Implementation Of Cluster System Based On LabVIEW In Dynamic Pattern Recognition

Under the disturbance of air flow,rotating stall and surge usually occur in the air engine.The surge will make accelerated fatigue and expand the cracks of the parts in the engine,which will not only influence the performance of the engine,but also cause the engine unable to be repaired,even explosion.In general,the rotating stall occurs before the surge.Thus,it will be greatly significant for the safety of the airplane,if the signal can be given when the rotating stall occurring to warn the surge.A new learning mechanism is proposed for the control in a dynamic environment called deterministic learning theory.The theory is applied in the RBF neural network,adaptive control,dynamic systems and other fields to acquire,express,store and reuse the knowledge of the unknown systems.Based on deterministic learning theory,Wang Cong et al propose dynamic patterns recognition method for engine rotating stall detection.In this research,RBF neural network is used to identify normal modes and stall modes of the dynamic system,and the weights of neural network are preserved in the form of constant value.The patterns library will be constructed after modes identification.By comparing the unknown system dynamics with the pattern library,the residuals are obtained.According to the criterion of minimum residual error,the current running state of axial compressor can be judged.In the engineering realization,the rotating stall detection system based on deterministic learning theory and dynamic pattern recognition method involves a large number of computational problems.The computations include neural network internal calculations and residual calculations.The computation of the system increases greatly with the increase of the mode in the pattern library,which leads to the long detection time.Therefore,the parallel computing method of single node cannot meet the demand of the system.In order to solve the problem of accelerating,this paper proposes a multi-node SMP cluster detection system based on LabVIEW,which combines the characteristics of message passing and shared memory.Between the nodes,the parallel computing of message passing is realized by the LabVIEW communication technology.In the node,shared memory parallel computing is realized by the Lab VIEW multi-core programming environment.The results show that compared with the single parallel algorithm,the system can greatly reduce the computation time,the acceleration effect is remarkable,and the performance is stable.The system realizes the fast performance of dynamic pattern recognition,and it is valuable for engineering application.