The Application of Singular Value Decomposition to Determine the Sources of Far Field Diesel Engine Noise

The identification of the dominant noise sources in diesel engines and the assessment of their contribution to far-field noise is a process that can involve both fired and motored testing. In the present work, the cross-spectral densities of signals from cylinder pressure transducers, accelerometers mounted on the engine surface, and microphones (in the near and far fields), were used to identify dominant noise sources and estimate the transfer paths from the various “inputs” (i.e., the cylinder pressures, the accelerometers and the near field microphones) to the far field microphones. The method is based on singular value decomposition of the input cross-spectral matrix to relate the input measurements to independent virtual sources. The frequencies at which a particular input is strongly affected by an independent source are highlighted, and with knowledge of transducer locations, inferences can be drawn as to possible noise source mechanisms. It was found that results from an analysis of fired test data were similar to those from motored test data: thus it is possible to reduce or eliminate motored testing. Also, since the transfer paths between the near- and far-field measurements can be estimated, convolutions of the input time histories and the transfer path impulse response functions can be combined to yield accurate output predictions. Furthermore, the contribution from each input measurement reveals which noise sources contribute significantly to the far field. Conclusions can then be drawn regarding source locations and spectral characteristics.