Spiral steel pipes are extensively used in water conveyance projects, and welding quality directly influences the operational safety of pipelines. Drawing on engineering practice, this paper introduces online monitoring technology for the welding of large-diameter, thick-walled spiral steel pipes and develops a closed-loop quality control framework of real-time monitoring, intelligent diagnosis, and dynamic regulation, thereby achieving refined welding control, lowering defect incidence, and providing technical support for the safe operation of major water conservancy projects.

Driven by rapid urbanization and the expansion of water conservancy infrastructure, large-diameter, thick-walled spiral steel pipes have emerged as a principal material for long-distance water conveyance pipelines because of their high strength, excellent sealing capability, corrosion resistance, mature production technology, and controlled cost. However, their large diameter and heavy wall thickness significantly increase the complexity of the welding process. As the core manufacturing process, welding is prone to defects such as incomplete penetration, lack of fusion, and cracking if process control is inadequate, thereby compromising structural safety and service life and even threatening the stable operation of the water conveyance system. Therefore, the application of online welding process monitoring to enable real-time quality surveillance and timely defect mitigation is of great significance for improving welding quality and ensuring project safety.

The Jiulongjiang Water Diversion Project is a major water conservancy initiative undertaken by Zhangzhou City during the 14th Five-Year Plan period and forms part of the Minnan–Southwest Fujian Water Resources Allocation Project. It is classified as a Class III medium-sized project and mainly comprises a water-intake booster pumping station, water diversion tunnels and pipelines, pressure-regulating structures, and a regulating reservoir. The project is executed in two phases: a near-term water supply of 360,000 m³/d, primarily for industrial use, and a long-term target of 1 million m³/d. The total water transmission route spans 122.65 km, comprising 55.09 km of tunnels and 67.56 km of pipelines. The main water transmission pipeline consists of Φ2820 mm × 24 mm spiral-welded steel pipes (Q235B), with each pipe with a length of 9 m. External corrosion protection consists of reinforced 3PE with a total thickness of at least 4.2 mm, while the internal wall is sandblasted to Sa2.5 grade and subsequently coated with three layers of solvent-free epoxy, achieving a dry film thickness of 400 μm. The pipeline has a working pressure of 0.7 MPa, a design pressure of 1.0 MPa, and a hydrostatic test pressure of 1.2 MPa. Manufacturing of the steel pipes is carried out by a qualified pressure-pipe manufacturer through a competitive bidding process to ensure product quality.

The welding process for spiral steel pipes involves rolling strip steel into a pipe blank at a predetermined helical angle, followed by welding to form the longitudinal pipe seam. During welding, the high temperature produced by the welding equipment rapidly melts and fuses the edges of the pipe blank, forming a strong, continuous weld seam. Among the various welding methods, submerged arc welding (SAW) is widely employed for large-diameter, thick-walled spiral steel pipes due to its distinct advantages. Submerged arc welding (SAW) is a process in which an electric arc operates beneath a layer of flux. During welding, the flux not only shields the arc and molten pool from the surrounding air but also participates in metallurgical reactions, enhancing the chemical composition and mechanical properties of the weld metal. This process offers significant advantages, including high welding quality, high production efficiency, and favorable working conditions, allowing it to meet the dual requirements of quality and efficiency for large-diameter, thick-walled spiral steel pipes.

The large diameter and thick walls of these pipes pose numerous technical challenges during welding operations, particularly:

1) High Welding Workload: The large diameter and thick walls substantially increase the welding length, significantly raising the workload and placing greater demands on both welding equipment and operators.

2) Complex Molten Pool Shape: The helical angle of the spiral weld seam produces a complex molten pool geometry and stress distribution during welding. Under the combined effects of gravity and arc force, defects such as molten metal sagging and undercut are highly likely to occur.

3) High Risk of Internal Defects: Thick-walled steel pipes require multi-layer, multi-pass welding, where the quality of each weld layer directly influences the overall performance of the weld joint.

Improper control of welding parameters can readily cause internal defects, such as lack of fusion between layers and slag inclusions, which can severely compromise the structural strength and sealing performance of the steel pipe.

