Researchers at Villanova University are developing a new method for optimizing the knock control threshold that significantly improves the performance of a standard knock controller in a spark ignition gasoline engine. Using the method with no modifications other than optimizing the parameters of a standard controller, they have shown that it is possible to operate closer to the knock limit, thereby improving fuel efficiency, emissions, and output torque.
The most recent paper on their work appears in the International Journal of Engine Research.
Knock (spontaneous detonation) results in a shock wave accompanied by high temperatures, pressures, and acoustic resonances within a cylinder. In the worst cases at high speed and load, these can damage the engine; in more moderate cases, knock can sound alarming, and is associated with higher NOx output and lower torque output.
Engine knock thus is an undesirable phenomenon, which requires feedback control in order to maximize engine efficiency and avoid damage to the engine, they note. Traditional approaches to knock control assume that in order to control potentially damaging knock events, it is necessary to use thresholds set to detect such events.
The knock control problem is unlike many process control problems because knock intensity behaves like an independent random process. It is therefore only possible to control some aspect of the knock intensity probability density function, such as the probability of knock events. Traditional knock control strategies regulate knock probability by responding deterministically to every knock event, even if these events are arriving at close to the desired rate. The traditional controller is, therefore, somewhat overactive, with a relatively high degree of cyclic disperson in close-loop spark advance and a mean spark advance, which was significantly retarded relative to the optimum achievable.
By casting the problem in a more stochastic framework, it is possible to design stochastic controllers which allow knock to occur at close to the specified rate, and only to react when the observations of knock events do not match the target in some statistical sense.—Jones et al. (2012)
The proposed new method takes a more stochastic view and sets the threshold such that it maximizes the sensitivity to changes in the knock intensity distribution. In their paper, they showed that optimizing the threshold and controller parameters in the manner proposed results in a controller with fast transient response, improved mean spark advance, and reduced cyclic dispersion.
Peyton Jones, J. C., Spelina, J. M., Frey, J. (2013) Optimizing Knock Thresholds for Improved Knock Control. International Journal of Engine Research doi: 10.1177/1468087413482321
Jones, J. C. P.; Spelina, J. M.; Frey, J. (2012) Likelihood-Based Control of Engine Knock, Control Systems Technology, IEEE Transactions on , vol.PP, no.99 doi: 10.1109/TCST.2012.2229280