Optimization Design And Research Of Control Valve’s Core Surface Based On Improved BP Neural Network

The spool is the most important part of the control valve to control the pipeline flow.The reasonable design of the spool shape determines the quality and working efficiency of the entire regulating valve.At present,most of the domestic control valve designs still use the flow test method combined with repeated repairs,it has a long development cycle,low efficiency,and high costs.what’s more,it can easily lead to system control ability to be inaccurate and too large energy energy consumption.Which will be too difficult to meet the high-performance adjustment functional requirements of the valve.In view of this,this text takes a certain type of compact single-seated control valve as the research object,proposes an improved BP neural network method based on genetic algorithm,BP neural network and orthogonal test to design and optimize the structure of valve core surface.At the same time,the flow field characteristics of the optimized regulating valve were analyzed and studied.The main work and conclusions are as follows:First of all,using Solidworks and Fluent to establish the internal flow channel structure of the control valve and performing flow field simulation,the internal flow field pressure,velocity values and distributions are obtained.The flow characteristics and flow resistance characteristics of the control valve were analyzed to establish a mathematical model of the valve core design and determine the design parameters.Selecting the valve core surface design parameters for input samples of the network,and adjusting the valve performance parameters for the network output samples,the use of orthogonal test method for random combination of input parameters.Repeatedly modifying the structural dimensions of the valve core profile,establishing a model and performing flow field analysis,and finally completing neural network training sample set acquisition based on the data obtained from the analysis.Secondly,since BP neural network has the inherent defect of being easy to fall into local minima,the genetic algorithm can be globally optimized in the solution set space,an improvement based on the combination of genetic algorithm,BP neural network and orthogonal experiment is proposed.BP neural network method is used tooptimize the design of valve core surface structure.Using the selected sample set,trained the predictive model of the performance of the regulating valve based on the improved neural network.Finally,the improved neural network model was used to obtain the regulating valve.Non-linear mapping relationship between valve core surface structure parameters and performance.Finally,these four main design parameters of the valve core profile,design variables,the flow coefficient and flow resistance coefficient are selected as objective functions,the valve core profile boundary parameters and leakage amount as constraints,and an optimization mathematical model is established in which the nonlinear relationship of the objective function is given in the improved BP neural network model of the valve core surface.The established mathematical model is solved using the fmincon function to determine the optimal design parameters of the valve core surface optimization and the optimized valve core is obtained.Profile;contrast analysis of the flow characteristics and flow resistance characteristics of the regulating valve before and after optimization to verify the rationality of the optimization method.In summary,this paper proposes an improved BP neural network to establish a nonlinear mapping relationship between the design parameters and performance of the valve core profile,and applies this constraint relationship to the design of the valve spool profile of the regulator valve.The method to the valve core surface optimization design.Under the premise of ensuring the stable flow field characteristics of the regulating valve,the flow resistance characteristics are effectively improved and the flow resistance coefficient is reduced at the same time.