1) Weld Seam Tracking Technology
In welding large-diameter, thick-walled spiral steel pipes, weld seam tracking is a critical step for ensuring weld quality. Vision-sensing–based tracking technology employs cameras to capture real-time images of the welding zone. Advanced image-processing algorithms accurately detect and analyze the weld seam position, enabling precise control of the welding equipment to ensure that the heat source follows the centerline of the weld seam. This technology typically employs laser or vision sensors to capture image data. Once transmitted to the image-processing system, algorithms such as edge detection and feature extraction are used to identify the position and geometry of the weld seam. The controller then adjusts the actuator—such as the welding torch of a robotic welder—in real time to achieve dynamic weld seam tracking. In practical applications, adaptive algorithms are employed to optimize tracking parameters in response to dynamic welding conditions, effectively enhancing weld seam tracking accuracy and stability while ensuring consistent welding quality. The camera captures images of the welding area, which are then transmitted to the image-processing system. Edge detection and feature extraction algorithms are applied to identify the position and geometry of the weld seam. The controller adjusts the actuator, enabling dynamic weld seam tracking.

Weld tracking workflow: 

Camera acquisition of welding-zone images → Image processing system reception of image data → Weld seam position and geometry identified via edge detection and feature extraction → Controller commands actuator adjustment → Dynamic weld seam tracking achieved

2) Welding Defect Visual Recognition Technology
Welding defect visual recognition technology leverages visual sensing to enable online detection of weld quality defects. This technology captures real-time images of the welding area and employs image processing and pattern recognition techniques to analyze and identify defect features. To enhance recognition accuracy and reliability, this technology often employs a combination of multiple image-processing methods. Initially, a threshold segmentation algorithm separates the defect area from the background, followed by morphological analysis, feature extraction, and classification of the segmented region. During actual welding, real-time images are compared with a pre-trained model to rapidly identify the presence, type, and location of defects. In recent years, deep learning algorithms have advanced significantly in this field. By employing a Convolutional Neural Network (CNN) trained on a large dataset of labeled welding defect images, the model can automatically learn defect feature patterns, substantially improving both the accuracy and efficiency of defect recognition.

Welding current and voltage are critical parameters that directly influence weld quality. Real-time monitoring enables prompt detection of welding anomalies. In welding large-diameter, thick-walled spiral steel pipes, the stability of welding current and voltage directly impacts weld penetration, weld bead width, and overall forming quality. Hall-effect current sensors are commonly employed to monitor welding current. Based on the Hall effect, the measured current is converted into a proportional voltage signal, and the real-time current value is obtained through precise measurement and processing. Voltage monitoring can employ devices such as resistive voltage divider sensors, which convert high voltage into a low-voltage signal suitable for accurate measurement. The collected current and voltage signals are first amplified and filtered by conditioning circuits, then transmitted to a data acquisition card for conversion into digital signals, and finally sent to a computer for analysis and processing. The monitoring software on the computer can display, store, and analyze data in real time. If the data exceed pre-set threshold values, the system immediately issues an alert, indicating potential welding defects or equipment malfunctions. Simultaneously, trend analysis of the current and voltage data provides direct insight into the stability of the welding process, offering a data basis for optimizing process parameters. The monitoring workflow is illustrated as follows:

Current and voltage sensor signals → amplification and preprocessing by conditioning circuits → conversion to digital signals by data acquisition card → analysis and processing by computer → real-time display, storage, and analysis by monitoring software → automatic alerts triggered when values exceed preset thresholds

Welding temperature is a critical factor affecting weld quality, as excessively high or low temperatures can cause defects. Therefore, real-time monitoring of the welding temperature in large-diameter, thick-walled spiral steel pipes is highly significant. Currently, commonly used temperature monitoring methods include thermocouples and infrared thermometry. Thermocouple measurement relies on the thermoelectric effect between two different metals, converting temperature changes into a corresponding thermoelectric potential. The measuring junction is placed in direct contact with the welding area to enable real-time temperature measurement.

