Hi, I'm Dawar Azhar, a Mechanical Engineer (MS in Manufacturing) with a deep passion for software development, AI/ML, and computer vision. This unique blend of traditional engineering and modern technology drives my exploration at the intersection of physical systems and digital innovation.
Here you'll find insights on how artificial intelligence and computer vision are transforming manufacturing, deep dives into computational engineering (CFD/FEA), practical implementations of machine learning algorithms, and my journey bridging the mechanical and software worlds. Whether you're interested in Industry 4.0, intelligent automation, or the latest in AI research, I hope you'll find something valuable here.
Just finished debugging a simulation that was giving weird pressure gradients. Turned out the boundary layer mesh had aspect ratios exceeding 100:1 in some regions.
Quick rule: Keep aspect ratios under 3:1 for general flows, or max 10:1 for well-behaved boundary layers. Your solver (and your sanity) will thank you.
Sometimes the basics matter more than fancy turbulence models. 🔧
TIL: ParaView’s Calculator filter can do way more than basic arithmetic. You can compute vorticity, Q-criterion, and custom flow metrics on the fly.
Just used it to calculate directly from velocity gradients without exporting to MATLAB.
The syntax is: 0.5*(Vorticity_0*Vorticity_0 + Vorticity_1*Vorticity_1 + Vorticity_2*Vorticity_2 - strain_rate)
Game changer for flow visualization! 🌊
Been reviewing fundamentals for a new project and rediscovered how elegant the Reynolds Transport Theorem really is:
It’s literally the bridge between Lagrangian (following particles) and Eulerian (fixed control volume) descriptions of flow.
Every conservation equation in fluid mechanics stems from this beauty. Yet most courses rush through it to get to Navier-Stokes. We should appreciate the fundamentals more! ⚡
Working on a tricky multiphase flow that kept diverging after 500 iterations. Instead of tweaking the mesh again, tried gradually reducing under-relaxation factors:
Simulation converged beautifully! Sometimes patience beats brute force. The physics doesn’t care about our deadlines. 🕐
Pro tip: Monitor residuals AND physical quantities. Residuals can lie, but mass flow rates don’t.
Watching the rapid evolution of AI-assisted engineering tools. Fusion 360’s generative design is impressive, but I’m more excited about AI for simulation setup.
Imagine describing a flow problem in natural language: “Model airflow around a car at highway speeds with crosswind” and having the AI set up geometry, mesh, boundary conditions, and solver settings.
We’re probably 2-3 years away from this being reliable. The challenge isn’t the AI—it’s encoding decades of engineering judgment into training data.
Still skeptical about AI replacing engineers, but excited about AI amplifying engineering productivity. 🤖