Bond Graph Paper A K Samantary

Journal of Systems and Control Engineering
Fo
Multi-body Dynamics and Control of a Twin Rotor System
Based on an Integrated Multi-energy Domain Bond Graph
Model
rP
Journal: Part I: Journal of Systems and Control Engineering

ee
Manuscript ID JSCE-19-0116
Manuscript Type: Original article

rR
Date Submitted by the

06-Apr-2019
Author:
Complete List of Authors: Srinivasarao, Gopisetti ; Birla Institute of Technology, Department of

ev
Mechanical Engineering
Samantaray, A; IIT Kharagpur, Mechanical Engg.
Ghoshal, Sanjoy; Indian Institute of Technology (IIT ISM) Dhanbad,
Department of Mechanical Engineering
iew
dynamic modelling, mechanical control systems, system simulation,

Keywords:
numerical modelling/ simulation, physical modelling
The dynamics of a Twin Rotor Multi-Input Multi-Output (MIMO) System

(TRMS), which is similar to that of a helicopter in many ways, is highly
nonlinear in nature. In this paper, a detailed dynamical model of TRMS is
developed and simulated by using bond graph (BG) approach. Nonlinear
nature of the interface gain, thrust and drag forces, and the stiffness of
cable attached to support column joint are estimated. The rotors are
modeled by using Newton-Euler equations. The BG model is created by
Abstract: using the generic sub-models and the same set of sub-models can be
assembled differently to model many other similar systems such as
tricopters and quadcoptors. Inertial forces and moments, rotor thrust
and drag forces, active and reactive motor torques, and DC motor
dynamics are considered in the model. The responses from the model
are compared with the test data for validation. Thereafter, PID control is
applied on the validated model for tracking the desired pitch and yaw
angles of the TRMS.
http://mc.manuscriptcentral.com/JSCE
Page 1 of 29 Journal of Systems and Control Engineering
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
Fo
20
21
22
23
rP
24
25
26
ee
27
28
29
rR
30
31
32
ev
33
34
35
36
iew
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60 http://mc.manuscriptcentral.com/JSCE
Journal of Systems and Control Engineering Page 2 of 29
Multi-body Dynamics and Control of a Twin Rotor System Based on Bond Graph Model
1
2
3 Multi-body Dynamics and Control of a Twin Rotor System Based on an Integrated
4
5
Multi-energy Domain Bond Graph Model
6
7 Gopisetti Srinivasaraoa Arun K.Samantarayb,1 Sanjoy K.Ghoshalc
8
9 aDepartment of Mechanical Engineering, Birla Institute of Technology (BIT), Mesra, India
10 bDepartment of Mechanical Engineering, Indian Institute of Technology (IIT) Kharagpur, India
11 cDepartment of Mechanical Engineering, Indian Institute of Technology (IIT-ISM) Dhanbad, India
12
13
14 Abstract
15
16 The dynamics of a Twin Rotor Multi-Input Multi-Output (MIMO) System (TRMS),
17 which is similar to that of a helicopter in many ways, is highly nonlinear in nature. In this
18 paper, a detailed dynamical model of TRMS is developed and simulated by using bond
19 graph (BG) approach. Nonlinear nature of the interface gain, thrust and drag forces, and
Fo
20
21
the stiffness of cable attached to support column joint are estimated. The rotors are
22 modeled by using Newton-Euler equations. The BG model is created by using the
23 generic sub-models and the same set of sub-models can be assembled differently to
rP
24 model many other similar systems such as tricopters and quadcoptors. Inertial forces and
25 moments, rotor thrust and drag forces, active and reactive motor torques, and DC motor
26 dynamics are considered in the model. The responses from the model are compared with
ee
27
the test data for validation. Thereafter, PID control is applied on the validated model for
28
29 tracking the desired pitch and yaw angles of the TRMS.
rR
30
31 Keywords: Multi-body dynamics, Bond graph, Twin rotor system, Pitch and yaw control.
32
ev
33 1. Introduction
34
35
36
The twin-rotor MIMO system (TRMS) considered in this study resembles a helicopter to some
iew
37 extent due to its use of two rotors for pitch and yaw control. In comparison to a helicopter, some
38 simplifications are made in the TRMS. Firstly, the TRMS is pivoted to a tower and hence it does
39 not move up or down at the pivot point. Secondly, the TRMS position and velocity are controlled
40 by the rotor velocity variation whereas in a helicopter, the speeds of the rotors remain constant
41 and the propulsive force is changed by varying the angle of the rotor blades. Nevertheless, some
42 imperative nonlinear dynamic characteristics in a helicopter are realized in the TRMS model.
