The invention discloses an aircraft fuel
centrifugal pump impeller parameterization design method, and relates to the technical field of aircraft fuel
centrifugal pump impeller parameterization design, and the aircraft fuel
centrifugal pump impeller parameterization design method comprises the steps that a
point cloud conversion terminal carries out Poisson sampling on the surface of each impeller model, and a
point cloud data set containing coordinates, normal vectors and curvatures is generated; in a physical constraint
neural network architecture construction stage, a neural network terminal inputs working condition vectors and performance requirements into a PointNet + +
encoder to extract global features, then geometric
point cloud coordinates are generated by using Transform in combination with MLP, and eddy dynamics physical constraints including a vorticity intensity suppression item and a vorticity gradient constraint item are applied in training. According to the method, the multi-objective
loss function combination is designed through the training optimization terminal, the point cloud
geometric error, the fluid equation residual error and the
performance index deviation are fused, multi-dimensional collaborative optimization of the geometric accuracy, the fluid
mechanics compliance, the performance standard degree and the
structural reliability of the impeller is achieved, and finally the
aviation fuel centrifugal pump impeller with better comprehensive performance is obtained.