Research On Theory And Method Of Error Tracing And Accuracy-Loss Diagnosis Of Dynamic Measurement System

The rapid development of science and technology constantly injects new blood to the measurement technology. At the same time, it also advances new and higher demands to accelerate the developing speed of the measurement technology. Dynamic measurement has gradually been the main trends in modern measurement. However, because the measuring process is influenced by many factors, it is difficult to improve the accuracy of dynamic measurement. It is the focus of the measurers and the designers all through the ages. Thus, the research on dynamic error and accuracy theory is always heavily concerned. In order to improve the accuracy of dynamic measurement effectively, one must know each unit within the system and master their own dynamic characteristics and the transmission characters between them, thus to know the sources of the measurement errors. Therefore, according to the project "Research on the Theory and Technology Basis of the Error Tracing and Accuracy Diagnosis of Dynamic Measurement" (sponsored by Doctor's degree Foundation of the University of China) and the project "Research on the Theory of Dynamic Measuring Accuracy" (sponsored by National Natural Science Foundation), a new theory is initiated in this dissertation. It is called the theory of error tracing and accuracy-loss diagnosis of dynamic measurement system. Here the research is performed systematically and deeply from the theory and the practice.Theoretically, on the basis of the analyses of the errors of typical units in measurement systems, the concept of the separating and tracing of the output errors is advanced. Its methods are discussed and compared. The analysis method of generalized wavelet-neural network (WNN) is applied. Then, the theory and concept of accuracy-loss are advanced. In order to prove above two theories, two practical measuring systems are selected for the experiments. Firstly, the transfer-chain functions and error whitening models are established. Secondly, according to these models, the measuring errors of the systems are decomposed into the component errors and traced to their sources, i.e. the units within the measurement system. Lastly, accuracy-loss diagnosis of each system is performed based on the decomposing and tracing results of the output errors. The units are diagnosed, which are the main factors of the accuracy loss of the whole system. It can provide the base of the optimization design for the system designers.The main originalities of this dissertation are as follows:The dynamic error models of the typical units of measurement systems are discussed. On this basis, the theory and method are developed of the decomposition and tracing of dynamic measurement errors. According to the transfer-chain function and error"whitening model" or "quasi-whitening model" of the dynamic measuring system,the output errors are decomposed into the component ones of the inner parts and the outer noises. Then the error tracing is realized. The selection of the decomposition method is also a key problem. After the comparison of the merits and faults between Fourier spectrum analysis method and wavelet-neural network analysis method,the latter is selected.After the research on the principle of accuracy-loss diagnosis of the dynamic measurement,the diagnosis theory and method are advanced. According to the error model established of the dynamic system and the results of the error tracing,using BP neural network,the accuracy-loss law of each inner unit is acknowledged. Then,its time-variant model is obtained. Thus,the inner parts are found which are the main factors of the total accuracy loss. This result can offer the basis to improve the design of dynamic measurement system and to increase its measuring accuracy.In order to prove the theory of the error tracing and accuracy-loss diagnosis of the dynamic measurement,two simple measuring system are selected,that is,pressure transducer system and the dynamic measuring system based on differential mutual inductive displacement transducer. Firstly,the "quasi-whiten...