NEURAL NETWORK ANALOGY OVER NUMERICAL ANALYSIS ON THERMOHYDRAULIC PERFORMANCE FACTORS j AND f CORRELATIONS DEVELOPMENT FOR COMPACT HEAT EXCHANGERS

Abstract

A compact heat exchanger is a component designed to transfer heat energy between two fluids either mixing or separated by a solid wall, which is a vital role for efficient energy transfer. The design and optimization for a low pressure drop and highly efficient heat transfer is a challenging endeavor. Development of thermo-hydraulic performance factors are typically achieved through experimental or numerical analysis. Correlations for the performance of fins, whether developed through experimental or numerical methods, are commonly presented in the form of dimensionless groups. These groups include the Colburn factor '<i>j</i>', Fanning friction factor '<i>f</i>', the Reynolds number, Nusselt number, and various geometric parameters, as found in the open literature. In this study, the plate fin model is utilized in the numerical analysis to address the governing equations and simulate the flow and heat transfer phenomena. The performance of the fin is evaluated by examining varying Reynolds numbers and geometric parameters for the generation of '<i>j</i>' and '<i>f</i>' correlations. A total of 144 fin geometric parameters were used in the numerical model to develop correlations. The numerical model is analyzed using Ansys Fluent®. Numerical analysis, however, is computationally intensive and may necessitate prior knowledge in computational techniques and expertise in physics. This tremendous process of correlation development is expedited by employing the use of artificial neural networks, which can prove to be especially advantageous when the physics of the system is poorly understood or difficult to model numerically. This paper focuses on developing design data requirements for rectangular plain fin compact heat exchanger using neural networks and computational fluid dynamics. The performance correlations are verified and validated using the open literature. Development of algorithms through the combined use of neural network and computational fluid dynamics can open a gateway to innovate new fin shapes or surfaces yielding higher efficiency for heat energy transfer and thereby more optimized designs for compact heat exchangers.