Comparison of Fouling Detection between a Physical Method and a Black Box Model

The paper presents two methods that are aimed at detecting fouling in cross-flow heat exchangers under normal operating conditions, using measurements that can easily be obtained, i.e. inlet and outlet temperatures and the mass flows of the hot and cold fluids. The methods are a) a physical method that is based on number of transfer units (NTU) relations and b) a black box model that is estimated using a clean heat exchanger. In method a) the overall heat transfer coefficient is estimated from the NTU relations. It is known that the heat transfer coefficients is mass flow dependent and it is therefore convenient to use empirical relations to filter variations caused by mass flow changes from the estimated values. In method b) a model of the heat exchanger is identified, based on past and present measurements from the heat exchanger. The model is then used to make one step predictions of the outputs of the heat exchanger. Currently the methods have only been tested on simulated data from a Computational fluid dynamic (CFD) model. Although the CFD model has not been validated for a cross-flow heat exchanger it has been validated for a counter-flow heat exchanger with good results. It is concluded that it is possible to detect fouling in cross-flow heat exchangers with the derived methods, within a reasonable accuracy and consistency. INTRODUCTION Heat exchangers are commonly used in domestic and industrial applications involving transfer of energy from one fluid to another. General classifications of heat exchangers are parallel flow, counter-flow and cross-flow. Their operating conditions can be put into two main classes, steady state operation where mass flow and temperatures are relatively constant and dynamic operation where mass flow and temperatures can vary greatly with time. In most cases heat exchangers accumulate fouling on the heat transfer area during usage. As the fouling increases, the efficiency of the heat exchanger decreases. Because of decreasing efficiency it is important to have some knowledge of the condition of the heat exchange. Since fouling in heat exchangers is very common, heat exchangers are commonly designed to withstand mild fouling, see (Pope et al., 1978). It has also been shown that surface coating and heat transfer enhancements that intensify heat transfer can increase the fouling induction time, see (Nejim et al., 1999; Liu et al., 2009; Somerscales and Bergles, 1997). Fouling can be described as a process where the separating metal inside the heat exchanger gathers deposits from the fluids. As fouling accumulates on the heat transfer area the resistance to heat transfer increases and the effectiveness of the heat exchanger diminishes. The cost of fouling is a waste of energy and cost due to downtime because of cleaning procedures and additionally due to product faults. These costs can become significant, and also the waste of energy which is an environmental issue that should be taken seriously. If not only because of the high cost associated with fouling, it is important to have some knowledge of the condition of the heat exchanger. If the condition is known, fouling mitigation techniques may become more focused and cleaning of the heat exchanger can be scheduled at optimum intervals. There are number of ways to detect fouling but many of the classical methods require the process to be operating in a steady state condition or to be stopped. These restrictions can be too strict or costly. If the steady condition can be met the detection has been proven to be relatively simple, since analytical and empirical relations can be derived for different heat exchanger types and used for all necessary calculations regarding time invariant conditions, see e.g. (Holman, 2002). In a dynamic operation, it becomes more complicated to monitor the condition of the heat exchanger and more complex models are used, see (Mishra et al., 2008) where finite difference method is used to model a cross flow heat exchanger. It is important that a fouling detection method can follow the process in real time, for it to become useful. The detection can be done for example by monitoring possible drift in parameters or variables in a model. Then the discrepancies between the model predictions and what is actually measured will indicate fouling. Fouling detection has been an active field for the last years and many methods have been developed. Examples of methods that monitor parameters for drift using standard and extended Kalman filter are presented in (Gudmundsson et al., 2009a) and (Gudmundsson et al., 2009b) and also in (Delmotte et al., 2008) where Fuzzy models are used. An alternative example of a method monitoring drift in variables can be Proceedings of International Conference on Heat Exchanger Fouling and Cleaning 2011 (Peer-reviewed) June 05 10, 2011, Crete Island, Greece Editors: M.R. Malayeri, A.P. Watkinson and H. Müller-Steinhagen Published online www.heatexchanger-fouling.com 391 H Müller-Steinh gen and A.P. Watkinso