Challenges for robotics in nuclear decommissioning
Decommissioning the UK’s legacy nuclear facilities that can be over 50 years old is one of the most complex engineering tasks that the UK currently faces. Nuclear decommissioning is difficult as the radiation levels in a vast number of facilities are too high to allow manual operations, so remote or robotic solutions are required.
Author(s): Tom Robinson
While there have been many successes in remote decommissioning, there are continuing challenges to overcome. Remote solutions need to: be deployed through small penetrations but operate at heights or distances >10 m; operate in restricted and congested environments while completing aggressive operations (such as cutting, shearing or scabbling); and work in high-radiation fields yet be reliable, reusable and disposable. Advanced robotic techniques may provide benefit by reducing the need for centralised communication, enhancing adaptability in uncertain conditions or increasing the productivity and reliability of deployments. This article discusses the common difficulties of remote and robotic decommissioning through examples of legacy operations and current research and development. Though this article does not come to any conclusions, in identifying the challenges of decommissioning legacy facilities, it is anticipated that optimised solutions can be developed.
The nuclear industry in the UK is over 60 years old with the first nuclear facilities, the windscale pile reactors, commissioned in 1952. A large number of nuclear facilities have reached the end of their operational lifetimes and such as all industrial facilities, need to be decommissioned and demolished. Sellafield Ltd. decommissioning operations are increasing with the onset broad-front decommissioning around 2020; the Magnox reactor fleet has ceased power generation and is now in initial decommissioning or care and maintenance; and Dounreay and other nuclear research sites are making significant strides toward their final end states. Decommissioning in the nuclear industry is the term for the actions taken to allow the removal of regulatory controls . In practical terms, decommissioning is the retrieval, dismantling and packaging of nuclear material, process equipment and civil infrastructure from facilities such that the facilities have no nuclear implications. Some of the legacy facilities at the Sellafield site have been described as ‘significant risks to people and the environment’  and pose the greatest decommissioning challenge. The high-radiation levels and ageing infrastructure of the legacy facilities do not allow man-access and require a remote or robotic decommissioning solution. This article details some of the challenges in the remote and robotic decommissioning of legacy facilities at Sellafield Ltd. (the workplace of the author). This article will not aim to solve these problems but provide lessons learned from the previous work and ongoing research and development to act as a stimulus for future decommissioning operations.
Nuclear is different
You will often hear that ‘nuclear is different’ . Exposure to radioactivity and a high absorbed dose can have significant health effects. This means the nuclear industry has a traditional conservatism against operational risk to ensure safety and right-first-time operations whilst protecting the workforce and the general public against potential hazards. A clear example of this was in decommissioning the prototype fast reactor (PFR) fuel facility. The facility had a number of gloveboxes within a secondary containment used to manufacture PFR fuel. The plan was to use a remote demolition machine to dismantle the gloveboxes. However, it was thought that in a fault scenario, the demolition machine could swing around and hit the secondary containment with enough force to breach it. A breach in the secondary containment would have caused a release of contamination into the operators’ working environment and pose a significant health risk. Despite extensive training and testing with the demolition machines, the risk and consequence of the containment breach was perceived to be too great for the demolition machines to be used and so decommissioning was completed manually Fig 1.
Systems need to be fail-safe. In the event of any failure, robotic systems must fail in predictable and manageable ways, unlike the planned remote demolition machines in the PFR fuel facility, to ensure safety during decommissioning. In addition to safety concerns, failures of systems will have high consequences in programme cost and schedule. This pressure is greatest in operational facilities where extending outages (periods where systems are offline for maintenance) can cost £millions. Sellafield Ltd. is currently running a research project called Reconfigurable Autonomy with the Universities of Liverpool, Sheffield and Surrey Space Centre that is looking to develop a robust architecture that is reconfigurable , i.e. it can change its setup and mode of operation if a fault occurs. This system would go beyond fail-safe solutions to allow graceful degradation and the completion of tasks even in a fault scenario. Current developments in the Reconfigurable Autonomy project use a robotic arm for autonomous sort and segregation of nuclear material. Should a joint fail in the robotic arm, the system will reconfigure and operate the arm without using the failed joint and operating with reduced degrees of freedom. Such a system or operational principle will be highly beneficial in the high consequence scenarios of nuclear decommissioning.
