Content Preface 5 Profile 6 Facts and Figures 7 Equal­Opportunities­and­Diversity 8 SCAIlights 9 Selected Software Solutions 12 Bioinformatics 14 Optimization 16 Multiphysics 18 Fast­Solvers 20 High­Performance­Computing 22 Network­Evaluation­Technologies 24 Virtual­Material­Design 26 Numerical­Data-Driven­Prediction 28 Meshfree Multiscale Methods 30 Computational­Finance 32 The­Fraunhofer-Gesellschaft 34 Academic­Partnerships 35 Online Sections 36 Editorial Notes 37 4

Advisory Board meeting in May 2022: Dr. Ulrich Leiner, Dr. Jan Hamaekers, Dr. Richard Trethewey, Dr. Anton Schüller, Dr. Thomas Soddemann, Klaus Wolf, Prof. Dr. Babette Dellen, Victoria Appelbe, Prof. Dr. Martin Hofmann-Apitius, Prof. Dr. Petra Mutzel, Dr. Johannes Landes, Dr. Daniel Oeltz, André Gemünd, Dr. Bram Metsch, Prof. Dr. Hans-Joachim Bungartz, Prof. Dr. Jochen Garcke, Prof. Dr. Michael Griebel, Christiane Stoll, Prof. Dr. Marc Alexander Schweitzer, Prof. Dr. Reinhard Klein (from left) Profile Director Prof. Dr. Michael Griebel Prof. Dr. Marc Alexander Schweitzer (Deputy) Branch Lab Bonn Dr. Jan Hamaekers Advisory­Board Dr. Bernhard Thomas, chairman, im.pulse, Köln Victoria Appelbe, Stadt Bonn Prof. Dr. Babette Dellen, Hochschule Koblenz, RheinAhrCampus Remagen, Dr. Bernd Eck, BASF SE, Ludwigshafen am Rhein Prof. Dr. Reinhard Klein, Universität Bonn Prof. Dr. Petra Mutzel, Universität Bonn Prof. Dr. Petra Ritter, Charité – Universitätsmedizin, Berlin Spin-offs adiutaByte GmbH, Sankt Augustin BioSolveIT GmbH, Sankt Augustin Causality Biomodels Ltd., Kalamassery, Cochin, Indien scapos AG, Sankt Augustin SIDACT GmbH, Sankt Augustin Business Areas Bioinformatics Optimization Multiphysics Fast Solvers High Performance Computing Network Evaluation Technologies Virtual Material Design Numerical Data-Driven Prediction Meshfree Multiscale Methods Computational Finance Prof. Dr. Martin Hofmann -Apitius Dr. Ralf Heckmann Klaus Wolf Dr. Hans-Joachim Plum Dr. Thomas Soddemann Dr. Mehrnaz Anvari Dr. Jan Hamaekers Prof. Dr. Jochen Garcke Prof. Dr. Marc Alexander Schweitzer Dr. Daniel Oeltz Central­Services Planning and Controlling Marketing and Communications IT-Services Christiane Stoll Michael Krapp André Gemünd 6

Equal Opportunities and Diversity Diversity­is­an­essential­premise­for­the­self-image­ and­prospects­of­Fraunhofer­SCAI Fraunhofer SCAI highly values diversity in its workforce. Mixed teams are more creative and better at problem- solving. Diversity management aims to create a working environment where all employees can participate equally. Equal opportunities for all employees regardless of ethnic origin, religion, world view, gender, sexual identity, disability, or age is an essential premise for the self-image and prospects of SCAI. The equal rights officer, Dr. Vanessa Lage- Rupprecht, elected by the employees, supports the institute’s management in pursuing these goals. Meeting the demands of childcare and work can be challenging. Fraunhofer SCAI offers assistance for families. The­TALENTA­development­program­is­a­career­ booster for women in science and management Fraunhofer SCAI supports and takes advantage of the opportunities offered by the Fraunhofer-internal support and development program TALENTA for female scientists. TALENTA focuses on the individual career development of female participants and qualifies them for leadership positions. Dr. Daniela Steffes-lai is promoted in TALENTA speed-up, a program tailored to accelerate the careers of professionally experienced female scientists. Sophia Krix participates in TALENTA start, which offers advice and specific qualification opportunities to women starting their scientific careers. 8 Dr. Vanessa Lage-Rupprecht, Bioinformatics department, is the institute‘s equal rights officer. Flexible­working­hours­and­comprehensive­support­ services­help­to­balance­work­and­private­life The institute promotes balancing work, family, and private life through flexible working hours. In addition to emergency childcare, homecare / eldercare, and life coaching services provided by pme Familienservice, Fraunhofer offers various information and training courses. Other services at the Fraunhofer Institute Center Schloss Birlinghoven include all-day holiday care during the summer school vacations and play and childcare materials provided for children brought into the office. Workshops and courses on health support and disease prevention complete the offerings. Dr. Daniela Steffes-lai (l.), Numerical Data-Driven Prediction department, and Sophia Krix, Bioinformatics department

