Agenda

A Holistic Approach to Engineering Training: Applying Tolerance Analysis, GD&T, and Dimensional Management

Authored & Presented By
Matthew Savage (EnginSoft UK Ltd)

In the ever-evolving landscape of engineering, manufacturing & CAE, the cornerstone of excellence consistently rests upon the robustness of processes and the proficiency of skills within teams to ensure precision, efficiency, and reliability. In an era marked by rapid innovation, complex project scopes and growing teams with hybrid working environments across the globe, traditional training's inadequacy to fill the skill gaps in engineering teams has become evident. Fostering a strong core foundation of engineering skill not only ensures that those quality needs are met by their products, but reduces business costs to contribute to sustainable success. Often courses are taught in isolation and don't utilise knowledge gained elsewhere in order to maximise the capability of their trainees. Segmented training courses result in a holistic understanding not being developed. If multiple training courses are led by the same provider, each training course can lead on from its predecessor, allowing case studies shared by the trainees to be explored in their entirety all interrelated requirements simultaneously. This is particularly significant with respect to mechanical variation, with three key pillars supporting a complete quality package. Tolerance Analysis, Geometric Dimensioning & Tolerancing (GD&T), and Dimensional Management must be applied to ensure that designs are cost effective and functional. Using GD&T definitions requires the impact of those tolerance definitions to be understood through 3D statistical tolerance analysis, and the implementation of both of those methodologies through Dimensional Management is key to cost and project workflows. This presentation explores the application of multi-disciplinary training with respect to those three disciplines, where engineers will learn the benefits of a holistic training program that is tailored to their specific needs. We will outline the process we used to develop material & enable this approach to facilitate the development of the next generation of simulation capability in engineers.

Full Name & Title: Mr Matthew Savage Project Engineer at EnginSoft UK Trainer of GD&T and 3D Tolerance Analysis Currently actively supporting clients with design precision for innovative products Experienced in applying GD&T and Tolerance Analysis practices in a range of industries including the medical and aerospace sectors. Supported a number of companies expanding their GD&T and tolerance analysis knowledge. Masters in Mechanical Engineering from University of Huddersfield

Industrial Applications of Isogeometric Analysis

Authored & Presented By
Michael Scott (Coreform LLC)

In contrast to the laborious and error-prone process of translating computer-aided design (CAD) into computer-aided engineering (CAE) models, isogeometric analysis (IGA) performs the finite element analysis (FEA) simulation directly on smooth CAD geometry, using smooth spline basis functions. The idea of IGA was first introduced over 12 years ago by Dr. Thomas J.R. Hughes, and is now among the most active academic fields of FEA, with thousands of papers published and multiple international conferences. IGA has generated academic interest because it consistently performs faster and more accurately than FEA and is more robust. These benefits are especially pronounced when running simulations where high accuracy in the shape description is of value, such as contact or fatigue simulations. IGA promises more accuracy, not less, while providing new opportunities for higher efficiency. The biggest obstacle to the commercial adoption of IGA has been the lack of a suitably general and robust analysis-suitable geometry description. Coreform has developed a proprietary new spline format U-splines, which is being integrated with Cubit's meshing technology for a general commercial solution. This presentation will introduce a native spline preprocessor, Coreform Flex, and a native spline FEA solver,Coreform IGA, in workflows currently used by commercial customers.

- Dr. Michael A. Scott is globally recognized as a leading authority in IGA. - He has been named a "Highly Cited Researcher" by Thomson Reuters yearly from 2016-2022. - He received his PhD from the University of Texas-Austin in 2011, and founded Coreform LLC in 2014. - Coreform has gone on to be recognized as the fastest-growing CAE company in the US every year from 2019 to 2023. - Through Coreform, he has developed the world's first commercial general-use IGA solver.

Wind Turbine Direct Drive PMG Machines and Airgap Prediction

Authored & Presented By
Xinmin Fan (GE Power Ltd)

In traditional wind turbine farm, gearing box maintenance and worn out is a problem. Direct Permanent Magnet Generators (PMG) where the turbine directly drives the generator without use gear box can solve this problem. However, to obtain high tangential rotor speed for electricity generation efficiently, the PMG normally has large diameter, for example up to 6 meters diameter or more for a 6 MW machine. With such large diameter, the designed air gap must be kept within a quite small value for generator efficiency. Due to the manufacturing and assembly error, there is initial unbalanced magnetic force. This magnetic force will cause generator deformation and reduce the airgap at some area. Reduced air gap will cause increased unbalanced magnetic force and more structure deformation. If the structure is not properly designed with sufficient stiffness, the airgap may disappear, and the rotor will clash with stator and damage the generator. The interaction between the magnet force and the mechanical deformation makes the stabilised airgap calculation a challenge. This paper presents a method for airgap calculation use FE for wind turbine PMGs.

