Zhen Hu

Confidence in the Prediction of Unmeasured System Output Using Roll-Up Methodology

This research is concerned with how to use available experimental data from tests of lower complexity to inform the prediction regarding a complicated system where no test data is available. Typically, simpler test configurations are used to infer the unknown parameters of an engineering system. Then the calibration results are propagated through the system model to predict the uncertainty in the system response. However, it is important to note that parameter estimation results are affected by the quality of the model used to represent the test configuration. Therefore, it is necessary that the model of the test configuration be also subjected to rigorous validation testing. Then the calibration and validation results for the test configurations need to be integrated to produce the distributions of the parameters to be used in the system-level prediction. Such a systematic roll-up methodology that integrates calibration and validation results at multiple levels of test configurations has been previously established (Sankararaman and Mahadevan, Reliab Eng Syst Saf 138:194–209, 2015).

Bayesian Network Learning for Data-Driven Design

Bayesian networks (BNs) are being studied in recent years for system diagnosis, reliability analysis, and design of complex engineered systems. In several practical applications, BNs need to be lea...