Multiple Misfire: Detection and Cylinder Isolation Based on Engine Speed Measurement

The diagnosis of misfire events (or missing combustions) is enforced by On-Board Diagnostics regulations (such as CARB OBD II or European OBD) over the whole engine operating range, for all vehicles equipped with spark ignition engines. Such regulations define both the minimum misfire frequency that is to be detected (related to catalyst damage and/or increased hydrocarbons emissions), and the various misfire patterns that the diagnostic algorithm should be able to detect. In particular, single (no more than one missing combustion per engine cycle) and multiple (more than one misfiring cylinder within the same engine cycle) misfire patterns are to be diagnosed, and the cylinder in which the misfire took place is to be isolated only when single misfires take place (cylinder identification is still not mandatory for multiple misfires). Various single misfire detection methodologies have been successfully developed in recent years (mostly based on the engine speed signal), and this type of misfire diagnosis is still challenging for engines with a high number of cylinders, especially during operating conditions characterized by high engine speed and low load. On the other hand, the detection of multiple misfires is still difficult even for the typical four cylinder engine, since their effects on the engine speed trend have not yet been clarified. In fact, a misfire occurrence is characterized by a sudden engine speed decrease and a subsequent damped torsional vibration. In case of multiple misfires, the engine speed oscillation induced by the first misfiring cylinder may still be present when the second missing combustion takes place, and the resulting engine speed waveform may be erroneously interpreted by the diagnostic algorithm, thus resulting in the improper cylinder being identified or missed detection of a misfiring cylinder. This paper deals with the identification of a specific pattern in the instantaneous engine speed trend, induced by a missing combustion and characteristic of the system under study, that allows performing the desired multiple misfire detection. The methodology has been designed in order to be run on-board, thus requiring low computational power and memory allocation. Its implementation has shown that false alarms can be avoided and correct cylinder isolation is possible, also in presence of multiple misfires. Experimental tests have been performed on a 1.2 liter spark ignition engine mounted in a test cell. Various multiple misfire patterns have been induced by controlling ignition and injection of the various cylinders. In-cylinder pressure signals have been acquired together with the instantaneous engine speed, in order to verify the capability of the methodology.