Experimental and multivariate analysis methods for sound quality evaluation of diesel engines

Sammanfattning: Noise control and sound quality analysis are important, since noise has been registered to be a predominant factor in stress and a source of great annoyance. Traffic noise is a problem and a major part of this noise comes from heavy vehicles. The only legislative requirement for heavy-duty trucks regarding noise emissions, is that the noise level does not exceed an Aweighted sound pressure level of 80 dB. The specification of an A-weighted sound pressure level is, however, not an adequate description of psychoacoustic annoyance and therefore work towards defining a better description of loudness is one of the principal fields of acoustics today. Sound radiation from trucks is speed-related. At medium and high speeds, the overall noise level is comprised mainly of the tyre noise, whereas at low speed and during acceleration, exhaust noise and noise from the engine and transmission structure are predominant. In front of the truck, the noise from the engine and especially that from the timing transmission cover, the torsional damper and the oil sump, comprises a greater proportion of the total noise. The aim of sound quality analysis of diesel engines is to find cost-efficient methods of reducing sound radiation and of changing the character of the sound in order to minimise annoyance. This thesis concerns the development of experimental methods for analysing the sound quality of diesel engines, and focuses on measurement of acoustic intensity, multivariate data analysis, structural modification and subjective assessment of engine noise. The applicability of the FFT-based sound intensity method is evaluated. It is found that the intensity measurements may be influenced by high reactivity, interference due to partlycoherent sources, difficulties in performing the spatial average, real-time limitations and engine speed variations. Scanning the intensity probe, preferably by a robot, is necessary when measuring within narrow bands to avoid interference problems. Scanning achieves more reliable estimates of sound power and intensity vectors. Experimental design and the multivariate techniques, principal components analysis (PCA) and partial least squares (PLS) were utilised to facilitate interpretation of intensity measurements. The results show that PCA and PLS enable independent phenomena in the sound field to be extracted and which can thereby be visualised by principal spectra and principal radiating patterns. The characteristics of sound radiation are determined by designed experiments, sound intensity measurements and operational deflection shape estimations. These methods enable the effects on sound radiation of structure modifications to be predicted. An annoyance index for in-line 6-cylinder diesel engines in stationary running conditions was developed using multivariate statistics. The index is based on engine sounds resulting from structure modifications and changes in fuel. The annoyance level was measured during listening tests of sound stimuli recorded in stereo and reproduced by loudspeakers under anechoic conditions. The different sound stimuli were ranked using paired comparisons or the method of successive intervals. It was found that 94% of the variance of annoyance can be explained by a model based on loudness (Sone), sharpness (Acum) and harmonic ratio (rumble). Impulsiveness, roughness and tonality were other important criteria used in the study and which were found to have a relationship with specific speed ranges. The annoyance was minimised by an increase in stiffness in the lower part of the engine achieved by using a ladder frame in combination with a bearing beam.