High radiation levels
Most legacy facilities have such high-radiation levels that manual operations are not possible and remote solutions are required. The environment means that robust solutions are needed but equipment is likely to fail. This poses a strategic question in the use of robotics: whether to develop robust, long-lasting solutions or use cheap, one-use solutions. The task at hand will ultimately determine the robotic characteristics.
Reuse and single use
Robust, long-lasting solutions will have a greater operational lifetime, which is essential in in situ operations such as retrievals and sequential decommissioning, where individual, neighbouring facilities are decommissioned one after another. Solutions can use radiation hardened electronics, resolvers (mechanical measurement devices) instead of encoders (electrical), or shielding to limit the effects of radiation. These types of solutions tend to be expensive and highly engineered to guarantee successful operations. Maintenance is crucial. Replacing whole systems is not commercially viable, so maintenance needs to be planned in the design stage. It is important to remember that any repairs will be completed in the highly radioactive environment, so remote solutions will maintain remote equipment. This leads to complex design solutions such as detachable cabling, single bolts or quick release connectors (Fig 1).
Fig 1: Manual decommissioning of a glovebox in the PFR fuel facility
The alternative methodology is to produce cheap, one-use solutions. The main benefit is the reduced commissioning and maintenance requirements but the short lifetime only permits completion of certain tasks. Sellafield Ltd. is actively engaged in the development of disposable robotic solutions. The Avexis project with the University of Manchester is developing a small-scale, underwater remotely operated vehicle (ROV) for inspection in the ponds, see Fig 2. The solution is 3D printed with a diameter of 150 mm and is expected to cost around £5000 per unit . The system is designed for ease of production, deployment and disposal. These single use systems are ideal for inspection and characterisation challenges but not for extended periods of decommissioning.
Fig 2: Avexis: a small-scale, underwater ROV
In addition to radiation, contamination can pose restrictions on robotic solutions. Contamination is radioactive particulate but the same analysis holds for the chemical hazards present in nuclear facilities. If solutions are to be reused, contamination must be managed. Often, protective covers are used to keep contamination off equipment and are disposed of after use. However, when doing physical operations, covers can restrict movement and pose extra risks of fire (for example) and cannot be used in the presence of certain chemicals. So, reusable solutions need to be cleanable: wires should be contained; crevices and pockets eliminated; and joints sealed. Clean robots are common in the medical industry and some of that design principle can be reused in contamination control.
Two bomb disposal robots were deployed in the First Generation Reprocessing Plant (FGRP) to dismantle process plant and remove equipment from a cell. The robots were operationally successful but were compromised by fine dust contamination. When the robots were washed to remove the fine dust contamination, water entered the robot casing. The water ingress led to a reduction in performance and meant contamination entered the casing. This increased the maintenance required, the downtime for the project and the overall costs.
Robots as waste
Ultimately, any equipment used in decommissioning will be disposed of as nuclear waste. The storage cost of nuclear waste is significant when considering the (potentially) thousands of years of storage. Any solution must take into account the lifetime cost of the equipment from design, manufacture, use, decommissioning and storage. One £50,000 device may have a lower lifetime cost than three £10,000 devices due to a lower storage volume. Or, one £10,000 device that will be stored as intermediate level waste may have a greater lifetime cost than three £20,000 devices that will be stored as low-level waste.
Shielding in nuclear facilities
Nuclear facilities are designed to contain the radiation and contamination generated during operations. Historically, they were not designed with decommissioning in mind. The thick concrete walls of most nuclear facilities contain radiation well but pose a number of issues for decommissioning.
The simplest but usually most pressing concern is how do you get your equipment into a facility to decommission it? Radiation fields require any openings to be small and have double, shielded doors to contain radiation. Gaining access can be achieved in two ways: creating a new access or using existing openings.
Creating a new opening into a facility is the method chosen for the Pile Fuel Cladding Silo (PFCS) retrievals at Sellafield Ltd. PFCS was commissioned in 1952 and has six silos, large cells for storing nuclear material. The silos were filled with nuclear fuel cladding and miscellaneous beta/gamma waste until they reached capacity in 1964. PFCS poses a significant risk and will be emptied of nuclear material before final decommissioning. To retrieve the material, new openings will be drilled into each of the six silos. Steel frame doors will be installed half-way up the facility in a sizeable metal frame, see Fig 3. The doors will maintain containment during drilling and retrievals. The building was not designed for and is not capable of taking the weight of the new doors, door frames and retrievals equipment. So, a separate retrievals building is being built beside the PFCS for the commissioning and operation of the retrievals systems. All of this will be completed while maintaining an argon atmosphere in the silos, to mitigate any fire hazard. Though this demonstrates one of the most complex access requirements at Sellafield site, the challenge of creating new openings should not be underestimated. Once access is obtained, the actual retrieval operations in PFCS are relatively simple and will be completed by grabbing the waste and transferring it into shielded containers in the new retrievals building.
