Kushal Sahare

Great insight from MathWorks' Application Engineering Manager Paul Barnard, who shared some tips to help engineers elevate their careers to the next level. 🔗 https://spr.ly/6042CI2OA

1

Recently, I began exploring Robot Operating System (ROS) by building a small project in turtlesim. The setup involved a controller node that subscribes to pose topics, computes relative positions, and publishes velocity commands to guide one turtle toward others. Interactions such as removing turtles were handled through ROS services. This project gave me practical exposure to core ROS concepts including node architecture, publisher/subscriber communication, service calls, and basic feedback control. Still early in the process, but it was a useful exercise in understanding how these components integrate within a ROS system. #ROS #Robotics

30

3 Comments

Join me this Thursday for the Advanced eMotor Design Conference organized by Powersys and JSOL Corporation at Southfield, MI, where I will present our research on the "Accelerated Multi-Criteria Design Optimization of Axial Flux Motors with 3D FEA". This is a joint investigation with my fellow colleagues in Europe, Dr. Radu Andrei Negrila and Dr. Eberhard Kull. ⚡💡 Event details: https://lnkd.in/e6_Eq3Dn Presentation abstract: Several factors and considerations affect the design of electric machines for traction applications. In the case of axial flux machines, this exercise is even more challenging given the need for 3D FEA simulations and that torque/power scaling is not as straightforward as industry-standard radial flux topologies. Data-driven design processes coupled with detailed analyses & representative models can effectively reduce the engineering time needed to arrive at optimal designs for electric machines. In this presentation, a case study for the accelerated multi-criteria design optimization of axial flux machines with 3D FEA simulations will be showcased using the Python/MATLAB scripting capabilities of JMAG Designer. 💻⏳ #WeAreSchaeffler #WePioneerMotion #AxialFluxMotors #JMAG

48

I just published an analysis that many mechanical engineers overlook: how lead-in chamfers in thin materials dramatically reduce usable thread engagement, and why ignoring this can lead to undersized fasteners that fail in production. The scenario: a #8-32 tapped hole in 1/8" aluminum plate. Many assume the full 0.125" provides thread engagement for load calculations. But that chamfer you need for assembly? It's eating into your capacity more than you think. In my analysis, I compared a conservative chamfer design against standard industry practice. The difference? Just 0.0075" in chamfer size. The result? An 18% reduction in achievable clamp force, enough to turn a robust joint into a field failure. The article walks through: Why unchamfered holes fail in production (and why partial threads don't bear load). Three scenarios with actual calculations using ThreadCalc Pro. Specific design recommendations for countersink tolerancing. If you design threaded joints in thin materials, this one's worth a read.

10

🎯 DAY 28  –  " 🔀Crosstalk isn’t noise. It’s interference you invited." 👋Hey folks! Not all noise comes from the outside. Sometimes your own traces turn against each other. That’s Crosstalk. 🧠 What is Crosstalk? Crosstalk is unwanted coupling between adjacent signal traces — like one wire “whispering” into another. It happens due to: 🔹Capacitive coupling (electric fields) 🔹Inductive coupling (magnetic fields) The faster the signal (rise time), and the longer the traces run together, the worse the crosstalk gets. 🎯 Where Crosstalk Matters • High-speed digital buses (DDR, SPI, LVDS) • ADC/DAC analog signals next to clocks or digital lines • Long, parallel traces on same layer • Poor return path (disrupted ground) 🚨 Types of Crosstalk ✅ NEXT – Near-End Crosstalk 🔹Noise appears at the same end as the transmitter. 🔹Caused by inductive and capacitive coupling. ✅ FEXT – Far-End Crosstalk 🔹Noise shows up at the receiving end of the victim line. 🔹Also due to field coupling & Usually weaker than NEXT, but still disruptive — especially in long or matched lines 📏 Spacing Rules to Reduce Crosstalk ✅ Rule of Thumb: Keep 3W to 5W spacing between signal traces (where W = width of the trace) 🔧 Design Scenario: Differential pair width (W) = 5 mil Pair spacing (S) = 7 mil (center-to-center spacing between + and – lines) Adjacent trace → keep it ≥ 3W to 5W from the edge of the pair. Spacing between differential pair and nearby signals should be ≥ 15 mil to 25 mil 🛡️ This prevents aggressor traces from coupling into either line of the pair, preserving common-mode noise immunity. 🛡️ Additional Tips to Reduce Crosstalk ✅ Use Solid Reference Planes Ensure every signal trace has a consistent ground return path beneath it. ✅ Increase Spacing The simplest and most effective: don’t route traces too close. ✅ Avoid Long Parallel Runs If signals must cross paths, route them perpendicularly on adjacent layers (90°). ✅ Use Guard Traces Ground traces placed between signals (with vias) help isolate. ✅ Differential Pairs Keep pairs tightly coupled (small spacing) and isolated from other signals. Crosstalk isn’t external EMI. It’s a design issue — not an environmental one. Fix it with spacing 📌 Like this post if you found it helpful 🔍 Learning never stops — Stay tuned! #PCBDesign #HardwareDesign #LearnElectronics #MixedSignalDesign #CircuitDesign #STEM #TechExplained #Microcontrollers #EmbeddedSystems #ElectronicsBasics #DailyLearning #BuildInPublic #TechInsights #ElectronicsEngineering #LinkedInCreators #TechExplained #EngineerOnLinkedIn #EsteemPCB #PowerBudget #ElectronicsDesign #DesignTips #EngineeringInsights #50Ohms #ControlledImpedance #Crosstalk #NEXT #FEXT #HighSpeedDesign

