A diesel engine electrical generator set (‘gen-set’) was instrumented with in-cylinder pressure indicating sensors as well as a nearby microphone. Conventional jet fuel plus high (Cetane Number CN55) and low (CN35) secondary reference fuels were operated during which comprehensive engine and acoustic data were collected. Fast Fourier Transforms (FFTs) were analyzed on the acoustic data. FFT peaks were then applied to machine learning neural network analysis with MATLAB based tools. Detection of the low and high cetane fuel operation was audibly determined with correlation coefficients greater than 98% on test data sets. Further, unsupervised machine learning Self Organizing Maps (SOMs) were produced during normal-baseline operation of the engine with jet fuel. Application of the high and low cetane fuel operational acoustic data was then applied to the normal SOM. The quantization error of various fueled acoustic data showed clear statistical differentiation from the normal baseline jet fueled operational data map. This unsupervised SOM based approach does not know the engine degradation behavior in advance, yet shows promise as a method to monitor and detect changing engine operation. Companion in-cylinder combustion data additionally shows the changing nature of the engine’s combustion with the different cetane fuels tested.
