Integration Of Fault Diagnosis Method And System Design On Integrated Navigation System

With the development of science and technology, the integrated navigation system has become the necessary navigation system on ship. Nowadays, many important discoveries in science and the emergence of new technology have promoted the development of the integrated navigation system greatly. Summarily speaking, the integrated navigation system is developing towards miniaturization, multi-functional, multi-mode, high precision, and high reliability direction. The appearance of the fault diagnosis technique offers a new approach which can improve the reliability and navigation precision of the integrated navigation system.Although the existing fault diagnosis approaches have been made many achievements in the actual application, there are still many problems: expert system diagnoses faults based on the domain knowledge and experience of expert, but the application of this method need to obtain experts'domain knowledge. This is also recognized as the "bottleneck" problems in the expert system research and development, which makes the application of expert system restricted. Fuzzy fault diagnosis methods solve the uncertain relationship between fault and sign by using fuzzy theory, and can give the probability of fault occurrence. The conclusion is clear and intuitive. The subordinate function is artificial structure and it contains some subjective factors. If function is structured unreasonably, it can lead to diagnosis accuracy decline. The neural network has powerful classification ability, which can classify failure modes and study, then diagnose fault, but it has also slow training speed and it is easy to fall into the local optimal value. In order to overcome the limitations of single fault diagnosis method, hybrid intelligent diagnostic technique integrated by various diagnosis methods is the study hotspot in fault diagnosis fields.Taking the integrated navigation system as the research object, this paper firstly discuss the development situation of the fault diagnosis technique and the main research contents. Then the specific methods of fault diagnosis were given. After that, the integrated navigation system's composition principle, main navigation system's working principle and error index were illuminated briefly.The basic principle of the immune algorithm, the detailed process and the steps of network study training are illuminated. Then the parameter selection rules of BP network were explained. After that the immune algorithm was introduced into the neural network. To cope with the problems of the neural network's structure, each layer's activation function and the training methods is difficult to determine. This will result that the neural network is not optimal. Then the fault diagnosis method based on immune BP network was utilized. This method uses immune algorithm to code and immune operation for the network structure and the training methods, and obtain optimal or times optimal solution. The wavelet packet algorithm was utilized to decompose gyroscope signal into three layers, and then extract node feature coefficients at the bottom. Then the eigenvector was constructed as the input of immune BP network, training and testing network. At last, the simulation results show that this method for fault diagnosis has good effect.The integrated navigation system's state equation and observed equation were established. And the learning and training principle of RBFNN are given, and then the specific design steps. By analyzed the shortcomings of the neural network which were easy to fall into the local optimal and slow convergent speed, a method which utilizes the genetic algorithm to optimize neural network parameters was bring forward. This method was based on the analysis of genetic algorithms genetic principles and the main factors. At last, the integrated navigation system failure diagnosis was regard as the object and the effects were validated by simulation.To cope with the potential misdiagnosis or capsulorrhaphy shortcomings of single fault diagnosis methods, the integrate diagnosis system of SINS/GPS/DVL integrated navigation system was built based on D-S evidence theory, including BP network, wavelet neural network and genetic RBF networks. The system can make decisions fusion for diagnosis result, and get the diagnosis. It can improve the reliability and accuracy of the fault diagnosis, and meanwhile overcome defects and deficiencies of single fault diagnosis method.