Laser Decommissioning of Reactor Cores and Structures: U.K. Dragon and Trawsfynydd Reactors

The Dragon reactor site in Winfrith, England.

At Dragon, which is currently being decommissioned, laser cutting is being used to remotely segment the reactor pressure vessel (RPV), the remaining reactor internal components, and the outer thermal shields. A 10-kW ytterbium-doped fiber laser, deployed from a remotely operated mast-mounted manipulator, is being used to size reduce the reactor within a shielded cell. The laser was selected over other cutting techniques because of its tolerance to “standoff” distance (the distance between the cutting head and the work face), reliability, and ease of deployment.

Laser cutting is a technique that NRS has deployed on a number of small-scale projects to achieve specific objectives over the past decade. The benefits of this technology to nuclear decommissioning applications are numerous and are already being demonstrated at the Dragon reactor during remote segmentation operations. NRS is now undertaking laboratory trials to prove the capability and benefits of laser segmentation on representative samples of structures that will be encountered in the Magnox reactor at Trawsfynydd.

The Dragon project

Fig. 1. Dragon reactor cutaway and inner containment photograph during the operational phase.

The Dragon project began in the late 1950s by a consortium of 13 European countries to construct and operate an experimental reactor at Winfrith. Criticality was first achieved in 1964, and the reactor operated until 1975. Aside from its role in proving the concept of a high-temperature, gas-cooled design, research principally focused on the design and testing of different fuel arrangements and chemistries [1].

A 30.48-meter-diameter cylindrical building was constructed to house the reactor and provided several layers of protection to workers and the environment. Principally, this consisted of outer reinforced concrete walls, a steel shell that hermetically sealed the inner containment of the reactor building during operation, and a biological shield wall around the RPV. Thermal shield tanks, through which cooling water was fed during reactor operation, surrounded the sides, base, and top of the reactor.

Due to the level of contamination and dose rates associated with the working environment, along with the inability to safely gain access using more conventional methods, decommissioning work completed since 2018 at Dragon has relied more heavily upon remote techniques. It is the intention of the project to complete decommissioning of the RPV and reactor core using a top-down and inside-out approach, with waste recovery campaigns arranged into distinct low- and intermediate-level radioactive waste campaigns.

Figure 1 shows an original sectional view of the reactor alongside a historic photograph of the upper biological shield wall housing the charge machine and upper part of the RPV during the operational phase.

Between 2020 and 2024, a head cell and associated waste segmentation and retrieval plant was constructed and commissioned within the Dragon building. The primary waste retrieval and segmentation tool being used by the project is a Walischmiller A1000S mast-mounted manipulator, referred to as the Reactor Core Segmentation Machine (RCSM). This has been adapted to enable deployment of the 10-kW laser-cutting system.

Separate works and site testing was conducted on each of these subsystems before integrated testing was completed on-site in 2023. Currently, the project is in an early operations phase where more benign structures inside of the biological shield are being segmented and removed.

The design of the waste segmentation and retrieval plant is illustrated in Fig. 2. Waste segmented and retrieved by the RCSM is deposited into a pair of bins, packaged according to predetermined plans. A remotely driven trolley, with an empty concrete box loaded onto it, is then deployed into the head cell through a set of shield doors. An 18-ton-capacity crane is then used to deposit the bins into the concrete box and apply a temporary lid. During placement into the box, the waste packages are transferred through a removable assay detector array to confirm activity levels are in accordance with the anticipated values. Shield doors at either end of the head cell enable the crane and RCSM to be withdrawn into maintenance areas where personnel access is available.

Fig. 2. Dragon reactor decommissioning scheme design.

Laser-cutting system

During the development of the reactor decommissioning scheme, an options study was undertaken to establish the most suitable tooling to segment the reactor. A solution was required that could be flexibly applied to a series of different geometries and materials, and which could practically be deployed by the RCSM.

Laser cutting offers a series of potential benefits when compared with other methods, such as the use of diamond wire sawing and plasma cutting. Most attractively for the Dragon project, the possibility of cutting through thick steel sections in a way that is relatively insensitive to the cutting-head offset proved to be the vital factor as the structure mainly consists of convex and concave surfaces. The ability to cut through multiple layers of material is also extremely useful in a decommissioning environment.

