The Realization Of The Monitoring System For Multi-mode Dynamic Recognition Of Axial-flow Compressors Stall Based On LabVIEW

The military and civilian aircraft often encounter flight safety accident. When the engine is running, the unstable disturbances of the air flow causes the rotating stall and surge pheno-menon, which will damage the engine performance, and even cause the flame of combustion chamber counter current, leading to the rejection of all the compressor blades. Normally, the rotating stall always occurs before the engine surge. So if we could predict the occurrence of rotating stall in advance and further prevent surge phenomenon, when the engine is running, the loss of human and material resources caused by the safety accident will be greatly reduced.In recent years, deterministic learning theory proposed by Wang et al is a feasible method to solve this problem. The main content of the theory is to identify the unknown system of dynamic environment, which is to obtain, express, store, and reuse the knowledge of the unknown system. To detect the problem of rotating stall, deterministic learning theory which uses RBF neural network identifies the normal mode and the stall mode of dynamic systems respectively, keeps the neural network weights in the form of constant value, and then builds the multi-mode library. In the detection process, deterministic learning theory will compare the unknown system dynamics with the normal mode and the stall mode in the multi pattern library, to calculate the residual and then conclude the similarity according to the judgment criteria. Then it can determine the current conditions of the axial flow compressor, to achieve the early warning of rotating stall.The first study content of this paper is the preliminary designation and realization of the rotating stall detection system based on deterministic learning and dynamic pattern recognition theory, which generally realizes on-line detection in advance. However, as the manner of selecting modes is different, the detection results is always very random. The paper expands the model library on the basis of the first stage work, and then designs a monitoring system of engine rotating stall based on the multi-mode, to express the characteristics of the compressor dynamic system more comprehensive. This phase of work includes the following several aspects: The optimization of the algorithm which is used in multi-mode recognition, including the designation of concurrent mode group scheme and the improvement of judging criterion algorithm, which ensure the accuracy and real-time performance of the system in the process of coping with more pattern; The designation and realization of the solenoid valve pressure relief and control scheme, through the LabVIEW serial communication. It is verified that, after expanding the model library, the redesigned detection systems can early warn accurately at given speed, based on the measurement data of BUAA compressor experimental platform. It provide sufficient time for the operator to control the solenoid valve, which has good practical application value and feasibility.