Consequences of Modular Controller Development for Automotive Powertrains A Case Study

A modular controller structure for automotive power trains has certain bene ts These include improved pro ductivity through module reuse seamless integration of new features transparent removal of obsolete features and module sharing across powertrain platforms Mod ular architecture also potentially reduces the complexity in the design and calibration process in that controller modules for di erent subsystems are developed indepen dently Due to the fact that the automotive powertrain system contains many highly interactive sub systems it is not clear that a modular controller development pro cess can yield acceptable feedback controller performance with respect to emissions fuel economy and drivability In this paper we describe the engineering design issues associated with a decentralized development process and the impact that the resulting decentralized controller has upon the dynamic response of the feedback system We describe the possible detrimental consequences of subsys tem interaction and the potential of coordinated multi variable feedback for alleviating these limitations Control of a spark ignition engine incorporating variable camshaft timing is used as a case study Introduction The automotive powertrain controller is tasked with regulating exhaust emissions to meet increasingly strin gent standards without sacri cing good drivability and providing increased fuel economy to satisfy customer de sires and comply with Corporate Average Fuel Economy CAFE regulations First designers are developing in novative mechanical enhancements of the spark ignition engine to achieve these goals New features provide ad ditional design parameters control variables needed to Control Systems Laboratory Department of Electrical Engineering and Computer Science University of Michigan Ann Arbor MI J Grizzle and A Stefanopoulou are supported in part by the National Science Foundation un der contract NSF ECS and J Freudenberg is sup ported by the National Science Foundationunder contractNSF ECS matching funds to these grants were provided by FORD MO CO Ford Motor Company Scienti c Research Laboratory PO Box Mail Drop SRL Dearborn MI improve engine performance over a wide range of operat ing conditions Tuning these parameters is a complicated problem because they interact with various powertrain subsystems such as the breathing process combustion process and exhaust generation process Second it is in creasingly important to achieve control over transient be havior for example to rapidly reject disturbances in air fuel ratio A F in order to minimize tailpipe emissions during transient operations Developing and implementing a powertrain manage ment system is a complex and multifaceted engineering task From the perspective of a controls engineer it is natural to approach this problem by developing a dynamic model of the complete powertrain A dynamic model fa cilitates study of such phenomena as transient response and subsystem interaction The information thus pro vided enables the engineer to make informed decisions and tradeo s that a ect several components of the power train with the design goal of achieving satisfactory overall system performance On the other hand the powertrain control problem is complex and one way to manage this complexity is to divide it into subtasks The goal of each subtask is to develop a controller module for a speci c component of the powertrain subsystem such as exhaust gas recirculation spark ignition timing and A F control Modular controller development reduces complexity of the design and yields a modular controller architecture From the software standpoint a modular architecture refers to a software organization consisting of a collection of inde pendent program components with well de ned interfaces specifying the information ow across module boundaries Such an architecture has many bene ts improved pro ductivity through module reuse seamless integration of new features transparent removal of obsolete features and module sharing across powertrain platforms Addi tionally maintainability is substantially enhanced in that modules can be modi ed independently of other parts of the powertrain control strategy That is one might re move and replace the EGR or fuel control module without a ecting the rest of the strategy The various advantages of modular controller design render it common practice for the design and calibration of control modules for di erent subsystems to be performed independently A potential caveat associated with modular controller design is that it naturally leads to a decentralized con trol architecture Were each of the powertrain subsys tems associated with the software modules independent of the others decentralized control would work well Con ventional automotive design practice has been to assume independence and apply typically classical control sys tem design techniques to individual engine and powertrain subsystems This is adequate for collections of subsystems where there is only weak dynamic coupling or where in teractions can be minimized by de tuning or calibrating a subsystem controller to avoid unintentional excitation It may be appreciated however that this approach of ten results in less than optimal system performance and imposes a large calibration burden in time and e ort for the very reason that there are in fact strong interactions among the various subsystems These interactions limit the ability of decentralized control to achieve the level of performance obtained with centralized multivariable con trol Furthermore even if a decentralized control strategy is satisfactory in implementation it may prove necessary to coordinate the design and analysis of the individual control modules as well as the calibration of their con troller parameters The issues associated with modular controller devel opment will be illustrated in subsequent sections using a system model that describes an engine equipped with variable cam timing This system has signi cant interac tion between the dynamics of the variable cam mechanism and those of the air fuel ratio subsystem We shall dis cuss the relative utility of a modular decentralized con trol architecture versus a multivariable control strategy It will be shown that allowing the fuel command used to regulate air fuel ratio to depend upon the cam phasing results in smaller transients in air fuel ratio This im provement in dynamic performance is at the expense of a more complex control architecture because the cam tim ing controller and the air fuel ratio controller are no longer self contained software modules Finally it will be argued that even if a decentralized control design is possible it is necessary to design and analyze the controller from a multivariable viewpoint in order to manage the tradeo between software complexity and controller performance Background on the VCT Engine and the Control Problem The purpose of this section is to brie y explain the dynamics of a spark ignition engine equipped with a vari able cam timing VCT mechanism with special empha sis on the cam phasing mechanism and its interactions with several subsystems of the engine Variable cam tim ing is a promising new feature for automotive engines be cause preliminary investigations show potential bene ts in fuel economy combined with emissions reduc tion It also obviates the requirement for external exhaust gas recirculation systems commonly used for NOx reduc tion Cam timing is used to reduce the base HC and NOx feedgas emissions levels of the engine with respect to a conventional powerplant By retarding the cam timing combustion products which would otherwise be expelled during the exhaust stroke are retained in the cylinder dur ing the subsequent intake stroke The contribution of this diluent to the mixture in the cylinder suppresses NOx for mation but also a ects the mass charge in the cylinders which in turn a ects the air fuel ratio A F response and makes the A F response highly coupled with the cam phasing activity Another important feature of the vari able cam timing is its e ect on the manifold lling dynam ics and ultimately on the engine torque response Figure shows the block diagram of the VCT engine The static and dynamic relations and its interactions are described in