Report on the Course in Mechanical Fault Diagnosis*

Thread Content

Report on the Course of Mechanical Fault Diagnosis* – Application of Vibration Signal Analysis Techniques in the Fault Diagnosis of Power Plant Equipment 【Abstract】 It introduces the typical faults of the main equipment in thermal power plants as well as the fault diagnosis methods used. It discusses the existing fault diagnosis systems and the problems present in the fault diagnosis of power plant equipment, and points out the development trends of such diagnosis systems. 【Keywords】Vibration; Power plant; Fault diagnosis. I. Characteristics of faults in power plant equipment 1. Main faults of boilers The common faults in boilers include: overheater leakage, economizer leakage, water wall leakage, and dust collector failures. 2. Main faults of steam turbine units (1) Main faults of the steam turbine ① Imbalance. It is mainly caused by machining errors and permanent bending of the rotor, blade detachment, coupling drift, poor manufacturing of the extension shaft, thermal warping due to uneven heating, or material defects. ②Not aligned. The main reasons include jamming of the sliding pin system, excessive temperature differences in the cylinders, too rapid acceleration of the unit, significant drops in vacuum, uneven settlement of the foundation, etc. These factors lead to poor expansion of the unit and uneven expansion of the bearing housings, as well as coupling eccentricity. ③Rotor rubbing. The main reasons include improper design and adjustment of the steam seal clearance, poor expansion of the cylinder, excessive thermal deformation of the cylinder, excessive vibration of the unit, and increased axial displacement. ④ Leaves fall off. This is mainly due to insufficient design strength, large fluctuations in steam parameters, the unit operating for extended periods under conditions of significant changes in grid frequency and at low loads, as well as the unit staying at a speed lower than its rated speed for too long. ⑤Steam excitation. The main reasons are the low critical speed of the high and medium pressure rotors, significant changes in their deflection under hot conditions, substantial variations in bearing levels, poor quality in the machining and installation of the extension shafts, excessive radial or axial errors in the couplings, as well as deteriorated shafting balance and alignment, all of which lead to the instability of the high and medium pressure rotors. (2) Main faults of the generator: ① Oil film oscillation. This is mainly due to the excessively low first-order critical speed of the generator rotor, as well as the poor stability of the three-axis wedge bearings; as the load is reduced, their instability speed becomes too low. ②Stator coil insulation failure. It is mainly due to insulation failure caused by wear, aging, contamination, and corrosion, which leads to partial discharges and increased temperatures; as well as damage to the insulation layer resulting from loose binding, cooling water leakage, and fatigue wear, which causes wire rod displacement and inter-turn short circuits. ③Stator coil overheating. It is mainly due to certain defects during manufacturing or installation that cause inter-turn short circuits, resulting in localized overheating. ④Rotor winding fault. Including ground faults, inter-turn short circuits, and open circuit faults. Grounding and inter-turn short-circuit faults are mainly caused by reduced insulation and damage, while weld separation at joints and thermal deformation can lead to wire breaks. ⑤Cooling water system failure. It is mainly due to reduced or interrupted cooling water flow caused by cooling water leaks, blockages in pipes by foreign objects, malfunctions, etc., as well as stator leakage resulting from material and installation defects. 3. Main faults of transformers The main faults of transformers include the following eight types: short circuits between turns of the windings ； Winding open circuit ； Winding-to-ground breakdown ； Inter-winding short circuit ； Local short circuit or local melting of the iron chip ； Poor contact in the tap switch, with melting of the contact surfaces ； Casing ground flashover or explosion ； The oil temperature rises when the load is normal. II. Methods for Vibration Fault Diagnosis. Currently, there are many methods for vibration fault diagnosis, which can be roughly divided into 3 categories: fault diagnosis based on control models, fault diagnosis based on pattern recognition, and fault diagnosis based on artificial intelligence. (1) Fault diagnosis based on control models: For dynamic systems, if a model can be established through theoretical or experimental methods, changes in system parameters or states can directly reflect changes in the physical system or physical processes of the equipment, providing a basis for fault diagnosis. This method involves techniques such as model establishment, parameter estimation, state estimation, and the use of observers; among these, parameter and state estimation techniques are crucial, as an accurate model of the system is required. In actual production environments, for complex equipment, this method is not economically viable. (2) Fault diagnosis based on pattern recognition: Pattern recognition involves classifying or describing a series of processes or events, and it is mainly divided into two categories: statistical methods and linguistic structure methods. The fault diagnosis of equipment can be regarded as a pattern recognition process: the operating condition parameters of the equipment are measured and recorded, from which fault