Fault Detection In Rotating Machines Research Articles

Vibration Image Representations for Fault Diagnosis of Rotating Machines: A Review

Rotating machine vibration signals typically represent a large collection of responses from various sources in a machine, along with some background noise. This makes it challenging to precisely utilise the collected vibration signals for machine fault diagnosis. Much of the research in this area has focused on computing certain features of the original vibration signal in the time domain, frequency domain, and time–frequency domain, which can sufficiently describe the signal in essence. Yet, computing useful features from noisy fault signals, including measurement errors, needs expert prior knowledge and human labour. The past two decades have seen rapid developments in the application of feature-learning or representation-learning techniques that can automatically learn representations of time series vibration datasets to address this problem. These include supervised learning techniques with known data classes and unsupervised learning or clustering techniques with data classes or class boundaries that are not obtainable. More recent developments in the field of computer vision have led to a renewed interest in transforming the 1D time series vibration signal into a 2D image, which can often offer discriminative descriptions of vibration signals. Several forms of features can be learned from the vibration images, including shape, colour, texture, pixel intensity, etc. Given its high performance in fault diagnosis, the image representation of vibration signals is receiving growing attention from researchers. In this paper, we review the works associated with vibration image representation-based fault detection and diagnosis for rotating machines in order to chart the progress in this field. We present the first comprehensive survey of this topic by summarising and categorising existing vibration image representation techniques based on their characteristics and the processing domain of the vibration signal. In addition, we also analyse the application of these techniques in rotating machine fault detection and classification. Finally, we briefly outline future research directions based on the reviewed works.

An adaptive vibrational resonance method based on cascaded varying stable-state nonlinear systems and its application in rotating machine fault detection

A weak character signal with low frequency can be detected based on the mechanism of vibrational resonance (VR). The detection performance of VR is determined by the synergy of a weak low-frequency input signal, an injected high-frequency sinusoidal interference and the nonlinear system(s). In engineering applications, there are many weak fault signals with high character frequencies. These fault signals are usually submerged in strong background noise. To detect these weak signals, an adaptive detection method for a weak high-frequency fault signal is proposed in this paper. This method is based on the mechanics of VR and cascaded varying stable-state nonlinear systems (VSSNSs). Partial background noise with high frequency is regarded as a special type of high-frequency interference and an energy source that protrudes a weak fault signal. In this way, high-frequency background noise is utilized in a VSSNS. To improve the detection ability, manually generated high-frequency interference is injected into another VSSNS. The VSSNS can be transformed into a monostable state, bistable state or tristable state by tuning the system parameters. The proposed method is validated by a simulation signal and industrial applications. The results show the effectiveness of the proposed method to detect a weak high-frequency character signal in engineering problems.

Rotating Machine Fault Detection based on Fuzzy Logic and Improved Adaptive Filter

Rotating Machine Fault Detection based on Fuzzy Logic and Improved Adaptive Filter

On-Line Fault Detection in Wind Turbine Transmission System using Adaptive Filter and Robust Statistical Features

To reduce the maintenance cost, avoid catastrophic failure, and improve the wind transmission system reliability, online condition monitoring system is important. Developing effective online fault detection methodology is important. In this paper, an adaptive filtering technique is applied for enhancing the fault impulse signals-to-noise ratio in wind turbine gear transmission systems. Multiple statistical features designed to quantify the impulse signals of the processed signal are extracted for rotating machine fault detection. The multiple dimensional features are then transformed into one dimensional feature. A minimum error rate classifier will be designed based on the transformed one dimensional feature to identify the gear transmission system with defect. Vibration signals collected from wind turbines in the real operation will be used to demonstrate the effectiveness of the presented methodology.

