A prediction method for the real-time remaining useful life of wind turbine bearings based on the Wiener process

Highlights

• Degradation speed of bearings changes with time and uncertain external factors.

• Physics-based models of RUL prediction for bearings are built difficultly.

• Degradation model of bearings is built with Wiener process using data-driven method.

• Proposed method acquires effective predicted RUL only based on temperature data.

Abstract

A performance degradation model and a real-time remaining useful life (RUL) prediction method are proposed on the basis of temperature characteristic parameters to determine the RUL of wind turbine bearings. First, using the moving average method, the relative temperature data of wind turbine bearings are smoothed, and the temperature trend data are obtained on the basis of the uncertainty of wind speed and wind direction that causes the temperature of wind turbine bearings to vary widely. Second, given that the degradation speed of bearings changes with operational time and uncertain external factors, the performance degradation model is established with the Wiener process. The parameters of this model are obtained through the maximum likelihood estimation method. Third, according to the failure principle of the first temperature monitoring value beyond the first warning threshold, the RUL prediction model for wind turbine bearings is established on the basis of an inverse Gaussian distribution. Finally, the performance degradation process and real-time RUL prediction are demonstrated by predicting the RUL of a practical rear bearing of a wind turbine generator. The comparison of the predicted RUL and actual RUL shows that the proposed model and prediction method are correct and effective.

Introduction

At present, the premature failure rates, restoration times, and operation and maintenance (O&M) costs of wind power are higher than desirable [1]. The operational unavailability of wind turbines reaches 3% of wind turbine lifetime. The O&M costs can account for 10%–20% of the total energy cost for a wind turbine project, and this percentage can reach 35% for a wind turbine at the end of its lifetime [2]. Wind turbines are unmanned remote power plants. Unlike conventional power stations, wind turbines are exposed to highly variable and harsh weather conditions, including calm to severe winds, tropical heat, lightning, Arctic cold, hail, and snow. Wind turbines undergo constantly changing loads because of these external variations, thereby resulting in gradual changes in the performance of their critical components. In particular, wind turbine bearings in drive train are under high stresses, including thermal and mechanical stresses. In general, the rolling element bearings are more vulnerable than other mechanical components in wind turbine drivetrains. Bearing failure is the most common failure mode associated with a long-term downtime [3]. Therefore, monitoring the degradation process, evaluating the state of health, and predicting the remaining useful life (RUL) of wind turbine bearings have become increasingly important in the development of maintenance strategies [4,5].

Prediction and health management (PHM) has emerged as one of the key ways to improve system safety, increase system operations reliability and mission availability, decrease unnecessary maintenance actions, and reduce system life costs [6]. As a very important step of PHM, RUL prediction based on condition monitoring (CM) information plays an important role in maintenance strategy selection, inspection optimization, and spare part provision [2]. The RUL of wind turbine bearings is defined as the length of time from the present time to the end time of useful life. Most bearing manufacturers and researchers today predict the RUL of wind turbine bearings through their field experience and research results [7,8] and not through available prediction methods for wind turbine bearings. A few studies have paid close attention to the RUL prediction methods for wind turbine bearings. In the literature, using the finite element analysis method with load calculation or the national standard RUL formula of bearing with compensation factors [9,10], the RUL of wind turbine bearings is determined. However, the aforementioned RUL prediction methods can only be used to obtain theoretical RUL as they do not consider the degradation speed of wind turbine bearings over time. Therefore, applying these methods to PHM systems in real time to improve wind turbine maintenance strategies may be difficult.

The existing RUL prediction methods can be roughly classified into the following two types: physics-based and data-driven [11,12]. The physics-based methods describe the growth trend of a failure mode quantitatively using the physical laws, predicting the RUL of a component by solving one or a number of deterministic equations derived from empirical data [13]. However, it is often difficult to build accurate physical models for RUL prediction in the practical applications, especially when the failure mechanism is complicated. Compared with the physics-based methods, the data-driven RUL prediction method in PHM has become common with the development of signal processing technology. Owing to the rapid of development of sensor and instrumentation technologies, the condition monitoring data which are highly correlated with the health condition and lifetime of an engineering system can be measured and collected easily [14].

The performance degradation of wind turbine bearings is typically caused by misalignments or other adverse factors in the drive train that cause abnormal loading and accelerate bearing wear. As a result of their construction, wind turbine bearings generate identifiable signatures on temperature and vibration with characteristic frequencies. These signatures present an effective route for the monitoring of progressive bearing degradation. Unlike vibration, temperature can be used to indicate the deterioration of wind turbine bearings because of its thermal inertia and strong anti-interference capacity. A large amount of temperature data available for research can be easily acquired with the supervisory control and data acquisition (SCADA) system [15,16]. Therefore, effectively determining the RUL of wind turbine bearings by analyzing temperature characteristic parameters is feasible.

The performance degradation trend of wind turbine bearings increases with operational time, whereas the temperature trend of these bearings increases with degradation trend. However, changes in temperature monitoring value reflect a wide range of performance degradation trends with varying rotor speeds because of the uncertain nature of the fluctuation and intermittence of wind. Moreover, temperature trend data fluctuate increasingly. Therefore, a stochastic method is more suitable than an intelligent learning method in predicting the RUL of wind turbine bearings without the need for a time-consuming training process. The stochastic models such as Markov chain, Gamma process, and Wiener process are frequently used to characterize the evolution of degradation process [17]. Due to its important physical interpretation and nice mathematical properties, the Wiener process model becomes one of the most popular stochastic models [18]. The Wiener process model is also called a Brownian motion model with linearity, which characterizes a non-monotonic degradation process, and has been applied to model the degradation process and to predict the RUL of a variety of industrial components, such as self-regulating heating cables [19], LED lamps [20], batteries [21], and laser generator [22]. In Ref. [20], Tseng et al. used a Wiener process to determine the lifetime for the light intensity of LED lamps of contact image scanners. In Ref. [21], S. J. Tang et al. proposed a novel RUL prediction method for lithium-ion batteries based on the Wiener process with measurement error. In Ref. [23], S. Mishra et al. proposed a new approach for predicting the RUL of a structure from lamb wave sensor data using principal component regression and Wiener process degradation modeling. As an extension, according to the bidirectional characteristic of Brownian motion, the Wiener process can be used to describe the performance degradation of wind turbine bearings. The performance degradation model of wind turbine bearings based on the Wiener process has an advantage in mathematics [22], as it determines the RUL through the inverse probability distribution function.

The contribution of this paper is to propose a real-time RUL prediction method for wind turbine bearings. This idea is unlike the existing finite element analysis method with load calculation or the national standard RUL formula of bearing with compensation factors, and the proposed method can be used to predict real-time RUL of wind turbine bearings just based on temperature. In the present study, the operational conditions of variable speed and the uncertainty of wind turbines are considered. First, we use the moving average method to smooth the relative temperature data of wind turbine bearings and to obtain the temperature trend data. Second, with consideration of the changes in the degradation speed of bearings with operational time and uncertain harmful factors, we establish a performance degradation model on the basis of the Wiener process. The maximum likelihood estimation method is also used to obtain the degradation parameters. Third, a real-time RUL prediction model is established according to the failure principle of the first temperature monitoring value beyond the first warning threshold (FWT). Finally, the validity of the proposed method is verified on the basis of the temperature data of an actual 1.5 MW wind turbine through a case of a generator rear bearing.