An Integrated Approach to Determine a Phenomenological Equation to Predict Yield Strength in Titanium Alloys

Since composition and microstructure considerably affect the mechanical properties of metallic materials, it is highly desirable to be able to predict mechanical properties based upon these features. However, due to the complexity of real, multi-component, multi-phase engineering alloys, it is extremely difficult to develop constituent-based phenomenological equations. An accepted solution to the problem is to use Neural Networks. Unfortunately, while the developed model is quantitative it is not phenomenological. Thus, we propose that one must use new approaches in parallel with neural networks to derive the phenomenological equations.  This talk will highlight a new method based upon the integration of three separate modeling approaches, specifically Artificial Neural Networks, Genetic Algorithms, and the Monte Carlo method to derive phenomenological equations with a simultaneous analysis of uncertainty. This approach has been applied to derive a phenomenological equation for the prediction of tensile strength in a variety of Ti-6-4 microstructures from databases containing compositional and microstructural features of the alloy.