Application of EMDWVD and particle filter for gearbox fault feature extraction and remaining useful life prediction
Xin Liu^{1} , Yunxian Jia^{2} , Zewen He^{3} , Jie Zhou^{4}
^{1, 2, 4}Department of Equipment Command and Management, Mechanical Engineering College, Shijiazhuang, China
^{3}Bureau of Beijing Military Representation, Beijing, China
^{1}Corresponding author
Journal of Vibroengineering, Vol. 19, Issue 3, 2017, p. 17931808.
https://doi.org/10.21595/jve.2017.17680
Received 4 September 2016; received in revised form 3 January 2017; accepted 4 January 2017; published 15 May 2017
JVE Conferences
Fault feature extraction and remaining useful life (RUL) prediction are important to condition based maintenance (CBM). In order to realize the fault feature extraction of gearbox vibration signal presenting nonlinear and nonGaussian, the integration of empirical mode decomposition (EMD) and WignerVille distribution (WVD) are proposed in this paper. Taking the kurtosis as standard, the WVD is applied to some IMFs with larger kurtosis to calculate the timefrequency distribution, with an effective suppress on mode mixing and the crossterm interference. Afterwards, particle filter (PF) with the state space model based on Wiener process is proposed to predict the RUL of gearbox considering degradation feature, gearbox teeth wear and nonlinear and nonGaussian system. The gearbox life cycle test shows that the EMDWVD method can extract the valued characteristics of vibration signal accurately, and the particle filter can provide an effective way to predict the RUL of gearbox.
Keywords: EMDWVD, fault feature extraction, particle filter, remaining useful life prediction, vibration signals.
1. Introduction
Gearbox is an important and common transmission component in the rotating machinery. Fault feature extraction and remaining useful life prediction of gearbox under variable conditions is a hot and relatively difficult topic. The vibration analysis techniques have been widely used in the detection of gear faults and the crux is how to extract the fault feature and predict the RUL from the vibration signals. The methods for early detection of faults in a gearbox have already been proposed with the condition monitoring of machinery [1].
EMD and WVD are widely used in the engineering [26]. They are especially suitable for processing the nonlinear and nonGaussian vibration signals. However, they have their own limits; the analysis results are respectively influenced by mode mixing and crossterm interference. A variety of methods are proposed to reduce the mode mixing of EMD. For example, the vibration signals are preprocessed with the Support Vector Machines to suppress the interference of highfrequency intermittent components of EMD [7]. Nevertheless, these methods can only be used to restrain the mode mixing for special frequency. Hence, a novel method is needed to be proposed for mode mixing of EMD.
The WignerVille distribution has been widely used in a variety of fields, such as the diagnosis of induction motors fed at constant frequency under variable speed transient conditions [8] and in an engine fault diagnosis system for feature extraction [9]. Although the potential of the WVD has long been recognized in engineering, its applications are limited mainly due to the crossterm interference [10]. These crossterms mean that the distribution shows energy which does not actually exist at these particular time/frequency coordinates. Many methods have been developed to reduce the existence of crossterms interference [1114]. For example, a novel method based on signal decomposition and modified magnitude group delay function, in which vibration signal decomposition achieved by perfect reconstruction filter bank, is proposed to reduce the existence of crossterms [15]. These methods can restrain the crossterm interference, but another function is introduced into the process of WVD, making the timefrequency distribution complex and different to calculate.
In this case, the integration of EMD and WVD is proposed, which can take the full use of the advantages of EMD and WVD to extract the fault feature of gearbox. Firstly, the EMD is used to obtain the IMFs of vibration signals. Additionally, after we obtain the IMF components, the WVD is applied to some IMFs with larger kurtosis to calculate the timefrequency distribution. Thus, the proposed method can restrain the frequency interference and mode mixing effectively to improve the effectiveness of vibration signal feature extraction.
Accurate prediction of the remaining useful life of the gearbox is important to the CBM of a rotating machinery system. Machine prognostic methods can roughly be categorized into two major classes: modelbased and datadriven methods [16]. Among methods for performing prognostics such as support vector regression [17] and artificial neural networks, particle filters are emerging as a technique with considerable potential. Particle filtering is a Monte Carlobased computational tool particularly useful for Bayesianframed prognostics of nonlinear and/or nonGaussian processes [18, 19]. Particle filter avoids making the simplifying assumptions of linearity and Gaussian noise typical of Kalman filtering (KF) [20], which has been proved very successful for nonlinear and nonGaussian estimation problems [21]. The approach has been used in many areas, including computer vision, target tracking, and robotics. The use of particle filters as a tool for prognostics has been increasing in recent years, and they have been applied to a range of applications, including Lithiumion battery capacity depletion [22], rolling bearings [23] and gearbox plate crack growth prediction [24]. In addition, many new methods are proposed based on particle filtering. For example, a new prognostic method is developed using adaptive neurofuzzy inference systems (ANFISs) and highorder particle filtering [25]. In this paper, the integration of particle filter and the state space model based on Wiener process is proposed to predict the remaining useful life of gearbox considering degradation feature, gearbox teeth wear and nonlinear and nonGaussian system. Finally, the results are compared with the Kalman filter.
