Towards a Probabilistic Framework for Integrated Computational Materials Engineering

Abstract

Integrated Computational Materials Engineering (ICME) “has been defined as the integration of materials information, captured in computational tools, with engineering product performance analysis and manufacturing process simulation”. As with other multiphysics, multi-scale modeling approaches, ICME deals with uncertainty. Managing uncertainty is critical to the effective management of risk and is critical to explaining the benefits of the ICME modeling approach. So, to successfully fulfill ICME’s promise, “an appropriate probabilistic framework is required for successful verification and validation of ICME solutions.” The risks associated with development of complex, novel systems are numerous and, consequently, drive high cost. In the aircraft engine industry, for example, high development costs are associated with developing design confidence. During engine development, many decisions are made using combinations of analytical engineering models, experimental data, and expert knowledge. This approach to decision making and risk management has led to safe, widespread, and affordable air travel. Even so, development costs remain very high. Based upon recent work we propose a probabilistic framework based on ICME principles and Bayesian networks to integrate expert knowledge, available data, engineering model results, and test results from across the engineering development process. We claim that the qualitatively superior estimates produced by this framework will significantly reduce the costs of engineering development by identifying high-risk issues early in the design process and enabling the elimination of costly tests that are shown to be of little value toward adding design confidence. We motivate and illustrate the proposed framework with an example from component design with materials and process modeling..