indicator parameters are extracted. For different fault conditions, corresponding indicator parameters form distinct patterns; by matching the system’s state pattern with the fault pattern samples in the fault dictionary, the fault of the equipment can be identified. When the model of the system is unknown or highly complex, this method provides a simple and effective means to address fault diagnosis problems. (3) Artificial intelligence-based fault diagnosis. Research on artificial intelligence-based fault diagnosis is mainly divided into two categories: knowledge-based (symbolic reasoning) fault diagnosis and neural network-based (numerical calculation) fault diagnosis. ①Knowledge-based fault diagnosis has roughly gone through two development stages: expert systems based on shallow knowledge (rules) and expert systems based on deep knowledge (model knowledge). An expert system is an artificial intelligence software system that utilizes the experiential knowledge of domain experts; based on the information and data provided by the user regarding a problem, it uses certain reasoning mechanisms to select the most appropriate explanation for that problem from a knowledge base. Knowledge-based fault diagnosis expert systems have many advantages, such as: they are suitable for simulating the logical thinking process of humans and for solving complex diagnostic problems that require logical reasoning ； Knowledge can be represented by symbols; with known basic rules, extensive detailed knowledge is not required ； It facilitates integration with traditional symbol database interfaces, etc. Although many mature commercial software solutions have been developed and are used in engineering practice, there are still some issues: the \"bottleneck\" problem in knowledge acquisition remains difficult to resolve ； “The “narrow steps of knowledge” problem ； It is prone to problems such as “combination explosion” and “infinite recursion” ； Real-time online diagnosis can vary in quality. ② Fault diagnosis based on neural networks: Neural networks are adaptive nonlinear dynamic systems created by simulating biological nervous systems. They possess learning capabilities and parallel computing abilities, and can perform functions such as classification, self-organization, associative memory, and nonlinear optimization. Neural networks are used in the field of fault diagnosis, and can address problems related to trend prediction and diagnostic reasoning. Currently, networks such as Multilayer Perceptrons (MLP), Adaptive Resonance Theory (ART), Self-Organizing Feature Mapping (FM), and Bidirectional Associative Memory (BAM) are widely used in fault diagnosis. Intelligent fault diagnosis based on neural networks has many advantages: unified form of knowledge representation, easy organization and management of knowledge bases, strong versatility, and ease of transplantation and expansion ； Knowledge acquisition can be easily automated (such as self-organizing self-learning*) ； It enables parallel association and adaptive reasoning, with strong fault tolerance ； Capable of representing complex relationships between things (such as fuzzy relationships) ； It can avoid the \"combination explosion\" and \"infinite recursion\" problems of traditional expert systems ； The reasoning process is simple, enabling real-time online diagnosis. However, there are also some problems: it is difficult to obtain training samples ； The experiential knowledge of domain experts was ignored ； Knowledge representation in the form of connection weights is difficult to understand, among other issues. (4) Other methods: As fault diagnosis is moving toward greater intelligence, other methods from the field of artificial intelligence, such as fuzzy theory, fuzzy neural networks, and rough set theory, have also been applied in areas such as diagnostic reasoning and diagnostic data processing. 3. Application of vibration diagnosis technology in the fault diagnosis of power plant equipment. In the fault diagnosis of steam turbine units, the vibration method is one of the most widely used and well-established approaches. Vibration faults exhibit complex vibration characteristics in the field, with multiple faults often occurring simultaneously. Therefore, to analyze and diagnose the specific causes of unit vibration quickly and accurately based on the characteristics and patterns of vibration, analysis and handling can be carried out from the following five aspects. 1 Data processing of vibration signals: The data processing of vibration signals involves the acquisition of these signals as well as various spectral analyses; it is the first step in vibration fault diagnosis. A preliminary classification of vibrations can be determined based on various spectral analyses of the vibrations. 2 Determining the faulty location Identifying the location of the fault is an important aspect of diagnostic knowledge. Correctly identifying the location of the fault plays an important role in narrowing down the scope of diagnosis and enabling a rapid and accurate diagnosis. For vibration faults at asynchronous speeds, the rotor and bearing where the fault lies are determined based on the location where excessive vibration occurs and the type of vibration. For faults related to synchronous speed vibration, the location of the fault can be determined based on the transmission characteristics of rotor vibration. The vibration amplitude transfer characteristics, combined with phase transfer characteristics and unbalance response calculations, can be used to identify the location of faults. 