Revision as of 06:56, 15 March 2023 by Stephan Kallee at AluStir (talk | contribs) (→‎Summary)
(diff) ← Older revision | Latest revision (diff) | Newer revision → (diff)
Jump to navigation Jump to search
The weld in the centre doesn't look good, the best weld is shown below
Three friction stir welds made during a training course in Munich
The weld in the centre doesn't look good, the best weld is shown below. View from the top
Three friction stir welds made during a training course in Munich
The roots of the three welds look very similar
The roots of tree friction stir welds made during a training course in Munich. The start is at the right hand side, the stop is a the left hand side

The Vision of the aiCAMstir project was determined during a friction stir welding training course in Munich.


Some years ago, Stephan Kallee had worked on a project on making tailor welded blanks by friction stir welding. A new spindle, a new jig, two FSW experts and a new technician were available, who was experienced in operating milling machines but had not received in-depth training on parameter optimisation. Due to an unforeseen business trip, they worked on three different sites: in the lab, in the office and in a hotel. They could only communicate by mobile phone, sharing photos and information on the parameter settings. After a few iterations, optimised parameters could be found.

A few years later, he considered using neuro-networks for optimising FSW parameters. During a training course in Munich, the concept of using artificial intelligence in computer aided manufacturing was discussed in more detail and a project with very promising results was conducted.

In the aiCAMstir project, we want to go one step further. We want to create a cloud, into which FSW operators can upload images and information on parameter settings during feasibility studies, prototyping, production ramp-up and series production, and get feedback about the weld quality and recommendations on optimising the parameters. In the final stage, such a system would be integrated into the FSW machine, and the machine would optimise the parameters itself within boundaries set by the operator.

The vision was explained for the first time in public during an on-line discussion with students of the University of Liège on 19 April 2021.

The first welds

The first welds were made by a group of experts and trainees in Munich. The trainees could choose the parameters after discussions with the experts and found visually acceptable parameter settings after a few iterations. The parameter matrix investigated during the course is shown below:

Visual appearance of five friction stir welds made during a training course at Munich
Photo Rotation Speed Welding Speed Downward Force Commment
A lot of flash was expelled 5000 rev/min 500 mm/min Not documented Too hot
Weld looks good, possibly a bit too hot 1700 rev/min 500 mm/min Not documented Visually acceptable but too slow
A lot of flash was expelled and the weld surface looks inhomogenious 5000 rev/min 1000 mm/min 7.6kN Too hot
Reasonable amount of flash but inhomogeneous weld surface 3000 rev/min 1000 mm/min 9.2kN Too hot
Weld looks good 1300 rev/min 1000 mm/min 9.8kN Visually acceptable, best weld so far

The next step: Automated PAUT-TOFD

Screenshot of PAUT-TOFD examination with the Sonaflex system, © Pavel Pashkov, Nordinkraft AG, CC BY-SA 4.0

The aiCAMstir concept might also be useful for PAUT-TOFD examination with the Sonaflex system. The Sonaflex system makes use of an industrially proven, automatic ultrasonic inspection system combining Phased Array Ultrasonic Testing (PAUT) and Time of Flight Diffraction (TOFD) developed by Nordinkraft AG. The aiCAMstir software could be modified to evaluate screenshots of the NDT data in the same way as processing optical photographs for automated quality control during friction stir welding. Please contact to discuss this in more detail.

Non-destructive testing concept for automatic examination by LRUT, PAUT or X-ray

Concept for Long Range Ultrasonic Testing or Phased Array Ultrasonic Testing during Orbital Welding of Pipelines, © Stephan Kallee, CC BY 4.0 using a photo of X Niu et al, CC BY 3.0.

The aiCAMstir project proposes to investigate a new concept for in-process long-distance ultrasonic examination or film-free X-ray examination during friction stir welding or other welding processes with adaptive control of the process parameters.

Especially when laying pipelines on construction sites or on the high seas using the J-lay or S-lay method, the quality of the weld should be examined during welding. If weld defects are detected during the process, the parameters can be adaptively controlled, for example the welding speed can be reduced or the contact pressure increased, or a repair weld can be carried out in the same clamping, e.g. in case of wear or breakage of a friction stir welding tool.

State of the art

After welding of tubes and pipelines, non-destructive testing is required for many applications in order to evaluate and document the quality of the weld seam. Up to now, mainly X-ray examinations or phased array ultrasonic examinations have been carried out.

Phased array ultrasonic testing (PAUT) and Time of Flight Diffraction (TOFD) are usually performed after welding, because a coupling medium is needed between the piezoelectric sensors of the transducers and the workpieces to examined. The high temperatures of the workieces during welding causes the coupling media to evaporate as they are usually water, oil or some sort of glycerin based gel.

