Repetitively enhanced neural networks method for complex engineering design optimisation problems

Based on current capabilities, we examine the feasibility of creating a carbohydrate-based regenerative fuel cell (CRFC) as the primary power source for unmanned aerial vehicles (UAV) for long endurance missions where station keeping is required. The CRFC power system evaluated in this research is based on a closed-loop construct where carbohydrates are generated from zooxanthellae, algae which create excess carbohydrates during photosynthesis. The carbohydrates are then fed to a carbohydrate fuel cell where electric power is generated for the UAV’s propulsion, flight control, payload, and accessory systems. The waste products from the fuel cell, carbon dioxide and water, are used by the zooxanthellae to create more carbohydrates, therefore mass is conserved in the process of power generation. The overall goal of this research is to examine the potential of CRFCs as a viable power source for UAV systems, to look at scaling issues related to different vehicle sizes and missions, and to identify sensitivities in the CRFC system to different system parameters, indicating the areas where technology improvements may make CRFCs a viable technology. Through simulations, a UAV is sized to determine if greater than 24 hour endurance flight is possible and these results are compared to UAVs using more traditional photo-cell based power systems. The initial results suggest that CRFCs have potential as a power system for long endurance UAVs, and could offer significant improvements to the overall system performance. The final outcome of this research is to identify the most important areas for more detailed follow-on work in designing a production-ready CRFC power system for long endurance UAVs.

Unmanned aerial vehicles (UAVs) have been applied to not only military missions like surveillance and reconnaissance but also commercial missions like meteorological observation, aerial photograph, communication relay, internet network build and disaster observation. Fuel cells make UAVs eco-friendly by using hydrogen. Proton exchange membrane fuel cells (PEMFCs) show low operation temperature, high efficiency, low noise and high energy density and those characterisitcs are well fitted with UAVs. Thus Fuel cell based UAVs have been actively developed in the world. Recently, fuel cell UAVs have started to develope for high altitude UAVs because target altitude of UAVs is expanded upto stratosphere altitude. Long endurance of UAVs is essential to improve effects of the missions. Improvement of UAV endurance time could be fulfilled by developing a hydrogen fuel storage system with high energy density and reducing the weight of UAVs. In this paper, research trend and analysis of fuel cell UAVs are introduced in terms of their altitude and endurance time and then the prospect of fuel cell UAVs are shown.

In recent years, unmanned aerial vehicles (UAVs) have become a dominant force in the aerospace industry. Recent technological developments have moved these aircraft from remote operation roles to more active response missions. Of particular interest is the possibility of applying UAVs toward solving complex problems in long-endurance missions. Under that belief, the feasibility of utilizing UAVs for wildfire detection and response was investigated in a partnership that included NASA’s Aeronautics Research Mission Directorate and Science Mission Directorate, and the United States Forest Service. Under NASA’s Intelligent Mission Management (IMM) project, research was conducted to develop a mission architecture that would enable use of a high altitude UAV to search for reported wildfires with a separate low altitude UAV supporting ground assets. This research proposes a “straw man” concept incorporating both a High Altitude Long Endurance (HALE) UAV and a Low Altitude Short Endurance (LASE) UAV in a loosely coupled, low cost solution tailored towards wildfire response. This report identifies the communications architecture, algorithms, and required system configuration that meets the outlined goals of the IMM project by mitigating wildfires and addressing the United States Forest Service immediate needs. The end product is a defined parent-child framework capable of meeting all wildfire mission goals. The concept has been implemented in simulation, the results of which are presented in this report.

Unmanned Aerial Vehicles (UAVs) stand out as of late because of their different applications in observation, ethereal planning, and information assortment. Aerodynamic productivity is pivotal in boosting UAV execution, especially for flying wing setups. One promising way to improve aerodynamic performance is using dimples on the UAV's wing surface, which alters the flow behavior and diminishes drag. This study plans to examine the effect of different dimple shapes on the aerodynamic characteristics of a flying wing UAV. This paper conducts a comprehensive numerical analysis using Computational Fluid Dynamics (CFD) simulations. The flying wing UAV model is exposed to various dimple arrangements, including round, square, and octagon shapes, put decisively along the wing surface. The CFD simulations utilize the SST k-w turbulence model to capture the flow features accurately. The aerodynamic assessment of each dimple shape is assessed by inspecting essential boundaries like lift coefficient, drag coefficient, and the coefficient of moment for UAV. Preliminary results indicate that dimples significantly affect the aerodynamic behavior of the flying wing UAV. Dimple shape prominently affects drag decrease, stream strength, and, by and large, lift-to-drag ratio. A similar examination gives experiences into the benefits and burdens related to various dimple shapes, aiding the plan streamlining process for flying wing UAVs. The optimized dimple shape recognized through this exploration might upgrade the UAV's endurance, range, and payload limit, making it a resource in applications requiring delayed or broadened flight activities.

