| Popis: |
The imaging of very small acoustic impedance variations in a medium cannot be performed by echography. On the contrary, the transmitted signals are sufficiently sensitive to these variations to allow the extraction of information on the medium. As the amplitude and time of arrival of transmitted signals through an air flow can be determined with sufficient accuracy, this information may be used to realize images. As the parameters of each individual transmitted signal are the result of the whole history of the acoustic ray along its path, an inversion procedure must be performed. Tomography is achieved by the use of multiple acoustic paths networking the area of investigation. The different inversion methods are discussed as well as their tractability in air flow visualization. The technique requires a very accurate determination of the amplitude and time of arrival of the acoustical signal. Although piezoelectric transducers provide easily this precision when exploited in immersion in liquids, their use is notoriously difficult when exploited in air where the huge impedance mismatch requires high excitation voltages and high amplification when pulse mode is used. Furthermore, in order to launch enough energy, the emitting surface of these transducers has to be fairly large. This large surface induces a high directivity impairing the use of piezoelectric transducers as a source emitting toward many receivers. Therefore, we have developed spark emitters perfectly suited to the problem. Very stable acoustic pulses, with a wide frequency band (audio frequencies to 1MHz) are generated. The effective bandwidth in a given experiment is only limited by the viscous damping in air. In that frequency range, the sources are highly non directive due to cylindrical symmetry and the small size (few millimeters) of the linear, sound emitting zone. The receivers are small radius capacitive microphones. Emitters and receivers are placed on a rotating wheel. 1200 acoustic paths are recorded in a tomography experiment. Stationary components of the velocity and turbulence in the flow fields are obtained using an iterative algorithm. The pixel resolution size is of a few centimeters for an investigation zone with a 1.8 m diameter size. The theoretical aspects of the problem and the assumptions made are discussed. |
