Engineering Simulation Tools Overview

PINNs for Electromagnetic Wave Propagation Nilufer K. Bulut https://lnkd.in/dcu36pbM This paper tackles some of the persistent challenges in applying Physics-Informed Neural Networks (PINNs) to electromagnetic wave propagation, specifically focusing on improving accuracy and energy conservation. The core idea is to introduce a hybrid training strategy that combines time marching, interface continuity losses, and a Poynting vector-based regularizer. Technically, the approach addresses three key issues: 1) Causality collapse: Solved by using a time-marching scheme, essentially breaking the problem into smaller time intervals. 2) Discontinuities at time interfaces: Mitigated by a two-stage interface continuity loss function, ensuring smoother transitions between time steps. 3) Energy drift: Controlled via a local Poynting vector regularizer, penalizing deviations from expected energy flow based on Maxwell's equations. The PINN architecture itself isn't explicitly detailed in the abstract, but the focus is clearly on the loss function engineering. The reported NRMSE of 0.09% and L^2 error of 1.01% are impressive, and the 0.024% relative energy mismatch suggests a significant improvement in energy conservation compared to naive PINN implementations. It's also important to note that the training is purely physics-informed, without relying on labeled field data. Electromagnetics is a computationally intensive field, and while FDTD and FEM are well-established, they can be expensive for large-scale or inverse problems. PINNs offer a mesh-free alternative, but their accuracy and stability have been a concern. This work demonstrates that by carefully crafting the loss function and training strategy, PINNs can achieve results comparable to FDTD in canonical electromagnetic examples. This opens the door for using PINNs in scenarios where mesh generation is difficult or where real-time solutions are needed, potentially impacting areas like antenna design, metamaterials, and medical imaging. The focus on energy conservation is also crucial, as it ensures the physical plausibility of the solutions, a key requirement for scientific applications.

