The Realization Of Rotating Stall Detection System For Axial-flow Compressor Based On Deterministic Learning Under The LabVIEW Platform

In recent years, research on rotating stall and surge of axial flow compressor gains moreand more attention. As one of the core components of turbofan engine, the flow instabilitywill make air compressor into unstable working condition such as rotating stall and surge inthe process of operation, and the unstable working condition can cause the engine flameout,air parking and other serious accidents.Rotating stall is considered as a precursor to surge,and effective monitoring and early warning for the compressor of starting stall will greatlyimprove the reliability and stability of the engine,also the engines our country use mostlycome from foreign countries. So it’s very important to pay more attention to the research oncompressor’s rotating stall and surge.Deterministic learning theory is put forward under the dynamic environment of the theoryof machine learning in recent years. It uses the concepts and methods of the field of adaptivecontrol and proposes a good method for the expression of dynamic model, similarity andquick identification. The realization of axial compressor’s rotating stall detection systembased on deterministic learning use RBF neural network to learn compressor’s data which isin stall condition, stored patterns in the form of constant weight and then compare the datathat need to be measured with patterns store in the library. According to the given strategy thatthe axial compressor’s state matches corresponding mode best when residual error is the least,the system will execute the right step such as alarm and display, thus achieving the goal ofdetection of rotating stall.This article will focus on the exploration and research of project application for axialcompressor rotating stall detection system, and the system is based on the existingdeterministic learning theory and dynamic pattern recognition algorithm. LabVIEW is anexcellent graphical programming language platform，with its advantage of powerful graphicdisplay and easy operation, this article will focus on the design work of the whole system.Until now, we have achieved the following main results:(1) Design the offline learningsystem for axial compressor’s rotating stall base on LabVIEW and ArrayFire, the system uses the GPU technology to optimize part of the algorithm, thus protects the relevant code andguarantee the consistency of the system.(2) Design the online detection system for axialcompressor’s rotating stall base on LabVIEW, it is also a key work of this paper. We improvedthe dynamic pattern recognition algorithm under the LabVIEW platform, and use themulti-core parallel computing technology to guarantee the system function.(3)Experiment onthe engine platform of the Beijing University of Aeronautics and Astronautics many times,and verify the reliability of the algorithm and the application value on engineering.