This fellowship programme will apply state-of-the-art 3D image processing and machine learning methods, developing them further where necessary, to deliver a new software tool that performs industrial production line 'virtual qualification' using part-specific simulations from 3D X-ray imaging in high-value manufacturing (HVM). Qualification is when manufactured parts are verified fit for purpose, often achieved by performing experimental tests representative of in-service conditions. Virtual qualification will verify by modelling micro-accurate digital replicas of the final part (flaws included) replacing costly and time-consuming experimental methods. Additionally, this will assess defects for performance impact (rather than expensive but unspecific pass/fail testing). The challenge is that image-based modelling currently requires significant human interaction over a timescale of weeks. Applying this to many parts takes significant time to complete unless methodology can be changed. The novelty of this proposal is to use machine learning with foreknowledge, due to production line parts being similar, to automate conversion of microresolution 3D images into part-specific models that simulate in-service conditions. This automation is required for the technique to scale for deployment in industrial manufacturing. Additionally, because much of the decision making entailed is subjective, and therefore prone to human error, a consequential benefit of automation is consistent outputs by removing this variability. This proposal focuses on image-based finite element methods (IBFEM), which merge real and virtual worlds to account for deviations caused by manufacturing processes not considered by design-based finite element methods (FEM), e.g. due to tolerancing or micro-defects. This implementation of part-specific modelling has applications in advanced manufacturing wherever there is variability from one component to another e.g. additive manufacturing or composites. A case study will be undertaken with the UK Atomic Energy Authority (UKAEA) for a heat exchange component. This will showcase the capabilities of the technique to automatically produce a report that estimates the impact of deviations from design on performance. Unlike FEM, which have undergone extensive certification and are industry-wide trusted methods, there has not been a systematic approach which can be used to benchmark image-based modelling workflows against verified experimental data. This work will produce benchmarks based on standards for experimental measurements of thermomechanical material properties to give confidence in the technique for industrial adoption. The database of benchmarks will be useful for those wishing to use image-based modelling to validate workflows and could contribute towards establishing new standards in the field. Central to this proposal is the use of FEM, the de-facto tool for predicting thermomechanical performance in engineering. Prof Zienkiewicz's research at Swansea University established it as a birthplace for FEM, and is now recognised as a leading research centre in the field. The team undertaking this fellowship, led by Dr Llion Evans, will be based at the Zienkiewicz Centre for Computational Engineering, Swansea University and will work in collaboration with the centre's head, Prof Nithiarasu, an expert in image-based modelling for biomechanics. Access to the equipment required for all aspects of this highly multidisciplinary work i.e. thermomechanical characterisation, 3D imaging and computing is available through complementary centres at the College of Engineering, Swansea University. To support this extremely multidisciplinary work, key industrial organisations will be collaborating on this project. Nikon Metrology Ltd. (X-ray imaging systems), Synopsys Inc. (image processing software), TWI (non-destructive testing and industrial standards), UKAEA (energy generation end-user) and Airbus (aerospace end-user).

Robots with legs and arms are likely replace most manual labour, especially in environments that are dangerous for humans, and revolutionize multiple services domains in the long-term. One of the main advantages of legged robots is that they can discretely make and break contact with the environment, in contrast to wheeled or tracked systems that require continuous contact with the ground. This way, robots with legs can modify their area of support from step to step, a requirement when negotiating challenging terrain and environments primarily built for humans. Also, the use of legs decouples the body from the robot's foot-print. This allows for wide areas of support with only small footprints, a major advantage when navigating passages, tight spaces, cluttered environments, etc. The high articulation of legged systems also allows them to manipulate their center of mass, so that the system's dynamics can be exploited for the task at hand, and to dynamically reconfigure their workspace for the benefit of their payload, i.e., increase a manipulator arm's reach or position a sensor suite in a preferred pose. The autonomous locomotion framework that we will develop will enable current technology to be used in industrial scenarios, especially in hazardous environments that are primarily built for humans. Examples of such places are nuclear power plants, factories, oil & gas facilities, etc., where typically industrial stairs are used and a system will need to overcome various terrain difficulties, such as step over pipes, gaps, climb up/down stairs, manoeuvre through narrow passageways. Legged systems in such settings can have a large variety of roles; starting from inspection, automated monitoring of the condition of a facility; maintenance, periodic recurring tasks that need to be performed typically by a human, to intervention when an anomaly is detected.

