Assessing the mechanical performance of different cable bolts based on design of experiments techniques and analysis of variance

Many experimental studies have been carried out to characterise the mechanical performance of cable bolts under axial or shear loading. The common limitation in all these studies is that only single-variate descriptive analysis was included leading to an insufficient understanding of interactive influences of a range of parameters on the mechanical performance of cable bolt. Also, defining a reasonable number of experiments for the testing program that includes wide range of parameters has been a challenge for the scholars and practitioners. Thus, to address these limitations in a sound and logical manner, two statistical techniques being Taguchi and Response Surface Methodology (RSM) were introduced to demonstrate their effectiveness for a proper design of experiments leading to a minimum number of tests without compromising the output quality. Such a task was performed on the data reported in the literature. In addition, the quantitative post-data processing was conducted using Analysis of Variance (ANOVA). To be specific, ANOVA was mainly performed to quantify the contribution of different independent parameters (e.g. confining medium and grout strengths as well as borehole dimeter) on the dependent ones (e.g. peak and residual loads as well as the initial stiffness) individually and interactively. RSM was also used for the post-data multivariate statistical analysis. Finally, it was concluded that with the aid of Design of Experiments techniques such as Taguchi it is feasible to develop a minimum number of experiments to obtain the most representative tests results without compromising the quality of research outcome along with some quantitative post-data processing using ANOVA. Also, it was demonstrated that one of the most important specifications of RSM is to provide the quantitative and visual interactive effects of independent parameters on the mechanical performance of different cable bolts using multiple regression modelling.