SCAIlights Innovative­process­simulation­for­the­food­ ­industry­with­the­software­MESHFREE April­2022 Whether cleaning, stirring, kneading, or injection processes – MESHFREE uses a grid-free approach to reliably simulate food technology processes. The Fraunhofer Institutes for Industrial Mathematics ITWM and SCAI have demonstrated the jointly developed software at the Anuga Food-Tec trade fair in Cologne from April 26 to 29. In this partnership, the business area Meshfree Multiscale Methods of SCAI is responsible for the highly efficient solution of large systems of equations and performance optimization. ITWM develops the underlying numerical methods for the simulation of physical processes. Dr. Thomas Soddemann (SCAI) explains his exhibit to visitors, including Victoria Appelbe, SCAI‘s Advisory Board member. The­MpCCI­toolbox­supports­engineers­in­creating­ and­assembling­customized­digital­twins May­/­June­2022 The Hannover Messe 2022 was a hybrid event from May 30 to June 2. The institute presented its research work on-site in Hannover and at a digital booth. The business area Multiphysics of Fraunhofer SCAI showed digital twin solutions for industrial production. The MpCCI software toolbox s upports engineers in creating and assembling customized digital twins. The goal is better interoperability and smarter engineering workflows. An exhibit from the business area High- Performance Computing demonstrated the benefits of material requirements and production planning. 10 A spraying full-cone nozzle: left in the experiment, right in the computational simulation Research­projects­with­a­focus­on­sustainability­ attract­visitors­to­the­Bonn­Science­Night Mai 2022 “Sustainable research” was the motto of the 13th Bonn Science Night from May 12 to 13. Fraunhofer SCAI presented two exhibits on Bonn’s Münsterplatz. The first exhibit explained how better batteries can be developed to store electricity from renewable energy sources. A central research aspect is a computer-aided search for suitable organic electroactive materials for redox flow batteries. The second exhibit used the example of a brewery to demonstrate how costs and energy can be saved through efficient production planning with the help of innovative optimization algorithms. Priyanka Gulati, business area Multiphysics, presents the benefits of interoperability in computer-aided engineering.

Selected Software Solutions Optimized­3D­ packing­­design PackAssistant is a software for optimizing packaging configurations of identical parts in containers. With PackAssistant, the exact 3D arrangement of parts with complex shapes can be calculated and displayed already in the design phase. www.packassistant.com Automatic marker creation The AutoNester software package optimizes the automatic placement of markers on fabrics, sheet metal, wood, and leather hides. By configuring the optima l nesting pattern, the software minimizes the amount of material waste. wwwscaifraunhoferde/en/autonester Reduction­of­ material waste and minimization­of­ manufacturing costs Optimized­cutting­ plans­for­the­cutting­ of­steel­profiles Automatic cut order­planning­ in textile manufacturing Generation­of­ optimized­plans­ for guillotine cuts Optimized­ arrangement of cuboid­parts­in­boxes AutoPanelSizer determines cutting layouts for producing rectangular parts from rectangular stock material and minimizes waste, production times, and manufacturing costs. www.scai.fraunhofer.de/en/autopanelsizer AutoBarSizer determines optimized plans for cutting steel profiles and other bar materials into shorter parts. The generated solutions are characterized by a high degree of material utilization, and cutting losses are minimized. www.scai.fraunhofer.de/en/autobarsizer CutPlanner is a software package used in textile manufacturing for automatically planning cut orders. CutPlanner takes a customer’s order for a clothing item and creates a cut plan, including different sizes, fabric types, or colors. www.scai.fraunhofer.de/en/cutplanner CuboNester optimizes the arrangement of cuboid parts in three dimensions. CuboNester-C is designed for cutting blocks of material, for example, from metal, marble, graphite, and foam. wwwscaifraunhoferde/en/cubonester CuboNester-P optimizes packing, for example, of disassembled furniture elements into packages. The software is also perfect for online retail that involves ever-varying parts to be shipped. wwwscaifraunhoferde/en/cubonester Algebraic multigri d methods­­for­systems SAMG is a software library for the highly efficient solution of large linear systems of equations. Such systems occur, e.g., in fluid and structural mechanics , oil reservoir and groundwater simulation, and process and device simulation. wwwscaifraunhoferde/en/samg Simulation and ­optimization­of­ energy­transport­ networks MYNTS is a multiphysical network simulator that can be used to analyze and optimize electrical circuits, gas and energy transport, and water distribution. Users can program subnetworks and individual special elements. www.scai.fraunhofer.de/en/mynts AUTO BAR SIZER 12

