Quality analysis of CMT lap welding based on welding electronic parameters and welding sound

https://doi.org/10.1016/j.jmapro.2021.11.055Get rights and content

Abstract

Monitoring of weld joint quality is a significant issue in Cold Metal Transfer (CMT) lap welding. In this paper, CMT lap welding experiment of low carbon steel sheet was carried out, the sound characteristics of CMT are studied. Further analysis of the “special two-step mode” of welding sound shows that the faster the change of arc energy, the greater the corresponding sound pressure value. Furthermore, the feature extraction and fusion methods of welding electrical parameters and welding sound signals were investigated based on the two abnormal welding states: gas feeding error and welding wear. In the aspect of electric signal, welding current, welding voltage, line energy were studied, and in the aspect of sound signal, MFCC is extracted after de-framing and windowing. BiLSTM-CTC algorithm has been used to identify welding process gas feeding error and welding wear. For the recognition model the classification error rate based on sound feature is the lowest at 0.389, and the classification error rate based on electrical signal and acoustic signal fusion feature is at 0.295.

Introduction

Cold Metal Transfer (CMT) is an advanced technology used in welding, which was first put forward by Fronius in 2002. Because of its advantages such as no spatter and low heat input [1], [2], [3], it is preferable in thin sheet welding. At present, CMT has been applied in welding single-pass 2-mm-thick aluminium [4] and multi-pass 15-mm-thick steel [5], [6]. At the same time, some achievements have been made in numerical simulation and process parameter research. For example, Cao et al. utilized software ANSYS to analyze temperature field of CMT welding [7]. Zhang et al. used high-speed video photography to study the CMT arc characteristics and its droplet transfer process [8]. Pickin et al. proposed a cladding process technology and applied it to ternary alloyed (Al–Cu–Mg) aluminium plate [9].

There is wealthy information can be obtained in welding process by sensing and recognition systems, which includes welding images [10], [11], [12],welding sound signal [13], [14], [15], [16], [17] and welding electrical signal [18], [19], [20], etc. The fusion method of multi-sensor information and the full mining of welding production data are currently applied to analyze data in welding process. Specifically, On-line diagnosis of welding production anomalies is a focus of research. In the process of CMT welding, there are often some abnormal problems in actual production, for example, welding leakage caused by excessive assembly clearance and insufficient protective gas, among which welding leakage has a direct adverse effect on the welding quality. To detect these defects, some methods like image processing and electronic signal processing are used in welding process. In particular, the research of welding arc sound signal is also developing rapidly. As an “audible image”, welding sound signal is used widely in the research of GTAW/GMAW dynamic process [14], [15], [16]. Nevertheless, there are few reports on the application of arc sound signal in the CMT welding process.

On the other hand, speech recognition methods are often used in the research of welding sound, for example, the HMM model [21]. With the further development of speech analysis, a circulating neural network called BiLSTM-CTC [22] has gradually developed and has been proved that it could analyze time series and recognize the speech signals more accurately [23]. However, it has not been reported that this method has been used in the study of various defects in CMT welding process.

Aiming to solve problems in welding, we attempt to design the CMT lap welding experiment, the sound feature of CMT arc is first extracted, then fused with the electric signal feature, and at last is analyzed with the tools of BiLSTM-CTC. The remainder of the paper is organized as follows. Section 2 presents the experiment setup. In Section 3, we have analyzed CMT sound signals combined with welding experiment. In Section 4, for CMT lap welding, we specifically analyzed insufficient gas supply and welding leakage these two abnormal states based on welding electronic parameters and welding sound. In Section 5, LSTM-CTC is used to model the recognition of two abnormal welding states. Finally, Section 6 summarizes the main conclusions.

Section snippets

Experimental multi-sensor system

The experimental multi-sensor system is shown in Fig. 1. The system consists of CMT welding module, welding signal acquisition module and analysis and processing module, which can realize the functions of welding, signal acquisition measurement and real-time quality analysis. The hardware of the system includes CMT welder, protective gas cylinder, FANUC robot and control cabinet, high-speed camera, electric signal acquisition module, sound acquisition system, host computer, etc.

The welder is

Welding sound signal of different CMT operation mode

The CMT welder provide us with different operation modes for welding steel, specifically, the ‘2-Step’ mode and ‘Special 2-Step’ mode. In order to study the different sound characteristic of CMT modes, we perform the thin plate lap welding experiments. We record the welding current and welding voltage at the same time.

As shown in Fig. 2, for ‘Special 2-Step’ mode, GPr means stable gas feeding all the time during welding procedure. Isingle bondS is the arcing stage, in this stage, the base metal is heated

CMT lap welding experiments

The welding power supply is set to “special two-step mode”. The welding lap experiment platform is shown in Fig. 9. In actual welding, the angle between torch and welding platform is 45 degrees. In welding, the dry elongation is fixed to 12 mm. The welding process of overlap fillet weld is completed on the test board by teaching-running mode.

The wire used in the experiment is YM-80A, and the diameter of the wire is 1.2 mm. The specific composition of the welding wire is shown in Table 1.

The

Long short term memory networks, BiLSTM and CTC

Long-term and short-term memory network is an improved version of RNN [25]. It can learn to depend on for a long time. They were proposed by Hochreiter &Schmidhuber and later improved and promoted by many people. LSTM has succeeded in many problems and has been widely used now [27], [28].

BiLSTM refers to bidirectional LSTM, which inputs time series in both positive and reverse order. The hidden layer of BiLSTM has two values. A is involved in forward LSTM calculation and A' is involved in

Results

Fig. 21 is the error rate curve in the training process of model2. It can be observed that when Epoch reaches 8, the error rate reaches a minimum of 0.295. By comparing the two models, which only use MFCC feature model and fusion feature model, we can find that the error rate of the model is reduced after the current RMS, voltage RMS and heat input feature of welding electric signal are fused.

Finally, there is still much room to improve the recognition of abnormal state in this welding process.

Conclusion

The sound of lap welding of CMT low carbon steel sheet was studied. In different operation modes of CMT, the short-term welding voltage, welding current and arc sound are compared. Under the special two-step mode, the welding voltage waveform has two obvious platforms in one pulse period, and the welding voltage waveform of the two-step mode is closer to a square wave signal. In the special two-step mode, the welding power pulse is more similar to triangular wave, while in the two-step mode,

Declaration of competing interest

We declare that we have no financial and personal relationships with other people or organizations that can inappropriately influence our work, there is no professional or other personal interest of any nature or kind in any product, service and/or company that could be construed as influencing the position presented in, or the review of, the manuscript entitled.

Acknowledgment

This work is partly supported by the National Natural Science Foundation of China under the Grant No. 61873164.

References (27)

Cited by (16)

  • Properties of TiC-reinforced Cu-W alloys prepared by wire arc additive manufacturing

    2023, International Journal of Refractory Metals and Hard Materials
View all citing articles on Scopus
View full text