Uncertainty quantification, verification and validation are crucial to establish the reliability and reproducibility of all forms of computer-based simulation. We propose to establish an open source and open development VVUQ toolkit optimised for efficient execution at current pre- and emerging exascale, which will raise new challenges and new opportunities for simulations in fields as diverse as fusion and climate modelling. Computer simulation results are validated compared with experiment in several ways, ranging from qualitative to quantitative measures which apply a validation metric. Likewise, verification is concerned with confirmation that the mathematical model and corresponding algorithm have been coded correctly. Uncertainty quantification (UQ) is concerned with understanding the origins of and assessing the magnitudes of the errors which accompany computer simulations, whether epistemic or aleatoric. VVUQ is necessary for any simulation that makes predictions in advance of an event to become actionable - that is, for its output to be useful in any form of decision-making process, from government interventions in pandemics to the choice of materials to combine for aircraft wing production. Here, exascale computing offers more opportunities to make actionable predictions. Moreover, because VVUQ is intrinsically compute intensive due to its ensemble-based execution pattern, it too requires exascale resources, as well as advanced resource management strategies to efficiently manage the large numbers of concurrent runs necessary. We propose to establish an open source and open development VVUQ toolkit optimised for efficient execution at current pre- and emerging exascale. This will include advanced approaches for surrogate modelling in order to minimise the expense and time needed to perform the most compute-intensive calculations and will demonstrate its efficiency gains for a diverse array of VVUQ workflows within multiple scientific applications, and on architecturally and geographically diverse emerging exascale environments. The software developed, implemented and benchmarked in this project will become an open and invaluable asset to the UK ExCALIBUR community but also much more widely within UK and internationally as high-performance computing enters the exascale era. The proposed exascale toolkit will be built on a combination of widely used tools and services which will be evolved to handle systems of increasing levels of complexity. These include components from the VECMA project (EasyVVUQ, FabSim3, QCG-PJ and EasySurrogate), as well as the UCL-Alan Turing Institute Multi-Output Gaussian Process Emulator (MOGP). We will apply these capabilities to several applications, including: (i) the UKAEA's tokamak fusion modelling use case for which a working software environment will be produced; (ii) weather and climate forecasting for the Met Office; (iii) turbulent flow simulation for environmental science; (iv) prediction of advanced materials properties of graphene-polymer based nanocomposites for aerospace applications; (v) high-fidelity patient-specific virtual human blood flow system for medical research; (vi) drug discovery; and (vii) human migration.

The vision for this Programme is to deliver the step changes in Robotics and Autonomous Systems (RAS) capability that are necessary to overcome crucial challenges facing the nuclear industry in the coming decades. The RAS challenges faced in the nuclear industry are extremely demanding and complex. Many nuclear installations, particularly the legacy facilities, present highly unstructured and uncertain environments. Additionally, these "high consequence" environments may contain radiological, chemical, thermal and other hazards. To minimise risks of contamination and radiological shine paths, many nuclear facilities have very small access ports (150 mm - 250 mm diameter), which prevent large robotic systems being deployed. Smaller robots have inherent limitations with power, sensing, communications and processing power, which remain unsolved. Thick concrete walls mean that communication bandwidths may be severely limited, necessitating increased levels of autonomy. Grasping and manipulation challenges, and the associated computer vision and perception challenges are profound; a huge variety of legacy waste materials must be sorted, segregated, and often also disrupted (cut or sheared). Some materials, such as plastic sheeting, contaminated suits/gloves/respirators, ropes, chains can be deformed and often present as chaotic self-occluding piles. Even known rigid objects (e.g. fuel rod casings) may present as partially visible or fragmented. Trivial tasks are complicated by the fact that the material properties of the waste, the dose rates and the layout of the facility within which the waste is stored may all be uncertain. It is therefore vital that any robotic solution be capable of robustly responding to uncertainties. The problems are compounded further by contamination risks, which typically mean that once deployed, human interaction with the robot will be limited at best, autonomy and fault tolerance are therefore