An improved reconstruction method of the reflected dynamic pressure in shock tube system based on inverse sensing model identification

The shock tube is capable of producing approximately ideal step pressure from the reflected shock wave and has been widely used in the dynamic calibration of pressure sensors in the aerospace industry. However, the measurement accuracy of the reflected dynamic pressure (RDP) is limited by the vibration effect and the unreasonable model which is established by the ideal gas theory, leading to the dynamic calibration results with large errors. In this paper, an improved reconstruction method of the RDP in a shock tube is proposed based on the identification of the inverse sensing model. The RDP is detected first by a reference pressure sensor to achieve the dynamic response signal and its normalized result. Based on adaptive decompositions of the normalized response signal and mode clustering strategies, the ringing component, trend component, and noise components caused by end -wall vibration are subsequently extracted to construct the datasets of the inverse sensing model of the reference pressure sensor. A hyper -parameters optimized bidirectional long short-term memory network (HPOBiLSTM) is finally applied to establish the inverse sensing model to achieve the accurate reconstruction of the RDP in the shock tube. Simulations with different -order pressure sensors and actual experiments with different pressure conditions are carried out to verify the performance of the proposed method. The results show that the proposed method is able to effectively reduce the influence of the ringing characteristic of the pressure sensor and the end -wall vibration on the RDP reconstruction. Furthermore, comparative experiments also demonstrate the superiority of the proposed method over both the Mach number method and the signal filtering method for RDP reconstruction in the shock tube system in both accuracy and applicability.