Bioinformatics The­fight­against­SARS-CoV-2­has­significantly­impacted­research­since­ the­spring­of­2020,­and­it­continued­when­the­pandemic­was­over.­ 2022­has­seen­an­impressive­transfer­of­technology­and­problem- solving­approaches­developed­initially­for­translational­computational­ biomedicine­in­the­neurodegeneration­and­neuropsychiatry­context­to­ the­COVID-19­topic. Research in the business area Bioinformatics of Fraunhofer SCAI represents the entire data-based value chain of translational biomedical research in academia and industry – from unstructured information mining to knowledge graph technologies and models that combine data and knowledge to actionable insights and decision support. Automated processes extract biomedical knowledge from the specialist literature and make it available in a searchable, structured form. Semantic technologies help to represent complex biological and medical knowledge in rich, disease- specific knowledge graphs. Widely recognized are the knowledge assemblies representing the complex biology of neurodegenerative diseases such as Alzheimer’s or Parkinson’s in computable cause-and-effect models. These knowledge- based models are then used to interpret and model patient- related data and to make individualized predictions. This means an essential step towards precision medicine. Tightly linked to this approach are data-driven models in drug development. Cause-and-effect models integrate knowledge about drug-target relationships in a disease-specific fashion. Big data architectures, modern machine learning methods, and artificial intelligence (AI) are used here. When combining data and knowledge-driven approaches, things get particularly exciting in science and industrial applications. This work is attracting significant interest from pharmaceutical and biotechnology companies. The health research industry (biotech to pharma to eHealth) is investing massively in “AI in biomedicine.” SCAI works closely with industrial partners, including small and medium-sized enterprises, to improve competitiveness by transferring knowledge and technology from academic 14 research to industrial application. The collaborative research and development projects provide solutions for the pharmaceutical, biotech, and life sciences software industries. COVID­infections­cause­detectable­changes­in­the­ human brain SCAI has a clear profile in computational biomedicine in neurodegenerative diseases and neuropsychiatry. At first sight, research on COVID-19 does not fit in with this. But COVID-19 is still a major threat to the world. Though the pandemic is over, alarming signs and symptoms can be observed following SARS-CoV-2 infections, which may link the virus to the original research topic of the business area Bioinformatics: neurology. Recent studies suggest that COVID-19 infections cause detectable changes in the human brain and induce neuroinflammatory processes. Based on the internal Fraunhofer project COPERIMOplus, SCAI has developed an entire project portfolio that addresses various aspects of the pandemic. Among them is the German-French project Neurodegenerative diseases affect million patients in Western societies – with an upward trend.

