Research On NO_x Emission Algorithm Of Heavy Duty Vehicle On Real Road

Nitrogen Oxides(NO_x) exhausted from motor vehicles is one of the most harmful atmospheric pollutants.There are two main methods for NO_x measurement: engine bench emission test and road emission test.Bench emission test is a laboratory test method to measure the emission of engine,it evaluates the NO_x emission of heavy vehicles in indoor environment.Road emission test is carried out by whole vehicles on actual road according to specified test process and road sequence.It has the characteristics of short test time,fixed test conditions and fixed road state.While in actual running process of heavy vehicles,due to the unstable load,road and environment,the actual running conditions and emission status are more complex and diverse.Comparing actual running status with the bench and road emission tests,it is found that the test conditions of both engine bench emission test and road emission test are not enough to completely cover the complex operation conditions of heavy vehicles in actual running process.At the same time,engine bench emission test is difficult to characterize the effect of the external state on vehicle emissions,road emissions test can only represent the emission of vehicles within the certain conditions.Therefore,in order to evaluate the NO_x emission of heavy vehicles in complex and unstable actual status,it is necessary to extract the representative samples that characterize heavy vehicles operation and emission state.The representative samples can be used to calculate and evaluate the NO_x emission in actual vehicle operation process.This paper focuses on the representative samples of heavy vehicles.Key research contents of vehicle emission influence characteristic parameters,samples that represents NO_x emission,and optimization of the representative samples are presented.The main research contents of this paper are as follow.(1)Vehicle characteristic parameters and their influence on NO_x emission.For the heavy vehicles,the necessary parameters are collected through CAN and added sensors.Characteristic parameters are divided into basic parameters and extended parameters.The influence law of vehicle basic parameters and NO_x emissions has been thoroughly studied in other research,this paper verify the situation based on the actual data of the heavy vehicle.The influence of vehicle extended parameters and NO_x emission is unknown.In order to analyze the correlation between extended parameters and NO_x emission,the paper construct a BP neural network.Through the the BP neural network,the influence weight of extended parameters to NO_x emission are analyzed.The irrelevant parameters are removed according to the size of the weight,the strongly correlated parameters are considered as the factors that affect the vehicle NO_x emission.(2)Data partitioning and characteristics description of sequence fragments.For 12 influence parameters related to vehicle emissions,factor analysis is conducted after data standardization.The 12 emission characteristic parameters are condensed into 5 common factors,and the physical meaning of the common factors is explained according to the correlation coefficients.Since each of the 5 common factors contains the information of the 12 emission characteristic parameters,engine factor is selected as the basis for vehicle emission status interval partition according to the amount of emission characteristic parameters information contained in each common factor.After the state intervals are determined with a certain step size,the vehicle original data is divided into sequence segments and put into the corresponding intervals.Average,variance and slope are used to describe the characteristics of each sequence segment,which can be used as the basis for judging similar segments in clustering analysis.For the sequence fragments divided from the original data,two-step clustering is conducted through the descriptive parameters of the sequence features.According to the calculated BIC value,the optimal clustering number is determined and then introduced into fuzzy c-means clustering.The chaotic sequence fragments in each interval are clustered into several different categories through membership grade,and the representative sequence sets are constructed.(3)Research of extraction and combination method of representative sequence fragment,optimization of sample sequences representing actual NO_x emission characteristics of heavy vehicles,and analysis of sample sequences method and bench test and road test.For each interval,according to the distance between the sequence fragment and the clustering center,the representative sequence fragments in each category are extracted and combined by a certain method to obtain the initial sample sequence representing the actual emission state of the vehicle.To improve the representational ability of the sample sequence to the actual vehicle working condition and emission status,the influence factors of sample sequence structure are analyzed,the influence factors include step length of interval partition,sequence extraction principle and sequence fragments combination.Then,the dynamic time warping method is introduced to optimize the sample sequence.When the step length of interval partition is 5,the sequence extraction principle abides by proportion extraction,and the sequence fragments combination use Markov approach,the sample sequence is the best data sequence that cover various heavy vehicle driving conditions and emission statuses,and the NO_x emission of vehicles can be reflected by the operation and emission parameters of sequence.Comparing the sample sequence method,bench test and road driving emission test,the coverage of actual operation and NO_x emission value are being analyzed,the results show that the representative sample sequence can better cover actual operation of vehicle,and the calculated NO_x emission value is higher than the bench test and road driving emission test.