Model based driveline analysis and controller development for improved driveability

Market demand has been pushing the truck manufacturers to provide trucks with more engine power and torque to the market. This demand is essentially related to the performance expectation of the customers to have high wheel torques available. In addition to the performance increase, also the customer expectations have increased on other attributes such as Noise Vibration Harshness (NVH), vehicle dynamics and etc. Driveability has also become an important attribute of the vehicle. A product needs to be developed to satisfy these expectations related to the stated attributes. This necessity brings up the need of analysis of driveability and driveability error states, which will be more obvious because of high torques in the driveline. High torques in the driveline is a result of the high transmission ratio and engine torques in the driveline. Driveability term contains responses to several maneuvers such as launch, acceleration and shifting. These maneuvers can be further extended to more detailed maneuver types. Examples to these maneuvers are in-gear acceleration, acceleration after shifting, up-shifting, down-shifting, comfort launch and performance launch. It is possible to analyse the driveability attribute of the vehicle separately for all these maneuvers. One of the most important error states of the driveability are the oscillations in the driveline and jerk which are caused by sudden tip−in and tip−out maneuvers of the driver which result in sudden load changes in the driveline. The presented work will focus to the tip-in maneuvers which covers also the tip-out maneuvers as the difference between these maneuvers are only the direction of torque change. Within the scope of the presented work, three vehicle models reflecting driveline components in different detail levels are created in Matlab environment in order to perform the mentioned analysis of driveability and driveability error states. The transient maneuver results obtained from these models are compared to test results of the real vehicle. The results show that in order to analyze the driveability and driveability error states, the elastic driveline model where the drive shafts are modelled is sufficient. The contribution of the tire models to reflect driveability is limited. In the next step of the work, in order to improve driveability, the effect of changing the physical characteristics such as inertia, stiffness and damping of the driveline components is investigated. According to the results of the work, it is shown that driveability error states during tip−in and tip−out maneuvers cannot be solved by changing physical properties of the driveline components. It is shown that the solution is not on the system side, which is the driveline in this case but in the excitation side. Another contribution of this part of the work is that it has shown how the changes of the physical properties of the driveline components may affect driveability. As the error states occurring during tip−in and tip−out maneuvers in terms of oscillations and jerks are shown to be not evitable by changing the physical properties of the driveline components, the necessity of controlling the demanded torque is arisen in order to solve the problem in the excitation source. It is shown that by applying filters or ramp functions will cause performance complaints by the driver and modern control methods are investigated. The control methods which are investigated in the context of this work mainly consist of three steps; defining metrics for two contradicting requirements such as performance and comfort, establishing a cost function using these metrics and applying control to put the cost function at minimum. In these methods which incorporate cost functions, the weighting values are of importance because they directly affect the optimal solution thus the control gain which is to be used. Of course there are methods which define how to choose the weighting functions but these methods do not ensure the satisfaction of the customer and rely on subjective assessment of the one who is tuning the weighting factors. Another disadvantage of the proposed controllers are either they do not consider different driver inputs such as differentiation between low tip−in and high tip−in or rely on drivers requested torque which is not directly linked with drivers input to the vehicle. Due to the disadvantages states above, the necessity for a new controller design is seen. The proposed "Driving Input Dependent Optimal (DIDO) Controller" is designed not to rely only on the weighting factors which will be chosen by the engine calibration engineer, who will calibrate this function but it will ensure that the customer expectations both from comfort and performance perspective are satisfied according to brand strategy and characteristics. The designed controller consists of a tip−in detection algorithm which will classify the tip−in maneuver depending on the accelerator pedal position and its rate, a reference acceleration generation algorithm which will generate reference acceleration of the vehicle based on the tip−in type and brand characteristics, three optimal controllers which are performance oriented, comfort oriented and the hybrid of these modes and a state machine which will govern the switching between these three controllers with respect to the driver inputs, acceleration of the vehicle and brand characteristics. During the controller development, the non−linear driveline model is converted to a linear model by making some assumptions which are explained in detail within this work. The effect of the linearization onto the simulation results are investigated before continuation with the controller development. Once the controller is designed and shown to work as it is intended, the LTI (Linear Time Invariant) plant model is removed and the non−linear plant model is used. By evaluation of the results with the non−linear plant model, it is seen that a Kalman filter is required to compensate the differences between the linear and non−linear driveline models. With the proposed Kalman filter, controller is shown to operate successfully. In the work, also the sensors which will be used in a real vehicle application are also identified and it is seen that the available sensors will not be enough to obtain all the necessary state information which are used in the controller for calculating control inputs. Therefore another Kalman filter is designed which can estimate the states from the information obtained from the sensors, known states in this case. It is also shown that the controller can operate properly by estimating states from the sensors. In the next step of the work, controller sensitivity analysis is performed. The sensitivity analysis is required because controller is designed based on driveline model parameters. It is expected from a controller to be able to compensate the difference between the real driveline parameters and the parameters to be used in the model. It is also shown that the proposed controller can work with the deviation of the driveline parameters as well. As the last step of the controller design, a process map for calibrating this function is generated. The process map shows that calibration of this controller is more efficient and effective compared to the other controllers proposed in the literature because it considers the expected outcome and tracks it according to its design. The contribution of this work to the literature can be summarized by its capability to propose a non−linear driveline model which includes all of the elastic elements in the driveline by incorporating submodels for the driveline backlash and tires for a rear wheel driven commercial vehicle; its results showing the effects of changes of physical properties of the driveline components on the driveability and the resulting "Driving Input Dependent Optimal (DIDO) Controller" which is capable to overcome driveability error states with an ease of application, ensuring delivering brand characteristic to the customer and adaptation to the different driver inputs.