Abstract
Keywords - Genetic algorithm, calibration, hybrids, hydraulic system, parameter tuning
The electronically controlled brake system (ECB) realizes high performance control functions for devices that contribute to improving fuel efficiency and vehicle safety, such as regenerative friction brake coordination, Vehicle Dynamics Integrated Management (VDIM), ABS, and the like. This system is currently applied in various cars, for example, "Prius".
In order to install ECB in other global cars, each component that constitutes the ECB ECU software must be calibrated in accordance with the requirements of the vehicle, in the same way as other vehicle systems like the power train. One of these calibration methods is the servo gain curve tuning process. This procedure, which is conventionally carried out manually, requires repetitions of trial and error. This is because brake calipers have many hardware specifications, the specific demands of each car or target destination, and the many evaluation points and parameters to be modified. Therefore, the demand for automatic calibration has increased to reduce the expense of this tuning.
The proposed method described in this paper is based on a genetic algorithm (GA), which has been constructed in accordance with the servo gain curve ECB calibration procedure. GA is usually adopted in discrete off-line processing, and to apply it to a continuous online fluid system like ECB there are three points should be embedded within the algorithm. First, tuned parameters should be expressed as a genetic pattern. Second, controllability of the servo gain curve should be defined as a scalar quantity according to the demands of the target vehicle, considering the particular behavior of a fluid system. Third, effective modification of the tuned parameters should be carried out according to the defined controllability.
Furthermore, in order to verify the validity of the proposed algorithm, the following four points should be examined. First, the target modified parameters should be convergent toward the desired area within the finite span. Second, the controllability of the automatic tuning should be superior to the controllability of manual tuning. Third, this algorithm should have repeatable results. Fourth, this algorithm should be practical with different calipers that have various characters.
Finally, considering the above requirements and confirmation of the above results, it was verified that the proposed algorithm is capable of computing the desired gain curve automatically a very short period of time.