3 Characteristics of vibration variation The characteristics of vibration variation refer to the patterns of change in the vibration vector, including the rate of change in vibration amplitude and the patterns of change in vibration phase; these are important aspects in the diagnosis of vibration-related faults. Especially for unstable vibrations, both the amplitude and phase of the vibration change over time under varying conditions, either increasing or decreasing, or accelerating or decelerating. Studying the characteristics of vibration variation patterns for different vibration faults is of great significance for fault identification. 4 Relationship between vibration and related quantities. Each type of fault in a power unit is characterized not only by its vibration parameters but also by certain related quantities, such as the pressure and temperature of main steam and reheat steam, steam flow rate, exhaust steam temperature, condenser vacuum, lubricating oil pressure, absolute and relative expansion of the unit, active and reactive loads of the generator, rotor current, cooling air temperature, cooling water temperature, and flow rate. Therefore, it is essential to understand the relationship between the causes of vibration faults and these related quantities in the diagnosis of vibration faults. A diagnostic system that only includes simple vibration parameters, without the operating parameters of the unit and other related quantities, cannot be considered a complete fault diagnosis system. Since the relevant parameters of large-scale units are already available in the power plant’s data acquisition systems (DAS or DCS systems), data communication technologies can be used to transmit these parameters to the fault diagnosis system, where they can be combined with vibration parameters for the diagnosis of unit vibration faults. Structural analysis of 5-unit systems: The failure characteristics of these systems are closely related to their structural features; some failures can only occur under certain structural conditions ； The same type of fault exhibits different symptoms in units with different structural configurations ； The vibration transmission characteristics of a unit are closely related to its structure. The unit structure includes aspects such as the bearing housing type, support method, coupling type, bearing bush type, and expansion method. Experience in vibration control on site shows that different structural configurations of the units can lead to various vibration faults, and they also have a significant impact on the critical speed of the shafting as well as its sensitivity to imbalance. A correct analysis of the unit’s structure helps in diagnosing the location of the fault and its specific causes. Only by including the structural analysis of the unit in the fault analysis system can the specific cause of a fault be determined based on different vibration patterns. The above five aspects can serve as the basic framework for developing an expert system for automatic diagnosis of unit vibration faults in the future. In the future, more in-depth research should be conducted to identify the exact signs of each cause of failure in every aspect, and to establish rules for determining the relationship between these signs and the failures. This is of great significance for establishing a comprehensive fault diagnosis system. IV. New Developments in Vibration Diagnosis Techniques
1. Diversification of Diagnostic Objects
Vibration diagnosis techniques have expanded beyond the initial focus on fault diagnosis in rotating machinery to other application areas. Today, these techniques are used for monitoring in various sectors such as machinery, process industries, nuclear energy, aerospace systems, and transportation. Examples include monitoring the attitude of spacecraft, inspecting the structural integrity of bridges and buildings, and overseeing production processes.

2. Diversity of Diagnostic Techniques
In terms of diagnostic methods, in addition to those that rely on a single parameter or single type of fault, comprehensive diagnostics using multiple parameters and multiple types of faults have become increasingly common. Diagnostic techniques can utilize various types of information such as vibrations, noise, oil conditions, stress levels, electromagnetic signals, and radiation. Moreover, modern diagnostic techniques and signal processing methods such as expert systems, artificial neural networks, wavelet transforms, chaos theory, fractal geometry, and data fusion technologies are also being applied in fault diagnosis to improve accuracy. Each method for diagnosing machine failures has its own advantages and disadvantages; therefore, a combined use of various methods to complement each other’s strengths is likely to be an important development direction.

3. Networking of Diagnostic Systems
With the advent of Internet and Intranet technologies, it has become necessary to network diagnostic systems in order to meet the needs of large enterprises for networked equipment management. By using various communication methods, diagnostic systems can be connected to data collection systems, enterprise management systems, and equipment control systems via the network. This allows different departments within an enterprise to obtain information about the operating status of various devices, as well as to monitor and diagnose equipment in different locations and companies around the world. As machines and equipment become larger, more automated, faster, and more complex, remote monitoring and fault diagnosis based on the Internet are developing rapidly worldwide.

