A fan design second-generation algorithm multi-objective optimization method based on variable learning rate network modeling

Abstract

A fan design second-generation algorithm multi-objective optimization method based on variable learning rate network modeling comprises the following steps that 1, the wind pressure and the wind volume are given values, the efficiency and the cost are target variables, and data samples of the structure variables and the target variables are obtained through experiments; 2, taking the structure variable as an input variable and the target variable as an output variable, training the data sample to complete the establishment of a variable learning rate network model, and updating the weight andthe threshold value by adopting a variable learning rate method; Step 3, establishing a second-generation algorithm model, wherein a non-dominated sorting operator and an elitist strategy design operator are adopted; And 4, predicting the energy consumption and the cost of the fan through the established variable learning rate network, and using the predicted value to solve an objective function value in the second generation genetic algorithm model to obtain a pareto leading edge; And finally, applying the structure variable value subjected to the reverse normalization to the actual design ofthe fan. The method is comprehensive in target and high in precision.

Description

technical field [0001] The invention belongs to the field of fan operating parameter design technology and industrial process simulation, and relates to a multi-objective optimization method of fan design second-generation algorithm for network modeling with variable learning rate. Background technique [0002] The fan belongs to fluid machinery, and its function is to compress gas and transport gas, which belongs to fluid machinery. The operation process of the fan is a turbulent fluid flow process and a complex energy transfer process. [0003] The change of each structural variable of the fan will have a comprehensive effect on the fan. That is to say, when the structural variables change, the changing trends of the target variables of the fan (generally efficiency and cost) are not consistent. And we are not only optimizing efficiency, but also optimizing cost. Therefore, we need to obtain a set of optimal design variable combinations under the condition of satisfying...

AI Technical Summary

Problems solved by technology

[0004] The traditional multi-objective calculation is actually a weighted calculation of the single-objective calculation. The value of the weight is closely related to the experience of the staff, and it is difficult to achieve accurate and effect...

Images