Robotic Laser Inspection of Airplane Wings Using Quadrotor
Tatyana Kruglova1, a,Daher Sayfeddine2,b, Kovalenko Vitaliy1,c
1,2 Platov South Russian State Polytechnic University, Novocherkassk, Russia
[email protected],
[email protected] (corresponding author),
c
[email protected]Keywords:robotic inspection, quadrotor, fuzzy logic, swarm optimization, aircraft diagnostic.
Abstract. In course of the new development in construction materials, micro and nanoelectromechanical systems, a new sophisticated class of machinery are rising. This has created the need of diagnostic and inspection algorithm matching with the level of artificial intelligence implemented in the end product. In this paper, we present a case study, a simulation on inspecting airplane wings using unmanned quadrotor. The aim is to automate the inspection process, maximize the efficient allocation of resources and minimize possible risks caused by human errors. In order to achieve wing inspection, a laser setup is mounted on the quadrotor, which trajectory and flight stabilization is controlled by optimized fuzzy logic position controllers using particle swarm optimization.
Introduction
Laser inspection has been used to identify cracks on a smooth surface. In particular, such technology is implemented in car assembly lines [1,2] and glass, metal and plastic manufacturing [3]. In parallel, inspection methods were developed based on machine vision algorithms or X-ray scanning [4,5]. The monitoring resultsaresatisfactory. On the other hand, such solutions are immobile, need preparation and activation time and are dependent on site conditions. Per instance, setups used to scan granite-building facades cannot reach difficult corners, require additional manpower, smooth working ground and perfect climatic conditions. In Order to achieve the aim of automating any process, it is necessary to use resources efficiently. Here, multitasking is an essential pillar to consider. It allows reducing time consumption to achieve a project milestone and maximize the technical and financial earned value.
Particularly speaking about aerospace industry, statistics shows that aircraft crashes are increasing. The reasons seem to be hidden. This number of accidents cannot be allowed in parallel with the advances in information technology, communication and artificial intelligence. Human errors and operating inappropriate interventions are being pinpointed to be the reason behind 17% of the accidents [6].
Underlining these facts and limitations, we suggest in this paper a solution based on miniature unmanned aerial vehicle (UAV). A quadrotor, basically a vertical take off and landing rotorcraft (VTOL), equipped with laser, can perform aircraft wing inspection for cracks. In order to achieve reliable results, sensors readings have to be recovered from noises caused by rotors’ vibration and external disturbance. In addition, the quadrotor flight has to be stabilized. Optimized fuzzy logic controllers using particle swarm optimization achieve precise positioning of the UAV.
Related Works
With the new development in technology, robots became cheaper and they are implemented more frequently is civil field. Researchers lunched a variety of special inspecting robots. Typically, these robots are inspired by the nature creatures and their type of movement: sliding, swinging, extending, flying and jumping. A new-sophisticated field has born: the Bionics. As end product, we got new generation robots. For instance, The StickyBot has a hierarchical adhesive structure to hold itself on any kind of surfaces [7], the climbing RiSE V3 robot is designed for high-speed climbing of a uniformly convex cylindrical structure, such as a telephone or electricity pole [8]. The efficiency of these robots was satisfactory but still their acquisition is not justifiable financially.The success achieved by using the UAV in military tasks has expedite the migration of such technology to the civil market. Nowadays, UAV is being used to perform several diagnostic tasks such as inspection of building facades [9], high-rise structure [10], municipality lights [11], bridges and performing indoor radiation safety tests [12]. Quadrotors are a good alternative for the climbing robots; they are cheaper and more service friendly. An autonomous quadrotor is not Wi-Fi dependent and can fly for longer range. Also, it can carry several sensors due to the new development in nanotechnology. Nowadays, quadrotors can be equipped with double or multiple high definition cameras, digital and auto calibrated, enough internal storage to capture long videos and necessary position sensors that assist in achieving better flight control.
Regarding the topic, X-ray setups and 3D laser immobile machines are used to scan airplane wings. Firstly, human-inspector carries out the diagnostic process, searches for viewable cracks. The second step consists of passing manually laser waves on the wing surface (fig.1). Readings on crack depth are acquired and filtered out according to safely norms and conditions. Hereafter, a decision is taken to send for maintenance team or to allow using. This process is made also for helicopter blades and others rotorcrafts. As it can be seen, the workflow requires high precision diagnostic, time and repeated steps. Automating the inspection process using quadrotor can be adequate solution. Moreover, quadrotors can scan multiple airplanes at the same time.