43
44
The TRMS set-up, developed by Feedback Instruments, consists of a horizontal beam which is
45 pivoted to a vertical tower at a point near the middle such that the beam can swivel in both the
46 horizontal and the vertical planes. When the TRMS is idle, the main rotor side of the horizontal
47 beam leans downward. There are two permanent magnet DC motors attached to both the ends of
48 the beam and their axes are perpendicular to each other. A propeller is attached to each of the
49 motors, one is larger (main rotor) and the other is smaller (tail rotor). A counterweight is attached
50
to the beam at the pivot point, perpendicular to the beam, which helps in balancing the system in
51
52 the static condition. The system can be controlled by changing the supply voltages to the DC
53 motors. The TRMS tip motion is constrained to remain on a sphere. The measured output signals
54
55
56 1Corresponding Author. E-mail: samantaray@mech.iitkgp.ac.in, Tel: +91 3222 282999, Fax: +91 3222 282778.
57
58 1
59
1
2
3 from the setup are the two position angles (i.e. two generalized co-ordinates yaw and pitch) of
4
5
the beam.
6 The model provided in the product manual of Feedback Instruments (1998) is empirical
7 and dynamically non-invertible. In the past, many researchers have developed some dynamical
8 1
models of the TRMS. Rahideh et al. employed the Newtonian and Lagrangian based analytical
9 models and also a neural network based empirical model to describe the dynamical behaviour of
10
11
the TRMS. An adaptive neuro-fuzzy inference system (ANFIS) network design technique was
2 3
12 used by Toha et al. for modeling of the TRMS. Ahmad et al. proposed a black box system
13 identification technique to estimate a linear time invariant (LTI) model. The approach requires
14 measurement of the natural frequencies of the system. A nonlinear system identification
15 4
16 technique based on the radial basis function (RBF) networks was proposed by Ahmad et al. A
17 neural network (NN) based nonlinear dynamic modeling approach was proposed by Rahideh et
5
18 al. where resilient propagation (RPROP) learning algorithm was used for training a feed
19 forward neural network. Toha and Tokhi6 proposed a real coded genetic algorithm (GA)
Fo
20
technique for parametric modeling of the TRMS. Later on, a dynamic particle swarm
21 7
22 optimization (PSO) was proposed by Toha et al. for parametric modeling of TRMS. Rahideh
23 8
rP
and Shaheed presented mathematical models of the TRMS dynamics based on Newtonian and
24
Lagrangian methods and the responses from both the models were compared with the real
25 9
26 TRMS. A gray box approach was proposed by Moness and Mostafa for the modeling of the
ee
27 TRMS. An algorithm was proposed by dividing the system into independent subsystems that can
28 be estimated for a given input. The proposed algorithm used trust region optimization methods
29 for successive estimations. A quasi-linear parameter varying (quasi-LPV) modeling,
rR
30 10
31 identification and control of the TRMS was proposed by Rotondo et al. where nonlinear least
32 square identification approach was used to identify the unknown model parameters. In the
ev
33 aforesaid models, cross coupling effect was modeled based on the angular acceleration of the
34 rotors which would disappear if the rotors were rotated with uniform angular velocity, thus
35 contradicting the physical reality. A complete nonlinear model of the similar system was
36
iew
11
37 developed by Mullhaupt et al. based on Lagrangian method. In this model also the cross
38 coupling was modeled in asimilar way. In a comprehensive dynamic model presented by
39 12
Tastemirov et al. , a flat cable at the base of the setup was modelled by a spring having certain
40 stiffness influencing the yaw angle. Thereafter, they designed a suitable controller for yaw
41
42
stabilization. In all these existing models, the DC motor reaction torque, which is also a cause of
43 cross-coupling, is not considered explicitly. Also, the dynamics of the main and tail rotors have
44 not been addressed in detail. The flat cable attached to the supporting beam, which may provide
45 nonlinear stiffness to the system in the horizontal plane motion, has not been properly modeled.
