Application of the hottest artificial intelligence

2022-07-30
  • Detail

The application of artificial intelligence in mechanical fault diagnosis

1 the application direction of artificial intelligence in mechanical fault diagnosis

the so-called mechanical fault diagnosis is to identify whether the technical state of a machine is normal, determine the nature and location of the fault, find the cause of the fault, predict the fault trend, and put forward corresponding countermeasures through the relevant information in the operation of the machine; It is based on fault mechanism and technical detection, and takes signal processing and pattern recognition as its basic theory and method. The general mechanical system fault diagnosis system is physically divided into five parts: mechanical measurement, monitoring and protection, data acquisition, vibration state analysis and network data transmission; In terms of function, the mechanical system condition monitoring and fault diagnosis system can be divided into three parts: data acquisition, condition monitoring and fault diagnosis

with the increasing scale and complexity of modern industrial equipment and systems, the problems of reliability, availability, maintainability and safety of mechanical equipment have become increasingly prominent, which has promoted the research on fault mechanism and diagnosis technology of mechanical equipment. With the rapid development of computer technology and digital signal processing technology, mechanical equipment vibration monitoring and fault diagnosis technology has been widely used in large and high-speed rotating machinery in electric power, petrochemical, metallurgy and other industries. At present, this technology has become the technical basis for modern equipment management and improving the comprehensive benefits of enterprises. Practice at home and abroad shows that equipment predictive maintenance based on vibration monitoring and fault diagnosis technology can save a lot of maintenance costs and achieve significant economic benefits. It can also ensure the safe operation of equipment, prevent and reduce the occurrence of malignant accidents, eliminate hidden troubles, ensure the safety of personnel and equipment, and improve productivity

traditional diagnosis methods and theories can play a good role in simple systems with single process, single fault and gradual fault, and have great limitations in multi process, multi fault and sudden fault, as well as large, complex and highly automated equipment and systems, such as steam turbine engine units. At present, typical mechatronic products such as CNC machine tools and AC servo drives are developing towards digitalization, miniaturization and high precision, which brings new challenges to monitoring and is expected to change the current situation of high dependence on imports of high-end caprolactam in China. Since fuzzy neural network control does not rely on control objects and mathematical models, it has strong robustness and is a nonlinear control method, It has good advantages in solving such problems. Expert system is mainly used in complex mechanical systems, which can overcome the over dependence of model-based fault diagnosis methods on models. The artificial neural network has unique advantages for fault pattern recognition. Applying the theory and method of artificial intelligence to mechanical fault diagnosis and developing intelligent mechanical fault diagnosis technology is a new way of mechanical fault diagnosis. Intelligent mechanical fault diagnosis expert system has been widely used and has become an important direction of mechanical fault diagnosis

2 application of artificial intelligence in mechanical fault diagnosis

artificial intelligence mainly studies the use of artificial methods and technologies to imitate, extend and expand human intelligence, so as to achieve machine intelligence. Traditionally, AI technology using mechanical fault diagnosis system can be divided into three categories: expert system (ES), artificial neural network (ANN) and fuzzy set theory (FST)

2.1 expert system BASF will also launch the following innovative automobile utilization: expert system. es is a practical discipline produced in the early 1960s. At present, it is one of the more active and successful fields in artificial intelligence technology. It is a computer software system composed of three main parts: knowledge base, inference engine and man-machine interface. In terms of knowledge expression, the use of production rules for knowledge expression is beneficial to the existing artificial intelligence language on the one hand, and on the other hand, its expression conforms to human psychological logic, which is convenient for knowledge acquisition and acceptance. The use of framework for knowledge expression has been more and more applied. In the aspect of diagnosis and reasoning, it is mainly manifested in the research on reasoning logic and reasoning model. In the field of artificial intelligence, there are many reasoning logic. Fuzzy reasoning logic is widely used in expert systems to reduce the complexity of the system and produce good results in mechanical system fault diagnosis. Its power lies in its expert knowledge and the reasoning mechanism of using knowledge to solve problems

based on the von Neumann computer architecture, the expert system gradually exposed the following problems in its development process: the bottleneck of knowledge acquisition, the narrow step of knowledge, the explosion of reasoning combination and infinite recursion, the low level of intelligence, the low level of system and the poor practicability

research and development of mechanical fault diagnosis expert system the emergence and gradual maturity of mechanical fault diagnosis expert system is one of the most remarkable achievements in the field of mechanical fault diagnosis. Because the scientific knowledge of mechanical fault diagnosis and maintenance often lags behind the practical and empirical knowledge of experts, it provides a broad application prospect for the expert system

