Research On Basic Calibration Of Engine Based On Operating Conditions

With the implementation of increasingly strict emission regulations,higher requirements have been put forward on the power,economy,emission and combustion stability of vehicle engines.That makes the design of vehicle engines more and more complex,needing more and more functions that can be controlled.As a result,the parameters that the engine needs to calibrate are also increasing.The traditional engine calibration method was mainly manual calibration so that the engineer needed to carry out a large number of engine tests through changing the special calibration parameters and comparing the response of different calibration values to the engine where real-time bench experiment was required.This situation would lead to that the engineer selected the control parameters from their own calibration experience.This calibration method relies on the level of the engineer's experience heavily.As engine control parameters increase,manual calibration is difficult to put into practice at the same time,and will lead to long project cycles.Therefore,the calibration of the engine is facing severe challenges.In order to improve this situation,the main work of this paper is as follows:Firstly,the basic calibration environment of the engine is introduced,including engine technical parameters,EMS(Engine Management System)and bench test system.The basic calibration problem of the engine is scientifically described,and the three factors of gas,oil and fire are decomposed,and then transformed and analyzed.The parameters of the rail pressure and the SOI(start of angle)in the stage of engine performance development are taken as the final parameters.At the same time,the ignition angle and the air-fuel ratio are analyzed,and the input strategy of optimal ignition angle and optimal air-fuel ratio under any working conditions are formulated.However,the calibration problem is focused on the VVT(Variable Valve Timing)parameters,which translates the basic calibration problem into a multi-objective optimization problem.Then this paper analyzes the influence of VVT parameters on the performance of the engine,and develops a calibration strategy based on data modeling.Based on the space filling method for DOE(Design Of Experiment),completing the scientific test design,can significantly shorten the test cycle and cut down the test times on expensive bench.Secondly,based on the GPR(Gaussian process regression)principle,the GPR models are established where engine basic control parameters responses to engine performance and boundary.Through data inspection,data transformation and outlier detection,the accuracy of the model after final training is improved,and all models are also checked by the additional verified data.Finally,based on the established GPR models,the VVT parameters are calibrated by NSGA-?(Non-dominated sorting genetic algorithm-II).Some GPR models are used as fitness functions,and the rest of GPR models are used as constraints,this approach achieves the goal of combining the GPR model with the NSGA-? algorithm.Under the optimization framework of GPR model combining with NSGA-?,the VVT parameters were calibrated by running algorithm and following the "trade-off" strategy.Finally,the calibration results were verified to be reasonable and effective on not only the virtual GPR model but also the real engine test bench.The engine economy,emissions and combustion stability are improved equally.The final calibration results are presented in the form of universal performance maps.