Showing papers by "Yaguo Lei published in 2017"

Remaining Useful Life Prediction Based on a General Expression of Stochastic Process Models

Abstract: In remaining useful life (RUL) prediction, stochastic process models are widely used to describe the degradation processes of systems. For age-dependent stochastic process models, the RUL probability density function (PDF) can be calculated using a closed-form solution. For state-dependent models, however, it is difficult to calculate such a closed-form solution. Therefore, the RUL is always approximately estimated using a sequential Monte Carlo-based method, but this method has some limitations. First, it only provides a numerical approximation result whose accuracy highly relies on the quality and quantity of the simulated degradation trajectories. Second, the time interval is unable to be adjusted during the state transition process, resulting in too few discrete probability densities in the result near the end-of-life. This paper describes the degradation processes using a general expression of age- and state-dependent models. The analytical solution of the RUL PDF is derived from the general expression. After that, a new RUL prediction method is proposed. In this method, a series of degradation trajectories are generated through degradation process simulation. The RUL PDF is estimated by inputting the state values of the degradation trajectories into the analytical solution. The validity of the proposed method is verified using fatigue-crack-growth data.

A fault diagnosis method of rolling element bearings based on CEEMDAN

Abstract: Ensemble empirical mode decomposition (EEMD) represents a valuable aid in empirical mode decomposition (EMD) and has been widely used in fault diagnosis of rolling element bearings. However, the intrinsic mode functions (IMFs) generated by EEMD often contain residual noise. In addition, adding different white Gaussian noise to the signal to be analyzed probably produces a different number of IMFs, and different number of IMFs makes difficult the averaging. To alleviate these two drawbacks, complete ensemble empirical mode decomposition with adaptive noise (CEEMDAN) was previously presented. Utilizing the advantages of CEEMDAN in extracting weak characteristics from noisy signals, a new fault diagnosis method of rolling element bearings based on CEEMDAN is proposed. With this method, a particular noise is added at each stage and after each IMF extraction, a unique residue is computed. In this way, this method solves the problem of the final averaging and obtains IMFs with less noise. A simulated signal is ...

Remaining useful life prediction

Abstract: This chapter focuses on the remaining useful life (RUL) prediction of rotating machinery. It first gives an explicit description to the concept of RUL prediction and illustrates the major processes of the RUL prediction. Then, two categories of RUL prediction methods, that is, data-driven methods and model-based methods are discussed in details, respectively. In the section of data-driven methods, the commonly strategy of this kind of methods is explained. And a relevance vector machine based prediction method is described as an example. In the section of model-based methods, the classification of the models is first introduced. Then, three prediction methods are described in details, which are based on the exponential model, polynomial model, and Paris–Erdogan model, respectively. By the end of the chapter, the prospects of the RUL prediction for rotating machinery are provided.

Deep convolution feature learning for health indicator construction of bearings

Abstract: In the field of data-driven prognostics of bearings, considerable research effort has been taken to construct an effective health indicator. However, existing health indicator construction methods are mainly based on manual feature extraction and feature fusion techniques. Such manual techniques are generally designed for specific tasks and need the help of experts' prior knowledge, resulting in labor-consuming and time-costing. So it is desirable to automatically construct health indicators. To deal with this problem, this paper presents a deep convolution feature learning based method to construct health indicators of bearings. The proposed method first learns features from the raw vibration signals through several convolution and pooling operations. Then the learned features are mapped to the health indicator through a nonlinear transformation. At last, the proposed method is validated by a bearing dataset. The results demonstrate that the proposed method is able to effectively construct the health indicator directly from the raw vibration signals, which is superior to that based on self organizing map. Additionally, because the proposed health indicator is constructed automatically, it significantly reduces the need of experts' prior knowledge and labor resources.

