Content uploaded by Josef Schleupen
Author content
All content in this area was uploaded by Josef Schleupen on Nov 17, 2016
Content may be subject to copyright.
adfa, p. 1, 2011.
© Springer-Verlag Berlin Heidelberg 2011
Developing a climbing maintenance robot for tower and
rotor blade service of wind turbines
Josef Schleupen
1
, Heiko Engemann
1
, Mohsen Bagheri
2
, Stephan Kallweit
1
and Peter
Dahmann
2
1
University of Applied Sciences Aachen, Institute for Mobile Autonomous Systems and Cogni-
tive Robotics (MASKOR), Aachen, Germany
{schleupen,engemann, kallweit}@fh-aachen.de
2
University of Applied Sciences Aachen, Faculty for Aerospace Engineering, Aachen, Germany
{bagheri,dahmann}@fh-aachen.de
Abstract. Today, more than 275.000 wind turbines generate over 400 GW elec-
trical power worldwide. So the demand for maintenance constantly raises. Since
September 2014 the University of Applied Sciences Aachen and partners develop
SMART (Scanning, Monitoring, Analyzing, Repair and Transportation), a mainte-
nance platform for wind turbines. The research project is funded by the German
federal ministry of economic affairs (BMWi), to support the upcoming industrial
needs. While the reliability of the mechanical parts, like main bearing, generator,
gears and main shaft increased during the recent years, the maintenance and im-
provement of rotor blades should be improved. A weatherproof cabin for rotor
blade maintenance can extend the annual maintenance period from eight to twelve
months, a major goal of the SMART development. In addition, a climbing mech-
anism for conical shaped, thin and slippery surfaces is generated and tested.
SMART successfully completed the proof-of-concept milestone by demonstrating
climbing in December 2015.
Keywords: climbing, robot, rotor blade, wind turbine, maintenance, service,
tracking, ADAMS, chain drive, ROS, ar_track_alvar
1 Introduction
The wind market is growing rapidly. In China, the annual growth is about 45 % [1]. The
technology of generating electricity from wind power is still young. The amount of in-
stalled wind turbines raised the need for maintenance.
The reliability of mechanical parts, e.g. main bearing, generator, gears and main shaft
evolved during the recent years, while the rotor blade maintenance needs to be im-proved
[2]. The SMART demonstrator is a downscaled model (Fig. 1) for research and devel-
opment. The main goal is the design of a fully functional prototype for a 2.5 MW wind
turbine, including a weatherproof cabin. Current rotor blade maintenance is limited to
convenient weather conditions: wind speed lower than twelve meter per second, low
humidity, temperature above 10 degrees Celsius. These requirements currently limit the
maintenance period to the warmer period of the year. A weatherproof cabin for rotor
2
blade maintenance extends the annual maintenance period from eight to twelve months
and from three to 24 hours a day.
Fig. 1. SMART demonstrator, scale one to three
In addition SMART can increase the quality of monitoring rotor blades: state of the art
technologies for inspection, like e.g. ultrasonic- and terahertz-spectroscopy, X-ray and
thermography, may be established to support the engineers and technicians. Rotor blade
manufacturing procedures can be scaled down and integrated into the platform, in order
to avoid expensive and inefficient dismounting of the rotor blades for full-inspection,
repair and replacement. Further research and development is focusing the possibilities
of a cooperative or stand-alone robot systems for inspection and repair duties. Customi-
zation of the platform for special applications, e.g. RBE - rotor blade extensions (Ener-
giekontor), fully autonomous inspection and turbine tower maintenance are part of the
challenging development.
The following two subchapters intend to summarize general principles for the climbing
robot design and associated research and development topics.
1.1 Kinematics of the climbing robot motion
SMART is a novel mobile robot design. One application for this kind of robotic sys-tem
is the monitoring and maintenance for the rotor blades and the tower of wind turbines.
There are two possible solutions to climb a wind turbine – either by using ropes from
the top or by climbing based on friction.
The robotic system SMART uses a frictional connection to the tower. Such mechanism
can be split into two subsystems: the tensioning system and the climbing system. The
intention of a tensioning system is to provide the essential normal force for static friction
between the tower surface of the wind turbine and the climbing system. The climbing
system can either be intermittent or continuous.
3
Fig. 2. Dual tracked drive of SMART robot
The following approach introduces a continuously climbing system, based on a tracked
vehicle design (Fig. 2). Parts of the tracked drives are: chain disk, harmonic drive motor,
guiding rails, chain tensioning mechanism, body, bearings and a 6D force sensor.
Many wind turbine towers are conically shaped. Therefore a major requirement for the
SMART robot is to continuously decrease the perimeter of the tensioning system, while
climbing up. In addition, the attachment and suspension of the tracked drive must permit
a pitch angle towards the tower axis around 5 to 15 degrees. There are several other
scenarios and circumstances that will require further degrees of freedom, e.g. skid-steer-
ing to climb horizontally.
