Machine learning applied to the prediction of trumpet bifurcation diagrams: towards a tool for trumpet designers - Archive ouverte HAL Accéder directement au contenu
Communication Dans Un Congrès Année : 2023

Machine learning applied to the prediction of trumpet bifurcation diagrams: towards a tool for trumpet designers

Résumé

This work aims to develop a fast and easy-to-use program for the prediction of brass instrument bifurcation diagrams, with minimal supervision from the user. Using numerical continuation, more than ten thousand bifurcation diagrams are generated to train a machine learning model with trumpet impedances as inputs, and descriptors associated with the bifurcation diagrams as outputs. Our approach is based on the definition of virtual players and virtual trumpets to generate the training data. Different regression models are then considered and their performance is compared. The model finally selected shows high speed and great accuracy in predicting the descriptors. Moreover, the regression approach includes regularization which promotes sparsity, hence improving the interpretability of the model. This program then constitutes a potential tool for music instrument designers to easily predict the dynamical behavior of numerical prototypes.
Fichier principal
Vignette du fichier
FA2023___Version_r_vis_e (1).pdf (1.62 Mo) Télécharger le fichier
Origine : Fichiers produits par l'(les) auteur(s)

Dates et versions

hal-04262291 , version 1 (27-10-2023)

Identifiants

  • HAL Id : hal-04262291 , version 1

Citer

Vincent Fréour, Mimoun Mohamed, Keita Arimoto, Valentin Emiya, Bruno Cochelin, et al.. Machine learning applied to the prediction of trumpet bifurcation diagrams: towards a tool for trumpet designers. Forum Acusticum 2023 - 10th Convention of the European Acoustics Association, European Acoustics Association; Acoustical Society of Italy, Sep 2023, Torino, Italy. ⟨hal-04262291⟩
50 Consultations
25 Téléchargements

Partager

Gmail Facebook X LinkedIn More