The rest of this paper is organized as follows. The main EMD and WVD steps are discussed, as well as the method of fault feature extraction for gearbox based on the integration of EMD and WVD is proposed in Section 2. In Section 3, the particle filter method is presented, which will be used for remaining useful life prediction, and a simulated example is given. In Section 4, the life cycle test of gearbox is carried out. Then, the proposed methods are applied for fault feature extraction and remaining useful life prediction of gearbox. Finally, the conclusions are drawn in Section 5.
2. Fault feature extraction method for gearbox based on EMDWVD
2.1. Short overview of EMD
The EMD is a widelyused timefrequency analysis method in the signal processing field, which mainly includes two steps. Firstly, the EMD is used to decompose nonlinear and nonGaussian time series into a number of IMFs and a residual. Second, each IMF is Hilbert transformed to calculate the timefrequency distribution. The first step is the most important. It makes the original signal linear and stationary time series, as well as produces a series of data sequences with different characteristics, that each sequence is an IMF. The EMD method based on such assumptions: (1) the original signal has at least two extreme points, which are a maximum and a minimum, (2) the characteristic time scale is defined as a time interval between the local extreme value points, (3) if the original signal does not have an extreme point but there is an inflection point, the extreme point can be obtained by differential calculus once or more times, then the decomposed results will be get by integral calculus. Therefore, we can see that EMD is a selfadaptive signal processing method, which especially suits for the nonlinear and nonGaussian signals.
The EMD algorithm exists in two stop conditions: one is aiming at the IMF, the repeating subtracting process will stop until the IMF satisfied the condition which is called component termination condition, the other one is the EMD decomposition termination condition for the original signal, which is namely decomposition termination condition. The EMD decomposition is terminated until the residual is less than the predetermined value or becomes a monotonic function, the EMD decomposition process is over.
The steps of EMD decomposition are shown as follows.
(1) Identify all the local minima and local maxima of the given signal $x\left(t\right)$.
(2) According to the local extrema of the signal, connect the local maxima and minima to construct the upper and lower envelopes $u\left(t\right)$ and $v\left(t\right)$ by a cubic spline interpolation.
(3) Calculate the mean of signal: $m\left(t\right)=\left(u\left(t\right)+v\left(t\right)\right)/2$.
(4) Obtain the new series: $h\left(t\right)=x\left(t\right)m\left(t\right)$.
(5) If $h\left(t\right)$ satisfies the two conditions, take it as the first IMF. Otherwise take $h\left(t\right)$ as the original signal and the above processes are repeated until it satisfies the conditions of IMF.
(6) The residual $r\left(t\right)$ is got after separating ${c}_{1}\left(t\right)$ from the $x\left(t\right)$, until the final residue becomes a constant value, or $x\left(t\right)$ is replaced by $r\left(t\right)$ and the above process is repeated.
The signal $x\left(t\right)$ can be decomposed into $n$ IMFs ${c}_{i}\left(t\right)$ and a residual ${r}_{n}\left(t\right)$ via EMD. The original signal $x\left(t\right)$ can be finally expressed as:
In the process of EMD decomposition, components of different time scales can be found in the same one of the IMFs when decompose a complex signal directly, which is called as mode mixing. In simple terms, there is a rapid change time scale in the signal which makes the different mode composition not to be able being separated effectively according to the characteristic time scales. Then the same IMF contains multiple mode composition inside, so the original signal can’t be clearly reflected the intrinsic nature. And the reasons for this phenomenon has the relationship with the EMD algorithm, signal sampling rate and signal change in a certain time scale.
2.2. Short overview of WVD
Timefrequency distributions have been used extensively for nonlinear and nonGaussian signal analysis. The WignerVille distribution is the best known one for a timefrequency analysis. The WignerVille distribution can be viewed as the signal energy timefrequency joint distribution, and is an important tool for analyzing nonlinear and nonGaussian signal. The autoWignerVille distribution of continuous signal $x\left(t\right)$ is defined as:
Accordingly, assumed another continuous signal $y\left(t\right)$, the crossWignerVille distribution of it is defined as:
where $\omega $ represents frequency, $\tau $ represents time variables. ${x}^{*}$ and ${y}^{*}$ are the complex conjugate of continuous signal $x\left(t\right)$ and $y\left(t\right)$.
The disadvantages of WignerVille distribution are the crossterm interference, which is caused by the time and frequency interference caused by nonlinear transformation. When the original signal contains multiple components, there will be no physical meaning of oscillation component at the midpoint of the timefrequency center coordinates between each of the two components in the WignerVille distribution, which provides a false energy distribution and is affected by the physical interpretation of WignerVille distribution. Take the signal $x\left(t\right)={x}_{1}\left(t\right)+{x}_{2}\left(t\right)$ as an example, the WignerVille distribution is defined:
