CLEAR
„ Classification of Emissions from Automobiles in Real Driving
“
Erstellt im Auftrag von
BMW
Bericht Nr. FVT-85/2012/ Fu Em 27/2011 - 6790 vom 03.05.2013
Abstract
Stefan Hausberger
Gernot Kager / Report No: Bericht Nr. FVT-xx/2012/ Fu Em 27/2011 - 6790 vom 03.05.2013
Title: CLEAR - „ Classification of Emissions from Automobiles in Real Driving
Orderer: BMW / Contract:
Publication date: 03.05.2013
Abstract:
CLEAR is a software tool for the after treatment of emission data collected at PEMS test cycles. The main purpose is to normalize results over different driving styles to a representative load cycle and also test cycle layouts. In particular this means that emission results from very moderate driving should be enhanced and results from very aggressive driving should be lowered to correctly analyze the emission behaviors of vehicles.
The actual report is the user manual for CLEAR
Key Words: PEMS tests, LDV, Emission Aftertreatment
Edited / Univ.-Prof. Dr. Stefan Hausberger
Dr. Nikolaus Furian / 03.05.2013
File J:\TE-Em\Projekte\2011_27_PEMS_Weighting_Factors_BMW\Reports\Doku_CLEAR\Doku_Clear_FVT_Format_V4.docx
1 CONTENT
1 CONTENT 3
2 General Concept 4
2.1 Weighted Emission [g/h] 4
2.1.1 Rasterize PEMS Test data into a GP 5
2.2 Weighted Emission [g/km] 6
2.3 Weighted Emission per CO2 [g/kg] 6
2.4 Mean Emission 6
3 CLEAR Tool 6
3.1 Load Goal Matrix 7
3.1.1 Loading Goal Pattern from a File 8
3.1.2 Goal pattern created from normalized pattern 8
3.2 Load Data 10
3.3 Choose Settings 10
3.4 Calculation of Weighted Emissions 10
3.5 Output 11
3.5.1 Result Information 11
3.5.2 Warnings and Logs 11
3.5.3 Goal Pattern Analysis 12
3.6 General Settings 14
3.6.1 Engine Power Tag 14
3.6.2 Engine Speed Tag 14
3.6.3 Distribution Tag 14
3.6.4 MinimumRequirementsTimeShare Tag 15
3.6.5 MinimumRequirementsAbsoluteSeconds Tag 15
3.6.6 Input Definition tags 16
3.7 Target histogram(s) for the Engine power 16
2 General Concept
To analyse vehicles and compare results three types of result values are computed. Weighted emission results are calculated in [g/h], in [g/km] and per CO2 in [g/kg]. In addition, the mean emission values over the whole test-drive is reported. In the following each of these quantities and the calculation scheme are discussed separately, where the main idea is demonstrated in the section for weighted emission in [g/h] and per CO2 [g/kg].
2.1 Weighted Emission [g/h]
The basic methodology is the use of a generalized standard-driving-pattern, or goal pattern, which represents the typical driving behaviour of a standard driver with respect to engine speed[1] and engine power. To do so interval patterns of engine speed and power are defined, N=(n0,n1,n1,n2,…,nk-1,nk) and P=(p0,p1,…,pl-1,pl), where
po:minimum power
p1:end of first and start of second interval
⋮
pl:maximum power
and
no:idling speed
n1:end of first and start of second interval
⋮
nk:maximum engine speed
A goal pattern (GP) is a frequency distribution in matrix form that assigns a relative frequency to each combination of these intervals. Hence it is a k×l Matrix where GPi,j, with 1≤i≤k, 1≤j≤l, represents the ratio of driving time in the combination of intervals ni-1,ni×(pj-1 ,pj), see for example Figure 1.
Figure 1: Example of a Goal Pattern (normalisation of the engine power see chapter 3.7)
2.1.1 Rasterize PEMS Test data into a GP
In the first step, the collected data from a test-drive is grouped according to the matrix pattern given by the goal pattern. Therefore, at every time point the engine speed and power, as well as emission data, is required. For smoothing purposes averages of data sets over short time periods (e.g. 3 seconds) are built first. These average-data sets are then partitioned and stored in a data-structure that represents the structure of the goal pattern, as illustrated in Figure 2.
