In this, Part 1 of the paper, the sound generation of a diesel engine is modelled based upon the combustion process, and time-frequency analysis is used to reveal the underlying characteristics of the sound waves. Simulation shows that the frequency bandwidth of the generated acoustic signals is significantly widened around the engine's top dead centre (TDC) positions, with the energy concentrated predominantly at the firing frequency and its harmonics. As anticipated, the model predicts an increase in sound level with increasing engine load and speed, and the model-predicted noise generation is correlated with waveforms extracted from intrusively-monitored cylinder pressure.Real monitored data, taken in an ordinary engine test-bay environment and without special acoustic monitoring precautions, is shown to be highly contaminated due to adverse environmental acoustics and intrusive background noise. The representation of acoustic signals using the smoothed pseudo-Wigner-Ville distribution (SPWVD) and continuous wavelet transform (CWT), however, is found to permit recognition of the adverse influences of the measurement environment. This subsequently allows the monitored sound characteristics to be closely correlated to the combustion process. Part 1 concludes with an investigation of the influences of the measurement environment upon the acoustic data, and of the signal conditioning and representation techniques required to reveal the condition-indicating content of the monitored acoustic data.The sister paper to this (“Part 2 - Fault Detection and Diagnosis”), puts the developed methodology to the test by investigating its capability to detect and distinguish between a range of realistic yet incipient engine faults on a standard production engine in an uncontrolled industrial environment.