On-Board Tailpipe Particulate Number Modeling AUTHORS: Chen Zhang, Lisa Aultman-Hall, Britt Holmén, Eric Jackson
ABSTRACT
ANALYSIS
This study focused on assessing relationships between tailpipe particulate numbers (PN) and second-bysecond vehicle operating characteristics, including speed, acceleration, vehicle specific power (VSP) etc. This study represents an advance in terms of detail over typical emission studies in that a PN prediction model is estimated based on the continuous real-world PN data collected at a second-by-second level. The results of this study contribute efforts for a new generation transportation emission models including movement towards inclusion of particle number in EPA's MOVES model.
Figure 1 PN rate compared to speed and accelerations (for one randomly selected run)
Figure 4 Normalized PN estimates versus VSP Bin by mode
Table 1 Spearman Correlation Mix NormPN NormPN Speed bin
1
Speed bin
VSP bin
Mode
0.40932
0.73721
0.08841
<.0001
<.0001
<.0001
0.25848
0.49907
<.0001
<.0001
1
0.08773 VSP bin
1
<.0001
Mode
1
Figure 3 Plot of Normalized PN Rate vs. VSP by Mode Figure 5 Residual vs. VSP plots a. Model of all modes together b. Model of accelerating mode
Figure 2 PN rate compared to VSP on a Loge scale
DATA
Condensation Particle Counter
Accelerometer
Figure 1 PN rate compared to speed and accelerations (for one randomly selected run)
GPS Receivers Ambient Temperature and Relative Humidity Sensor
Mini-Diluter
Thermocouples
Tailpipe Adapter
Video Camera
5-Gas Analyzer Pitot Tube
ScanTool 4 Pressure Sensors to Calculate Exhaust Flow rate
Desktop Computer
ACKNOWLEDGMENTS This study was funded by National Science Foundation (NSF). It is a continuation of prior studies completed at the University of Connecticut by Eric Jackson, Lisa Aultman-Hall, and Britt Holmén.
CONCLUSIONS • Vehicle specific power (VSP) is the most relevant factor when predicting PN rate • Different modes (accelerating, decelerating, cruising, and idling) corresponds to distinctive relationships between PN rate and VSP • Fixed effect ANOVA models seem to provide better prediction results than continuous models • Model diagnostics results show better model results are obtained by vehicle mode • Potential future research direction – 1) spatial analysis for PN prediction; 2) assessing the relationship between instrumentation accuracy, time resolution and model accuracy
UNIVERSITY OF VERMONT TRANSPORTATION RESEARCH CENTER
BURLINGTON, VERMONT
Average normalized PN emissions by location
www.uvm.edu/~transctr