Bulletin May/June 2011
Camweld is a collaboration between UK companies and research organisations with an objective to develop a multi-sensor monitoring system which could facilitate the mechanisation and automation of the arc welding process. Mechanised and automatic welding, which would greatly reduce overall manufacturing costs, cannot be applied in many cases because sensor systems are not able to record the behaviour of the weld pool in the presence of an intense arc light.
Most complex components are manually welded so that the welder can control the weld pool and accommodate any component's dimensional and material variations. In the Camweld system, an innovative camera technique combined with optical and electrical sensors was used to achieve real-time imaging of the weld pool behaviour, and facilitate automation of the welding process.
As weld defects can be caused by variations in electrical parameters or weld pool instability, a multi-sensor system is required to enable defects to be detected as they form, or give assurance that the produced welds are defect-free. Repair costs when fabricating high quality components can be as much as 30% of the total manufacturing cost. Many industries require a sensor system which can be used with existing welding equipment to reduce the repair rate by detecting defects at an early stage. Assurance on weld quality is essential in safety critical applications in nuclear, oil and gas and chemical industries. See Fig.1.
Figure 1. Photograph of TIG welding process taken with the CAMWELD system. The arc light is completely absent providing a clear view of the weld pool as it is formed.
The innovative feature of the Camweld system is the combination of vision and electrical sensors which have previously been shown individually to provide information on arc stability or behaviour of the weld pool. One of the Camweld partners, the University of Liverpool used two novel vision techniques, namely laser diode illumination, and software 'windowing' which offered the potential to prevent the arc from masking the information on the profile.
The parameter monitoring system was based on an AMV 4000 arc monitoring system produced by Triton Electronics, another Camweld partner. The basic system was modified to receive the signals from the vision and arc sensors and to display them side by side on a real time basis. The capability of the system was verified by conducting welding trials both within the laboratory and in industry. The industrial partners of Camweld representing the oil and gas, automotive and nuclear industries have indicated that the sensor system would have the potential in their industry sectors of reducing overall manufacturing costs significantly.
System details
The Camweld system successfully combined both electrical and visual sensors to monitor the weld formation and improve confidence on weld quality.
The basic principle of the arc monitoring system was to use a laser illumination source operating in the infrared region to brighten the weld pool and to filter out all other radiation using a band pass filter, so that the arc light effectively disappears from the camera image. This is effectly maximised by closely synchronising the laser pulse with the camera shutter. The images were taken with a CMOS camera with sensitivity in the near infrared region.
The arc monitoring system comprised a CMOS camera with laser illumination and associated electronics. For illumination, 32 laser diodes, each pulsed to provide 17 Watts average power was used. This laser illumination system was certified as a Class 3B laser. The laser diode cluster was connected to the camera using optical fibre. The maximum length of this optical fibre cable was 3m. The weld pool was photographed at 25 frames per second.
Fig. 2. Integrated Camweld system
The parameter monitoring system was based on an AMV 4000 arc monitor produced by Triton Electronics. It was modified to receive signals from the vision system, and to display the major arc welding parameters; the arc voltage, the welding current, the filler wire feed rate and the welding speed. The vision system output was displayed side by side on a real time basis on a computer screen. See Figure 2.
Welding trials
The purpose of the laboratory testing at TWI was to ensure that the Camweld system was capable of picking up the electrical and vision signals simultaneously, corresponding to variations in welding process, process parameters, joint configuration, and formation of defects. The welding experiments comprised bead-on-plate melt runs, as well as a number of welds on selected joint configurations, on different materials such as stainless steel, carbon steel and aluminium alloys with widely differing physical and metallurgical properties. Experiments were also performed with different welding processes such as tungsten inert gas (TIG) welding with direct current straight polarity, TIG welding with alternating current, TIG welding with current pulsing, and metal inert/active gas (MIG/MAG) welding with globular transfer, pulsed transfer and spray transfer conditions. Different joint configurations ranging from simple square butt joint to narrow groove pipe joints were used. Images recorded with the system on different materials, joint configuration, and welding processes are shown in Figure 3.
Fig. 3. Images recorded with the Camweld system on different materials and joints:
a) TIG welding of stainless steel; b) Carbon steel; c)Aluminium; d)MIG - droplet transfer in stainless steel; e)TIG welding of stainless steel withV groove; f) Burn through defect formation in stainless steel; g) Stainless steel spray transfer; h) Narrow groove pipe joint; j) Narrow groove pipe joint;
Discussion
The Camweld system is produced with the objective of providing a system capable of observing the weld pool without the interference of arc light, and the major welding parameters on a real time basis, so that any abnormalities occurring during welding with respect to the change in the welding parameters and weld pool can be dealt with manually to maintain the weld quality and consistency.
To evaluate the system's performance, welding experiments were carried out using different welding processes and process variants on a range of materials and joint configurations. The arc monitoring module was based on the AMV4000 platform which is a proven system and is being used extensively in industry. The majority of the testing was focused on identifying the capability of the vision module to record the weld pool image with different welding processes, materials and joint configurations.
The selection of base materials for welding was based on different degrees of reflectivity and surface finish. Carbon steel was at the lower end with respect to surface cleanliness and reflectivity, and aluminium was at the higher end. Surface preparation through grinding could improve the image quality of carbon steel welds; however the grinding marks interfered with the image quality. The absorption of the laser radiation by the molten metal and the HAZ was also a major factor influencing the contrast at the fusion boundary.
