COMPUTATIONAL ANALYSIS OF A KNUCKLE JOINTAND IMPLEMENTATION OF THE GENERALIZED REGRESSION NEURAL NETWORK

A Knuckle joint can be used to connect two rods that are under tensile force. This joint permits an angular misalignment between the rods and can support compressive load when it is controlled. The joints are used to make different kinds of connection i.e. tie rods, tension links in bridge structure. In this case, one of the rods has eyes at the rod's ends and the one end is forked with eyes at both legs. Pin (knuckle pin) is put through the rod's end and fork eye and held by a collar as well as split pin. Normally, empirical relationships can be found to determine the various sizes of the joint, and they are secure from a design perspective. The purpose of this paper is to determine the stresses within Knuckle joint using the analytical method. Further studies in this direction is possible by using different ways of contact, as well as the capability to handle loads. This research focuses on the type of meshing that is the best for components. The knuckle joint model is created using CATIA V6 R20. Later the model is loaded into the ANSYS 15.0 and is then implemented in both meshes, which are hexagonal and tetra mesh. A lot of industrial systems utilize knuckle joints, which ismade up of two components which are cast iron and steel. We are proposing modifications of the materials Steel, AL 6061-T6 and Teflon. The structural analysis was conducted for the Knuckle Joint at loads of 100N, 105N and 110N as well as 115N. The most effective combination of parameters such as Von Misses stress and equivalent shear stresses, shear stress, deformation, as well as weight loss for the knuckle joint was determined using ANSYS software. Teflon offers more factors of security, it is lighter and stiffness, as well as reduce stress, and is more rigid than other materials. Based on the findings, the machine learning method i.e. the neural network program that studies deformation, shear stress and von-Miss stresses, widely called Generalized Regression Neural Network (GRNN) was developed. The process involves defining specific variables for input (Different Materials and Loads) as well as output parameters which have been pre?defined and are readily available (Shear Stress, Von-Mises stresses and deformation)