Ansys SimAI - Customer-facing Presentation 2024 - Dark Theme.pptx

Ansys SimAI customer presentation

1. Ansys SimAI: Artificial Intelligence for Simulation Solution Overview

2. Ansys AI – Transforming Simulation at the Speed of AI CTRL + Click To Play The Videos

3. Ansys AI – Transforming Simulation at the Speed of AI AI Add-ons to Ansys products across portfolio AI Add-ons to various Ansys simulation products that enhance simulation Various Improvements ML platform for simulation across the physics Extremely fast and reliable physics predictions which learns from existing data 10x to 1000x Faster Virtual assistant to Ansys products Natural language assistant for documentation, training Simple & Natural UX

4. Predict at the Speed of AI Simulation result … Simulation result 4 Simulation result 3 Simulation result 2 Simulation result 1 TRAINING . . . MODEL READY 2- Your AI Model 3- In Seconds MODEL CONFIGURATION Fast. Reliable. Accessible. Performance Prediction New Design

5. Unleash the power of AI for your design with Ansys SimAI Ansys SimAI is our new cloud-based AI platform for Simulation: Train the AI without having to parametrize your geometry Predict performance across design changes, even when the geometry structure is inconsistent Leverage previously generated simulation results  to train the model Ansys SimAI is physics-neutral: Any physics – Fluids, Structures, Emag , Optics Across all industry segments – Aerospace, Automotive, Semiconductor, etc... Works with any 3D simulation data, whether it is Ansys or not 

6. Ansys SimAI – Typical workflow Data SimAI training SimAI prediction SimAI prediction SimAI prediction SimAI prediction Existing simulation data New simulation data Ansys or other solvers Test data ( future ) Simulation E xperts Stylists Bid Managers Design T eams Simulation E xperts Data conversion to VTP Minutes to upload 2 days to train the AI model Model 1 Model 2 Model … Model N Create Catalogue of trained AI models Validate Selected Best Designs with Full Fidelity Simulation Simulation E xperts Validate Explore Interactive Design and Optimization Prepare

7. Ansys SimAI – In your organization 3D simulation and test results Upload your data 1min per Gb Simulation Analyst Test Engineer Method Engineer Train your AI models Your catalog of trained AI models 48h Sedan Aero SUV EMC SUV Lidar SUV Battery SUV HVAC Compact Aero MPV Crash Sports Crash SUV Wheel Sedan Radar Compact E-motor Micro Gearbox Predict new designs AI prediction 30s Simulation Analyst Designer Project Manager Chief Architect System Engineer

8. Apply AI to different physics Electronics Optics Fluids Structures Impact performance Wire forming process Crystal plasticity homogenization Generative design Stress + deformation CFD comparison Thermal management Cooling design Antenna design & placement Magnet placement PCB EM losses and forces Electric motor design Illumination Fluent Mechanical LS-DYNA HFSS Maxwell Icepak Speos

9. Quantified benefits from validated use cases Structures Bumper Impact Performance Predict safety performance across design changes faster: >50x compared to classical crash simulation Evaluate an increased variety of hook designs : ~15x more than conventional simulation techniques Crane Hook Design Electronics Fast answers to antenna design problems for extensive trial & error, performance optimization Antenna Design and Integration Telecommunication Antenna Array Evaluate more antenna placement topologies to drive innovation Fluids Hull Design Exploration SimAI resistance error compared to CFD:  less than 4% and perfect wave pattern prediction. SimAI  Prediction on new SUV geometry in less than 1 min SUV Aero Performance Optics The fast evaluation capabilities of the SimAI model empower designers and engineers to explore more environments Optical Systems in Harsh Environments

10. SimAI In Action

11. Testimonial – Renault Group “With Ansys SimAI , we will be able to easily test a design within minutes and rapidly analyze the results, ultimately redefining our digital engineering workflow and reshaping our perception of what is possible. By enhancing simulation speed, we can explore more technical possibilities during the upstream phase of our projects and reduce the overall time-to-market.” William Becamel Expert Leader in Numerical Modelling and Simulation | Renault Group Source: https://www.ansys.com/news-center/press-releases/1-9-24-ansys-launches-simai

12. Guidance for suitable SimAI evaluation All physics are supported 30-100 simulations for a use case, require surface and optionally volume data Training dataset must be uploaded to the cloud (SaaS only solution) Data can be converted to SimAI format - . vtp and . vtu (we can help) Formulas available to compute global coefficients from surface fields Steady data supported Transient data supported with some conditions Next step : Technical Dive?

