Xiaoting Liu
Electrical Power Testing and Research Academe of Hubei & Central China,
Wuhan, 430077, China
TelePhone: 86-27-86764134, FAX: 86-27-86794495
Fuzhou Feng
Department of Precision Instrument, Tsinghua university, Beijing, 100084, China
Telephone: 86-10-62788308, Fax: 86-10-62784691
E-mail: fengfz@post.pim.tsinghua.edu.cn
Abstract: In the presence of good developing chance for modern scientific management of hydroelectric industry in China, according to the requirements of management mode on implementing “no one on site, a few on duty” in hydro-plant, an efficient management mode is proposed for diagnosis and maintenance of operating unit in hydro-plant. And also an integrated system for condition monitoring, diagnosis, maintenance and management(CMDMM) of all operating units is formed so as to adapt to the development characteristics of large-scale electric power network in China. Architecture and contents of the mode is discussed, and considering its architecture characteristic, some key technique on implementing such a system is emphasized.
Keywords: operating unit, diagnosis and maintenance, efficient mode, key technique
At present, conforming to the sprit of document (the No.484 document of the former Electrical Ministry) of Chinese Electrical Corporation, large-scale and medium-size water power station in China are heading for the management mode of “ no one on site, a few on duty”, which will improve safety and economical operation management and also increase automation degree in water power station step by step. It will bring obvious economical efficiency and social efficiency to the water power station and electric power system if equipment maintenance based on condition can be carried out to establish predictive maintenance system and to improve management. Therefore, it is an urgent task to realize integrated system of CMDMM for all the equipment in service, especially for large scale water-turbine generator set, some key technique has been explored in order to put the system into practice in a hydro-plant.
It is a system that integrates CMDMM (including dispatching) for operating unit, which will supply high performance, high-usage, high efficiency. The integration of CMDMM is a sign of large scale power network and high efficient management, and also the tendency of dispatching automation of network system.
The aim to realize the integration system of CMDMM (including dispatching) for operating unit in a hydro-plant is to supply an interface for connecting the equipment condition monitoring and diagnosis system(CMDS) to the plant management information system (MIS), which can supply enough information for leaders in the hydro-plant or corporation to make reasonable decisions. On the other hand, the information can be transmitted to the electric control center in both power network and bureau of a province, and also to the long distance diagnosis center by means of WAN (Wide Area Network). Then long distance diagnosis, maintenance and high effectiveness management can be realized. In short, the integration can be said as a modern commercial running mechanism which conforms to the principle of electric power production and adapts to the principle of electric power market.(see figure 1)
The general arrangement of such a system is consisted of signal source, signal input, condition monitoring, diagnostic system, signal communication, management and dispatching etc, in which the main component is CMDS for the equipment. Signal sources are captured from equipment supervision and control system for main unit which can be listed as hydraulic turbine and water-pass part, guide bearing and thrust bearing, generator, field regulator, speed governor and lubricant, etc. Data acquisition can be completed by unit condition monitoring device. Typical characteristics extracted from the selected data will be transmitted to the diagnostic system so as to predict the operating unit condition, service life and determine whether to repair, replace or change the operating condition of the unit. These procedures will help to supply maintenance plan, management decision and operation dispatching.
However, to put the CMDMM system into practice in a hydro-plant, there is still a great deal of work to be done, especially some key technique about how to implement the system. Network construction of the system, operating unit fault diagnosis system, equipment maintenance and decision system are emphasized in this paper respectively. Other implementing technique about the integrated system, such as measuring point selection and location, sensor selection and usage, signal characteristics of the operating unit, analysis and processing, and so on which can be referenced in paper[2][3][4].
Seeing from fig.1, the whole system is an integrated system based on an opening distributed on-line monitoring system. Considering the network construction, the system can be divided into three parts: the first one is to connect all monitoring units for different operating unit in a hydro-plant to the bus on field. Then signal from different unit can be captured and analyzed before transmitted to and saved in a general monitoring and diagnosis workstation (Data server for one operating unit). The second one is to transmit signals from different general monitoring diagnosis workstation to data server and maintenance workstation (server for all operating units in a hydro-plant) by 10M/100M high speed Ethernet. The third one is to transmit data from server on plant LAN to long distance diagnosis and maintenance workstation by means of communication server and WAN or private line. This kind of arrangement assures the flexibility, practicability and easy to operate of the network construction, which can help to realize not only the diagnosis, maintenance and management on the level of plant LAN, but also the long distance diagnosis, maintenance and management by WAN. At the same time, by means of Plant LAN and MIS, relevant information can be transmitted to electric control system so as to supply enough information for dispatching management decision of electric system. Therefore, the problems of data communication and resource sharing should be solved firstly to complete the network construction of the system. Once the system is put into practice in the power network, as a local area control system in a hydro-plant, it should include several subsystem, such as supervising system for equipment safety, MIS for management of enterprise, dispatching system for power production dispatching, condition monitoring and fault diagnosis for operating unit and so on. It is very difficult to realize data exchange, information communication and resource sharing among these subsystems in the case of assuring not only non-interfering, but also good compatibility. To solve these problems, data communication and database appropriate for integrated management mode, that is to say, distributed database should be explored.