Figure 2: Concept of Data Processing
In the second step, the grouped data-sets are summed up to build the average value in each interval combination. These average values are then weighted with the corresponding probabilities of the goal pattern. Thereby, combinations of engine speed and power that occur more often in driving patterns are assigned bigger weights and contribute more to the overall weighted emission result. Whereas regions with less probability, e.g. high power regions are taken less into account as they are reached less while standard test-drives. Figure 3 summarizes the principal of this procedure.
Figure 3: Concept of Weighted Emission Method
2.2 Weighted Emission [g/km]
To obtain weighted emission data in gram per kilometre, the average speed for each interval combination is computed in the same manner as the emission data was derived. Further, these average values are weighted in the same way as the emission data, illustrated in Figure 3. The quotient of both results leads to the desired quantity of weighted emission per distance,:
Weighted Emission gkm=Weighted EmissionghWeighted Speed kmh.
2.3 Weighted Emission per CO2
To obtain weighted emission data in gram per kilogram CO2, the average CO2 in [g/h] for each interval combination is computed in the same manner as the emission data was derived. Further, these average values are weighted in the same way as the emission data, illustrated in Figure 3. The quotient of both results leads to the desired quantity of weighted emission per CO2:
Weighted Emission gkg=Weighted EmissionghWeighted CO2 kgh.
2.4 Mean Emission
To compare corrected results of the weighted methods with the actual emission data the mean emission values over the entire PMES test time might also be of interest. Therefore, the emission data at each time point are summed up and simply divided by the number of time-points or other mean results of calculated per km or CO2:
Mean Emission gkm=Mean EmissionghMean Speedkmh,
Mean Emission per CO2 gkg=Mean EmissionghMean CO2kgh.
3 CLEAR Tool
The CLEAR software tool implements the procedure described above and provides simple user-interaction features to alter settings and parameters of the algorithm, as well as basic result presentation- and visualization features. At the execution of the CLEAR.exe the main window and the settings tab appear, which is illustrated in Figure 4.
Figure 4: Main Window of CLEAR Tool
In the following, the basic steps of calculations (load goal pattern, load data, choose settings and initiate calculation) are described separately and the handling of the CLEAR is tool is described in detail.
3.1 Load Goal Matrix
There are two ways to load a goal pattern for calculations:
- Loading from a file
2. Using a standard-normalized pattern and de-normalize with chosen settings
The preferred method can be chosen with the radio buttons in the Goal Pattern Settings-Group, as shown in Figure 5. Either the chosen file name or the message “Goal pattern from vehicle Data” will appear in the “Goal Pattern” text-box in the file information section.
Figure 5: Goal Pattern Settings Section
3.1.1 Loading Goal Pattern from a File
Any goal pattern with any corresponding interval structure over engine speed and power may be loaded and used during calculations. However the format of the file has to follow strict rules to ensure that it can be read properly by the tool:
· the file format is limited to CSV files
· the delimiter has to be set to the comma ‘,’
· there must not be any empty rows and columns before the input data starts
· the first two rows, starting in the third column represent the interval pattern (n0,…,nk) for the engine speed
· the first two column, starting in the third row, represent the interval pattern for the engine power (p0,…,pl)
· both interval patterns have to be strict monotone increasing
· The fields enclosed by these two patterns represent the corresponding probabilities (or frequencies), where the limits of the interval are specified by the first two rows and columns in a “from to” manner.
Figure 6 gives an example, note that the column- and row names are only reported for demonstration purposes and MUST BE LEFT OUT in actual files.
Figure 6: Input Structure of Goal Pattern (power intervals in kW are defined in c1&c2, the corresponding engine speed in rpm I defined in r1&r2 from c3 on, the shares of each grid is written from r3/c3 on, the sum of all shares shall be 1.0)
The probability in the cell (c3,r12)denotes the frequency of speed/power combinations of the intervals 780,941×[0.0,12.4). By loading a goal pattern from a file any interval pattern (as long as it is strict monotone increasing) and any frequency distribution of those intervals can be loaded.
3.1.2 Goal pattern created from normalized pattern
To alter goal patterns quickly over different vehicles the user is able to define a standard normalized goal pattern and de-normalize it using specified settings. The standard normalized patterns can be defined via the GeneralSettings XML file, which is described in detail later. However, the standard normalized pattern must be a histogram over power where no classification and grouping of data is done over engine speed test-data values. The dimension of the power intervals structure and its limits are free to choose.