13. Appendix Validated use cases

14. Structures

15. Bumper Impact Performance Bumper design is driven by several factors: A lightweight design improves range and fuel efficiency.  Safety regulations for cars require the evaluation for robustness and manufacturing variations . Performing physical crash tests are very expensive and time consuming. A Multi-disciplinary virtual optimization approach  is needed to get the best performance in safety, durability and NVH.  ~ 50  different crash models with varying part thicknesses were evaluated to generate a surrogate AI model for bumper impact. (SimAI, LS-DYNA) SimAI model accurately predicts the bumper deformation and barrier forces as a transient response . (SimAI) SimAI Prediction on new bumper thickness in less than 1 min.  Overall crush predictions have an error of less that 0.5% and barrier force error is within 10%. (SimAI) Assess more car designs: 20x compared to traditional simulation methods and optimize quicker. Predict consistent safety performance across design changes faster (>50x compared to classical crash simulation) , ​even when the geometry structure is inconsistent by leveraging on past crash database (earlier design phase, previous car generations). Shift Left: Cut-down your design process duration and cost by allowing designers to use fast and meaningful crash prediction . Engineering Goals Solution Benefits SimAI Generated Animation Rigid-wall Force Prediction Crush comparison Overall Crush Prediction Accuracy

16. Wire Forming Process The shape of hairpin wires for electric motors depends heavily on forming process parameters. Typically, a time-consuming, manual trial-and-error process is used to fine-tune forming parameters to ensure the as-manufactured shape meets the design requirements. Need a fast and easy-to-use simulation approach to determine an optimal set of forming process parameters to produce the designed wire shapes. Automated simulation workflow to model the forming process and create a database of simulation results. (Workbench, LS-DYNA , optiSLang)   Train  AI model on stored simulation results.   (SimAI)   SimAI efficiently predicts as-manufactured shape of wire centerline in <1min , delivering a less than 5% shape difference compared to Ansys LS-Dyna. (SimAI, LS-DYNA) Rapid evaluation of different process parameter combinations. Significant time savings: 1hr of simulation vs. <1min for AI prediction. Can be integrated to an optimization framework to identify parameters to match the as-manufactured shape with the designed shape. Reduce both the time and cost of the manufacturing process by getting process parameters right early on. Engineering Goals Solution Benefits Seamless training data generation/preparation workflow Final wire shape characterized by four parameters GC0, GC1, GC2 and GC3

17. Crystal Plasticity Homogenization Microstructures have a significant impact on the mechanical material properties and performance of AM parts. Material properties for different microstructures are typically obtained from costly and time-consuming experimental set-ups. Alternatively, a  crystal plasticity simulation can be performed . Accelerate simulations with an efficient and easy-to-use surrogate model to predict the homogenized stress-strain curve for different microstructures.  Training : FE simulation results on 5 0 microstructures at 10 different strain levels were generated to train the AI model. (SimAI, Mechanical) SimAI efficiently predicts the mean stress-strain curve and the stress distribution in <1min , delivering a less than 3% stress difference compared to Mechanical. (SimAI, Mechanical) Rapid prediction of homogenized stress-strain curve, homogenized mechanical properties can be easily extracted from the predictions. Significant time savings (4min of simulation in 2D, ~1h for 3D vs. <1min for AI prediction). Reduce both the time and cost of the micro-scale crystal plasticity simulations by creating an accurate AI surrogate model. Engineering Goals Solution Benefits Stress distribution Microstructure Crystal plasticity simulation