According to the definition of WVD, $WV{D}_{x1,x2}\left(t,\omega \right)=WV{D}_{x2,x1}*\left(t,\omega \right)$.
$WV{D}_{x1,x2}\left(t,\omega \right)$ and $WV{D}_{x2,x1}\left(t,\omega \right)$ is the conjugate, then:
This suggests that the WVD of two signals summed is not the sum of their each WVD, in addition to two autoWVDs, there is a crossWVD which is named as a crossterm interference. Because the crossterm interference is usually an oscillation, which amplitude can be reached twice as much the autoWVD, which caused the timefrequency signal characteristics is blur.
2.3. The integration of EMD and WVD for gearbox fault feature extraction
The WignerVille distribution can deal with nonlinear and nonGaussian signals to get the time–frequency distribution, which is a powerful tool for signal processing. In order to reduce the effects of its crossterm interference, the WVD and EMD are combined together, as well as, this method can also reduce the mode mixing of EMD. Firstly, the original signal is decomposed into IMFs which frequency components are relatively single via EMD, then some IMFs that are selected according to the specific rules are processed by WVD. As we know, the original vibration signal can be decomposed into an IMF component by the EMD, and the IMF component signal frequency is single, so the WVD of the IMFs can restrain the frequency interference. Meanwhile, the WignerVille distribution makes the different mode composition of IMF separated, resulting the mode mixing restrained effectively. Finally, the method can be used to improve the effectiveness of vibration signal feature extraction.
Based on this, a fault feature extraction method based on EMDWVD is proposed in this paper to realize the nonlinear and nonGaussian signal analysis, the steps are as follows:
(1) The original signal is decomposed via the EMD to get a set of IMFs which contains the different frequencies.
(2) Selecting the appropriate IMFs, they will be analyzed by WVD to get the signal ${C}_{{c}_{i}}$, $i=$ 1, 2,….
(3) The ${C}_{{c}_{i}}$ will be beaded together to get the WignerVille distribution of signal, which is shown as ${C}_{x}=\sum _{i=1}^{n}{C}_{{c}_{i}}$.
During the signal processing, the WVD analysis was conducted for each of the IMFs, so in the final results the restraint for crossterm interference caused by the EMD method will still retain. Such the WignerVille distribution of the superposition original signal can effectively characterize the characteristics in time and frequency domain.
The IMFs which are analyzed by WVD are selected according to kurtosis in this paper. The kurtosis is the degree of a signal waveform peak parameters, which is defined as:
where $\sigma $ represents the standard deviation of signal $x$, $\mu $ represents the mean of signal $x$.
When the gearbox is in the normal state, the distribution of its monitored vibration signal can approximate to the normal one; at this moment, its kurtosis is approximately equal to 3. When the gearbox degenerates to a failure, the monitored vibration signal is no longer than the normal distribution because of an impact component of gearbox, which can lead to a bigger kurtosis. On the other hand, the degradation information of gearbox is often contained in the modulation signal of amplitude which with a more impact component. The vibration caused by the state degradation will cause the gearbox shaking at a different frequency, and each of the IMFs contains different natural vibration frequency caused by degradation. Therefore, the IMF is selected to get the fault feature of gearbox vibration signal according to the kurtosis criterion in this paper.
3. The remaining useful life prediction method for gearbox
3.1. Bayesian estimation for particle filter
Particle filter is a hybrid prediction method combined with datedriven and modeldriven two kinds of methods. The particle filter employs Monte Carlo simulation of a state dynamic model and Bayesian estimation for estimating the posterior probability density function (PDF) of the state of a degrading component at future times. Therefore, the method is an effective tool to predict the remaining life of nonlinear and nonGaussian systems.
The adaptive particle filtered Bayesian inference probability index is established for gearbox remaining life prediction in this paper. The state equation and measurement equation of particle filter are expressed as follows:
where ${x}_{k}$ is the hidden state to be estimated at time $k$, ${y}_{k}$ is the observation at time $k$, $f$ and $h$ represent the known process and observation functions respectively, $\left\{{n}_{k1},k\in N\right\}$ and $\left\{{\omega}_{k},k\in N\right\}$ are the dimensions of the state and process noise.
When the initial PDF of the state is $p\left({x}_{k1}{z}_{0:k1}\right)$, in order to obtain the posterior probability density function $p\left({x}_{k}{z}_{0:k}\right)$ of gearbox RUL, the state prediction is defined as:
$=\int p\left({x}_{k}{x}_{k1}\right)p\left({x}_{k1}{z}_{0:k1}\right)d{x}_{k1}.$
The new observations ${z}_{k}$ are collected at time $k$. According to the rule of Bayesian, the posterior probability density for system state is updating over time, so posterior probability distribution of the current state ${x}_{k}$ is obtained:
where the constant is $p\left({z}_{k}{z}_{0:k1}\right)=\int p\left({x}_{k}{z}_{0:k1}\right)p\left({z}_{k}{x}_{k}\right)d{x}_{k}$.
Above is the prediction process and update process for Bayesian estimation of particle filter. However, the gearbox vibration signal is nonlinear and nonGaussian, which makes it difficult to obtain the distribution. In this paper, we utilize Bayesian filtering and Monte Carlo algorithm to obtain the distribution.
