Wenlong Zhu

Fault diagnosis based on dependent feature vector and probability neural network for rolling element bearings

Rolling element bearings (REB) are crucial mechanical parts of most rotary machineries, and REB failures often cause terrible accidents and serious economic losses. Therefore, REB fault diagnosis is very important for ensuring the safe operation of rotary machineries. In previous researches on REB fault diagnosis, achieving the accurate description of faults has always been a difficult problem, which seriously restricts the reliability and accuracy of the diagnosis results. In order to improve the precision of fault description and provide strong basis for fault diagnosis, dependent feature vector (DFV) is proposed to denote the fault symptom attributes of the six REB faults in this paper, and this is a self-adaptive fault representation method which describes each fault sample according to its own characteristics. Because of its unique feature selection technique and particular structural property, DFV is excellent in fault description, and could lay a good foundation for fault diagnosis. The advantages of DFV are theoretically proved via the Euclidean distance evaluation technique. Finally, a fault diagnosis method combining DFV and probability neural network (PNN) is proposed and applied to 708 REB fault samples. The experimental results indicate that the proposed method can achieve an efficient accuracy in REB fault diagnosis.

Multifault Diagnosis for Rolling Element Bearings Based on Intrinsic Mode Permutation Entropy and Ensemble Optimal Extreme Learning Machine

This paper presented a novel procedure based on the ensemble empirical mode decomposition and extreme learning machine. Firstly, EEMD was utilized to decompose the vibration signals into a number of IMFs adaptively and the permutation entropy of each IMF was calculated to generate the fault feature matrix. Secondly, a new extreme learning machine was proposed by combining ensemble extreme learning machine and the evolutionary extreme learning machine which used an artificial bee colony algorithm to optimize the input weights and hidden bias. The proposed diagnosis algorithm was applied on the three rolling bearing fault diagnosis experiments. The numerical experimental results demonstrated that the proposed method had an improved generalization performance than traditional extreme and other variants.

Design of a multi-mode intelligent model predictive control strategy for hydroelectric generating unit

This paper proposes a nonlinear multi-mode intelligent model predictive control (MPC) strategy for hydroelectric generating unit (HGU). In this multi-mode MPC scheme, excitation MPC mode and integrated MPC mode work for excitation control process and load scheduling condition, respectively. Every control mode is built on the basis of a tree-seed algorithm based model predictive control (TSA-MPC) scheme, which introduces a newly proposed tree-seed algorithm (TSA) and the stability-guaranteed measures into rolling optimization mechanism of nonlinear MPC (NMPC) to replace the existing complex numerical differential geometric solutions. Simulation experiments of the proposed multi-mode MPC and the comparative methods are undertaken under diverse operating conditions in a HGU control system as case studies. Experimental results indicate the superiority in voltage regulation and damping performance as well as the effectiveness of the comprehensive control of turbine governing and generator excitation.

A state tendency measurement for a hydro-turbine generating unit based on aggregated EEMD and SVR

The reliable measurement of state tendency for a hydro-turbine generating unit (HGU) is significant in guaranteeing the security of the unit and promoting stability of the power system. For this purpose, an aggregated ensemble empirical mode decomposition (AEEMD) and optimized support vector regression (SVR)-based hybrid model is developed in this paper in order to enhance the measuring accuracy of state tendency for a HGU. First of all, the non-stationary time series of the state signal are decomposed into a collection of intrinsic mode functions (IMFs) by EEMD. Subsequently, to obtain the refactored intrinsic mode functions (RIMFs), the IMFs with different scales are aggregated with the proposed reconstruction strategy in consideration of the frequency and energy conditions. Later, the phase–space matrix in accordance with each RIMF is deduced by phase–space reconstruction and all the RIMFs are predicted through establishing homologous optimal SVR forecasting models with a grid search. Finally, the ultimate measuring values of state tendency can be determined through the accumulation of all the RIMF forecasting values. Furthermore, the effectiveness of the proposed method is validated in engineering experiments and comparative analyses.

