RF‐LSTM‐based method for prediction and diagnosis of fouling in heat exchanger

Abstract

Fouling degrades the thermal and hydraulic performances of the heat exchanger (HE), leading to failure if undetected. It occurs due to the accumulation of undesired material on the heat transfer surface. Knowledge about the HE fouling dynamics is required to plan mitigation strategies, ensuring a sustainable and safe operation. This paper aims to propose a feature‐based technique to predict the fouling status of the HE based on historical data. Three thermal and two hydraulic features are extracted from the HE. Random forest (RF) is employed to detect the dominant features using the Gini index. These dominant features are used to predict the fouling resistance using a deep neural network based on the Long Short‐Term Memory (LSTM) model. Also, these dominant features provide reliable inferences to reason out the fouling dynamics. A diagnostic flag is derived based on the dominant feature and is used to diagnose the ongoing fouling phenomena, which is vital to formulate mitigation. The proposed technique is investigated on the data acquired from the HEs used in a thermal power plant (Case 1) and petroleum refinery (Case 2). Prediction accuracy of 99% and 97% is observed for Cases 1 and 2, respectively. Experimental results illustrate that RF enables the LSTM to achieve faster training and reliable prediction of fouling.

Iterative quality weighted interpolation for LPV‐MPC control of industrial heat exchanger under varying fouling conditions

“…It is determined using the temperature and flow rate profile at the shell and tube sides. In the proposed work, the RF‐LSTM (random forest‐based long‐term short‐term memory) based FR predictor reported in the literature is adopted to estimate the instantaneous FR 27 as in Figure 4. It uses the RF to select the optimal features based on the dominant effect of the HE operations.…”

Artificial Intelligence (AI) Based Predictive Maintenance of Waste Heat Recovery System

“…The underlying root cause of fouling is the impurities in the inlet streams and their affinity for the heat exchanger surface. Unfortunately, the fouling cannot be measured directly in real-time; some of the available handheld devices support offline measurement that requires perforating the equipment ( [5], [6]). Notable studies investigated different heat exchangers and their fouling characteristics.…”

“…The challenge is measuring the fouling thickness on the interior and exterior of the tubes. Real-time measurement is not possible to measure fouling readily; some of the available hand-held devices require dismantling the system, which is expensive, and regular measurement is not possible [6]. In such cases, leveraging thermodynamic models is very relevant for the user to generate system performance at various inlet, surrounding, and fouling conditions.…”

“…cases). Unfortunately, in the actual system, the possibility of collecting broad operational data is difficult because (i) fouling occurs in the long run, (ii) all variables cannot be measured online, especially fouling thickness [6], and (iii) the actual operating range is limited, not wide-ranging. Hence, this study exploits a thermodynamic model to generate complete data under diverse operational conditions.…”

Statistical Model Identification and Variable Selection for Prediction of Heat Exchanger Fouling

“…Te neural network model and a Kalman flter model are compared when applied for fouling detection, and from the result analysis, it is recommended to use the neural model to deal with fast drifts and the Kalman flter model to deal with slow drifts [17]. Fouling degrades the hydraulic and thermal performances of the heat exchanger, and if undetected, it will lead to failure of the system [18]. It has been concluded that the random forest (RF) with Gini index which is used to predict fouling resistance as a deep leaning neural network-based (LSTM, long short term memory) is a biased model [19].…”