuzzy logic qualitative approach. ANFIS have been successfully applied for automated fault detection and diagnosis of induction machines (Altug et al., 1999, Shukri et al., 2004). Recently, ANFIS and its combination with other methods were also employed as an enhanced tool for fault classification. Some examples of the combined algorithms are ANFIS with genetic algorithms (Lei, He, Zi, & Hu, 2007) and ANFIS with wavelet transform (Lou & Loparo, 2004) for bearing fault diagnosis. ANFIS has been applied for classifying the faults of induction motor with variable driving speed (Ye, Sadeghian, & Wu, 2006).

The data obtained from measurements is normally high dimension and has a large amount of redundant features. If the data is directly inputted into the classifier, the performance will be significantly decreased. Feature extraction and selection have been utilized for reducing the dimension of data by selecting important features wherein feature extraction means transforming the existing features into a lower dimensional space (Yang, Han, & Yin, 2006). Nevertheless, each feature set contains many redundant or irrelevant features as well as salient features in feature space after the feature extraction has been done. Consequently, there is a need for feature selection procedure to select minimum features which can characterize the machine conditions from the whole feature set (Lei et al., 2007).

In this study, decision tree is utilized as feature selection procedure to remove irrelevant features for the purpose of reducing the amount of data needed to achieve good learning, classification accuracy, compact and easily understood knowledge-base, and a reduction in computational time (Kumar, Jayaraman, & Kulkarni, 2005). It involves an integrated method which combines classification and regression trees (CART) and ANFIS for use in fault diagnosis of induction motors. The proposed approach consists of two stages. First, the CART is performed as a feature selection tool to obtain the valuable features and identifies the structure of classifier in the next iterative step. Second, the ANFIS classifier is used to diagnose the faults of induction motors wherein the parameters of membership functions which are tuned throughout the learning process.

Section snippets

Classification and regression trees (CART)

CART algorithm (Breiman, Friedman, Olshen, & Stone, 1984) is similar to those used in decision tree induction such as ID3 and C4.5 (Quinlan, 1986). One of the major distinctions is that CART induces strictly binary trees through a process of binary recursively partitioning of feature space of a data set (Jang, Sun, & Mizutani, 1996). The trees produced by CART also consist of internal nodes (with two children) and terminal nodes or leaf nodes (without children). Each internal node is associated

Proposed system fault diagnosis

In this work, the vibration signals and current signals are utilized for detecting the faults of induction motors. The proposed system consists of four procedures as in Fig. 3: data acquisition, feature calculation, feature reduction and fault classification which are specifically explained in the next section. In this section, the summary role of each procedure is described as follows:

  • Data acquisition: this procedure is used to attain the vibration signals and current signals. Furthermore,

Data acquisition

To validate CART–ANFIS model, experiment was carried out using a test-rig which consists of a motor, pulleys, belt, shaft and fan with changeable blade pitch angle that represents the load. The load can be changed by adjusting blade pitch angle or the number of blades. Six induction motors of 0.5 kW, 60 Hz, 4-pole were used to generate data. One of the motors in good condition (healthy) is used for comparison with faulty motors. The others are faulty motors, with rotor unbalance, broken rotor

Conclusion

A combined classification and regression tree (CART) algorithm and adaptive neuro-fuzzy inference system (ANFIS) have been presented to perform fault diagnosis of induction motors. The implementation of CART–ANFIS based classifier requires two consecutive steps. Firstly, CART is utilized to select the relevant features in data set obtained from feature calculation part. The output of CART is decision tree that is employed to product the crisp if-then rule set. Secondly, the structure of ANFIS

References (21)

There are more references available in the full text version of this article.

Cited by (180)

  • A deep transfer learning method based on stacked autoencoder for cross-domain fault diagnosis

    2021, Applied Mathematics and Computation
    Citation Excerpt :

    In addition, some other machine learning methods are also introduced to fault diagnosis. For example, support vector machine (SVM) [13–16], logistic regression (LR), decision tree (DT) [17], random forest (RF) [18,19], deep neural network (DNN) [20–22] etc. These methods transform fault diagnosis problems into machine learning problems and use historical data to train models to achieve fault diagnosis goals.

  • AI techniques in EV motor and inverter fault detection and diagnosis

    2024, AI Techniques in EV Motor and Inverter Fault Detection and Diagnosis
View all citing articles on Scopus
View full text