NiTiHf Shape Memory Alloys

Because of the time necessary for performing experiments on high dimensionality design space in manufacturing processes, exploring the entire design space is often prohibitively costly and time-consuming. To accelerate the development of NiTiHf shape memory alloys, we have proposed a sequential learning design framework to strategically identify experimental candidates who maximize information gain using fewest possible experiments. A Kriging regression model scales to high-dimensional search space in terms of processing features, part properties and alloy compositions. An active selector based on Maximum Expected Improvement (MEI) strategy was used to pick experiments candidate by balancing exploration of highly-uncertain candidates and exploiting high-performance candidates. The developed framework can effectively guide the practitioner to test the most promising candidates (i.e. processing parameters and/or material compositions) earlier for achieving desirable designed properties, in such a way as to greatly reduce the time and cost for developments of NiTiHf shape memory alloys.