ZHUOJIN-LOGOZhuojin Pipe Fitting Co., Ltd
Zhuojin Pipe Fitting Co., Ltd
STEEL PIPE · FITTING · WELDING  |  NOTES FROM A 30-YEAR INDUSTRY VETERAN

Acoustic Signal Monitoring for Penetration Control in MIG Welding

Why Penetration Monitoring Matters

Penetration is the single most important quality parameter in a fusion weld. Too little penetration means lack of fusion and a weak joint; too much means burn-through and weld collapse. In automated welding, the operator cannot see the weld pool directly, so we rely on indirect indicators — arc voltage, current, travel speed. But these electrical parameters do not always correlate perfectly with penetration. Liu Lijun and colleagues at Zhejiang University and Harbin University of Science and Technology explored an alternative: the acoustic signal generated by the arc itself. The idea is that the sound of the arc changes as the weld pool geometry and penetration state change, and this acoustic signature can be captured and analysed to monitor penetration in real time.

Signal Analysis and Correlation Findings

The authors studied the non-stationary arc sound signal during MIG welding of butt joints in the spray transfer mode. They used wavelet transform to achieve effective signal-to-noise separation, suppressing the noise that typically contaminates acoustic signals in a welding environment. Then, by carefully selecting sample data, they analysed the signal from three perspectives: time domain, frequency domain and time-frequency joint analysis. The results are striking: the arc sound signal energy and its spectral characteristics are highly correlated with the penetration state. Specifically, the energy in the 1500-4500 Hz frequency band changes detectably when the penetration state changes. This means that a simple acoustic sensor — essentially a microphone — can be used to detect penetration changes that would otherwise go unnoticed until the weld is inspected.

Implementation Considerations for Production

The practical appeal of this method is its low cost and simplicity. A microphone is far cheaper than a laser vision sensor or an X-ray system, and it does not require a clear line of sight to the weld pool. However, I have seen enough welding shops to know that acoustic monitoring has a serious drawback: background noise. A typical fabrication shop has grinders, cranes, other welding arcs, and general industrial noise that can swamp the arc sound signal. The wavelet transform approach used in this paper is effective, but it requires careful signal processing and a well-designed sensor placement strategy.

Factor Acoustic Monitoring Vision-Based Monitoring
Cost Low High
Sensor robustness Good Moderate
Susceptibility to background noise High Low
Information content Moderate High
Suitability for closed-loop control Promising Established

Reflections on Non-Destructive Process Monitoring

This paper is an important contribution to the field of welding process monitoring. The finding that the 1500-4500 Hz frequency band is particularly sensitive to penetration changes gives us a specific, actionable signal feature that can be used for monitoring. In my experience, the most successful implementations of this type of monitoring are those that combine acoustic sensing with other signals — arc voltage, current and even high-speed imaging — to build a multi-sensor picture of the process. The acoustic signal alone may not be sufficient for full closed-loop control, but it is an excellent complementary signal that can detect trends and anomalies. For a production line running long welds, the ability to detect a gradual loss of penetration before it becomes a defect is worth a great deal.

Conclusion

This study demonstrates that arc sound is a viable, low-cost signal for monitoring penetration state in MIG welding. The 1500-4500 Hz band is the key frequency range to watch. For production engineers, this opens the door to affordable, real-time quality monitoring on automated welding cells. The challenge is not the sensing principle — it is the practical engineering of sensor placement, noise suppression and signal interpretation in a real shop environment. This paper provides a solid foundation for that engineering work.