Although there is a high initial capital cost that must be recognized, the main components of the system are effectively solid state and have been proven extensively within other industries to be reliable for the long term. Fiber lasers such as the unit being used at Dragon are typically used in “lights-out,” 24-hour manufacturing applications and have projected operating lifetimes in the region of 100,000 hours. There is therefore anticipated to be longer-term benefits in terms of reduced maintenance burden and an increase in overall safety and waste production rate by virtue of scaled-down direct interaction with potentially contaminated equipment. The extended useful life of the laser and its relative portability mean its cost may be offset by applications on other projects. It is only the relatively inexpensive components that become contaminated during the decommissioning operations, whilst the laser generator can be located in a remote, clean area.

Particularly during the design phase of the project, where the feasibility of laser cutting could only be established theoretically, it was recognized that there were some fundamental risks with proceeding with laser cutting as such an integral part of the overall design solution. To reduce the level of risk exposure and to provide a firm basis for the approach, a series of trials were commissioned by NRS that were performed by The Welding Institute (TWI) in Cambridge, U.K. During the trials, TWI tested representative samples of reactor components that were manufactured from original drawings. The aim of the research was to both confirm the overall feasibility of using laser cutting on each component as well as to develop laser setup parameters that could be deployed operationally. There are four continuously variable parameters that can be manipulated during laser cutting:

1. The laser beam power in kilowatts.

2. The position of the process head nozzle from the workpiece, known as the standoff, measured in millimeters.

3. The assist gas pressure setting at the cutting head, measured in bar.

4. The rate at which the laser is moved across the workpiece, measured in millimeters per minute.

Each of these parameters can be adjusted to suit a specific use case. There is, however, some crossover in the effect of each of these. For example, the rate at which the work piece is cut may generally be increased either by increasing the beam power or reducing the standoff. Depending on the specimen being segmented, maximum achievable cutting rates between about 75 mm/min and about 350 mm/min were demonstrated in trials. Testing was also completed for certain specimens on both the horizontal and vertical orientations to account for differences in the way molten material is cleared from the kerf.

One of the attractive features of laser cutting is its ability to segment relatively thick work pieces without generating a wide kerf. Trials data from other studies also demonstrate the potential for laser cutting to produce fewer secondary emissions than other thermal techniques. Unlike other applications, where cut quality and speed are the optimization parameters, in a nuclear setting it is generally advantageous if kerf material adheres to the work piece as dross to reduce the airborne fraction as much as possible. This goal may be achieved by tuning the parameters listed above, with an increase in standoff distance and a decrease in assist gas pressure being positively correlated to the degree of dross adhesion.

System design

Fig. 3. Dragon reactor decommissioning project laser cutting system.

Figure 3 shows the configuration of the Dragon laser-cutting system. The specification of the system by NRS required the designer, TWI, to make use of a majority of off-the-shelf components to reduce the technology risk to the lowest level possible and ensure reliable operation. Interfaces with the RCSM and the rest of the segmentation plant are intentionally minimized.

The system can be grouped into five main areas:

The laser process head. This is the termination point of the optical chain and consists of a robust outer casing that can be handled by the RCSM; a cutting nozzle that fires a high-velocity jet of air concentrically with the laser beam; a collimator and focusing lens assembly that focuses the laser beam to a spot size of approximately 0.5 mm diameter, 15 mm in front of the cutting nozzle (the optics are protected by a sacrificial cover slide); a laser rangefinder for the precision setting of the cutting head standoff; and a pressure transducer to provide feedback on the assist gas pressure inside of the head.

The RCSM. This was a custom solution, enabling integration of the laser-cutting deployment system and deployment into the RPV.

The laser source, optical fibers, and the chiller. The optical fibers are more than 80 m in length and have a core diameter of 200 micrometers. Chilled water is fed to the process head optics through the energy chains.

The laser safety interlock system.

The compressed air system. This is a bespoke system based on off-the-shelf components. Moisture is removed from the air to prevent misting of the optics.

