Numerical Prediction of Experimentally Obtained Thermal Performance and Capacity of EC Fan Coil with Artificial Neural Networks During Cooling Process


Uğuz B., Gemici Z., Tonya D., Çolak A. B., Dalkiliç A. S.

ASES VIII. INTERNATIONAL HEALTH, ENGINEERING AND SCIENCES CONFERENCE, İzmir, Turkey, 6 - 07 April 2024, pp.43-44

  • Publication Type: Conference Paper / Summary Text
  • City: İzmir
  • Country: Turkey
  • Page Numbers: pp.43-44
  • Yıldız Technical University Affiliated: Yes

Abstract

Fan coil devices are one of the most widely used heating, ventilation, and air conditioning (HVAC) system devices for heating and cooling interior areas. The selection of fan coil devices and determining the actual operating conditions according to the needs are of great importance in terms of indoor comfort and energy savings. In this regard, studies on the prediction of fan coil units' performance outputs at different operating points are not sufficient. In this study, two distinct artificial neural networks (ANNs) were used to numerically predict the experimentally obtained thermal performance and capacity of a concealed ceiling-type fan coil with no cabinet. The experiments were carried out in a test device designed in accordance with the AMCA 210 test standard and in an indoor air and heat exchanger fluid regime in accordance with Eurovent test norms. In the first ANN, important performance output factors of the fan coil unit, such as air outlet dry bulb temperature, total and sensible cooling capacities and consumed electrical power, were estimated by an artificial neural network method using 1700 test points. The effect of eight different input factors, such as the number of heat exchanger tubes, number of rows, finned length, number of circuits, EC motor voltage, fan external static pressure, air flow rate, and water flow rate, on the outputs was analyzed. Levenberg Marquardt training algorithm was used in the network model with 10 neurons in the hidden layer. In the second ANN, in order to predict the heat exchanger fluid side pressure loss as the output, a separate ANN was performed using the input factors of heat exchanger pipe number, number of rows, finned length, number of circuits, and water flow rate. While deviations were obtained as 0.045%, - 0,014%, -0.01% and 0,283% for air outlet dry bulb temperature, total and sensible cooling capacities, and consumed electrical power, respectively, in the first ANN, the heat exchanger fluid side pressure loss’s deviation was obtained as -0.014% in the second developed ANN.