Optimization The­efficient­utilization­of­resources­and­associated­cost­savings­are­of­ fundamental­importance­in­industry.­When­optimizing­transport­and­storage­ capacities,­material­usage,­or­operating­production­facilities,­complex­ optimization­tasks­have­to­be­solved.­Algorithms­can­bring­about­notable­ improvements­by­solving­application­problems­and­help­to­reduce­costs,­ save­energy,­and­reduce­material­consumption­and­environmental­impact.­ The business area Optimization of Fraunhofer SCAI has developed several sophisticated software solutions for optimization problems arising in many branches of industry, commerce, or transport. Typical areas of application are: Production: cutting and packing, machine scheduling, work schedules, material consumption Logistics: transport optimization, route planning, choice of location Material­flow:­utilization of transportation means, machines, and workers, cycle times of workpieces, inventory of buffers and intermediate storages, dimensions of resources Planning: optimal utilization of area and space The CuboNester software can optimally arrange disassembled parts of furniture in outer packaging. 16 Optimized­3D­packaging­planning­for­increased­ transportation­and­storage­capacity­ Optimal container filling saves time and money regarding transportation and storage capacity. That’s why experienced packaging planners sometimes fiddle around for a long time when packing parts with complex geometries. The world’s leading software PackAssistant helps packaging planners find the ideal way to fill containers with identical parts in just a few minutes. PackAssistant facilitates packaging with rigid or flexible interlayers, with compartments or in stacks, as well as the simulation of bulk goods. The latest development is an automation module that allows data to be transferred to PackAssistant without user interaction. Companies can use it to automatically calculate a large number of containers and parts and determine critical key figures for their logistics. Software­solutions­for­optimized­cutting­patterns­ in­1D­and­2D AutoNester is a world-leading software package for the automatic creation of markers on fabrics, leather, sheet metal, wood, or other materials. AutoNester is capable of optimally nesting any quantity of parts in a very short time, minimizing material waste and respecting various constraints. The efficiency of the markers achieved by AutoNester-T is competitive to experienced human nesters. Software companies can seamlessly integrate the AutoNester optimization kernel into their software (such as CAD systems). Recently, SCAI achieved significant speed increases and improved material utilization, as well as developing special algorithms for incremental nesting and regular cutting patterns.

Multiphysics Designing­sustainable­products­virtually­and­developing­material-saving­ production­procedures­are­key­goals­of­the­industry.­For­this­purpose,­ simulation­models­from­different­engineering­disciplines­and­tools­from­ several­software­providers­are­combined­in­virtual­engineering­workflows.­ Multi-physical­simulations­consider­the­influence­of­manufacturing­history­ on­product­properties­and­the­interdependencies­of­physical­effects.­ One of the main focuses of the business area Multiphysics of Fraunhofer SCAI is the development of methods and software solutions for tasks in which engineers have to evaluate the effects of several physical disciplines. In the center of these developments stand the software packages MpCCI CouplingEnvironment and MpCCI Mapper. These tools provide a vendor-independent solution for simulation coupling and file-based data transfer and support commercial as well as research codes from numerous application disciplines. MpCCI provides an open application program interface (API) for connecting to internal or new commercial software codes. SCAI collaborates with the leading vendors of commercial simulation programs as well as developers of university research codes. In further developing MpCCI, the institute cooperates with the world’s leading software providers. The aim is to continuously improve the MpCCI interface and to develop coupling solutions for new application fields in multiphysics. These applications include creating and using high- performance ceramics in gas turbines that can withstand significantly higher temperatures (joint project BonoKeram, funded by the German Federal Ministry of Economic Affairs and Climate Action) or developing a highly efficient electric motor for compressors in hydrogen fuel cells (Fraunhofer- internal project HABICHT). Work with a methodological-algorithmic focus addresses the development of adaptive and self-learning coupling methods for fluid-structure interactions and the continuous development of robust methods for model correlation and data transfer in integrated process chains. Broad­range­for­multiphysics­simulations­in­the­ industry­ Multiphysics problems arise in numerous industrial applications: Aeroelasticity and mechanical engineering (aerodynamics in racing cars, aeroelastic effects in aeronautics, deformation of turbine blades) Vehicle and machine dynamics (extreme loads on particular components, dynamic behavior of elastic components under flow loading, for example, in valve flaps or hoses) Thermal stress analysis in vehicles, engines, or turbines (influence of hot flows on components) Thermal management in automotive engineering (combination of computational fluid dynamics with efficient radiation models, whole vehicle models) Electrical systems (dynamic loads and vibrations in electric motors, thermal stresses in electronic devices) Integrated process chains (complex and multi-step manufacturing processes) Digital­twins­rely­on­interoperable­data­to­unleash­ their­full­potential­ Multiphysics phenomena are paramount to many industrial applications and production lines. They are directly related to efficient and safe production processes or the integrity of produced parts. As multiphysics simulations are embedded in process chains and design tasks, they are successively undergoing enormous automation and optimization. In this way, they are part of Digital Twins, i.e., virtual versions of real systems. The Digital Twins contain the properties of a given 18

