Accurate measurement of high-frequency blast waves through dynamic compensation of miniature piezoelectric pressure sensors

摘要

To suppress the parasitic effects in explosion test, a blast wave pressure sensor component composed of a pressure sensor, an additional mechanical structure and silicone greases on the sensitive surface is built up for blast wave measurements. If the dynamic characteristics of the sensor component cannot meet the requirements of so-called undistorted measurements, it will result in measurement errors. In this paper, we present a dynamic compensation approach which considers all the components within the frequency range of interest to correct measured blast wave pressure signals. Via the frequency characteristic analysis of the sensor component, we know that the additional mechanical structure and silicone greases introduce a formant within the low frequency range. The amplitude of the formant within the low frequency range is much smaller than that at the resonance frequency, making it difficult to establish the model of the sensor component to reflect this small formant accurately. Therefore, based on dynamic calibration on a shock tube, considering that the effective bandwidth of blast wave pressure signals is generally within 100 kHz, we propose a partial modelling method and establish the partial model of the sensor component. Compared to the full-bandwidth model, the partial model is more accurate within the low frequency range. Then a dynamic compensation approach is presented for accurate measurements. When designing the dynamic compensation filter, all the poles and zeros on the dynamic characteristics of the sensor component are analyzed as influence factors. Using the presented dynamic compensation approach, the working bandwidth of the sensor component is increased from 14.5 kHz to 120 kHz. Compared with the step response signal measured in the dynamic calibration experiment, the step response signal passing through the dynamic compensation filter contains less noises and smaller overshoot. Furthermore, comparing the peak overpressures of the simulated b