Fig. 3. Kinematic model of tracked drive chassis frame
In the following a kinematic model of the climbing robot will be explained, simulated
and experimentally tested for scientific prove. Figure 3 illustrates the degrees of freedom
of the current robot design that belong to the chassis frame of the tracks. Axis one and
4
two are required to compensate different tower perimeters and to allow the tracks to
contact the tower tangentially. Axis 3 is required for the relative pitching between two
tracks during the steering process. The necessity will be underlined later with the virtual
model in ADAMS ATV. Finally, the spherical ball joint (axis 4, 5, 6) is required to let
the tracked drives move relative to the tensioning system.
The interface to the tensioning system is mounted on top of a 6D force sensor to measure
the forces and torques induced by the robot movement. The kinematic model for the
tensioning system, derived from the “Nurnberg Scissors” principal, is displayed in 2D
in figure 4.
Fig. 4. Kinematic model of tensioning system
The climbing robot consists of around 95 moving parts, 49 revolute joints, 6 screw joints,
5 spherical joints, 12 translational joints and 17 fixed joints.
1.2 System dynamics of SMART demonstrator
Similar to the experimental SMART demonstrator, shown in figure 1, a model is de-
rived utilizing ADAMS (acronym of Automated Dynamic Analysis of Mechanical Sys-
tems). Current demands from use cases suggest that this device will have a mass of up
to 5 tons, including payload and power supply. A strong commitment to composites
could reduce the weight by 20%.
Fig. 5. Virtual SMART model (ADAMS ATV simulation)
The virtual model (Fig. 5) possesses the same links, joints, masses and forces like the
experimental model. The virtual model will be verified by the experimental demonstra-
tor. The final goal of this process is to establish a valid virtual model to support upscaling
5
into a one-to-one prototype. The prototype will be designed for a 2.5 MW wind turbine.
More than 65% of the weight is distributed concentrically around the wind turbine tower
axis as part of the power supply and the climbing mechanism. This will reduce the bend-
ing moment on the tower.
There are three different ways of power supply in the evaluation process: electric power
cable to ground, on board generator, on board LiFePo4 batteries similar to electric cars.
The most promising supply is based on batteries with a total weight of up to 400 kg in a
worst-case scenario as part of the 5 tons total mass. The batteries and the BMS will be
stored inside the tracked drives to support the weight distribution concept.
2 Modell based kinematic and dynamic simulation
Essential for analyzing the climbing process of SMART are static and dynamic loads
(Fig. 6) in combination with a variety of different movements, so called scenarios. A
standard scenario for a climbing robot is moving up and down. The movement path is
described as a vertical path, parallel to the tower axis. Further scenarios consists of steer-
ing or rather horizontally climbing, described by a spiral path around the tower axis.
Finally, steering on a conical shaped surface must be analyzed. The static forces shown
in figure 6 are expected to change significantly during the steering processes. Subse-
quently to the dynamic simulation many results are shared with ANSYS Workbench, an
FEM analysis tool, to optimize the design and the structure of the climbing robot.
Fig. 6. Static forces climbing robot SMART
ADAMS Tracked Vehicle (ATV) supports the modelling of the climbing robot with a
framework to simulate track drives. Table 1 illustrates two similar scenarios, steering
towards right and left, regarding normal forces and lifting forces.
The ADAMS simulation shows that the forces are equally distributed between the tracks
in both scenarios. The model offers the possibility to track all motions and forces of
SMART virtually. To validate these results several sensors are integrated in the experi-
mental demonstrator.
6
Table 1. Steering towards left and right, „horizontally climbing”
3 Experimental analyses of SMART demonstrator
A major goal of the experimental analyses is to validate the simulation and theoretical
calculations of the climbing robot. The measurements are split into two parts: Motion
tracking and force acquisition.
3.1 Setting up a 3D vision system for motion tracking
The motion tracking is done by a 3D vision system. It is based on four high resolution
cameras, which are used to detect specific landmarks attached to the SMART robot. The
task of the tracking system is to identify the 6D pose of each individual landmark in
relation to a fixed frame in space – the world frame. As a result the motion tracking
system provides a punctual observation of the motion process of the SMART robot.
A Manta G609B camera offers Mono8 images with a resolution of 2752x2206 pixels at
a maximum of 15 fps. The cameras are mounted around the SMART robot, with indi-
vidual distances to the robot greater than three meters. Camera lenses with a focal length
of 25 mm are used. As a result the ROI is covering the whole motion area of the robot.
The identification of different landmarks is possible by a binary coding of the markers.
Multiple landmarks are mounted at fixed positions of the SMART robot.
Fig. 7. Landmark for vision tracking [3]
The software ar_track_alvar of the Robot Operating System ROS includes an open
source AR tag tracking library. It is used to determine the 6D poses of the landmarks in
relation to the world frame of the camera. The software camera_calibration is used to
provide the parameters of the camera system.
