On the Performance of Unconventional Knock Control Variants With Time-Dependent Gains

Recently, an interesting new type of classical knock control strategy has been proposed which adapts the controller gains based on the number of cycles since the last knock event. In this work, the performance of such strategies is critically assessed using both time-history simulations as well as new analytical stochastic methods which are developed for this purpose. Initial results suggest that the particular time-dependent adaptation laws (logarithmic advance and likelihood-ratio-based retard) offer improved steady-state performance with respect to a typically tuned standard advance-retard strategy. However, further investigation shows that this performance is not robust and that the transient response time in particular can degrade rapidly in the presence of disturbances.