Predicting compressive strength of consolidated molecular solids using computer vision and deep learning. (May 2020)

• Record Type: Journal Article
• Authors: Gallagher, Brian; Rever, Matthew; Loveland, Donald; Mundhenk, T. Nathan; Beauchamp, Brock; Robertson, Emily; Jaman, Golam G.; Hiszpanski, Anna M.; Han, T. Yong-Jin
• Is Part Of: Materials & design. Volume 190(2020)
• Publication Date: 2020-05
• Subjects: Mechanical performance prediction -- Image analysis -- Random forest -- Deep neural network -- Machine learning; Materials -- Periodicals; Engineering design -- Periodicals; Matériaux -- Périodiques; Conception technique -- Périodiques; Electronic journals; 620.11
• DOI: 10.1016/j.matdes.2020.108541
• Languages: English
• ISSNs: 0264-1275

Abstract

Abstract: We explore the application of computer vision and machine learning (ML) techniques to predict material properties (e.g., compressive strength) based on SEM images. We show that it's possible to train ML models to predict materials performance based on SEM images alone, demonstrating this capability on the real-world problem of predicting uniaxially compressed peak stress of consolidated molecular solids samples. Our image-based ML approach reduces mean absolute percentage error (MAPE) by an average of 24% over baselines representative of the current state-of-the-practice (i.e., domain-expert's analysis and correlation). We compared two complementary approaches to this problem: (1) a traditional ML approach, random forest (RF), using state-of-the-art computer vision features and (2) an end-to-end deep learning (DL) approach, where features are learned automatically from raw images. We demonstrate the complementarity of these approaches, showing that RF performs best in the "small data" regime in which many real-world scientific applications reside (up to 24% lower RMSE than DL), whereas DL outpaces RF in the "big data" regime, where abundant training samples are available (up to 24% lower RMSE than RF). Finally, we demonstrate that models trained using machine learning techniques are capable of discovering and utilizing informative crystal attributes previously underutilized by domain experts. Graphical abstract: Unlabelled Image Highlights: Mechanical performance of uniaxially compressed solids can be predicted using machine learning on SEM image data. Computer vision is an effective approach to extract materials attributes for correlating to their performance Traditional computer vision and machine learning methods are compared with end-to-end deep learning methods. Deep Learning is the more powerful method, provided you have sufficient amount of data Random forest model performs best in the "small data" regime, whereas deep learning outpaces random forest in the "big data" regime. In the case of TATB, fine crystal attributes including pores and defects in few micron ranges are strong indicators of material strength