𝗗𝗼𝗻’𝘁 𝗝𝘂𝘀𝘁 𝗥𝗲𝗮𝗱 𝗔𝗯𝗼𝘂𝘁 𝗔𝗜 𝗶𝗻 𝗠𝗮𝗻𝘂𝗳𝗮𝗰𝘁𝘂𝗿𝗶𝗻𝗴. 𝗔𝗽𝗽𝗹𝘆 𝗜𝘁. The AI headlines are exciting. But if you're a founder, engineer, or educator in manufacturing, here's the question that actually matters: 𝗪𝗵𝗮𝘁 𝗰𝗮𝗻 𝘆𝗼𝘂 𝗱𝗼 𝘵𝘰𝘥𝘢𝘺 𝘁𝗼 𝘁𝘂𝗿𝗻 𝘁𝗵𝗲𝘀𝗲 𝗶𝗻𝗻𝗼𝘃𝗮𝘁𝗶𝗼𝗻𝘀 𝗶𝗻𝘁𝗼 𝗲𝘅𝗲𝗰𝘂𝘁𝗶𝗼𝗻? Let’s get tactical. 𝟭. 𝗦𝘁𝗮𝗿𝘁 𝘄𝗶𝘁𝗵 𝗔𝗜 𝗱𝗲𝗺𝗮𝗻𝗱 𝗳𝗼𝗿𝗲𝗰𝗮𝘀𝘁𝗶𝗻𝗴 Tool to try: Lenovo’s LeForecast A foundation model for time-series forecasting. Trained on manufacturing-specific datasets. 𝗨𝘀𝗲 𝗶𝘁 𝗶𝗳: You’re battling supply chain volatility and need better inventory planning. 👉 Tip: Start by connecting your ERP data. Don’t wait for perfect integration: small wins snowball. 𝟮. 𝗕𝘂𝗶𝗹𝗱 𝗮 𝗱𝗶𝗴𝗶𝘁𝗮𝗹 𝘁𝘄𝗶𝗻 𝗯𝗲𝗳𝗼𝗿𝗲 𝗯𝘂𝘆𝗶𝗻𝗴 𝘁𝗵𝗮𝘁 𝗻𝗲𝘅𝘁 𝗿𝗼𝗯𝗼𝘁 Tools behind the scenes: NVIDIA Omniverse, Microsoft Azure Digital Twins Schaeffler + Accenture used these to simulate humanoid robots (like Agility’s Digit) inside full-scale virtual factories. 𝗨𝘀𝗲 𝗶𝘁 𝗶𝗳: You’re considering automation but can’t afford to mess up your live floor. 👉 Tip: Simulate your current workflows first. Even without a robot, you’ll find inefficiencies you didn’t know existed. 𝟯. 𝗕𝗿𝗶𝗻𝗴 𝘆𝗼𝘂𝗿 𝗤𝗔 𝗽𝗿𝗼𝗰𝗲𝘀𝘀 𝗶𝗻𝘁𝗼 𝘁𝗵𝗲 𝟮𝟬𝟮𝟬𝘀 Example: GM uses AI to scan weld quality, detect microcracks, and spot battery defects: before they become recalls. 𝗨𝘀𝗲 𝗶𝘁 𝗶𝗳: You’re relying on spot checks or human-only inspections. 👉 Tip: Start with one defect type. Use computer vision (CV) models trained with edge devices like NVIDIA Jetson or AWS Panorama. 𝟰. 𝗘𝗱𝗴𝗲 𝗶𝘀 𝗻𝗼𝘁 𝗼𝗽𝘁𝗶𝗼𝗻𝗮𝗹 𝗮𝗻𝘆𝗺𝗼𝗿𝗲 Why it matters: If your AI system reacts in seconds instead of milliseconds, it's too late for safety-critical tasks. 𝗨𝘀𝗲 𝗶𝘁 𝗶𝗳: You're in high-speed assembly lines, robotics, or anything safety-regulated. 👉 Tip: Evaluate edge-ready AI platforms like Lenovo ThinkEdge or Honeywell’s new containerized UOC systems. 𝟱. 𝗕𝗲 𝗲𝗮𝗿𝗹𝘆 𝗼𝗻 𝗰𝗼𝗺𝗽𝗹𝗶𝗮𝗻𝗰𝗲 The EU AI Act is live. China is doubling down on "self-reliant AI." The U.S.? Deregulating. 𝗨𝘀𝗲 𝗶𝘁 𝗶𝗳: You're deploying GenAI, predictive models, or automation tools across borders. 👉 Tip: Start tagging your AI systems by risk level. This will save you time (and fines) later. Here are 5 actionable moves manufacturers can make today to level up with AI: pulled straight from the trenches of Hannover Messe, GM's plant floor, and what we’re building at DigiFab.ai. ✅ Forecast with tools like LeForecast ✅ Simulate before automating with digital twins ✅ Bring AI into your QA pipeline ✅ Push intelligence to the edge ✅ Get ahead of compliance rules (especially if you operate globally) 🧠 Each of these is something you can pilot now: not next quarter. Happy to share what’s worked (and what hasn’t). 👇 Save and repost. #AI #Manufacturing #DigitalTwins #EdgeAI #IndustrialAI #DigiFabAI

People often ask how I learned so many ECAD tools. The answer is simple: I do not start by learning the tool. I start by understanding the engineering problem the tool is trying to solve. Every ECAD platform is just automating a solution to a problem. Routing tools solve connectivity and manufacturability problems. Constraint managers solve electrical, physical, and timing problems. DFM and DFA checks solve fabrication and assembly risk problems. Library systems solve repeatability and data integrity problems. Output generation solves communication between engineering, fabrication, assembly, and test. Once you understand the actual problem deeply enough, the specific software becomes the medium. That is why I tell my students: Don't focus only on memorizing where buttons are. Focus on the root cause of the problem, the engineering decision being made, and the type of solution the tool is offering. Then when you open a new ECAD tool, you are not lost. You become the evaluator. You start asking: Where are the design rules? Where are the assembly rules? Where is stackup controlled? Where are impedance constraints defined? How does this tool handle variants, fabrication outputs, libraries, and collaboration? That is the difference between learning software and becoming an engineer. The tool helps implement the solution. The engineer understands why the solution is needed in the first place.

