Expert System of Fault Diagnosis for Gear Box in Wind Turbine Systems Engineering Procedia 4 (2012) 189 – 195 2211-3819 © 2011 Published by Elsevier Ltd. Selection and peer-review under responsibility of Desheng Dash Wu. doi:10.1016/j.sepro.2011.11.065 Available online at www.sciencedirect.com Systems Engineering Procedia Systems Engineering Procedia 00 (2011) 000–000 www.elsevier.com/locate/procedia The 2nd Expert System of Fault Diagnosis for Gear Box in Wind Turbine International Conference on Complexity Science & Information Engineering YANG Zhi-Ling∗ North China Electric Power University, Beijing,102206, China , WANG Bin, DONG Xing-Hui, LIU Hao Abstract Gear box is one of the key components in wind turbine. To make timely and accurate diagnosis for gear box, an expert system based on fault tree analysis is developed in this paper. Firstly, a fault tree model is established in accordance with the structural features of wind turbine gearbox. On basis of the fault tree model, qualitative analysis and quantitative analysis of the gearbox faults are carried out afterwards. Finally，a Web-oriented expert system is developed by C# on the .NET platform which will save the fault diagnosis time significantly and make the expert solution for the fault of gear box to achieve the precise and quick maintenance more effectively. © 2011 Published by Elsevier Ltd. Selection and peer-review under responsibility of Desheng Dash Wu Keywords: expert system, fault diagnosis, industrial engineering, gear box, wind turbine; 1. Introduction Gear box is one of the key components in wind turbine and the fault of gear box will lead to wind turbine shut down inevitably. So what must be done is making timely and accurate diagnosis to gear box. However, due to rapid development of wind power industry, there is no fault diagnosis system specifically for wind turbine gearbox on the market in China. Currently, during the wind turbine gearbox fault diagnosis, workers generally analyze the phenomenon by rule of thumb. This fault diagnosis approach will spend more time and the accuracy is low. Statistics show that the determining time for the fault diagnosis takes up 70% to 90% of the total time, while the repair time takes up only about 10% to 30%[1]. Fault diagnosis systems have been successfully applied in many kinds of technical processes to improve operation reliability and safety. Wu introduced an expert system for fault diagnosis in internal combustion engines using wavelet packet transform and neural network[2]. Lei proposed a new multidimensional hybrid intelligent diagnosis method to identify different categories and levels of gear damage automatically[3]. Saravanan dealt with the effectiveness of wavelet-based features for fault diagnosis using support vector machines and proximal support vector machines[4]. In another paper, he introduced the use of decision tree for selecting best statistical features that discriminate the fault conditions of the gear box from the signals extracted and the decision tree was also used to generate the rules automatically from the feature set [5]. ∗ Corresponding author. E-mail address: yzhil@ncepu.edu.cn. © 2011 Published by Elsevier Ltd. Selection and peer-review under responsibility of Desheng Dash Wu. Open access under CC BY-NC-ND license. Open access under CC BY-NC-ND license. http://creativecommons.org/licenses/by-nc-nd/3.0/ http://creativecommons.org/licenses/by-nc-nd/3.0/ 190 YANG Zhi-Ling et al. / Systems Engineering Procedia 4 (2012) 189 – 195 2 YANG Zhi-Ling / Systems Engineering Procedia 00 (2011) 000–000 Fault diagnosis systems are relatively new and are flourishing on a rapid scale in wind energy technology, especially for the development of large-scale wind turbines. Barszcz presents the application of the spectral kurtosis technique for detection of a tooth crack in the planetary gear of a wind turbine[6]. Tang presents a wind turbine fault diagnosis method based on the Morlet wavelet transformation and Wigner-Ville distribution[7]. Jiang applied a new denoising method based on adaptive Morletwavelet and singular value decomposition (SVD) to feature extraction for wind turbine vibration signals[8]. Hussain proposed a novel method for real time fault detection in gearboxes by using