46 Most researchers have plugged-in one or more first and/or second order transfer functions during
47 model identification stage for model and measured system response matching. Such ad-hoc
48
subsystems and black/grey-box artificial intelligence models do not have direct physical
49
50
correspondence with the real system.
51 The present work is focused on detailed dynamic modelling of the TRMS system through
52 bond graph (BG) method. Due to various attractive features such as graphical nature and unified
53 approach to model the multi energy domain and complex multi body systems, BG method has
54 been successfully used over the years in the modeling of various mechanical and
55 13-16 17
56
electromechanical systems with nonlinear constrains. BG technique was used by Engja to
57
58 2
59
1
2
3 18
4 develop a nonlinear model of a reciprocating compressor. Drozdz & Pacejka used it to develop
5 a car model in order to examine the influence of tire characteristics on vehicle handling. Pathak
19
6 et al. designed an impedance controller for space robots using bond graphs. A bond graph
7 model of a planar slider component with two moving contact points was developed by Bera et
8 20
9 al. where reaction forces, contact forces and friction forces acting at different revolute and
10 prismatic joints in a hydraulic cylinder-piston assembly were calculated by using the BG method.
11 Further models of multi-body dynamics by using bond graph can be found in literature.
21-23
12
BG modeling allows integration of different energy domains and is ideally suited for
13 24-26
14 modeling and control of mechatronic systems. A complete solid oxide fuel cell (SOFC)
15 27
system with heat exhanger and control valves etc. is modelled by Vijay et al. in bond graph
16 framework with complex cascaded controller implementation.Therein, the hydrogen and air flow
17
18
rates are simultaneously controlled to minimize thermal stresses and energy loss during load
28
19 transients. Sharma and Bera presented a new bond graph model of the autonomic nervous
Fo
20 system embedded with baroreflex system. In healthy and faulty states, a quadruped robot was
21 29
modeled and simulated through bond graphs by Gor et al. Bond graph model for a redundant
22 30-31
23 space robot was created by Dalla and Pathal. A fixed-causality diagnostic bond graph of
rP
24 linear models with ideal switches was used for designing fault indicators by Borutzky.
32
25
BG modeling has also been implemented for model based fault detection, isolation and
26
ee
33-35 36
27 prognosis by various authors. Bicausality principle is used by Loureiro et al. to generate the
28 fault idicators from a bond graph model. The method is applied for fault diagnosis and isolation
29 of the elctro-mechanical part of a robotino mobile robot. In a safety critical system like a
rR
30 helicopter which is similar to the twin-rotor set-up, the BG model-based diagnosis and prognosis
31 37-38
32 can be applied to monitor the health of the motors, rotors, control system and sensors.
ev
33
34 In literature, different kinds of nonlinear controllers for the TRMS were proposed by
39
35 various authors. Li et al. proposed a fuzzy adaptive backstepping controller to control a similar
36
iew
40
type of 3 DOF helicopter system. Faris et al. implemented a decentralized sliding mode
37 41
38 controller for the TRMS in a real-time control. Pandey et al. decoupled the TRMS using
39 generalized decoupling technique and then implemented two individual PID controllers. They
40 used the Kharitonov theorem to find out the ranges of parameters of the controllers to ensure
41 robustness and also adopted the bacterial foraging optimization technique to obtain the
42
parameters values from the robust ranges.
43
44 The dynamical analysis of TRMS or similar type of systems by using BG method with
45 incorporation of all the dynamical forces such as thrust forces, drag forces, flat cable nonlinear
46 stiffness torque, motors and their reaction torques, and the damping forces are not available in
47 the literature as per the knowledge of the authors.
