Deep Learning Approach for Vibration Signals Applications

This study discusses convolutional neural networks (CNNs) for vibration signals analysis, including applications in machining surface roughness estimation, bearing faults diagnosis, and tool wear detection. The one-dimensional CNNs (1DCNN) and two-dimensional CNNs (2DCNN) are applied for regression and classification applications using different types of inputs, e.g., raw signals, and time-frequency spectra images by short time Fourier transform. In the application of regression and the estimation of machining surface roughness, the 1DCNN is utilized and the corresponding CNN structure (hyper parameters) optimization is proposed by using uniform experimental design (UED), neural network, multiple regression, and particle swarm optimization. It demonstrates the effectiveness of the proposed approach to obtain a structure with better performance. In applications of classification, bearing faults and tool wear classification are carried out by vibration signals analysis and CNN. Finally, the experimental results are shown to demonstrate the effectiveness and performance of our approach.

Download Full-text

Wavelet Analysis and Envelope Detection For Rolling Element Bearing Fault Diagnosis—Their Effectiveness and Flexibilities

Journal of Vibration and Acoustics ◽

10.1115/1.1379745 ◽

2001 ◽

Vol 123 (3) ◽

pp. 303-310 ◽

Cited By ~ 272

Author(s):

Peter W. Tse ◽

Y. H. Peng ◽

Richard Yam

Keyword(s):

Fault Diagnosis ◽

Wavelet Analysis ◽

Rolling Element Bearing ◽

Frequency Spectra ◽

Bearing Faults ◽

Time Frequency ◽

Envelope Detection ◽

Rolling Element ◽

Element Bearing ◽

Bearing Characteristic

The components which often fail in a rolling element bearing are the outer-race, the inner-race, the rollers, and the cage. Such failures generate a series of impact vibrations in short time intervals, which occur at Bearing Characteristic Frequencies (BCF). Since BCF contain very little energy, and are usually overwhelmed by noise and higher levels of macro-structural vibrations, they are difficult to find in their frequency spectra when using the common technique of Fast Fourier Transforms (FFT). Therefore, Envelope Detection (ED) is always used with FFT to identify faults occurring at the BCF. However, the computation of ED is complicated, and requires expensive equipment and experienced operators to process. This, coupled with the incapacity of FFT to detect nonstationary signals, makes wavelet analysis a popular alternative for machine fault diagnosis. Wavelet analysis provides multi-resolution in time-frequency distribution for easier detection of abnormal vibration signals. From the results of extensive experiments performed in a series of motor-pump driven systems, the methods of wavelet analysis and FFT with ED are proven to be efficient in detecting some types of bearing faults. Since wavelet analysis can detect both periodic and nonperiodic signals, it allows the machine operator to more easily detect the remaining types of bearing faults which are impossible by the method of FFT with ED. Hence, wavelet analysis is a better fault diagnostic tool for the practice in maintenance.

Download Full-text

Feature Extraction of Impulse Faults for Vibration Signals Based on Sparse Non-Negative Tensor Factorization

Applied Sciences ◽

10.3390/app9183642 ◽

2019 ◽

Vol 9 (18) ◽

pp. 3642

Author(s):

Lin Liang ◽

Haobin Wen ◽

Fei Liu ◽

Guang Li ◽

Maolin Li

Keyword(s):

Feature Extraction ◽

Fourier Transform ◽

Mechanical Equipment ◽

Tensor Factorization ◽

Short Time Fourier Transform ◽

Vibration Signals ◽

Frequency Distributions ◽

Time Frequency ◽

Value Decomposition ◽

Short Time

The incipient damages of mechanical equipment excite weak impulse vibration, which is hidden, almost unobservable, in the collected signal, making fault detection and failure prevention at the inchoate stage rather challenging. Traditional feature extraction techniques, such as bandpass filtering and time-frequency analysis, are suitable for matrix processing but challenged by the higher-order data. To tackle these problems, a novel method of impulse feature extraction for vibration signals, based on sparse non-negative tensor factorization is presented in this paper. Primarily, the phase space reconstruction and the short time Fourier transform are successively employed to convert the original signal into time-frequency distributions, which are further arranged into a three-way tensor to obtain a time-frequency multi-aspect array. The tensor is decomposed by sparse non-negative tensor factorization via hierarchical alternating least squares algorithm, after which the latent components are reconstructed from the factors by the inverse short time Fourier transform and eventually help extract the impulse feature through envelope analysis. For performance verification, the experimental analysis on the bearing datasets and the swashplate piston pump has confirmed the effectiveness of the proposed method. Comparisons to the traditional methods, including maximum correlated kurtosis deconvolution, singular value decomposition, and maximum spectrum kurtosis, also suggest its better performance of feature extraction.

