Naïve Bayes classifier, multivariate linear regression and experimental testing for classification and characterization of wheat straw based on mechanical properties

Abstract

Mechanical, chemical and microstructure characterization of a particular type of wheat straw is conducted in this study. Uniaxial tension tests on different gage length wheat straws are conducted to determine the mechanical properties. Fourier transform infrared test and pH tests are performed to confirm the chemical composition and acidity/alkalinity of wheat straw, respectively. The microstructural morphology of untested wheat straw and failure modes in failed wheat straw are investigated. Naïve Bayes (NB) algorithm is trained and is used to classify or determine the gage length of the wheat straw based on target mechanical properties. In the end, a multivariate linear regression (MLR) model is calibrated to predict the ultimate tensile strength (UTS) of wheat straw. The UTS and fracture strain decreased with increase in gage length. Wheat straw is found to have a pH value of 7.43 and presence of lignocellulosic functional groups is verified through FTIR. Both mode-I and mixed mode failures are observed in wheat straws. The trained NB classifier is able to categorize the wheat straws in to different gage lengths with an accuracy of 92% and the calibrated MLR model is able to predict the UTS of wheat straw fairly well. Considering the mechanical and chemical properties obtained in this study, wheat straw can be used as a reinforcement in concrete.