Design optimization of an heat exchanger using Gaussian process

Mehta

Food Processing Preservation

et al. 2021

This study was aimed to explore the effect of ultrasonic treatment on osmotic dehydration of aonla slices. The Box–Behnken design (BBD) was utilized to design the experiment to study the effect of three process variables: ultrasonication time of 10–30 min, ultrasonication temperature of 30–50°C, and sugar concentration of 40–60°Bx. Artificial neural network (ANN), Gaussian process regression (GPR), and response surface methodology (RSM) (in terms of a second‐order polynomial) were used to predict the mass transfer parameters (water loss, solid gain, and weight reduction), shrinkage ratio and color changes during osmotic dehydration of aonla slices. The predictive capabilities of the ANN, GPR, and RSM were compared by means of mean square error (RMSE), mean absolute error (MAE), average absolute deviation (AAD), standard error of prediction (SEP), chi‐square statistic and coefficient of determination (R2). Results showed that the ANN model was found to be a more accurate predictor than the other models for all responses except shrinkage ratio where the GPR model was found to be better in predicting model. RSM also suggested the optimum processing conditions leading to maximizing water loss and weight reduction and minimizing solid gain to be at ultrasonication time of 10 min, ultrasonication temperature of 30°C, and sugar concentration of 60°Bx. Novelty impact statement Ultrasonic pretreated Aonla slices were found to have improved mass transfer dynamics during osmotic dehydration. Integrated Process modeling was explored and ANN was found to be a better process model compared to GPR and RSM. RSM predicted the optimum process parameters such as pretreatment time: 10 min, temperature: 30°C and sugar concentration: 60°Bx.

Section: Predictive Modeling and Optimization Methodsmentioning

confidence: 99%

Ultrasonic modulated osmotic dehydration of Aonla ( Phyllanthus emblica L.) slices: An integrated modeling through ANN, GPR, and RSM

Mehta

Food Processing Preservation

et al. 2021

“…The purpose of the third to sixth design is to study the effect of the rib shape on the flow separation behavior, which can further affect the pressure drop and heat transfer. The sharp edge rib geometry is selected based on the optimization work by Campet et al 39 Figure 3 also shows the cross‐section of the geometries under investigation, where the rib height‐to‐diameter ratio is 0.036, and the pitch‐to‐height ratio is 11.67. For a clear comparison between the actual coil and the designed ribs, the transverse view of bare tube and two MERT designs by Kubota 40,41 are shown in Figure 4.…”

Section: Numerical Setupmentioning

confidence: 99%

Wall‐resolved large eddy simulation of turbulent flows in helically ribbed steam cracker reactors

2022

Helically ribbed coils are commonly applied in steam cracking furnaces. To fully understand the impact of these ribbed wall modifications on the local heat transfer and associated pressure drop throughout the reactor, detailed experimental, and numerical studies have been performed. Experimental data based on stereo-particle image velocimetry (S-PIV) and liquid crystal thermography have been used to validate the numerical results from wall-resolved large eddy simulations using OpenFOAM.The validation shows an excellent agreement in terms of mean and fluctuating velocities, pressure drop, and heat transfer behavior in a discontinuously ribbed tube. Compared with the pressure drop in a continuously ribbed tube, an approximately 40% lower pressure drop is obtained with a discontinuously ribbed tube, at the cost of a slightly decreased heat transfer enhancement. This makes the discontinuously ribbed tube design particularly interesting for steam cracking applications. The results also show that the nonuniform heat transfer at the wall is inherently linked to the flow reattachment and recirculation zones caused by the rib. Finally, the validated numerical model was used to study comparable designs and propose novel helical rib designs. Based on the results of the study, enlarging the trailing edge of the conventional ribbed geometry will improve the thermal enhancement performance, and is therefore found most promising for steam cracking reactor design.

“…This approach is similar to Bayesian optimisation, where an acquisition function (also referred to as the infill sampling criterion) is used to select the new query points. Numerous works have used this technique using Gaussian processes (GPs), which was found effective when coupled to high-fidelity computational fluid dynamics (CFD) simulations, for which query points are expensive to compute (Roy et al 2018;Campet et al 2020). Active learning has been applied for optimisation in various scientific fields, from biology (Pandi et al 2022) to materials science (Wang et al 2022), including fluid mechanics to optimise turbine shapes (Wang et al 2022).…”

Section: Introductionmentioning

confidence: 99%

Discovering optimal flapping wing kinematics using active deep learning

Corban,

Bauerheim,

Jardin

2023

J. Fluid Mech.

This paper focuses on the discovery of optimal flapping wing kinematics using a deep learning surrogate model for unsteady aerodynamics and multi-objective optimisation. First, a surrogate model of the unsteady forces experienced by a 3-D flapping wing is built, based on deep neural networks. The model is trained on a dataset of randomly generated kinematics simulated using direct numerical simulation (DNS). Once trained, the neural networks can quickly predict the unsteady lift and torques experienced by the wing, using sparse information on the kinematics. This fast surrogate model allows multi-objective optimisation to be performed. The resulting Pareto front consists of new kinematics that may be very different from the kinematics of the initial dataset. A few arbitrarily chosen kinematics on the Pareto front are thus simulated using DNS and used to enhance the database. The new dataset is used to train again the networks, and this active deep learning/optimisation framework is performed until convergence, obtained after only two iterations. Overall, this method reduced the cost of optimisation by 83 %. Results reveal two distinct families of motions. Kinematics promoting high efficiency are characterised by large stroke amplitudes and relatively low angles of attack, as observed for fruit flies, honeybees or hawkmoths. For those, lift production is driven by quasi-steady effects and the formation of a stable leading edge vortex. Kinematics promoting high lift are characterised by small stroke amplitudes and high angles of attack, reminiscent of mosquitoes. Lift production is driven by the rapid generation of vorticity at the trailing edge.