Are we looking at software engineering the wrong way? What if it’s less about writing code and more about making better decisions? Learn a revolutionary approach to understanding complex software systems in my conversation with Tudor Girba, the CEO of feenk. We explore 𝙈𝙤𝙡𝙙𝙖𝙗𝙡𝙚 𝘿𝙚𝙫𝙚𝙡𝙤𝙥𝙢𝙚𝙣𝙩, a groundbreaking concept that challenges traditional views of software engineering. Learn why treating development as a decision-making process, supported by custom tools, is crucial for tackling today’s software challenges, especially when dealing with legacy systems. Key topics discussed: ⤷ 𝗦𝗼𝗳𝘁𝘄𝗮𝗿𝗲 𝗘𝗻𝗴𝗶𝗻𝗲𝗲𝗿𝗶𝗻𝗴 𝗮𝘀 𝗗𝗲𝗰𝗶𝘀𝗶𝗼𝗻-𝗠𝗮𝗸𝗶𝗻𝗴: Why software development is fundamentally about making informed decisions rather than just constructing systems. ⤷ 𝗧𝗵𝗲 𝗜𝗻𝗲𝗳𝗳𝗶𝗰𝗶𝗲𝗻𝗰𝘆 𝗼𝗳 𝗥𝗲𝗮𝗱𝗶𝗻𝗴 𝗖𝗼𝗱𝗲: Developers spend over 50% of their time reading code, yet this activity remains unoptimized. ⤷ 𝗠𝗼𝗹𝗱𝗮𝗯𝗹𝗲 𝗗𝗲𝘃𝗲𝗹𝗼𝗽𝗺𝗲𝗻𝘁: Learn how creating custom tools tailored to specific problems can revolutionize your workflow and decision-making process. ⤷ 𝗟𝗲𝗴𝗮𝗰𝘆 𝗦𝘆𝘀𝘁𝗲𝗺𝘀 𝗮𝘀 𝗢𝗽𝗽𝗼𝗿𝘁𝘂𝗻𝗶𝘁𝗶𝗲𝘀: Reframe legacy systems as value-creation opportunities instead of burdens. ⤷ 𝗚𝗹𝗮𝗺𝗼𝗿𝗼𝘂𝘀 𝗧𝗼𝗼𝗹𝗸𝗶𝘁: Discover the innovative development environment enabling thousands of micro-tools for better system understanding. ⤷ 𝗧𝗵𝗲 𝗙𝘂𝘁𝘂𝗿𝗲 𝗼𝗳 𝗗𝗲𝘃𝗲𝗹𝗼𝗽𝗺𝗲𝗻𝘁 𝗘𝗻𝘃𝗶𝗿𝗼𝗻𝗺𝗲𝗻𝘁𝘀: Explore how AI, moldable development, and tools like Glamorous Toolkit can coexist to solve diverse class of problems. This conversation will completely transform how you think about software development!

How does automakers make sure the #electronics inside of your car is safe against external electromagnetic noise? We have #emc tests such as radiated immunity where we have an #antenna outside the car generating noise and we need to make sure all the electronics such as #pcb and ECUs inside the car are working as expected. Here we have a test where we measure the electric field on the dashboard of the car and calibrate the antenna to generate a 70V/m incident E-field from 30 to 100Mhz. The goal is to identify if a critical system is placed in an area where it will be susceptible to high intensity fields. The simulation results with #ansys #hfss are very close to measurements (I would say this is a good correlation considering the complexity of the test and numerical model as well as uncertainties) where we can see peaks of around 200V/m. Simulation of course can provide lots of insight, since it does not need a physical prototype or an anechoic chamber (which can be very expensive and difficult to get access to) so it can be used early in the design stage evaluating different scenarios. We published the results here in case you need more details: J. Mologni, et al., "The significance of specific vehicle parts on automotive radiated immunity numerical simulations," 2015 SBMO/IEEE MTT-S International Microwave and Optoelectronics Conference, doi: 10.1109/IMOC.2015.7369065.

