Designing the floor heating system in a 3 dimensional room by using of neural network

Abstract

Since many parameters effect on the floor heating System<br><br>designs, so system Simulation and design by using of dominant<br><br>equation through numerical methods and achieving desirable<br><br>results needs much time and expense.<br><br>In this Study, the neural network used as the floor heating<br><br>system which we can use it as a new method to design a<br><br>quicker and Simpler floor heating system. For this purpose,<br><br>first, the model of System which was a 3 dimensional room<br><br>with different boundary Conditions Stimulated by Gambit &<br><br>Fluent Soft wares. Considering, the inner air of the room<br><br>incompressibility, boussinesq model, reformed model of k −ε,<br><br>radioactive model DO and simple algorithm, the stimulation of<br><br>fluid flow and temperature distribution performed. After<br><br>change of effective various parameters on the floor heating<br><br>system in the model (room) such as pipe depth, water<br><br>temperature, internal emissivity, air absorption Coefficient,<br><br>window sizes, object size(object was in room) , distance<br><br>between object and floor are used as educational data for neural<br><br>network. The neural network produced by two layers: hidden<br><br>layer, with 20 neural neurons and arbitrary function, outer<br><br>layer, with 1 neural neuron and liner Function, in the Matlab<br><br>software with 6 inputs and 1 output.<br><br>Comparing the achieved results with the Corresponding<br><br>results by numerical methods which done by the new data<br><br>showed that The difference between the results was very<br><br>insignificant So the neural network Can be used as a Simpler<br><br>and quicker design for floor heating system of rooms.