Controller Design for a Gearbox Oil ConditioningTestbed Through Data-Driven Modeling

Abstract: With the exponential development of more sustainable automotive powertrains, new gearbox technologies must also be created and tested extensively. Scania employs dynamometer testbeds to conduct such tests, but this plethora of new and rapidly developed gearboxes pose many problems for testbed technicians. Regulating oil temperature during tests is vital and controllers must be developed for each gearbox configuration; this is difficult given system complexity, nonlinear dynamics, and time limitations. Therefore, technicians currently resort to a manually tuned controller based on real-time observations; a time-intensive process with sub-par performance. This master thesis breaks down this predicament into two research questions. The first employs a replicate study to investigate whether linear system identification methods can model the oil conditioning system adequately. A test procedure is developed and executed on one gearbox setup to capture system behavior around a reference point and the resulting models are compared for best fitment. Results from this study show that such data-driven modeling methods can sufficiently represent the system. The second research question investigates whether the derived model can then be used to create a better-performing model-based controller through pole placement design. To draw a comparison between old and new controllers, both are implemented on the testbed PLC while conducting a nominal test procedure varying torque and oil flow. Results from this study show that the developed controller does regulate temperature sufficiently, but the original controller is more robust in this specific test case.