Fig 3: Steel frame doors during installation for access into PFCS
The alternative is to use existing openings. In high hazard facilities, openings are restricted to inspection ports, penetrations that typically range from 10 to 30 cm in diameter. This severely limits the solutions available and requires small robots, thin robots (such as robotic arms) or innovative shape changing or self-assembling robots. Sellafield Ltd. is involved in the LaserSnake2 project with OC Robotics, The Welding Institute (TWI), the National Nuclear Laboratory (NNL), Laser Optical Engineering and ULO Optics. LaserSnake2 combines a self-supporting snake-arm robot from OC Robotics with a laser cutting system developed by TWI. The system can reach through small penetrations while retaining the flexibility and tool deployment necessary for nuclear decommissioning, see Fig 4. This innovative system is undergoing testing on Sellafield site in June 2016.
Fig 4: LaserSnake2: the system can reach through small penetrations while retaining the flexibility and tool deployment necessary for nuclear decommissioning
Communicating with your robotic device is vital to ensure operability, safety and successful operations. The obvious solution is to use wireless fidelity (Wi-Fi) but this has a few drawbacks in a nuclear environment. First, nuclear facilities are designed to contain radiation. This works for radiation coming out of the facility and communication signals going into the facility. So, any Wi-Fi or radio transmitter needs to be placed within the facility and is subject to the same radiation levels and hazards as the robotic solution. Second, the security of any transmitted signal must be guaranteed. The high consequence of the operations and sensitive information transmitted means signals must not be intercepted. Developments in encryption may negate this risk in the near future. Third, facilities are large, complex and congested. Wi-Fi signals will have a finite range and this is exacerbated with metre thick concrete walls, steel process equipment or lead bricks. These challenges are relatively simple to overcome but pose the debate for the deployment of technology both now and in the future.
Current operations favour the use of tethers. Tethers are wired links to a device and can provide benefits when compared with wireless operations. Tethers can provide robust communications through a permanent link, a power supply for longer operations, increased computing power and use of tools, and act as a fail-safe mechanism, literally acting as a tow to drag failed equipment out of cells. The difficulty with tethers is they severely limit the capability, manoeuvrability and reach of any system.
One potential answer to both the restricted movements of wired systems and the inconsistent communications of wireless systems is to introduce more autonomy to the robotic solution, i.e. allow the system to make decisions independent of an operator. This will reduce and potentially eliminate the need for communication to a central point. Autonomous systems are at the cutting edge of new research and so it is unlikely we will see anything like it decommissioning nuclear facilities in the next 10 years. However, work is currently underway and Sellafield Ltd. is involved in the Autonomous Intelligent Systems Partnership, a cross-sector programme with the Engineering and Physical Sciences Research Council funding a range of academic research projects. The autonomous projects look to enhance the capability of robotic solutions while reducing the burden on both communications and manual operations.
Characteristics of nuclear facilities
Legacy nuclear facilities were not designed with decommissioning in mind; they were designed for industrial operations. This means facilities are large, congested and complex and require flexible solutions.
Industrial facilities are large. To decommission large facilities you need large equipment, mobile solutions or innovative deployments. The tallest facility at the Sellafield site is the FGRP; it is ten storeys and has cells over 50.5 m tall, see Fig 5. The height of these cells and the large volume of other cells mean deployment of tools to the workface is not straightforward. Historically, decommissioning operations have made use of existing infrastructure such as cranes. For example, when decommissioning the Windscale Advanced Gas-cooled Reactor (WAGR), the reactor refuelling machine was adapted to remotely deploy tools in the reactor core. However, only certain facilities have useful infrastructure so other deployment methods such as telescopic booms have been used. These large-scale deployment methods can deploy tools within large cells but they lack flexibility so require rigorous design and testing.