Fast Solvers Emerging­hardware­resources­allow­for­more­complex­computer­ simulations­in­different­industrial­design­processes.­This­further­raises­ the­quality­of­innovative­products,­reduces­the­need­for­experiments­ and­prototypes,­and­improves­product­and­process­design­cost-efficiency.­ The­pre-requisites­for­such­advanced­simulations­are­efficient­and­robust­ linear­solvers.­ Linear solvers are the kernel of numerical simulations in various industrial applications. Such simulations transfer a physical model into a linear problem with a finite number of unknowns to approximate a solution that is not analytically known. These linear systems need to be solved frequently in a simulation: for each time step and, possibly, linearization step. Therefore, the linear solution is the most time-consuming part and requires fast and robust solver methods. This is especially true as emerging hardware developments allow for more complex physics at higher model resolutions. The linear systems feature millions and even billions of unknowns and structures that often enough prevent the robust application of textbook algorithms. The business area Fast Solvers of Fraunhofer SCAI develops the library SAMG (Algebraic Multigrid for Systems). The software contains a vast variety of linear solver algorithms for the efficient and robust solution of linear systems. Hierarchical multigrid methods are used as the core strategy, offering numerical scalability: Solution efforts for doubling the number of unknowns only roughly double. This is unique to hierarchical methods and makes them superior to classical solver methods, particularly for large problem sizes. SAMG­is­a­toolbox­for­application-tailored­solver­ solutions SAMG provides not only a fixed solution algorithm but a versatile toolbox of solver ingredients that users can combine to create a solver strategy tailored to a specific kind of simulation. These components are implemented in a highly tuned manner and feature parallelization options that provide computational scalability. 20 SAMG is a licensable software solution applied in industrial and scientific practice to reduce computational time for increasingly complex simulations drastically. In several applications, it is the key to allowing for a certain level of complexity. The fields of application of SAMG comprise flow processes such as fluid dynamics in the automotive and machinery industry, optimization processes in material design, geological simulations of oil reservoirs and groundwater processes in the energy and construction sector as well as the simulation of casting, plating, and molding processes. Moreover, SAMG is applied in circuit and battery aging simulations and various applications from structural mechanics in numerous applications from electronics to construction engineering to the automotive and machinery industry. Visualization of the error smoothing procedure in AMG, which is fundamental for the overall hierarchical solution procedure.

High Performance Computing Optimization­problems­arise­in­many­fields­of­industry.­For­example,­ material­requirements­and­production­planning­are­essential­for­the­ efficient­operation­of­manufacturing­facilities.­Modern­mathematical­ methods­allow­tailor-made­solutions­to­be­engineered­for­specific­ customer­problems.­Quantum­computing­has­the­potential­to­solve­ particular­optimization­problems­in­seconds­instead­of­days­or­years.­ The research work in the business area High-Performance Computing of Fraunhofer SCAI concentrates especially on two domains: Quantum­computers­could­solve­optimization­ problems­faster­and­more­precisely­in­the­future­ optimization in the mathematical sense with a focus on combinatorial optimization and heuristic solvers optimization of code performance on various hardware systems The goal is to develop tailored solutions to the optimization problems of various industries. SCAI has supported customers in several projects in solving their specific optimization problems. This includes optimization problems for finding the optimal position of articles on the shelves of a warehouse, e.g., by minimizing distances traveled by the workers during the picking process. Another task was to assist dispatchers when stocking shelves or assigning loading bays to loading actions. Over the last two years, SCAI has created an algorithm suite that tackles the problem of solving the so-called Job Shop Scheduling Problem when liquids are processed. This suite can be successfully applied to food manufacturers (e.g., breweries) or in processing chemicals, such as paint manufacturing. SCAI demonstrated the optimization solutions at the Hannover Fair 2022, which sparked considerable interest from potential customers. For the retailer Galeria Karstadt Kaufhof in Essen, SCAI created a solution for the optimal distribution of returned articles to online shops and department stores. This project is ongoing, and SCAI is working closely with its partner Fraunhofer IAIS and the customer’s logistics provider. The combinatoric problem, in this case, is a combination or variant of a multi-knapsack and a bin packing problem. 22 Many scientists believe that quantum computing will eventually enable us to solve the above optimization problems more precisely and in less time. In the project QuantumQAP, SCAI is porting its patented solver for the so-called Quadratic Assignment Problem (QAP) to include quantum computers in finding the optimal solution. The German Federal Ministry of Education and Research funds the three-year research project, which began in January 2022. QuantumQAP aims to apply post-quantum cryptography algorithms to protect the encryption against being broken by Shore’s Algorithm with the help of quantum computers. In this project, not the algorithms are developed, but their implementation is optimized. Post-quantum cryptography algorithms can protect encryption from being broken by quantum computers.

