Predicting the optimum spark advance in a spark-ignition engine using artificial neural network

Spark advance is an important operating variable in a spark-ignition engine that affects nearly all engine outputs. Therefore, an accurate prediction of optimum spark advance (OSA) at a given operating condition is important for optimal engine performance. In the present work, a single-cylinder gasoline-fueled spark-ignition engine is operated over a wide range of loads with different equivalence ratios. The measured data is then used to predict the optimum spark advance for a given operating condition using artificial neural network (ANN). The ANN model is developed based on 29 operating points. Randomly, 80% of data was used to train the ANN model using Levenberg-Marquardt algorithm. In order to obtain best performance, number of neurons and transfer function of the hidden layer were changed. The ANN model, incorporated logarithmic sigmoid function in the hidden layer with 34 neurons, showed the best performance - with mean square error and correlation coefficient of 0 and 1, respectively. The OSA of remaining 20% of data was determined using the ANN model. It was found that the ANN model compared well with the measured data at different operating conditions.