HVAC
NABLA
FARM
Livestock building climate,
designed rather than guessed
CLIENT
AGCO Corporation
CHALLENGE
Animal health and productivity depend on conditions nobody can see: temperature, humidity, air quality, ventilation
RESULT
A parametric model that became the training set for a neural network now controlling the HVAC in real time
01
CHALLENGE
Whether the farm holds chickens, pigs, horses or anything else, productivity follows the environment the animals live in. Temperature, humidity, air quality and ventilation are the levers, and on a real building they interact in ways intuition gets wrong.
The design problem has four constraints that pull against each other: heat and moisture management, ventilation requirements, biosecurity, and energy efficiency. Improve one by guesswork and another usually pays for it.
02
SOLUTION
NablaFARM is a parametric model of the building and of what happens inside it. The geometry of the building is parametrised — its dimensions, the fans and where they sit, the air inlets and outlets, the heating and cooling systems, the circulating fans, the outside conditions — and so are the animals: their position, their number and what each group emits, as a heat source and a source of CO2.
That means a configuration is not drawn, it is specified. Change a fan position or a stocking density and the model regenerates, so one ventilation layout can be compared with another before either is built.
03
RESULTS
A run takes from a few minutes to several hours depending on the hardware. What comes back: contour plots of temperature, velocity and pressure; streamlines; humidity levels; concentration of species such as CO2 and N2O; and the air recirculation rate — collected into an automatic report.
Then the model was put to a use it was not built for. Because it is parametric, it can be run hundreds of times over: a parametric model is a generator of synthetic data, which AGCO used to train a neural network.
That network was coupled with the sensors already in their buildings, so it predicts how the internal environment will evolve rather than reporting where it is now — and it actively controls the HVAC devices, heading off a problem instead of responding to one.
Applications: poultry houses, swine barns and horse stables; intensive farms.