Dr Xinmin Fan 1. Lead Mechanical Engineer, (Since 2008, GE Vernova UK, Previously GE) Electrical rotating machines static, dynamic and fatigue analysis 2. Project Engineer, (2004 to 2008, International Automotive Research Centre, Coventry) Joining simulation of Body in White on the PAG (Premium Automotive Program) 3. Principle Engineer, (1990 to 2000, ASN Tech Group, China) IC Engine design/analysis/test PhD: The University of Birmingham ------------------------------------------------11/00-11/03 School of manufacturing and mechanical engineering MSc: Xi’an Jiaotong University (China) --------------------------------------------09/87-06/90 Department of power machinery engineer BSc: Xi’an Jiaotong University (China) -------------------------------------------09/80-06/84 Department of power machinery engineering

The Epistemic Benchmark Challenges

Authored & Presented By
Jack Reijmers (NAFEMS Benelux Contacts)

The focus of the NAFEMS Stochastics Working Group is to champion and improve best practices that relate to stochastic engineering analysis and simulation methods and tools. It will promote the extension of current engineering analysis and simulation practices to include stochastic methods and tools to enable the virtual product development processes to be closer to the real-world behaviour of the modelled systems and components. Over the last decade the SWG challenged the engineering community with benchmark problems in the field of uncertainty, and the most recent set of problems encompasses epistemic uncertainty. This means that the uncertainty can be reduced by gathering as much data as possible, however, the participant in the challenge is provided with just a small data set. Remarkably, the solutions received show a large scatter in the required probability of failure, and it can be concluded that the different approaches of the problem do not end up in a comparable outcome. A Benchmark article in progress will elaborate on the variation in solutions. The solution given in this presentation is based on the application of a Student or t-distribution. The limited input data gives a bell-shaped distribution that resembles a normal distribution, however, with thicker tails. This approach gives probabilities of failure that are in the lower region of the solution space.

• Jack Reijmers is a retired naval architect. • He graduated from the Delft University of Technology in 1982 and started his career at Nevesbu B.V. in The Netherlands. • In the eighties his activities comprised naval projects with a special focus on submarines, such as the Walrus and Sea Dragon class. In the years ninety, activities shifted to offshore engineering, however, with strong involvement in submarine engineering (the Moray class). • In the beginning of this century, he worked for seven and a half years as a consultant for Navantia at the Cartagena Shipyard in Spain, assisting in submarine engineering. • In 2015 he joined the Stochastics Working Group, and at the World Congress 2019 in Quebec he was awarded the status of NAFEMS Technical Fellow. • Although retired he is still active within the Nevesbu company, not in the least by a PhD study at the Delft University of Technology on the theoretical basis of pressure hull analysis.

Statistical qualification of complex systems based on zero-failure evidence

Authored & Presented By
Andrew Halfpenny (HBK UK Ltd.)

Complex systems are susceptible to many failure modes. Conveniently, their reliable life is often governed by relatively few 'critical' modes. As it is uneconomic to perform separate reliability tests for every failure mode, designers typically concentrate on testing the critical modes. The question arises: how confident are we that those 'non-critical' modes are, indeed, non-critical? All tests contribute a degree of 'damage' to every failure mode. In exercising the critical modes, some tests may exaggerate the damaging effect on other non-critical modes. In this case, they exceed 100% of their expected operational damage. Other tests may under-achieve the damaging effect, so the test damage is less than 100%. The damage percentage may be calculated through simulation; therefore, combining simulation with physical tests allows us to calculate the confidence level on achieving a specified reliability target, based on aggregating zero-failure test evidence. The method is based on the Parametric Binomial method. The presentation starts by introducing FMEA (Failure Modes and Effects Analysis) and the RPN (Risk Priority Number). RPN is a risk based index derived through the subjective assessment of a team of engineers at the start of the design cycle. The presentation describes how a statistical confidence level is also calculated based on actual test evidence through the design process. Comparing the initial RPN with the calculated confidence level offers significant reassurance that failure modes initially designated non-critical are, indeed, non-critical. The presentation focuses on zero-failure test evidence. This is useful because zero-failures are common in non-critical failure modes, because test articles typically fail in the critical failure modes first. It also discusses how these methods can be extended to account for in-service data which offers considerable in-sight for future product designs.

