| Abstract: |
Heat exchangers are central to energy recovery in process industries, yet they are frequently operated at conservative, fixed set-points that ignore fouling and changing inlet conditions. This empirical study investigates whether a data-driven optimization framework can improve the energy performance of a counter-flow heat exchanger by tuning its operating parameters. A dataset of 1,500 operating records, generated from an effectiveness–NTU thermo-hydraulic model with 1% Gaussian measurement noise, covered hot- and cold-side flow rates, hot and cold inlet temperatures, and fouling resistance. Five surrogate learners were compared, and the best was coupled with a genetic algorithm and particle swarm optimization to maximize exergy efficiency subject to a minimum heat duty of 400 kW. Support vector regression gave the most accurate surrogate (R² = 0.994 for exergy efficiency on unseen data). Across clean, moderately fouled and severely fouled conditions, the optimized set-points raised exergy efficiency by 7.6%, 12.1% and 16.7%, cut pumping power by 73.1%, 67.3% and 51.4%, and reduced exergy destruction by up to 24%, while still delivering at least 405 kW. The two optimizers converged to practically identical solutions, within 1% of an exhaustive-search reference. The benefit grew with fouling severity, showing that adaptive set-points matter most when the unit is degraded. Pumping-energy savings of about 20 MWh per year were estimated for a single unit. Lower flow rates combined with a higher hot-inlet temperature, rather than customary high-flow practice, gave the best energy performance within the studied envelope. The findings should be confirmed on plant measurements before deployment. |