4. Database-Based Diagnostic Systems
Database-based dynamic monitoring systems provide the technical foundation for processing, querying, and utilizing large amounts of monitoring data. Relational and object-oriented databases, database management systems, distributed database management systems, data warehouse technologies, as well as advanced decision support systems and data mining techniques built on these technologies not only provide reliable bases for corporate decision-making but also offer new directions for fault diagnosis.

V. Problems and Development Directions in Fault Diagnosis of Power Plant Equipment
(A) Current Problems in Fault Diagnosis of Power Plant Equipment
Based on current research, the main problems encountered when applying fault diagnosis techniques in the thermal systems of thermal power plants are as follows: (1) Optimization of measurement points: Modern power plants have well-developed monitoring systems to ensure safe operation. How to make effective use of these existing measurement signals and information is a significant research topic that requires expertise and thorough analysis using historical data. (2) Fault location: Existing diagnostic systems mainly determine the cause of faults (such as vacuum leakage in condensers or dust accumulation on heat exchanger tubes), but they are unable to identify the exact location of the fault. (3) Multiple fault diagnosis: In actual production, abnormal system operation is often not caused by a single fault; it is more likely to result from two or more faults. However, the diagnostic models described in existing literature are trained using data related to single faults only, and they lack the ability to distinguish between multiple faults. Therefore, research is needed to develop methods for identifying multiple faults. (4) Fault prediction: It involves the diagnosis of faults in equipment and systems, with the ultimate goal of preventing faults from occurring, or replacing or repairing faulty components before they cause further damage, thereby avoiding increased losses. (II) Development of Equipment Fault Diagnosis in Thermal Power Plants 1. Development trends of fault diagnosis systems (1) Hierarchical distributed structure of fault diagnosis systems The structure and functions of the various subsystems in thermal power plants are distributed and multi-level; this hierarchical structure requires that the diagnostic systems be similarly distributed and multi-level, consisting of a global diagnostic system and sub-diagnostic systems. The global diagnosis system is responsible for managing diagnosis tasks, including breaking down the overall task into sub-tasks and assigning these tasks to various sub-diagnosis systems; these tasks are often interdependent. After the diagnostic subsystems are completed, a final conclusion is drawn by synthesizing the results from each of these subsystems. Distributed fault diagnosis expert systems feature high reasoning efficiency, fast diagnosis speed, system reliability, and good timeliness. (2) Integrated fault diagnosis system: Due to the simplicity of the reasoning methods used in current diagnosis systems, they face significant limitations when dealing with the diagnosis of complex systems. Future fault diagnosis systems for thermal power units will employ different inference models based on the characteristics of various subsystems; in some cases, multiple inference models will be used in combination. The advantages of each inference model will thus be fully utilized, thereby improving the speed and accuracy of inference. (3) Establish a large-scale monitoring and diagnosis center. Within the same power grid, many thermal power units of the same type operate simultaneously. The benefits of establishing a large-scale monitoring and diagnosis center are quite obvious: ① It facilitates the centralized storage of operational data on the units as well as information regarding their health status ； ②It facilitates the sharing of existing knowledge among multiple units and power plants, as well as the improvement of the knowledge base ； ③It facilitates the load scheduling of the unit. References: Wang Junxin, He Xiaoming. Current Status and Development Trends of Equipment Fault Diagnosis in Large Thermal Power Plants. Equipment Management & Maintenance, 2003, 8: 33–35. Yu Wenhу, Jiang Ruijin. Knowledge Scope of the Expert System for Vibration Fault Diagnosis of Ni Weidou Turbine Generators. China Electric Power, 1996, 29(10): 24–28. Wang Guangyang, Zhou Yilian. A Review of Vibration Fault Diagnosis for Fans. Journal of Anhui University of Technology, 2006, 23: 64–68. Xia Xiaolei, Zhang Yuanyue. Development of Mechanical Fault Diagnosis Techniques under Modern Scientific and Technological Conditions. Journal of Xuzhou Institute of Technology, 2006, 21: 67–69. Zheng Puyan, Cai Ningsheng. Applications and Existing Problems of Fault Diagnosis in the Thermal Systems of Thermal Power Plants. Turbine Technology, 2001, 43: 321–323