Detection of fouling in a cross-flow heat exchanger using a neural network based technique

This paper presents a method for the detection of fouling in a cross-flow heat exchanger. A numerical model is used to generate data when the heat exchanger is clean and corresponding data when fouling occurs. In a first step, the model is used to generate a long time series by simulating a clean heat exchanger. This allows the determination of a neural network model of the heat exchanger. Then, hundred sets of data are generated by simulating a fouled heat exchanger and it is checked that the simple Cusum test can be used to detect fouling without any false alarm, whatever the reference time series is.