Prediction and optimization of epoxy adhesive strength from a small dataset through active learning

Abstract

Machine learning is emerging as a powerful tool for the discovery of novel high-performance functional materials. However, experimental datasets in the polymer-science field are typically limited and they are expensive to build. Their size (< 100 samples) limits the development of chemical intuition from experimentalists, as it constrains the use of machine-learning algorithms for extracting relevant information. We tackle this issue to predict and optimize adhesive materials by combining laboratory experimental design, an active learning pipeline and Bayesian optimization. We start from an initial dataset of 32 adhesive samples that were prepared from various molecular-weight bisphenol A-based epoxy resins and polyetheramine curing agents, mixing ratios and curing temperatures, and our data-driven method allows us to propose an optimal preparation of an adhesive material with a very high adhesive joint strength measured at 35.8 ± 1.1 MPa after three active learning cycles (five proposed preparations per cycle). A Gradient boosting machine learning model was used for the successive prediction of the adhesive joint strength in the active learning pipeline, and the model achieved a respectable accuracy with a coefficient of determination, root mean square error and mean absolute error of 0.85, 4.0 MPa and 3.0 MPa, respectively. This study demonstrates the important impact of active learning to accelerate the design and development of tailored highly functional materials from very small datasets.

Prediction of Lap Shear Strength and Impact Peel Strength of Epoxy Adhesive by Machine Learning Approach

“…In this study, among various ML techniques, ANN was selected to build a prediction model from given experimental data sets for lap shear strength at room temperature and impact peel strength at −40 • C because ANN is the most effective technique for classifying and predicting complex nonlinear or linear relationships among numerous variables [20,21,23,24]. In the input data set for the ANN model, lap shear strength, and impact peel strength were set as target variables, and two separate ANN models were constructed to have only one target variable.…”

“…In particular, ANNs, which are also known as deep learning if the neural network has multiple deep layers, provide the most accurate predictions in a short time from highly disordered data through their high-level algorithms [ 18 ]. In recent reports, the use of ANNs in materials science has been successfully demonstrated [ 19 , 20 , 21 ]. For instance, the tensile shear strength of bonded beechwood was predicted by an ANN model trained using an experimental database [ 22 ].…”

Corrosion rate prediction and influencing factors evaluation of low-alloy steels in marine atmosphere using machine learning approach

“…In recent years, machine learning methods have attracted significant attention with their powerful data mining capabilities. Due to its enriched modeling packages and improved operability for non-professionals, it has also been abundantly used in material researches [ 18 ]. Some successful examples of the informatics-driven design of new materials include high-temperature alloys, low thermal hysteresis shape memory alloys, and metal additive manufacturing [ 19 – 21 ].…”

Stacking fault energy prediction for austenitic steels: thermodynamic modeling vs. machine learning

“…Thus, some models are often only capable of predicting trends due to simple alloying effects with a limited composition range. A promising way to leverage the wealth of data and circumvent the difficulty of SFE prediction is by applying data-driven methods [ 26 , 27 ]. Machine learning (ML) is useful in extracting knowledge from multi-dimensional data and modeling the relationships between the targeted property and its related features [ 28 ].…”