Comparison of neural networks and Kalman filters performances for fouling detection in a heat exchanger

This paper presents the comparison between a neural network model and a Kalman filter model when applied for fouling detection. The models are determined using data that do not require the heat exchanger to be in a steady state. These data are the inlet and outlet temperatures and the mass flow rates. It monitored how the difference between estimated values and actual values evolve with time. This difference is computed for predictions (one-step ahead) or simulations (the whole set of data is computed using past estimated values). The evolution is due to fouling (the fouling scenario is given in terms of a fouling factor). The detection of the drift is carried out using the Cusum test. It is shown that fouling is detected quite early. By the analysis of the results, it is recommended to use the neural network model when dealing with fast drifts, and to use the Kalman filter model when dealing with slow drifts.