Assuming that update monitoring information ${z}_{0:k}$ is known, the posterior probability density of system state ${x}_{0:k}$ can be expressed as follows:
where $\delta \left(\cdot \right)$ is the Dirac delta measure, $p\left({x}_{0:k}{z}_{0:k}\right)$ is the posterior probability function, and $p\left({\xi}_{0:k}{z}_{0:k}\right)$ is the priori probability function. If the real posterior probability function $p\left({x}_{0:k}{z}_{0:k}\right)$ can be obtained, then Eq. (9) can be calculated according to Eq. (10):
where ${x}_{0:k}^{i}$, $i=$ 1, 2,..., ${N}_{s}$ are independent random samples sampling from $p\left({x}_{0:k}{z}_{0:k}\right)$.
$p\left({x}_{0:k}{z}_{0:k}\right)$ are multitype and nonstandard, and it is difficult to get the results. Then importance function sampling method is proposed, which means the probability distribution and $p\left({x}_{0:k}{z}_{0:k}\right)$ is the same, and the PDF distribution $\pi \left({x}_{0:k}{z}_{0:k}\right)$ is known. Therefore, Eq. (10) can be transformed as follows:
where ${w}_{k}^{*i}=p\left({z}_{0:k}{x}_{0:k}^{i}\right)p\left({x}_{0:k}^{i}\right)/p\left({z}_{0:k}\right)\pi \left({x}_{0:k}^{i}{z}_{0:k}\right)$ is the weight of the system states, ${x}_{0:k}^{i}$, $i=$ 1, 2,..., ${N}_{s}$, which can be calculated by $\pi \left({x}_{0:k}^{i}{z}_{0:k}\right)$ sampling. $p\left({z}_{0:k}{x}_{0:k}^{i}\right)$ are likelihood value of observation sequence. In order to obtain the weight, we assume that $p\left({z}_{0:k}\right)=\int p\left({z}_{0:k}{x}_{0:k}\right)p\left({x}_{0:k}\right)d{x}_{0:k}\text{,}$ and the probability distribution $p\left({x}_{0:k}{z}_{0:k}\right)$ can be expressed as follows:
where:
The state distribution at time ${t}_{k}$ can be estimated according to the system state distribution $p\left({x}_{0:k1}{z}_{o:k1}\right)$. According to the theory of Bayesian filtering, $p\left({x}_{0:k}{z}_{o:k}\right)$ can be expressed as follows:
$=\frac{p\left({x}_{k}{x}_{0:k1},{z}_{0:k1}\right)p\left({x}_{0:k1}{z}_{o:k1}\right)p\left({z}_{k}{x}_{0:k},{z}_{o:k1}\right)}{p\left({z}_{k}{z}_{o:k1}\right)}$
$=\frac{p\left({x}_{k}{x}_{0:k1}\right)p\left({x}_{0:k1}{z}_{o:k1}\right)p\left({z}_{k}{x}_{0:k}\right)}{p\left({z}_{k}{z}_{o:k1}\right)}$
$=p\left({x}_{0:k1}{z}_{0:k1}\right)\frac{p\left({z}_{k}{x}_{o:k}\right)p\left({x}_{k}{x}_{0:k1}\right)}{p\left({z}_{k}{z}_{o:k1}\right)}.$
The implementation of the prediction algorithm based on PF is shown in Fig. 1.
Fig. 1. The flowchart for the implementation of PF
3.2. Simulated signal analysis compared with KF
In this section, we take the KF as an example and compare the state estimation performance with particle filter, which prove that PF is more effective under the condition of nonlinear and nonGaussian. A simulation test proposed with a discrete time adaptive noise of highorder nonlinear state space model, and the function is expressed as follows:
where $w$ and $v$ obey the standard normal distribution. Simulation repeats 50 times to get 50 real value and observation value of simulation state. The process is repeated twice, and the results are shown in Fig. 2.
Fig. 2. KF and PF simulation comparison
As shown in Fig. 2, KF state estimation results is bad, and has obvious deviation with the real data. On the other hand, PF has a higher precision, which indicates that PF has a better prediction performance for the nonlinear and nonGaussian state space.
4. Case studies and discussion
The gearbox is widely used in the transmission system of mechanical equipment, and the mechanical equipment fault is caused by gearbox failure mostly. Therefore, taking the gearbox as an example, the methods of fault feature extraction based on the EMDWVD and remaining useful life prediction based on PF proposed above are proven in this section.
4.1. The life cycle test of gearbox
The test object gearbox is shown in Fig. 3. The test bed is mainly composed of gearbox, speed and torque sensor (3056B4, produced by DYTRAN Company), magnetic powder electromotor (YCT1804A), a load (FZJ5) and other parts. Among them, the test bed is a steel platform (1500×600×700 mm) with a longitudinal $T$ slot; the output shaft axis of electromagnetic motor is parallel with the gearbox input shaft, where the power transfer is done through a coupling (6305). The gearbox output shaft is connected with the input shaft of the load through a coupling (6309). The vibration signal is acquired by the PXI1031 case that produced by NI Company and PXI4472B data acquisition card.
The most important part of the experiment is a secondary transmission gearbox, as shown in Fig. 4.
Fig. 3. Image of test bed. (1 – load, 2, 3 – speed and torque sensor, 4 – electromotor, 5 – test bed, 6 – gearbox system)
Fig. 4. Gearbox structure sketch map
The maximum capacity for lubricant is 2$L$, the rated speed is 1450 rpm, and rated transmission power is 0.75 kW. At particular parts of the gearbox there are four acceleration sensors numbered 14, the test load for gearbox is 22.5 times as much as the rated load. The signal is collected every five minutes with the sampling frequency for 20 kHz. The main parameters of the gearbox are shown in Table 1.
The life cycle test for gearbox is carried out; the vibration signal is acquired by four sensors. According to the monitoring vibration information, a failure of severe wear of the gear tooth surface occurs when the gearbox is working during about 400 hours, and the result is shown in Fig. 5. The incomplete data of wornteeth is collected, which means that the wear condition is monitored only at some run time, the wear data obtained is shown in Table 2.
Fig. 5. Comparison of gear after the experiment
a) Gear before experiment
b) Gear after experiment
Table 1. Prime parameters of gearbox
Name