• PhD in Mechanical Engineering and a Master’s Degree in Civil and Structural Engineering • Heads R&D activities for HBK’s nCode products • Over 25 years of experience in structural dynamics, vibration, fatigue and fracture • Introduced many new technologies to the industry including: FE-based vibration fatigue analysis, crack growth simulation and accelerated vibration testing • Holds a European patent for the ‘Damage monitoring tag’ and developed the new vibration standard used for qualifying UK military helicopters • Worked in consultancy with customers across the UK, Europe, Americas and the Far East • Many written publications on Fatigue, Digital Signal Processing and Structural Health Monitoring • Founding member of NAFEMS PSE Certification scheme and sits on the NAFEMS committee for Dynamic Testing

Probabilistic Machine Learning and Uncertainty Quantification to Guide Complex and Safety-critical Engineering

Authored & Presented By
Ross Allen (digiLab)

In the development of complex and safety-critical engineering systems, such as those in fusion energy, nuclear fission, and aerospace, accurate predictions of operational performance and lifespan degradation are essential. However, in the design stage the incorporation of novel materials and techniques introduces significant uncertainty into the expected operation. Research and development (R&D) efforts are increasingly focused on reducing this uncertainty to assure system reliability and to meet operational objectives even in the face of inherent system variances. This work explores the introduction of probabilistic machine learning to enable an understanding of the robustness of quantities of interest to uncertainties, and guide R&D pathways to sufficiently constrain uncertainty where required. This enables informed engineering decisions, regulatory approvals, and commercial endorsement based on the predicted performance. The application of 'design by analysis' for design assurance in safety critical environments is gaining prominence, by leveraging existing data, high-fidelity finite element analysis (FEA) and advanced computational techniques. These models predict material, component, and system responses in both normal and off-normal operation, able to efficiently probe rare and undesired events. Coupling FEA with robust probabilistic machine learning techniques such as Gaussian Process regression enables uncertainty quantification, whereby the distribution of all possible outcomes is determined. This enables assessment of the robustness of design points, operational modes, and understanding of which inputs are driving uncertainty, which in turn provides engineers an assessment of current risk, and guidance on an optimised route to reduce this uncertainty through the engineering programme.­­ This paper will explore digiLab's application of probabilistic machine learning and developments in systems level uncertainty quantification, with a particular focus on fusion energy, enabling the development of robust engineering design and determining optimised engineering testing programmes. ­­­

- Senior Solutions Engineer at digiLab, specialising in machine learning models for complex engineering, with a focus on fusion energy -Holds a PhD in Nuclear Physics, from the University of Birmingham, where he was focused on the intersection of expensive models and experiments, predicinting nuclear reaction probabilities. - Working in the fusion energy sector Ross has experience in fusion system design using system finite element simulation, working with both the public and private sector.

Probabilistic Design Approach in Fusion Reactor Development: A Case Study of High Heat Flux Component Design

Authored & Presented By
Sebastian Rosini (UK Atomic Energy Authority)

Future fusion reactor designs will heavily rely on numerical simulations for improvements in efficiency, safety, and reliability. Plasma-facing components (PFCs) in particular, must withstand harsh complex loading scenarios such as extreme thermal loads combined with repeated intense thermal shocks. While ITER and CHIMERA will act as test devices for PFCs in the coming years, the historical data currently available is limited and experiments focus on the most essential material properties. As a consequence, design of fusion systems must rely more heavily on simulation and as such historical processes of discrete deterministic simulations will not be sufficient. Integrated simulations of appropriate fidelity, adopting a probabilistic approach, are needed to provide a more rigorous quantification of reliability. This must be conducted using tools and processes which are flexible and easy to adopt and integrated within the engineering design process. In this study, a novel application of a comprehensive probabilistic approach to fusion reactor design is presented. The work outlines the essential parameters for executing probabilistic calculations, underscores the algorithms to adopt, and identifies potential avenues for further investigation. A preliminary case study, based on the model of a divertor high heat flux component design, highlights the significance of enhancing confidence in the uncertainty surrounding input parameters. The study emphasises the importance of adopting a probabilistic design approach over a deterministic one, marking a significant shift in fusion reactor design methodology and showcasing the transformative impact of this approach in other scientific and engineering fields.