Network Evaluation Technologies Efficient­and­high-performance­transport­networks­for­gas,­electricity,­ heat,­and­water­are­crucial­for­transforming­European­energy­use­toward­ climate­neutrality.­Smart­grids­are­becoming­increasingly­important,­as­ are­future­hydrogen­transport­networks.­The­intelligent­networking­and­ control­of­power­generators,­storage­facilities,­consumers,­and­network­ resources­are­major­economic­and­ecological­challenges. The methods and software packages developed in the business area Network Evaluation Technologies of Fraunhofer SCAI are suitable for modeling, simulating, analyzing, and optimizing networks. These are energy networks for transporting gas (including hydrogen), water, electricity, district heating, and oil. The tools from SCAI are also suitable for modeling CO2 in its liquid state. For this purpose, SCAI develops and implements methods from physical modeling, numerical mathematics, data mining, and machine learning. MYNTS­software­also­models­transport­networks­ for­hydrogen­and­liquefied­carbon­dioxide Analyzing and optimizing their networks is an essential competitive factor for energy suppliers and large consumers. For this purpose, SCAI has developed the MYNTS (MultiphYsical NeTwork Simulator) software package. MYNTS can be used to plan complex networks and predict, analyze, and optimize their behavior during operation. The software models transport networks for energy or water. The simulation shows how changes in certain factors affect the network. MYNTS can be used, for example, to calculate how temperature fluctuations change flow rates and how the failure of subnetworks affects the other network components. For network operators, this makes energy conversion and grid expansion more flexible. They save energy and money and also increase the security of their infrastructure. MYNTS is also suitable for networks transporting hydrogen or liquefied carbon dioxide. For this reason, the software is a sought-after tool for modeling the transition to hydrogen infrastructure in research projects or industrial planning. The future infrastructure of energy transport networks is 24 also the subject of the joint project TransHyDE-Sys, funded by the German Federal Ministry of Education and Research. MYNTS plays a central role in the toolchain, and several other institutes in the project use the software from SCAI. SCAI’s portfolio also comprises the simulation and analysis of electrical power networks, such as power flow analysis.

Virtual Material Design New­materials­with­tailored­properties­are­becoming­increasingly important­in­many­industrial­application­areas.­Developing­them­quickly­ and­efficiently­is­achieved­by­applying­mathematical­methods­and­ combining­scientific­and­engineering­knowledge­with­modern­ data-driven­approaches.­In­this­process,­machine­learning­techniques­ play­an­essential­role.­ The work in the business area Virtual Material Design of Fraunhofer SCAI aims to use the computer as a virtual laboratory within a data-driven design approach to develop and investigate new materials and molecules. This strategy can significantly reduce development costs for new materials and optimize manufacturing processes. For most materials design problems, the design space size increases exponentially with the number of parameters. Hence, a conventional trial-and-error approach is not applicable. To this end, SCAI follows a so-called data-driven design technique to efficiently explore the design space, including an iterative feedback loop to elaborate the optimum systematically. However, some potentially required material properties are not directly available by straightforward, efficient physical model numerical simulation. Therefore, the workflow involves an informed machine learning approach based on a smart mixture of black-box and white-box models, i.e., so-called grey-box models. Note here that the aim of white-box models is an accurate physical description of the modeled system. In contrast, a black- box model’s purpose is to reproduce just the input-output behavior of a system. However, in many cases, establishing white-box models is too complex. The amount and quality of data available to create black-box models is often insufficient, but their mixture can potentially be applied. For the outer optimization loop, optimization methods such as Bayesian optimization, are employed to balance the exploration and exploitation of the design space. In each iteration, based on the current surrogate models, new potential candidates for the most practical experiments will be suggested by the inverse design step. 26 Then, the newly generated experimental data are utilized to improve the involved predictive surrogate models leading to new suggestions of promising candidates for experiments in the next iteration. This way, the data-driven workflow will exploit existing data and lead to the application of further investigations within the materials development and design process. Note here, in particular, that learn-on-the-fly machine learning methods are applied such that the involved predictive models will be automatically improved using the workflow. This substantially reduces the number of necessary tests and simulations. Tremolo-X includes computations in nanotechnology: Simulation of the bending of a double-walled carbon nanotube.