First level transmission

Second level transmission

Transmission ratio


Highspeed shaft teeth number

Middle gear teeth number

Middle gear teeth number

Big gear teeth number


35

64

18

81

8.23

Table 2. Gear teeth wear at different inspection points
Time (h)

0

300

325

380

400

442

Gear teeth wear (mm)

0

0.05

0.08

0.1

0.19

0.3

4.2. Fault feature extraction for gearbox vibration signal
First of all, a vibration signal of gearbox is analyzed, as the original vibration signal of channel 1 at a different monitoring time is shown in Fig. 6.
Fig. 6. Original vibration signal of channel 1 at different monitoring time
a)$t=$ 1 h
b)$t=$ 100 h
c)$t=$ 200 h
d)$t=$ 300 h
e)$t=$ 400 h
f)$t=$ 442 h
However, the health status of gearbox can’t be directly identified by the time domain analysis. Thus, fault feature extraction of vibration signal based on EMDWVD is proposed in this paper.
The vibration signal of the gearbox shown in Fig. 6 at a different monitoring time is decomposed by the EMD, and the IMFs and a residual are obtained, as shown in Fig. 7. ${c}_{1}$${c}_{7}$ are the IMFs decomposed from the vibration signal of gearbox and ${c}_{8}$ is the residual.
The kurtosis of the IMFs ${c}_{1}$${c}_{8}$ obtained by EMD is shown in Table 3. ${c}_{1}$ and ${c}_{2}$, which have bigger kurtosis, are selected to be analyzed by the WVD. All the timefrequency distributions of different states are obtained, as shown in Fig. 8.
Fig. 7. Gearbox vibration signal EMD results at $t=$ 200 h
Fig. 8. EMDWVD results of channel 1 vibration signal at different condition time
a)$t=$ 1 h
b)$t=$ 100 h
c)$t=$ 200 h
d)$t=$ 300 h
e)$t=$ 400 h
f)$t=$ 442 h
It can be seen in Fig. 8 that the amplitude is increasing within 0.17 kHz to 0.37 kHz over time, so this frequency band is taken as a specific frequency band of gearbox, and the special frequency band energy collected from vibration signal is also taken as the health status value of gearbox.
In order to confirm the effect of feature extraction for vibration signal of gearbox, taking the vibration signal of the gearbox under the normal state without EMD as an example, the original signal is analyzed via WVD directly. For comparison, the result of WVD is given according to the contour, as shown in Fig. 9.
As compared with Fig. 9(b) and Fig. 9(a), it can be seen that the energy concentration of Fig. 9(b) whose distribution is got from the signal obtained by EMD is more obvious than Fig. 9(a) whose distribution is got from the original signal in the timefrequency domain. This is mainly because the original vibration signal is decomposed into IMFs with a single frequency by EMD, so the signal WVD obtained by EMD can restrain frequency interference and mode mixing effectively.
Table 3. IMF kurtosis of gearbox vibration signal at $t=$ 200 h
IMF