Relational data base may be used for data saving and querying so as to realize distribution database. And the data should be transformed into a unified specification for sharing; Under allowable circumstances, in order to increase accuracy and reliability, data base should be established step by step; On the aspect of long distance diagnosis and maintenance management, to choose a reasonable operating platform based on network, protocol and integration problem of multiple protocol, especially those data relevant to process control system, are very important for realizing real-time communication control of the whole system.
It is a very important procedure to establish fault diagnosis system appropriate for hydroelectric operating unit. From the viewpoint of implementation, there are four specific technical problems to be solved: the first one is typical signal acquisition and symptom extraction based on signal analysis; the second one is to determine the fault type of the operating unit; the third one is0 to design a practical diagnosis model; the last one is to establish a reliable, accurate and credible expert.
Since the condition data acquisition is the base of fault diagnosis, it will directly influence the predictive diagnosis results. Condition data and state parameters at different work condition should be firstly captured, analyzed, then symptom extracted to diagnose, maintain and manage the condition and performance of the operating unit.
Generally, state signal of the operating unit is displayed with two formats: one is the characteristic signal indicated as energy format, the other is the characteristic signal indicated as physical state format. For the first kind of signal, such as vibration, swing, pressure, temperature, shaft, voltage, current and so on, symptoms can be extracted in time domain, as well as in frequency domain, amplitude domain or phase domain; for the second kind of signal, symptoms can be extracted directly by flaw, smoke, oil quality, etc, as well as by means of physical or chemical methods, such as detection analysis, spectroscopic analysis, etc.). For signals of a hydro-turbine unit are indicated as energy format, so frequency is the most important symptom. Table.1 list some characteristic frequencies of a hydro-turbine unit.
Because of the complexity and multiplicity of failure in operating unit, while doing predictive diagnosis, failure type, indicating format and cause should be carefully explored. Based on mechanism research, to determine failure type and property rapidly and accurately will help to establish failure diagnosis model and expert system so as to realize monitoring, diagnosis, maintenance and management of operating unit. Here still taking water turbine vertically mounted as an example, the common failure types are listed in Table.2.
For the main auxiliary equipment, the failure types speed governor and magnetization control system is not listed in Table.2. The failure type listed in Table.2 can be cascaded analyzed to produce multi-level failure modes which can be determined as “fault tree”.
Equipment fault diagnosis model is the bridge that connects engineering expert to diagnosis expert system. To some degree, the model can determine the basic problems, such as knowledge organization, knowledge acquisition, knowledge presentation, inference strategy and control strategy, and so on. At present, because of the development of diagnosis technique, more and more diagnosis models have been proposed, and also many expert systems have been established based on these models, such as the intelligent mode based on numerical calculation and knowledge (including deep knowledge and superficial knowledge). Some achievement has been obtained with the application of this mode. But for some natural shortages, other shortages are also existed while applying this mode into fault diagnosis:
(1) The reference signal of different failures in diagnosis system are given artificially, not directly captured from practical unit, so it can not reflect the operating characteristics and regularity faithfully.
(2) Diagnosis technique basically stay on the conventional level of extracting frequency symptom, it is too difficult to identify the fault location and fault severity.
(3) Degree of intelligent is not high, and diagnosis results are too fuzzy, almost completely depending on engineer expert.
Therefore, deep research of diagnosis model should go further to find out the appropriate model for fault diagnosis for water turbine generator set of natural characteristic and fault symptom of the operating unit, the three item mentioned above should be firstly considered. Then considering the practical situation of operating unit, a diagnosis model appropriate for water turbine generator set will be obtained.
Condition monitoring and fault diagnosis system for operating unit in hydro-plant can be said as four levels from the viewpoint of function, that is to say, condition monitoring---analysis and diagnosis----predictive diagnosis---decision. Among the four levels, it is not only the hard core of the whole system, but also the base of implementing the integrated system of monitoring, diagnosis, maintenance and management to establish a reliable, accurate and credible diagnosis expert system.
Generally, an expert system is consisted of these components, such as data base, knowledge base, inference engine, context, knowledge acquisition, explanation system, learning system, editing, compilation and so on. Whatever structure of expert system to be adopted is not only relevant to purpose, function and content, but also relevant to the equipment, its performance and its characteristic of implementing the monitoring, diagnosis system. As a monitoring and diagnosis system for water turbine generator set, according to the basic principle of establishing expert system, the structure of the future expert system may be designed as Fig.2.
(1) Real-time data base. It is a complete software consisted of dynamic and static data base in which all kinds of data are the main evidence while diagnosing. It includes data in time domain from equipment condition monitoring components, structure parameters of the equipment, data from assembling, operating and testing and data management. To assure completion, validation and practicability of the data and flexibility of data management are the necessary condition for real-time data base.