18. Jet Engine Bracket Generative Design Coming up with new jet engine bracket designs as quickly as possible  that meet structure requirements and constraints . Moreover, it is important to make use of existing designs that have accumulated over years, retaining the knowledge from previous projects . Up to 250 training samples, with different and topologically diverse bracket designs, were run to build a global AI model. ( SimAI , optiSLang, Discovery) Based on knowledge from previous projects and simulations, optiSLang can predict a new shape , while SimAI shows you its physical behavior in less than a few seconds . The AI model can be extended and retrained with new designs at any time to capture additional physical behavior .  (optiSLang, SimAI ) 0 scraps , use all the designs you have from the past to move forward, even a bad design is a source of knowledge. Reduce your current workflow by 90% , go directly from CAD to physics. Use 100% of your team and let AI guide them with new ideas, no deep knowledge of physics required. Engineering Goals Solution Benefits D isplacement Von Mises stress 95% training samples prediction confidence 50 100 250 Discovery SimAI Discovery SimAI Displacement Von Mises stress Unseen Shape Verification Diverse Training Data

19. Crane Hook Design Optimizing crane hook design often involves multiple iterations to achieve desired performance and safety factors.  Time-consuming and may require adjustments to the geometry, material properties, and loading conditions. Drastically faster stress predictions for design ideation are required to stay competitive. Utilizing accurate FEM simulation results , an AI model was created based on 50 diverse hook design simulations . (Mechanical, SimAI) SimAI swiftly predicts stress and deformation results for new geometries within a minute , delivering a less than 1% stress difference compared to Mechanical while accurately pinpointing critical locations . (Mechanical, SimAI) Evaluate an increased variety of hook designs ( ~15x more than conventional simulation techniques ) and expedite the optimization process. Accelerated and consistent prediction of stress and deformation across design alterations ( 20 to 80 times faster ). Implement a Shift Left strategy to reduce both the time and cost of the design process by enabling designers to leverage rapid and accurate predictions . Expedite the development life cycle of the next generation heavy equipment’s design. Engineering Goals Solution Benefits Mechanical Computation: 20mins on 4 cores Sim AI Prediction Diff ~ 0.0216e-3 (0.32%) Sim AI Prediction Diff ~ 0.0205e9 (0.5%) SimAI prediction: ~50s Mechanical Mechanical

20. Fluids

21. Hull Design Exploration New environmental regulations require to reduce carbon emissions from international shipping (IMO, 2021). One of the methods is to improve ship resistance by optimizing the hull form, which directly translates into energy efficiency and fuel saving. For evaluating several concept designs fast and efficient simulations are needed. 288 accurate CFD results are used to create the AI model, including hull shape variations and operational conditions (draft, boat speed). (SimAI) SimAI Prediction on new hull geometry in less than 1 min. (SimAI) SimAI Resistance error compared to CFD:  less than 4% and perfect wave pattern prediction . (SimAI) Assess the performance of new hull designs in a few minutes instead of days . Make informed decisions in the early stages of a design project . Save time for value-added tasks: explore a range of solutions and innovate. Engineering Goals Solution Benefits SimAI prediction: ~50s CFD Calculation: 4h on 32 cores Isostatic Pressure X- WallShearStress Wave Patterns (elevation) Image : Shutterstock/studio concept

22. SUV Aero Performance CO2   emission reduction  plans (e. g.  WLTP ) require automotive manufacturers to  assess the aerodynamic performance of all design variants  of a new car family. Wind tunnel tests are too slow and costly to solve this challenge. And it is important to  shrink the development time of the next generation of electric cars .  Drastically faster predictions of aerodynamic performance are required to stay competitive. ~ 50 accurate CFD results are used to create the AI model, including car exterior shape variations and topological changes (rear mirror, ski rack, spoiler,  etc ). (Fluent, SimAI ) SimAI   Prediction on new SUV geometry in less than 1 min . ( SimAI ) SimAI Drag error compared to CFD:  less than 0.5% (5 to 10 drag counts)  and accurate skin friction field and wake topology prediction. ( SimAI ) Assess more car designs: 20x compared to traditional simulation methods and optimize quicker. Predict consistent aero performance across design changes faster (10 to 100x) , ​even when the geometry structure is inconsistent by leveraging on past CFD simulations database (earlier design phase, previous car generations). Shift Left: Cut-down your design process duration and cost by allowing designers to use fast and meaningful aero prediction. Engineering Goals Solution Benefits SimAI prediction: ~30s Fluent Calculation: 5h on 200 of cores Velocity magnitude X-Velocity Drag Solver SimAI

