Applied Mechanics and Materials Submitted: 2014-08-12

ISSN: 1662-7482, Vols. 773-774, pp 139-143 Revised: 2014-11-19

doi:10.4028/www.scientific.net/AMM.773-774.139 Accepted: 2014-11-20
© 2015 The Author(s). Published by Trans Tech Publications Ltd, Switzerland. Online: 2015-07-15

Vibration Condition Monitoring: Latest Trend and Review

K.H. Hui1,a, L.M. Hee2,b, M. Salman Leong3,c and Ahmed M. Abdelrhman4,d
1,2
RAZAK School of Engineering & Advanced Technology, Universiti Teknologi Malaysia, Malaysia
3,4
Faculty of Mechanical Engineering, Universiti Teknologi Malaysia, Malaysia
a
[email protected], [email protected], [email protected], [email protected]

Keywords: Condition Monitoring; Vibration; Condition Monitoring Research.

Abstract. Vibration analysis has proven to be the most effective method for machine condition
monitoring to date. Various effective signal analysis methods to analyze and extract fault signature
that embedded in the raw vibration signals have been introduced in the past few decades such as
fast Fourier transform (FFT), short time Fourier transform (STFT), wavelets analysis, empirical
mode decomposition (EMD), Hilbert-Huang transform (HHT), etc. however, these is still a need for
human to interpret vibration signature of faults and it is regarded as one of the major challenge in
vibration condition monitoring. Thus, most recent researches in vibration condition monitoring
revolved around using Artificial Intelligence (AI) techniques to automate machinery faults detection
and diagnosis. The most recent literatures in this area show that researches are mainly focus on
using machine learning techniques for data fusion, features fusion, and also decisions fusion in
order to achieve a higher accuracy of decision making in vibration condition monitoring. This paper
provides a review on the most recent development in vibration signal analysis methods as well as
the AI techniques used for automated decision making in vibration condition monitoring in the past
two years.

Introduction
Research in machinery condition monitoring has begun since several decades ago. This is a
result of more awareness of the importance of keeping the critical machines such as turbines,
pumps, compressors, etc. to run continuously and safely in order to meet its production demands.
These critical machines are extensively used in the industries such as in power generation, oil and
gas industry, aviation that aim for zero unscheduled downtime. The success of condition monitoring
has thus transformed maintenance strategy in certain industries from corrective and preventive to
condition-based maintenance [1]. Various condition monitoring and faults diagnosis techniques
have been developed such as acoustic emission, pressure, strain gauge measurements, etc. [2] Faults
diagnosis by monitoring machine vibration is proven to be the most effective method for machinery
condition monitoring. Both the conventional vibration analysis methods such as FFT, STFT and the
more recent signal processing methods such as wavelets analysis, EMD, HHT, etc. heavily
dependent on the human experience and interpretation of the results. As a result, the current
researches in vibration condition monitoring are thus mainly on using AI techniques in order to
substitute or reduce the human involvement in results interpretation and decision making in
condition monitoring of machines. This review provides a brief discussion of the latest development
in these subjects.

Latest Development in Vibration Signal Analysis Methods

Vibration signal of machines are directly related to the health condition and the working
condition of the machines [3]. Latest development in vibration signal analysis has seem the trend of
shifting from the frequency based analysis to the time-frequency based analysis, and from the non-
adaptive signal analysis to self-adaptive signal analysis. The latest researches in this area are
highlighted as follow.
Cong et al. [4] proposed a new signal processing method, namely S-SVDR, which is the short-
time matrix series (STMS) based singular value decomposition (SVD) method using singular value

This article is an open access article under the terms and conditions of the Creative Commons Attribution (CC BY) license
(https://creativecommons.org/licenses/by/4.0)
140 International Integrated Engineering Summit 2014

ratio (SVR). A new way to create a matrix for SVD process was proposed by the researchers and
named as STMS. STMS is capable of extracting fault information from raw vibration signals. SVR
spectrum is introduced subsequently by combining both STMS and SVD, and it was reported to
have good local identifications for incipient detection for rolling bearings.
Zheng et al. [5] have proposed a new non-stationary signal analysis method namely the local
characteristic-scale decomposition (LCD). This method was used to decompose raw vibration
signals into a number of intrinsic scale components (ISC) in different scales. LCD is comparable to
empirical mode decomposition (EMD) as shown in Fig. 1. Thus, it is more suitable for non-
stationary signal analysis purposes. It stated that the LCD is leading in running time, reducing
fitting errors, restraining end effect and mode mixing as compared to EMD approach.

