Exploration of data science techniques to predict fatigue strength of steel from composition and processing parameters

Abstract

This paper describes the use of data analytics tools for predicting the fatigue strength of steels. Several physics-based as well as data-driven approaches have been used to arrive at correlations between various properties of alloys and their compositions and manufacturing process parameters. Data-driven approaches are of significant interest to materials engineers especially in arriving at extreme value properties such as cyclic fatigue, where the current state-of-the-art physics based models have severe limitations. Unfortunately, there is limited amount of documented success in these efforts. In this paper, we explore the application of different data science techniques, including feature selection and predictive modeling, to the fatigue properties of steels, utilizing the data from the National Institute for Material Science (NIMS) public domain database, and present a systematic end-to-end framework for exploring materials informatics. Results demonstrate that several advanced data analytics techniques such as neural networks, decision trees, and multivariate polynomial regression can achieve significant improvement in the prediction accuracy over previous efforts, with R 2 values over 0.97. The results have successfully demonstrated the utility of such data mining tools for ranking the composition and process parameters in the order of their potential for predicting fatigue strength of steels, and actually develop predictive models for the same.

High-Dimensional Materials and Process Optimization Using Data-Driven Experimental Design with Well-Calibrated Uncertainty Estimates

“…These data sets include a magnetocalorics data set [19], a superconductor data set compiled by the internal Citrine team, a thermoelectrics data set [20], and a steel fatigue strength data set [21], which will all be described in more detail in the "Results on Test Cases" section. Models were trained to predict the magnetic deformation, superconducting critical temperature, figure of merit ZT, and fatigue strength, respectively, on these four data sets.…”

Solving Materials’ Small Data Problem with Dynamic Experimental Databases

“…Small data can be curated in two main approaches, namely (1) controlled experiments via the design of experiments or high-throughput methods within a single laboratory, and/or (2) mining the literature to leverage experiments conducted by the community. The former datasets are often well-structured, allowing process-property information extraction via material informatics and/or machine learning methodologies [9,11]. The latter approach involves a significantly wider scope of potential design variables, resulting in an unstructured database rife with missing and noisy data.…”

“…A decision tree classifier considering 27 synthesis variables was trained to predict whether or not a titania synthesis route would produce nanotubes, achieving 82% accuracy [8]. Agrawal et al compared a range of machine learning approaches, including multivariate polynomial regression (R 2 = 0.9801), support vector machines (R 2 = 0.9594), and artificial neural networks (R 2 = 0.9724), to predict the fatigue strength of steels [9]. Input variables describing chemical composition, processing temperatures and times, and upstream processing details were taken from the National Institute of Materials Science (NIMS) MatNavi [10].…”

A Survey on Data Mining in Steel Industries

“…The following paper revealed [16] the use of data analytics tools for predicting the fatigue strength of steels. Several physics-based as wellas data-driven approaches have been used to arrive at correlations between various properties of alloys and their compositions and manufacturing process parameters.…”