Review on condition monitoring and fault diagnosis

Review of FMS condition monitoring and fault diagnosis

on the basis of a large number of literature, this paper analyzes the necessity, difficulties and needs of FMS condition monitoring and fault diagnosis, describes the development status, achievements and research characteristics at home and abroad, evaluates and summarizes if the rotational resistance drops by 30%, and points out the main problems at the same time. The future development trend of multi-sensor, multi parameter and multi model comprehensive decision-making system with self-learning training and self-adaptive adjustment function, which is universal, modular, integrated and intelligent, is put forward, as well as research suggestions

keywords FMS processing equipment processing process status monitoring fault diagnosis

1 the necessity and difficulties of flexible processing equipment status monitoring and fault diagnosis

status monitoring and fault diagnosis is one of the key and bottleneck technologies that must be solved in all links and levels of FMS, CIM, intelligent manufacturing and other advanced manufacturing technologies. Condition monitoring and fault diagnosis technology is a common technology in many aspects, such as production process and control, process planning, product quality assurance and control, equipment performance, maintenance and life management. It is of great significance to improve the reliability of weapon equipment and advanced equipment manufacturing system, prolong service life, shorten development cycle, reduce costs, and strengthen the foundation of manufacturing industry and national defense science and technology industry

the high automation, flexibility and intelligence of manufacturing system, and the increasingly complex equipment and functions put forward higher requirements for condition monitoring and fault diagnosis technology. For the processing equipment and processing process, the main performances are as follows: ① the system equipment is diverse and complex, the processing takes flexible multitasking as the goal, and the processing types, processes and working conditions are diverse, so it is difficult to comprehensively collect a priori samples and mode samples of various normal and abnormal states; ② The discontinuity, abruptness, fuzziness, relevance and timely variability of process state and fault; ③ Dynamic linkage and discreteness among the components of processing equipment; ④ Random interference factors such as workpiece size and even misoperation have great influence; ⑤ The amount of information in the processing process is large and complex, and the information resources suitable for monitoring, diagnosis and early warning need to be mined; ⑥ Installation restrictions of monitoring system and coordination requirements between monitoring system and manufacturing unit; ⑦ The level and degree of monitoring and diagnosis are subject to the automation and intelligence level of the processing equipment itself. These must be paid attention to and solved in the research. Therefore, how to quickly and effectively identify the state in the changeable and complex working conditions and processes, how to carry out the state monitoring and fault diagnosis of dynamic, correlated and discrete equipment, how to reasonably eliminate the influence of the timely variability of changing working conditions on the monitoring and decision-making, and how to identify, classify and make decisions when the prior samples and fault categories of States and faults cannot be completely obtained, How to simply realize the monitoring and diagnosis in the actual processing field environment is the technical key that must be broken through in the research of processing equipment and processing process condition monitoring and fault diagnosis technology

2 review of domestic and foreign literature

the research on FMS condition monitoring and fault diagnosis technology is being carried out in depth in two aspects: first, the conceptual design, architecture, functional composition, management and decision-making mode, information integration and transmission and other basic, conceptual and systematic problems of intelligent condition monitoring and fault diagnosis system are studied; The second is the specific and in-depth research on various monitoring and diagnosis methods, as well as the establishment of practical monitoring and diagnosis systems with different functions

2.1 main contents of FMS condition monitoring and fault diagnosis technology

throughout a large number of literatures, the main research contents of FMS condition monitoring and fault diagnosis can be summarized into the layered structure shown in Figure 1 according to the division of objects. Each module includes not only the composition, design and implementation of the system, but also the realization of each functional module and the specific method research

Figure 1 main research contents of condition monitoring and fault diagnosis

2.2 common methods of FMS condition monitoring and fault diagnosis theory and technology

(1) various single function monitoring and diagnosis methods, such as acoustic emission, force or torque monitoring, current and power monitoring, vibration and spectrum characteristic analysis, time sequence or time domain characteristic analysis and other functional parameters and threshold comparison and inspection. This is only partially applicable to flexible large-scale systems

(2) the application of training models such as pattern recognition and neural network and their combination with other methods. It is considered to be one of the models with special advantages in classification and recognition. It is limited by the orthogonality and completeness of the training sample data, as well as the consistency of the training sample and the detection sample format. In the application of monitoring and diagnosis, it mainly aims at the anomalies and faults of a small range or a few fixed processes

(3) the application of fuzzy mathematics and decision-making methods includes fuzzy pattern recognition, fuzzy logic reasoning, fuzzy comprehensive decision-making, and the combination of fuzzy theory with artificial neural network, expert system and other methods. The key is to determine the fuzzy relationship. This method is suitable for both specific objects and complex large-scale systems

(4) the combination of artificial intelligence and expert system method and graph theory, fuzzy theory and artificial neural network is a research hotspot in recent years, but for state monitoring and fault diagnosis, qualitative reasoning is mainly used more. In recent years, the fusion model combining symbol and value is being favored, such as the fusion expert system combining expert system and neural network. The application of intelligent models with self-learning, self-organization and self adaptation in monitoring and diagnosis is a hot topic, such as intelligent information fusion theory