(2) Knowledge base. Knowledge base is the hard core of diagnosis expert system to save knowledge of domain expert. Diagnostic ability and efficiency of an expert system are determined by completeness of the knowledge base. Diagnostic knowledge base is consisted of knowledge base from expert experience, rule base and other relevant material base, then it is presented with a certain format. With the development of science and technology, knowledge base will be continuously expanded, modified and renewed. Therefore, methods of knowledge presentation should allow the addition and acquisition of knowledge conveniently.
(3) Inference engine. It is the key component of realizing a diagnostic expert system. Its performance will directly affect the validity of decision. Generally, there are three kinds of inference engine: forward-inference, backward-inference, forward and backward mixing inference. Considering the structure characteristic and specialty of different kinds of failure, it is better for a reasonable diagnostic model to combine forward and backward mixing inference engine.
(4) Explanation system. Tracing explanation system is to explain the behavior and diagnostic results of the system for customers. The explanation of the inference rules can be indicated as the format of frame, data table and text.
Except for the main module mentioned above, compilation system and symptom extractor should be carefully considered during establishment of an expert system. Formal knowledge base and diagnostic knowledge base may be obtained through compilation system. Symptom extractor will capture the on-line monitoring condition data, analyze and process the data to yield characteristic signal. Then inference engine will use these characteristic signals to complete diagnosis.
4.4.2 Diagnostic decision for operating unit
The reasonable conclusion may be drawn for operating water turbine generator set by condition monitoring and fault diagnosis and predictive diagnosis. Decision can be transmitted to decision department or electric power network-province dispatching control center and long distance diagnostic center by decision report software module. Then maintenance and management of the operating unit can be determined, the maintenance based on condition can be implemented. In other words, hereto, the high efficient integrated mode for condition monitoring, diagnosis, maintenance and management has been formed.
Condition based maintenance (CBM) is a maintenance mode based on equipment condition monitoring analysis and diagnosis, which depends on cost-effectiveness analysis to make decision. To implement management of CBM involves several problems, such as maintenance risk analysis and decision, integrated management of equipment information, management of spare parts and components, management of maintenance staff and maintenance efficient analysis, and so on. It is the core problem to implement CBM based on the decision depending on maintenance risk analysis, which includes the following problems: whether to repair, how to repair, what to repair, when to repair and how to support maintenance and evaluate maintenance effect, etc. Therefore, based on the theoretical analysis for maintenance mentioned above, many key problems should be further explored, such as information management of CBM, maintenance decision and optimization of cost-effectiveness analysis, maintenance decision technique of expert system based on knowledge, etc. So a predictive diagnosis, maintenance and overhaul decision system adapting to the practical situation of power production in China will be established in the near future.
(1) The integrated system of CMDMM for hydro-turbine unit considers the operating characteristics and development of large-scale electric network in China. And it also indicates the characteristics of high-voltage network and high efficient energy, and specifies the mode of MIS, supervising system, production dispatching system and unit condition monitoring and diagnostic system in some large-scale and medium size hydro-plant.
(2) To assure reliable data exchange, information communication , resource sharing and excellent compatibility, the design of distributed data base should be first solved. While implementing the system, Data saving and querying should adopt relational data base, and all data should be transferred into a unified specification.
(3) It is a widely used and efficient method to establish a reliable, accurate and credible expert system. Design and development of an expert system should closely combine with the practical situation of unit in hydro-plant. Meanwhile, in order to design the hard core and key components ofof an expert system,that is to say, knowledge base and infenrence engine, the operating performance and structure characteristics of the unit should be considered carefully.
(4) Based on maintenance theoretical analysis, many key problems should be further explored, such as information management of CBM, maintenance decision and optimization of cost-effectiveness analysis, maintenance decision technique of expert system based on knowledge, etc. These key techniques are the fundamentality to develop a diagnosis, maintenance and overhaul decision system adapting to the practical situation of power production in China.
References
[1] Liu XiaoTing, A new mode of condition monitoring and diagnosis for operating unit, Proceeding of the XIX IAHR Symposium (Hydraulic Machinery and Cavitation), Singapore Vol.2 P576,1998.
[2] Liu XiaoTing, The placement exploration of diagnostic measuring points on operation unit equipment, proceedings of the XVIII IAHR symposium (Hydraulic machinery and cavitation), Valencial Spain, Vol. 2 P1172, 1996.
[3] Liu XiaoTing, Li. W. F, Water power unit site measurement handbook, Hydraulic and electric power press, Beijing, 1993.
[4] Liu XiaoTing, Condition monitoring and diagnostic technology for operation of hydro-generator set, Journal of Vibration engineering, Vol.13, No.3, p392, 2000.
Fig. 1 General arrangement of the integrated system of diagnosis and maintenance management for operating unit
Fig. 2 Structure diagram of diagnosis expert system