Sebastian Rosini, an Engineering Analyst (Simulation Research) at the UK Atomic Energy Authority, specialises in developing new simulation capabilities. His current primary focus is on creating a probabilistic methodology to complement deterministic structural integrity assessments of in-vessel components. During his tenure at UKAEA, he has led the development of innovative multiphysics analysis workflows, but also inspection campaigns for additively manufactured components. Prior to joining UKAEA, Sebastian completed an industrial PhD at the muVIS X-ray Imaging Centre, University of Southampton, funded by Mitsubishi Corporation. His research focused on understanding the damage in carbon fibre composites. As a consultant at the centre, he conducted numerous damage investigations for aerospace applications, collaborating with industry leaders such as Boeing and Rolls-Royce.

Structural Reliability Analysis: Random Sampling or Machine Learning: it is only a matter of time (and accuracy).

Authored & Presented By
James Lee (Stochastics Ltd)

Solving multi variate equations is fundamental to all structural reliability analyses. For decades random sampling methods, such as the Monte Carlo method and its derivatives, have being used to solve probabilistic multi variate equations. Although random sampling methods are robust, they are not practical for many real-world applications as they can take too long to solve reliability scenarios to an acceptable level of accuracy. Stochastics Ltd. have developed a machine learning algorithm which substitutes the random sampling method in structural reliability analyses. This work compares the results and solve times of the random sampling method and the machine learning algorithm for several analytical and numerical (finite element method) structural reliability analyses. All examples are taken from published material. Where possible theoretical answers are given for comparative purposes. It will be shown that on a standard desktop computer, to a specified accuracy, the machine learning algorithm solves multi variate equations significantly faster than random sampling methods. The analytical examples compare the accuracy and solve times for a standard normal distribution, two structural reliability analyses and a non-linear material analysis. The numerical example presents a probabilistic based finite element incorporating the machine learning algorithm. This element is used in two finite element analyses which compare the solve times and accuracy of using the machine learning algorithm and random sampling method in finite element analyses. In conclusion this work shows that the machine learning algorithm is, based on solving time and accuracy, is a more efficient method than random sampling. It also shows that the machine learning algorithm can be explicitly coupled with the finite element method allowing rapid and accurate structural reliability analyses to be performed. Finally, this works shows that given the choice between random sampling and machine learning picking both methods, machine learning to do the bulk of the work and random sampling as an independent approximate verification tool, is the best approach.

BE in structural engineering from University of Galway, Ireland, MSc in bridge engineering and PhD in computational mechanics from the University of Surrey, UK. Member of the Institute of Structural Engineers and the NAFEMS Stochastics Working Group. Worked as a structural engineer with Hyder in Guildford and in sales and technical support with LUSAS finite element software in Kingston Upon Thames. Worked in the Oil and Gas industry, this is where the idea for coupled finite element structural reliability analysis software evolved. Currently divides time between distributing finite element and geotechnical software and developing structural reliability analysis software.

An Innovative Inquiry into Weld Durability Challenges in Heavy Commercial Vehicles: A Paradigm Shift Towards Robust Design through Strategic Material Optimization.

Authored & Presented By
Doğukan Elibol (Ford Otosan)

This study investigates a strength issue in the tandem structure of a heavy-duty commercial truck's rear axle. A crack emerging from the Heat Affected Zone (HAZ) in a failed rear axle housing prompted theoretical considerations and assumptions about the factors influencing strength in this region. Implementing vehicle architecture considerations, simulation environments incorporated specific boundary conditions. Engineering hand calculation were conducted to predict loads on the axle, and the model's behavior was examined in a Computer-Aided Engineering (CAE) environment. The model's response in simulation identified a hotspot from actual conditions where damage occurred. Post-static analyses included weld fatigue calculations, emphasizing the challenging nature of weld fatigue. Instrumentation with strain gauges, accelerometers, and displacement sensors was performed on the vehicle. Data collected from these sensors were used to transiently extract force data from six points, revealing a non-proportional loading nature. Dynamic analysis methods were then adopted. Simulating the durability track with these loading scenarios, strain data on housing from the actual durability parkour were compared with CAE-simulated strains, achieving a 70% correlation in both frequency and time domains. An accelerated loading scenario as well as renewed duty cycle based on strain data was produced. Resulting lifetime from fatigue analyses correlated well with real-world damage occurrences. This successful correlation validates the developed methodology. Due to time and cost considerations, the study's established boundary conditions and loading scenarios were transferred to a multiaxial test rig, instead of another vehicle parkour test which takes 2 months to complete and a lot of cost. In the multi-axial test rig that was designed particularly for this study, the rear axle underwent successful testing, accurately representing real-world damages. These studies instilled high confidence in the developed methodology. In improvement scenarios, a non-conventional approach involved reducing material instead of adding it, resulting in a successful example of engineering optimization?a lighter and more durable structure.