48 The objective of this research is to develop a completely physics-based model of the TRMS
49 setup without using any ad-hoc elements in the model. In this paper, the detail dynamics of a
50
51
TRMS are explored through bond graphs (BG). The TRMS model is divided into eight
52 subsystems and all these subsystems are modeled as rigid body submodels. These submodels are
53 assembled through the appropriate constraints. The DC motors are also explicitly modeled
54 through BG. Thereafter, PID control is applied on the said model for tracking a desired pitch and
55 yaw angle of the TRMS and the results are compared with experimental test data. It is possible to
56
57
58 3
59
1
2
3 model the controller with the system on a single platform by using BG. In addition, the BG
4
5
model is open-ended and can be converted to a block diagram or direct simulation code from the
6 equations of state and those can be solved with various numerical integration procedures.
7 The motivation for this work is to develop a validated BG model that can be used in future
8 for various safety-critical applications related to helicopters such as its robust trajectory control
9 during automatic navigation, rescue operation, advance warning system with state-of-the-art
10 health monitoring or prognosis, disturbance rejection in presence of strong cross-wind and load
11
variation, etc. Bond graph models have been proved to be very useful by various researchers in
12
13
these application areas. Note that the sub-models developed in this article are generic and can be
14 assembled with appropriate constraints to model various other systems such as helicopter,
15 tricopter, quadcopter, etc. with full roll, pitch, yaw, heave, surge and sway motions.
16 This paper is organized as follows: The next section describes the TRMS set-up and
17 various forces and couples experienced by it. Section 3 presents the modelling of the TRMS
18 based on the BG approach. Section 4 discusses the results from the developed model and
19
compares with the experimental test data. Finally, conclusions and perspectives are drawn in
Fo
20
21 Section 5.
22
23 2. Description of the TRMS set-up
rP
24
25 The TRMS comprises various components as shown in Figure 1(a). The shields, DC motors and
26
ee
counterweight are rigidly attached to the supporting beam. The TRMS support beam is restrained
27
28
in such a way that it can have only two independent rotational motions; one is rotational motion
29 in the horizontal plane (yawing) and the other is rotational motion in the vertical plane (pitching).
rR
30 The global inertial coordinate system and the body-fixed frames used to model the setup are
31 shown in Figure 1(b) where points m, t, c and b are markers for assembly of various subsystems.
32
ev
33
34 (a) (b)
Flat cable Main
35 Supporting rotor
36 y
iew
beam
37 Hub m xmmr
Tail rotor shield
38 xb xmr
Tail motor
39 b
40 xt ym
yb zmr
41 ztr
42 Main rotor zm Main motor
t yt zb shield X
43 c xcb O
44 ytr
45 zt ycb
46 xtr Tail rotor zcb Counter weight Y
Z
47 Supporting
column
48
49 Figure 1: (a) photograph of the actual TRMS set-up and (b) coordinate frames for modeling
50
51 The dynamics of the TRMS resembles a hovering helicopter. The purpose of the counter-
52
weight is to balance the pitch of the TRMS in steady state. Main and tail rotors can rotate about
53
54 their own axes with respect to the support beam. These rotors are powered by two permanent
55 magnet DC motors which are controlled by the signals generated through a computer interface.
56 The revolving rotors produce aerodynamic thrust and drag forces which produce angular
57
58 4
59
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
Fo
20
21
22
23
rP
24
25
26
ee
27
28
29
rR
30
31
32
ev
33
34
35
36
iew
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
1
2
3 2.2 Interfacing
4
5
6 The input voltages to the main and tail motors of the system given in real-time workshop of
7 Simulink model (Um and Ut) are not identical to the respective actual terminal voltages (Vm and
8 8
Vt) of DC motors. It means the control signals or input voltages given in Simulink environment
9 and the respective measured terminal voltages at DC motor are not equal and there are nonlinear
10
11
interface gain/amplifier functions, as shown schematically in Figure 3. The terminal voltages of
12 the motors are measured and plotted against the input voltages (given in Simulink real-time
13 workshop) in Figures 4 and 5. The non-linear relationships between the input-output voltages are
14 expressed by 4-th order polynomial fits.
