Combustion noise and piston slap noise are major part of engine's noise source, and thus the identification, separation and quantification of them are of farreaching significance to evaluating and controlling engine's noise. However, it is difficult to separate all kinds of noise sources from each other due to the aliasing problems existing in both time and frequency domain. Therefore, this paper explores the separation of engine combustion noise and slap noise based on a multi-input/single-output model and a partial coherence algorithm. Firstly, the influences on separation performance of the partial coherence algorithm brought by various noise sources' aliasing level and different signal-to-noise ratio are investigated with simulation signals and results show that the partial coherence algorithm can efficiently eliminate the mutual coupling effects between inputs. Besides, it can also effectively separate the parts generated from a single input as well as the ones caused by the interaction of multiple inputs. Finally, the algorithm is applied when using a diesel engine's measured signals to separate the combustion noise, piston slap noise, etc., and then the structural response function of combustion noise is further computed in order to obtain the main controlling frequency range.