A Modeling the coefficient of friction on various surfaces using multiple linear regressions

Engineering properties of agricultural products

Abstract

When analyzing and designing equipment for post-harvest handling, food processing, and storage, it is crucial to understand the coefficient of friction of legume crops on different structural surfaces. The article examines the effects of various treatments on the static friction coefficient and dynamic friction coefficient of Phaseolus vulgaris L. Statistical analytic techniques were used to determine the significance of the single effect, dual interaction effect, and triple interaction effect of treatments (moisture content, contact surface, and sliding velocity) on static and dynamic friction coefficients, respectively. The static and dynamic friction coefficients on various contact surfaces were predicted using multiple linear regression modeling. Mean relative deviation modulus, root mean square error, and coefficient of determination (R²) were three statistical measures used to assess the predictive power of constructed multiple linear regression models. The significant increases in moisture content and contact surface single effect were slightly smaller than the significant increases in the dual interaction effect of treatments on the static friction coefficient, which were found to be 3.2 and 3 times the former, respectively. With regard to the dynamic friction coefficient, the treatments' significant increases in the triple interaction effect were 8.8, 3.7, and 8.9 times larger than the corresponding single effects for sliding velocity, moisture content, and contact surface. The single effect of the contact surface on the dynamic friction coefficient was similarly found to be 2.4 times greater than that of sliding velocity or moisture content, while the single effect of the contact surface on the static friction coefficient was found to be 1.1 times greater than that of moisture content. Static and dynamic friction coefficients could be precisely predicted by proposed multiple linear regression models, based on a reasonable average of statistical parameters (R² = 0.955, RMSE = 0.01788, and MRDM = 3.152%). Using the models for direct prediction of static and dynamic friction coefficients, corresponding to each contact surface, based on sliding velocity and moisture content is suggested in practice.