${c}_{1}$

${c}_{2}$

${c}_{3}$

${c}_{4}$

${c}_{5}$

${c}_{6}$

${c}_{7}$

${c}_{8}$

$K$

4.524

4.035

3.774

1.832

2.030

2.740

2.1158

2.016

Fig. 9. EMDWVD comparison results of gearbox vibration signal at $t=$ 200 h
a) WVD of the original signal
b) WVD of the signal obtained by EMD
Analyzing the result of Fig. 8, we can see that the amplitude of frequency band (0.71 kHz0.37 kHz) is increasing with the gearbox working longer, the specific frequency band energy of the vibration signals is extracted based on energy analysis using wavelet packet as the special frequency band of gearbox, and the special frequency band energy of vibration signal $\left\{{E}_{i},i=\text{1,}\text{}\text{2,...}\text{,}\text{}\text{442}\right\}$ is taking as the health status characteristic of gearbox, which is shown in Fig. 10.
Fig. 10. Specific band energy of gearbox vibration signal
As we can see in Fig. 10, the green curve represents the real degradation process of gearbox, which is the wear of the gear tooth. The red curve and blue curve represent the fault feature extracted based on the EMDWVD and fault feature extracted based on the WVD respectively. They all correspond with the degradation trend of gearbox, which demonstrates the accuracy of the methods. In addition, the fault feature based on the proposed method has a better stability compared with the fault feature based on the WVD, especially during the early failure and serious failure occurred. In the proposed method, some IMFs with larger kurtosis were selected, and the WVD was applied to these selected IMFs to attain the timefrequency distribution, so the noise, mode mixing and the crossterm interference were restrained. However, the WVD method can’t avoid the influence of these factors. Therefore, the proposed method has a better accuracy and effectiveness than the traditional method.
4.3. Gearbox remaining life prediction based on particle filter
The training samples for gearbox remaining useful life prediction based on particle filter are specific frequency band energy $E$ shown in Fig. 10 and gear teeth wear $wl$ shown in Table 2. To describe the degradation process of gearbox, the state space model based on Wiener process (WienerSSM) is proposed, and gearbox remaining useful life is obtained based on the degradation process model. The state space model for the gear teeth wear is as follows:
where ${x}_{1}$ is the gearbox worn state, ${x}_{2}$ is the unknown parameters related to the state which can’t be observable, $y\left(t\right)$is extracted specific frequency band energy, noise $w~N(0,Q)$ and $e~N(0,R)$is Gaussian distributed. Eq. (15) is used to describe the gearbox worn over time, parameters updating during the prediction process and the relationship between the gearbox worn and the fault feature, respectively. Take gearbox teeth wear at 442 hours as the failure threshold, namely $W{L}_{f}=\text{0.3}\text{}\text{mm}$. Assuming that the initial model parameters are as follows: $A=\text{1.00}$, $H=\text{1.00}$, $Q=\text{0}$, $R=\text{0}$, ${x}_{0}=\text{8000}$.
Taking the parameters estimation at 300 hours as an example, initial value ${x}_{0}=\text{8000}$ and historical data are both iterative based on PF to get the best estimation for the gearbox state $({x}_{300}^{s},{p}_{300}^{s})$. Afterwards, all parameters of the state space model are calculated at 300 hours, as shown in Fig. 11.
Fig. 11. The updated parameters at monitoring time 300 h
Model parameters are convergence to the optimum value quickly, which is the response of the gearbox SSM degradation trend, and will be used in the gearbox state feature prediction.
The parameters have randomness during the solving process, 1000 times repeated calculation is carried through to obtain the mean and standard deviation of parameters in this section, as shown in Table 4.
The probability density can be obtained based on the model parameters:
where $p\left({x}_{k}\right{x}_{k1}^{s})$ is the posterior probability under the conditions of ${x}_{k1}^{s}$ at time ${t}_{k}$, and ${w}_{k1}^{s}$ is the corresponding weight of particle ${x}_{k1}^{s}$. According to the Eq. (13), the estimated parameters are shown in Table 5.
The probability density function of RUL and mean remaining life can be obtained based on the model parameters at different time, as shown in Fig. 12. The mean remaining life of gearbox is closer to the real remaining life with the increase of monitoring data. The variances of the probability density function are smaller over time, with more condition monitoring information is used, which means that the prediction results are more and more accurate.
Table 4. Results of parameter estimation
Parameters