Dogukan Elibol : I am from Turkiye, representing Ford Otosan. I am a graduate from Istanbul Technical University from Mechanical Engineering. My field of interest is Solid Mechanics. In Ford Otosan, my role is Powertrain Durability CAE Engineer.

High-fidelity digital twin for Laser Metal Deposition by Wire

Presented By
Joanna Tumelty (CFMS Services Ltd)

Authored By
Joanna Tumelty (CFMS Services Ltd)
Will Alden
Alex Ballisat (CFMS Services Ltd)

Laser Metal Deposition by wire (LMDw) is an additive manufacturing process used to produce high-value products in the aerospace industry. The technique enables precision layer-by-layer deposition to produce parts with complex three-dimensional geometries. LMDw involves a range of multi-scale and multi-physics processes involving many process parameters that ultimately affect the quality of the printed part. Process optimisation is therefore essential to consistently produce high rates of high quality parts needed for industrialisation of this technology. Performing experimental trials to achieve this is time, cost and resource intensive; using simulations instead can greatly reduce the expense. To this end, a digital twin of the LMDw manufacturing process, integrating several high-fidelity simulations and modelling the end-to-end process, has been developed. The digital twin includes a robot model that uses toolpath information to determine the physical locations and time at which material is deposited. This information is fed into higher fidelity multi-physics simulations that capture the phase changes and material addition near the melt pool, and the post-print thermal distortion of the final part. This allows properties of custom printed parts to be determined. Using this digital twin facilitates process optimisation by enabling a much greater exploration of design spaces than would be achievable through experimental trials. This can be used to develop deeper understanding of process parameters, which enables faster and easier identification of those required to manufacture high quality parts.

Joanna Tumelty is a Research Engineer at The Centre for Modelling and Simulation (CFMS). She completed a BA and MMath in Mathematics at the University of Cambridge. Following this, she undertook a PhD in Applied Mathematics at the University of Leeds, where she investigated pattern formation in thermosolutal convection using numerical simulations. At CFMS, she has applied a range of computational methods to simulate multi-physics systems and developed system models used to simulate and optimise factory designs.

Bayesian Optimisation for Manufacturing and Engineering Design

Authored & Presented By
Greg Nelson (Frazer-Nash Consultancy Ltd)

Constrained Multi-Concept Bayesian Optimisation is a powerful tool that is well suited to design/process optimisation where individual design points are costly to evaluate; where there are many varying constraints; and where it is advantageous to keep multiple, and potentially significantly differing, concepts within the design loop. In this presentation we apply the method to three engineering problems: a demonstrator study into the design of Additively Manufactured divertor targets for nuclear fusion reactors, running FEA structural and thermal models in the loop, an optimisation of diffusion bonded pressure vessel windows, and tuning a laser cladding manufacturing process to meet all of the process requirements. This illustrates the breadth of problems that this method can be applied to, and highlights the advantages that can be gained in design and manufacturing through the use of the method.

Greg is a chartered Mechanical Engineer with a strong background in fusion, structural analysis and design. He has been involved with and led a number of concept design projects for the STEP project, including designs of liquid lithium divertor targets for additive manufacture, including design and multi-physics coupled analyses He has been developing Bayesian Optimisation techniques to support concept selection and optimisation in a fusion design context and wider, which have also been applied to the design of windows for optical measurements and to manufacturing methods.

Enabling high-rate manufacture of carbon fibre wings through simulation driven process optimisation

Presented By
Alex Ballisat (CFMS Services Ltd)

Authored By
Alex Ballisat (CFMS Services Ltd)
Leo Nolan (CFMS Services Ltd.)

The use of carbon fibre composites is growing in the aerospace industry due to their desirable strength and weight properties, however, manufacturing methods are less mature than those for traditional materials. In particular, the resin infusion process is under heavy development for producing components to be used in the next generation of aircraft wings. Complex physical phenomena, such as cure, racetracking, and void formation, make this a challenging process to design and operate consistently. To accelerate the development of the process, a custom resin infusion model was built using the finite element framework MOOSE, generating a tightly coupled simulation. Temperature dependent resin properties were captured in the cure kinetics and rheology models, and an advection stabilisation scheme was implemented to fix these properties relative to the fluid flow. Shell elements, combined with an analytical formula for the effective permeability of small channels, were used to model the racetracks. Simulation of void formation via mechanical air entrapment was also investigated, through a two-phase model at the macro scale and analysis of the capillary number at the micro scale. The model was validated against experimental infusions of aerospace wing components, reaching agreement within 1%, while also highlighting the sensitivity of the process to material property variation. It will be used in conjunction with parallel, high performance compute to explore the design space of process parameters. This enables sensitivity analysis to isolate crucial parameters and process optimisation to improve quality and repeatability. It can also be used to aid design of passive open-loop control systems, or integrated into an active closed-loop control system via a trained machine learning model. This would allow the process to be more finely tuned, ultimately resulting in higher quality, more consistent parts.

