Realization Of Axial Compressor Instability Detection System Based On Deterministic Learning And Permutation Entropy

With the development of axial compressors toward higher bypass ratio and fewer stages,the load on their blades is increasing,and the anti-interference ability decreases.The non-uniform airflow（inlet distortion）of the compressor inlet channel has become one of the important factors that induce the instability of the internal airflow,which severely limits the performance of the axial compressor.Therefore,it is necessary to study the compressor instability detection method,detect the instability signal at the initial stage,and further select an appropriate controller to avoid the development of instability.This paper is based on the deterministic learning theory and information entropy theory to study the nonlinear feature extraction and rapid detection of the instability data,and develops a compressor instability nonlinear feature extraction and detection system on the LabVIEW platform.The specific work is as follows:（1）Based on deterministic learning theory and information entropy theory,extract the non-linear characteristics of the compressor flow data,in other words,obtain its mathematical characteristics under different operating conditions:the normal operation stage,the instability precursor stage and the complete instability stage.In the deterministic learning theory,the mathematical features are expressed as the weights of the constant RBF neural network and stored in the offline pattern library;in the information entropy theory,they are expressed in the form of entropy value.（2）On the basis of extracting non-linear features,the l₁ norm is selected to measure the similarity between the dynamic trajectory of each model in the model library and the state trajectory of the compressor system,and the model with the highest similarity is required to last for a certain duration.So as to determine whether the compressor has entered a state of instability or complete instability.At the same time,the method based on information entropy combined with the sliding window design can realize the detection of the complete instability of the compressor.The results show that both the deterministic learning theory and the information entropy theory can accurately identify the operating state of the compressor.（3）Design a compressor instability detection system based on LabVIEW,and combine the GPU Analysis Toolkit to optimize the calculation process.The parallel calculation originally simulated by the CPU is transferred to the GPU,which overcomes the inability of the CPU to perform large-scale parallel calculations.The results show that the compressor instability detection system can still maintain a high parallel computing efficiency even with a large-scale expansion of the model library.The realization of the detection of the compressor system’s instability has important practical significance and engineering application value for improving the safety and stability of the compressor and prolonging the service life of the engine.