Sparse feature extraction based on periodical convolutional sparse representation for fault detection of rotating machinery

Sparse fault transient extraction is the primary step in rotating machine fault detection. In the present paper, periodical convolutional sparse representation (PCSR) is proposed for reliable separation of fault transients imbedded in raw vibration signals. Specifically, a sparse optimization problem of PCSR is constructed, in which periodical fault transients and harmonic components are sparsely represented by a learned dictionary and Fourier dictionary, and the periodicity and group sparsity of sparse coefficients related to sparse fault transients are also incorporated. Meanwhile, to further promote the sparsity of the sparse coefficients, a non-convex function is also introduced into the optimization problem. In addition, an iterative algorithm is developed to resolve the constructed sparse optimization problem, and the parameter selection method is also investigated to ensure the fault transient extraction ability of PCSR. The performance of the proposed PCSR is assessed via a synthetic and actual vibration signal. The results illustrate that the proposed PCSR has an excellent ability in fault transient extraction and machine fault detection.

A review of stochastic resonance in rotating machine fault detection

Abstract Condition-based monitoring and machine fault detection play important roles in industry as they can ensure safety and reduce breakdown loss. Weak signal detection is an essential stage in many signal processing-based machine fault detection methods because the acquired machine signals are always corrupted by heavy background noise. Stochastic resonance (SR) is a nonlinear phenomenon in which the weak signal can be enhanced with the assistance of proper noise. Due to this distinct merit, SR has been extensively investigated in rotating machine fault detection. Given this, the present study is committed to providing a comprehensive review of SR from history to state-of-the-art methods and finally to research prospects, along with the applications in rotating machine fault detection. First, the classical SR theory including the history, merits and limitations is introduced and discussed, and the basic research progress of SR is reviewed. Second, the modified SR methods designed for processing the rotating machine signals are reviewed and summarized. Third, applications of SR for analyzing different kinds of rotating machine fault signals are introduced. Finally, the open problems, challenges and research prospects of SR in rotating machine fault detection are discussed.

Multipoint Optimal Minimum Entropy Deconvolution and Convolution Fix: Application to vibration fault detection

Minimum Entropy Deconvolution (MED) has been applied successfully to rotating machine fault detection from vibration data, however this method has limitations. A convolution adjustment to the MED definition and solution is proposed in this paper to address the discontinuity at the start of the signal – in some cases causing spurious impulses to be erroneously deconvolved. A problem with the MED solution is that it is an iterative selection process, and will not necessarily design an optimal filter for the posed problem. Additionally, the problem goal in MED prefers to deconvolve a single-impulse, while in rotating machine faults we expect one impulse-like vibration source per rotational period of the faulty element. Maximum Correlated Kurtosis Deconvolution was proposed to address some of these problems, and although it solves the target goal of multiple periodic impulses, it is still an iterative non-optimal solution to the posed problem and only solves for a limited set of impulses in a row. Ideally, the problem goal should target an impulse train as the output goal, and should directly solve for the optimal filter in a non-iterative manner. To meet these goals, we propose a non-iterative deconvolution approach called Multipoint Optimal Minimum Entropy Deconvolution Adjusted (MOMEDA). MOMEDA proposes a deconvolution problem with an infinite impulse train as the goal and the optimal filter solution can be solved for directly. From experimental data on a gearbox with and without a gear tooth chip, we show that MOMEDA and its deconvolution spectrums according to the period between the impulses can be used to detect faults and study the health of rotating machine elements effectively.

Dynamic linear models-based time series decomposition and its application on bearing fault diagnosis

Although various methods have been presented in the literature for rotating machine fault detection, it still remains a huge challenge to accurately extract features from non-stationary vibration signals with a high noise level, typically in the case of rolling element bearings faults diagnosis. Due to its random and non-stationary nature, fault related features are difficult to extract by common techniques and are usually overwhelmed by noise and macro-structural vibrations. In this paper, a new time series decomposition method based on a dynamic linear model is proposed, which provides time domain decomposition of a non-stationary signal into collections of latent components. The signal of a damaged bearing consists of exponentially decaying ringing that occurs periodically at the bearing characteristic defect frequency. Dynamic linear models with time-varying cyclical components provide a more generalized and adaptive description on rolling element bearings’ vibration signal with time-varying cyclical behavior. This allows the precise isolation of latent, quasi-cyclical bearing fault components via inferences on the time-varying parameters which characterize these components. An accelerated whole lifetime test of bearing has been performed to collect vibration data, which is utilized to validate the effectiveness of the proposed method.