Download Full-text

Stationarity test of vibration signals with surrogate data and time–frequency analysis

Advances in Structural Engineering ◽

10.1177/1369433216677602 ◽

2016 ◽

Vol 20 (8) ◽

pp. 1143-1154

Author(s):

Zuo-Cai Wang ◽

Feng Wu ◽

Wei-Xin Ren

Keyword(s):

Frequency Analysis ◽

Surrogate Data ◽

Transform Method ◽

Vibration Signals ◽

Time Frequency Analysis ◽

Time Frequency ◽

Analysis Methods ◽

Stationarity Test ◽

Modulated Signals ◽

Short Time

The stationarity test of vibration signals is critical for the extraction of the signal features. In this article, the surrogate data with various time–frequency analysis methods are proposed for stationary test of vibration signals. The surrogate data are first generated from the Fourier spectrum of the original signal with keeping the magnitude of the spectrum unchanged and replacing its phase by a random sequence. The local and global spectra of the original signal and the surrogate data are then estimated by four time–frequency analysis methods, which are short-time Fourier transform, multitaper spectrograms, wavelet transform, and S-transform methods. The index of nonstationarity is then defined based on the distances between the local and global spectra. Three kinds of synthetic signals, which are stationary signals, frequency-modulated signals, and amplitude-modulated signals, are tested to compare the efficiency of the four time–frequency analysis methods as mentioned. The results show that with a certain observation scale value, the index of nonstationarity based on the short-time Fourier transform or wavelet transform method may fail to test the stationarity of the signal. The parametric studies and sensitivity analysis of the observation scale and noise-level effect are also extensively conducted. The results show that the index of nonstationarity calculated using the multitaper spectrograms’ method is more suitable for stationarity test of frequency-modulated signals, while the index of nonstationarity calculated using the S-transform method is more suitable for stationarity test of amplitude-modulated signals. The results also show that the noise has a significant effect on the stationarity test results. Finally, the stationarity of a real vibration signal measured from a cable is tested, and the results show that the proposed index of nonstationarity can effectively test the stationarity of real vibration signals.

Download Full-text

Application of Choi—Williams Reduced Interference Time Frequency Distribution to Machinery Diagnostics

Shock and Vibration ◽

10.1155/1995/721969 ◽

1995 ◽

Vol 2 (6) ◽

pp. 437-444 ◽

Cited By ~ 2

Author(s):

Howard A. Gaberson

Keyword(s):

Fourier Transform ◽

Vibration Signal ◽

Test Cases ◽

Short Time Fourier Transform ◽

Vibration Signals ◽

Time Frequency ◽

Machinery Diagnostics ◽

Short Time ◽

Amplitude And Frequency Modulation ◽

Time Frequency Distribution

This article discusses time frequency analysis of machinery diagnostic vibration signals. The short time Fourier transform, the Wigner, and the Choi–Williams distributions are explained and illustrated with test cases. Examples of Choi—Williams analyses of machinery vibration signals are presented. The analyses detect discontinuities in the signals and their timing, amplitude and frequency modulation, and the presence of different components in a vibration signal.