15 Vm
16 Input voltage to main motor Um f(Um) Main
17
18 Vt
Input voltage to tail motor Ut f(Ut) Tail
19
Fo
20
21 Interface gain DC motors
22 functions
23
rP
Figure 3: Block diagram of interface gain or amplifier function

24
25
26
The range of the input voltage signal to the motors given in Simulink real-time workshop is
ee
27 J8 A V to +2.5 V. The corresponding range of actual terminal voltages of the main motor is
28 J2: 67 V to +17.58 V and that of the tail motor is J27 28 V to 12.71 V.
29
rR
30
31
32
ev
33
34
35
36
iew
37
38
39
40
41
42
43
44 Figure 4: Relation between Um and Vm Figure 5: Relation between Ut and Vt
45
46 2.3 Aerodynamic propulsive forces
47
48 The thrust forces and drag forces acting on the rotors at various speeds of the rotors are obtained
49 through an experimental parameter estimation procedure. The thrust and drag forces at various
50
51
positive and negative voltage inputs or rotor speeds are measured by using an electronic balance
52 as shown in Figure 6. An initial weight is applied and the reduction in weight due to string
53 tension is measured. For reverse rotation of main rotor, a pulley arrangement is used.
54
55
56
57
58 6
59
1
2
3 Fm
4
5
6
7
8
9 String
10
11 Balance weight Electronic balancing
12 machine
13 020.02 grms
14 Electronic balancing machine
15
Figure 6: Arrangement for estimation of thrust and drag forces, and non-linear cable stiffness.
16
17
18 Similarly, pulley arrangement is used to convert horizontal force due to tail rotor actuation
19 to vertical force applied on the electronic balance.
Fo
20 The measured thrust forces for main and tail rotor at various speeds are plotted in Figs. 7
21 and 8, respectively. Least square method is adopted to obtain the thrust coefficients by using the
22 equations (6) and (7). Note that due to rotor blade twist, the drag and lift coefficients also depend
23
rP
24
on the direction of rotation.
25 The thrust force on the main rotor (Fm) is given by
26 Fmn Cmtn m for 0 (6)
ee
m m
27
28 Fmp Cmtp m m 0. for (7)m
29
In equations (6) and (7), Fmn and Fmp are the thrust forces on the main rotor for clockwise
rR
30
31 and counterclockwise rotations, respectively and Cmtn and Cmtp are two constant coefficients.
32
ev
33
34
35
36
iew
37
38
39
40
41
42
43
44
45
46
47
48 Figure 7: Thrust force versus main rotor speed Figure 8: Thrust force versus tail rotor speed
49
50 Likewise, the thrust force on the tail rotor (Ft) is given by
51
52 Ftn Cttn t t for t 0 (8)
53
54
Ftp Cttp t t for t 0 (9)
55 where Ftn and Ftp are the thrust forces on the tail rotor for clockwise and counterclockwise
56 rotations, respectively, and Cttn and Cttp are two constant coefficients.
57
58 7
59
1
2
3 The rotor blade drag forces are estimated by locking the support beam joints at its pivot and
4 1
5 then applying constant input voltages in steps. By using the DC motor characteristics and Eqs.
6 (4) and (5), the drag coefficients in main and tail rotor blades are estimated at steady blade
7 rotation speeds ( d- / dt 0 ).
8
9 2.4 Non-linear flat cable moment
10
11
12 In the TRMS set-up, a flat cable is attached to the TRMS hub, as shown in Figure 1. The cable
13 opposes angular movement of the TRMS in the horizontal plane (yawing), as it applies a
14 nonlinear moment on the hub. This nonlinear moment is difficult to estimate, as the cable
15 behaves like torsional spring at some instants and rope at other instants. The cable’s one end is
16 attached to support beam through the hub and the other end passes through a hole in the support
17
column, where the cable end can have slack movement. However, the moment due to cable can
18
19 be modeled as a restoring force developed by a non-linear torsional spring whose spring stiffness
varies with the angle of twist/yaw ( ).The stiffness Kyaw of spring is estimated with 3rd order
Fo
20
21 nonlinear equation as
22 3
23
K yaw 0.0046 0.0016 2 0.021 0.032 (10)
rP
24 The nonlinear stiffness is experimentally measured by using electronic balancing machine

25 at various yaw angle positions of the beam. The restoring yaw moment is plotted versus the yaw
26
ee
angle and a fourth order best-fit curve is generated. The first derivative of the fitted expression
27
28
for yaw moment with respect to the yaw angle is then used to define the non-linear relation given
29 in equation (10). Note that the torsion spring stiffness decreases with increase in yaw angle up to
rR
30 1.4 radians and thereafter it increases.