7
The determination of the 6D pose of a landmark with reference to the world frame de-
pends on the 6D pose of the specific camera system in relation to the world frame. The
software tf of the Robot Operating System is used to calculate these trans-formations. In
addition, a gyroscope is added to each track drive to validate the tilt angles while steer-
ing.
3.2 Force sensors
The SMART demonstrator consists of five similar drives distributed over the tower sur-
face (Fig. 8). The z-axis of the coordinate frame is parallel to the tower axis and is point-
ing upwards. The Y-axis points in radial direction of the tower towards drive A1. In each
drive, A1 to A5, a 6D force sensor, K6D68 10kN/500Nm from ME-Systems, is imple-
mented to measure Fx, Fy, Fz, Rx, Ry and Rz. This customized sensor can measure loads
of up to 20kN in z direction. In the current design a ball joint is mounted on top of the
6D sensor. Therefore, the torques are currently not significant. In addition to the 6D
force sensors ten strain gauges are installed in the tensioning system to measure the stress
during the climbing process. The feedback from the Harmonic Drive motors delivers
absolute position, torque, acceleration and velocity. In fact, the multiturn absolut encoder
feedback is not relevant at this point, because there is still slip between the drives and
the tower surface.
Fig. 8. System overview SMART robot
The following test results display the current state of development. An advanced con-
troller is currently established to reduce slip and unbalanced forces for the track drives.
8
4 Results
The data is generated during a standard climbing scenario test run. The SMART demon-
strator climbs 0.6m up and down in around 52s.
Fig. 9. Motion tracking of SMART demonstrator
All drives are controlled manually and started at the same time (Fig. 9). The real-time
controller is disabled to show the effects of non-linear force distribution. The velocity
profile characteristics are: 5% from 7.5s to 22.5s, 15% from 22.5s to 42.5s, 5% from
42.5s to 55.0s.
Fig. 10. Force measurements of SMART demonstrator
Figure 10 displays the forces Fx, Fy and Fz for all drives. In addition, a 1D force sensor
is measuring the force between the tensioning system and the vertical connection to drive
A1 (FzA1). This force reflects the share of the lifting-force that is generated by drive A4
and A5 on the other side of the tower. This force should be the same for all drives climb-
ing at the same speed. When all front drives A1, A2 and A3, are generate slip – the 1D
9
forces FzAn raises. The critical point is reached at 27.5s. The height difference of all
drives need to be compensated thus resulting in a different force distribution.
5 Conclusion and future aspects
SMART demonstrator concept is able to climb vertically based on friction between the
tracked vehicle rubber parts and the tower surface. The results from the tracking system
are similar to the movement in ADAMS ATV. Results from the force sensor – without
the force feedback controller – differs significantly from the expected values generated
by the dynamic simulation.
This specific test run shows the necessity for advanced model based control [4]. To gen-
erate an equally distributed force over all drives, each motor speed needs to be accurately
controlled. The feedback loop actuator consists of the speed control of the motors from
the tracked vehicles as well as the motors from the tensioning system.
Pressure sensors are currently implemented on top of the rubber-chain-parts of the
tracked vehicle to validate the surface pressure between the tower and the tracked drives.
Subsequently it will be possible to investigate the steering process for further improve-
ment of the SMART kinematics. The validation of the analytical model of the SMART
by experimental data will allow the future upscaling of the robot system.
The SMART prototype is designed for wind turbines of up to 2.5 MW in the test phase.
Therefore the maximum distance between the tower and the rotor blade is around 12 m.
Current FEM analyses [5] and some tests on the 1:3 climbing robot demonstrator al-
lowed lifting a 140 kg payload, corresponding to a payload of around 500 kg for the 1:1
prototype. The demonstrator presented in the paper is able to carry the load at the end of
the cabin arms and has not yet reached its limit. The cabin arms are based on a multiple
triangle framework that support guide rails for the cabin movement. Both cabin arms are
connected to each other in order to compensate torque. The cabin structure is lightweight
and made of composites. Therefore all forces generated by the rotor blade movement or
coming from the working platform inside, are guided directly into to cabin arms. The
cabin structure will absorb the wind forces and its own weight.
Bibliography
1. Global Wind Energy Council (GWEC): Global Wind Statistics 2015,
http://www.gwec.net
2. Deutsche Windtechnik: On- and Offshore Services, http://www.deutsche-
windtechnik.de
3. Quigley, M., Conley, K., Gerkey, B., Faust, J., Foote, T., Leibs, J., Wheeler, R., Ng, A.: ROS:
an open-source Robot Operating System, ICRA Workshop on Open Source Software, 2009.
4. Chitta, S., Sucan, I., Cousins, S.: MoveIt!, IEEE Robot. Automat. Mag. , volume 19, 1, pages
18-19, (2012)
5. Kallweit, S., Dahmann P., Bagheri, M., Schleupen, J., Engemann, H.: 13. AALE Conference:
Entwicklung eines Kletterroboters zur Diagnose und Instandsetzung von Windenergieanla-
gen (2016)