adaptive features extraction algorithm to deal with non-stationary faulty signals[9]. This paper applies the fault diagnosis system on wind turbine gear box, in order to make timely, accurate diagnosis for gear box and provide timely solutions. In this way, it can not only meet the requirements of rapid response to wind turbine fault diagnosis, but also prevent the emergence of human error, the valuable experience of experts can be retained to prevent lost, too. Fault tree analysis is an intuitive, clear, logical, and applicable system compares to the relatively simple ones. The gear box of wind turbine has less hierarchical structure, so fault tree is used to analyze the component faults and express their reasons, and an expert system based on fault tree analysis is chosen to make diagnosis on wind turbine gear box in this paper. 2. Fault Tree of Gear Box in Wind Turbine 2.1. Fault Tree Analysis Fault tree analysis is a graphical interpretation method which can refine the system fault causes from whole to part by dendrite progressively. It is also an important method of analyzing system reliability and safety. Fault tree is drawn through analyzing a variety of factors (including hardware, software, environment, human factors, etc.) which should cause system fault, refining system fault event from whole to part by dendrite progressively, identifying the cause of system fault, determining the probability of fault, and evaluating the important degree related to the cause[10]. Fault tree model is a behavior and qualitative causal model based on studying the structure and functional characteristics of fault. The top event is the event which is to be avoided firstly. The basic events are those can cause top event while the intermediate events are the nodes. Fault tree is a dendritic logic diagram which shows relevance between events by using logic gates. As shown in Fig. 1. System fault Module A fault Module B fault C1 C2 and C3 C4 or or Fig.1 Fault tree diagram 191 YANG Zhi-Ling et al. / Systems Engineering Procedia 4 ( 2012 ) 189 – 195 YANG Zhi-Ling / Systems Engineering Procedia 00 (2011) 000–000 3 In Fig. 1, the top event is system fault caused by intermediate events of component A fault or component B fault. Component A fault is caused by basic event of part 1 fault or part 2 fault. The basic events of part 3 fault and part 4 fault happen will cause the component B fault. 2.2. Fault Tree of Gear Box in Wind Turbine From the current structural characteristics of gear box in wind turbine and its actual fault conditions, faults usually occur in parts such as gears, shafts, bearings and box. But in order to perfect the fault tree, fastener fault and oil seal fault are added to the fault tree. In the fault tree of gear box, wind turbine gear box fault is the top event; the intermediate events include gear fault, shaft fault, bearing fault, the box fault, fasteners fault, seals fault and other events which can be further refined; the damage of parts or improper operation is taken as the basic event. As fault of any component of the gear box will cause a system fault, the main logical relationship between the fault is "or". The final fault tree is shown in Figure 2(M stands for intermediate event, while C stands for basic event). Wind turbine gear box fault Seal faultFastener faultBox faultBearing faultShaft faultGear fault M1 Shaft misalignment Shaft bending Shaft loose C7 C9C8C6C5C3 C4...C2C1 C10 C10C11 ... w1 w5w4w2 w3 assembling error Shaft deformation Uneven settlement Box wear or deformation Bearing wear or deformation Load too large Improper placement Speed ​​too fast Unequal selection Heating less Fig.2 Wind turbine gear box fault tree diagram In order to get an extended fault tree, auxiliary information including the weight and solution of fault is added to each basic event. Fig.2 only shows corresponding weight of intermediate event M3. The relationship of weights is: (1) In which, ωi Finally the diagnostic knowledge base is obtained by using production rules to put the fault tree information into knowledge base [11]. According to the diagnostic knowledge base, not only the specific locations and causes of fault can be obtained, but also the reference solution can be given by experts. stands for the