Date of Award

5-2017

Document Type

Dissertation

Degree Name

Doctor of Philosophy (PhD)

Legacy Department

Automotive Engineering

Committee Member

Zoran Filipi, Committee Chair

Committee Member

Andrej Ivanco, Co-chair, Research Advisor

Committee Member

Beshah Ayalew

Committee Member

Robert Schalkoff

Abstract

Commercial vehicles fulfill the majority of inland freight transportation in the United States, and they are very large consumers of fuels. The increasingly stringent regulation on greenhouse-gas emission has driven manufacturers to adopt new fuel efficient technologies. Among others, advanced transmission control strategy can provide tangible improvement with low incremental cost. An adaptive shift strategy is proposed in this work to optimize the shift maps on-the-fly based on the road load and driver behavior while reducing the initial calibration efforts. In addition, the adaptive shift strategy provides the fleet owner a mean to select a tradeoff between fuel economy and drivability, since the drivers are often not the owner of the vehicle. In an attempt to develop the adaptive shift strategy, the vehicle parameters and driver behavior need to be evaluated first. Therefore, three research questions are addressed in this dissertation: (i) vehicle parameters estimation; (ii) driver behavior classification; (iii) online shift strategy adaption. In vehicle parameters estimation, a model-based vehicle rolling resistance and aerodynamic drag coefficient online estimator is proposed. A new Weighted Recursive Least Square algorithm was developed. It uses a supervisor to extracts data during the constant-speed event and saves the average road load at each speed segment. The algorithm was tested in the simulation with real-world driving data. The results have shown a more robust performance compared with the original Recursive Least Square algorithm, and high accuracy of aerodynamic drag estimation. To classify the driver behavior, a driver score algorithm was proposed. A new method is developed to represent the time-series driving data into events represented by symbolic data. The algorithm is tested with real-world driving data and shows a high classification accuracy across different vehicles and driving cycles. Finally, a new adaptive shift scheme was developed, which synthesizes the information about vehicle parameters and driver score developed in the previous steps. The driver score is used as a proxy to match the driving characteristics in real time. Drivability objective is included in the optimization through a torque reserve and it is subsequently evaluated via a newly developed metric. The impact of the shift maps on the objective drivability and fuel economy metrics is evaluated quantitatively in the vehicle simulation. The algorithms proposed in this dissertation are developed with practical implementation in mind. The methods can reduce the initial calibration effort and provide the fleet owner a mean to select an appropriate tradeoff between fuel economy and drivability depending on the vocation.

Share

COinS