Dyad is a modeling and simulation platform. You describe a physical system — a brake, a car suspension, a house losing heat — as a diagram of connected parts. Dyad turns that diagram into equations it can simulate, tune, and fit to real measurements.
This repository holds full-sized demos of that workflow. Each one is a standalone project: open the folder, run it, and read its own README.
These demos use the dyad-3.5.0 Julia distribution (Julia 1.13.2) and target the 3.5.1 Dyad CLI. See the toolchain and validation notes for setup and results.
| Demo | Description | |
|---|---|---|
| FrictionBrakeDemo | A vehicle friction brake system with thermal effects and a cycle test. | |
| TurkeyDemo | A discretized thermal model of a turkey cooking in an oven, with an interactive dashboard. | |
| CoffeeMugDemo | A modular thermal simulation of espresso cooling in a cup, with optional spoon and steam subsystems. | |
| HybridTransmission | A torque-split hybrid transmission model with ECMS control strategy over an EPA highway drive cycle. | |
| MakieWebinar | Demonstrations of using Makie interactive plotting with Dyad models, including dashboards and parameter sweeps. | |
| ThermalHouseDemo | A residential house thermal model covering envelope heat loss, infiltration, solar gains, and HVAC control. | |
| DynamicSteadyState | A three-zone office building thermal model with steady-state HVAC sizing and 24-hour diurnal transient operation. | |
| QuarterTruckSciML | A quarter-truck ride-comfort model showcasing two SciML workflows: training a neural network to recover suspension nonlinearities, and calibrating physical parameters from measurements. | |
| MotorTemperatureSciML | A Thermal Neural Network estimating four motor temperatures on the public Paderborn dataset: neural networks inside a lumped-parameter thermal ODE, trained with stochastic mini-batch multiple shooting and validated free-running against the PyTorch reference. | |
| DrivelineSciML | An electric car's driveline with a rubbery coupling nobody has measured: the demo works out how that coupling behaves from two shake tests, then checks the answer against a hard acceleration it never saw. | |
| PassiveSuspension | A quarter of a car driving over a rough road, where the shock absorbers are tuned until the ride is comfortable enough for the person in the seat. | |
| HL20Demo | An aerodynamic core and design brief for building an HL-20 spaceplane simulation with the Dyad Agent; full-flight analyses remain to be implemented. | |
| SatelliteGNC | A shoebox-sized satellite turning to point at a new target: it works out how fast it is already spinning, then fires eight small thrusters to swing around and hold still, both on its own and while orbiting. | |
| FIRFilterComponents | A clocked finite impulse response filter, showing sampled signals, delay history, and discrete filtering alongside continuous dynamics. | |
| CLibraryIntegration | Six small stages for bringing a C tire library into Julia and Dyad, including derivatives and stateful co-simulation. | |
| CFDReducedOrderModeling | Reduced-order thermal models learned from synthetic CFD snapshots, comparing SVD networks, neural ODE networks, and POD/DMDc. | |
| MultibodyDemo | Four planar multibody simulations: a pendulum, spring–damper, balancing robot, and car with tire slip. | |
| LinearActuatorSystem | A 24 V DC motor drives a lead screw that pushes a carriage back and forth, held at speed by a cascaded speed and current controller with Hall-sensor feedback. |