Figure 1. Manual laser wing scanner
Quadrotor Model
Quadrotor dynamics
As it is name underlines, a quadrotor has four rotors working in two sets to achieve six degrees of freedom. The quadrotor changes its positioning by modifying the rotation speed of the rotors. This allows a change in positioning according to one of the following flight regimes: roll, pitch, yaw and hover. The first three regimes are rotational movements and correspond to variation in position along axis X, Y and Z respectively. A change in a value of one of the three angles causes horizontal linear movement of the quadrotor. The hovering is the fact of vertically standing over an assigned area. Hence the quadrotor dynamics can be described using the following equations:
X ̈=(sinψsinφ+cosψsinθcosφ)U_1/m (1)
Y ̈=(-cosψsinφ+sinψsinθcosφ) U_1/m; (2)
Z ̈=-g+(cosθcosφ)U_1/m ; (3)
p ̇=(I_YY-I_ZZ)/I_XX qr-J_TP/I_XX qΩ+U_2/I_XX ; (4)
q ̇=(I_ZZ-I_XX)/I_YY pr-J_TP/I_XX pΩ+U_3/I_YY ; (5)
r ̇=(I_XX-I_YY)/I_ZZ pq+U_4/I_ZZ . (6)
Where, X ̈,Y ̈ andZ ̈are the projection of the linear acceleration of the quadrotor in the Earth fixed axis, p ̇ ,q ̇ andr ̇ are projection rotational acceleration of the quadrotor in the body fixed axis, g is the gravitational acceleration, the torque generated from the rotors, m is the mass of the quadrotor, I_XX,I_YY andI_ZZare the projection of the Inertia of the quadrotor, Ω is the rotational speed of the propellers, , φ, θandψ are the roll, pitch and yaw angle consequently, U_1,U_2 ,U_3 andU_4 are the torques requirement for the hover, roll, pitch and yaw flight regimes respectively.
Observer
Observers are used to recover the vectors of the controlled object. In accordance with this paper, an adaptive observer shall identify sensors noise and regenerate recovered noise-free signals. To identify the observer, we will be using the system of equations (1-6) and reshape it into differential equation for each state. Hence system (7) can be written as follows:
[■(υ ̇_x@x ̇@■(υ ̇_y@y ̇@■(υ ̇_z@z ̇@■(ω ̇_ϕ@ϕ ̇@■(ω ̇_θ@θ ̇@■(ω ̇_ψ@ψ ̇ ))))))]=[■(1/m C_t aU(1)∙θ@υ_x@■(-1/m C_t aU(1)∙ϕ@υ_y@■(-1/m C_t aU(1)@υ_z@■(1/I_xx C_t laU(3)@ω_ϕ@■(1/I_yy C_t laU(4)@ω_θ@■(1/I_zz C_q aRU(2)+J_r U ̇(5)@ω_ψ ))))))]
(7)
Where C_t ais the thrust coefficient of the propeller, l – is the distance between the center of the quadrotor to each motor, C_q is the coefficient of the peripheral thrust, R – propeller radius, Jr – is the moment of inertia of the drive, υ ̇_x,υ ̇_y andυ ̇_zare the projection of the linear velocity, ω ̇_ϕ, ω ̇_θ and ω ̇_ψ are the projection of the rotational velocity.
The differential equation that describes the six states of degree of freedom of the quadrotor and their derivatives is illustrated in (8)
x ̂ ̇=Ax ̂+Bu+K(y-Cx ̂ )+K_(-1) J∫_0^t▒〖(y-Cx ̂ )dt.〗 (8)
Where x is the state vector, y- is the a scalar output signal, u- control vector, K- matrix size of nx1, К–1 matrix size of (nxn), J- matrix size of (nx1).
The completion condition of coincidence between the observer equation and the quadrotor equation can be identified using the following equation
K_(-1) J∫_0^∞▒〖(y-Cx ̂ )dt=w=const.〗 (9)
Thus the output of the integrator in equation (8) provides an estimate of the unknown external influence w. The simulation results illustrated in figure 5 show the effectiveness of the observer to recover the original signal from the applied noises.
Figure 2- Adaptation of the observer with reference to the applied external noises w
The gyroscope signal to left (red curve), the integrated filtered signal to the right (red curve) and the roll angle (blue curve)
From figure 2, it can be clearly seen that the observer has successfully recover the sensor signals from the disturbance. It assures better stabilization of the roll angles as shown by the purple curve.
Optimized Fuzzy logic controllers
As it is important to know the value of the error, it is similarly critical to understand how it is changing over time. The error and its derivative in time are one of the possibilities to configure a Fuzzy logic controller. It is an artificial intelligence approach that computes mathematical operations based on degree of truth rather than the conventional True-False Boolean logic. The Fuzzy logic allows having more adaptable controller specially when dealing with nonlinearities (i.e. nonlinear aerodynamic mode