23. Battery Thermal Management Conceptual Design Optimize cooling design for safety and performance by reducing expensive hardware iterations. Identify hot spots & maintain temperature uniformity in battery pack in order to prolong battery life . In the global race for electric vehicles, a fast turnaround time is a necessity to stand out from the intense competition. ~ 10 accurate CFD results are used to create the AI model of the battery module, including cold plate for cooling. Training data include different configuration of cell packs, liquid coolant heat capacity, cooling inflow speed and cooling channel diameters . (SimAI) AI Prediction on 8 new cooling designs in less than 1min. (SimAI) AI temperature compared to CFD: max error 2K in the battery module, average error 0.2K .   (SimAI) Trade off analysis in early design of battery module , including different configuration of packs. Design exploration and fast assessment of cooling system performance : test dozens of new cold plate designs and coolant properties in less than 10 minutes . Engineering Goals Solution Benefits Wall temperature (K) on Battey module for an unseen new design CFD Difference Module + cooling system Example of 4 packs module Example of 5 packs module Cooling system (cold plate)

24. Power Inverter Cooling Ensure thermal reliability of electric components in critical environments. With increasing power-densities, classical thermal design (air cooled heat sink with bonded fins) are not enough. New cooling systems, including cold plates with non-conventional pin/fin designs (irregular / non-organic shapes) are built to answer thermal requirements for a large operational envelope. ~ 30 accurate CFD results are used to create the AI model of a forced convection cooling channel, installed on top of a hot power inverter, with large topology changes in the design of the pins. (Fluent, SimAI ) AI Prediction on new cooling channel in 30s. ( SimAI ) AI temperature error compared to CFD: less than 2% and accurate flow field and pressure drop predictions.  ( SimAI ) Reduce electronic cooling system development time, typically from 2 months to less than a week. Increased product reliability by testing more operational conditions , including those outside of nominal mode of operation and by optimizing heat sink cooling design for combined thermal and pressure losses. Engineering Goals Solution Benefits SimAI : ~30s Fluent : 1h on 64 of cores SimAI Fluent Examples of old cooling devices Flow temperature (unseen new design)

25. Centrifugal Pump Design Exploration Involves many design parameters, including leading and trailing edge angle at hub and shroud, number of blades. Engineering parameter of interest: Head, Torque, Efficiency. The optimization process requires an economical approach to investigating many design scenarios. ~ 113 CFD simulation results are used to create the AI model of Centrifugal Pump, with topology changes in terms of number of impeller blades, leading and trailing edge angle at hub and shroud. (Fluent, SimAI ) AI Prediction on Pressure and Velocity Field in 30s. ( SimAI ) AI Torque error compared to CFD: ~0.5% and accurate 3D pressure and velocity field predictions.  ( SimAI ) Accurate AI model for faster design space exploration and optimization for topology variations in a non-parametric way, reduces design cycle time significantly. Engineering Goals Solution Benefits SimAI : ~30s Fluent : 1h on 32 of cores Pressure Contour on Unseen Design Variation in Number of Blades