49

Every engineer faces the same decision: when is “good enough” good enough? Hand calcs teach intuition and move fast. Spreadsheets and scripts help speed iterations with a little upfront cost. FEA/CFD gives higher precision at a sometimes heavy speed cost. I often revisit this trade-off between speed and precision. I'm interested in how you decide where to stop, what benefits do these bring you in your field, and where "good enough" is for you?

2

1 Comment

Can you tolerance "invisible" features? 🤔 I recently got asked: "Can virtual features be controlled by geometric tolerances?" The answer might surprise many engineers: ABSOLUTELY YES! Here's what most people miss about derived/centerline features: ✓ Straightness → Controls both surfaces AND centerlines ✓ Flatness → Works on center planes too ✓ Orientation tolerances (⊥ ∠ //) → Perfectly suited for axes and center planes ✓ Position, Concentricity, Symmetry → ASME Y14.5: center features ONLY! The reality check: ⭕ 10 out of 14 GD&T symbols can control derived center features ⭕ They can also serve as datums (like Datum B, C in the image) ⭕ Combined with material condition modifiers (MMC/LMC), they unlock powerful design flexibility Hot take: If you're only using GD&T to control surfaces, you're missing half the power of tolerance engineering! 🎯 💬 Question for the community: What's your go-to approach for tolerancing hole patterns - surface controls or center feature controls? Why? _______________ We make excellent parts in China for customers worldwide. Need a quote? Give us a shout! #MechanicalEngineering #GDT #ToleranceAnalysis #DesignEngineering #ASME #Manufacturing

80

11 Comments

High-performance inference in a sealed enclosure often fails the first real-world stress test. Everyone focuses on the model's accuracy, but they ignore the thermodynamics. When the silicon hits its limit, the system governor aggressively throttles clocks to prevent damage. We recently measured a scenario where sustained inference frame rates dropped by 40% simply because the case temperature crossed 75°C. This introduces a severe operational risk that simulations often miss. If thermal protection triggers during a critical security event, the device doesn't just slow down; it effectively goes blind. The ethical hacker mindset tells us to find the breaking point before the field does. We have to design for the worst-case thermal envelope, not the ideal lab benchmark. Pragmatic engineering accepts slightly lower peak performance for guaranteed availability. You cannot secure a system that shuts itself down to stay cool. #EmbeddedSystems #EdgeAI #ThermalManagement #RnD #IoT

1

🔬 Coming Soon: A Structured Training Program for Developing #CFD Solvers in #Fortran After two decades of hands-on experience in numerical methods, #PDE solvers, linear algebra kernels, and CFD algorithm development, I have designed a methodical teaching framework that guides learners step-by-step toward building real CFD codes. In an upcoming training program, I will teach the complete process of developing a 2D incompressible Navier–Stokes solver in Fortran, covering: • Domain discretization and mesh handling • Implementation of spatial discretization schemes • Pressure–velocity coupling algorithms • Time integration strategies • Boundary condition treatment • Solver optimization and code performance • Debugging workflows and verification steps 🎯 Outcome: Participants will be able to independently implement and simulate a full 2D incompressible flow problem, using a solver they develop during the course. 🎥 Note: The example simulation (https://lnkd.in/e_T3gAfB) (https://lnkd.in/dKDSMm3g) shown is illustrative—actual proficiency will depend on your practice and commitment. And let me share a line often linked to Winston Churchill — perhaps said in another context, but perfectly fits the learning curve of CFD programming: “This is not the end. It is not even the beginning of the end. But it is, perhaps, the end of a beginning.” Stay tuned for the full announcement. Wadud Miah Prof. Rafael Gabler Gontijo

81

4 Comments

Try the GPU Return On Investment Estimator tool to see which GPU configuration offers the best combination of speedup, payback, and workload fit for your Ansys Fluent and Ansys Mechanical simulations. https://bit.ly/4u0Z8gk

11

See how the Vanderhall team built a vehicle model using Simulink helping engineers deploy the Brawley all-electric off-road UTV prototype in just eight months. ⚡ #VirtualVehicle #EV #ModelBasedDesign https://spr.ly/6043A8fWF

