Fault Detection And Diagnosis Of Liquid Rocket Engine Based On Genetic Algorithm

On the basis of genetic algorithm, fault detection and fault diagnosis technique for liquid rocket engine were studied on a certain type of liquid-propellant rocket engine. A new genetic neural network was proposed, based on which system identification and pattern recognition were used to analyze engine fault detection. A great deal of analysis were obtained combining with test data. Fuzzy C-means genetic technology was put forward to, which was applied to engine fault diagnosis. Simulation results demonstrated this method was better in real-time and accuracy. Quantum genetic algorithm was firstly applied in liquid rocket engine fault detection, which was helpful for quantum genetic algorithm used for liquid rocket engine monitoring processes. Decision logic for fault detection and fault diagnosis were studied and a method was put forward to determine the threshold value.The main contents of the thesis include:1. The key technical problems of liquid rocket engine fault detection and fault diagnosis were transformed into corresponding optimization problem in which genetic algorithm and quantum genetic algorithm could be used to solve them.2. Other new algorithms based on genetic algorithm were studied. (1) A new genetic neural network which came from the deep cross-meeting of a genetic algorithm and BP neural network was put forward to. According to the performance comparison we found the algorithm was better than the BP algorithm to avoid local optimal solution, and better than the GA in the time and speed of searching for optimal solutions because of GA working similar to the form of exhaustive. On the other hand, the algorithm avoided relying on the experience and trial to determine the value of the threshold and would do a full, comprehensive and accurate expression of the fault diagnosis "knowledge" to enhance nonlinear neural network mapping. (2) A genetic fuzzy C-means algorithm was put forward to which could make clustering algorithm have a better search results, comparing with the traditional fuzzy C-means algorithm the algorithm had less classification error rate and the better value of the objective function. Through the performance comparison, the genetic Fuzzy C-means algorithm had a better effect in mode classification.3. The fault detection algorithm based on genetic neural networks nonlinear identification and pattern recognition techniques in liquid rocket engine start and steady process was realized. How to select the monitoring parameters was discussed, it was proved that this algorithm basically monitored the normal test data and accurately predicted abnormal test data third times in the test data analysis of a certain type of rocket engine. The simulation results demonstrated that the real-time ability and the response time of the detection algorithm met the requirements.4. The fault diagnosis algorithm based on genetic fuzzy C-means algorithm in liquid rocket engine steady process was realized. It proved genetic fuzzy C-means valid in classification problem and tentatively this method was used in liquid rocket engine fault diagnosis. How to select the monitoring parameters was discussed, it was proved that this algorithm normally monitored the normal test data and accurately diagnosed engine failures twice by simulation analysis.5. Quantum genetic algorithm was firstly applied in liquid rocket engine fault detection. Nonlinear regression algorithm based on quantum genetic algorithm was put forward to and applied in liquid rocket engine fault detection. It was proved simple, fast operation and practical by simulation analysis.6. Decision logic for the fault detection and diagnosis suggested in this thesis was studied and a method was put forward to determine the threshold value.