Fig 5: The construction of the tallest facility at the Sellafield site, FGRP. The concrete structure is the nuclear cells and the steel frame is the building cladding and man-accessible areas
Sellafield Ltd. has recently deployed small unmanned aerial systems (UASs) for radiation mapping. The RISER system developed by Createc and Blue Bear Systems successfully mapped the radiation levels of the Pile Chimney, a 110 m tall ventilation shaft of the UK's first nuclear reactor, see Fig 6. UASs have the flexibility to investigate such large structures and are likely to become an important tool for inspection and characterisation surveys.
Fig 6: RISER system
Chemical processing plants have vast arrays of complex pipework, see Fig 7. ‘Pipe nests’ are common with labyrinths of pipework reaching a planned 22,000 m in Evaporator D  and 307,000 m in ThORP . The highly congested environments mean deployment of tools can be very difficult. As part of the South Dissolver Refurbishment Project in Magnox Reprocessing Plant, a pipeline was remotely diverted from the North Dissolver Cell to the South Dissolver Cell. The work, completed by engineers from BNFL’s R&T Department (British Nuclear Fuels Limited Research and Technology Department, now NNL) , cut and removed a section of pipework and welded a new section onto a different line using a manipulator system called Raffman. However, to gain access to the pipework to be diverted, redundant pipework had to be removed first using a separate system, Pipeman, see Fig 8. Though this project was not in decommissioning, the highly congested facility required a secondary system to allow the remote equipment to get access to the workface and this is likely to be repeated across decommissioning operations.
Fig 7: Congested facilities
Fig. 8: Pipeman
A number of legacy nuclear facilities are over 50 years old and the first nuclear facilities built, the windscale pile reactors, began construction in 1947 and were operational in 1952. The buildings were built before computers, computer-aided design and electronic record keeping. So design records, operational records and inventories are hand written and often incomplete. The high radioactivity and consequences of a radioactive release means that facilities may not have been accessed for years or even decades. The outcome of the aged records and lack of access is that the condition and contents of facilities may be unknown. The unknown forces decommissioning operations to plan against worst case scenarios to ensure safety is maintained and there are no significant radiological (or conventional safety) events. The uncertainties demand solutions that are flexible, adaptable and robust to eliminate the need for over-engineered solutions, overcome the pessimistic assumptions and allow successful operations in spite of loosely defined operational requirements.
As designed and as built facilities
Even when the records of legacy facilities are complete, there can be a significant difference between what is designed and what needs to be decommissioned. Any industrial structure over 50 years old will have evolved over time: there may be high rates of corrosion; weak points in the civil structures or facility furniture; or degraded and failed service equipment. This must be taken into account at the decommissioning planning stage, particularly when using heavy or destructive remote equipment as facilities have floor loading safety limits that become stricter over time.
The built facilities can differ from the actual design of facilities. Legacy facilities were first-of-a-kind across the world and their design evolved during construction and commissioning. This means slightly different equipment may have been installed, pipework can be located in different positions or entire additional systems may be installed from that on design drawings. Decommissioning solutions need the flexibility to allow for the changes in design and build. This prohibits the use of fixed mechanical deployments, preprogrammed routines and offline optimisation. Characterisation of facilities, i.e. the mapping of facilities’ geometric, radiological, chemotoxic, structural and other characteristics, can overcome this uncertainty and is a prominent phase in the decommissioning lifecycle, e.g. in Fig 9. Yet, characterisation operations need to overcome the same difficulties as decommissioning operations described in this article, effectively moving and not replacing the operational risks.
Fig 9: Characterisation of facilities. Here, radiological information is overlaid on a geometric model
Restricted facilities not designed for access
The high-radiation precludes manual access into facilities, as previously discussed. Therefore, facilities were not designed or built to be man accessible. There are few footpaths or corridors, limited flat floors and small gaps between process plant and vessels. Robotic solutions need to be able to traverse steps, move around obstacles and navigate through uneven terrain. Mechanical process cells are particularly restricted as narrow columns, troughs and chutes funnel material between dense mechanical equipments, e.g. in Fig 10. This sort of environment does not lend itself to large or ground-based solutions.
Fig 10: Restricted facilities not designed for access
This article thus far has focused on the deployment of equipment to the workface. The actual process of decommissioning deploys tools to complete a task.