2.2 artificial neural network (ANN)

artificial neural network is referred to as neural network for short. It is a complex network formed by a large number of simple processing units (called neurons). If it is found that the pressure is not enough, it is a simulation of biological neural system. Its information processing function is determined by the input and output characteristics (activation characteristics) of the units of the network and the topology of the network (connection mode of neurons). In order to make the system transparent, fuzzy rules are used in neural network reasoning, which makes it convenient to establish a good interpretation mechanism for artificial neural network

because neural network has the functions of fault tolerance in principle, structural topology robustness, association, speculation, memory, self adaptation, self-learning, parallel and complex mode processing, it plays a great role in the monitoring and diagnosis of a large number of multi fault, multi process, sudden fault, huge and complex machines and systems in engineering practice

system fault has hierarchy, correlation, delay and uncertainty, which makes the problem of equipment fault diagnosis very complex and difficult. Using a single sub neural network to solve the problem requires a large number of fault samples, and it is difficult to determine the network structure suitable for diagnosing multiple types of faults. Even if it is determined, it is easy to fall into local minimum, self-adaptive adjustment, improvement of error function and acceleration of convergence; The initial random weight is limited in order to overcome the local minimum problem. The applications of neural network in mechanical fault diagnosis are as follows: from the perspective of pattern recognition, neural network is used as a classifier for fault diagnosis; From the perspective of prediction, neural network is used as a dynamic prediction model for fault prediction; Fault diagnosis based on structure mapping is carried out by using the strong nonlinear dynamic tracking ability of neural network; From the point of view of knowledge processing, the diagnosis expert system based on neural network is established. At present, in order to improve the learning and diagnostic performance of neural network in practice, the research is mainly carried out from two aspects: the improvement of neural network model itself and the diagnostic strategy of modular model; At the same time, the combination of fuzzy logic and fuzzy logic is also a research hotspot

2.3 fuzzy sets theory (FSN)

researchers have been trying to find an effective way to deal with the incompleteness and uncertainty scientifically. Practice has proved that the fuzzy sets theory established by Zadeh in 1965 is a good method to deal with the uncertainty. When people's cognitive world contains a large number of uncertainties, it is necessary to fuzzify the obtained information to reduce the complexity of the problem. Fuzzy logic can be considered as an extension of multi valued logic, which can complete approximate reasoning that is difficult to achieve by traditional mathematical methods. An analog circuit fault diagnosis method based on fuzzy fusion of multi class electric quantity test information has been proposed. The construction method of component fault membership function based on K fault node diagnosis method and minimum standard deviation method, and the fuzzy information fusion diagnosis algorithm based on measurable point voltage and circuit gain at different test frequencies have also been described. Using these two kinds of test information, K fault

diagnosis method and minimum standard deviation method, the circuit is preliminarily diagnosed, and then the fused fault diagnosis results are obtained by using fuzzy transformation and fault location rules. Simulation results show that the proposed method greatly improves the accuracy of fault location in mechanical systems

3 development trend of artificial intelligence in mechanical fault diagnosis

the four main tools in artificial intelligence, namely expert system, artificial neural network and fuzzy set theory, have their own advantages and limitations

although ES has been widely used in many fields, there are still some problems, such as the bottleneck of knowledge acquisition, the difficulty of knowledge maintenance, the narrow application range, and the weak diagnostic ability. However, with the development and penetration of related disciplines and technologies, the theory and methods of expert system have also been greatly improved, and the above problems have been gradually alleviated or eliminated. We should pay attention to the combination with fuzzy logic, fault tree, machine learning and other methods

although ANN has strong self-organization, self-learning ability, high robustness and avoids the construction of inference engine, and the inference speed has no obvious relationship with the size, it soon attracted people's attention. And the application of neural network technology can make up for the problems encountered in the application of traditional expert system. However, there are still many limitations in fault diagnosis, such as:

(1) Ann extrapolation error is large, which is difficult to ensure the accuracy and fault tolerance of the solution

(2) if the system structure changes, it may be necessary to change the composition structure of ANN, or add new samples to re learn and acquire new knowledge

(3) Ann is difficult to realize logical reasoning based on structured knowledge

(4) lack of explanation ability, and the diagnostic results are not easy for operators to understand

in addition, how to ensure the rapidity of ANN training convergence and avoid falling into local minimum is also a problem that every ANN based diagnosis system must face. With the addition of FST, the corresponding intelligent diagnosis systems have more mature principle, more perfect technology and improved fault tolerance in the analysis of uncertain factors in mechanical system fault diagnosis. However, there are still maintainability problems, and the treatment of uncertain factors can only be limited

at present, there is a lack of a universal and effective method applied to various fields of mechanical systems. Hybrid intelligence, which integrates multiple intelligent technologies, has become one of the important development directions of AI. It is a new development trend to combine various intelligent technologies to design, control and monitor mechanical systems. The combination methods mainly include rule-based expert system and neural network, CBR and rule-based system and neural network, fuzzy logic, neural network and expert system. Among them, the diagnosis model which combines fuzzy logic, neural network and expert system is the most promising, and it is also one of the research hotspots in the field of artificial intelligence. For example, the combination mechanism of fuzzy logic and neural network, the combined algorithm, and the expression of fuzzy knowledge that is convenient for neural network processing. The application of hybrid intelligence in mechanical system fault diagnosis has the following development trends: from rule-based system to hybrid model system, from domain experts to machine learning, from non real-time diagnosis to real-time diagnosis, from single reasoning control to hybrid reasoning control strategy, etc<

Copyright © 2011 JIN SHI