This method offers high accuracy and a fast response, but contact measurement can interfere with the welding process. Infrared thermometry, on the other hand, measures temperature based on the infrared radiation emitted by the object. Objects emit infrared radiation at a given temperature, with the intensity of the radiation corresponding to the temperature. Infrared thermometers detect this radiation and convert it into electrical signals, which are then processed to determine the temperature. This non-contact method offers a wide measurement range and fast response without interfering with the welding process, although its accuracy is slightly lower than that of thermocouple measurements.

In practical applications, the appropriate method can be selected according to welding conditions, or a combination of both methods can be employed to enhance the accuracy and reliability of temperature monitoring. Real-time temperature monitoring enables characterization of the thermal cycle in the welding area, providing a basis for evaluating the rationality of the welding process. For instance, in multi-layer, multi-pass welding, monitoring the interpass temperature to ensure it remains within an appropriate range can effectively prevent defects such as cold cracking. A comparison of the two temperature measurement methods is presented in Table 1.

Table 1 Comparison of Temperature Measurement Methods

Internal defects in large-diameter, thick-walled spiral steel pipes can significantly compromise their structural strength and sealing performance. Ultrasonic testing, a widely used non-destructive method, can be effectively applied for online monitoring of internal defects during the welding process. The principle of ultrasonic testing is based on the behavior of ultrasonic waves as they propagate through materials, where encountering defects causes reflection, refraction, and scattering. By receiving and analyzing the reflected wave signals, the presence, location, size, and shape of internal defects can be determined. During welding, the ultrasonic probe is positioned near the steel pipe surface to emit ultrasonic waves toward the welding area and capture the reflected signals. To enhance detection reliability and accuracy, ultrasonic testing parameters—such as frequency, probe angle, and sensitivity—must be appropriately selected. Different defects produce distinct ultrasonic reflection behaviors, and optimizing these parameters improves the detection of various defect types. Simultaneously, multi-channel ultrasonic testing from multiple angles can increase the likelihood of detecting defects. In practical monitoring, the received ultrasonic signals are processed and analyzed, then compared with preset defect signatures to determine the presence and characteristics of internal defects. When the system detects a defect, it immediately issues an alert and provides detailed information to guide operator intervention. The ultrasonic probe emits ultrasonic waves near the surface of the steel pipe.

Ultrasonic Testing Workflow:

Ultrasonic probe emits waves near pipe surface → Receive reflected signals → Process and analyze signals → Compare with preset defect signatures → Determine internal defects → Trigger alert if defects detected

Online monitoring of the welding process for large-diameter, thick-walled spiral steel pipes generates diverse and high-volume data, including visual images, welding current and voltage, temperature measurements, and ultrasonic testing signals. Constructing an efficient data acquisition and storage system is essential for ensuring stable operation of the monitoring system and effective utilization of data. Data acquisition is performed by specialized equipment that interfaces seamlessly with the monitoring sensors, capturing sensor output signals in real time. Analog signals are first converted into digital form before processing and storage. During acquisition, the sampling frequency and accuracy must be appropriately configured according to the characteristics and requirements of each monitoring parameter. For example, a higher sampling frequency can be applied to accurately capture dynamic changes in welding current and voltage, whereas temperature data can be sampled at an appropriate frequency according to welding process requirements. The acquired data is systematically stored for subsequent analysis and retrieval, typically using a database management system. Relational databases (e.g., MySQL) or non-relational databases (e.g., MongoDB) can be chosen according to the characteristics of the data and the requirements of the application. During storage, data should be systematically classified and labeled, recording information such as acquisition time, location, and welding process parameters to enable efficient management and comprehensive analysis. A comprehensive data backup and recovery strategy should be implemented to ensure data security and reliability, preventing loss caused by hardware failures or software errors.

In-depth analysis of the acquired monitoring data can reveal insights into welding quality, providing a foundation for precise process control and optimization. Commonly employed data analysis methods include statistical analysis, signal processing, and machine learning techniques.