26. Solid Suspension in Stirred Tanks Distribution of the solid phase (a product, a reactant or a catalyst) affects mixing scales and availability of solids to chemical reactions, and therefore overall performance of the tank. Cloud height , defined as the location of the clear liquid interface, is a critical measure of process performance. Need to evaluate range of configuration and operating scenarios. ~ 28 accurate Liquid-Solid Multiphase CFD simulation results are used to create the AI model of Stirred Tanks, with topology changes in terms of number of impellers. Agitation Rate and % Solid Loading variations are also captured. (Fluent, SimAI ) AI Prediction on Solid Volume Fraction Field in 30s. ( SimAI ) AI Cloud Height error compared to CFD: ~2% and accurate 3D Volume Fraction field predictions.  ( SimAI ) Faster design space exploration and optimization for variation in Stirred Tank component configuration and operating parameters of Agitation Rate and % Solid Loading. Use historical data to build an accurate AI model that understands change in Impeller type and its effect on Cloud Height. Engineering Goals Solution Benefits SimAI : ~30s Fluent : 12h on 64 of cores Solid Phase Volume Fraction on Unseen Design

27. Shell and Tube Heat Exchanger A heat exchanger must be as efficient as possible , and the energy required to pump the medium through it should be minimal . The number of possible designs can be very high due to the freedom of design, which makes the search for the optimal design very time consuming . Design an efficient yet economical heat exchanger . Compare multiple designs and make situational trade-offs, making sure your design is optimal . 250 conjugate heat transfer (CHT) simulation results are used to create an AI model of the shell and tube heat exchanger, with topology changes in terms of the number of tubes, baffles, their orientation, relative positioning and inlet coolant velocity. (Discovery, SimAI ) SimAI   prediction on a new heat exchanger geometry in less than 1 min . ( SimAI ) SimAI outlet temperature error compared to the CHT simulation is less than 0.2% for the unseen design. ( SimAI ) Reduce computational time by 99%. Evaluate 3,600 designs in the same time it takes to run a standard simulation and focus on multi-objective optimization instead. Engineering Goals Solution Benefits Simulation 2 hours SimAI 20 seconds outlet temperature 57.3 °C outlet temperature 57.4 °C A Complex System with Hundreds of Pipes and Baffles Hot liquid and coolant separated by a thin steel layer hot liquid · very slow · needs to cool down coolant · very fast · takes the heat away click to play

28. Electro-Magnetics

29. Antenna Design and Integration Antennas and their integration on an aircraft platform are designed for multiple mission purposes. The broad range of requirements requires the validation of many design parameters and mission scenarios. Furthermore, antennas are often tailor-made to fit a specific customer requirement. A fast and efficient approach to quickly evaluate a large number of design variations and their performance in multiple mission scenarios. Training dataset: from 30 to 300 accurate simulation results ( HFSS ) are used to create different AI models to predict the far-field radiation pattern of different antenna designs (length and separation between dipoles in an array). (SimAI) Each new AI Prediction for new design takes ~30s. (SimAI) Far-field radiation patterns are accurately predicted by the AI model build provided that the training dataset contain at least 200 simulations.  (SimAI) Faster Time to Market : Enable fast answers to antenna design problems for extensive trial & error, performance optimization and controlled risks. Simulation data full leverage : Leverage past HFSS simulations to continuously increase prediction accuracy. Engineering Goals Solution Benefits Far-field radiation pattern indicates how well the antenna radiates SimAI: 30 training data SimAI: 100 training data SimAI: 200 training data SimAI: 50 training data Antennas are mounted on the exterior of today's airliners

30. Telecommunication Antenna Array Systems Telecom OEMs and operators require a high degree of accuracy of antenna systems and hardware before deployment. Furthermore, the antenna systems are also becoming ever more complex as they accommodate the integration of more frequency bands. It is therefore required that the antenna systems pass the antenna and system requirement specifications before manufacturing.  System simulations are complex and may take 2-4 days to simulate (HFSS). Highly fidelity antenna models. Fast and highly accurate evaluation of validations is required. Training dataset: 5 simulations that generate 30 training data sets (5 element array and frequency sweep).  (HFSS)   Training time: 14h. (SimAI) SimAI Prediction: 10 seconds. (SimAI)   Far-field radiation patterns are accurately predicted. (SimAI) SimAI was also able to predict the far-field pattern on an array with 6 elements i.e., using an additional element that was not included in the training.  (SimAI) Innovation: Evaluate more antenna placement topologies to drive innovation. Decision making: In the early product development phase, there is a strong need to know antenna performance. SimAI could be used to quickly predict performances for different scenarios on the fly. Engineering Goals Solution Benefits Far-field radiation pattern for different antenna systems (training) 4 Elements. 5-degree beam steering 5 Elements. 10-degree beam steering 6 Elements. 10-degree beam steering HFSS SimAI

