Study On Optimization Of Ignition Angle Parameters Based On Neural Network Model

The calibration of the electronic control parameters of the engine is at the back end of the development process of the electronic control system,and the whole research and development process is sorted out,which is of great significance for the high efficiency and high quality of the electronic control parameters.This paper summarizes the comprehensive development management process,so that the project development can be completed successfully according to the predetermined cost,the expected schedule and the controllable quality.In the process of establishing the electronic control strategy,the engineers develop the coordination mechanism of ECU to meet the specific requirements.The basic ignition advance angle data under steady-state conditions is obtained through the engine bench test.The MAP required for electronic control is obtained by using the MATLAB platform to calculate the electronic control parameter data and engine output data during the experiment.The system can coordinate and determine the set value of the three-dimensional MAP under the corresponding conditions adjusted by the experiment according to the information of the external sensor,and complete the output for the demand through the actuator(solenoid valve).Control parameters for the electronic control system are so much that the coupling degree between the parameters is so high.This paper attempts to establish a steady-state neural network model of input and output with a small number of control parameters and performance parameter data collected on the engine gantry.The neural network is used for learning the response relationship between input and output data so that it could predict some date we have not tested yet.The model prediction results show that the 3/4 speed distribution data training model can predict that the output index error percentage can be controlled within 5%.The 1/2 speed distribution model can make the predicted value roughly follow the experimental test value,and the efficiency of the bench test can be modeled by neural network.The efficiency can be improved,the model’s prediction results can give guidance to the bench experimental engineers.In this paper,the particle swarm optimization algorithm is used to optimize the neural network model between the key output index and input parameters of the engine running time,and the initial optimal solution of the neural network model parameters is obtained.The optimized model predicts the 3/4 speed.The average error percentage of the model’s prediction accuracy for fuel consumption is optimized from 1.85% to 1.23%.The stability of the 1/2 speed distribution model for torque prediction is improved,and the prediction accuracy is also improved to some extent.Through the modeling of neural network and the optimization of particle swarm optimization algorithm,the efficiency of the bench experiment can be improved,and the prediction result of the model can give guidance to the bench experimental engineers.