For aluminium the melting range was the smallest, resulting in somewhat similar levels of laser absorption by the HAZ and the molten metal, leading to poor contrast even though aluminium had the highest reflectivity. In carbon steel, the poor surface conduction resulting from the manufacturing process resulted in poor reflectivity and inferior quality images. In spite of these, the images recorded by the system on these materials were clear enough to be used for the detection and measurement of the weld pool size, and if required to be used in an adaptive control system which may be developed in the future.
The ability of the system to record the weld pool images with TIG and MIG welding processes were investigated. Different welding current, current pulsing, and metal transfer conditions were used. TIG welding images were clearer than MIG/MAG welding images. The TIG welding images provided a clear picture of the electrode tip which was useful in condition monitoring of the TIG welding electrodes, or positioning the electrode accurately with respect to the seam. One of the observations in these experiments was the influence of the welding current on the image quality. At very high welding currents (>200A) the arc was partially visible in the recorded images.
Welding experiments were carried out on different types of joints such as closed square butt joint, square butt with a varying root gap, lap joint , T butt joint, V groove joint, and compound bevel narrow groove joints. The system could record good quality images on different joint types and materials with both with TIG and MIG welding processes. The laser illumination system was powerful enough to illuminate the root of a narrow groove joint on 20mm thick carbon steel pipe. Good quality images of the root welding passes were produced demonstrating the potential of the system for use in line pipe welding applications.
Welding experiments with the introduction of defects showed the capability of the system to be used for monitoring weld quality. For example a change in welding current resulted in a change in the width of the weld pool which was clearly recorded in the images. A change in welding speed also resulted in a similar variation which could be picked up in the images produced by the monitoring system. Other defects such as burn though, weld pool flooding, incorrect positioning of the electrode, degradation of the tungsten electrode, formation of undercut, gas defects, lack of fusion and lack of sidewall fusion could be noticed in the images. This showed the capability of the system to be used for providing input signals for possible future applications such as adaptive control of the welding processes.
The integrated system provided the capability of a multi sensor system for effective monitoring of the welding process. The system was able to measure four major welding parameters; the arc voltage, the current, welding speed, and the filler wire feed rate even though in the laboratory only the voltage and the current were measured. The availability of real time images of the weld pool along with the welding parameters helped in observing the welding process more closely, and to control the process through manual corrective action.
At this point control function is not incorporated in the system however when it is done; planned in a follow-on project, it could significantly enhance the consistency of the welding process and its controllability. Results of the experiments with the integrated Camweld showed the capability of the system to be used for capturing information about the electrical parameters as well as the resulting weld quality.
The arc monitoring and vision module could be operated simultaneously and each frame generated could be correlated to the corresponding welding parameters. The voltage and current could be displaced in detail in the play back mode. This means that when the formation of the defect is detected, the welding parameters corresponding to that instant could be identified and may be controlled manually at this stage. Any variation in the welding process, for example a change in the metal transfer mode could be detected from the voltage and current wave form display. Other set up variations, for example a change in torch height, result in a change in arc voltage which was accurately recorded by the arc monitor.
Conclusions
Laboratory scale experiments were carried out using Camweld on different materials by using different welding processes. The following conclusions were made based on these experiments:
- Camweld is capable of measuring the major welding parameters and recording the corresponding weld pool simultaneously.
- The system produced clear enough images of the weld pool on material of different reflectivity and surface characteristics.
- Good quality images were produced with MIG and TIG welding process, however the TIG welding images were better than MIG welding images.
- The image quality is affected by the surface condition of the material, its reflectivity, and the absorptivity of the molten metal and the adjacent region. Stainless steel provided the best quality images. Images were clear enough to detect the formation of defects.
- The system has the ability to correlate each frame to the corresponding welding parameters.
- The system could record good quality images on different joints and materials with both MIG/MAG and TIG welding process.
To sum up, the Camweld system is capable of measuring the major welding parameters and recording the weld pool images simultaneously, thus enabling close monitoring of the weld quality produced.
Further work
The present Camweld system has the ability to record the weld pool images and welding parameters simultaneously. To facilitate complete mechanisation of a welding process, the system should have the ability not only to view the weld and process parameters, but also have the ability to indicate any instability/defect formation. This requires a system capable of measuring the weld dimensions, intelligence to distinguish between a good and a bad weld, and the ability to identify the most important parameters requiring change to bring the process under control.
TWI is keen to develop the Camweld system further though a Group Sponsored Project (GSP) or a collaborative project incorporating the above mentioned capabilities so that the system can be used as a feedback control system enabling the automation/mechanisation of welding processes. For further information on this proposed project, please contact Vinod Kumar (vinod.kumar@twi.co.uk)
Acknowledgement
This work was carried out under the Camweld project with Technology Strategy Board (TSB) funding. The author would like to thank the members of the consortium; Subsea 7 Pipeline Production Ltd, Serimax Ltd, Sellafield Ltd, National Nuclear Laboratory, Triton electronics, University of Liverpool, ThyssenKrupp Talent Limited, and TWI Ltd for permission to publish this article. The authors would also like to thank Mr Jitendra Patel the TSB Monitoring Officer for his support throughout the project.
About the author
Vinod Kumar is a Principal Project Leader in the Manufacturing Support Group of TWI. He joined TWI in 2006 with a PhD in engineering after working in the aerospace industry and at Cranfield University for more than 17 years. He has managed several projects for a wide range of industries, mainly related to materials and joining processes. He has expertise in many arc welding processes, and has carried out several research projects for optimising the process procedure combination for a given application. He is a Chartered Engineer, a senior member of the Welding Institute, and a member of the Institution of Mechanical Engineers.