The laser source itself is a 10-kW IPG YLS-10000 Yb continuous-wave fiber laser that operates at a wavelength of 1070 nanometers, in the near-infrared part of the spectrum. Internally, the laser contains 11 independently selectable laser modules that are then combined to produce a single-output feed fiber. In normal operation, all 11 modules are activated and their pump current attenuated according to the required level of laser output power. Should a fault occur with any module (the Dragon project has yet to experience this), they can be deactivated to allow uninterrupted laser operation at reduced power until a replacement can be made. The feeding fiber is connected to a two-way beam switch inside the laser source, which allows the user to select an output channel for the laser into which the process fibers can be installed and which are responsible for transmitting the laser energy to the process head. Chilled water is continuously fed through the laser source and process head optics by a dedicated unit.

The process fibers are considered consumable items by the manufacturer. Due to the degree of integration required with the RCSM and the difficulty in replacement during active operations, however, it was recognized at an early stage that it was advisable for at least one replacement fiber to be pre-routed to the process head. In the experience of the Dragon project, although the process fibers are remarkably robust when in use by the RCSM, they are susceptible to damage under maintenance conditions if not handled correctly.

High-power fiber-optic filaments are commonly terminated with a relatively large-diameter quartz block, which is almost exclusively the part of the fiber damaged on the project. Modest mechanical impact to the connector or the block itself is often sufficient to cause chipping or fracturing, with the only recourse available being to replace the entire fiber. During the process of commissioning, two additional fibers (taking the total to four) were routed to the RCSM to provide additional redundancy should failure occur during active operations. During the remote operations phase, failure of the optical fibers may also occur due to, for example, inadvertent cutting of a highly reflective material, such as one of the many pure silver gaskets used in the construction of the reactor.

Cleanliness of the optical elements is of utmost importance for fault-free operation of the laser-cutting system. Exposure of these surfaces under maintenance conditions must be tightly controlled to avoid even the most minor foreign debris from contaminating the surfaces. To this end, a laminar airflow hood is available within the maintenance cell should the process head optics need to be exposed as part of maintenance work. Under normal operational conditions, a low-pressure purge stream of air is fed through the nozzle even when not cutting to ensure contamination cannot backflow toward the process head optics. If contamination were to occur, then this can potentially lead to burn-in and subsequently to rapid and catastrophic damage of the optics.

Alongside the main operational system, a power meter is also permanently installed into the maintenance cell to enable periodic testing of the optics. This will provide indication of degraded system performance that could either prompt remedial work on the optics components or provide advanced warning of failure.

With the exception of the laser process head and the RCSM, the entire system is located outside of the head cell containment in a radiologically supervised area. This enables convenient access for maintenance and troubleshooting tasks. The laser services comprising the laser fibers, chilled water lines, compressed air, and signal cables are routed to the process head through a series of energy chains that accommodate the movements of the machine. A bespoke reeving system on the RCSM itself enables deployment of the laser services into the core, following the position of the laser process head. These are encased in an IGUS Triflex R chain, which provides mechanical protection.

Laser safety is achieved by designation of the entire Dragon building as a Class 1 laser safety enclosure. Aside from the obvious hazards present if one was to be in the direct vicinity of the cutting operation, the main risk is ocular damage resulting from reflected light. This does not require line-of-sight visibility to the laser. In general, the laser is operated using a key-enabled handheld control pendant located in the control room that must be pressed by the operator for the duration of the cut. Emission control can be conducted internally to the software, which allows for very precise control of laser on time and power.

Early operating experience

Remote dismantling of the reactor and ancillary structures began at Dragon in July 2024 after a four-year period of construction, commissioning, and nonactive trials. Early work is focused on removal of the top ring thermal shield (TRTS) tank, which is a 50-t lamellar structure of plates with interstitial air gaps positioned around the upper neck of the RPV.

Fig. 4. TRTS top plate segmentation (TRTS highlighted in yellow).

The segmentation strategy for the TRTS involves removing the top layer of plates, which are welded onto the underlying structure, to expose the inner plates, which are then removed without bulk segmentation. The TRTS is identified in Fig. 4, along with a photograph showing cutting of the 25-mm-thick top layer of plates. Areas of the TRTS are up to 75 mm thick, which the Dragon project team have demonstrated can be segmented in a single pass with the appropriate laser settings.

