Today, 99% of the two wheelers in India operate with carburetor based fuel delivery system. But with implementation of Bharath Stage VI emission norms, compliance to emission limits along with monitoring of components in the system that contributes towards tail pipe emissions would be challenging. With the introduction of the OBD II (On-Board Diagnostics) and emission durability, mass migration to electronically controlled fuel delivery system is very much expected. The new emission norms also call for precise metering of the injected fuel and therefore demands extended calibration effort.
The calibration of engine management system starts with the generation of pre-calibration dataset capable of operating the engine at all operating points followed by base calibration of the main parameters such as air charge estimation, fuel injection quantity, injection timing and ignition angles relative to the piston position. Finally, the vehicle calibration is executed keeping drivability and compliance to legislative norms as prime requirements. The quality of the pre-calibration data and base calibration decides the number of iterations required to arrive at the final dataset that meets the emission targets. Currently, the pre-calibration data is ported from datasets belonging to engines of similar displacement calibrated before; as a result of which the data do not fit well at all engine operation points.
This paper elucidates a model based approach that generates pre-calibration dataset closest in match to the dataset obtained after base calibration at engine dynamometer using limited measurement logs from the engine. This is achieved through modelling the system using identified geometrical information of the engine, intake and exhaust systems and then introducing the physics of engine operation into it. Using the geometrical information, MATLAB based models are built to calculate the critical parameters like pressure drop across air filter, resonant frequency of the Helmholtz resonator in the intake path, throttle and valve flow coefficients and friction torque. The output of these individual MATLAB models are then fed into a predictive model that estimates the combustion parameters. These in turn serve as inputs to a one dimensional engine model built in GT Suite which then predicts the air charge entering the cylinder, optimum ignition angles, brake torque and exhaust gas temperature at the manifold.
A case study was done with a 200cc air cooled engine as reference, for which the outcome of the GT Suite model is compared against the actual calibration dataset. The model is found to predict the air-charge at an accuracy of 85%, optimum ignition angles within ± 4.5° CA, brake torque at 85% accuracy and exhaust temperatures within ±20° C.