A prediction model for keyhole geometry and acoustic signatures during variable polarity plasma arc welding based on extreme learning machine

Abstract

Purpose The purpose of this paper is to investigate the relationship between the keyhole geometry and acoustic signatures from the backside of a workpiece. It lays a solid foundation for monitoring the penetration state in variable polarity keyhole plasma arc welding. Design/methodology/approach The experiment system is conducted on 6-mm-thick aluminum alloy plates based on a dual-sensor system including a sound sensor and a charge coupled device (CCD) camera. The first step is to extract the keyhole boundary from the acquired keyhole images based on median filtering and edge extraction. The second step is to process the acquired acoustic signal to obtain some typical time domain features. Finally, a prediction model based on the extreme learning machine (ELM) technique is built to recognize different keyhole geometries through the acoustic signatures and then identify the welding penetration status according to the recognition results. Findings The keyhole geometry and acoustic features after processing can be closely related to dynamic change information of keyhole. These acoustic features can predict the keyhole geometry accurately based on the ELM model. Meanwhile, the predict results also can identify different welding penetration status. Originality/value This paper tries to make a foundation work to achieve the monitoring of keyhole condition and penetration status through image and acoustic signals. A useful model, ELM, is built based on these features for predicting the keyhole geometry. Compared with back-propagating neural network and support vector machine, this proposed model is faster and has better generalization performance in the case studied in this paper.

Application of SVM Models for Classification of Welded Joints

“…The SVM model finds a hyperplane that has the highest margin of class separation. The model itself has already proved as efficient in classifying welded joints, as confirmed by the authors [7,13,14,16].…”

“…It is possible to evaluate successfully keyhole geometry prediction and welding penetration using several machine languages such as extreme learning machine (ELM) technique, back-propagating neural network (BPNN) and support vector machine (SVM) [13].…”

“…The authors used SVM regression analysis where the kernel function is a radial basis function, whereas the activation function is a simple sigmoidal function and a model is used to identify the classification of different geometric features [13].…”

Compliant training control of ankle joint by exoskeleton with human EMG-torque interface

“…By making use of principles of extreme learning machine (ELM) (Huang et al, 2006(Huang et al, , 2012Lan et al, 2010;Huang and Chen, 2007;Yu et al, 2014;Huang et al, 2015aHuang et al, ,2015bWu et al, 2016), we attempt to describe the relationship between input and output of black-box model by using the feedforward neural network with single hidden layer. The output function of ELM for the single layer forward network (SLFN) is depicted as follows:…”

Effect of Arc Pressure on the Digging Process in Variable Polarity Plasma Arc Welding of A5052P Aluminum Alloy

“…Saad et al [12] achieved identification between three models of VPPA keyhole welding (no-keyhole, keyhole and cutting) using acoustic signal measurement. Wu et al [13] also investigated the relationship between the keyhole geometry and acoustic signatures with a dual-sensor system. An extreme learning machine model was built for predicting keyhole geometry.…”