An improved fusion prognostics method for remaining useful life prediction of bearings

Abstract: The remaining useful life (RUL) prediction of bearings has emerged as a critical technique for providing failure warnings in advance, reducing costly unscheduled maintenance and enhancing the reliability of bearings. Recently, a fusion prognostics method combining exponential model and relevance vector machine (RVM) has been proposed and applied to the RUL prediction of bearings. This fusion prognostics method integrates the advantages of RVM and exponential model and so has better prediction performance than other exponential model-based methods. However, selecting the appropriate value of kernel parameter is very difficult for this fusion prognostics method because of the lack of an explicit prior knowledge. which reduces the prediction accuracy of the fusion prognostics method and affects its generalization performance. To solve this problem, an improved fusion prognostics method is proposed in this paper. In the improved fusion prognostics method, RVM regressions with different kernel parameter values are first applied to obtaining different sparse datasets. Then, using the exponential model of bearing degradation, the different degradation curves are got by fitting the obtained sparse datasets and the Frechet distance is employed to select the optimum degradation curve from those fitted curves. Finally, the RUL is predicted by extrapolating the selected degradation curve to reach the failure threshold. To verify the superiority of the proposed method compared with the original fusion prognostics method, a real bearing degradation data is used for the RUL prediction. The results show that the improved fusion prognostics method outperforms the original fusion prognostics method in the RUL prediction of bearings.

Incipient Fault Detection for Rolling Element Bearings under Varying Speed Conditions

Abstract: Varying speed conditions bring a huge challenge to incipient fault detection of rolling element bearings because both the change of speed and faults could lead to the amplitude fluctuation of vibration signals. Effective detection methods need to be developed to eliminate the influence of speed variation. This paper proposes an incipient fault detection method for bearings under varying speed conditions. Firstly, relative residual (RR) features are extracted, which are insensitive to the varying speed conditions and are able to reflect the degradation trend of bearings. Then, a health indicator named selected negative log-likelihood probability (SNLLP) is constructed to fuse a feature set including RR features and non-dimensional features. Finally, based on the constructed SNLLP health indicator, a novel alarm trigger mechanism is designed to detect the incipient fault. The proposed method is demonstrated using vibration signals from bearing tests and industrial wind turbines. The results verify the effectiveness of the proposed method for incipient fault detection of rolling element bearings under varying speed conditions.

Fault Diagnosis of Rotating Machinery Based on Empirical Mode Decomposition

Abstract: Rotating machinery covers a broad range of mechanical equipment in industrial applications. It generally operates under tough working environment and is therefore subject to faults easily. Vibration signals collected in the working process have valuable contributions for the presentation of conditions of the rotating machinery. Consequently, using signal processing techniques, these faults could be detected and diagnosed. Empirical mode decomposition (EMD) is one of the most powerful signal processing techniques and has been widely applied in fault diagnosis of rotating machinery. This chapter attempts to introduce the recent research and development of EMD in fault diagnosis of rotating machinery, including basic concepts and fundamental theories about EMD methods and improved EMD methods. Moreover, the applications of EMD methods and improved EMD methods in fault diagnosis of common and key components of rotating machinery, like rotors, gears and rolling element bearings, are described in details.

Clustering algorithm–based fault diagnosis

Abstract: Clustering algorithms can automatically recognize the pattern inside the data so as to analyze the collected data without their labels. Using this advantage, three clustering-based fault diagnosis methods are presented to deal with some diagnosis cases of rotating machinery in which the labeled data are limited. In the first method, compensation distance evaluation technique and the weight K nearest neighbor are used to recognize the mechanical faults, harnessing the merits that the computation of feature weights is simpler and the weights are easier to understand. The second method is presented based on weight fuzzy c -means, which is robust to the local structure of the data and reflects the level of uncertainty over the most appropriate assignment. Finally, a Hybrid clustering algorithm–based fault diagnosis method is introduced, considering the problems like the sample influence for clustering and the automatic setting of the cluster number. The results of the diagnosis cases verify that these diagnosis methods take full advantage of unlabeled data and reduce the human labor in fault diagnosis.

