Samuel Jacobs

An NSF-funded project led by Lehigh University Mechanical Engineering researcher Parisa Khodabakhshi aims to streamline machine learning models that incorporate physical laws, facilitating alloy design for high-performance parts in #aerospace, #automotive, and #healthcareindustries. https://lnkd.in/ezGWDe4J

14

Wall-modeled LES helps bring high-fidelity CFD into practical engineering workflows by capturing unsteady flow physics without the cost of wall-resolved LES. This carousel highlights current WMLES applications across aerodynamics, aeroacoustics, combustion, heat transfer, and turbomachinery. For anyone interested in more detail, this technical brief goes a bit deeper into the workflow behind these results: https://ow.ly/ghqI50XTgCL

3

1 Comment

Meso-Fluidics? Is that a thing? Well, here's a instrument I designed and built for a UMASS Lowell Physical chemist way back in about 2004. It's still in use in the physical chemistry classroom. Designed using Solidworks, user interface in LabVIEW, some EE stuff, some optics, some spectroscopy. Kind of a fun project. I called this instrument a 'Continuous Flow Moving Photometer'. It's used to study moderately fast (i.e. ~1s to completion) chemical reactions. It incorporates many principals from the science of micro-fluidics, but at a 'meso' scale. The flow channel is about 4mm diameter. But that's small enough to demonstrate many microfluidics principles.  You can read more about it here:  Bisson, P.; Whitten, J. E. Studying Fast Reactions: Construction and Use of a Low-Cost Continuous-Flow Instrument. J. Chem. Educ. 2006, 83 (12), 1860–1863. https://lnkd.in/eXsiqmhv. Below: The whole 4' long instrument, left. Closeup of the moving photometer, right. The quartz reaction tube is rendered in yellow to aid seeing. 

3

When I started teaching Cornell's mechanical design course two years ago, I wanted to showcase that mechanical engineering isn't just engines and blocks of steel; I wanted to highlight the elegance of good design work while also calling out that mechanical complexity can fit into all sorts of products. That concept turned into sleepless nights putting an assignment together about a goofy dinosaur. Of course, I got attached to said dinosaur and spun the project out into a year-long prototyping endeavor as I tweaked, revisited, and refined. Young engineers often feel limited by rapid prototyping technologies - that they don't teach them about manufacturing and DFX and while I agree to some extent... there was also a lot of DFX built into this 3hr FDM print. Capital "E" Engineering meets lowercase "e" engineering. https://lnkd.in/edR-YAPB #engineeringdesign #mechanicalengineering #opensource #prototyping

52

2 Comments

❌ ❌ Step-by-Step Modeling and Simulation of Pressure Vessel ❌ ❌ 💥Part 2: Material Definition💥 😱 I have created, edited and finally published this tutorial under the bombardments of Israel-US in Tehran. I hope to have the opportunity of creating the next parts of these free tutorials. The internet is highly restricted in Iran by the government (as usual for every crisis 💩) and I should pay near 7 USD for each GB of internet traffic to connect via VPN 🤯 . Some people even pay much more 🤬 👁️The link to the first part is in the comments. In the first video, I have created the geometry based on schematic pictures. The problem is based on Example 9.11 of the API Example Problems Manual. In the second part, elastoplastic material is defined for the pressure vessel based on the data given in Example 9.11. 🎯This tutorial requires no prerequisites❗and is even suitable for engineering bachelor's students ✔️ . You can watch these series of videos as a starting point for learning Structural Integrity via FEA, or to increase your skills and knowledge in this regard. 🎁 We have a lot of tutorials on Structural Integrity and assessment via FEA, based on standards such as ASME Section 8 Division 2, API 579, and BS 7910. Links to them are mentioned in the comments. #abaqus #simulia #simulation #cae #fem #fea #femex #flange #mesh #asme #api #oil #tutorial #discount #structure

17

1 Comment

MATLAB Coding Guidelines: https://lnkd.in/eYfk7_AF I know it's always tempting to write fast and dirty code to put a prototype together, or do some quick testing. But it makes sense to adhere to some standard or guideline if you want to be a professional. Honestly, in most cases, it is you who will later thank yourself for writing well-structured and readable code! My suggestions: - The PDF is a long document, as any good standard should be! Take a quick look and refer to it later if you are in doubt in any practical case. After a while, it becomes natural to you. - Use the MATLAB Code Analyzer to detect and fix the issues quickly.

