Arc Sound Signal Research Articles

Machine learning-based weld porosity detection using frequency analysis of arc sound in the pulsed gas tungsten arc welding process

Automatic welding equipment has replaced human welders in the nuclear industry for safety issues and uniform and high welding quality. However, automatic welding equipment cannot predict porosity defects. So, the weldment must be inspected by non-destructive testing. This inspection was a costly and time-consuming process, and it applies to each weldment even if it welded same material. To improve the welding efficiency, a weld porosity detection system of the same weld material with different material thicknesses was needed. This paper proposed a machine-learned porosity detection system for 3.0 mm plates with welding arc sound data from the pulsed gas tungsten arc welding (P-GTAW) process of 1.6 mm plates. Ensemble-Empirical Mode Decomposition (EEMD) was used to divide the arc sound signal according to the pulse period of P-GTAW. Fast Fourier transform (FFT) was used to convert the arc sound into frequencies for features extraction according to porosity. The validity of these weld frequency features was confirmed through k-fold cross-validation across various machine learning techniques, with evaluation of F-1 scores against experimental weld sounds.

Penetration State Identification of Aluminum Alloy Cold Metal Transfer Based on Arc Sound Signals Using Multi-Spectrogram Fusion Inception Convolutional Neural Network

The CMT welding process has been widely used for aluminum alloy welding. The weld’s penetration state is essential for evaluating the welding quality. Arc sound signals contain a wealth of information related to the penetration state of the weld. This paper studies the correlation between the frequency domain features of arc sound signals and the weld penetration state, as well as the correlation between Mel spectrograms, Gammatone spectrograms and Bark spectrograms and the weld penetration state. Arc sound features fused with multilingual spectrograms are constructed as inputs to a custom Inception CNN model that is optimized based on GoogleNet for CMT weld penetration state recognition. The experimental results show that the accuracy of the method proposed in this paper for identifying the fusion state of CMT welds in aluminum alloy plates is 97.7%, which is higher than the identification accuracy of a single spectrogram as the input. The recognition accuracy of the customized Inception CNN is improved by 0.93% over the recognition accuracy of GoogleNet. The customized Inception CNN also has high recognition results compared to AlexNet and ResNet.

Welding Penetration Monitoring for Ship Robotic GMAW Using Arc Sound Sensing Based on Improved Wavelet Denoising

The arc sound signal is one of the most important aspects of information related to pattern identification regarding the penetration state of ship robotic GMAW; however, arc sound is inevitably affected by noise interference during the signal acquisition process. In this paper, an improved wavelet threshold denoising method is proposed to eliminate interference and purify the arc sound signal. The non-stationary random distribution characteristics of GMAW noise interference are also estimated by using the high-frequency detail coefficients in different domains after wavelet transformation, and a mode of measuring scale that is logarithmically negatively correlated with the wavelet decomposition scale is created to update the threshold. The gradient convergent threshold function is established using the natural logarithmic function structure and concave–convex gradient to enable the nonlinear adjustment of the asymptotic rate. Further, some property theorems related to the optimized threshold function are proposed and theoretically proven, and the effectiveness and adaptability of the improved method are verified via the denoising simulation of speech synthesis signals. The four traditional denoising methods and our improved version are applied in the pretreatment of the GMAW arc sound signal, respectively. Statistical analysis and short-time Fourier transform are used to extract eight-dimensional time and frequency domain feature parameters from the denoised signals with randomly time-varying characteristics, and the extracted joint feature parameters are used to establish a nonlinear mapping model of penetration state identification for ship robotic GMAW using the pattern classifiers of RBFNN, PNN and PSO-SVM. The simulation results yielded by visual penetration classification and the multi-dimensional evaluation index of the confusion matrix indicate that the improved denoising method proposed in this paper achieves a higher accuracy in the extraction of penetration state features and greater precision in the identification of pattern classification.

