Regression analysis framework for material selection of natural fibre-reinforced polymer composites

Material selection is one of the important processes to the automakers in producing and manufacturing parts for the automotive industry. The conventional material selections tool of Multiple Criteria Decision Making (MCDM) is based mostly on inconsistent judgement and preference subjectivity over the process selection. The biasness through the process selection can produce unreliable final decision. By using Statistical Package for Social Science (SPSS), it can manage quick processing, huge volume of data and save time and cost. In this study, a regression analysis framework is introduced to select the natural fibres (NF), polymer matrix and composites. Simple and multiple linear regression is used to construct statistical modelling of natural fibre reinforced polymer composite (NFRPC). In advanced approach, stepwise regression is used to construct the best regression modelling with significant mechanical properties that influence the performance score (PS) of the materials. Inferential statistical methods, such as estimation, hypothesis testing and confidence interval, are used to test the sample data of NFRPC to make a conclusion on the final decision. In addition, Pearson coefficient of correlation (r) is used to identify the relationship between the mechanical properties and PS. In addition, Multicollinearity issue is resolved by calculating the Tolerance (Tol) and Variance Inflation Factors (VIF). Three types of errors, such as mean absolute error (MAE), mean square error (MSE) and root mean squared error (RMSE) are used to evaluate the estimation process. All statistical measurements in stepwise regression are used during the screening process on material selection. The ranking process will finalize the material based on the maximum value of PS and minimum value of estimation error. Product design specification (PDS) of hand-brake lever parking is used as a case study to select the natural fibre, polymer matrix and the composite at the early stage. The ideal composition of fibre loading that optimises the PDS is identified by using analytical approach, namely the rule of mixtures (ROM), at the final stage. The results reveal that regression modelling can assist the design engineers to identify the best material in automotive application. Tensile strength (TS) is the significant mechanical property in the model proposed by stepwise regression to evaluate PS. The adequacy checking on the statistical model is performed by plotting the normal probability of regression standardized residual and normality plot. Regression analysis on 12 types of natural fibres and 10 types of polymers matrix shows that the top three best materials were coir, kenaf, cotton and polypropylene (PP), polystyrene (PS), high-density polyethylene (HDPE), respectively, to manufacture hand-brake lever parking. The estimation of density, Young’ modulus and tensile strength in various fibre loading using simplest micromechanical model showed kenaf/PS composite with 40% fibre loading offer better composition to manufacture hand-brake lever parking. A well corporation between regression and analytical approach is proven. Overall, this work can act as a guideline for the selection of the most suitable natural fibre, polymer matrix and composite candidate for an engineering application. This study has contributed to material selection process field which can provide more options of method to be chosen by practitioners especially for automotive product development application.

Text FK 2020 35 ir.pdf Download (1MB)

Actions (login required)

View Item