The novel concept to detect cooling coil fouling automatically with an embedded datadriven agent using expert signature maps, introduced in a companion paper (Veronica 2010), is exercised here on data from computer simulations of clean and fouled cooling coils. The companion paper describes in detail a crucial element of its fouling detection concept: that responses from an adaptive, predictive datadriven model of the cooling coil sum to an appropriate value when the datadriven model is queried offline from the realtime data acquisition. This query consists of a timesequence of input vectors, passed through the model as a triangular pulse of water flow. This report shows rudimental success, a foundation for further research, is obtained using a relatively simple, threenode multilayered perceptron as the datadriven coil model. Also discussed here are factors critical to successfully employing a datadriven model of any architecture to detect coil fouling.
Citation: International Journal of Heating, Ventilating, Air-conditioning and Refrigerating Research
Pub Type: Journals
automatic, fault detection, FDD, fouling, coil, heat exchanger, neural network, data-driven, machine learning