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

Signal Processing for Rotating Arc TIG Welding Seam Tracking

Literature Overview and Core Findings

This paper by Jia Jianping, Liu Yunlong, Chen Jianping, Peng Liang and Liu Dan from Nanchang University addresses a problem that every pipe fabricator faces sooner or later: how to keep a welding torch centred on the joint when the fit-up is not perfect. The authors built a seam tracking system based on a rotating arc TIG welding head and developed a signal processing method to extract the weld seam position from the arc voltage signal.

The core contribution is the combination of median filtering and multiple passes of mean filtering to remove noise from the arc voltage signal. The authors found that this combined approach provides better denoising than either method alone. This is exactly what I would expect from practical experience. Median filtering is excellent for removing spike noise, which is common in arc welding due to arc instability, droplet transfer and electromagnetic interference. Mean filtering is good for smoothing out high-frequency fluctuations but it can also smear out the signal features you are trying to detect. By combining them, you get the best of both worlds.

Technical Interpretation and Process Significance

Let me explain why this matters in real pipe welding. When you are welding a long seam on a large diameter pipe, the joint line is never perfectly straight. There is always some mismatch, some high-low, some gap variation. If you are using an automated welding carriage, you need to know where the seam is at all times. The rotating arc method uses the fact that the arc voltage changes as the arc rotates around the torch axis. When the arc is pointing at the centre of the joint, the voltage is at a baseline value. When it points to one side or the other, the voltage changes because the arc length changes.

The problem is that the arc voltage signal is noisy. It contains not just the seam position information but also the arc instability noise, the power supply ripple, the electromagnetic interference from the welding current and the mechanical vibration of the torch. If you try to use the raw signal, you get a lot of false seam position readings. The filtering method described in this paper cleans up the signal so that the seam tracking controller can make reliable decisions.

Practical Implications for Pipe Welding

In my experience, seam tracking is one of the most underrated aspects of automated pipe welding. Everyone focuses on the welding parameters, but if the torch is not tracking the seam, all the parameter optimisation in the world will not help. I have seen automated welding systems produce beautiful weld beads on test coupons and then produce complete chaos on real pipe because the seam tracking could not handle the fit-up variation.

The rotating arc method has an advantage over other seam tracking methods because it does not require any additional sensors. The arc itself is the sensor. This means there is no need for a laser sensor, a tactile sensor or a vision system. The cost is lower and the system is simpler. However, the signal processing is critical. Without good filtering, the rotating arc method produces a signal that is too noisy to be useful.

The authors used multiple passes of mean filtering. In practice, I would recommend starting with a median filter to remove the spikes, then applying a moving average filter with a window size that matches the rotational frequency of the arc. The key is to filter out the noise without filtering out the signal. If you over-filter, you lose the seam position information. If you under-filter, you get false readings.

Key Questions and Reflections

One question that the paper does not fully address is how the filtering parameters should be adjusted for different welding conditions. The optimal filter window size depends on the arc rotation speed, the welding travel speed, the sampling rate and the noise characteristics. In a production environment, you would need to tune these parameters for each welding application.

Another question is how the system performs when the joint geometry changes. A V-groove joint has a different arc voltage signature than a square butt joint. The seam tracking algorithm needs to be robust to these variations. In pipe welding, we deal with a wide range of joint geometries, from thin wall square butt joints to thick wall V-groove joints with a backing ring.

Study Insights and Practical Recommendations

The signal processing method described in this paper is a solid engineering contribution. It addresses a real problem with a practical solution. For pipe fabricators who are considering automated welding, this is a useful reference. The combination of median and mean filtering is simple, robust and effective.

However, I would add a word of caution. Seam tracking is not just about signal processing. It is also about mechanical rigidity, torch positioning accuracy and the welding carriage design. If the carriage is not rigid, the torch will vibrate and the seam tracking will not work well no matter how good the signal processing is. In my experience, the mechanical system is often the weak link, not the electronics.

The bottom line is this: the rotating arc method with proper signal processing is a cost-effective way to achieve seam tracking in automated TIG welding of pipe. The key is to get the filtering right and to ensure that the mechanical system is rigid enough to support the tracking accuracy. This paper provides a good foundation for engineers who are developing such systems.