Research Based On Parallel Computing Turbofan Engine Monitoring System For Rotating Stall

Rotating stall and surge has been a complex and challenging issues in research areas ofthe engine. The deterministic learning algorithm is a new algorithm which used thecharacteristics of RBF neural network and pattern recognition. It is applied to systems andcontrol field in a dynamic environment is a new theory of machine learning. It puts forward avery useful way based on the expression of the dynamic model, similarity, rapid identification,classification and intelligent control. It reflects a dynamic distributed computing informationprocessing in the dynamic pattern recognition, overcomes the defects of repeated learning,poor capacity of local approximation and improve the efficient in the traditional algorithm.Traditional pattern recognition algorithms require large-scale and very long calculation, solooking for some ways to reduce the computation time of pattern recognition is a very urgenttask.The emergence and growing popularity of multi-core processors, the maturing of theparallel programming language and the mechanism to solve the multi-threaded parallelalgorithms running on multi-core processors, provide a good foundation conditions for sharedmemory multi-core programming. OpenMP is an industry standard of the shared memorysystem programming and has many advantages such as simplicity, much better portability andscalability. So most users prefers to use OpenMP to improve computation efficiency ofalgorithmsIn this paper we proposed a parallel multi-modal identification algorithm based ondeterministic learning, the purpose of this algorithm is to improve the computing speed ofmulti-pattern recognition and prediction. After analyzing the algorithm, we use parallelcomputing design patterns and the OpenMP parallel programming interface to achievemulti-threaded parallel identification of large amounts of data model calculation in theenvironment of stand-alone multi-core platforms.,and given the graphical interface of thevisualization. The corresponding experimental results show that parallel deterministiclearning algorithm greatly accelerating the speed of the dynamic pattern recognition in thecase of full use of multi-core resources, and the visual interface to ensure real-time display ofthe identification results.