Optimizing complex lightweight structures through Simulation, Automation and Design Optimization

Presented By
Ruoyu Huang (Strathclyde University)

Authored By
Ruoyu Huang (Strathclyde University)
Maximilien Landrain (Noesis Solutions)

Material, structural, and process optimizations are the key challenges for lightweighting. The Lightweight Manufacturing Centre (LMC) is a specialist technology centre within the National Manufacturing Institute Scotland (NMIS) at the University of Strathclyde, Glasgow. LMC is collaborating with Noesis Solutions (NOESIS) to develop a model-based optimization capability for designing and manufacturing lightweight structures in Scotland. NMIS has extensive experience in material and process simulations for industry, and with the support of NOESIS and Optimus, an automation and optimization tool, different pieces of finite element software can be integrated into one single optimization platform, which fills the capability gap and expands the modelling capability in LMC. A shell-like lightweight structure can be made of a metallic liner and a non-metallic composite cover. Optimizing such a structure involves material optimization of the composite material using micromechanical modelling and structural optimization using FEA. The integration with Optimus automates the process, explores the design space, and finds an optimal solution, leading to substantial time and resource savings compared to the previous manual iteration method. This paper demonstrates how the collaboration between LMC/NMIS and NOESIS contributes to help industries effectively design and manufacture lightweight components/structures using advanced simulation, automation, and optimization tools. The collaboration addresses complex design challenges by blending various multiphysics processes and multidisciplinary simulations of FEA and other non-FEA software. The results promote cost and resource efficiency, which can benefit the industry in the long run.

Analytical Multi-physics of Electric Motor Vibroacoustics

Presented By
Panagiotis Andreou (Loughborough University)

Authored By
Panagiotis Andreou (Loughborough University)
Stephanos Theodossiades
Dr Marcos Ricardo Souza
Dr Amal Z. Hajjaj
Dr Mahdi Mohammadpour (Loughborough University)

In the realm of vehicle development, electric motors have become indispensable components due to the ongoing electrification of the automotive industry. Understanding the vibrational behaviour and noise characteristics of these motors is now paramount to ensuring optimal performance and passenger comfort. Hence, Noise, Vibration, and Harshness (NVH) analysis and optimisation have become a critical aspect of motor design. Such an analysis demands a comprehensive multi-physics methodology, capable of accurately computing the high-frequency electromagnetic excitations under diverse operating conditions, as well as the interaction of these excitations with the constrained structure that ultimately results in NVH generation. Traditionally, vibroacoustic predictions have relied on Finite Element Analysis (FEA) simulation tools. While powerful, these tools often impose formidable computational demands, especially when considering a wide range of operating conditions and frequencies. Consequently, more recent commercial numerical software packages have begun integrating alternative reduced-order sub-routines. This allows users to perform rapid computations, albeit with a compromise in prediction quality due to necessary assumptions and simplifications. Following this trajectory, this study introduces an innovative approach to simulating noise in electric motors, employing a reduced order 3D cylindrical model to predict the stator's response under electromagnetic excitation. Representation of the stator's complex mechanical properties is accommodated through bi-directional material and dimensional variations, allowing for accurate prediction of the system's dynamics. The electromagnetic excitations are obtained from a semi-analytical electromagnetic model, and the steady-state vibration response of the system is evaluated analytically. The proposed methodology significantly improves computational efficiency when compared with numerical techniques, expediting the early design phase, where rapid iterations and trade-off assessments are crucial. The results demonstrate a practical solution that balances computational efficiency with accurate predictions for improved decision-making during the development cycle, offering tangible benefits for industry professionals engaged in motor design and optimization.

Panagiotis Andreou received his master’s in engineering (Meng) degree in Mechanical Engineering from Loughborough University, United Kingdom, in 2021, where he is currently pursuing the Ph.D. degree, on Noise, Vibration and Harshness (NVH) prediction and characterisation in electric powertrains. Throughout his academic career, he has been involved in a number of projects with industrial partners, developing analytical models and assessing the tribo-dynamic behaviour of coated valvetrains, as well as performing experimental fault diagnosis in high performance powertrains. His research interests include the development of reduced order methodologies for fast prediction of electromagnetic and mechanical noise generation by electric powertrains.

