Limits...
A predictive machine learning approach for microstructure optimization and materials design.

Liu R, Kumar A, Chen Z, Agrawal A, Sundararaghavan V, Choudhary A - Sci Rep (2015)

Bottom Line: These challenges render traditional search-based optimization methods incompetent in terms of both searching efficiency and result optimality.A systematic framework consisting of random data generation, feature selection and classification algorithms is developed.Experiments with five design problems that involve identification of microstructures that satisfy both linear and nonlinear property constraints show that our framework outperforms traditional optimization methods with the average running time reduced by as much as 80% and with optimality that would not be achieved otherwise.

View Article: PubMed Central - PubMed

Affiliation: EECS Department, Northwestern University, Evanston IL, USA.

ABSTRACT
This paper addresses an important materials engineering question: How can one identify the complete space (or as much of it as possible) of microstructures that are theoretically predicted to yield the desired combination of properties demanded by a selected application? We present a problem involving design of magnetoelastic Fe-Ga alloy microstructure for enhanced elastic, plastic and magnetostrictive properties. While theoretical models for computing properties given the microstructure are known for this alloy, inversion of these relationships to obtain microstructures that lead to desired properties is challenging, primarily due to the high dimensionality of microstructure space, multi-objective design requirement and non-uniqueness of solutions. These challenges render traditional search-based optimization methods incompetent in terms of both searching efficiency and result optimality. In this paper, a route to address these challenges using a machine learning methodology is proposed. A systematic framework consisting of random data generation, feature selection and classification algorithms is developed. Experiments with five design problems that involve identification of microstructures that satisfy both linear and nonlinear property constraints show that our framework outperforms traditional optimization methods with the average running time reduced by as much as 80% and with optimality that would not be achieved otherwise.

No MeSH data available.


Related in: MedlinePlus

Visualization of ODF solutions to F2.All 26 cases that maximize the objective function F2 are shown.
© Copyright Policy - open-access
Related In: Results  -  Collection

License
getmorefigures.php?uid=PMC4477370&req=5

f6: Visualization of ODF solutions to F2.All 26 cases that maximize the objective function F2 are shown.

Mentions: Table 1 shows optimal values of each design function obtained by the aforementioned methods. As we can deduce from this comparison, for linear properties as Y and ms, LP always provides a valid solution. It is the nonlinear problems as E, F1 and F2 that pose a challenge for traditional optimization methods. Exhaustive search gives largely unstable results; one almost has to rely on pure luck to bet on a fair result. Guided search has a tendency to get stuck at local optima, which can sometimes be even worse than the naive exhaustive search (although the running time is much less). Surprisingly, GA has worked poorly on nonlinear problems (we also tried GA on linear problems and the answer is never as good as LP). In fact, for nonlinear problems the guided search is often a better choice than GA. ML methods have been proven successful achieving the best answers throughout all problems, linear and nonlinear.


A predictive machine learning approach for microstructure optimization and materials design.

Liu R, Kumar A, Chen Z, Agrawal A, Sundararaghavan V, Choudhary A - Sci Rep (2015)

Visualization of ODF solutions to F2.All 26 cases that maximize the objective function F2 are shown.
© Copyright Policy - open-access
Related In: Results  -  Collection

License
Show All Figures
getmorefigures.php?uid=PMC4477370&req=5

f6: Visualization of ODF solutions to F2.All 26 cases that maximize the objective function F2 are shown.
Mentions: Table 1 shows optimal values of each design function obtained by the aforementioned methods. As we can deduce from this comparison, for linear properties as Y and ms, LP always provides a valid solution. It is the nonlinear problems as E, F1 and F2 that pose a challenge for traditional optimization methods. Exhaustive search gives largely unstable results; one almost has to rely on pure luck to bet on a fair result. Guided search has a tendency to get stuck at local optima, which can sometimes be even worse than the naive exhaustive search (although the running time is much less). Surprisingly, GA has worked poorly on nonlinear problems (we also tried GA on linear problems and the answer is never as good as LP). In fact, for nonlinear problems the guided search is often a better choice than GA. ML methods have been proven successful achieving the best answers throughout all problems, linear and nonlinear.

Bottom Line: These challenges render traditional search-based optimization methods incompetent in terms of both searching efficiency and result optimality.A systematic framework consisting of random data generation, feature selection and classification algorithms is developed.Experiments with five design problems that involve identification of microstructures that satisfy both linear and nonlinear property constraints show that our framework outperforms traditional optimization methods with the average running time reduced by as much as 80% and with optimality that would not be achieved otherwise.

View Article: PubMed Central - PubMed

Affiliation: EECS Department, Northwestern University, Evanston IL, USA.

ABSTRACT
This paper addresses an important materials engineering question: How can one identify the complete space (or as much of it as possible) of microstructures that are theoretically predicted to yield the desired combination of properties demanded by a selected application? We present a problem involving design of magnetoelastic Fe-Ga alloy microstructure for enhanced elastic, plastic and magnetostrictive properties. While theoretical models for computing properties given the microstructure are known for this alloy, inversion of these relationships to obtain microstructures that lead to desired properties is challenging, primarily due to the high dimensionality of microstructure space, multi-objective design requirement and non-uniqueness of solutions. These challenges render traditional search-based optimization methods incompetent in terms of both searching efficiency and result optimality. In this paper, a route to address these challenges using a machine learning methodology is proposed. A systematic framework consisting of random data generation, feature selection and classification algorithms is developed. Experiments with five design problems that involve identification of microstructures that satisfy both linear and nonlinear property constraints show that our framework outperforms traditional optimization methods with the average running time reduced by as much as 80% and with optimality that would not be achieved otherwise.

No MeSH data available.


Related in: MedlinePlus