Algorithm to generate Synthetic Driving Cycle using Real driving data
Typ
Examensarbete för masterexamen
Program
Automotive engineering (MPAUT), MSc
Publicerad
2021
Författare
Nagaraj, Subramanya
Modellbyggare
Tidskriftstitel
ISSN
Volymtitel
Utgivare
Sammanfattning
The growth of technology has led to much increase in pollution levels. The European
Union has enforced strict rules for car manufacturers to reduce the emission
levels for vehicles. The regulation of the European Union includes a test for Real
Driving Emissions. The automobile manufacturers are forced to test their vehicles
for Real Driving Emissions. The available driving cycles like WLTC or NEDC lack
real-world driving characteristics. This makes it is highly essential to develop a driving
cycle by using real driving data. An algorithm is created to produce a driving
cycle delivering the parameters within the Real Driving Emissions test parameters.
In this master thesis, micro-trip based construction model is applied for the vast
data collected from real driving trips. The process includes use of unsupervised
learning algorithm by utilizing k-means clustering technique to group the data. The
statistical CH index is used to evaluate the performance of clustering and the trips
are filtered with the Real Driving Emissions parameters before deploying D-optimal
design to maximize the created design matrix from the filtered data. The microtrips
are selected in a ratio of 7:1:1 with urban, rural and motorway sections to stay
within the required duration limits. The selected micro-trips are combined to form
complete driving cycles, and are simulated using a simulation model constructed
by using QSS toolbox in Simulink. The model comprises a normal IC engine with
manual transmission, capable enough to determine the fuel consumption.
The developed driving cycles are analyzed and their parameters are compared with
real driving emission test criteria. The results show that the cycles are valid. The
results of the simulation are dependent on the engine operating points. The transmission
model needs to be calibrated and evaluated with the real scenario to increase
accuracy. The regression analysis carried forward from the simulations, predicts the
relation of VApos with fuel consumption. The aggressiveness of the cycle tends to
increase fuel consumption. Hence, it helps to understand the variation in fuel consumption
based on the driving cycle parameters
Beskrivning
Ämne/nyckelord
Micro trip based construction model , K-means clustering , CH Index , D-optimal design , Driving cycle , fuel consumption