Exploring the Future of Electromagnetic Certification in Consumer Electronics Through Simulation

Presented By
Richard Chippendale (Element Digital Engineering)

Authored By
Richard Chippendale (Element Digital Engineering)
William Caton (Element Digital Engineering)

In today's world, the surge of portable electronic devices has led to an innovative trend: the convenience of high-power wireless charging. This method depends on the effective interaction of electromagnetic fields between a transmitter and receiver. However, these fields extend beyond the devices and into the environment. To safeguard consumer electronics, the industry adheres to specific emission limits, measured by the Specific Absorption Rate (SAR). This critical standard applies not only to wireless charging but also to the communication antennas found in IoT devices, ensuring their safety and compliance. The certification process for these devices traditionally involves extensive physical testing to demonstrate adherence to SAR standards, simulating real-world usage. At Element, our laboratories boast over 15 years of expertise in conducting such essential testing and certification. A significant advancement in this field is the introduction of the IEC standard 62704-4, which recognizes the role of digital simulation in the SAR certification process. This innovation promises a faster, cost-effective approach, allowing for early integration in the product design lifecycle. Our paper will illustrate how Element is pioneering the combination of modelling and physical testing. This dual approach not only streamlines certification but also enhances device optimization, marking a new era in the electromagnetic certification of consumer electronics. Join us as we delve into the transformative impact of simulation on the future of consumer electronics certification, showcasing our journey towards integrating cutting-edge technology and traditional testing to set new standards in device safety and efficiency.

An Advanced Data Science & Optimisation Based Approach for Wastewater Network Planning

Authored & Presented By
Ben Simms (EnginSoft UK Ltd)

The UK water industry faces increasing pressure & regulatory demands to improve wastewater network performance. In addition to managing flood risk with changing rainfall patterns there are new mandates to cut Combined Sewer Overflow (CSO) spills into rivers & seas. In collaboration with water companies, data science and optimisation techniques have been applied to inform "Price Review"5 year investment plans for the most complex catchments. A multi-objective optimisation based on 1D sewer network models first generates a dataset of thousands of optimal upgrade plans balancing performance against cost. The complex nature of these models necessitates innovative methods. Firstly transforming long CSO spill analysis simulations into short 'Extract Events', cutting runtime from days to minutes. Secondly aggregating tens of thousands of individual upgrades to a manageable number (200-300) of optimization parameters. This pushes the bounds of even best-in-class algorithms' scalability, yet convergence is achieved within a realistic timeframe. A bespoke, web-based data analysis platform is key to this approach, enabling all stakeholders to gain understanding from the dataset. Pareto charts demonstrate cost-performance trade-offs and include filters to enable live 'what-if' scenario testing, instantly finding next-best solutions when buildability constraints are discovered late on. Heatmaps prove simple to calculate but invaluable to visualise designs geographically and find low-risk "No Regrets"upgrades. Finally, summary efficiency metrics are readily determined, to be applied to estimating requirements on a business level without studying every catchment. This data-led approach has provided far greater detail and accuracy to investment planning, reducing business risk. Additionally this gave a head start to more detailed design, allowing engineers to cope with rapidly increasing workloads. Though developed around this specific need, the concepts and reporting tools are equally applicable to other industries.

Simulation and physical validation of triply periodic minimal surfaces-based scaffolds for biomedical applications

Presented By
Mudassar Khalil (Loughborough University)

Authored By
Mudassar Khalil (Loughborough University)
Matt Burton
Paul Conway
Carmen Torres-Sanchez (Loughborough University)

Metallic scaffolds are considered in the biomedical field due to their ability to mimic bone structures. Sheet-based Triply Periodic Minimal Surfaces (TPMS) structures are good candidates for this application because of their high surface-to-volume ratio, tailorable stiffness, and manufacturability by Additive Manufacturing. However, there are well-known discrepancies between the as-designed and as-manufactured physical and mechanical properties due to manufacturing limitations of the Selective Laser Melting (SLM) process. Many studies have reported and quantified these gaps. Porosity and pore size are important features from a Bioengineering viewpoint, but what many studies fail to consider in addition to those is surface area per unit volume (S/V), which is a crucial parameter for guiding cellular behaviour. This study focuses on three types of SLM TPMS scaffolds made using commercially pure Titanium (cp-Ti) powder and addresses the discrepancies in the as-designed/as-manufactured structures, as well as considers S/V as an important design specification. The as-designed scaffolds had 70-75% porosity and 25-35cm-1 S/V, achieved by varying the unit cell size and sheet thickness. Their stiffness was evaluated using finite element analysis (FEA) and resulted in 6.7-9.3 GPa. The as-manufactured scaffolds featured a 59-70% porosity and 35-42 cm-1 S/V. Their stiffness was 8.6-15.5 GPa, evaluated by a meshless method. Finally, the manufactured scaffolds were tested in the lab via quasi-static compression testing, yielding an effective Youngs modulus in the range of 5-9 GPa and yield strength 66-92 MPa. These values fall within range of trabecular bones. The difference between the as-designed and the as-manufactured moduli was 12% in average; and the difference between as-manufactured simulations and lab experiments was 57% average. It is hypothesised that the differences stem from the secondary roughness on the scaffolds walls deposited during to the manufacturing process. It was found that unit cell size correlates negatively with deviations between as-designed and as-manufactured specimens in terms of physical properties. The cyber-physical validation methods presented in this study are useful to quantify these deviations as the information obtained is then employed to feedback to the design stages and obtain more predictable as-manufactured structures.

