Acoustic imaging in enclosed spaces

摘要: This thesis is concerned with the problem of noise source identification in closed spaces. The main motivation was to propose a technique which allows locate and quantify sources within industrial vehicles, time-effective manner. In turn, might be used by manufacturers for abatement purposes such as provide quieter vehicles. A simplified model based on equivalent formulation tackle problem. It shown that ill-conditioned, sense it very sensitive errors measurement data, thus regularization techniques were required. detailed study this issue, particular tuning so-called parameter, importance ensure stability solution. particular, Bayesian criterion robust approach optimally adjust parameter an automated way. target application concerns large interior environments, imposes additional difficulties, namely: (a) positioning array inside enclosure; (b) number unknowns ("candidate" sources) much larger than positions. An iterative weighted then proposed overcome above issues by: first correct enclosure second iteratively solve order obtain quantification. addition, has provided results enhanced spatial resolution dynamic range. Several numerical studies have been carried out validate method well evaluate its sensitivity modeling errors. affected non-anechoic conditions, reflections are identified "real" sources. post-processing helps distinguish between direct reverberant paths discussed. last part experimental validations practical applications method. custom spherical consisting rigid sphere 31 microphones built tests. academic semi-anechoic illustrated advantages limits Finally, tested application, consisted identifying bus at driving conditions.