Fig. 1: The Decomposition Result Generated by (a) EMD (b) LCD [5]
Li et al. [6] proposed a multiscale slope feature extraction method by utilising wavelet-based
multiresolution analysis. The proposed method was implemented in three steps: Firstly, the
measured vibration signals were processed using discrete wavelet transform (DWT) to obtain a
series of detailed signals at different scales. Secondly, the variances of multiscale detailed signals
were calculated accordingly. Lastly, the wavelet-based multiscale slope features were estimated
from the slope of logarithmic variances. To verify the new proposed method, two experiments; the
researchers conducted the bearing defect identification and gear wear diagnosis. The results showed
that the proposed method was high in accuracy and stability for machinery fault diagnosis
application.
Liu and Han [7] proposed a novel fault feature extraction method based on the local mean
decomposition (LMD) and the multi-scale entropy. The researchers conducted a study on faulty
rolling bearing. The proposed method decomposed the non-stationary bearing faults signal into a
number of product functions. Subsequently, the multi-scale entropy of each product function was
calculated, namely feature vectors. It was reported that the newly proposed method is useful for
rolling bearing faults feature extraction.
Summary of these newly proposed vibration signal analysis methods is showed in Table 1.

Table 1: Summary of the Recently Published Vibration Signal Analysis Methods

Author Proposed Vibration Analysis Method Application

Cong et al. S-SVDR – Short-Time Matrix Series (STMS) based Rolling bearing
Singular Value Decomposition (SVD) method using fault diagnosis
Singular Value Ratio (SVR)
Zheng et al. Local Characteristic-Scale Decomposition (LCD) Rolling bearing
fault diagnosis
Li et al. Multiscale slope feature extraction method Rotating machinery
fault diagnosis
Liu and Han Local Mean Decomposition (LMD) Rolling bearing
fault diagnosis
 Applied Mechanics and Materials Vols. 773-774 141

Review on these papers shows that in general, the latest advancement in vibration signal analysis
techniques revolved around methods for signal decomposition. The results were subsequently used
for fault features extraction from a complex vibration signal self-adaptively. Physical meaning of
the instantaneous frequency (IF) is the major challenge of adaptive time-frequency signal analysis
method [5]. The following section discusses AI techniques used in vibration condition monitoring.

Latest Development in Artificial Intelligence Techniques for Vibration Signal Analysis

AI techniques such as artificial neural networks (ANN), support vector machines (SVM) and
Bayesian networks (BN) have been widely adopted in vibration condition monitoring and many
other fields. In vibration condition monitoring, AI techniques are widely used for faults
classification, data fusion, features reduction, decision making, etc. The latest advancement in AI
techniques employed in machinery condition monitoring is discussed as follow.
Chen et al. [8] proposed a novel intelligent fault diagnosis model namely the multi-kernel
support vector machine (MSVM) optimised by chaotic particle swarm optimization (CPSO). This
method was tested for the diagnosis of roller bearing fault. The multi-kernel function used by the
researchers in this study is constructed using Gaussian function and polynomial functions. The
benefits of MSVM as reported are it only required small sampling; it has a nonlinearity and high
dimension. CPSO is used to determine the optimal parameters for MSVM with high accuracy and
great generalization ability. It was reported that the proposed model is superior as compared to
single kernel or randomly selected parameters MSVM classifier for faults classification purposes.
Tran et al. [9] applied a new type of machine learning architecture namely the deep belief
networks (DBN) to diagnose faults in reciprocating compressor valves. By stacking a number of the
restricted Boltzmann machines (RBM) layer by layer, DBN is formed as shown in Fig. 2. One layer
of binary stochastic hidden units and one layer of binary stochastic visible units formed the RBM.
DBN was reported to be effective for pattern recognition. It was reported that the DBN classifier
was capable of improving valve faults identification.

Fig. 2: Architecture of DBN [9]

Zhao et al. [10] proposed the used of trace ratio criterion linear discriminant analysis (TR-LDA)
to diagnose rolling element bearing faults. Two new scatter matrices to address rolling element
bearing faults dataset with non-Gaussian distribution were developed in the study. The new scatter
matrixes were used to characterize the intra-class compactness and the inter-class separability of
dataset by local labelled information. The advantages of this method were reported to be able to
handle the nonlinear bearing data and also to improve pattern classification performance of faults.
Seera et al. [11] introduced hybrid intelligent models for induction motors condition monitoring.
The proposed model combined the fuzzy min-max (FMM) neural network and the random forest
(RF) model. The proposed models carried out fault diagnosis in four steps. Firstly, provide input
142 International Integrated Engineering Summit 2014

data samples. Then, initiate the FMM learning procedure. Subsequently, the hyperboxes generated
in the previous step were used as the input samples for classification and regression tree (CART).
Lastly, the majority voting scheme used to combine the predictions from an ensemble of trees. The
experiment was conducted in both noise-free and noisy environments. Experimental results showed
good faults classification performances in both environments.
Summary of the latest development of AI techniques in vibration condition monitoring was
shown in Table 2.