important. The need for RAS in the nuclear industry is spread across the entire fuel cycle: reactor operations; new build reactors; decommissioning and waste storage and this Programme will address generic problems across all these areas. It is anticipated that the research will have a significant impact on many other areas of robotics: space, sub-sea, mining, bomb-disposal and health care, for example and cross sector initiatives will be pursued to ensure that there is a two-way transfer of knowledge and technology between these sectors, which have many challenges in common with the nuclear industry. The work will build on the robotics and nuclear engineering expertise available within the three academic organisations, who are each involved in cutting-edge, internationally leading research in relevant areas. This expertise will be complemented by the industrial and technology transfer experience and expertise of the National Nuclear Laboratory who have a proven track record of successfully delivering innovation in to the nuclear industry. The partners in the Programme will work jointly to develop new RAS related technologies (hardware and software), with delivery of nuclear focused demonstrators that will illustrate the successful outcomes of the Programme. Thus we will provide the nuclear supply chain and end-users with the confidence to apply RAS in the nuclear sector. To develop RAS technology that is suitable for the nuclear industry, it is essential that the partners work closely with the nuclear supply chain. To achieve this, the Programme will be based in west Cumbria, the centre of much of the UK's nuclear industry. Working with researchers at the home campuses of the academic institutions, the Programme will create a clear pipeline that propels early stage research from TRL 1 through to industrially relevant technology at TRL 3/4. Utilising the established mechanisms already available in west Cumbria, this technology can then be taken through to TRL 9 and commercial deployment.

In highly engineered materials, microscale defects can determine failure modes at the compo-nent/system scale. While X-ray CT is unique in being able to image, find, and follow defects non-destructively at the microscale, currently it can only do so for mm sized samples. This currently presents a significant limitation for manufacturing design and safe life prediction where the nature and location of the defects are a direct consequence of the manufacturing process. For example, in additive manufacturing, the defects made when manufacturing a test-piece may be quite different from those in a three dimensionally complex additively manufactured engineering component. Similarly, for composite materials, small-scale samples are commonly not large enough to properly represent all the hierarchical scales that control structural behaviour. This collaboration between the European Research Radiation Facility (ESRF) and the National Research Facility for laboratory CT (NRF) will lead to a million-fold increase in the volume of material that can be X-ray imaged at micrometre resolution through the development and exploitation of a new beamline (BM18). Further, this unparalleled resolution for X-rays at energies up to 400keV enables high Z materials to be probed as well as complex environmental stages. This represents a paradigm shift allowing us to move from defects in sub-scale test-pieces, to those in manufactured components and devices. This will be complemented by a better understanding of how such defects are introduced during manufacture and assembly. It will also allow us to scout and zoom manufactured structures to identify the broader defect distribution and then to follow the evolution of specific defects in a time-lapse manner as a function of mechanical or environmental loads, to learn how they lead to rapid failure in service. This will help to steer the design of smarter manufacturing processes tailored to the individual part geometry/architecture and help to establish a digital twin of additive and composite manufacturing processes. Secondly, we will exploit high frame rate imaging on ID19 exploiting the increased flux available due to the new ESRF-extremely bright source upgrade to study the mechanisms by which defects are introduced during additive manufacture and how defects can lead to very rapid failures, such as thermal runaway in batteries In this project, we will specifically focus on additive manufacturing, composite materials manufacturing and battery manufacturing and the in situ and operando performance and degradation of such manufactured articles, with the capabilities being disseminated and made more widely available to UK academics and industry through the NRF. The collaboration will also lead to the development of new data handling and analysis processes able to handle the very significant uplift in data that will be obtained and will lead to multiple site collaboration on experiments in real-time. This will enable us to work together as a multisite team on projects thereby involving less travelling and off-setting some of the constraints on demanding experiments posed by COVID-19.