Figure 7: Settings for GP from Vehicle Data
The de-normalized entries of the interval pattern, pd, for engine power are derived from the normalized values, pn, using the following expression (normalisation method see chapter 3.7):
pdi=pni⋅R0+R1⋅Speed+ R2⋅Speed2+SMK⋅RefAcc⋅Speed3600, for 0≤i≤l.
Speed…RefSpeed kmh, m in kg, RefAcc in ms2 , pdi in [kw]
Changing parameters in the settings does not automatically trigger a recalculation of the goal pattern. The user has to do this manually by clicking on the “Calc Goal Matrix” button. However, changed parameters are highlighted with yellow boxes to alert the user that the calculation is missing . Reference speed and acceleration are currently not adjustable.
Figure 8: Edited Vehicle Data
3.2 Load Data
In the second step the user can load data from a test drive. Again the format in which data is loaded is very important. The required file-format is again CVS with ‘,’ as the delimiter. There are some mandatory columns that have to be loaded to ensure meaningful results:
- Time [s]
- Engine Speed [rpm]
- Engine Power [kw]
- Speed [km/h]
- CO2 [g/h]
The positions of these columns are free to choose in the GeneralSettings XML. In addition several emission values can be loaded to the tool, their names (identifiers) and position in the input files have also to be specified in the GeneralSettings XML.
Note that at least one emission value has to be specified. All emissions have to be provided in [g/h]. The file-names of loaded files are listed in the “Data Loaded” text-box in the file information section.
3.3 Choose Settings
Before launching the calculation the user is able to adjust some basic settings in the settings area, see Figure 9.
Figure 9: Calculation Settings Section
As described in the first section the average of test-data values over a short time periods are built before those values are assigned to the interval-data structure. Adjusting the setting averaging width the user is able to define the size of these sub-samples.
Further, the user can chose whether data values with engine power above the maximum value and below the minimum value should be included in the calculation. According to the pattern showed in Figure 6, this means values with power smaller than -37.2 and bigger 123.9. Values outside these limits are then assigned to the first (p0,p1) or last interval(pl-1,pl).
3.4 Calculation of Weighted Emissions
The last button in the command section triggers the calculation of weighted emission results, mean emission results and moving averages. Results are computed in [g/h], [g/km] and [g/kg] according to the calculation scheme described in the first section. All specified emission quantities are considered for computation of weighted and mean results, where moving averages are only reported for the first emission quantity
3.5 Output
3.5.1 Result Information
The computational results are represented in two ways, they are visualized within result information section in the main window of the tool, and they are exported to a CSV file in the file-system.
Figure 10: Result Information Section
The user is able to select a file in the “Data Loaded” section and the corresponding results appear in the “Emission Info” text box.
Further, in the containing folder of the CLEAR.exe a CSV named “ExportData.csv” file is created that contains all information.
3.5.2 Warnings and Logs
In addition to analysis of emission results the CLEAR tool warns the user if loaded test drives do not fulfil minimum requirements in terms of goal pattern coverage. This is done in two separate ways:
3.5.2.1 Minimum Required Time Share per Power Interval
The user is able to specify in the ConfigSettings.xml to which extent the share in the goal pattern has to be reached by a test drive. Let’s assume that the goal pattern is histogram over power and the overrun segment (negative power) is assigned a frequency of 20% in the goal pattern. If the user specifies in the ConfigSettings XML that at least 25% of this value must be reached, CLEAR will report a warning if not more than 5% of all test values occurred in this segment. The warning is reported in the Logs and Warnings section of the Settings and Results tab.
Figure 11: Warnings if Time Share Requirements no met
In this view it is not described in detail in which interval the requirements are not met, if one power region fails the whole file is reported. A more detailed observation can be obtained in the “Goal Pattern Analysis” tab, where regions which passed the test are marked with a green “yes” and regions that failed with a red “no”, see Figure 12.
3.5.2.2 Minimum Required Absolute Times per Power Interval
In addition to the time share requirements an absolute minimum of time steps measured in a specific power region can be specified in the ConfigSettings XML.