Research On Intelligent Fault Diagnosis Techniques For Radar System

In the present, the deficiency and need of fault diagnosis abilities for the radar system on field operations, so it starves for study the radar fault diagnosis to improve the level of maintenance support and efficiency of the radar system. In this paper, the radar fault diagnosis research is developed according to three levels with system engineer method, which are radar system-level fault diagnosis, radar circuit board-level fault diagnosis and radar circuit component-level fault diagnosis, specially three intelligent fault diagnosis approaches are discussed for the radar circuit component-level fault diagnosis. The most distinctive parts can be described in the following aspects:1) To detecting and diagnosis for radar system on field operations, a radar system-level fault diagnosis method is presented. A radar system failures diagnostic fault tree model is established for radar system expert diagnosis system.2) For on-line testing and diagnosis of radar circuit board on field operations, a board-level fault diagnosis approach of the radar is proposed. Radar signals are used to replace signal generator that is fit for outfield radar circuit board-level fault diagnosis, input and output signals for sampled signals are used spectrum analysis to judge the relativities between the nomal signals, then to determine whether the corresponding circuit board is faulty. The fault diagnosis experiment is conducted in a scan circuits of certain a type of radar. The results show the validity of the proposed fault diagnosis method.3) Based on the proposed radar system-level and circuit board-level fault diagnosis methods, a scheme and its soft structure of the radar fault diagnosis system is advanced, the software function module and the implement principle of the fault diagnosis system are established.4) Aiming to improve the fault diagnosis ratio of radar circuit component-level, a wavelet neural network method for fault diagnosis is presented. Firstly, output voltage signals under faulty conditions are obtained with sampling. Then wavelet coefficients of output voltage signals are gained by wavelet lifting transform, and faulty feature vectors are extracted from coefficients with wavelet lifting decomposition. After training the networks by faulty feature vectors, the wavelet neural networks model of the circuit fault diagnosis system is built with Back Propagation (BP) algorithm. The simulation results of the high power circuit of the radar show the fault diagnosis method of the radar component-level circuits with wavelet neural network is effective.5) For complexity of network structure and samples influencing the performance of wavelet neural network, a systematic approach for fault diagnosis of radar circuit component-level based on support vector machines is presented. Firstly, sampling voltage signals under faulty conditions are obtained from analog circuits test points. Then wavelet coefficients of output voltage signals are gained by wavelet lifting decomposition, and faulty feature vectors are extracted from the coefficients. Bayesian evidence framework and quantum-inspired evolutionary algorithm are used to select the parameters of the least squares wavelet support vector machines model. After training the least squares wavelet support vector machines by faulty feature vectors, the least squares wavelet support vector machines model of the radar circuit fault diagnosis system is built. The simulation results of the radar scan circuit show the fault diagnosis method of the radar circuit component-level with wavelet lifting transform and least.squares wavelet support vector machines is effective.6) Because of the difficulty of disposing of fuzzy inference rules for fault diagnosis, a systematic approach for fault diagnosis of radar circuit component-level based on optimized fuzzy inference is presented. The fuzzy logic system for based on fuzzy inference radar circuit component-level fault diagnosis, quantum-inspired evolutionary algorithm (QEA) is used to optimize the membership function of the rules of the fault diagnosis fuzzy logic system, then self adapted genetic algorithm to select the optimum fuzzy rule aggregate, so the number of fuzzy rules is decreased to make it easy to dispose fault diagnosis of radar circuit component-level fault diagnosis. The simulation results of the radar power magnifier circuit show the fault diagnosis method of the radar circuit component-leve! with optimized fuzzy inference is effective.