Automotive Science and Engineering

One of the factors that affects the efficiency and lifetime of spark ignited internal combustion engine is “knock”. Knock sensor is a commonly used to detect this phenomenon. However, noise, limits detection accuracy of this sensor. In this study, Empirical Mode Decomposition (EMD) method is introduced as a fully adaptive signal-based analysis. Then, based on weighting decompositions, a method for reducing knock signal noise to enhance detection accuracy of knock, has been proposed. Then, the presented method has been evaluated using recorded signals from four engine cylinders. Internal pressure of each cylinder were recorded and used as reference for knock detection. Test results verifies that knock detection accuracy improved by about 11.3%. The results of optimization method were consistent with our expectations and the weights of middle levels are higher than other levels, which means that the proposed method not only extracts the main frequencies of knock, but also assigns reasonable weights to them.