Development of Advanced Automated Mapping/Calibration Techniques for Complex Engine Control Systems, 03-R9495Printer Friendly Version
Inclusive Dates: 07/01/04 07/01/06
Background - The objective of this project is to develop, implement, and demonstrate an advanced, transient engine calibration technique on a modern, port-injected, spark-ignited gasoline engine. The technique is to be automatic in nature, such that great time and cost savings may be realized when calibrating the engine. The typical automotive spark-ignited engine has only two main actuators: ignition timing and Exhaust Gas Recirculation (EGR) rate. In a move to promote better fuel economy and better performance, additional actuators like cam phasers that control the relative position of the camshaft in relation to the crankshaft are being used. Initially, only one phaser on the intake camshaft was utilized, but more engine manufacturers are moving toward dual camshaft phasers with phasers on the intake and exhaust camshaft. A full factorial calibration with four actuators may take 10 to 20 times as long. Therein lies the need for an automated mapping routine to run the engine, log engine performance parameters, make intelligent decisions, and pick the best operating point in a minimum amount of time.
Approach - A major automotive manufacturer has supplied a V6 engine for use on this project. The engine is representative of the majority of engines available in the marketplace today. It has distributorless ignition, EGR, and a three-way catalyst. This platform represents the current actuator set that the manufacturers must understand to achieve optimum mapping/calibration. It is important to prove the technique with a known platform so that the manufacturers can understand the results. After successful completion, the technique will be applied to an engine with a larger actuator set for evaluation.
An SwRI Rapid Prototyping Electronic Control System (RPECS) will be used to read and record test cell information, control engine parameters, transition the engine from one operating point to the next, and make decisions on which transition is next. Initially, offline models will be used to select the operation condition that best trades off emissions, fuel economy, and performance. As the work progresses, the models will be transitioned to run real-time to further reduce the operating time necessary for calibration.
Accomplishments - The engine that was supplied for the project was installed in a test cell with a low inertia motoring dynamometer. An SwRI RPECS was built to handle all engine control and data acquisition. The automated calibration /optimization control code previously developed on R9323 was modified to work with the engine and available actuator set, and the optimization suite was rewritten to be more generic for application to many engines. One of the main differences between this gasoline application and previous applications on diesel engines is that the gasoline engine can misfire or knock. Code was developed to quickly take a snapshot of the engine conditions when either misfire or knock was encountered, and then move on to protect the engine and data acquisition equipment.
The automated mapping technique was applied to two representative speed/load conditions, one representative of highway cruise and the other moderate acceleration. For each condition, the actuators were swept within a space using a full factorial method to determine the optimum settings. That data were compared to similar data taken using the SwRI automated optimization and calibration technique where the actuators were moved within a similar space, but with fewer set points between boundaries. Using the SwRI method, linear models are fit to the sparse data set and mathematically manipulated to find the optimum. Great timesaving was achieved as the required number of points was cut by two-thirds. Further improvements were shown by finding a near optimum by analyzing data only recorded at the boundaries of the actuator space, and then performing a more refined sweep around the near optimum. The end result was an optimum could be found at each speed/ load condition with about 10 minutes of engine run time.