31. Permanent Magnets (PMs) in Consumer Electronics Permanent magnets are widely used in consumer electronics devices, such as smart phones, tablets, earbuds, VR devices, and laptops, wireless chargers Sizing, shaping and PM material selection for various electromagnetic functions while reducing the cost Optimal placement of PM, hall sensors, shields to avoid undesirable electromagnetic interference Training dataset: tens to hundreds accurate simulation results ( Ansys Maxwell ) are used to create AI models of different fidelities ( Ansys SimAI ) to predict the magnetic field distribution and magnetic forces. Each new design is evaluated in seconds with the AI model build. Accurate magnetic field prediction by AI model facilitates device level design to place permanent magnets, sensors, and shields. Reuse of historical data: Leverage all existing data generated in early design stages to enhance model accuracy and reduce simulation time for later stages. Faster Time to Market : Provide fast evaluation and visualization of magnetic designs for design space exploration, performance optimization, EMC compliance. High Tech | Permanent Magnets Engineering Goals Solution Benefits Difference SimAI Maxwell 3D Magnitude of magnetic flux density on the container wall due to a cylindrical magnet inside Difference SimAI Maxwell 3D Magnitude of magnetic flux density inside the container Hall sensor chip Magnets Note: these four images have been purchased by Ansys and used in a previous corporate webinar -- Get the Right Permanent Magnet Latching System in Portable Devices (ansys.com) This image was created by myself using Ansys AEDT.

32. Traction Motors for Electric Powertrains Electric motors are core components of electric powertrains to provide traction in industrial, automotive and aerospace applications Complex multiphysics design needs demand computationally intensive simulations to optimize EM torque and efficiency, cooling and reduce motor noise Training dataset: tens to hundreds accurate simulation results ( Ansys Maxwell ) are used to create AI models of different fidelities ( Ansys SimAI ) Each new design is evaluated in minutes with the AI model build Smooth magnetic field and electric current distributions are accurately predicted by the AI model Faster Time to Market : Quick design space exploration and performance optimization with a focus on field distribution results in reduced design cycle Reuse of historical data: Leverage data generated in early design stages to enhance model accuracy and reduce simulation time for later stages Automotive / Aerospace | Electric Traction Motors Engineering Goals Solution Benefits Difference SimAI Maxwell 3D Magnitude of magnetic flux density in the stator and rotor cores Difference SimAI Maxwell 3D Magnitude of current density in the rotor cage of an induction motor Image source: Cutaway View Electric Vehicle Motor Suspension Stock Illustration 1976442890 | Shutterstock This image was created by myself using Ansys AEDT.

33. Optics

34. Optical Systems in Harsh Environments The design of new optical systems for automotive and aerospace application needs to take into account diverse environmental conditions, particularly under the influence of factors such as fog and haze. Fast optical system design methodologies are needed to ensure its applicability to a variety of setups commonly encountered in practical scenarios. Leverage existing simulation datasets comprising optical simulation data under various environmental conditions ( 10 simulation results ). Integrate discrete variables, such as fog density, haze levels, and weather conditions, as input features to the AI model. ( SimAI , Speos) AI Prediction: accurate illuminance field prediction with 2% error in the calculation of the max illuminance for unseen new design. ( SimAI ) By accurately predicting signal integrity issues in advance, the AI model helps minimize downtime by allowing operators to take preventive measures. The fast evaluation capabilities of the SimAI model empower designers and engineers to explore more operational environments without needing several hours to get simulation results. Engineering Goals Solution Benefits Training dataset Max illumination prediction (x axis): SimAI vs Speos Ansys SimAI Ansys Speos