Neural-network-based analog simulation system of vehicle gas emissions

Abstract

The invention relates to a neural-network-based analog simulation system of vehicle gas emissions. The system includes a basic simulation model building module, a neural network input data processingmodule, a neural network module, a neural network training module and an emission calculation and output module. Firstly, the basic simulation model building module is used for building modules for engines, transmission systems and the overall vehicle; secondly, the neural network input data processing module is used for simulation of the modules, and original emission data is obtained. The original emission data, catalytic converter parameters and the emission influence conditions are combined, and input data of the neural network training module is obtained. The neural network training module completes analog calculation on the input data, and final emission data of the overall vehicle is obtained; finally, special working conditions are added in the emission calculation and output module for correction of the final emission data. Through the combination of the neural network technology and the generation and transformation rule of the emissions of the overall vehicle during the running test cycle, the emission result of the whole vehicle is accurately predicted.

Description

technical field [0001] The invention relates to the performance simulation of a vehicle, in particular to a neural network-based simulation system for vehicle gas emissions. Background technique [0002] Among the requirements for energy saving and emission reduction in the automotive field, the two most important indicators of vehicle performance are fuel consumption and emissions. The design, model selection and matching work in the process of vehicle performance development are also carried out around these two key indicators. In the whole vehicle development process, CAE (Computer Aided Engineering) tools have been widely used. With the increasingly mature application of CAE, the performance can be analyzed and evaluated in the early stage of vehicle development, so as to guide the design, selection and matching The work is more reasonable, the development rounds and the workload of the whole vehicle test are reduced, the development cycle is shortened and the developme...

AI Technical Summary

Problems solved by technology

However, both methods have certain limitations. The first method needs to simulate the physical and chemical processes in the complex combustion process, and needs to establish a detailed three-dimensional model of the engine and combustion-related structures, and divide a large number of grids. It leads to long calculation time, large cumulative error, many factors affecting the final result in the intermediate process, the overall accuracy is not high, and the factors affecting the conversion efficiency of the catalytic converter cannot be well reflected

The second method mainly simulates the chemical reaction in the catalytic converter. It is necessary to define the chemical reaction equation and use the existing test data to calibrate the chemical reaction coefficient, determine the catalytic conversion efficiency, and combine it with the original engine emission test data to calculate the final emission. It is predicted that this method does not consider enough factors such as the ignition process and the warm-up process, and the calibration process of the chemical reaction coefficient is also very complicated. The chemical reaction coefficient calibrated by the existing data only corresponds to the structure and formula of the current catalyst. The representativeness of other catalytic converter schemes is poor, and this method uses the test data of the engine's original emission cannot well reflect the engine's transient emission performance, and the applicability is not good

Images