In order to improve the endurance of unmanned aerial vehicles (UAVs), a power supply controller with automatic switching between solar energy and battery was designed. The controller is mainly composed of an AT89C51 single chip microcomputer. By collecting ambient light intensity and comparing it with the preset light intensity of the system, the controller completes automatic switching of the power supply for solar unmanned aerial vehicles, thereby achieving continuous and stable power supply for solar unmanned aerial vehicles. In addition, the implementation method of the single-chip microcomputer program in the controller is also introduced. The control core is to use Schmidt de shake control to switch between solar energy and battery, and to test the various functions of the controller through the flight experiment of a small solar powered unmanned aerial vehicle. Tests have found that solar powered drones using this controller can navigate for a long time in weather with varying light intensity, and the device operates well, effectively improving the power stability of solar powered drones.

In the coming sixth-generation era, unmanned aerial vehicles (UAVs) have been explored for convenient air access and strong environmental adaptability to provide communication services for ground Internet-of-Things (IoT) devices. However, the inherent broadcast nature of wireless makes UAV communications vulnerable to eavesdropping attacks. This paper explores a secure uplink communication in a heterogeneous network, where a UAV provides communication services to ground IoT nodes in the presence of an eavesdropper. We propose a transmission scheme to maximize the overall sum secrecy rate, in which the cooperative jamming technique and coalition formation game are employed, and the UAV endurance is considered. Specifically, a double-layer optimization problem is formulated, including a non-convex optimization problem in the lower layer, and a game theory problem in the upper layer. We use block coordinate descent and successive convex approximation methods to decompose the lower layer problem into UAV deployment and power allocation sub-problems, and convert them into a difference of two convex functions programming problem and a second-order cone programming problem, respectively. Moreover, we develop an algorithm to solve the upper layer problem, obtaining a stable coalition formation. Simulation results show that the proposed transmission scheme can effectively improve the overall sum secrecy rate, and has dynamic adaptability to eavesdropping attacks.

"One of the main problems in micro unmanned aerial vehicles is endurance or flight time since the general domain aircraft uses conventional fuel, which is a pollutant, has a limited life and is costly. Then, there is a huge demand for using an unlimited non-exhaustible source of energy. Solar energy is one of the unlimited available renewable energy which can be used to increase the endurance of unmanned aerial vehicle without adding significant mass or increasing the size of the fuel system. This paper aims to encourage research on renewable energy sources for aviation considering the basic challenges for a solar-powered aircraft: geographical area of operation, energy collection and storage, payload and design parameters. Hence, a plane is designed for 2 kg, including payload, and is analyzed in various aspects. Besides, the design is optimized starting from airfoil to complete structure for better performance."

This paper presents a comparative analysis of hybrid hydrogen fuel cell and lithium battery power systems for enhancing the endurance of unmanned aerial vehicles (UAVs). Compared to a battery-only configuration, the study evaluates the performance of a UAV powered by a 500 − W proton exchange membrane (PEM) fuel cell integrated with a lithium polymer battery. Experimental tests, including static bench evaluations and flight trials, were conducted to assess key performance metrics such as flight duration, energy density, power distribution efficiency, and hydrogen consumption rate. The results demonstrate that the hybrid system extends flight endurance by 25%, increasing operational time from 20 to 25 minutes. However, challenges related to hydrogen storage constraints and limited peak power output necessitate further optimization. The cost-benefit analysis highlights a higher initial investment for hydrogen fuel cells than battery-only systems, yet fuel cells offer a longer operational lifespan and reduced battery replacement frequency. The break-even point for the hydrogen system, at which a fuel cell-only configuration achieves the same energy density as the battery-only system, is observed at 3.5 − L of hydrogen stored at 350bar. The findings suggest that hybrid hydrogen fuel cell-powered UAVs offer a promising alternative for long-endurance missions, provided that key storage and power management limitations are addressed.

CFD study of the effect of leading-edge tubercles on the aerodynamic characteristics of a small UAV based on eppler 186 airfoils

The momentum method has become the workhorse in the deep learning community. To theoretical understand its success, researchers put efforts in demystifying its convergence properties when optimizing neural networks. For the convex problem, it is well-known that the triple momentum (TM) method owns the fastest theoretical convergence rate among all first-order methods. However, there exists no theoretical convergence results about the TM method in solving the non-convex neural networks training problem, let alone its acceleration guarantee. In this paper, we focus on the training process of the TM method for a two-layer ReLU neural network. Inspired by the accurate characterization of the high-resolution dynamical system, we consider the high-resolution ordinary differential equation (ODE) of the TM method. Under the over-parameterized assumption, we derive that the original non-convex optimization problem can be transformed to a strongly convex task. By applying an appropriate Lyapunov function, we prove that the TM method can linearly converge to a global minimum. Compared to the heavy ball method and Nesterov’s accelerated gradient method, our result provides the first guarantee for the acceleration of the TM method in training neural networks. Through empirical experiments, the accelerated convergence of the TM method and the effect of over-parameterization are validated.

