The operating parameter optimization of spark duration effect on the performance and emission characteristics of direct-injection propane by genetic algorithm

It is challenging to optimize of the effect of spark discharge duration on low carbon direct injection combustion of propane from a specific aspect of the internal combustion engine. The strategy in this paper combines a genetic algorithm (GA) with an artificial neural network (ANN) to enhance efficiency of a large bore diesel engine modified with spark plug and fueled with propane direct injection under various spark timing durations and identify its engine performance parameters and corresponding conditions for operational control purpose. In-cylinder performance experiments were used to create 756 datasets for training, testing, and validating the ANN with 16 neurons on each hidden layer, respectively. The selected performance parameters were the heat release rate, turbulent kinetic energy, indicated mean effective pressure, and indicated thermal efficiency. The NOx and CO emissions are improved by 7.14 % and 0.74 %, respectively, by the optimized ANN model. In addition, compared to an experimental result, the model improves the indicated thermal efficiency, heat release rate, and indicated mean effective pressure by 1.35 %, 0.37 %, and 0.04 %, respectively. The combined of ANN-GA is effective in identifying the optimal operating parameters for in-cylinder performance and emission.