31
32 2.5 Torque due to the viscous damping force
ev
33
34 Air resisting force on the TRMS (excluding the rotor blades) due to its movement in the air can
35
36
be modeled as a viscous damping force for small angular velocities (pitch and yaw rates). The
iew
37 viscous damping forces produce damping torques (Th and Tv) in the horizontal and vertical
38 planes, with the damping coefficients as r and r , respectively, as given below
39
40 = sign( ) (11)
41 = sign( ) (12)
42 The damping coefficients are tuned in the model to match the response time, decay rate and
43 phase of the simulated and experimental responses.
44
45
46 3. Modeling of the TRMS dynamics
47
48 The TRMS model is described in the Cartesian coordinate system by considering the horizontal
49 position of the support beam as an inertial frame of reference aligned in such a way that the z-
50 axis is vertical, the y-axis is normal to the axis of the support beam and x-axis is parallel to the
51
axis of the support beam. Rotor thrust and drag forces are applied on the rotors in the body frame
52
53 of reference of the rotors. The thrust forces generated by main and tail rotors act along the body
54 fixed z-axis and y-axis, respectively. The viscous damping forces on the supporting beam in the
55 horizontal plane (yawing) and the vertical plane (pitching) are applied in the inertial frame of
56 reference.
57
58 8
59
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
Fo
20
21
22
23
rP
24
25
26
ee
27
28
29
rR
30
31
32
ev
33
34
35
36
iew
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
1
2
3 defined about the coordinate axes which are attached to the center of gravity and aligned to the
4
5
principal axes of the body. The sub-models are integrated through co-ordinate transformation
6 sub-models.
7
8 3.2 Coordinate transformation sub-model (CTF)
9
10 Rigid body dynamics in a multi-body system is generally modeled in the body-fixed reference
11
frame aligned with the principal axes so that the Euler’s equations do not contain product of
12
13 inertia terms. However, constraints between two or more rigid bodies with different body-fixed
14 frames are implemented by transforming all motions to a common reference frame, i.e., the
42
15 inertial frame. The constraint forces and moments are then transformed back to the respective
16 body-fixed frames. Coordinate transformation (CTF) sub-model is used for transformation of
17
linear and angular velocities from the body-fixed frame of reference to the inertial frame of
18
19 reference. This transformation is performed by simple multiplication with the transformation
matrix
Fo
20
21 c c s s c c s c s c s s
22
23 CTF c s s s s c c c s s s c (13)
rP
, ,
24 s s c c c
25
26
ee
27 where c is shorthand for cos , s for sin and so on. Note that , and are Z-Y-X Euler
28 angles (yaw, pitch, roll angles). This transformation is modeled as shown in Fig. 10
29
rR
30
31
Moving or body frame CTF Fixed or inertia frame
32
ev
33
34 Figure 10: Schematic diagram of coordinate transformation from moving frame to fixed frame.
35
36
iew
The bond graph in Figure 10 models two things together: it transforms body-fixed
37 velocities to inertial frame velocities and transforms inertial frame forces to body-fixed frame
38 forces. The detailed model is constructed with a set of TF, 0 and 1 junctions and it is a power
39 43
40 conserving transformation. The moduli of TF elements used inside the CTF sub-model are the
41 elements of the matrix given in equation (13).
42
43 Similarly, CTF for transformation of velocities from the inertial frame to the body-fixed
44 frame of reference is given by
45
46 c c s c s
47 CTF 1
, , c s s s c s s s c c c s (14)
48
49 c s c s s c s s s c c c
50
51 This inverse transformation is modeled as shown in Figure 11. It also models
52 transformation of forces from body-fixed frame to inertial frame and is implemented internally
53 through a power conserving junction structure composed of TF, 0 and 1 elements.
