Some Studies On Nonlinear Estimation And Control Of Automotive Drivetrain

To improve fuel economy, reduce emission and enhance driving performance, manynew technologies have been introduced in the automotive drivetrain area in recent years,which results in a more and more complex powertrain control system. Because of thecomplex nonlinearities (such as the engine torque characteristics, driving resistance andactuator dynamic characteristics) and the uncertainties (such as the mass of vehicle, roadslope and environment temperature), the designed control system should be able to workwell at any operation point under the existence of modeling errors. In order to reduce thecalibration of the development of control software and improve the control performance,this thesis uses the newly developed robust nonlinear control theory, including input-to-state stability (ISS) and backstepping, to address some typical estimation and controlissues of automotive drivetrain. The following issues are investigated: estimation of driveshaft torque, clutch pressure estimation of clutch-to-clutch shift and clutch speed controlof clutch-to-clutch shift.Mechanical resonances of vehicle drivetrain may occur due to the elasticity of thedrivetrain parts, such as clutch spring, propeller shaft and drive axle shaft. How to avoidor reduce the oscillations of drivetrain is an important problem, especially for heavy dutyvehicles which have relatively large drivetrain torsion. Although the knowledge of axleshaft torque is necessary for improving vehicle's longitudinal speed control performanceand shift quality, the shaft torque sensors are seldom used in production vehicles becauseof the cost and durability. Hence, it is required to estimate the axle shaft torque. Anobserver of drive axle shaft torque is proposed for vehicles with stepped ratio transmis-sions, which is a class of switching systems. Model uncertainties including steady stateerrors and unmodelled dynamics are considered as additive disturbance inputs and theobserver is designed such that the error dynamics is input-to-state stable (ISS) for allgear positions. The proposed observer adopts the easily and precisely measured rota-tional speeds to constitute the correction term, while complex nonlinear characteristics ofengine systems appear in their original form of lookup tables. Hence, it is easily imple- mented. A systematic design procedure is proposed where the requirements on decay rateof estimation error and error o?set are easily and explicitly considered. Lowest possibleobserver gains are obtained through convex optimization, which increases the robustnessagainst noises and reduces the estimated upper bound of the error o?set. The designedobserver is tested on an AMESim powertrain simulation model, which represents a typicalmedium-duty truck with a 6.2L diesel engine and a 6-speed automated manual transmis-sion (AMT). Simulation results show that the proposed observer is robust to drivingcondition variations, such as change of vehicle mass and road grade. The observer withconstant gain provides satisfying estimation error o?set for all gear positions. The track-ing overshoot of the high gear driving motivates us to design a switching observer in thefuture work.In the case of passenger cars, it is predicted that hydraulic automatic transmission(AT) and dual clutch transmission (DCT) will possess the most market share of automo-tive automatic transmissions of China. In both DCTs and new ATs, the change of thespeed ratio can be regarded as a process of one clutch to be engaged while the other beingdisengaged, namely, clutch-to-clutch shift. Clutch to clutch shift greatly simplifies themechanical contents of the transmissions and improves the control ?exibility, but makesthe robust shift control a challenge. For vehicles adopting hydraulic cylinder as clutchactuator, which is ubiquitous in the present transmissions, the cylinder pressure controlbecomes very important for good shift quality. Sensors for measuring the clutch cylinderpressure, however, are seldom used because of the cost and durability. Hence, it is re-quired to estimate the shaft torque or the cylinder pressure, in order to enhance controlperformance. A reduced-order clutch pressure observer is proposed in the concept of ISSand the designed observer is tested on an AMESim powertrain simulation model, whichrepresents a mid-size passenger car with a 2000cc gasoline engine. Comparison resultswith the existing sliding mode observer also verify the potential benefits of the proposedISS observer in eliminating chatters and in achieving satisfying estimation performance.During the torque phase, however, the peak error becomes larger when the driving con-dition is changed to a large extent. This problem is solved after the rotational freedom ofthe drive axle shaft is introduced into the model-based observer design.During the shift inertia phase of clutch-to-clutch shift, the applying(oncoming) clutchslips, and the rotational speeds change intensively. The clutch slip control during inertiaphase e?ects shift quality(smoothness and e?ciency) greatly. This thesis, therefore, fo-cuses on the clutch speed tracking control during the shift inertia phase of a stepped ratioautomatic transmission, with taking into account the system nonlinearities and uncertain-ties. A clutch speed controller is proposed based on"Backstepping", and ISS is introduced into the derivation of backstepping, which analyze the stability of the designed controllerunder modeling errors. Comparisons with a designed 2 degree of freedom (DOF) linearcontroller verify the potential benefits of the proposed nonlinear controller in achievingless tracking error. Moreover, the application of dynamic surface control (DSC) solvesthe problem of"explosion of terms"and makes the controller easy to be implemented.In this thesis, ISS is applied for the torque estimation and the pressure estimationof automotive drivetrain, which is a supplement and an improvement of the present in-vestigating methods. Nonlinear method of"Backstepping + ISS"is proposed, and it isused for the clutch slip control problem. Introduction of ISS into backstepping derivationguarantees the controller's robustness against the modeling uncertainties. In order to getan in-vehicle assessment of the proposed observers and controllers, discrete simulationsare carried out as well as the continuous implementations. The proposed observers andcontrollers are discretized at the sampling frequency of electric control system of presentproduction drivetrains. The discrete models of speed sensors are also included. The ro-bustness of the proposed observers and controllers is verified through a large amount ofsimulations. Because the proposed observers and controllers take the system nonlineari-ties and uncertainties into considerations, they can greatly reduce the calibration of thedevelopment of control software and improve the control performance. Finally, becausein-vehicle test has not been carried out until now, the major future work should be theimplementations of Hardware-In-the-Loop Simulation (HILS) and in-vehicle experiments.