Weld Defect Detection of a CMT Arc-Welded Aluminum Alloy Sheet Based on Arc Sound Signal Processing

The cold metal transfer (CMT) process is widely used in thin plate welding because of its characteristics of low heat input and stable arc. In actual production, a larger weld gap, misalignment, or other problems due to assembly error lead to serious welding defects, such as burn-through and a lack of fusion. The arc sound contains a wealth of information related to the quality of the weld. This work analyzes the mechanism of CMT arc sound generation, as well as the correlation between the time–frequency spectrum of the arc sound signal and welding quality. This paper studies the extraction of the multi-channel time–frequency spectrum of an arc sound and inputs it to a custom convolutional neural network for the CMT welding defect identification of thin aluminum alloy plates. The experimental result shows that the average accuracy of the proposed model is 91.49% in the defect identification of a CMT arc-welded aluminum alloy sheet, which is higher than that of the single-channel time–frequency convolutional neural network and other traditional classification models.

Weld Seam Deviation Prediction of Gas Metal Arc Welding Based on Arc Sound Signal

Weld seam deviation prediction is the key to weld seam tracking control, which is of great significance for realizing welding automation and ensuring welding quality. Aiming at the problem of weld seam deviation prediction in GMAW (gas metal arc welding), a method of weld seam deviation prediction based on arc sound signal is proposed. By analyzing the feature of the arc sound signal waveform, the time domain feature of the arc sound signal is extracted. The wavelet packet analysis method is used to analyze the time-fre- quency domain feature of the arc sound signal, and the wavelet packet energy feature is extracted. The time domain feature and wavelet packet energy feature are used to establish the feature vector, and the BP (back propagation) neural network is used to realize the weld seam deviation prediction. The results show that the method proposed in this paper has a good weld seam deviation prediction effect, with a mean absolute error of 0.234 mm, which provides a new method for GMAW weld seam recognition.

Arc sound model for pulsed GTAW and recognition of different penetration states

Weld defect detection based on arc sound signal in pulsed gas tungsten arc welding (GTAW) has been a hot research topic in industry and academia. However, arc vocal mechanism model of pulsed GTAW remains to be studied further. In this paper, a sensing system is developed to collect arc voltage, welding current, arc sound, and weld pool images synchronously during pulsed GTAW process. Theoretical researches and experiments verify that the arc sound signal of pulsed GTAW is proportional to the change rate of the instantaneous arc power input. Furthermore, an arc sound excitation model is proposed to explain the mechanism of arc sound signal. The model inputs are arc voltage and welding current; the model output is the arc sound signal. The experiments prove that the proposed excitation model is a non-linear model. Therefore, a dynamic long short-term memory (DLSTM) network model is designed to identify the non-linear model. Furthermore, the effects of different penetration states on arc sound signal are discussed in combination with the proposed arc sound excitation model. Finally, a novel method based on DLSTM model is built to recognize different penetration states: lack of fusion, normal penetration, and burn through. The proposed method was verified to be effective with high accuracy and robustness.

Penetration state identification of lap joints in gas tungsten arc welding process based on two channel arc sounds

Abstract Lap joints commonly exist in the welding processes during manufacturing of automobiles and ships. Their asymmetric structures make it extraordinary difficult to monitor the penetration states online. In this paper, a two-channel arc sound signal acquisition device is developed and a binaural auditory perception model is proposed to identify the penetration states of lap joints. Firstly, an auditory peripheral model is built to filter the two-channel arc sound signals and decompose them into a series of frequency bands. After that, with simulating the energy transformation and lateral inhabitation of auditory system, the frequency distributions of the two-channel arc sound signals are obtained. Subsequently, the energy differences of the two-channel sounds are obtained and decomposed into multi-scales by an auditory cortex model. Finally, a support vector machine is built to recognize the lap bead penetration states. This study provides a new feasible approach to identify penetration states of lap weld joints.