54
55
56
57
58 10
59
1
2
3
4 Fixed or inertia frame CTF Moving or body frame
5
6
7
Figure 11: Schematic diagram of coordinate transformation from fixed frame to moving frame.
8
9 The model also requires the Euler angles defining the orientation of body-fixed frame with
10 respect to the inertial frame. The Euler angle rates are computed from the angular velocities in
11 43
the body-fixed frame through a transformation called Euler Angle Transformation (EATF). The
12 42
13 rate of change of Euler angles are integrated with respect to time to obtain the Euler angles.
14 The detailed modeling of CTF and EATF sub-model structures can be consulted in standard text
15 books on the subject
13, 14, 43
and archival literature.
18, 42
16
17
18 3.3 Dynamics of the support beam
19
Fo
20 The support beam acts as a base for all the components of the TRMS and all components are
21 mounted on it. The beam is pivoted to the vertical static supporting column as shown in Fig.
22 1.The joint is articulated in such a way that the beam can rotate about the inertial vertical axis
23
rP
(i.e. about the Z-axis or yawing) and the horizontal axis which is normal to the beam axis, i.e. the
24
25
body fixed y-axis (pitching). Thus, the supporting beam is having two degrees of freedom and its
26 configuration at any instant can be defined by yaw angle ( ) and pitch angle ( ). The rigid
ee
27 body model of the support beam is shown in Fig. 12. To get only this 2 DOF motion, four
28 constrains are applied. The beam translational motions along three axes are constrained by pivot
29 at the fixed vertical column. This constraint is included in the BG model through pads #+J2J
rR
30
31
structures) in Fig. 13. The rotational motion of the beam about its own axis is also constrained.
32 The Euler’s third rotation angle or roll angle is locked by a pad as shown in Fig. 12. Note that
these constraints can be removed from the model when modeling a free-flying object like a
ev
33
34 helicopter, tricopter or quadcopter.
35 Coordinate transformation of the different points on the beam has been made to integrate
36
iew
(or constrain) the model with other components of the TRMS. The weight of the beam is applied
37 in the inertial reference frame. The position of support beam is controlled by thrust forces
38
39
applied by the rotors attached at the ends of the beam. Instead of considering a specific case of
40 TRMS, we develop the model of a generalized floating body and then constrain it appropriately.
41 Thus, the model developed here can be modified to represent the main body of other systems
42 such as a helicopter or a quad-copter by relaxing the constraints.
43
44 Newton-Euler equations for spatial dynamics of the rigid beam in the scalar form are
45
46 = + ( ) (15)
47 = + ( ) (16)
48
49
= + ( ) (17)
50 and
51 = ( ) (18)
52 = ( ) (19)
53
54
= ( ) (20)
55
56
57
58 11
59
1
2
3 In equations (15) through (20), x, y, z are the displacements of supporting beam along the
4
5
principal axes and their successive derivatives are linear velocities and accelerations.
T
6 - -x -y -z and =[ ] are angular velocity and angular acceleration
7
vectors, respectively, m is mass of the supporting beam, I x , I y and I z are mass moment of inertias
8
9 of the supporting beam about its principal axes, and F Fx Fy Fz
T
and
10
T
11 M Mx My Mz are the force and moment vectors, respectively. These equations are
12
13 represented in a bond graph model as shown in Figure 12, whose central or core part containing
13, 42, 43
14 six GY elements is often called an Euler junction structure (EJS). Rotational inertias of the
15 body are defined in the body fixed coordinate frame, which are fixed at the centre of gravity of
16 the body and its axes are aligned along the principal axes of the body.
17
C 1 R
18 0
19 Sf 0 1Q
.
Fo
20 Velocities at point ‘p’ on .R:rY Euler Angle
support beam S Transformation
21 (To articulation joint with S T Df 1S
.