(5) model-based monitoring and diagnosis method and its combination with neural network and fuzzy theory. Model-based state monitoring and fault diagnosis is an effective method with good adaptability and robustness [1, 2]. Model-based state monitoring and fault diagnosis are widely used in the fault monitoring and diagnosis of process control. It can be divided into process dynamics model and process filtering model. At present, the latter is the main application, while the application of dynamic model is less. The main reason is that the dynamic model of processing equipment has not been well established, and the specific physical parameters that constitute the model parameters are not clear, so it cannot be further diagnosed. In recent years, the penetration of intelligent theories such as artificial neural network and fuzzy theory has made this field more active

(6) the application of new information processing methods and new decision-making methods the application of new information processing methods (such as wavelet analysis, time-frequency analysis, high-order spectrum, high-order time-frequency, fractal and other non-linear information processing, etc.) for specific objects such as mechanical equipment and processing processes, has achieved a good effect of feature extraction of state description and classification. The application of new decision-making methods and frameworks (such as multi-sensor fusion technology) in condition monitoring and fault diagnosis of processing equipment and processes has achieved gratifying results

2.3 typical achievements in FMS system and processing equipment status monitoring and fault diagnosis

Huazhong University of technology has carried out research on status monitoring and fault diagnosis earlier in China. Under the leadership of academician Yang Shuzi, it has carried out extensive research on many fields and objects involved in mechanical system, especially in machine tool cutting status monitoring and fault diagnosis based on time series model, fault diagnosis system architecture It has made great achievements in the research of monitoring and diagnosis models and new methods in intelligent manufacturing and control

Xiao Xiangsheng and Ma Yulin from Harbin Institute of technology have made some achievements in the field of FMS condition monitoring and fault diagnosis during the Eighth Five Year Plan Period: ① establish a detection and monitoring architecture and software system prototype suitable for FMS; ② Develop the prototype of intelligent data acquisition device and its software prototype; ③ Break through the difficulty of extracting AE signal when AE technology is applied in machining center, etc

for the FMS flexible processing equipment of Changchun 55 Institute, the Key Laboratory of national defense science and technology, Beijing University of technology proposed the overall framework and specific implementation strategy of intelligent condition monitoring and fault diagnosis for the whole line; Nanjing University of technology proposed the framework and implementation strategy of FMS fault diagnosis expert system, as well as the scheme of neural network based real-time monitoring system

since the University of national defense science and technology began to undertake a number of national 863 High Technology Program projects, national defense pre research and national defense pre research fund and other topics in 1986, it has carried out in-depth and systematic research in many aspects, such as the architecture and composition of condition monitoring and fault diagnosis and fault early warning system, state identification and decision-making under the background of variable working conditions, fault diagnosis expert system and so on, and has made gratifying achievements

abroad, the idea and scheme of "universal flexible monitoring system" proposed by German weck [3] in 1986 has typical significance. It uses various conventional and special characteristic parameters to judge the quality of the processing process, NC program execution and other system operation processes, and takes pattern recognition technology as the core to realize the detection and prediction of collision, drilling tool wear and tear, etc. Tonshoff [4] proposed the future trends, namely multi-sensor synthesis, multi model system, learning and decision system, rule-based system. Monostori [5] proposed a knowledge-based hierarchical state monitoring and fault diagnosis model for machine tools and manufacturing units in 1993. In 1995, Du et al. [6] deeply analyzed the methods of manufacturing process and equipment monitoring, and proposed a multiprocessor structure monitoring system scheme based on global data base and knowledge base

in terms of review and summary literature, Matrin [7] of the UK summarized the current status monitoring and fault diagnosis of processing equipment and processes in Europe, Japan and the United States; Lee Jay [8] summarized and commented on modern computer-aided maintenance in manufacturing equipment and systems

2.4 typical achievements in machining process and tool cutting state monitoring and diagnosis

in China, Wanjun [9] of Tsinghua University commented on cutting state monitoring, gave a list of state signals and characteristic parameters of direct and indirect methods of tool state detection, and made a specific analysis. Xi'an Jiaotong University has made many beneficial discussions in this regard, and put forward many monitoring models, such as turning component ratio monitoring, frequency band coherent function method and so on. Huazhong University of technology has done a lot of pioneering work in AR model monitoring of cutting state, monitoring and control of cutting chatter and so on. Huang Ren and Zhong Binglin of Southeast University have made in-depth research on tool life management, grinding burn identification, turning condition monitoring and other aspects using pattern recognition theory. Yaoyingxue and yuanzhejun of Harbin Institute of technology proposed multi parameter, multi model monitoring system and acoustic emission monitoring system for drilling process and cutting state monitoring. Shanghai Jiaotong University has also carried out in-depth research on acoustic emission tool cutting condition monitoring

abroad, in 1983, tlysty [10] made an authoritative summary of the sensor systems used to monitor the machining process of unmanned machining systems, mainly involving shape and position sensors, cutting force sensors, spindle and feed shaft force sensors, spindle motor sensors and acoustic emission sensors, and gave their dynamic ranges and applications. In 1983 and 1986, bertok et al. Of Japan proposed the relationship model identification between torque monitoring and geometric dimensions to realize machining process monitoring, and established an AR model based on cutting torque signal to obtain residuals and realize tool damage monitoring. Emel et al. Used AE signals for pattern recognition to realize the monitoring of turning tool wear damage in 1988, elbestawi et al. Used cutting force, torque, spindle vibration signals and other multi parameter input pattern recognition models to realize milling in 1989