Multidisciplinary UAV Optimization image

Multidisciplinary UAV Optimization

Project Overview

Carrying out UAV optimization involves defining an optimization problem, selecting relevant variables, choosing software with an API capable of performing the necessary calculations, and developing a program that systematically adjusts inputs until the solution converges, ensuring an efficient and effective design.

Skills Used

Python Git

Introduction

Leading a two-person team, we developed a Python program that reduced a UAV’s drag by 15% while maintaining radar cross-section (RCS) by varying the geometry. This results in significant fuel savings, leaving more room for payloads or a longer flight time. I focused on designing the geometry modification loop, which adjusted airfoil shape, wing parameters, and tail sizing, and the aerodynamic analysis, which was used to find lift, drag, and stability.

Technical Approach

Web image

Structure Design Matrix

Literature Review

We began by conducting a literature review of similar projects. This gave us an idea of what had been done before, and helped shape our optimization approach.

Methodology

We chose Python because it offered libraries and APIs for all required tasks. Our final project used openMDAO, OpenVSP, and RadarSimPy to calculate all needed values.

The structure design matrix (above) visually represents the optimization process. We started with a random geometry and fed it into OpenVSP, which created a mesh. Since the Reynolds number (Re) depends on both size and velocity, we use the geometry specifications to calculate Re given the speed of the aircraft. VSPAero takes this information and calculates the lift (Cl) and drag (Cd) of the aircraft. Using the CG from OpenVSP and the CP from VSPAero, we compute the static margin to assess the stability of the aircraft. The mesh from OpenVSP is then used by RadarSimPy to give us the RCS of the aircraft. The algorithm iterates through geometry modifications until convergence.

Geometry

We allowed the program to create any four-digit NACA airfoil by varying camber, the location of camber, and thickness of the airfoil. The program also adjusted the wing size, angle of incidence, and location. To simplify the problem, the program adjusted tail sizing based on wing placement. We validated stability by calculating the static margin in each iteration, ensuring the tail surfaces were properly sized. With more time, tail sizing could have been an independent variable, but this would have exponentially increased computation time. The optimization already required 500 iterations over 26 hours to converge. Because the problem lacked a defined gradient, we could not use a gradient-based solver, which significantly increased computation time.

Constraints and Objective Function

We wanted to optimize RCS and the lift to drag ratio (L/D), as L/D defines the efficiency of the aircraft. Since our solver minimizes the objective function, we took the inverse of L/D (D/L) to ensure proper optimization. We then scaled D/L and RCS to comparable weights and combined them into a single objective function. To maintain feasibility, we constrained both the coefficient of lift and static margin, ensuring sufficient lift and stability.