corresponding weight of basic event C. 192 YANG Zhi-Ling et al. / Systems Engineering Procedia 4 (2012) 189 – 195 4 YANG Zhi-Ling / Systems Engineering Procedia 00 (2011) 000–000 3. Fault Analysis Based on Fault Tree 3.1. Qualitative Analysis The task of qualitative analysis is to identify the possible causes of system top event fault, or to find out the minimal cut sets of the fault tree [12]. The method of finding the minimum cut set is mainly upstream and downstream. According to logical relationship between faults of gear box in wind turbine, this paper adopts downstream. The idea of downstream is: according to the two adjacent terms in the fault tree, logic "or" gate increases the number of cut sets and logic "and" gate increases the capacity of the cut set [13]. Starting from the top event, the cut sets of fault tree is obtained by replacing the previous events into the next level event, writing event vertically when meeting "or" gate, until the gates are all converted into the top event. Fig. 3 shows the minimal cut sets are gotten by qualitative analysis which take the gearbox fault, axis fault and axis bending as the main line. The minimal cut sets of other components fault can be obtained by the same method. Due to the structural characteristics of gear box, the resulting cut sets are all the minimal cut sets, which are {C1},{C2},{C3},{C4},{C5},{C6},{C7},{...}. Minimal cut sets are independent of each other. Gear box fault Gear fault Bearing fault Shaft fault Box fault ... Shaft misalignment Shaft bending Shaft loose Unequal selection Improper placement Heating less Speed ​​too fast Load too large . Fig.3 Qualitative analysis of gear box fault 3.2. Quantitative analysis After obtaining all minimal cut sets, if there is enough data which can support inferring probability of each basic event, the further quantitative analysis can be made [14]. The main purpose of quantitative analysis is to find the characteristic quantities of the top event (such as the top event probability and system reliability) and the importance of the basic event, which can determine the weaknesses in the system. 1) Calculate the probability of top event The probability of occurrence of top event with "and" gate can be calculated by following formula: 11 ( ) ( ) ( ) n n i i ii p x p x p x == = =∏ (2) The probability of occurrence of top event with "or" gate can be calculated by following formula: 11 ( ) ( ) 1 [1 ( )] n n i i ii p x p x p x == = = − −∏ (3) 193YANG Zhi-Ling et al. / Systems Engineering Procedia 4 (2012) 189 – 195 YANG Zhi-Ling / Systems Engineering Procedia 00 (2011) 000–000 5 In which, p(x) is probability of the fault tree top event, p(xi) is probability of the i-th minimal cut set of fault tree, n is the total number of fault tree minimal cut set. 2) Importance analysis of basic event Importance analysis is to analyze the contribution of fault, which occurred in the minimal cut sets of a component, to the probability of top event. Probability importance degree of basic event describes the extent of probability change of top event caused by the probability change of i-th basic event fault, which is expressed by the formula as follow: ni xp xp xI i i ...1,)( )( )( = ∂ ∂ = (4) Ii(x) is probability importance degree of i-th basic event, p(x) is probability of the tree top event fault, p(xi) is probability of the i-th minimal cut set of fault tree, n is the amount of minimal cut set in the fault tree. 4. Expert System of Fault Diagnosis On the basis of fault tree analysis for gear box, a Web-oriented expert system for fault diagnosis of gear box is developed. The system is designed to have three grades, including symptoms page, fault causes page, solutions page. After entering the system, the default page is a symptoms page. As shown in Fig. 4. Fig.4 Qualitative analysis of gear box fault This system is developed by C# on the .NET platform. The overall framework of the system is as shown in Fig. 4. According to the structural of the gear box, the left side of page is made to accordion style, in which the six components fault of gear box are defined to the category name; the right side of the page defaults symptoms page, in which the previous column of symptoms column displays corresponding fault category and the next column links possible causes. 