Numerical Data-Driven Prediction Data-driven­approaches­are­becoming­increasingly­important­in­ Industry 4.0.­The­interplay­of­efficient­data­analysis­algorithms,­large­ amounts­of­sensor­data,­or­numerical­simulations­to­incorporate­domain­ knowledge­is­promising.­Machine­learning­methods­provide­targeted­ insights­from­data­that­support­decision-making­or­the­automation­of­ process­control. The work in the business area Numerical Data-Driven Prediction of Fraunhofer SCAI focuses on dealing with complex data from physical-technical applications, especially to meet the challenges of Industry 4.0. SCAI combines mathematics, machine learning, and engineering knowledge to develop robust, scalable, and domain-adapted data analysis concepts and methods. Applications can be found, for example, in virtual product development using computer-aided engineering, in condition monitoring, including predictive maintenance of wind turbines, or in the realization of digital twins. One research and development focus is the integration and use of engineering knowledge in machine learning (ML). Other topics are the efficient processing and analysis of large complex data sets, time series or numerical simulation results, the quantification of uncertainties, contributions to transfer learning, and approaches for robust design. impact, SCAI has extended its framework for automatic data analysis and event detection to the interactive exploration of many simulation results with the newly developed software SimExplore. It supports CAE engineers in their simulation data analysis workflow by providing a browsable overview of many crash simulation results, e.g., in terms of clusters and outliers based on deformations and mesh functions. SCAI uses its specifically developed, patented dimensionality reduction technique to compute geometry-driven features representing the displacements or mesh functions from the selected parts. Moreover, SimExplore provides a web-based application for interactive visualization. SimExplore thus represents a breakthrough towards automatic event detection for car crash simulations within the overall crashworthiness data analysis process. SimExplore­supports­engineers­by­providing­a­ browsable­overview­of­crash­simulation­results­ Computer-aided engineering (CAE), based on reliable numerical simulations, is a core pillar in many engineering domains. Particularly in the CAE-based crashworthiness analysis process, the large number of design possibilities combined with increasing requirements and regulations for crash safety lead to large so-called development trees. A key challenge is to learn useful designs from a set of model changes and corresponding simulation results, for example, to mitigate undesired deformations. In addition to the software tools ModelCompare and SimCompare, already on the market for comparing two design variations and their 28 The vision of the EVOLOPRO project is the automatic adaptation of production processes to rapidly changing requirements.

Meshfree Multiscale Methods The­pre-processing­step­of­computational­mesh­generation­and­ adaptation­in­established­simulation­workflows­is­time-consuming­and­ costly.­Meshfree­simulation­methods­overcome­these­problems­and­allow­ engineers­to­design­products­significantly­faster.­They­are­highly­attractive­ for­complex­industrial­use­cases,­for­example,­in­the­automotive,­chemical,­ and­hydropower­industries. Today’s product design processes typically involve highly complex geometries and fine details. The required resolution of these geometric details – which may deform and change over time – leads to the use of very fine meshes which need to be generated appropriately, i.e., adaptively refined and coarsened locally, in classical numerical simulation tools. The computational costs associated with mesh generation and local mesh adaptation pose a severe bottleneck so that simulation methods that do not require the availability of a computational mesh become highly attractive for such complex use cases. Thus, the work in the business area Meshfree Multiscale Methods of Fraunhofer SCAI focuses on developing efficient multiscale algorithms, e.g., the generalization of the finite element (FE) method or particle-based multiscale methods. MESHFREE­provides­customers­with­essential­ features to reduce cost Together with Fraunhofer ITWM in Kaiserslautern, SCAI develops the simulator MESHFREE which is built on a generalized finite difference approach and only involves a point or particle cloud instead of a classical mesh – which is far simpler to create and refine. MESHFREE is used by customers and project partners, e.g., in the automotive sector, chemical engineering, and hydropower industry, in many application areas typically involving fluid flow. The implicit MESHFREE solver employs SCAI’s SAMG fast linear solvers to attain robustness and efficiency independent of geometric details and material properties. During a laser powder bed fusion process in additive manufacturing, the metal vaporizes, forming a liquefied pool of molten metal. The PUMA software toolkit can be used to simulate this process. 30