Welding penetration recognition based on arc sound and electrical signals in K-TIG welding

Abstract This paper presents a novel methodology for welding penetration recognition in Keyhole Tungsten Inert Gas(K-TIG) welding. A multisensor sensing system is established to acquire the signal of arc sound, arc current, and arc voltage. The Spectral Noise Subtraction(SNS) method is introduced to extract the pure arc sound from the collected sound signal. The interference of electromagnetic field induced by the high current arc is investigated and thus the generation mechanism of arc sound in K-TIG welding is revealed. Some particular features are extracted and the Principal Component Analysis(PCA) is utilized to reduce the dimension of features. Finally, a Support Vector Machine with ten-fold cross validation, grid search optimization, and Error-Correcting Output Codes(ECOC-SVM-GSCV) is utilized to identify partial penetration, full penetration, and excessive penetration. It is effective with high accuracy and robustness under different penetration states.

Weld bead penetration identification based on human-welder subjective assessment on welding arc sound

Utilizing arc sound signals to monitor the penetration states of weld bead is a promising method. However, the correlation between penetration states and arc sounds is relatively weak, and arc sounds easily affected by background noise. To distinguish the penetration states of weld bead from arc sound signals, a novel method was developed in this paper. This method utilized the subjective assessments of human welders on arc sounds to build mathematical model, and to identify penetration states. Firstly, a dissimilarity matrix of arc sound pairs was made according to the subjective assessments of human welders on these sound pairs. Subsequently, the correlations of penetration states with auditory features of arc sounds were obtained based on this dissimilarity matrix. Finally, a support vector machine was built to identify the penetration states. The results show that the proposed method has a high correct rate in identifying the penetration states of weld bead.

Analysis of the frequency features of arc voltage and its application to the recognition of welding penetration in K-TIG welding

Keyhole tungsten inert gas (K-TIG) welding is a novel and high-efficiency welding method and is characterized by the formed keyhole. In this paper, a multisensor sensing system is set up to collect voltage, current and arc sound signals. K-TIG experiments on Ti alloy and 304 stainless steel are carried out. A distinct frequency band at 2-4 kHz is found, and its physical significance is analyzed, which has a correlation with weld pool oscillation. The weld pool oscillation frequency is then extracted by the proposed algorithm. The oscillation modes in different weld states are analyzed, and the special mode in full penetration is determined. The difference in oscillation frequency between partial penetration and full penetration is distinct, and it is believed that the combination of the discovered frequency band and the proposed algorithm can be utilized to conveniently identify weld penetration.

Weld bead penetration state recognition in GMAW process based on a central auditory perception model

Arc sound signals contain abundant information about the welding process, so they are widely adopted to monitor the penetration states of weld bead online. However, the arc sounds are vulnerable to back ground noises, which limits the practical application of them in industry. In this study, a central auditory perception model was proposed. This model through simulating human auditory system to identify the penetration states of weld bead in GMAW process. Firstly, the auditory spectrums of the arc sounds are obtained through auditory peripheral model. Then the shape of auditory spectrums is analyzed by an auditory cortex model. Finally, according to the auditory cortical response, character vectors are built and used to identify the penetration states. The results show that the proposed method has a high accuracy rate in the identification of penetration states of weld bead. This research provides a new method for the online monitoring of welding quality.

Research on the seam tracking of narrow gap P-GMAW based on arc sound sensing

Arc sensor is important in the automation of narrow gap welding process because only the arc signal can reflect the position of the arc in the groove in real time, as well as the heat input to the side wall. Except the current and voltage of the arc, the deviation state of the welding arc can be identified by experienced welders by the associated arc sound, therefore, the arc sound signal presents a potential welding seam tracking signal. In this study, arc sounds were acquired and processed to examine the correlation with weld deviation during narrow gap pulsed gas metal arc welding (P-GMAW). The influences of groove sidewall on the arc sound signal and the sensitive of arc sound features to welding torch deviation were studied, and the short-time energy was determined as the best parameter to reflect the arc deviation in the narrow gap grooves. Furthermore, a system was established for seam tracking using the arc sound short-time energy and the results show that the arc sound can be used as the control signal to realize seam tracking to a certain degree of precision, thus, providing a new control signal for narrow gap P-GMAW welding seam tracking based on multi-information fusion.