$A$

$H$

$Q$

$R$

Mean

0.2576

87.8

8913

2095

Standard deviation

0.0039

0.610

357.4

53.9

Table 5. Estimated parameters at different inspection time
Time (h)

$A$

$H$

$Q$

$R$

325

0.7400

187.8

12722

2483

350

0.5323

141.4

12316

2432

375

0.4653

113.3

9947

2420

400

0.2576

87.8

8913

2221

Fig. 12. Remaining useful life probability density function of gearbox
4.4. Performance evaluation
In order to evaluate the performance of remaining useful prediction based on PF, we take RMSE, MAE and VAE as indicators of absolute error, and take MARE and VRE as indicators of relative error. These indicators are shown as follows:
where ${x}_{t}$ is the real state, ${\widehat{x}}_{t}$ is the predicted state. The indicator is smaller means higher prediction precision.
The predicted results based on KF and PF are compared, as shown in Fig. 13, and the calculation results according to Eqs. (1721) is shown in Table 6. The main conclusions are as follows. The RUL prediction method based on PF has a better precision and accuracy than based on KF. The prediction method based on PF considers degradation feature extracted from vibration signal, gearbox teeth wear and nonlinear and nonGaussian system, which make it more suitable for gearbox remaining life prediction. However, the KP model is simple, and is easier to obtain the prediction results.
Fig. 13. Comparison of two RUL prediction methods
Table 6. Performance evaluation results of RUL prediction
The model