For in-line examination the PAUT or TOFD sensor should be kept at least several centimetres away from the rotating friction stir welding tool, if water or gel are used to couple the ultrasound, which is why the material to be examined must have cooled down at least to such an extent that the coupling medium does not evaporate in the area of the ultrasonic probe. For this reason, the phased array ultrasonic head runs at a distance of several centimetres or a pipe length away from the welding tool during in-line ultrasonic examination, making adaptive control of the process impossible.

An alternative solution would be to use electromagnetic acoustic transducers (EMATs), which do not require physical contact between the sensor and the test-piece and thus could operate with an air gap between sensor and the pipe (lift-off distance). However, the signal to noise ratio (SNR) of EMATs is not as good as the SNR of piezoelectrics and temperature will affect the impedance (and thus the response) of the sensors and some sort of compensation will be required if the parts are to be inspected while hot.

As a rule of thumb the wavelength of the ultrasonic wave must be approximately the size of the defect one is searching for. So, for example, transducers operating at 3 MHz can be used in steel (considering a longitudinal velocity of 5900 m/s) searching for a defect with an approximate size of 2 mm.[1]

For X-ray inspection of pipelines, in many cases a radioactive emitter or X-ray tube is inserted on a long rod into the pipe, which has previously been wrapped with an X-ray sensitive film. This is not possible during friction stir welding because the film cannot be wrapped 360° around the pipe because the welding tool makes the pipe circumference inaccessible in places. For this reason, the X-ray inspection is often staggered by one pipe length at the same time as the next weld. If defects are found, repair welding is difficult and time-consuming because accessibility for the welding head at the NDT station is usually obstructed by the NDT equipment.

Long-range ultrasonic testing

Concept for Long Range Ultrasonic Testing or Phased Array Ultrasonic Testing during Orbital Welding of Pipelines, © Stephan Kallee, CC BY 4.0 using an unrelated photo of Bundesanstalt für Wasserbau, CC BY 2.0.

Long-range ultrasonic testing is a non-destructive testing method using guided ultrasonic waves. Compared to other NDT methods, guided waves can propagate over several metres with a relatively high sensitivity to defects in the structure. The general sensitivity range is up to 3% of the cross-sectional area, depending on the signal-to-noise ratio. Further optimisation of the inspection process with guided ultrasonic waves is possible. This can be done by optimising the transducer array.

Guided waves are usually employed with lower frequencies (tens of kilohertz range) in order to achieve a larger area of inspection (lower frequencies provide lesser attenuation). In this scenario, they provide information regarding a loss of cross-section (3 % being an accepted limit for inspection; this can be lowered to 1.5 % if monitoring is performed, but it is not the case here) and might only detect larger defects. It is also possible to move the sensors closer to the area to be inspected (weld in this case) and to use higher frequencies in order to obtain a better resolution of small defects. There are solutions for tomography with lamb waves that could be interesting for this application. It is also possible to use EMATs in this scenario. Again, inspection while the component is hot will affect the speed of sound in the material, and although this does not impede inspection, additional processing of the data might be needed.[1]

Depending on the resonance frequencies of the system to be examined, it may be advantageous to use sound waves instead of ultrasonic waves for in-process non-destructive testing. In an advantageous concept variant, multiple transmitters and receivers are used. These transducers are placed on the left and right (or above and below) of the joint or weld.

Especially in friction stir welding, it is also advantageous to use a set of transducers in the rotating tool spindle or near the spindle bearings. The sound or ultrasound generated with these can then be received and processed, possibly with a time delay, at various points.

In arc or beam welding processes, it is advantageous to modulate the welding current or the laser or electron beam energy in the sonic or ultrasonic range to generate sonic or ultrasonic waves in the molten metal that can be received by various transducers or modulated for transmission.

Acoustic Emission

If the acoustic transducers are set to work only on passive mode (there is a chance that accelerometers/vibration sensors could also be used) and to monitor the welding process it might be possible to correlate the acoustic waves with weld quality. As a comparison, for example, there are some works of monitoring with acoustic emission (AE) the machining of parts with a lathe and the authors associated wear of the tool with acoustic signals.[1]


Thermography with automatic thermal imaging processing could be used for non destructive testing (NDT) of friction stir welding. The idea is simple: Heat the surface and the thermal response can be used to reveal subsurface features. An internal void (wormhole), for example, is likely to have a different thermal characteristic than an intact weld. The imaging could be based upon the welding heat or an additional heating could be applied behind the weld. Absolute temperature would not be as important as the distribution.[2]

The thermographic inspection could be carried out right after the weld in order to use the heat produced during the operation to make the analysis. Also, there is a possibility of associating the amount of heat and temperatures produced during welding with the final quality of the weld.[1]

