Artificial neural network-based resistance spot welding qualityassessment system

Abstract

On-line quality assessment has become one of the most critical requirements for improvingthe efficiency and the autonomy of automatic resistance spot welding (RSW) processes. Anaccurate and efficient model to perform non-destructive quality estimation is an essentialpart of the assessment process. This paper presents a structured and systematic approachdeveloped to design an effective ANN-based model for on-line quality assessment in RSW.The proposed approach examines welding parameters and conditions known to have aninfluence on weld quality, and builds a quality estimation model step by step. Themodeling procedure begins by examining, through a structured experimental design, theeffect of welding parameters (welding time, welding current, electrode force and sheetmetal thickness) and welding conditions represented by typical characteristics of thedynamic resistance curves on multiple welding quality indicators (indentation depth,nugget diameter and nugget penetration) and by analyzing their interactions and theirsensitivity to the variation of the dynamic process conditions. Using these results and bycombining an efficient modeling planning method, neural network paradigm, multi-criteriaoptimization and various statistical tools, the identification of the model form and thevariables to be included in the model is achieved by executing a systematic modeloptimization procedure. The results demonstrate that the proposed approach can lead to ageneral ANN-based model able to accurately and reliably provide an appropriate assessmentof the weld quality under diverse and variable welding conditions.