Mudassar Khalil is a Doctoral Researcher at Loughborough University, specializing in the design, manufacture, and validation of porous structures. In addition to his doctoral research, he serves as a Research Associate, focusing on the post-processing of 3D-printed metallic scaffolds to reduce surface roughness. Mudassar holds an MSc in Polymer Technology from Aalen University, Germany, and a BSc in Mechanical Engineering from UET Lahore, Pakistan. He has over 13 years of experience in academia and 5 years in manufacturing industries across the Middle East and Pakistan.

Parametric Models of the Human Knee for Efficient Workflows

Authored & Presented By
Laurence Marks (Laurence Marks Consultant)

Focusing on modelling strategies to improve turnaround times and productivity is arguably more important now than ever, even with the widespread availability of multicore solutions and cloud computing. In the field of biomechanics the use of segmented models created automatically from medical images has become the defacto standard for model generation, however in some circumstances the use of CAD models driven by medical images offers a significant advantage in terms of model creation times, the ability to explore the "designspace" and critically the time taken to achieve a converged solution. In this presentation we will look at some of the modelling challenges that the human knee presents. These cover material response, geometric detail and resolution, function, and how medical devices are designed and operate within this environment. Working with uncertainty is a given, so modelling strategies have to offer a means of investigating this. The design of a uni-compartmental knee replacement provides a useful case study when looking at optimal simulation strategies, and what can be achieved, especially with at the application of structural optimisation techniques, and how they can be used to directly improve patient outcomes.

AI-enabled engineering

Authored & Presented By
Nico Haag (PhysicsX)

Context and challenge: Our client, a MedTech start-up asked us to optimize a novel artificial heart. They were planning the first human trials in 2024 and wanted to see if we can optimize the pump and the housing to reduce the energy consumption while catering for all the constraints. This was a challenging problem as there were numerous constraints that had to be met. The problem required highly complex simulation that needed to address various blood damage mechanisms (e.g., shear stresses, complex flows, heating) in a transient setup. Existing simulation took long time (48-96h) while producing inaccurate results thereby, limiting the optimization. Results: We used fully automated and state-of-the-art CFD as well as advanced AI technology to simulate and optimize the design of the heart pump and the housing and improved shutoff pressure by 42% and efficiency 15%. In fact, as a result of our optimization the wire that was connecting the heart from inside the body to the battery was removed and the heart could operate wirelessly significantly improving the quality of life for the patients. The approach: We used an AI-based optimisation workflow in conjunction with our Large Physics Model (LPM) trained on simulation data and achieve 98% accuracy with a single GPU over 2 days. Next, using geometry generation engine (parametric CAD + mesh morpher) and the optimization engine we built an e2e loop comprising of: • Generate novel and practical geometries, • Assess performance of the geometry (both scalars and flow-fields) and • optimize the geometry Due to the efficiency of the geometry generation, LPM, and optimizer each geometry is generated, assessed, and optimized in less than 1 second. We let our e2e loop run for 5 days, generating ~ 1,000,000 geometries on a single A100 GPU leading us to the global optimum geometry in a high dimensional design space.

Nico leads the simulation engineering and data science teams at PhysicsX, delivering tailored software, methodologies and designs for some of the world’s most challenging engineering applications. Leveraging expertise across both computer-aided engineering and geometric deep learning, Nico orchestrates our cross-functional delivery, research, and products teams while shaping our pioneering PhysicsX's AI-driven platform, fundamentally reinventing the way simulation engineering is done today. Prior to co-founding PhysicsX, Nico held engineering roles at Mercedes-Benz, Audi Sport and Bentley Motors.