Film-free X-ray

Filmless X-ray, i.e. digital radiography, is an advanced technology based on digital detector systems where the X-ray image is displayed directly on a computer screen without the need to develop chemicals or perform intermediate scans. The incident X-ray radiation is converted into an equivalent electrical charge and then into a digital image by a detector sensor. Compared to other imaging devices, the screen detector provides high quality digital images with a better signal-to-noise ratio and improved dynamic range, which in turn provides high sensitivity for radiographic applications. Flat panel detectors use two different approaches, indirect conversion and direct conversion. Indirect conversion flat panel detectors use an amorphous silicon photodiode array. Instead of a conventional flat panel detector, the system described here can advantageously use a concave screen detector whose inner diameter is slightly larger than the outer diameter of the tube.

Direct conversion detectors use a photoconductor such as amorphous selenium (a-Se) or cadmium telluride (Cd-Te) on a multimicroscopic electrode plate and offer the highest sharpness and resolution. The information on both types of detectors is read out by thin-film transistors. In the direct conversion process, when photons such as amorphous selenium strike the photoconductor, they are directly converted into electronic signals that are amplified and digitised. Since there is no scintillator, the lateral distribution of the photons is missing here, which makes for a sharper image. This distinguishes it from the indirect design concept.

Automatic image evaluation can provide signals for adaptive control of welding process parameters. Although this seems to be a very good option, it might be more complicated and more expensive than ultrasonic examination, e.g. because of health and safety considerations and the effects of the temperature onto the system. Also, it would be necessary to check the resolution of the flat panel detectors and the size/type of defects that are being searched for.[1]

Microfocus X-ray inspection

Systems for microfocus X-ray examination differ from conventional X-ray examination in the precise focusing of the X-ray beam, typically in the millimetre range. The advantage is that it can produce radiographs with extremely high resolution. This means that images can be produced with high magnification, whereby even the smallest defects can still be detected well.

Macrofocus X-ray examination

Instead of microfocus X-ray examination, macrofocus X-ray examination can also be advantageous, e.g. with a 450kV Broad Focus system.

Eddy Currents

The eddy current technique could be used to evaluate defects that appear at or very near the surface of the weld. The difficulty is that the skin effect limits the penetration in the sample, although there is some room to work with by varying frequencies. And again, temperature will influence the impedance of the sensor and will require compensation.[1]

Adaptive control of welding parameters

The data obtained from the in-line non-destructive testing is automatically evaluated and used for adaptive control of the welding parameters.

Pipeline welding under construction site conditions

Liebherr SR 714 LGP, with a tent for arc welding and non-destructive inspection of pipelines, © Alf van Beem

The system should be designed so that it can be used not only in the laboratory but also under construction site conditions. Depending on the requirements on site, the pipeline is either horizontal, vertical or diagonal.

Typical applications are under laboratory conditions, in factory buildings when manufacturing tubular products, in trenches when laying pipelines or pipelines on land, or laterally or above the trench, if necessary on transportable laying platforms. The process described here offers particular advantages on ships or platforms used for the production of underwater pipelines, riser pipes or flowlines, especially in the J-lay or S-lay process. The process is applicable in the manufacture of long pipelines that are wound onto a spool prior to laying, in particular for pipelines for oil, gas, water, waste water or data transport (e.g. for pipes, conduits and cables laid on land or under water).

Other industrial sectors

In addition to its use in pipeline construction in the oil and gas industry, the system presented here is ideally suited for the manufacture and laying of pipelines for energy transport (GIS and GIL), for drinking water and waste water pipelines, for rocket, satellite and aircraft construction as well as for the manufacture and sealing of containers for the transport or storage of nuclear waste. The process can also be used in plant construction, e.g. in refineries, gas separation plants and chemical plant construction, as well as in chimney and stack construction. It can also be used for load-bearing tubular structures in oil rigs, bridges or houses. It is also suitable for welding railway rails or wire ropes of suspension bridges or cable cars.

The process offers great advantages for lining wells or boreholes, especially those in oil, gas and water production, CO2 storage and geothermal energy, as well as tunnels, horizontal boreholes and mine tunnels.