Intelligent fault diagnosis of rotating machinery using locally connected restricted boltzmann machine in big data era

Abstract: In intelligent fault diagnosis, unsupervised feature learning is a potential tool to replace the manual feature extraction in big data era. Therefore, we first develop a locally connected restricted Boltzmann machine (LCRBM) from the traditional RBM in order to handle the periodic appearance of fault characteristics in the raw signals of rotating machinery. Then, using LCRBM, we propose a method for intelligent fault diagnosis of rotating machinery. In the method, LCRBM is used to obtain features directly from raw signals. Based on the features learned by LCRBM, the method uses softmax regression to recognize faults. The proposed method is verified by the dataset of locomotive bearings and its superiority is demonstrated by the comparison with methods using the traditional RBM and eighteen widely used manual features. Results indicate that the proposed method is able to automatically learn fine features from raw signals of rotating machinery and achieves higher diagnosis accuracies.

A dirty data recognition method for machinery condition monitoring in big data era

Abstract: Condition monitoring of machinery has entered the big data era, while the existence of dirty data reduces the quality of the whole data. In order to recognize the dirty data included in machinery monitoring data, a new method is proposed in this paper. First, a feature named sampled power index (SPI) is designed to transform the dirty data recognition issue into the outlier recognition. Then the windowing technique, the difference operation and the logarithm transform are introduced to reduce the feature tendency and the feature volatility. Next, auto regression-generalized autoregressive conditional heteroskedasticity (AR-GARCH) model is applied to regress the feature series and produce the crippled local means and local volatilities. Finally, the features are normalized and the 3σ criterion is applied to recognize the dirty data. The performance and the feasibility of this proposed method are evaluated by a simulation and an experiment. The results validate the effectiveness of the proposed method.

Weak signal detection based on underdamped multistable stochastic resonance

Abstract: Traditional overdamped stochastic resonance (SR) methods are difficult to match with complicated and variable input signals due to single stable-state types. Moreover, their performance depends on the parameter selection of highpass filters. To further explore the potential of SR, this paper studies the behavior of underdamped SR in a multistable nonlinear system by analyzing its output frequency responses, and presents a promising underdamped multistable SR method for weak signal detection and further incipient fault diagnosis of machinery. Numerical analyses indicate that the proposed method is supposed to possess two advantages: 1) the stable-state diversity of the multistable potential makes it easily match with input signals and 2) underdamped multistable SR is equivalent to a bandpass filter as the rescaling ratio varies, which is able to suppress the interference from multiscale noise. Simulated and experimental data of rolling element bearings demonstrate the effectiveness of the proposed method. For comparison, ensemble empirical mode decomposition (EEMD) method and traditional overdamped bistable SR method are also employed to process the data. The comparison results show that the proposed method can effectively detect incipient fault characteristics and perform better than traditional SR and EEMD methods.

Machine remaining useful life prediction considering unit-to-unit variability

Abstract: Remaining useful life (RUL) prediction of machinery plays a significant role for predictive maintenance, thus attracting more and more attentions in recent years. Stochastic process model-based methods are widely used in the RUL prediction of machinery. One of the major issues in the stochastic process model-based methods is that how to deal with the unit-to-unit variability during the RUL prediction process. Traditional methods generally handle this issue by introducing a unit-to-unit variability parameter into the model expression and estimate the parameter using the maximum likelihood estimation (MLE) algorithm. There exist two major limitations in the traditional methods. 1) The degradation processes are assumed to be dependent on only the age, which restricts their implementation in the cases of the state-dependent degradation processes. 2) They do not discuss the influence of the unit-to-unit variability in the RUL prediction processes systematically. To deal with these two limitations, a new RUL prediction method based on age- and state-dependent stochastic process models is proposed in this paper. In the proposed method, a generalized expression of the age- and stage-dependent stochastic process models is generated. An enhanced MLE algorithm is developed to estimate the model parameters according to the measurements of the available training units. And the unit-to-unit variability parameter is updated according to the real-time measurements of the testing unit. The effectiveness of the proposed method is demonstrated using a numerical simulation dataset of fatigue crack-growth.