Download Full-text

Overview of Time-Frequency Analysis Techniques in Vibration Signals of Rotating Machinery

Applied Mechanics and Materials ◽

10.4028/www.scientific.net/amm.684.124 ◽

2014 ◽

Vol 684 ◽

pp. 124-130

Author(s):

Hong Li ◽

Qing He ◽

Zhao Zhang

Keyword(s):

Frequency Analysis ◽

Local Characteristic ◽

Vibration Signals ◽

Time Frequency Analysis ◽

Hilbert Huang Transform ◽

Time Frequency ◽

Analysis Techniques ◽

Advantages And Disadvantages ◽

Analysis Methods ◽

Short Time

There is very rich fault information in vibration signals of rotating machineries. The real vibration signals are nonlinear, non-stationary and time-varying signals mixed with many other factors. It is very useful for fault diagnosis to extract fault features by using time-frequency analysis techniques. Recent researches of time-frequency analysis methods including Short Time Fourier Transform, Wavelet Transform, Wigner-Ville Distribution, Hilbert-Huang Transform, Local Mean Decomposition, and Local Characteristic-scale Decomposition are introduced. The theories, properties, physical significance and applications, advantages and disadvantages of these methods are analyzed and compared. It is pointed that algorithms improvement and combined applications of time-frequency analysis methods should be researched in the future.

Download Full-text

Wavelet Transform in Vibration Analysis for Mechanical Fault Diagnosis

Shock and Vibration ◽

10.1155/1996/375635 ◽

1996 ◽

Vol 3 (1) ◽

pp. 17-26 ◽

Cited By ~ 10

Author(s):

W.J. Wang

Keyword(s):

Wavelet Transform ◽

Time Scale ◽

Vibration Analysis ◽

Three Dimensional ◽

Two Dimensional ◽

Vibration Signals ◽

Time Frequency ◽

Special Cases ◽

Mechanical Faults ◽

Short Time

The wavelet transform is introduced to indicate short-time fault effects in associated vibration signals. The time-frequency and time-scale representations are unified in a general form of a three-dimensional wavelet transform, from which two-dimensional transforms with different advantages are treated as special cases derived by fixing either the scale or frequency variable. The Gaussian enveloped oscillating wavelet is recommended to extract different sizes of features from the signal. It is shown that the time-frequency and time-scale distributions generated by the wavelet transform are effective in identifying mechanical faults.

Download Full-text

A New Hybrid Diagnosis of Bearing Faults Based on Time-Frequency Images and Sparse Representation

Traitement du signal ◽

10.18280/ts.370604 ◽

2020 ◽

Vol 37 (6) ◽

pp. 907-918

Author(s):

Ilhan Aydin ◽

Seyfullah Kaner

Keyword(s):

Deep Learning ◽

Induction Motors ◽

Early Stage ◽

Training Data ◽

Bearing Faults ◽

Time Frequency ◽

Histograms Of Oriented Gradients ◽

Diagnosis Method ◽

And Performance ◽

Learning Machine

Induction motors are an essential component of many applications in industry due to their robust and simple construction. Since bearing faults are the most occurred fault type in the induction motors, it is important to implement the fault detection procedure at an early stage to prevent a sudden interruption of industrial systems. In recent years, deep learning-based techniques have become important tools for converting raw data into images and for producing high-quality images. However, deep learning-based techniques are still difficult to apply in real-time because the techniques require large training data, which slows down the learning process. In the present study, we propose a novel bearing faults diagnosis method at different operating speeds and load conditions. We obtain the time-frequency (TF) representation by applying continuous wavelet analysis to the raw vibration signals. The results of TF representation is recorded as an image. We apply co-occurrence Histograms of Oriented Gradients (coHOG) to the image to obtain features and classify the features with extreme learning machine with a sparse classifier (ELMSRC) to diagnose faults. We obtained better results in terms of time and performance compared with the proposed method of other classification and deep learning techniques.

Download Full-text

Tool Wear Monitoring with Vibration Signals Based on Short-Time Fourier Transform and Deep Convolutional Neural Network in Milling

Mathematical Problems in Engineering ◽

10.1155/2021/9976939 ◽

2021 ◽

Vol 2021 ◽

pp. 1-14

Author(s):

Zhiwen Huang ◽

Jianmin Zhu ◽

Jingtao Lei ◽

Xiaoru Li ◽

Fengqing Tian

Keyword(s):