Claims and Benefits

The aiCAMstir project proposes to investigate a method for in-process long range ultrasonic examination or film-free X-ray examination during friction stir welding or other welding processes:

  • If desired, friction stir welding can be performed in-process with one, two or more non-destructive testing methods, i.e. simultaneously or sequentially with LRUT and X-ray.
  • Instead of friction stir welding, the concept can also be used in combination with other welding processes, such as electron beam welding, laser welding, MIG/MAG welding or in hybrid processes such as laser-MIG-MAG hybrid welding.
  • The results or signals of the in-line non-destructive testing are processed digitally or analogue and used as input variables for the adaptive control of the welding process used.
  • Instead of conventional long-distance ultrasonic testing with one set of transducers placed in a ring on the circumference of the pipe, it is advantageous to use several sets of transducers for in-process ultrasonic testing.
  • Instead of conventional ultrasound, low-frequency ultrasound or sound can also be used or detected. It is sometimes advantageous to change the frequency during the examination.
  • The sound or ultrasound can also be generated with the aid of an electromagnetic induction coil, e.g. in a leading head for preheating or in a trailing head for slower cooling of the weld.
  • The transducer sets can be placed, especially in friction stir welding, to the left and right (or above and below) of the weld, in the rotating spindle, near the spindle bearings or at suitable points on the welding machine.
  • In arc, laser beam or electron beam welding, it is useful to modulate the welding energy in the sonic or ultrasonic range in order to generate sound or ultrasound in the molten metal which can be used, inter alia, for in-process non-destructive testing.
  • Advantageously, the system described in the concept can also be used after the welding has been completed, for example to evaluate the quality of the weld end.
  • The in-process non-destructive testing according to the concept is not only suitable for pipelines in which gases or liquids are transported, but is also suitable, among other things, for the manufacture or installation of gas insulated switches and gas insulated lines for the transmission of electricity (GIS and GIL). It proves itself both in the production in a factory and in the laying on the construction site. Other applications include the fabrication of round, oval or polygonal pipe or tank structures, particularly in the aerospace and nuclear industries.
  • The concept can also be used in the welding of tanks for the storage of gases and liquids, in particular for the manufacture of liquefied natural gas (LNG) tanks made of aluminium or other materials, and in storage tanks for the storage of chemical or petrochemical gases or liquids.
  • The concept can also be used in plant construction and for load-bearing tubular structures in oil rigs, bridges or houses.
  • The concept can also be used for welding railway tracks or supporting cables of suspension bridges.
  • For digital radiography, instead of a conventional flat screen detector, the system described here can advantageously use a concave screen detector whose inner diameter is slightly larger than the outer diameter of the tube.
  • Two or more detectors may be used, since in orbital friction stir welding it is advantageous to use two friction stir welding tools offset by, for example, by 180°, or three offset by, for example, by 120°, or several offset by a suitable angle.

Please contact if you are interested in investigating this innovative concept.

Automated shearography

aiCAMstir concept for friction using shearography and artificial intelligence in friction stir welding, © Stephan Kallee, using photos of Shearo and TheGrenzebachGroup, CC-BY-SA 4.0

The aiCAMstir project participants are currently talking about setting-up a government-funded collaborative project on using in-line shearography for closed-loop parameter control during friction stir welding (FSW) or friction stir spot welding (FSSW) of thermoplastics in the high-volume production of mixed-material plastic/aluminium battery trays.

A growing interest in reducing the weight and cost of battery trays has been noted from OEMs and Tier 1 suppliers from the US, Europe and Asia. A hybrid structure makes the most sense for weight, stiffness and impact mitigation. FSW and FSSW allow that to happen, particularly when suitable parameter monitoring systems and control algorithms are being used in combination with automated non-destructive testing procedures.

Potential users of the aiCAMstir system

The most successful users of friction stir welding are divided in three groups:

  • A group of PhDs, who industrially apply the process for safety-focussed applications, such as space rockets, railway rolling stock, shipbuilding panels and pressure vessels according to well established regulations using expensive and highly-automated equipment
  • A group of experienced machine operators, who exploit the cost benefits of FSW in high-volume production, who commonly use single-purpose machines or robots after a few days training, to fulfil the requirements of ISO 25239.
  • A group of highly motivated students and academics, who investigate the benefits and limitations based on “learning by doing” as part of challenging projects often using conventional milling machines or home-made FSW machines.

All these groups will benefit from applying artificial intelligence during the following aspects of their FSW related work:

  • Feasibility studies, prototyping and production ramp-up
  • Development of Welding Procedure Specifications (WPSs) and Welding Procedure Qualifications (WPQs)
  • Day to day industrial production with often unforeseen problems caused by workpiece tolerances, alloy variations and FSW tool geometry variations

The project focusses on generating and sharing know-how to FSW process users, FSW machine builders, NDT equipment manufacturers, data mining experts, enthusiastic academics all over the world.

Notes and references

  1. 1.0 1.1 1.2 1.3 1.4 1.5 The draft of this article has kindly been reviewed by Prof Ricardo Callegari Jacques of Universidade Federal do Rio Grande do Sul (UFRGS) and Laboratório de Metalurgia Física (LAMEF) and subsequently been expanded, incorporating his comments.
  2. Dr Simon Smith contributed his idea on using thermography, following the discussions before and during the Second aiCAMstir Meeting, 30 September 2021.