Table 2: Summary of the Latest Development of AI Techniques in Vibration Condition Monitoring

Author Latest Researches of Artificial Intelligence Techniques Application

Chen et al. Multi-Kernel Support Vector Machine (MSVM) Rolling bearing
optimised by Chaotic Particle Swarm Optimization fault diagnosis
(CPSO)
Tran et al. Deep Belief Networks (DBN) Reciprocating
compressor valves
fault diagnosis
Zhao et al. Trace Ratio Criterion Linear Discriminant Analysis (TR- Rolling bearing
LDA) fault diagnosis
Seera et al. Hybrid intelligent models consist of Fuzzy Min-Max Induction motor
(FMM) neural network and Random Forest (RF) model fault diagnosis

Conclusion
The utilization of various AI techniques for faults detection and diagnosis was shown to be a
major and latest trend in vibration condition monitoring. It is used mainly to reduce human error
and to increase the performance of automated faults detection and diagnosis. Researches in this area
are anticipated to grow in tandem with the need for more intelligent and more autonomous
condition monitoring in the future.

Acknowledgements
This work is supported by the UTM Flagship Grant (Q.K130000.2409.01G44) and Research
University Grant of UTM (Q.K130000.2540.06H51 and Q.J130000.2524.05H01) financed by the
Ministry of Education, Malaysia. The main author is joint-funded by the Ministry of Education and
Universiti Tun Hussein Onn Malaysia for his Ph.D. study.

References

[1] K. H. Hui, M. H. Lim, M. S. Leong, and A. M. Abdelrhman, “Time-frequency Signal

Analysis in Machinery Fault Diagnosis: Review,” in Advanced Materials Research Vol. 845, 2014,
vol. 845, pp. 41–45.

[2] A. M. Abdelrhman, M. S. Leong, L. M. Hee, and W. K. Ngui, “Application of Wavelet

Analysis in Blade Faults Diagnosis for Multi-Stages Rotor System,” Appl. Mech. Mater., vol. 393,
pp. 959–964, Sep. 2013.

[3] A. M. Abdelrhman, M. S. Leong, M. H. Lim, and K. H. Hui, “Vibration Analysis of Multi

Stages Rotor for Blade Faults Diagnosis,” in Advanced Materials Research Vol. 845, 2014, vol.
845, pp. 133–137.
 Applied Mechanics and Materials Vols. 773-774 143

[4] F. Cong, J. Chen, G. Dong, and F. Zhao, “Short-time matrix series based singular value
decomposition for rolling bearing fault diagnosis,” Mech. Syst. Signal Process., vol. 34, no. 1–2, pp.
218–230, Jan. 2013.

[5] J. Zheng, J. Cheng, and Y. Yang, “A rolling bearing fault diagnosis approach based on LCD
and fuzzy entropy,” Mech. Mach. Theory, vol. 70, pp. 441–453, Dec. 2013.

[6] P. Li, F. Kong, Q. He, and Y. Liu, “Multiscale slope feature extraction for rotating
machinery fault diagnosis using wavelet analysis,” Measurement, vol. 46, no. 1, pp. 497–505, Jan.
2013.

[7] H. Liu and M. Han, “A fault diagnosis method based on local mean decomposition and
multi-scale entropy for roller bearings,” Mech. Mach. Theory, vol. 75, pp. 67–78, May 2014.

[8] F. Chen, B. Tang, T. Song, and L. Li, “Multi-fault diagnosis study on roller bearing based
on multi-kernel support vector machine with chaotic particle swarm optimization,” Measurement,
vol. 47, pp. 576–590, Jan. 2014.

[9] V. T. Tran, F. AlThobiani, and A. Ball, “An approach to fault diagnosis of reciprocating
compressor valves using Teager–Kaiser energy operator and deep belief networks,” Expert Syst.
Appl., vol. 41, no. 9, pp. 4113–4122, Jul. 2014.

[10] M. Zhao, X. Jin, Z. Zhang, and B. Li, “Fault diagnosis of rolling element bearings via
discriminative subspace learning: Visualization and classification,” Expert Syst. Appl., vol. 41, no.
7, pp. 3391–3401, Jun. 2014.

[11] M. Seera, C. P. Lim, S. Nahavandi, and C. K. Loo, “Condition monitoring of induction

motors: A review and an application of an ensemble of hybrid intelligent models,” Expert Syst.
Appl., vol. 41, no. 10, pp. 4891–4903, Aug. 2014.