|Institution:||The Ohio State University|
|Keywords:||Mechanical Engineering; diesel engine; inverse-distance interpolation; calibration; closed-loop combustion control; set-point generation; kernel-based interpolation|
|Full text PDF:||http://rave.ohiolink.edu/etdc/view?acc_num=osu1253553419|
Closed-loop control of combustion is of great importance for conventional diesel engines in order to reduce the deterioration in engine performance and emissions caused by different sources of variability. Diesel engine emissions and performance are affected by variability in the combustion root-cause variables like air mass, residual mass, injection parameters etc. If combustion can be referenced based on the root-cause combustion variables, the engine performance and emissions can be improved. Such a referencing, however, leads to increased calibration effort for the combustion controller due to increase in the number of scheduling variables for generating such references as well as due to the increase in the number of references. Conventional methods for generating the set-points for the closed-loop combustion controller are not suitable as they become impractical as the number of scheduling variables increases. In this work, inverse-distance based interpolation methods have been developed to generate the set-points for the closed-loop combustion controller. The inverse-distance interpolation method provides the advantage of reduced calibration effort and can be extended to multiple-dimensional interpolation without significant increase in computational effort. In this work, a closed-loop combustion control architecture has been developed for a heavy-duty diesel engine. The inverse-distance based calibration method has been demonstrated for a simplified version of the closed-loop combustion control architecture. The method involves an optimization approach that generates a scattered set of engine operating points where the engine should be calibrated. The inverse-distance interpolation method can interpolate with the scattered calibration data set to generate the set-points for the closed-loop combustion controller. The proposed method showed a great potential to improve the calibration effort when compared to conventional calibration methods that can be applied to the set-point generation problem. The inverse-distance based calibration approach resulted in a calibration effort that is 5 to 25 times less than that of a conventional method and resulted in engine performance and emissions that are comparable to those of the conventional method.