Nicolas LOPES, will defend his PhD work entitled: “Passive approach for the modelling, the simulation and the study of a robotised test bench for brass instruments.”
This work took place in the Project-Team Sound Signals and Systems and in the Analysis-Synthesis et Instrumental acoustics teams of Laboratoire Sciences et Technologies de la Musique et du Son, IRCAM-CNRS-UPMC.
It is part of two ANR (French National Research Agency) projects: ANR-HamecMopsSys and ANR-Cagima.
This work will be defended to the following jury:
Brigitte d’Andréa-Novel Rapporteur – Professeur Mines-ParisTech
- Christophe Vergez - Rapporteur – Directeur de Recherche, CNRS LMA
- Benoît Fabre - Examinateur – Professeur, LAM, Institut d’Alembert, UPMC
- Isabelle Terrasse - Examinateur – Directrice de Recherche, Airbus Group Innovations
- Bernhard Maschke - Examinateur – Professeur Université Claude Bernard, Lyon 1
- Thomas Hélie - Directeur de thèse – Chargé de Recherche, Laboratoire STMS, CNRS
- René Caussé - Co-directeur de thèse – Directeur de Recherche, Laboratoire STMS, Ircam
Abstract
This thesis is to be seen against the robotic, the automatic, and the musical acoustics backgrounds. It provides a study of a robotised test bench for brass instruments. This study is divided into three parts: the passive modelling of the system, its simulation and its development. The modelling is done using a passive formalism, namely, the ports-Hamiltonian systems. The main parts of the complete system are: an air supply for the breath, an acoustic exciter itself composed of a couple of artificial lip and an air jet, and an acoustic resonator. In this work, the acoustic resonator is a valve trombone. A new model for the air jet generated between the lips is proposed. This model aims at providing a power balance, which is closer to the real system than other commonly used models. Refinements are added to the jet model to obtain a self-oscillating complete model.
The discrete gradient method is presented to perform simulations. This method offers a discrete time description that verifies the power balance, and then the passivity during simulations. However, it does not generally guaranty the existence and uniqueness of a solution. Moreover, it is limited to the second order of numerical consistency, and its execution needs nonlinear optimisation algorithms, that are time-consuming processes. To compensate for these limitations, a multi-stage method of double Runge-Kutta type and based on a change of variable is proposed. Results from simulations are interpreted and compared to those coming from a Bernoulli type model. Finally, the test bench and the technical developments carried out in this thesis are presented. These developments are both about programming and mechanic. They enable the performance of repeatable cartographies experiments which can be used to characterise music instruments. Experimental and numerical results are compared. Comparisons enable the highlighting of the defaults and the qualities of the proposed model and lead to future choices for the modelling and the development of the test bench.