Performance evaluation


RMSE

MAE

VAE

MARE

VRE


KF

16.1169

35.5484

164.9626

0.4778

0.1024

PF

10.8169

20.2750

56.9469

0.2533

0.0205

5. Conclusions
This paper proposes the fault feature extraction and remaining useful life prediction methods for the gearbox under the condition of nonlinear and nonGaussian states. Firstly, the vibration signal of gearbox is decomposed into a set of IMFs by EMD. Then the timefrequency features can be obtained via the WignerVille distribution, to achieve the purpose of fault feature extraction for gearbox. Afterwards, the RUL is obtained based on PF considering degradation feature, gearbox teeth wear and nonlinear and nonGaussian system. The main findings are as follows.
1) The signal processing method based on EMDWVD can suppress the influence of mode mixing and crossterm interference, which has a better effect for the timefrequency analysis of the nonlinear and nonGaussian vibration signals. So, the proposed method can effectively extract the fault characteristics of the gearbox.
2) The remaining useful life prediction based on PF and state space model based on Wiener process considers degradation feature, gearbox teeth wear and nonlinear and nonGaussian system, which can make full use of all the monitoring information. Therefore, the method is suitable for gearbox remaining useful life prediction and can be used for stochastic degradation systems.
3) The life cycle test of gearbox demonstrates the accuracy and effectiveness of fault feature extraction method based on EMDWVD and remaining useful life prediction method based on PF in the engineering application.
Acknowledgements
The research is supported by the National Natural Science Foundation of China (No. 71401173) and the authors are grateful to all the reviewers and the editor for their valuable comments.
References
 Baydar N., Ball A. Case study detection of gear failures via vibration and acoustic signals using wavelet transform. Mechanical Systems and Signal Processing, Vol. 17, Issue 4, 2003, p. 787804. [Publisher]
 Xiuzhong Xu, Zhiyi Zhang, Hongxing Hua, et al. Timevarying modal parameter identification with timefrequency analysis methods. Journal of Shanghai Jiaotong University, Vol. 37, Issue 1, 2003, p. 122126. [Search CrossRef]
 Changhang Xu, Jifei Liu, Guoming Chen, et al. Application of EMD and WVD to feature extraction from vibration signal of reciprocating pump valves. Journal of China University of Petroleum, Vol. 34, Issue 3, 2010, p. 99103. [Search CrossRef]
 Zhengjia He, Jin Chen, Taiyong Wang, et al. Theories and Applications of Machinery Fault Diagnostics. Fourth Edition, Higher Education Press, Beijing, 2010. [Search CrossRef]
 Farook Sattar, Goran Salomonsson The use of a filterbank and the WignerVille distribution for timefrequency representation. IEEE Transactions Signal Process, Vol. 47, Issue 6, 1999, p. 17761783. [Publisher]
 Cong Wang, Meng Gan, Chang’an Zhu Nonnegative EMD manifold for feature extraction in machinery fault diagnosis. Measurement, Vol. 70, 2015, p. 188202. [Publisher]
 Xiaofeng Liu, Lin Bo, Honglin Luo Bearing faults diagnostics based on hybrid LSSVM and EMD method. Measurement, Vol. 59, 2015, p. 145166. [Publisher]
 Changhang Xu, Jifei Liu, Guoming Chen, et al. Application of EMD and WVD to feature extraction from vibration signal of reciprocating pump valves. Journal of China University of Petroleum, Vol. 34, Issue 3, 2010, p. 99103. [Search CrossRef]
 JianDa Wu, ChengKai Huang An engine fault diagnosis system using intake manifold pressure signal and WignerVille distribution technique. Expert Systems with Applications, Vol. 38, 2011, p. 536544. [Publisher]
 Ghofrani S., Mclernon D. C. AutoWignerVille distribution via nonadaptive and adaptive signal decomposition. Signal Processing, Vol. 89, 2009, p. 15401549. [Publisher]
 Chen G., Chen J., Dong G. M. Chirplet WignerVille distribution for timefrequency representation and its application. Mechanical Systems and Signal Processing, Vol. 41, 2013, p. 113. [Publisher]
 Ram Bilas Pachori, Anurag Nishad Crossterms reduction in the WignerVille distribution using tunableQ wavelet transform. Signal Processing, Vol. 120, 2016, p. 288304. [Publisher]
 Hai B. Huang, Ren X. Li, Xiao R. Huang, et al. Sound quality evaluation of vehicle suspension shock absorber rattling noise based on the WignerVille distribution. Applied Acoustics, Vol. 100, 2015, p. 1825. [Publisher]
 Wei Zeng, Haitao Wang, Guiyun Tian, et al. Application of laser ultrasound imaging technology in the frequency domain based on WignerWille algorithm for detecting defect. Optics and Laser Technology, Vol. 74, 2015, p. 7278. [Publisher]
 Narasimhana S. V., Malini Nayakb B. Improved WignerVille distribution performance by signal decomposition and modified group delay. Signal Processing, Vol. 83, 2003, p. 25232538. [Publisher]
 Jardine A. K. S., Lin D., Banjevic D. A review on machinery diagnostics and prognostics implementing conditionbased maintenance. Mechanical Systems and Signal Processing, Vol. 20, 2006, p. 14831510. [Publisher]
 Muhammad Farrukh Yaqub, Iqbal Gondal, Joarder Kamruzzaman Multistep support vector regression and optimally parameterized wavelet packettransform for machine residual life preciction. Journal of Vibration and Control, Vol. 19, Issue 7, 2012, p. 963974. [Search CrossRef]
 Arulampalam Ms., Maskell S., Gordon N., et al. A tutorial on particle filters for online nonlinear/nonGaussian Bayesian tracking. IEEE Transactions on Signal Processing, Vol. 50, 2002, p. 174188. [Publisher]
 Doucet A., Godsill S., Andreu C. On sequential Monte Carlo sampling methods for Bayesian filtering. Statistics and Computing, Vol. 10, 2000, p. 197208. [Publisher]
 Shane Butler, John Ringwood Particle filters for remaining useful life estimation of abatement equipment used in semiconductor manufacturing. Conference on Control and Fault Tolerant Systems Nice, France, 2010, p. 436441. [Search CrossRef]
 Katja Nummiaroa, Esther KollerMeierb, Luc Van Goola An adaptive colorbased particle filter. Image and Vision Computing, Vol. 21, 2003, p. 99110. [Publisher]
 Saha B., Goebel K., Poll S., et al. Prognostics methods for battery health monitoring using a Bayesian framework. IEEE Transactions on Instrumentation and Measurement, Vol. 58, Issue 2, 2009, p. 291296. [Publisher]
 Bin Fan, Lei Hu, Niaoqing Hu Remaining useful life prediction of rolling bearings by the particle filter method based on degradation Tate tracking. Journal of Vibroengineering, Vol. 17, Issue 2, 2015, p. 743756. [Search CrossRef]
 Orchard M. A Particle FilteringBased Framework for OnLine Fault Diagnosis and Failure Prognosis. Ph.D. Dissertation, Georgia Institute of Technology, 2007. [Search CrossRef]
 Chaochao Chen, Bin Zhang, George Vachtsevanos, et al. Machine condition prediction based on adaptive neurofuzzy and highorder particle filtering. IEEE Transactions on Industrial Electronics, Vol. 58, Issue 9, 2011, p. 43534364. [Publisher]
Cited By
Structural Health Monitoring
Yubin Pan, Rongjing Hong, Jie Chen, Jianshe Feng, Weiwei Wu

2020

Journal of Vibration and Control
Hongchao Wang, Deli Sun

2019

ISA Transactions
Bo Xu, Fengxing Zhou, Huipeng Li, Baokang Yan, Yi Liu

2019

Ekspolatacja i Niezawodnosc  Maintenance and Reliability
Xueyi Li, Jialin Li, David He, Yongzhi Qu

2019

Jiangting LIU, Kaifeng ZHOU, Yue HU 
2018

Circuits, Systems, and Signal Processing
Rishi Raj Sharma, Ram Bilas Pachori

2018