A novel method for detecting the pantograph–catenary arc based on the arc sound characteristics

In recent years, the speed and the current transmission density of high-speed railways are increased. As a result, the rate of the occurrence of pantograph–catenary arc also increases, which directly influences the service performance of the pantograph–catenary system. The detection and identification of the pantograph–catenary arc is of great significance for evaluating the off-line level to ensure the safe operation of the pantograph–catenary system. The currently used conventional optical arc detection method would be distorted by strong sunlight or other environmental lights in the actual operation, which would affect the accuracy of detection. On the contrary, the arc sound signal is not affected by the diurnal variation and can be easily obtained; it can also be distinguished well from other interfering sounds. In this paper, a new method that uses the arc sound is proposed to realize the effective detection of the pantograph–catenary arc. The frequency spectrum and the intensity characteristics of the pantograph–catenary arc sound are analyzed. The results show that the frequency band of the arc sound signal is wide and mainly gets distributed at 5–17 kHz. The curve of the sound waveform is drawn by calculating the short-time average energy, which is used to obtain the arc’s starting time, duration, and intensity, for calculating the off-line rate.

Automated control of welding penetration based on audio sensing technology

This paper presents a technology for welding quality control in pulse gas tungsten arc welding (GTAW). An automated welding penetration control system and effective controller are designed to achieve real-time collection and analysis of the welding acoustic signal. A special preprocessing method called “auditory attention” is proposed to optimize the extraction of the arc sound signal, which includes region of interest (ROI) extraction and denoising. The penetration feature extraction is implemented in the preprocessed signal. A sound channel feature based on linear prediction cepstrum coefficient (LPCC) is proposed for inclusion in the feature extraction method. Using these penetration features, a typical back propagation artificial neural network (BPANN) prediction model is introduced for identification of the penetration state during the welding process. Through training using a large number of data, the prediction rate reached 80–90%. The BPANN-piecewise (BPANN-PW) controller is used to achieve online control of welding penetration via arc sound signal for pulse GTAW welding using work-piece of different shapes. The results showed that this controller could adjust the welding current accurately and promptly depending on the variation of the arc sound signal. The controlling effect was good for the online monitoring of automated robotic GTAW welding.

Application of short time energy analysis in monitoring the stability of arc sound signal

Abstract This paper employed the short time energy of arc sound signals to online quantitatively describe the stability of arc sound signal. At first, the signal can be preprocessed by wavelet packet filtering, and then detailed information of the short time energy of the signal was obtained using hamming window. After statistical analyzed the short time energy, the energy distribution possibility and cumulative distribution function of the signal can be collected. Then a proposed stability evaluation criterion was employed to quantitatively describe the stability of arc sound signal. Relative experimental data showed that more stable signal corresponded lager value of the criterion. The proposed method which combined the short time energy and statistical analysis was supported by many actual experiments. This contribution can benefit the quantitative evaluation of the arc welding process, and instructed the future parameters optimization to obtain welding products with high quality.

Quantitative evaluation method of arc sound spectrum based on sample entropy

Arc sound analysis is an effective way to evaluate the stability of the arc welding process. Current methods cannot effectively quantify the disorder of the process. By studying the characteristics of the arc sound signal, we found that low frequency random mutation of arc sound power resulted from unstable factors, such as splashes or short circuits, increased the complexity and randomness of the arc sound signals. Then the arc sound signals were visualized on time-frequency interface by means of spectrogram, and it was found that the max power spectral density (PSD) distribution of spectrogram was closely related to the stability of arc welding process. Moreover, a method based on sample entropy was proposed to further quantify the relation. Finally, considering the factors such as averages of max PSD and the standard deviations of sample entropy, a compound quantitative evaluation indicator, arc sound sample entropy (ASSE), which can avoid the influence of different parameters on the quantitative results, was proposed, so that the stability of arc welding process can be quantitatively presented. Testing results showed that the accuracy rate of the method was more than 90 percent.

