Auto-Ignition Control in Spark-Ignition Engines Using Internal Model Control Structure

This paper analyzes the feasibility of applying model predictive control strategies for mitigation of the auto-ignition phenomenon, which affects the performance of spark-ignition internal combustion engines. The first part of this paper shows the implementation and experimental validation of a two-dimensional model, based on thermodynamic equations, to simulate operating conditions in engines fueled with natural gas. Over this validated model, several control strategies are studied in order to evaluate, through simulation analysis, which of these offer the best handling capacity of the auto-ignition phenomenon. In order to achieve this goal, multivariate control strategies are implemented for a simultaneous manipulation of the fuel/air ratio, the crank angle at ignition, and the inlet pressure. The controlled variable in this research is the temperature at the ignition point. This temperature is obtained through an estimation based on pressure in the combustion chamber at that point, which is located toward the end zone of the compression stroke. If the ignition temperature of the fuel–air mixture is reached during the compression process, then auto-ignition takes place. Proposed control strategies consist of maintaining the temperature in the ignition point below the fuel–air mixture auto-ignition temperature. The results show that auto-ignition is difficult to avoid using a single input–single output (SISO) strategy. However, a multiple input–single output (MISO) approach avoids the influence of the phenomenon without a significant impact over the engine's performance. Among the developed strategies, an approach based on model predictive control and feedforward control strategy shows the best performance, measured through the integral absolute error (IAE) index. These results open the possibility of new ways for improving the control capacity of auto-ignition phenomenon in engines compared to currently available feedback control systems.