Performance analysis of artificial neural networks for control in internal combustion engines

In the last years the Electronic Control Units (ECUs) technology has evolved both in terms of hardware and firmware. The tasks requested to ECUs are, as a matter of fact, ever more challenging, due to the growing complexity of the internal combustion engines control, causing increasing needs in terms of computational speed, as well as memory amount. In order to limit look up tables and white box models in ECUs, also Artificial Neural Networks (ANNs) were proposed for the use in different types of tasks, as On-Board Diagnosis, virtual sensors and ECU look up table replacement. ANNs showed a good performance in all these cases. In this paper, the application of ANNs as virtual sensors is further analyzed, in particular in order to verify their performance when used to set-up multiple sensors by using a single ANN module. In this work ANNs were used to evaluate pressure and flowrate in different points of a lubrication circuit of bench and connecting rod bearings.