Misfire detection on internal combustion engines using exhaust gas temperature with low sampling rate

Abstract

A novel misfire detection method for internal combustion engines at plants is presented. The main aim of the proposed method is focused on statistical detection of misfire as a precursor of ignition system malfunctions. Use of conventional exhaust temperature sensors installed in exhaust manifolds is proposed. Characteristic waveforms that appear in the measured sensor output are analyzed and classified as signs of misfire or normal fluctuations. A key characteristic of the proposed method is the use of an algorithm for discrimination of waveforms that involves a set of if/then empirical rules about waveform features. Parameter setting methods using laboratory or field engines are also presented. Parameter determination is carried out so as to maximize successful detection ratio (SDR) under the constraint of fixed allowable false alarm ratio (FAR). In case of an example using a field commissioned engine, an SDR as high as 75% (that is, three misfire events out of four) was attained, while FAR was 1.9 × 10 −7 , corresponding to three false alarms during a half year of continuous engine operation. The FAR is reasonably acceptable and the SDR is sufficient to allow condition trends to be understood in order to conduct decision-making in connection with preventive maintenance. ►Misfire detection method of internal combustion engines is presented. ►Use of conventional exhaust temperature sensors is proposed. ►Waveform classification using if/then rules is a key factor. ►Data-driven parameter optimization method is shown. ►The method allows statistical detection of misfire as a precursor of malfunctions.