Computational Finance Computational­Finance­is­an­interdisciplinary­area­of­scientific­computing.­ The­main­goal­is­to­develop­efficient­and­robust­algorithms­for­financial­ mathematics.­Machine­learning­methods,­which­form­the­basis­of­many­ applications­in­artificial­intelligence­(AI),­play­an­essential­role­here.­ SCAI­contributes­to­making­AI­models­understandable,­to­quantifying­ uncertainties­of­the­financial­market,­and­to­analyzing­blockchains. The focus in the business area Computational Finance of Fraunhofer SCAI is to develop efficient and robust algorithms for financial mathematics. The scopes of application include the valuation of financial and physical assets, the calculation of sensitivities and hedging strategies (risk management strategies used by short- to medium-term oriented traders and investors to hedge against unfavorable market developments), risk measurement and risk management, and big data analysis for market or blockchain data. For these applications, numerical methods such as (quasi-) Monte Carlo methods, techniques for the discretization and solution of partial differential equations, graph algorithms, and machine learning models are used and further developed. In its research and project work, SCAI focuses on the following main areas of activity. Energy­asset­optimization­and­valuation­of­green­ power­purchase­agreements­ The optimal control and hedging of portfolios of physical assets are gaining importance due to the advancing decentralization of power generation. Suppose one wants to take future price uncertainties into account during optimization. In that case, the solution to the problem often becomes computationally intensive. Additionally, it requires experienced developers who can make necessary adjustments to the parameterization of the algorithm when the portfolio composition changes. SCAI applies reinforcement learning techniques to develop a generic framework for describing and 32 The challenge for energy suppliers today is to create and manage a cost-optimal portfolio of renewables, flexible resources, and storage that guarantees customers a certain amount of green power.

The Fraunhofer-Gesellschaft The­Fraunhofer-Gesellschaft­based­in­Germany­is­the­world’s­leading­applied­ research­organization.­Prioritizing­key­future-relevant­technologies­and­ commercializing­its­findings­in­business­and­industry,­it­plays­a­major­role­ in­the­innovation­process.­It­is­a­trailblazer­and­trendsetter­in­innovative­ developments­and­research­excellence.­The­Fraunhofer-Gesellschaft­supports­ research­and­industry­with­inspiring­ideas­and­sustainable­scientific­and­ technological­solutions­and­is­helping­shape­our­society­and­our­future. advanced technology and educate and train the urgently needed next generation of research scientists and engineers. Highly motivated employees up on cutting-edge research constitute the most important success factor for us as a research organization. Fraunhofer consequently provides opportunities for independent, creative and goal-driven work and thus for professional and personal development, qualifying individuals for challenging positions at our institutes, at higher education institutions, in industry and in society. Practical training and early contacts with clients open outstanding opportunities for students to find jobs and experience growth in business and industry. The prestigious nonprofit Fraunhofer- Gesellschaft’s namesake is Munich scholar Joseph von Fraunhofer (1787–1826). He enjoyed equal success as a researcher, inventor and entrepreneur. The Fraunhofer-Gesellschaft’s interdisciplinary research teams turn original ideas into innovations together with contracting industry and public sector partners, coordinate and complete essential key research policy projects and strengthen the German and European economy with ethical value creation. International collaborative partnerships with outstanding research partners and businesses all over the world provide for direct dialogue with the most prominent scientific communities and most dominant economic regions. Founded in 1949, the Fraunhofer-Gesellschaft currently operates 76 institutes and research units throughout Germany. Over 30,000 employees, predominantly scientists and engineers, work with an annual research budget of €2.9 billion. Fraunhofer generates €2.5 billion of this from contract research. Industry contracts and publicly funded research projects account for around two thirds of that. The federal and state governments contribute around another third as base funding, enabling institutes to develop solutions now to problems that will become crucial to industry and society in the near future. The impact of applied research goes far beyond its direct benefits to clients: Fraunhofer institutes enhance businesses’ performance, improve social acceptance of 76 Fraunhofer institutes 34

Online Sections Publications www.scai.fraunhofer.de/en/media-center/publications.html Academic theses www.scai.fraunhofer.de/en/media-center/publications/academic-theses.html Travel­directions www.scai.fraunhofer.de/en/contact-fraunhofer-scai/travel-directions-birlinghoven.html Teaching www.scai.fraunhofer.de/en/about-us/Teaching.html Events www.scai.fraunhofer.de/en/events.html 36

Notes 38