SUMMARY

The future development of engineered products will require a blend of technical knowledge from multiple engineering domains that meet its relevant multidisciplinary design criteria with sufficient accuracy. However, modeling and simulating multidisciplinary engineering systems are challenging due to complexities such as interactions between various input parameters. Moreover, in order to accurately estimate risk and reliability of such complicated systems critical input parameters and the corresponding uncertainties must be correctly captured and propagated. Multidisciplinary engineering systems often require accurate representations of multivariate phenomena. Thus, it is essential to develop a framework that can handle multivariate phenomena of complex engineering systems under uncertainty.

The proposed research will develop a framework that can accurately capture and model input and output parameters under uncertainties for multidisciplinary systems. Specifically, the Artificial Neural Network (ANN) with Principal Component Analysis (PCA) and the Auto-Encoder (AE) algorithm will be developed to handle this issue. The Independent Features Test (IndFeaT) to select a critical subset of model features will also be utilized when using the Probabilistic Neural Network (PNN). In addition, a copula function will be employed to accurately model input uncertainties. The proposed method permits complicated and multiple properties to be represented effectively and realistically, leading to accurate response predictions. To demonstrate the efficacy of the proposed method, experimental results of a cantilever beam test and electro-mechanical systems such as a solder joint and stretchable patch antenna will be estimated.