20

1 Comment

New paper out! And a story about learning (whether the student is a person or a machine!) Three years ago, during my time in Dr. Prahalada Rao 's group at Virginia Tech, we began working hard on an ideal for how vision-based machine learning should be deployed for manufacturing quality control. That ideal is: use physics-inspired features to add morphological context and give AI a true opportunity to learn defect patterns, rather than simply feeding black-box models that memorize visual cues in a mindless way. Teaching a machine, after all, can be a lot like teaching a student: context matters. Pure memorization does not necessarily lead to strong performance outside the narrow constraints of a test. I’m glad we continue to see positive results from this ideal. In this new paper spearheaded by André Ramalho, we enhance our computer vision toolkit in several ways: we expand our work towards WA-DED additive manufacturing and its specific flaws, and we go beyond simple meltpool contour detection by estimating a full "meltpool skeleton", made up of four fiducial coordinates, their Euclidean distances, and a set of derived angles. Ultimately, we show that physics-aware AI can be more consistent and effective than black-box AI. Congratulations to the whole team: André Ramalho, Benjamin Bevans , Fernando Deschamps , Telmo G. Santos , João Pedro Oliveira , Prahalada Rao

30

🚀 New Research on Physics-Informed Operator Learning Efficiently solving partial differential equations has long been a central objective in computational mechanics. Classical numerical methods such as FEM, FDM, and FFT have played a foundational role and continue to evolve. But a simple question got our attention for a while: Can we learn the solution operator itself? That is, can we construct a function that maps any admissible parametric input directly to the solution field, without relying on labeled data from traditional solvers, and scalable to realistic 3D settings with nonlinear behavior? We investigate this promising direction through a physics-informed operator learning framework. Introducing SPiFOL: a Spectral Physics-Informed Finite Operator Learning framework for modeling the mechanical behavior of heterogeneous materials. 🔍 What’s new? SPiFOL builds a physics-informed surrogate by minimizing equilibrium residuals parametrically in Fourier space. This creates a continuous map from microstructure geometry to its mechanical response, bypassing the need for automatic differentiation or supervision from conventional solvers. Once trained, SPiFOL can predict full-field strain or stress responses across 2D/3D microstructures under arbitrary loading. 📊 Highlights: - No labeled data from FEM or other solvers required - Residuals of governing equations are minimized parametrically over Fourier-based microstructure samples - Efficient gradient construction via the Lippmann–Schwinger operator - FNO-enhanced for better generalization and zero-shot super-resolution - maximum error≤2%; average error<0.3% (see the discussions on out-of-distribution samples as well, where the errors get higher but still acceptable) - Extended to finite elasticity and 3D domains 🧩 Challenges & Outlook: SPiFOL’s zero-shot super-resolution (ZSSR) enables evaluation at finer resolutions than the training scale. While high phase contrast can increase maximum error, SPiFOL still outperforms interpolation and baseline FNOs, particularly in physically critical zones. Future improvements include meta-learning, attention-guided adaptation, and multi-resolution training to boost robustness and generalization. This is a fantastic collaboration between Access e.V. and RWTH Aachen University, led by ALI HARANDI. If you're interested in master's thesis topics on operator learning, surrogate modeling, or real-time mechanics, feel free to get in touch! 📎 For more scientific details and even code, see: https://lnkd.in/ebTCmtM4 https://lnkd.in/ejn_gjMD #OperatorLearning #PhysicsInformedAI #ComputationalMechanics #FourierMethods #DeepLearning #MaterialScience #SurrogateModeling #DigitalTwins #PDEs #ZeroShotLearning #FNO #microstrcutre

191

3 Comments

Jeff ERNO got us in deep thoughts today with his post. What if the possibility of AI includes the ability to learn from the path engineers took to get to the final design? Even looking at system iterations or rejected concepts. AI might begin to have its own design intent 🤯 What do you all think? #AI #3dexperience #designthinking #designintent

6

3 Comments

Check out our recent paper on radiation shielding of tungsten borides-epoxy composites just published in Composite Science and Technology in collaboration with Dr. Jessika V. Rojas, Andre Furkan Erdogan and Santiago Bermudez. We demonstrate that hard tungsten borides combined with epoxy provide a range of materials for various radiation shielding, with tungsten tetraboride composites being very effective for neutron attenuation and tungsten monoboride composites for blocking gamma-rays. These composites can form the next generation of shielding materials for nuclear and space applications. https://lnkd.in/eKgzWhPJ

21

3 Comments

Have you ever needed to use an unsightly variable names like DeltaF_DoublePrime in EngineeringPaper.xyz because of limitations in allowed variable names? No more, EP now supports variable accents (vector arrow, prime, dot, hat, etc.) and international characters in variable names! See examples of what is possible in the screenshot below: #Engineering #epxyz #EngineeringCalculator #EngineeringEducation #MechanicalEngineering #CivilEngineering #ElectricalEngineering #ChemicalEngineering

26