Study of real-time stochastic knock detection and control for spark-ignition engines

The ability to operate a spark-ignition (SI) engine near the knock limit provides a net reduction of engine fuel consumption. This work presents a real-time knock control system based on stochastic knock detection (SKD) algorithm. The real-time stochastic knock control (SKC) system is developed in Matlab Simulink, and the SKC software is integrated with the production engine control strategy through ATI's No-Hooks. The SKC system collects the stochastic knock information and estimates the knock level based on the distribution of knock intensities fit to a log-normal distribution. A desired knock level reference table is created under various engine speeds and loads, which allows the SKC to adapt to changing engine operating conditions. For the steady-state engine operation, a fixed-length weighted discrete FIR filter is used to estimate the knock factor (KF), an indicator of knock intensity level. Both offline simulation and engine dynamometer test result show that stochastic knock control with fixed length FIR filter has slow and excessive retard issue when significant knock event happens. To enhance the knock control response, an integrated feed-forward and feedback knock control strategy is employed. For the heavy knock events, a combination of gain scheduling and a fast retard (FR) is applied based on the detected KF. In addition, a variable length FIR filter is used to reduce the KF estimation time. The performance of the developed knock detection and control system is then evaluated on a V6 3.5L turbocharged engine on a dynamometer test stand.