Ruonan Liu

Fault Diagnosis for a Wind Turbine Generator Bearing via Sparse Representation and Shift-Invariant K-SVD

It is always a primary challenge in fault diagnosis of a wind turbine generator to extract fault character information under strong noise and nonstationary condition. As a novel signal processing method, sparse representation shows excellent performance in time–frequency analysis and feature extraction. However, its result is directly influenced by dictionary, whose atoms should be as similar with signals inner structure as possible. Due to the variability of operation environment and physical structure in industrial systems, the patterns of impulse signals are changing over time, which makes creating a proper dictionary even harder. To solve the problem, a novel data-driven fault diagnosis method based on sparse representation and shift-invariant dictionary learning is proposed. The impulse signals at different locations with the same characteristic can be represented by only one atom through shift operation. Then, the shift-invariant dictionary is generated by taking all the possible shifts of a few short atoms and, consequently, is more applicable to represent long signals that in the same pattern appear periodically. Based on the learnt shift-invariant dictionary, the coefficients obtained can be sparser, with the extracted impulse signal being closer to the real signal. Finally, the time–frequency representation of the impulse component is obtained with consideration of both the Wigner–Ville distribution of every atom and the corresponding sparse coefficient. The excellent performance of different fault diagnoses in a fault simulator and a wind turbine proves the effectiveness and robustness of the proposed method. Meanwhile, the comparison with the state-of-the-art method is illustrated, which highlights the superiority of the proposed method.

Dislocated Time Series Convolutional Neural Architecture: An Intelligent Fault Diagnosis Approach for Electric Machine

In most current intelligent diagnosis methods, fault classifiers of electric machine are built based on complex handcrafted features extractor from raw signals, which depend on prior knowledge and is difficult to implement intelligentization authentically. In addition, the increasingly complicated industrial structures and data make handcrafted features extractors less suited. Convolutional neural network (CNN) provides an efficient method to act on raw signals directly by weight sharing and local connections without feature extractors. However, effective as CNN works on image recognition, it does not work well in industrial applications due to the differences between image and industrial signals. Inspired by the idea of CNN, we develop a novel diagnosis framework based on the characteristics of industrial vibration signals, which is called dislocated time series CNN (DTS-CNN). The DTS-CNN architecture is composed of dislocate layer, convolutional layer, sub-sampling layer and fully connected layer. By adding a dislocate layer, this model can extract the relationship between signals with different intervals in periodic mechanical signals, thereby overcome the weaknesses of traditional CNNs and is more applicable for modern electric machines, especially under nonstationary conditions. Experiments under constant and nonstationary conditions are performed on a machine fault simulator to validate the proposed framework. The results and comparison with respect to the state of the art in the field is illustrated in detail, which highlights the superiority of the proposed method in industrial applications.

Data-driven discriminative K-SVD for bearing fault diagnosis

Rolling element bearing is an important component. As it is usually used in a complex environment, there are many failures occur on them. How to find the fault has become a pressing problem to be solved. The vibration signals generated by bearings are usually containing a variety of noise. The general diagnosis is divided into two stages: feature extraction and classification. Unlike conventional methods, there is no need to have a specific fault feature extraction step for sparse representation method. One of the dictionary learning methods which called the K-SVD is an algorithm does not need a defined dictionary but whose output is an over-complete dictionary studied by signals. The method that iteratively updating the K-SVD-trained dictionary based on the outcome of a linear classifier usually leads to the local minima. In order to train a dictionary that works well both in representation and classification, we use the Discriminative K-SVD which the labels are directly embedded in the dictionary learning step. Discriminant K-SVD can find all parameters of global optimum at the same time. The complexity of Discriminative K-SVD is related to that of K-SVD. Finally, the proposed method is applied in the practical bearing experiments, the results not only confirmed the accuracy of the proposed method for finding the fault types of bearings but also identified the damage degrees of bearings.

Sparse components separation-based operational reliability assessment approach

As a metric to quantize the engineering system and plants quality, reliability has developed as a scientific discipline which is mainly rely on statistical analysis and life tests. However, with the improvement of mechanical system quality and service time, access to life tests and historical failure data become more and more difficult and time-consuming. To overcome the dependence of statistical failure data, a novel operational reliability assessment approach is proposed. System vibration response varies from operational states to states. In a bearing-rotor system, the vibration response of failure system is the impulse component. Besides, the vibration caused by abrasion is the harmonic component. For impulse and harmonic components extraction, a morphological component analysis (MCA) method based on basis pursuit denoising (BPDN) is used to decompose the vibration signals and reconstruct the impulse signals. Then classical time domain indexes of impulse signals are used as the observation sequence of a corresponding Hidden Markov Models (HMM) to assessment operational reliability. Finally, BPDN, traditional time-features and the proposed method are respectively applied in the operational reliability assessment of an experiment carried out on an aerospace bearing test rig. Comparison results confirmed the effectiveness of the proposed method for operational reliability assessment in bearing-rotor system.

Fault diagnosis of wind turbine using local mean decomposition and synchrosqueezing transforms

Wind turbine is driven by natural wind, thus the vibration signal of gearbox is influenced and performs with nature of non-stationary and noisy. In order to analyze the variable speed signal, time-frequency representation is introduced as a powerful tool. The synchrosqueezing transform is a useful time-frequency representation method in many situation but can not restrain noise in high noise background. And local mean decomposition can decompose noise from the signal and remain the character of signals. In this paper, we proposed a new method for fault diagnosis of wind turbine with help of synchrosqueezing transform and local mean decomposition, which can denoise and represent the frequency of signals changing with time. This method is validated by simulated signals and experimental field test. The results showed the proposed method can restrain the noise and decrease the inaccuracy of time-frequency representation.

Sparse representation based on redundant dictionary and basis pursuit denoising for wind turbine gearbox fault diagnosis

Due to the dramatic growth of total installation and individual capacity make the failures of wind turbines costly or even unacceptable. Therefore, wind turbine fault diagnosis, which is considered as a useful tool to ensuring the safety of wind turbines and reducing costly system maintenances, is attracting increasing attention. In this paper, a novel fault diagnosis for wind turbine gearbox based on basis pursuit denoising (BPDN) and the union of redundant dictionary is proposed. The union of redundant dictionary is constructed based on the underlying prior information of vibrations signal with multicomponent coupling effect. Within the frame work of BPDN, sparse coefficient and corresponding time-frequency atoms can be obtained. By time-frequency representation of the reconstructed signal, fault information can be displayed. Finally, an engineering application of a wind turbine gearbox is used to verify the effectiveness of the proposed method.