Development of a hybrid model for reliably predicting the thermal performance of direct contact countercurrent cooling towers. (15th November 2022)

• Record Type: Journal Article
• Authors: Jayaweera, Chamanthi; Groot, Niels; Meul, Steven; Verliefde, Arne; Nopens, Ingmar; Hitsov, Ivaylo
• Is Part Of: International journal of heat and mass transfer. Volume 197(2022)
• Publication Date: 2022-11-15
• Subjects: Cooling tower -- Hybrid draft -- Merkel number -- Hybrid modelling; Heat -- Transmission -- Periodicals; Mass transfer -- Periodicals; Chaleur -- Transmission -- Périodiques; Transfert de masse -- Périodiques; Electronic journals; 621.4022
• DOI: 10.1016/j.ijheatmasstransfer.2022.123336
• Languages: English
• ISSNs: 0017-9310

Abstract

Highlights: A hybrid model for predicting outlet conditions of a cooling tower was developed. Calculation of air flow resistance in a hybrid draft cooling tower was simplified. Use of neural networks to account for the volumetric mass transfer coefficient. Abstract: Cooling towers are a primary and vital component of the cooling cycle in a chemical plant. The heat and mass transfer inside a cooling tower is governed by the complex geometry of the fill, flow conditions and turbulence degree of the air and liquid phases. These aspects of a cooling tower are generally modelled using computational fluid dynamics, which is time consuming and computationally intensive. The current study illustrates a methodology for modeling the heat and mass transfer of a full-scale hybrid-draft cooling tower and an induced draft cooling tower located in two separate plants. A convenient method was devised to account for the resistance encountered by the air stream passing through a full-scale hybrid-draft cooling tower. The devised method is notably simpler than engaging computational fluid dynamics. The model constitutes a mechanistic component that computes the variation of liquid and air properties along the height of the tower. The variation of the mass transfer coefficient with the contact area between the liquid and gas phases was predicted using an artificial neural network. As the model utilizes a mechanistic component developed based on heat and mass transfer principles, that borrows values for the mass transfer coefficient predicted by a neural network to simulate liquid and air properties, the entirety functions as a hybrid model. The developed neural network predicted the mass transfer coefficient with an R 2 > 0.94 for both cases. The overall model demonstrated a prediction accuracy ( R 2 ) of 0.99 in the year-round thermal performance of both towers. Therefore, the high prediction accuracy and simplicity of the model enables applications in real time monitoring of the thermal performance and optimization of operational parameters.