I designed an antenna in HFSS software, so how do we connect it in VLSI, Simulink? You’ve already designed the antenna in HFSS, which handles the electromagnetic (EM) side. Now you want to connect it in VLSI and Simulink, meaning: How does this antenna interact with RFIC/VLSI circuits, and How do you simulate system-level behavior in MATLAB/Simulink? OVERVIEW Think of your HFSS antenna, VLSI circuit, and Simulink system as three layers: 1. Layer - EM Level Tool - HFSS Function - Antenna radiation, impedance, S-parameters 2. Layer - Circuit Level Tool - Cadence/ADS/VLSI Function - LAN, Mixer, PA, Filters, etc. 3. Layer - System Level Tool - MATLAB/Simulink Function - Modulation, baseband, link performance So, you “connect” them by exchanging simulation data and models between these layers. STEP 1: From HFSS → VLSI (RF Front-End Integration) > (a) Export Antenna Results In HFSS, after simulation: Export S-parameters (like antenna.s2p file). It contains reflection (S11) and transmission characteristics. File → Export → Solution Data → S-parameters (.s2p) > (b) Import to Circuit Simulator In Cadence, ADS, or Virtuoso: Add your antenna as a 2-port S-parameter block. Connect it to your RFIC circuit (LNA, PA, etc.). Simulate impedance matching, gain, and noise figure. So, now your VLSI circuit "sees" the antenna as a real component. STEP 2: From VLSI → Simulink (System-Level Co-simulation) Once you have the antenna + circuit response, you move to Simulink to analyze full system behavior (like BER, modulation, signal power). > (a) Export VLSI or Circuit Model From Cadence/ADS: Export behavioral or equivalent model (like gain, noise, and nonlinearity parameters). Optionally generate Verilog-A model or use a baseband equivalent block. > (b) Import in Simulink In Simulink: Use the RF Blockset or Antenna Toolbox. Import: The antenna impedance model (from HFSS), The circuit model (from VLSI). Example blocks: RF Amplifier → for LNA/PA S-Parameter block → load your .s2p file Antenna block → for pattern and impedance matching Modulator/Demodulator → for system-level testing STEP 3: Validation Once integrated: Check S11 < –10 dB → good matching. Check Gain, Noise Figure, and Bandwidth. Use Smith Chart and Spectrum Analyzer in Simulink to verify performance. Real-World Example Domain - Antenna, RF Circuit, System, Result Tool - HFSS, Cadence, Simulink, BER, Gain, S11 Example - 28 GHz Patch antenna, LAN designs in 65 nm CMOS, 5G Transceiver Model, Evaluated Via co-simulation #RFIC #InnovationInElectronics #TechLearning #FutureEngineer #DesignAndSimulation #STEMEducation #HardwareDesign #EDAtools #MixedSignalDesign #RFSystemDesign #MicrowaveEngineering #RFSimulation #SystemOnChip #Electromagnetics

This is the moment simulation becomes more important than prototyping. In our last posts, Pascalis and I showed two things: First, how you can generate a full production and warehouse environment in NVIDIA Omniverse using Claude Code and the USDA data format. Second, how NVIDIA’s new Kimodo model can generate robot motions from simple text prompts. Now we are taking the next step: Transferring robot motion into Omniverse and merging both use cases. Omniverse is not just for static visualizations. It allows dynamic simulation of movements, interactions and behavior with CAD components inside a virtual environment. And this is where it gets interesting for future product development. The vision is clear: If we can model production environments, warehouses, and real operating environments of products, we can simulate mechatronic products in realistic conditions before they physically exist. Environment → Sensor & actuator interaction → Model-in-the-loop simulation. Very similar to how autonomous vehicles are developed today, but applied to all kinds of mechatronic products. The effects are huge: • Less physical prototyping • Earlier insights without building hardware • Faster iteration cycles • Better product decisions earlier in development • Simulation becomes the main development environment Omniverse already shows how granular these simulations can be created today. Not through months of manual modeling, but increasingly through prompts that generate environments, movements and soon maybe even control logic. We are moving from designing products to designing behavior in simulated worlds first. And that will fundamentally change how we develop products. Curious to hear your thoughts! When will simulation become the primary development environment in your industry? Vlad Larichev | Rüdiger Stern | Rick Bouter | Ruben Hetfleisch | Dr.-Ing. Tobias Guggenberger