A Probability Density Estimation for Fault Detection

In this paper, a time-domain analysis method based on probability density estimation is presented for rotating machine fault detection. Generally, the vibration signals obtained from a rotating machine are time-variant since they are strongly related to the rotational speed that is not constant even in the macro steady state. Since the mostly used signal processing method, the Fourier analysis is only suitable for stationary signals, the development of the joint time-frequency analysis is demanded. Here, the probability density estimation method based on Parzen window is introduced. The probability density function of the vibration signal of the rotating machine is estimated by Parzen window, and a threshold value is predefined to decide the state of the rotating machine. By inspecting the change of the probability density of the vibration signal, the condition of the machine is monitored. Air gap eccentricity and ball cage broken bearing are considered in the experiment section, they are difficult to be detected in frequency domain while the rotating speed is not constant. The validity of the method is proved by the experiment.

Gear Damage Detection Using Impulse Response Wavelet Transform

A novel method for detection the gear wear fault according to impulse response wavelet transform is presented. The continuous wavelet transform is widely recognized as an effective technique for rotating machine fault detection using vibration signal since it can be used to detect both stationary and transitory signals. This advantage makes it very suitable for the detection of singularity generated by localized damage in gearbox. In this study, the impulse response wavelet time-frequency joint representation is used to diagnose the gear wear fault. The experimental result shows that impulse response wavelet time-frequency joint representation is a sensitive indicator of the existence of damage in the gearbox.

Blind source separation: A new pre-processing tool for rotating machines monitoring?

Blind source separation (BSS) is a general signal processing method, which consists of recovering from a finite set of observations recorded by sensors, the contributions of different physical sources independently of the propagation medium and without any a priori knowledge of the sources. Recently, these methods paved a new way for the monitoring or the diagnosis of mechanical systems in a working environment. Actually, we show that BSS allows recovering the vibratory information issued from a single rotating machine working in a noisy environment by freeing the sensor signal from the contribution of other working machines. In that way, BSS can be used as a pre-processing step for rotating machine fault detection and diagnosis.

Rotating machine fault detection based on HOS and artificial neural networks

In order to identify the faults of rotating machinery, classification process can be divided into two stages: one is the signal preprocessing and the feature extraction; the other is the recognition process. In the preprocessing and feature extraction stage, the higher-order statistics (HOS) is used to extract features from the vibration signals. In the recognition process, two kinds of neural network classifier are used to evaluate the classification results. These two classifiers are self-organizing feature mapping (SOM) network for collecting data at the initial stage and learning vector quantization (LVQ) network at the identification stage. The experimental results obtained from HOS as preprocessor to extract the features of fault are clearer than those obtained from the power spectrum. In addition, the recognizable rate by using either SOM or LVQ as classifiers is 100%.

BLIND SOURCE SEPARATION: A TOOL FOR ROTATING MACHINE MONITORING BY VIBRATIONS ANALYSIS?

Blind source separation (BSS) is a general signal processing method, which consists of recovering, from a finite set of observations recorded by sensors, the contributions of different physical sources independently from the propagation medium and without any a priori knowledge on the sources. Such methods are attractive for the monitoring or the diagnosis of mechanical systems. It is shown that BSS allows the vibratory information generated from a single rotating machine working in a noisy environment to be recovered by freeing the sensor signal from the contribution of other working machines. In that way, BSS can be used as a pre-processing step to rotating machine fault detection and diagnosis. In this paper, two possible approaches to solve the BSS problem of rotating machine signals are compared; that is, the temporal or frequential approach. The first method developed initially for temporally white signals is used in an experimental context and it is shown that the results are comparable to the frequential domain approach specially developed for rotating machine signals. These two approaches are tested on real signals from a mechanical testing bench, and the implementations of the different methods as well as their performances are discussed. An example to bearing fault detection is given in the final part, to illustrate the potential of this approach as a pre-processing step to improve the diagnosis.