Structural parameter identification method based on radial basis function network and multi-objective optimization

Determining the optimal structural parameters is the key to improving the performance of mechanical equipment and breaking through operational bottlenecks. It is significant to optimizing the efficiency of full-cycle operation and maintenance. To identify the optimal structural parameters, the study takes the centrifugal pump machinery as the research object and designs an improved Latin hypercube sampling experimental design method to lay the foundation for constructing the fitting model. A surrogate model based on an improved radial basis function network is constructed. The fitting accuracy is improved through adding multiple points, combining basis functions and building a final model based on a low-precision model, and the final fitting objective function is output. An improved slime mold optimization strategy is designed to solve the objective function. The maximum optimal solution values of the improved locust optimization algorithm and the improved slime mold optimization algorithm on the Sphere test function were 4.74 × 10 –1 and 1.82 × 10 –3 , respectively, which were closer to the theoretical optimal value of 0. Under the improved Latin hypercube sampling experimental design method, the distribution of sample points was more uniform. Under the Sixhump test function, the relative average absolute error of the improved radial basis function network fitting model was 0.016, which was significantly lower than the 1.214, 0.917, 0.703, and 0.484 of the comparison model. After optimizing the centrifugal pump, the hydraulic efficiency increased by 8.3%, the head increased by 10.8m, and the NPSH decreased by 1.2m. The method can accurately identify the optimal structural parameters of centrifugal pumps, provide a reference for identifying the optimal structural parameters of similar rotating machinery, and help equipment in related fields operate with high efficiency and low consumption.