Experimental-based statistical models for the tensile characterization of synthetic fiber ropes: a machine learning approach.

This study investigated the tensile behavior of some prevalent synthetic fiber ropes made of polyester, polypropylene, and nylon polymeric fibers. The aim was to generate well-documented experimental statistics and develop simplified stress-strain constitutive laws that can describe the ropes' tensile response. The methodology involved analyzing the thermal history of the fibers using the DSC technique, tensile testing of fibers and yarn components of the rope, and conducting 196 rope tensile tests with optimum testing conditions. Based on the test results, an experimental database of the ropes' tensile characteristics was established, containing different parameters of material properties, rope construction, fiber processing, fiber tensile properties, and rope tensile responses. Subsequently, ANN models were developed and optimized using MATLAB based on the generated dataset's inputs and outputs to predict the studied ropes' tri-linear stress-strain profiles. The results showed that the ANN models accurately predicted the stress-strain properties of ropes represented by the tri-linear approximation with an error of about 5% for the failure strength and strain. The study provides insight into the process-structure-property relationship of synthetic fiber ropes and contributes to minimizing the cost and effort in designing and predicting their tensile properties while contributing to the practical industry.