Two methods are available for source identification by multiple-input/single-output modeling ; frequency response function(FRF) method and partial coherence function(PCF) method. The former is applicable to a case where the inputs are not correlated to each other while the latter is to a case where the inputs are correlated to some extent. In the second case, however, first of all the priority among the correlated inputs must be determined before the partial coherence technique is applied. This problem can be solved by causality checking between any two correlated inputs.
The causality checking idea is briefly explained in this thesis and applied to the correlated inputs due to measurement interference. The sources are identified using both FRF and PCF methods for the engin room noise of a passenger car. Inputs are measured by accelerometers on several components inside the engin room and also by microphones near the same components. The sound pressure level at the driver's seat is used as the output. PCF method is applied to the sound pressure measurements since they are correlated to each other to some extens. FRF method is used in case of the acceleration measurements since they are little correlated. In both cases, the contributions of each source are investigated and identified qualitatively as well as quantitatively.