»
» » »
Fouling
to the linear regression model. Additionally, its straightforward approach helps prevent overfitting, a common issue with more complex models such as decision trees and boosted decision trees.
Fouling prediction- Results validation After selecting the best regression model, the next step is predicting when the heat exchanger requires cleaning. In figures 6 and 7, red line represents the minimum overall heat transfer coefficient( UA) necessary for both fouled and clean heat exchanger to deliver the specified heat transfer. When the UA reaches this threshold, it indicates that the heat exchanger requires cleaning. The model predicts UA for the next three months, with the trendline shown by the dotted line. The intersection with the minimum UA line suggests cleaning around January 2022 for clean heat exchangers and February 2021 for fouled ones. Accuracy is confirmed by internal inspection reports. Figure 8 depicts the condition of E-21-02 and E-26-02 during the turnaround in June 2021. This observation aligns with the model’ s prediction, which recommended cleaning around January 2022 for clean heat exchangers and February 2021 for fouled heat exchangers. The implementation of digital twins in cooling water heat exchangers demonstrates significant potential for operational efficiency. Key benefits include: 1. Improved decision-making: Real-time data allows for planning cleaning period effectively. 2. Predictive maintenance: Detection of potential fouling reduces unplanned downtimes and maintenance costs.
Conclusion The implementation of digital twins in heat exchanger maintenance offers a compelling case for industry adoption. By leveraging real-time data and predictive capabilities, businesses can enhance performance and reliability while achieving substantial cost savings. However, several challenges must be addressed to achieve accurate fouling
Figure 8. Pre-cleaned clean heat exchanger.
Figure 6. Prediction results for clean heat exchanger.
Figure 7. Prediction results for fouled heat exchanger.
prediction results. These includes PI data accuracy and quality, HTRI design model, digital twin system integration, etc. Future research should focus on refining predictive algorithms and leveraging artificial intelligence to uncover further optimization opportunities. The team planned to continuously improve data accuracy, enhance user interface, and expand fouling prediction capabilities to other types of heat exchanger. Team appreciation: Upenthiran Ganeson, M Syahir Mohamad Zaki
Figure 9. Pre-cleaned fouled heat exchanger.
References:
1. Tubular Exchanger Manufacturers Association( TEMA)( 2023). TEMA standards for tubular heat exchangers( 11th Edition). 2. Experimental / Numerical Investigation and Prediction of Fouling in Multiphase Flow Heat Exchangers: A Review.
3. Heat Transfer Research Inc. Software.
4. Microsoft Azure Machine Leaning Software.
5. ICON Symmetry Process Simulation Software. www. heat-exchanger-world. com Heat Exchanger World July 2025
49