Hierarchical model calibration of the aerospace electronic cabinet based on the Bayesian network

Establishing a reliable model to predict the vibration response of the cabinet is fundamental to vibration isolation and damping design for the aerospace electronic cabinet (AEC). The AEC has multi-level characteristics and is subject to multi-source uncertainties, which pose challenges to accurate modeling of the cabinet. In the paper, a hierarchical model calibration method for the AEC is proposed based on the Bayesian Network (BN). The method divides the model calibration of the cabinet into three levels and integrates multi-source observation data. By strategically reducing the number of parameters to be identified simultaneously while enhancing information integration, the accuracy of model calibration has been significantly improved. The effectiveness of the calibration model is verified through the random vibration test. Results show that spring-damper elements can effectively simulate the dynamic behavior of the AEC with damping pads, and the response of the structure is most sensitive to the stiffness coefficient in the direction perpendicular to the base of the AEC. After model calibration, the max prediction error in the root mean square of the total acceleration (RTA) is <10.8% in all measurement points. Compared with the traditional method of identifying all model parameters simultaneously, the proposed method has improved by 39.8% in RTA. Moreover, the entire response is covered within the uncertainty interval of the predicted results.