3

1 Comment

"I would like to mention low code workflows and apps thae help you along the process." Watch how Paola Jaramillo, MathWorks Application Engineer, answers Jousef Murad in this podcast about #ai for #engineering #leaders ➡️ spr.ly/61106A6ugK

5

Living The Simulation I was knee-deep in Python code, building a tool to validate explicit dynamic simulation scripts—checking that vendors followed best practices for drop tests, shock analysis, modal extractions. Quality control. Process automation. Pretty standard CAE engineering stuff. But then something clicked. I was building a tool to check simulations. Simulations that predict how physical systems behave. Predictions we trust enough to base million-dollar design decisions on. And I started wondering: - Why do we simulate? - What drives this need to build virtual worlds before we build real ones? That question led me down a neuroscience rabbit hole I wasn't expecting. Turns out, your brain is already running simulations. Constantly. Right now. It's called predictive coding—your brain doesn't passively receive reality, it actively predicts what should happen next and checks those predictions against incoming sensory data. Every color you see, every sound you hear, every movement you make—it's all your brain's best guess, a controlled hallucination refined by prediction errors. We don't build FEA models and run drop tests because it's convenient. We build them because simulation is what we ARE. Your brain has been virtually "poking" the world since birth—running mental experiments, testing hypotheses, minimizing uncertainty to survive. Engineering simulations are just that same drive, externalized in silicon and code. Curiosity isn't a luxury. It's evolution's survival mechanism. Uncertainty is dangerous. Accurate predictions keep us alive. So when you discretize a mesh, apply boundary conditions, and hit "solve"—you're not doing something artificial. You're extending your brain's natural prediction machinery beyond what neurons alone can do. To simulate is human. I recorded a 45-minute audiobook exploring this connection between neuroscience, curiosity, and why we build virtual worlds. No fluff—just dense, mind-bending ideas grounded in real science. Fair warning: it will rewire portions of your brain. That's not Joe-speak, that's neuroplasticity. No worries—no animals were harmed during the making of this work. What's mind-altering is the fact that as you listen to this, portions of your brain are rewiring. That's learning. As with all my postings, my goal is to provoke thought, hopefully leading the curious to explore deeper and outside of their so-called swim lanes. That's the point. What do you think—does your curiosity feel different when you recognize it as a survival mechanism rather than a hobby? Audiobook -> https://lnkd.in/gYh3vFH4 #CAE #Simulation #Neuroscience #Engineering #FEA #PredictiveCoding #Curiosity

4

I spoke with Terry Denery of MathWorks at the Robotics Summit & Expo in Boston. If you’re building a robot system and you’re not using Matlab and Simulink you’re missing out. It’s a proven way for doing anything mechanical or mechatronic related and that’s why most automotive company and space company are customers. The top three companies that get the most comments and likes will get in depth follow up interviews and three months of free advertising on https://lnkd.in/g_dhJeiE. Comments are worth 3 points, likes are worth 1 point. Previous video: https://lnkd.in/gAKSYWAh Next video: https://lnkd.in/g6n7Mi-8 These videos are sponsored by MHI: The Association That Makes Supply Chains Work, MassRobotics, The Intelligent Machine Guild, and Mekable. Check them out! Most automation projects fail. Mekable.com can help. They bring the third party automation expertise that your team needs to start a new project or save a project that is going sideways. Reach out today! #Robotics #AI #MHI #Mekable #intelligentMachineGuild #MassRobotics #RoboticsSummit2026

38

3 Comments

ARM: another capstone project I am incredibly proud of. https://lnkd.in/gERT6BTu This open-source 5-DOF robotic manipulator was designed with AI capabilities while remaining affordable, with a manufacturing cost of just $2,350. It can be programmed in both ROS and Python. It can work standalone or be placed inside the mobile robot, ODD, from my previous post and together, they can work like a rover. The robot has a user-friendly GUI and can be controlled manually by entering the joint values directly (or via the sliders) or be programmed to perform autonomous tasks. They GUI also provides an animation of the robot when being remotely controlled. Its hardware includes a Jetson Orin Nano Super, an RGB-D camera, and a force sensor. It uses advanced Dynamixel servos with joint velocity feedback and is equipped with an E-Stop and Circuit overdraft protection. Sponsored by the robotics lab at Embry-Riddle Aeronautical University Prescott, this robot was built in just two semesters by a team of undergraduate robotics students: William Lindauer, E. Leilani Alvarado, Dakota Jacobs, Hunter C. Smith, Bruce Noble, Lauren Otto, and Joshua Butler. It will begin supporting robotics lab courses starting next academic year.

20

4 Comments

Check out this free Ansys Education Resource that dives deep into the analysis of a transfer chute using Ansys Rocky, the industry-leading particle dynamics simulation tool. https://bit.ly/4bEt3Vu

1