Tanya Morton, Richard Connors, Pete Maloney, David Sampson
1. Einleitung/ Abstract
Modern engines have an increasing number of control parameters that are having a dramatic impact on calibration time. Up-front prototyping of calibration processes can be used to reduce the burden of increasing powertrain complexity, however, such techniques are still in their infancy in the automotive field. In this paper we illustrate how flexible model-based calibration tools can be used to prototype a calibration process for an advanced engine type. Specifically, we produce optimal calibration tables for intake and exhaust cam timings that trade off brake specific fuel consumption against NOx emissions. The data used is from a GT-Power model of a production 2.2L naturally aspirated 4-valve overhead-cam spark ignition engine, modified for twin-independent variable valve-timing capability. All analysis techniques used are available in version 2.0 of the Model-Based Calibration Toolbox.
In recent years calibration time has been increasing due to the need to calibrate more advanced types of engines. To reduce this calibration time many companies have turned to a model-based calibration process combined with test bed improvements to allow automation of the data collection process. An additional way of reducing calibration time is to move some calibration tasks to earlier in the design process. Calibration tasks traditionally occur only on the right-hand side of the V design process shown in Figure 1. By performing calibration tasks in parallel with the design tasks on the left-hand side of the V design process, significant time savings can be made. A calibration process can be prototyped and process-related problems resolved at a stage when they are much less costly to correct. Informed decisions on the design of experiment, model type and optimisation routines expected can be made, allowing for early preparation of initial test plan templates, model templates and optimisation scripts, ready for when the actual engine arrives.
Figure 1: V design process for control development
In this paper we will produce calibration tables for a dual-independent variable valvetiming engine to illustrate a model-based calibration process. The data is collected from a GT-Power model. Physical models, such as those created by GT-Power can be helpful to gain an understanding of engine behaviour early in the design process and to identify data-gathering requirements before test. However, physical models can be timeconsuming to run, and this can limit the amount of analysis that can be performed with them. In this paper a statistical emulation of the physical model is produced. The statistical model evaluates much more quickly than the physical model or an engine test, typically in hundreds of microseconds. This opens the door to a greater amount of optimisation and analysis – all of which can be performed before the actual engine is available.
The problem addressed in this paper is to produce optimal calibration tables for the three main control parameters: spark advance, intake cam phase, and exhaust cam phase. The inputs to the tables are engine speed and load. Here, load is defined in the typical way as a fraction of the maximum cylinder air charge possible at a given RPM, and based on measured airflow. The values of spark advance and cam positions will be chosen to trade off brake-specific fuel consumption (BSFC) against the amount of engine-out NOx produced, subject to upper limits on catalyst-in exhaust temperature, and intake manifold pressure. The upper bound on intake manifold pressure ensures that there is sufficient vaccum in the intake manifold to allow the use of a brake-servo (booster). The bound on exhaust temperature is to prevent catalyst overheating.
Design of Experiment and Data