1) To query fault symptom In the symptoms page, the user can query page by page or do fuzzy query by writing keywords in the search box, and click the “search” button to get the results you want. 2) To query reason of fault 194 YANG Zhi-Ling et al. / Systems Engineering Procedia 4 (2012) 189 – 195 6 YANG Zhi-Ling / Systems Engineering Procedia 00 (2011) 000–000 The user can get corresponding possible causes of fault symptom by clicking on the details link of possible cause column in the above list. They can also directly click on the corresponding fault symptom in accordion structure to get a list of corresponding reasons. The query result is ordered by the weight of reason and the reason whose weight is large is sorted in the top of the list. 3) To query solution of fault After getting the list of fault reason, the user can click on the details link of solution column to get the recommended solution given by expert. Recommended solution will be ordered from top to bottom, if there are multiple. Maintenance staff can choose different solutions according to the circumstances. 5. Conclusion Gear box is one of the key components of wind turbine, whose fault will directly affects the safety and the operation of the overall wind turbine. This paper develops an expert system of fault diagnosis for gearbox in wind turbine based on fault tree analysis. With the gradual accumulation of fault and repair instance about wind turbine gear box, the expert system based on fault tree analysis will play a better role in the fault diagnosis of wind turbine gear box. Acknowledgements This study is supported by “the Fundamental Research Funds for the Central Universities” (09MG23). References [1]WANG Yi, FENT Xiaoyun. Electric locomotive fault diagnosis expert system based on fault tree. Electric Locomotives & Mass Transit Vehicles 2004; 27: 35-36. [2]Jian-Da Wu, Chiu-Hong Liu. An expert system for fault diagnosis in internal combustion engines using wavelet packet transform and neural network. Expert Systems with Applications 2009; 36: 4278-4286. [3]Yaguo Lei, Ming J. Zuo, Zhengjia He, Yanyang Zi. A multidimensional hybrid intelligent method for gear fault diagnosis. Expert Systems with Applications 2010; 37: 1419-1430. [4]N. Saravanan, V.N.S. Kumar Siddabattuni, K.I. Ramachandran. A comparative study on classification of features by SVM and PSVM extracted using Morlet wavelet for fault diagnosis of spur bevel gear box. Expert Systems with Applications 2008; 35: 1351-1366. [5]N. Saravanan, S. Cholairajan, K.I. Ramachandran. Vibration-based fault diagnosis of spur bevel gear box using fuzzy technique. Expert Systems with Applications 2009; 36: 3119-3135. [6]Tomasz Barszcz, RobertB Randall. Application of spectral kurtosis for detection of a tooth crack in the planetary gear of a wind turbine. Mechanical Systems and Signal Processing 2009; 23:1352–1365. [7]Baoping Tang, Wenyi Liu, Tao Song. Wind turbine fault diagnosis based on Morlet wavelet transformation and Wigner-Ville distribution. Renewable Energy 2010; 35: 2862-2866. [8]Yonghua Jiang, Baoping Tang, Yi Qin, Wenyi Liu. Feature extraction method of wind turbine based on adaptive Morlet wavelet and SVD. Renewable Energy 2011; 36: 2146-2153. [9]Sajid Hussain, HossamA.Gabbar. A novel method for real time gear fault detection based on pulse shape analysis. Mechanical Systems and Signal Processing 2011; 25: 1287–1298. [10]ZHU Xuejun,CHEN Yu. The research of intelligent fault diagnosing methods based on FTA. Microcomputer Information 2005; 21(6): 123-124. [11]YANG Bin,DING Hui,LUO Weimin,ZHANG Xiaofei.( 2002) Expert system of transformer fault diagnosis based on knowledge, Proceedings of the CSEE, 22(10): 121-124. 195YANG Zhi-Ling et al. / Systems Engineering Procedia 4 (2012) 189 – 195 YANG Zhi-Ling / Systems Engineering Procedia 00 (2011) 000–000 7 [12]YU Aihua. Fault tree analysis and application to hydraulic system maintenance. Hoisting and Conveying Machinery 2006; 6: 60-62. [13]CHEN Yongwang. Application of fault tree analysis method in fault diagnosis for 16V280 diesel engine fuel system. Mechanical Research & Application 2010;21(3): 85-86. [14]SHI Dinghua,WANG Songrui. Fault tree analysis methods and theories. Beijing Normal University Press 1993:.