Application of ANN methodology to optimize the criteria of performance evaluation with multiple v-cut twisted tape inserts

Abstract

This study investigates the impact of v-cut twisted tapes on Double Pipe Heat Exchanger (DP HE) thermo-hydraulic performance to optimize v-cut numbers. Eight tapes with a consistent width-to-depth ratio (1.2) were tested, varying v-cut configurations from one to eight. Heat transfer rate improved by up to 188.54% with eight v-cuts, while the friction factor rose by 244.04%. The highest Criterion of Performance Evaluation (CP E) (1.96) occurred with four v-cuts. Nusselt number (Nu) and friction factor (f) increased with v-cut numbers, but beyond four cuts, friction factor dominance led to decreased CP E. A Multilayer Feed-forward Neural Network (MLFNN) with backpropagation predicted CP E using normalized input–output data. Optimal ANN configuration (3_7_7_1) had a minimal Mean Relative Error (MRE) of 0.95%, Standard Deviation (ST D) of 1.11%, and R2 of 0.99. The model accurately predicted thermo-hydraulic performance, especially for the fourth v-cut, offering valuable insights into DP HE design and performance prediction.