It is of note that although information gained from laboratory trials may be used as a starting position for laser parameter setup, it is often the case that in a real-world setting, complicating factors exist that necessitate alterations. For example, adjacent structures may preclude access close to the cutting area and require a greater standoff during cutting.

Compensation may be required to the cutting speed, assist gas setting, and beam power. Larger standoff distances (in the order of 100–150 mm), compensated for by a higher beam power, are typically being used to release the plates than were determined in trials, as this widens the kerf and eases subsequent removal. For thicker areas, however, a 15-mm standoff is needed to provide the requisite beam intensity and assist gas velocity throughout the plate depth.

A remotely operated battery magnetic lifter is used to remove the plates once they have been severed from the structure. Due to its relatively benign nature and absence of significant levels of surface contamination or activation products, remote operation on the TRTS removal is being used by the Dragon team to complete elements of active decommissioning of the plant.

During this phase of the project, several important points of learning have already been established that are beginning to prompt possible enhancements to ease the process of setting up and performing cuts. Comprehensive records for each of the cuts made are being collected for trending purposes with a view to compiling an optimum set of cutting parameters that can be used for more activated and contaminated structures.

Initial findings indicate that the achievable cutting rates demonstrated in the laboratory trials for a representative sample of the TRTS do not match those found to be possible on the actual plant. Initial empirical analysis tends to indicate that this is mainly attributable to less efficient clearance of the molten material away from the back of the cut kerf between the layers of the TRTS. Significant levels of surface rust also exist both on top of and within the TRTS, and the rust has been observed to degrade cutting performance. Another notable factor is the manipulator’s inability to hold as accurate a cut path as the robot platform used during trials. Slight meandering of the beam path across the surface reduces the cumulative heat input within each part of the kerf, thus requiring a longer dwell time to melt the material and eject it from the kerf.

Achievable cutting speeds have been in the region of 50 percent lower for comparable laser settings. There is, however, significant variation. Depending on the margin between the cutting speed and achievable speed with the given settings, this may lead to a stop in material penetration. When this happens, it is necessary to reverse the cut path to a position at which penetration was achieved and then resume the cut to reestablish melt-through at the tip of the cut kerf. Exploratory techniques are currently under development on the project to dynamically alter the cutting speed according to visual feedback indicators to improve the efficiency of processing.

Perhaps the main point of learning for the Dragon team so far is that the rate of laser processing is generally not governed by the speed at which the laser-cutting system can perform the actual cutting operation. Each cut is generally programmed into the RCSM prior to its being performed using “teach and repeat” functionality, requiring the operator to position the machine tool center point on a series of locations along the desired cutting path. This activity requires the operator to first move the RCSM into the cutting area using the prepositioning system before using the manipulator arm to describe the cutting path. The average efficiency experienced in the earlier part of the TRTS removal for each cut was 16 percent for beam on time, although this has now improved significantly due to an increase in operator experience and process development.

Fig 5. Removal of the Dragon reactor RPV neck.

Continuous improvement techniques typically used in factory production, alongside improvements to the operational plant, are being applied by the Dragon project to increase the rate of operating efficiency over the longer term. Keeping comprehensive records of each of the cuts being performed is a cornerstone of the operational work being completed at Dragon. In-house tools have been developed and are being optimized that enable the operations team to log the positioning, angle, and speed of movement of the cutting head and to keep records of the laser settings being used to perform each cut. Variations are made to the system parameters to establish the relative effects on cutting performance associated with alteration of these parameters.

More recently, one of the structures the project has contended with removing is a section of the RPV neck that protrudes above the reactor plenum chamber (Fig. 5). Removal of this neck was originally scheduled to be completed following segmentation of the TRTS, however an opportunity to improve the available operating range of the RCSM was identified by executing this task early. Over the course of three days of operation, the neck was removed and exported in sections each weighing approximately 0.5 t. Use of the laser-cutting system to create handling holes for recovery of each section proved to be a particular point of success. The removal of this part of the RPV neck posed a very different challenge to the TRTS and was excellent confirmation of the realizable range of practical capabilities possessed by the laser-cutting solution and of its subsequent application to other structures within the RPV.