1) Statistical Analysis: By computing basic statistical metrics (e.g., mean, variance, and standard deviation), the central tendency and dispersion of the data can be clearly characterized. For example, statistical analysis of welding current and voltage data can indicate that a large variance reflects significant fluctuations in the welding process, which may adversely affect welding quality. Applying Statistical Process Control (SPC) to generate control charts enables real-time monitoring of the welding process, with immediate alerts triggered when data exceed control limits.

2) Signal Processing: Primarily used to process and analyze various welding signals, extracting useful feature information. For ultrasonic detection signals, techniques such as filtering, noise reduction, and spectral analysis are employed to suppress noise, enhance defect signal features, and accurately characterize defect conditions. For visual image signals, algorithms such as image enhancement, edge detection, and feature extraction are employed to identify weld seam location, shape, and defect characteristics.

3) Machine Learning: Widely applied in welding process monitoring data analysis, machine learning enables the development of welding quality prediction models and defect identification models through training on large datasets. For example, algorithms such as artificial neural networks and support vector machines can be employed, using welding current, voltage, temperature, and ultrasonic signals as input features, with welding quality or defect type as output labels for model training. A well-trained model can predict and analyze new data, accurately assess welding quality and defects, identify defect types and locations, automatically learn complex patterns, and enhance the accuracy and efficiency of data analysis.

Based on the processed and analyzed monitoring data, the welding quality of large-diameter, thick-walled spiral steel pipes can be objectively assessed, with timely alerts issued when anomalies are detected. Welding quality assessment relies on preset quality standards and evaluation indicators, comprehensively analyzing various monitoring parameters during the welding process to determine whether the weld meets design and specification requirements. Assessment indicators encompass weld appearance quality (e.g., width, reinforcement height, surface defects), internal quality (e.g., porosity, cracks, incomplete penetration), and the mechanical properties of the weld joint. By comparing and analyzing monitoring data against quality standards, welding quality can be quantitatively evaluated, and the corresponding quality level determined. When monitoring data indicate a potential quality issue, the system immediately triggers an early warning, employing methods such as audible and visual alarms, as well as SMS notifications, to ensure that operators can promptly detect the problem and take corrective action. The early warning system should establish appropriate thresholds and distinct warning levels based on the type and severity of quality issues. When parameters exceed the thresholds, corresponding levels of early warning are triggered, enabling operators to rapidly assess the severity of the issue and implement targeted corrective actions. Establishing an effective welding quality assessment and early warning system enables prompt identification and resolution of quality issues, ensuring consistent welding quality.

In the steel pipe manufacturing process, vision-based monitoring technology enables high-precision weld tracking, maintaining weld deviations within a minimal range and effectively ensuring weld uniformity and overall welding quality. Welding current and voltage monitoring technology can promptly capture abnormal current fluctuations, preventing welding defects caused by unstable current. Temperature monitoring technology enables real-time observation of temperature trends in the welding area, allowing optimization of welding process parameters based on the data and significantly reducing defects caused by improper temperature control. Ultrasonic testing-based internal defect monitoring technology can accurately detect internal defects such as porosity and incomplete penetration, and promptly schedule rework to ensure the pipeline’s internal quality. Practical applications demonstrate that the online monitoring system increases the accuracy of welding defect identification to over 92%, effectively reducing the occurrence of typical defects such as cold cracking and lack of fusion, and substantially enhancing overall welding quality. Simultaneously, by analyzing and providing feedback on monitoring data, the online monitoring system optimizes welding process parameters, enhances steel pipe welding efficiency, reduces production costs, and provides robust support for the safe and reliable operation of water pipelines.

From an engineering practice perspective, this technology offers significant application value: it can effectively reduce welding defect rates, enhance the mechanical and sealing performance of steel pipes, and ensure the safe operation of water pipelines. It reduces repair costs and rework time, enhances production efficiency, ensures on-time project delivery, and provides standardized quality control solutions for water conservancy projects, thereby promoting the advancement of water transmission projects toward greater refinement and intelligent operation. Online monitoring technology for welding processes is practical, efficient, and holds broad prospects for widespread application.