Abstract:Pulsating heat transfer is important in many engineering applications but is not fully understood or easily modeled. This work helps understand the phenomenon and demonstrates the efficacy and utility of deep learning modeling for pulsating heat transfer. The convective heat transfer coefficient (HTC) of pulsating flow through a copper tube was experimentally investigated at seventy-seven different operating conditions (500,000 time-series data points were collected at each operating condition) corresponding to different pulsating waveforms, frequencies, mass flow rates and pressures. The intensity of pulsation ranged from mild to strong enough to cause reverse flow. The literature on pulsating HTC is mixed, with several researchers reporting increased heat transfer and several other reporting decreased heat transfer relative to steady flow. The experimental data showed that both could be true depending on the ratio of the amplitude to mean flow velocity (amplitude ratio AR ). The enhancement ratio ( ER ) of pulsating HTC relative to steady-flow HTC was measured to vary between 0.48 and 2.18 for AR ranging from 0.64 to14.42. A numerical study showed that this was only a (positive) correlation; fundamentally the overall pulsating HTC could be approximately explained by the variation of instantaneous Reynold's number ( Re ) during the pulsating cycle. The distribution of instantaneous Re and hence the ER is primarily determined by the pulsating waveform, and this explains the disparate observations by different researchers. Long Short-Term Memory (LSTM) networks have been used to model pulsating HTC as a function of pulsating waveforms occuring across a measurement orifice. The 'Toy Model' concept where physics-based variables are used to increase generalization was implemented with the LSTMs for increased model robustness. These models could be used to maximize pulsating heat transfer by optimizing the waveform.

Prediction of Drag Reduction Effect by Machine Learning toward Optimization of Pulsating Control in Turbulent Pipe Flow

Numerical Analysis of the Drag Reduction for Turbulent Pulsating Pipe Flows Based on Large Eddy Simulations

Active Learning for Modeling and Prediction of Dynamical Fluid Processes

Probabilistic Modeling and Prediction of Dynamic Discharge Process in Multiphase Pumps

Machine Learning Based Developing Flow Control Technique Over Circular Cylinders

Experimental, numerical and deep learning modeling study of heat transfer in turbulent pulsating pipe flow

Machine-learning flow control with few sensor feedback and measurement noise

Experimental research on drag reduction by polymer additives in a turbulent pipe flow

Closed-loop plasma flow control of a turbulent cylinder wake flow using machine learning at Reynolds number of 28 000

Drag prediction of rough-wall turbulent flow using data-driven regression

Prediction accuracy improvement of pressure pulsation signals of reversible pump‐turbine: A LSTM and VMD‐based optimization approach

Machine-learning data-driven modeling of laminar-turbulent transition in compressor cascade

A combined active control method of restricted nonlinear model and machine learning technology for drag reduction in turbulent channel flow

Transition Modeling for Low Pressure Turbines Using Computational Fluid Dynamics Driven Machine Learning

Research into Prediction Method for Pressure Pulsations in a Centrifugal Pump Based on Variational Mode Decomposition-Particle Swarm Optimization and Hybrid Deep Learning Models

Synthetic turbulent inflow generator using machine learning

Influence of nonlinear pulsating characteristic on the simulation of the load rejection process in a pump-turbine

Influence of turbulence model on performance prediction of mixed-flow pump

A Predictive Model for Steady-State Multiphase Pipe Flow: Machine Learning on Lab Data

DIRECT NUMERICAL SIMULATION OF TURBULENT PIPE FLOW CONTROLLED BY MHD FOR DRAG REDUCTION

Experimental investigation of flows in a high-head pump-turbine draft tube at turbine mode