Comparison of the Kriging and neural network methods for modeling foF2 maps over North China region

Due to its simplicity and efficiency, the first-order gradient method has been extensively employed in training neural networks. Although the optimization problem of the neural network is non-convex, recent research has proved that the first-order method is capable of attaining a global minimum during training over-parameterized neural networks, where the number of parameters is significantly larger than that of training instances. Momentum methods, including the heavy ball (HB) method and Nesterov's accelerated gradient (NAG) method, are the workhorse of first-order gradient methods owning to their accelerated convergence. In practice, NAG often exhibits superior performance than HB. However, current theoretical works fail to distinguish their convergence difference in training neural networks. To fill this gap, we consider the training problem of the two-layer ReLU neural network under over-parameterization and random initialization. Leveraging high-resolution dynamical systems and neural tangent kernel (NTK) theory, our result not only establishes tighter upper bounds of the convergence rate for both HB and NAG, but also provides the first theoretical guarantee for the acceleration of NAG over HB in training neural networks. Finally, we validate our theoretical results on three benchmark datasets.

Today, unmanned aerial vehicles (UAV) are widely used across various fields, including civil, military, and social activities. Fixed-wing UAV operating at low altitudes typically have a wingspan ranging from 2 to 4 meters and can sustain flight for up to three hours. Their wing structure generally consists of three main components: a central wing that bears the primary structural load, and left and right side wings that are critical for determining flight endurance and speed. The lengths of these components vary, with the side wings designed at an angle to reduce wind resistance and enhance flight speed. In this study, three main parameter levels were identified for an average low-altitude drone, and the optimal dimensions were evaluated using RSM (Response Surface Methodology), the Taguchi method, FEM (Finite Element Method), and Analysis of Variance (ANOVA) analyses. The optimal design was achieved with a central wing length of 400 mm, side wing lengths of 700 mm, and a side wing angle of approximately 8°. Analysis results showed a maximum lift force (FZ) of 253 N, a minimum drag force of 10.2 N, a maximum lift coefficient (CL) of 0.66, and a lift-to-drag (CL/CD ) ratio of 17.6. Based on these findings, composite molds were manufactured for the aircraft, and a testing device was developed to measure speed during flight trials. Under 10 m/s wind conditions, the maximum speed recorded for this optimized geometry was 183 km/h.

Plasma flow control (PFC) is a promising active flow control method with its unique advantages including the absence of moving components, fast response, easy implementation, and stable operation. The effectiveness of plasma flow control by microsecond dielectric barrier discharge (μs-DBD), and by nanosecond dielectric barrier discharge (NS-DBD) are compared through the wind tunnel tests, showing a similar performance between μs-DBD and NS-DBD. Furthermore, the μs-DBD is implemented on an unmanned aerial vehicle (UAV), which is a scaled model of a newly developed amphibious plane. The wingspan of the model is 2.87m, and the airspeed is no less than 30m/s. The flight data, static pressure data, and Tufts images are recorded and analyzed in detail. Results of the flight test show that the μs-DBD works well on board without affecting the normal operation of the UAV model. When the actuators are turned on, the stall angle and maximum lift coefficient can be improved by 1.3° and 10.4%, and the static pressure at the leading edge of the wing can be reduced effectively in a proper range of angle of attack, which shows the ability of μs-DBD to act as plasma slats. The rolling moment produced by left-side μs-DBD actuation is greater than that produced by the maximum deflection of ailerons, which indicates the potential of μs-DBD to act as plasma ailerons. The results verify the feasibility and efficacy of μs-DBD plasma flow control in a real flight and lay the foundation for the full-sized airplane application.

This study presents a surrogate-model-assisted Quasi-Newton optimisation framework for simultaneously improving the aerodynamic performance and radar stealth characteristics of an unmanned aerial vehicle (UAV). High-fidelity computational fluid dynamics (CFD) and computational electromagnetics (CEM) simulations are integrated through surrogate models generated via a face-centred central composite design within a design of experiments framework. Quadratic polynomial response surface equations are constructed for key aerodynamic and radar cross-section (RCS) metrics, enabling analytical gradient evaluation. A gradient-based quasi-Newton method with Broyden–Fletcher–Goldfarb–Shanno Hessian updates is employed to minimise a scalarised objective function combining normalised maximum lift coefficient, overall RCS and frontal RCS. Constraints are imposed on the lift-to-drag ratio ( $L/D \geq 10$ ) and static longitudinal stability ( ${C_{m0}} \geq 0$ ). Analytical derivatives from the response surface equations (RSEs) eliminate the need for direct numerical differentiation of CFD/CEM outputs, reducing computational cost and eliminating simulation noise. An interior-point sequential quadratic programming strategy is used to ensure satisfaction of nonlinear constraints during the optimisation process. The optimised UAV design demonstrates a $12{\rm{\% }}$ increase in maximum lift coefficient and a $30{\rm{\% }}$ reduction in both overall and frontal RCS compared to the baseline configuration. The results are confirmed through high-fidelity CFD and RCS simulations and are further validated experimentally in an anechoic chamber, with close agreement across all measured frequencies. The proposed methodology provides an efficient and experimentally verified approach for integrated aerodynamic and stealth optimisation in UAV design.