Surge Detection Using Knock Sensors in a Heavy Duty Diesel Engine

Improving turbocharger performance to increase engine efficiency has the potential to help meet current and upcoming exhaust legislation. One limiting factor is compressor surge, an air flow instability phenomenon capable of causing severe vibration and noise. To avoid surge, the turbocharger is operated with a safety margin (surge margin) which, as well as avoiding surge in steady state operation, unfortunately also lowers engine performance. This paper investigates the possibility of detecting compressor surge with a conventional engine knock sensor. It further recommends a surge detection algorithm based on their signals during transient engine operation. Three knock sensors were mounted on the turbocharger and placed along the axes of three dimensions of movement. The engine was operated in load steps starting from steady state. The steady state points of operation covered the vital parts of the engine speed and load range. The collected data was analysed with the objective of extracting information of a surging or non-surging compressor. In the charging system studied, the knock sensors detected a profound frequency peak between 5.0 Hz to 7.0 Hz. Another surge related frequency component of about 25 kHz was also observed, dependent on the turbocharger speed. Two surge detection algorithms were evaluated, one based on short time Fourier transform (STFT) and one based on the correlation integral (CI). These algorithms where then validated against temperature measurements at the compressor inlet and visual observation of oscillations of the air inlet piping. The surge detection algorithms were compared for accuracy and repeatability. The accuracy of the methods was found to be 73 % and 71 % respectively when compared to the temperature rise in the compressor inlet.