Gasoline engines can be affected, under certain operating conditions, by knocking combustions, which can result in serious engine damage. Specific power and efficiency are influenced by factors such as compression ratio and spark advance regulation, that modify the combustion development over the crank angle: the trade-off between performance and the risk of irreversible damages is still a key factor in the design of both high-performance (racing) and low-consumption engines. New generation detection systems, especially based on ionization current technology, allow aggressive advance mapping and control, and future equipment, such as low-cost in-cylinder pressure transducers, will allow following that trend.
Also HCCI (Homogeneous Charge Compression Ignition) engines need a sophisticated combustion monitoring methodology, since increasing BMEP levels in HCCI mode force the combustion to approach the knocking operation.
Many methodologies can be found in the literature to recognize potentially dangerous combustions, usually based on the analysis of accelerometer, in-cylinder pressure or ionization current signals. Signals are sampled with high sample rates, then filtered, for a clear recognition of the phenomenon. Filtered signals can then be used to define damage-related indexes, by means of various types of mathematical operations. The indexes are then compared to pre-defined thresholds, for the diagnosis of dangerous combustion events.
Thresholds setting is a challenging task, since indexes are not usually intrinsically related to the damages caused by abnormal combustion events. Furthermore, the indexes values usually strongly depend on the engine operating conditions (speed and load), and thresholds must therefore vary with respect to speed and load. It can be said that indexes generally depend on the combustion development over the crank angle, and not necessarily on the knock phenomenon. This means that the index value is also influenced by the spark advance regulation, even without knock.
This paper shows why commonly used indexes values vary with engine running conditions, and how raw indexes can be modified in order to obtain an operating conditions independent information. In-cylinder pressure data are analyzed both in the time and frequency domains, in order to show how parameters such as window angular position and width, and band-pass filter characteristics influence the filtered in-cylinder pressure signals, and, as a consequence, knock indexes. Such parameters can affect both the signal to noise ratio and the operating condition dependence, thus they must be carefully optimized.
Once operating conditions effects on knock indexes are known, they can be taken into account for indexes evaluation, letting emerge the plain knock effect. This compensation can be carried out in many ways: in the paper two possible methodologies are considered and compared. Both of them perform an index normalization, the first one in the time domain, the second one in the frequency domain.
The frequency-domain normalization methodology proves to be efficient in finding operating conditions-independent knock indexes, allowing an a significant reduction of the calibration time. The time-domain methodology is more influenced by the window choice