Search

Deep reinforcement learning (DRL) is a promising outer-loop intelligence paradigm which can deploy problem solving strategies for complex tasks. Consequently, DRL has been utilized for several scientific applications, specifically in cases where classical optimization or control methods are limited. One key limitation of conventional DRL methods is their episode-hungry nature which proves to be a bottleneck for tasks which involve costly evaluations of a numerical forward model. In this article, we address this limitation of DRL by introducing a controlled transfer learning framework that leverages a multi-fidelity simulation setting. Our strategy is deployed for an airfoil shape optimization problem at high Reynolds numbers, where our framework can learn an optimal policy for generating efficient airfoil shapes by gathering knowledge from multi-fidelity environments and reduces computational costs by over 30\%. Furthermore, our formulation promotes policy exploration and generalization to new environments, thereby preventing over-fitting to data from solely one fidelity. Our results demonstrate this framework's applicability to other scientific DRL scenarios where multi-fidelity environments can be used for policy learning.

Direct simulations are carried out to investigate the influence of unsteady heat flux transfer on transonic shock-boundary layer interaction; for flow past SHM-1 airfoil at a free-stream Mach number \(M_{\infty }\) = 0.72 and angle of attack \(\alpha = 0.38^{\circ }\). Flux is added in a periodic manner through a region \((8{-}18\% \; of \;the \;chord)\) located on the suction side of the airfoil, with multiple values of exciter time period \((T_{\text {e}}=2,4)\) considered for our simulation. We show that addition of unsteady heat flux delayed shock formation, along with significant modifications in it’s structure. The time-averaged \(C_{\text {p}}\) distributions revealed a shift in the shock towards the aft, by approximately 5% of the chord; along with an increased lift near the trailing edge, suggesting a nose-down stabilizing influence. Primarily, it is noted that the additional heat flux resulted in an overall increase of the aerodynamic efficiency (lift to drag ratio) by approximately \(10\%\).

Various types of faults can occur in an air conditioner resulting in a decrease in efficiency, a rise in energy consumption, and increasing maintenance costs. Hence predictive maintenance becomes important. In this paper, the two most common types of faults – gas leakage and capacitor malfunction have been detected using the decision tree machine learning algorithm. The data for faulty and operating air conditioners have been collected using distributed sensors, microcontroller, and dedicated circuitry and analyzed using MATLAB Classification App Learner Toolbox. The results obtained by the decision tree for fault detection and diagnosis and load monitoring were then compared with results obtained by support vector machine and the prediction accuracy for the decision tree was found to be higher. The presented research work can identify the air conditioner which is faulty as well as predicts the type of fault at an early stage to do maintenance beforehand.

Implicit large eddy simulations are carried out to investigate the influence of the molecular bulk viscosity on transonic shock-wave boundary layer interaction for flow past a natural laminar flow airfoil at a free-stream Mach number M∞ = 0.72 and an angle of attack α = 0.38°. To quantify the nonlinear interactions, we have discarded the putative assumption made by Stokes and used a mathematical model derived from acoustic attenuation measurements to compute bulk viscosity terms. Circumventing the Stokes’ assumption, the time-averaged Cp distributions reveal a much better agreement of the shock strength and location with the available experimental data. Furthermore, due to the additional dissipative term in the energy balance equation, the strength of the upstream propagating Kutta-waves also diminishes along with a decrement in the frequency of the unsteady shock oscillation. Finally, it is observed that the additional bulk viscosity in the fluid flow resulted in an overall increase in the aerodynamic efficiency (the lift-drag ratio).