Application of Nonlinear Dynamical Methods for Arc Welding Quality Monitoring

Abstract

Owing to its diverse, the stability of arc signals in high- IntroductionWith the development and application of automation technology and computer technology, the digitization and intelligent of welding has become a focus of the development of welding technology nowadays.Welding process is a complex physical and chemical process with various interference, and the stability of welding quality control is a more complex and difficult nonlinear problem [1][2][3]. Electrical signals,such as welding current and welding voltage,contains abundant information of welding process.If we can extract the feature information accurately,and then using the scientific method to undertake statistical analysis and calculation,will surely can be better sight into the characteristics of the mechanism of welding process [4,5]. Many scholars have studied in this respect: to reduce the subjectivity of double wire pulsed welding welding process stability evaluation. Considering weld strength as the quality characteristic in the selection of process parameters, fume formation in a pulsed gas metal arc welding (GMAW) process is investigated by coupling a time-dependent axi-symmetric two-dimensional model, which takes into account both droplet detachment and production of metal vapour, with a model for fume formation and transport based on the method of moments for the solution of the aerosol general dynamic equation [6]. Three geometry changes to the inner bore of a welding nozzle and their effects on weld quality during gas metal arc welding (GMAW) were investigated through the use of computational fluid dynamic (CFD) models and experimental trials [7]. An adaptive technique based on estimation of signal parameters via rotational invariance technique (ESPRIT) is proposed that optimizes the accuracy and computation time for harmonic/interharmonic estimation of stationary as well as nonstationary power supply signals [8]. The use of fuzzy rule based systems to model the relationship between weld control parameters and the weld bead geometry features is explored in this paper. The system is tested on three datasets and the performance is found to be satisfactory compared to the multilayer perceptron (MLP) and radial basis function (RBF) neural networks based systems [9]. Simpson S W elucidates the signature image approach to welding fault detection, covering the calculation of signature image data objects from blocks of welding electrical data (voltage and current), the definition of appropriate vector operations, and the manipulation of the signatures to permit detection of welding faults [10].The above analysis of the status of the system is not comprehensive, the extracted characteristic information is more rely on the experience of the people,and overemphasize on the independence of the state. Loss time-varying characteristics of the signal, and ignore the