Development of a behavioral model of a CI engine for HIL implementation – Torque generation and supercharging

This thesis is centerd around the development of a 1.6L Diesel engine model in Simulink for hardware in the loop purposes. In the automotive industry, the need for a simpler way of testing with respect to a complete HIL structure is arising, mainly for cost-efficiency reasons. In said scenario, the need for an engine model that is parametric and can be accesed to evaluate various measurements is a key point. Thus, the focus for the development has been given to the parametric nature of the model so that it is flexible and adaptable to different motorizations, and the fact that its performancies are mainly focused on the behavioural side, but still respecting at a high degree the system’s physics. The job has been done at Kineton, in collaboration with my colleague Vincenzo Palmieri. The work started with a study of the engine structure and the model’s “external infrastructure” that serves as foundation for our engine model, providing the inputs and outputs of the system. With said study we collected all the knowledge regarding engine components and their theoretical models. The model has been succesively divided in subsytems for logic reasons and to develop in parallel with Mr. Palmieri. I focused on the “power generation” systems, comprising of: intake and exhaust manifold, turbocharger and torque generation. After a long development phase, we had the model ready to be tested. This required to additionally develop an environment to test the model on the PC, basing on the inputs of recordings made with the validated GT-Power model of the same engine. We prepared a structure in Simulink that could substitute the input signals and read and save the outputs. This is kinda a software in the loop testing, as we refer to it as such also in the thesis. Then we collected at the HIL machinery the input and output readings by performing specific manouvers that could cover the most significant behaviours of an engine such as: idle running, constant cruising speed, acceleration and braking. The SIL was thus ready to operate and with its results we started to calibrate and fix the model’s bugs. Once reached satisfying results, we had to confront ourselves with the HIL implementation. This mainly required to modify the model up to a point that no problems would arise while building and deploying it on the machinery. The two main problems have been the removal of algebraic loops and the optimization to run the model under the target time of one millisecond. After solving said issues we tested the model at the HIL so we were able to compare performances of the same manouvers made at the SIL stage, pinpointing the behaviours and problems with respect to the validated one. In conclusion the model has been built completly parametric so that with only changing constant datas the model is still correct, and showed a satisfying general behaviour but with some differencies from the GT-Power model due to the problematic interfacing of all the powertrain black-box models. The most fulfilling result is the low run time achieved, that is lower then the target one of at least one order of magnitude. The performance comparison is linked to the calibration of the model, but the overall demeanor is always respected by making the model very good on the behavioural aspect.