A natural starting point for the study of any physical system is linearisation–leading to great simplification is terms of analysis, and also, in the computer age, to design flexibility and algorithmic simplification in simulation. The acoustics of musical instruments is no exception. One question, then, is: how much of the behaviour of a given instrument can be linearised? The only clear answer is: definitely not all of it. The production of musical sound by an instrument, whether it is struck, blown, or bowed, relies critically on a nonlinear excitation mechanism. One standard model of the musical instrument, then, relies on a subdivision of the instrument into a nonlinear excitation mechanism, which is to a good approximation lumped, and a linear resonator which is distributed, and characterized by a number of natural frequencies, or modes. Such a model has been employed, for many particular cases, for some time–a powerful unified picture emerged, however, with the article by McIntyre, Schumacher and Woodhouse [1].
https://www.academia.edu/8027847/The_Changing_Picture_of_Nonlinearity_in_Musical_Instruments
