Wind Tunnel Process Mach Number Prediction Based on Modal, Stage, and Intra-Stages Three-Layer Partitioning

The wind tunnel experiment process is a nonlinear process with complex process characteristics. It is the primary task to master the key physical parameters and performance evaluation criteria during its operation. Aiming at the characteristics of multi-mode, multi-stage and intra-stage changes in the wind tunnel process, this paper proposes a Mach number prediction method based on mode, stage and intra-stage division. Firstly, mode division is carried out. The K-means clustering method is mainly used to cluster process data. The elbow rule is used to determine the cluster number K. The Mach number is used as the index variable to divide the process into phases, and divide the phases into stable parts and transitional parts according to different process characteristics. Considering the nonlinearity of the data, a kernel partial least squares prediction model is constructed for the stable process. Considering the dynamic characteristics of data, a dynamic partial least squares prediction model is constructed for the transitional process. The proposed method has been applied to multi-stage nonlinear wind tunnel experiments, and satisfactory results have been obtained.