Real-time monitoring of welding path in pulse metal-inert gas robotic welding using a dual-microphone array

Online monitoring of welding path is important as deviation from the welding path will significantly affect the welding quality. This paper proposed a welding path monitoring system using a dual-microphone array for the pulse metal-inert gas (MIG) robotic welding process. A calibration method was developed for microphones to find best locations for collectin arc sound signals. Arc sound features were analyzed to obtain the relationship between the arc sound signal and the welding path deviation. Prediction models for welding paths were then established in two ways: (1) the prediction model was based on individual microphone signals and (2) the model was based on signals from the dual-microphone array which proved to have better prediction accuracy in monitoring the welding path. The dual-microphone model showed its potential advantage in quality monitoring and control for the MIG welding process. The prediction error could reach 0.6708 and 1.1901 mm. The monitoring system based on a dual-microphone array could meet the position accuracy requirement of welding quality control in pulse MIG welding.

Porosity detection by analyzing arc sound signal acquired during the welding process of gas pipeline steel

Detection of severe defects such as porosity underneath the weld bead during the welding process is vital during installation of a gas pipeline network because such defects might lead to fatigue crack. In this study, the work associated with detection of porosity through analysis of the acquired arc sound is presented. Air-borne acoustic signal was acquired during the metal inert gas welding process on API 5L X70 gas pipeline steel. Then, the acquired signal was analyzed using Hilbert Huang transform (HHT), which uses empirical mode decomposition for the purpose of filtering unrelated-to-damage signal components, and Hilbert spectral analysis to obtain the energy-frequency-distance plot. Results showed a significant energy amplitude pattern at the region where both surface- and subsurface-pores existed. Thus, the application of HHT analysis to the acquired arc sound signal has significantly assisted in identifying hidden information that is related to the existence of defects. This finding would enhance the development of an online welding defect detection system during the welding process.

Acoustic Signature Based Weld Quality Monitoring for SMAW Process Using Data Mining Algorithms

The quality of weld depends upon welding parameters and exposed environment conditions. Improper selection of welding process parameter is one of the important reasons for the occurrence of weld defect. In this work, arc sound signals are captured during the welding of carbon steel plates. Statistical features of the sound signals are extracted during the welding process. Data mining algorithms such as Naive Bayes, Support Vector Machines and Neural Network were used to classify the weld conditions according to the features of the sound signal. Two weld conditions namely good weld and weld with defects namely lack of fusion, and burn through were considered in this study. Classification efficiencies of machine learning algorithms were compared. Neural network is found to be producing better classification efficiency comparing with other algorithms considered in this study.

Real-time seam penetration identification in arc welding based on fusion of sound, voltage and spectrum signals

Sensor technology application is the key for intelligent welding process. Multiple sensors fusion has shown their significant advantages over single sensor which can only provide limited information. In this paper, a feature-level data fusion methodology was presented to automatically evaluate seam quality in real time for Al alloy in gas tungsten arc welding by means of online arc sound, voltage and spectrum signals. Based on the developed algorithms in time and frequency domain, multiple feature parameters were successively extracted and selected from sound and voltage signals, while spectrum distribution of argon atoms related to seam penetration were carefully analyzed before feature parameters selection. After the synchronization of heterogeneous feature parameters, the feature-level-based data fusion was conducted by establishing a classifier using support vector machine and 10-fold cross validation. The test results indicate that multisensory-based classifier has higher accuracy i.e., 96.5873 %, than single sensor-based one in term of recognizing seam defects, like under penetration and burn through from normal penetration.