Nature of the decommissioning task
For legacy facilities, decommissioning is the retrieval, processing and packaging of nuclear material in containers suitable for long-term storage. Decommissioning processing typically includes the size reduction of facility inventory, process equipment and civil structure and is an aggressive, physical task that moves, cuts, shears, drills, crimps, crushes or manipulates material into small and safe waste forms for waste packaging. These operations are physical, destructive and need to deal with the dense steels, thick concrete and miscellaneous waste found in nuclear facilities. The physical tasks need high powered machinery which has high power consumption. Robotic devices’ capabilities are often limited by battery life or power distribution. This is the case in the nuclear industry and leads to widespread adoption of tethered solutions and their limitations – as described in the ‘Communications’ section.
Tools for deployment
The actual tools deployed in decommissioning vary widely according to the task at hand. Characterisation will deploy a suite of sensors to understand the environment; decontamination will deploy scabbling techniques or chemical washes to clean the environment; dismantling will deploy cutting tools for size reduction; and waste management will deploy grippers for waste manipulation. The range of tasks requires a range of tools and the choice of tools is not simple. The Caesium Extraction Plant (CEP) was decommissioned using a remote decommissioning machine, see Fig 11. The machine had two robotic arms, was mounted on a telescopic boom and deployed tools at the workface. In the design process, every possible decommissioning operation was planned and a tool identified for its completion; this resulted in a large collection of eight commercially bought and 12 bespoke designed tools. When it came to the actual operations, only five of the simpler, more robust commercial tools were extensively used. There are two points of learning from this example.
Fig 11: Caesium extraction plant
First is whether equipment should be bespoke or commercial-off-the-shelf (COTS). COTS equipment is often preferred because the high number of successful working hours proves the robust operations and replacements can be easily bought. However, in a nuclear environment, COTS equipment can result in non-optimal or failed operations and bespoke solutions are often required to fulfil the operational requirements.
Second is the need for fit-for-purpose solutions. In highly complex operations, fit-for-purpose may mean a precisely engineered, adaptive tool that can account for dynamic environmental conditions. However, a lot of decommissioning operations process low-risk material where basic saws or shears are sufficient. As is clearly demonstrated in the CEP decommissioning, why use two tools when one will do?
While there has been great success in UK decommissioning, with Sellafield Ltd. decommissioning the first full scale nuclear reactor in the world, WAGR, and over 69 buildings decommissioned and demolished (at the time of writing ), there are still challenges in remote and robotic decommissioning to be overcome. The breadth of challenges and variety of environments in the one-of-a-kind legacy facilities dictate that there is not one single catch-all solution to remote decommissioning. Instead, a toolbox of fit-for-purpose solutions are required combining COTS and bespoke equipment and disposable and robust solutions. This will be achieved through continued engineering developments and a growing research and development programme to underpin the successful decommissioning of legacy facilities across the UK nuclear estate.
- IAEA Safety Standards: ‘Decommissioning of facilities. General safety requirements part 6’, 2014.
- The Comptroller and Auditor General: ‘Managing risk reduction at Sellafield’ (National Audit Office, 2012), p. 5.
- Sellafield Ltd.: ‘It’s not all black and white. Annual review 2013/14’, p. 2. Available at http://www.sellafieldsites.com/wp-content/uploads/2014/06/SEL9817-3_ARAC_Combined.pdf, accessed August 2016.
- Dennis L. Fisher M. Aitken J. M. et al.: ‘Reconfigurable autonomy’, Künstliche Intelligenz, 2014, 28, (3), pp. 199–207 (doi: 10.1007/s13218-014-0308-1).
- Nuclear Decommissioning Authority: ‘R&D research and development’ (NDA, 2015), p. 8.
- Sellafield Ltd.: ‘Sellafield magazine: issue 2’, p. 13. Available at http://www.sellafieldsites.com/publications/sellafield-magazine/issue2/#12, accessed August 2016.
- Battenbo H.: ‘Pipe dreams lessons from ThORP’, Process Eng., 1991, 72, (11), pp. 42–46.
- Kniazewycz G. Webster B.: ‘Two decades of robotics, remote handling and automation applications in nuclear fuel cycle facilities’. The American Nuclear Society, Sixth Topical Meeting on Robotics and Remote Systems, Monterey, CA, USA, February 1995, pp. 41–49.
- ‘Sellafield Ltd. Decommissioning Facts’. Available at http://www.sellafieldsites.com/press-office/facts/decommissioningrisk-and-hazard-reduction-facts/, accessed July 2016.