Neural Network ◽

Fourier Transform ◽

Tool Wear ◽

Convolutional Neural Network ◽

Deep Convolutional Neural Network ◽

Tool Wear Monitoring ◽

Short Time Fourier Transform ◽

Vibration Signals ◽

Wear Monitoring ◽

Short Time

Tool wear monitoring is essential in precision manufacturing to improve surface quality, increase machining efficiency, and reduce manufacturing cost. Although tool wear can be reflected by measurable signals in automatic machining operations, with the increase of collected data, features are manually extracted and optimized, which lowers monitoring efficiency and increases prediction error. For addressing the aforementioned problems, this paper proposes a tool wear monitoring method using vibration signal based on short-time Fourier transform (STFT) and deep convolutional neural network (DCNN) in milling operations. First, the image representation of acquired vibration signals is obtained based on STFT, and then the DCNN model is designed to establish the relationship between obtained time-frequency maps and tool wear, which performs adaptive feature extraction and automatic tool wear prediction. Moreover, this method is demonstrated by employing three tool wear experimental datasets collected from three-flute ball nose tungsten carbide cutter of a high-speed CNC machine under dry milling. Finally, the experimental results prove that the proposed method is more accurate and relatively reliable than other compared methods.

Download Full-text

Lightweight Convolutional Neural Network and Its Application in Rolling Bearing Fault Diagnosis under Variable Working Conditions

Sensors ◽

10.3390/s19224827 ◽

2019 ◽

Vol 19 (22) ◽

pp. 4827 ◽

Cited By ~ 8

Author(s):

Hengchang Liu ◽

Dechen Yao ◽

Jianwei Yang ◽

Xi Li

Keyword(s):

Neural Network ◽

Fault Diagnosis ◽

Convolutional Neural Network ◽

Rolling Bearing ◽

Time Frequency ◽

Bearing Fault ◽

Bearing Fault Diagnosis ◽

Time Domain Signal ◽

The One ◽

Short Time

The rolling bearing is an important part of the train’s running gear, and its operating state determines the safety during the running of the train. Therefore, it is important to monitor and diagnose the health status of rolling bearings. A convolutional neural network is widely used in the field of fault diagnosis because it does not require feature extraction. Considering that the size of the network model is large and the requirements for monitoring equipment are high. This study proposes a novel bearing fault diagnosis method based on lightweight network ShuffleNet V2 with batch normalization and L2 regularization. In the experiment, the one-dimensional time-domain signal is converted into a two-dimensional Time-Frequency Graph (TFG) using a short-time Fourier transform, though the principle of graphics to enhance the TFG dataset. The model mainly consists of two units, one for extracting features and one for spatial down-sampling. The building units are repeatedly stacked to construct the whole model. By comparing the proposed method with the origin ShuffleNet V2, machine learning model and state-of-the-art fault diagnosis model, the generalization of the proposed method for bearing fault diagnosis is verified.

Download Full-text

SECOND CARDIAC SOUND ANALYSIS TECHNIQUES AND PERFORMANCE COMPARISON

Journal of Mechanics in Medicine and Biology ◽

10.1142/s021951940500162x ◽

2005 ◽

Vol 05 (03) ◽

pp. 429-442 ◽

Cited By ~ 1

Author(s):

S. M. DEBBAL ◽

F. BEREKSI-REGUIG

Keyword(s):

Fourier Transform ◽

Wavelet Transform ◽

Wigner Distribution ◽

Performance Comparison ◽

Frequency Characteristics ◽

Second Sound ◽

Time Frequency ◽

Cardiac Sound ◽

And Performance ◽

Short Time

This paper presents the applications of the spectrogram, Wigner distribution and wavelet transform analysis methods to the second cardiac sound S2 of the phonocardiogram signal (PCG). A comparison between these methods has shown the resolution differences between them. It is found that the spectrogram Short-Time Fourier Transform (STFT) cannot detect the two internals components of the second sound S2 (A2 and P2, atrial and pulmonary components respectively). The Wigner Distribution (WD) can provide time-frequency characteristics of the sound S2, but with insufficient diagnostic information as the two components (A2 and P2) are not accurately detected, appearing to be one component only. It is found that the wavelet transform (WT) is capable of detecting the two components, the aortic valve component A2 and pulmonary valve component P2, of the second cardiac sound S2. However, the standard Fourier transform can display these components in frequency but not the time delay between them. Furthermore, the wavelet transform provides more features and characteristics of the second sound S2 that will hemp physicians to obtain qualitative and quantitative measurements of the time-frequency characteristics.

Download Full-text