Optimization Of Engine Calibration By Particle Swarm Algorithm Combined With BP Neural Network Model

Today,the automotive market is still developing rapidly and the future is still very bright.However,the increasing number of cars has brought about problems such as the scarcity of oil resources and increasing air pollution.Therefore,to improve the economic performance of cars and improve their emission performance has become the goal of major automobile manufacturers.Calibration optimization technology is one of the most important methods to accomplish this goal,through which the control parameters of the engine under different operating conditions can be continuously optimized to improve its economy and emission performance.However,with the increasing complexity of automotive engine electronic control system,the traditional calibration optimization technique generates a very large and inefficient workload,which can no longer meet the calibration needs,and the model-based calibration optimization technique has become one of the important methods for engine calibration nowadays because of its ability to reduce test workload and improve calibration efficiency.Based on this background,this paper conducts the study of engine calibration optimization by particle swarm algorithm combined with BP neural network model,and the main research contents are as follows:The Sobol sequence sampling method was used to design the test points before the engine modeling,and a total of 520 test points were identified for the bench test.After completing the bench test phase,the input and output of the obtained test points were applied to the BP network modeling work.In determining the optimal BP network model,the cross-validation method is first used to improve the prediction accuracy of the model and avoid overfitting problems caused by too small a prediction set.In addition,the most suitable model for this paper was obtained by gradually determining the network model parameters.Finally,the BP neural network model was compared with other regression models,and it was found that the BP neural network model had the highest quality and was the most suitable for the modeling work of this paper,so that the final modeling model of this paper was successfully determined.The optimal BP network model and the particle swarm algorithm are jointly simulated to optimize the fuel consumption data at each operating point step by step under the premise of meeting the emissions,and the MAP of the control parameters corresponding to the final optimal fuel consumption is obtained,and then the bench validation is carried out to confirm that the MAP of the control parameters after the calibration optimization makes the fuel consumption curve of the engine smoother and the economy is significantly improved,which indicates that The study on calibration optimization in this paper has achieved substantial results.