マルチデータウインドウ化による管長計測の高精度化

抄録

In the field of measurement, we often encounter the situations where pipe length must be measured indirectly. The authors previously proposed a pipe length measurement system using stationary waves generated in straight pipes. In these papers, sound fluctuations were modeled as an output of a linear dynamic system. As the system includes unknown parameters such as pipe length and observation variance, adequate candidates were given to the set of unknown parameters and their posteriori probabilities were calculated using a bank of Kalman filters and Bayes' rule. The pipe length was thus measured as the expectation of the pipe length with the probabilities.
To realize a much more accurate measurement, it is considered to lengthen the data window on which the bank of Kalman filters are used. Taking this approach, however, requires many unknown parameters and consequently increases computational time, because attenuation in each mode of the stationary wave must be taken into account. The paper proposes to divide the data window into several subwindows where the attenuation is negligible, and applies the pipe length measurement proposed before.