Design of integrated synergetic controller for the excitation and governing system of hydraulic generator unit

Abstract Synergetic control theory is introduced into hydraulic generator excitation system (HGES) and hydraulic generator governing system (HGRS) in this paper. Synergetic excitation controller (SEC), synergetic governing controller (SGC) of HGU have been designed. In order to enhance the terminal voltage control and mechanical power tracking performances simultaneously, the integrated synergetic controller (ISC) is also proposed. ISC implements synergetic control of terminal voltage, rotor speed, mechanical input power and guide vane opening. Namely, the ISC is considering both of the excitation system and governing system of hydraulic generator unit (HGU), which can provide control function instead of SEC and SGC. In addition, the control rules of the aforementioned three controllers are deduced from the nonlinear mathematical analytic model of hydroelectric generator unit. At the end of this paper, comparative case studies between the proposed SGC, SEC, ISC and classic PID controller are presented. The results show that the proposed ISC improves the nonlinear HGU system performance with a more accurate precision and shorter settling time in different operating conditions.

Multi-fault classification based on the two-stage evolutionary extreme learning machine and improved artificial bee colony algorithm

Extreme Learning Machine (ELM) is a novel single-hidden-layer feed forward neural network with fast learning speed and better generalization performance compared with the traditional gradient-based learning algorithms. However, ELM has two issues: the hidden node number of ELM needs to be predefined and the random determination of the input weights and hidden biases lead to ill-condition problem. In this paper, a two-stage evolutionary extreme learning machine (TSE-ELM) algorithm was proposed to overcome the drawbacks of original ELM, which used an improved artificial bee colony (ABC) algorithm to optimize the input weights and hidden biases. The proposed TSE-ELM algorithm was applied on the UCI benchmark datasets and rolling bearing fault diagnosis. The numerical experimental results demonstrated that TSE-ELM had an improved generalization performance than traditional ELM and other evolutionary ELMs.

Residual diagnosis model based on wavelet neutral network and its application to hydroelectric generator unit

Most of current diagnosis methods of hydroelectric generator unit (HGU) performed not well when lacking domain expert knowledge, in order to address this problem, we propose a novel residual diagnosis model based on wavelet neural network (RDM-WNN) and weighed fuzzy set theory for quantitative diagnosis of HGU in this paper. First, the main working condition parameters (MWCP) are extracted according to the mutual information between the performance parameters and working condition parameters, and used as input feature vector to construct the RDM-WNN model. Second, relative residual are calculated by comparing the output vector of RDM-WNN model to the corresponding real values. Third, the relative residual values are used to implement quantitative diagnosis of HGU using weighted fuzzy set theory. The proposed method was verified on a real HGU with 100 normal working conditions, 200 slight faults working conditions, and 200 fully faults working conditions. Six groups of partial load experiments were implemented. The results demonstrate that the proposed method is an effective means for fault diagnosis of HGU.

An improved ensemble empirical mode decomposition method and its application to pressure pulsation analysis of hydroelectric generator unit

The noise-assisted method of ensemble empirical mode decomposition represents a significant improvement over the original empirical mode decomposition for eliminating the mode mixing problem. However, the ensemble empirical mode decomposition method will generate some additional problems, including the contamination of the residue noise in the signal reconstruction and the high computational cost. In this work, an improved ensemble empirical mode decomposition method, combining the complementary ensemble empirical mode decomposition and a time-saving ensemble empirical mode decomposition method by over-sampling the investigated signal, is proposed to solve these problems. By using the proposed method, the residue of the added white noise in the signal reconstruction can be eliminated completely by adding white noises in pairs with positive and negative signs to the targeted signal, and the computational cost can be saved drastically by processing the original signal with the cubic spline interpolation technique. Two simulation signals have been used to demonstrate the effectiveness of the proposed method in this article. The analysis results indicate that this method has good performance in eliminating the residue noise and reducing the costing time, which also provides more accurate decomposition results than the original ensemble empirical mode decomposition. Finally, the application to the feature extraction of pressure pulsation signal of hydroelectric generator unit shows that the proposed method has strong practicability.