An experimental study of vibration signal responses for defect classification in friction stir welded joints using optimal neural network

Abstract

In the fast growing industrial environment, condition monitoring is useful to determine the quality and smooth process. Two types of signals are emanated by the machines such as primary signals and secondary signals. Process parameters are called the primary signals which determine the performance of the machines. Defects and loss output signals are called the secondary signals like vibration, thermal, physical and chemical changes. Defect free products can be achieved by monitoring the secondary signals. In this experiment, vibration signal responses are used to classify the defects in Friction Stir Welding (FSW) process. FSW process is a developing solid state joining process, in which the material that is being welded does not liquefy and recast. Friction Stir Welding of aluminum alloys has clear route for the manufacturer to investigate the potential outcomes of utilizing commercial materials. Contrasted with the fusion welding technique those are routinely utilized for joining structural aluminum alloys. A portion of the focal points of the regular welding procedures are low twisting, no exhaust, porosity or scatter, no consumables, no extraordinary surface treatment, and no protecting gas prerequisites. This study of the FSW procedure was completed with a specific end goal to evaluate the influence of process parameters and to classify the defects in the welded joint of aluminum alloy (AA6063) 4 mm thickness material which is comprised of HSS M2 tool material newline