The Trawsfynydd project

Trawsfynydd is a double reactor site, with each reactor consisting of an approximately 8.5-m-high graphite core with support structure housed within an 18.5-m-diameter, 90-mm-thick spherical steel pressure vessel, with up to 300 mm of asbestos insulation covering the lower sections. The steel RPV is housed within a concrete bio-shield, whose walls are approximately 3 m thick with an approximately 4-m-thick pile cap.

To date, an initial concept design phase and subsequent optioneering has been completed for the decommissioning of Trawsfynydd. Three reactor dismantling approaches were separately developed to a concept level by three separate consultant organizations. Following completion of these designs and acceptance by NRS, a process of integrated optioneering was undertaken in-house by the decommissioning project team [2].

Different aspects of each of the individual concepts were assessed, with the favorable aspects of each taken forward as a basis for the design of the dismantling scheme. From the concept designs, a suite of waste campaigns was established that broke the reactor dismantling down into sequential steps based on the components for removal and processing. Figure 6 shows the sequence of work to dismantle the reactor.

Fig. 6. Overview of reactor decommissioning (shown on a sectional view).

The use of laser technology for component size reduction was initially identified in the submitted decommissioning concept designs. The technology underpinning and details of its deployment within the reactor dismantling were limited at the time, given the early stage of the designs and level of maturity. The intention of the laser trials was to bolster the decommissioning toolkit of approaches and technologies available by developing in-house knowledge of laser technology and assess feasibility of deployment for reactor dismantling.

The laser trials would ultimately support the development of the dismantling scheme through the incorporation of the conclusions of the technical underpinning. The areas of focus for the laser trials were the capability of laser technology to undertake the challenging cutting scenarios anticipated within the reactor dismantling, namely:

Cutting metallic components clad in insulating materials mimicking RPV construction.

Cutting large ducts, resulting in significant standoff of the cutting head from the component.

Cutting metallic waste components in close proximity to other components, with a focus on any effects of laser “overshoot.”

Trials were undertaken at Onet Technologies’ Technocenter in France using a 14-kW laser mounted to a STAUBLI robotic arm within a ventilated cutting cell. Based on the most challenging cutting operations anticipated as part of reactor dismantling, four general test pieces, each designed as a direct replica of a reactor dismantling cutting scenario, were cut. The test pieces were subjected to numerous trials with different cutting configurations to maximize useful data gathered from the trials. Data of interest for each test piece were minimum and maximum standoff distances, minimum laser power requirements, optimum cutting head travel speed, and any other limitations observed.

The laser-cutting trials concluded in October 2024 and all test pieces were successfully cut and data gathered. Overall conclusions and noteworthy observations for each of the test pieces are summarized in the table opposite.

The Trawsfynydd laser trials demonstrated that the application of laser-cutting systems is feasible for the commercial-scale dismantling of reactors and that the technology is suitably versatile for cutting the largest and smallest components anticipated within the reactor dismantling, while also being able to execute cuts of multiple materials in one pass.

Data from the trials conducted have been provided to NRS and are being shared with contract partners responsible for delivering decommissioning design schemes for inclusion within the overall decommissioning plan. n

REFERENCES

1. Price, M.S.T and L.R. Shepherd, “DPR1000: A Summary and Evaluation of the Achievements of the Dragon Project and Its Contribution to the Development of the High Temperature Reactor,” (1978).
2. Tilsley, Matthew, “Decommissioning of the U.K. Trawsfynydd Magnox Gas Cooled Reactor,” WM2022 Conference, Phoenix, Ariz., 22243 (2022).

3. ACKNOWLEDGMENTS

Walischmiller GmbH, The Welding Institute, Onet Technologies.

James Reed is a lead commissioning engineer and Chris Ewing is a senior project engineer with U.K. Nuclear Restoration Services.

This article is based on a paper presented at 2025 Waste Management Conference, presented by Waste Management Symposia, Mar. 9–13 in Phoenix, Ariz.