Open-source simulation tools can be difficult to use due to lack of a GUI and the need to get familiar with custom syntax. Tools like Claude and Codex change that. I'm using openEMS to run some initial geometry sweeps to begin tuning the cavity antenna shape for the open-source smartwatch. OpenEMS uses MATLAB/octave as an interface and especially the meshing step has been a very tedious and time consuming process. However with tools like Claude or codex the entry hurdle for anyone wanting to start experimenting with OpenEMS has been significantly reduced. Writing import scripts to convert custom STL files to workable rectilinear meshes has gotten very easy. The attached animation shows 20 simulation runs and their corresponding radiation patterns. I've varied the geometry of the dielectric insert that sits in the slot and seals the speaker against the watch frame to maintain the IP67 rating. I've automated mesh generation directly from the STL input geometry, allowing me to iterate quickly through geometry changes. For each simulation, I generate electric field and current density dumps in ASCII format so I can easily read the field distributions for use in later automated optimization loops. #simulation #electronics #hardware #hardwaredesign #design #opensource #automation

 🚀 Today I'm releasing py2femm — the FEA companion for open-source power-electronics design. A few years back I open-sourced pyplecs to automate PLECS simulations. py2femm is its younger brother: where pyplecs handles the topology, py2femm handles the finite-element world behind the schematic — thermal, magnetic, electrostatic. Power electronics design is never just the circuit. It's:  → Will this heatsink drop 20 °C or 200 °C?  → Does my inductor saturate under peak current?  → What's the parasitic capacitance between that bus bar and the chassis? FEMM answers all of these. The pain is doing it at scale — parametric sweeps, CI pipelines, shared licenses — from a Windows GUI.  py2femm gives you:  ✅ Pure Python API for magnetics, electrostatics, heat flow, current flow  ✅ REST server — run FEMM from a notebook, a Linux box, or CI  ✅ Cross-platform via Wine / Docker / Windows  ✅ Built-in parametric engine — 360-config factorial heatsink sweep in ~10 min  The open-source power-electronics toolbox keeps getting stronger:  🔹 #pyplecs topology-level simulation (PLECS)  🔹 OpenMagnetics — the gold standard for magnetic design  🔹 #py2femm — thermal, EM field, and boundary-condition workflows Together they let you go from topology → magnetic sizing → thermal sign-off. All in Python. All open  source. All reproducible.  🔗 Repo: https://lnkd.in/e6_C_UHg  📚 Docs: https://lnkd.in/evdzkdVS  📜 License: AGPL-3.0  If you design converters, what examples would you want to see next?  #PowerElectronics #OpenSource #FEMM #FEA #Python #MagneticDesign #ThermalDesign #ConverterDesign

𝗛𝗼𝘄 𝗮𝗰𝗰𝘂𝗿𝗮𝘁𝗲 𝘀𝗵𝗼𝘂𝗹𝗱 𝗮 𝗵𝗲𝗮𝗱 𝗺𝗼𝗱𝗲𝗹 𝗯𝗲 𝗳𝗼𝗿 𝗥𝗙 𝘀𝗶𝗺𝘂𝗹𝗮𝘁𝗶𝗼𝗻𝘀? Working frequency is important. At lower frequencies, energy penetrates deeper, while at mmWave it’s mostly superficial (skin-depth scale). Moreover, the human head is multilayer: skin, fat, skull, gray/white matter, etc. Each layer has different dielectric properties and loss that vary with frequency. There is always a trade-off. More anatomical detail increases simulation time and memory consumption. 🟠 What accuracy you actually need depending on use case: - ≤1 GHz: A homogeneous sphere with average head properties is usually sufficient for these frequencies. - Sub-6 GHz, wearable/SAR simulation: Use a voxel/segmented head with skin, skull and brain. Include hand/ear if relevant. - mmWave (24–100 GHz): Absorption is skin-limited. You need layered skin and realistic curvature at the site. ❓ Have you done simulations with a human body model?