A simple constrained machine learning model for predicting high-pressure-hydrogen-compressor materials

Here we present the results of using techno-economic analysis as constraints for machine learning guided studies of new metal hydride materials. Using existing databases for hydrogen storage alloys, a regression model to predict the enthalpy of hydrogenation was generated with a mean absolute error of 8.56 kJ mol−1 and a mean relative error of 28%. Model predictions for new hydride materials were constrained by techno-economic analysis and used to identify 6110 potential alloys matching the criteria required for hydrogen compressors. Additional constraints such as alloy cost, composition, and likely structure were used to reduce the number of possible alloys for experimental verification to less than 400. Finally, expert heuristics and a novel machine learning approach to approximating alloy stability were employed to select an Fe–Mn–Ti–X alloy system for future experimental studies.

[1]  Alok Choudhary,et al.  A General-Purpose Machine Learning Framework for Predicting Properties of Inorganic Materials , 2016 .

[2]  Kristin A. Persson,et al.  Commentary: The Materials Project: A materials genome approach to accelerating materials innovation , 2013 .

[3]  Ian H. Witten,et al.  Data mining in bioinformatics using Weka , 2004, Bioinform..

[4]  David B. Brough,et al.  Microstructure-based knowledge systems for capturing process-structure evolution linkages. , 2017, Acta materialia.

[5]  Sergei V. Kalinin,et al.  Knowledge Extraction from Atomically Resolved Images. , 2017, ACS nano.

[6]  G. Kresse,et al.  Efficiency of ab-initio total energy calculations for metals and semiconductors using a plane-wave basis set , 1996 .

[7]  Bruno G. Pollet,et al.  Metal hydride hydrogen compressors: A review , 2014 .

[8]  Richard G. Hennig,et al.  Grand-canonical evolutionary algorithm for the prediction of two-dimensional materials , 2016 .

[9]  S. Suram,et al.  Generating information-rich high-throughput experimental materials genomes using functional clustering via multitree genetic programming and information theory. , 2015, ACS combinatorial science.

[10]  Jianjun Hu,et al.  Semi-Supervised Approach to Phase Identification from Combinatorial Sample Diffraction Patterns , 2016 .

[11]  A. Oganov,et al.  How evolutionary crystal structure prediction works--and why. , 2011, Accounts of chemical research.

[12]  L. Kaufman,et al.  Ternary phase diagrams of the manganese-titanium-iron and the aluminum-titanium-iron systems: A comparison of computer calculations with experiment☆☆☆ , 1979 .

[13]  Kamal Choudhary,et al.  High-throughput Identification and Characterization of Two-dimensional Materials using Density functional theory , 2017, Scientific Reports.

[14]  Manh Cuong Nguyen,et al.  On-the-fly machine-learning for high-throughput experiments: search for rare-earth-free permanent magnets , 2014, Scientific Reports.

[15]  Xiang Li,et al.  The application of Principal Component Analysis to materials science data , 2002, Data Sci. J..

[16]  A. Corma,et al.  Discovery of a new catalytically active and selective zeolite (ITQ-30) by high-throughput synthesis techniques , 2006 .

[17]  I Takeuchi,et al.  High-throughput determination of structural phase diagram and constituent phases using GRENDEL , 2015, Nanotechnology.

[18]  Yousef Saad,et al.  Data mining for materials: Computational experiments with AB compounds , 2012 .

[19]  Julia Ling,et al.  High-Dimensional Materials and Process Optimization Using Data-Driven Experimental Design with Well-Calibrated Uncertainty Estimates , 2017, Integrating Materials and Manufacturing Innovation.

[20]  Ichiro Takeuchi,et al.  Combinatorial Experimentation and Materials Informatics , 2006 .

[21]  Anubhav Jain,et al.  Finding Nature’s Missing Ternary Oxide Compounds Using Machine Learning and Density Functional Theory , 2010 .

[22]  B. Hardy,et al.  Acceptability envelope for metal hydride-based hydrogen storage systems , 2012 .