1 function
P x Py Pz
22 support column, Fig.13) R:rR
8 O 9 .
23
R
rP
R T Df 1R. K:Kyaw
24
25 z
TF ..p
26 1/. .yc
ee
1/ym z. .t TF
27 TF
1/. y. t TF
28
1Pz z. . c 1Py
1/yp TF TF
zm
29 ..
rR
TF TF
30 0 0 0 0
31
32 mb:I 18b I:Jby
(To tail motor and shield, Fig.15)
(To main motor and shield, Fig.16)

ev
33 JbxPx
Velocities at point ‘t’ on
34 1Pz 1Py
Velocities at point ‘m’ on

GY
support beam
8 8
35 O 1Pz x O
support beam
U ..p 1/xm 1Py U
36
iew
Px TF 0 I:Jbz I:mb Px
Py 0 TF Py
37 Pz x. . m mbPx 1/xp Pz
38 TF 0 1Ob GY
I:Jbx
1 9b 0 TF
39 x. .t 1/xt
I:mb 0
40 TF 0 1Px TF 1/x
x. . c c
41 Wb TF 0 0 TF
m
TF yt
yp TF
.. t
TF
Xb
42
TF 1/Z
T Df 1W
TF y
1/Z TF
.. p
yc
TF
1/z
..c
Vb
TF
43
..m
Yb T Df 1V
1/Z
CTFBI
44 Xb
1Px 1Px
Zb T Df
45 1X
46
47 Se: mbg
8 O PxPyPz
U
48 Velocities at point ‘c’ on
49 support beam
(To counterweight, Fig.14)
50
51 Figure12: Bond graph model of TRMS Support beam
52
53 A point where one body is connected to another body is defined in its own body frame
54 since the points positions are fixed in the body fixed frame. The velocity of such a point p ( xp, yp,
55
56
57
58 12
59
1
2
3 zp) in the body frame (i.e., an inertial frame momentarily aligned with body-fixed frame) is
4
5
computed by the following expressions:
6 = + (21)
7
= + (22)
8
9 = + (23)
10 where subscript G indicates velocities at the CG and = , = and = .
11
12
13
These body frame point velocities are then transformed into the inertial frame velocities
14 through coordinate transformation (CTF). In Figure 12, four such points are modeled as
15 indicated within the rectangular boxes. The x, y and z variables appearing in the various
16 transformer moduli indicate the positions from the CG of the TRMS beam the subscript to those
17 coordinates indicates the point, e.g, p (pivot), t (tail motor and shield CG), c (counterweight CG),
18 and m (main rotor and shield CG). Euler angles are used in the CTF. Euler three dimensional z-y-
19
x axis rotation sequences have been considered in this article. In the Euler angle rotation
Fo
20
21 sequence, firstly the support beam rotates about the fixed inertial z-axis (yaw angle), followed by
22 a second rotation about the rotated y-axis (pitch angle) and finally the third rotation about body
43
23
rP
fixed x-axis (rolling). A quaternions representation can be used for CTF which does not suffer
24 from gimbal lock as happens in Euler angles representation. Although Euler angle is used in this
25
work as pitch angle (i.e. Euler second rotation angle) would never be 90°. So, this type of
26
ee
27 singularity will not arise.

28 The Euler angle transformations are shown in the top-right corner of Figure 12. But due
29 to the articulated joint between the beam and support column, the beam cannot rotate about body
rR
30 fixed x-axis which is aligned with the axis of the beam. So, the second Euler angle of rotation
31 will be the rotation about the body fixed y-axis. In order to impose the articulated joint constraint
32 on the beam, a high stiffness pad has been inserted at the Euler third angle rotation as shown in
ev
33
Figure12. In a helicopter, drone or quad-copter, this constraint on the third Euler angle is absent
34
35 and the rotor positions differ. The general model developed here can be adapted appropriately for
36 such flying objects.
iew
37 Similarly to the support beam, the remaining five rigid body sub-models are developed.
38 These sub-models are connected with the supporting beam as per the physical structure of the
39 TRMS plant.
40
41
42
3.4 Support Column and Joints
43
44 The support beam is pivoted to the support column at the articulation point. At this point, relative
45 translation between the bodies is zero. Assuming a small flexibility/clearance at this point, the
46 bond graph model of the support column is developed in Figure 13. The zero flow sources (SF:0)
47 indicate ground fixation and the pad structures (R-1-C) indicate joint flexibility. To avoid too
48
much joint flexibility, the values of pad parameters CP and RP are kept large. For clearance
49
43
50 joints, the contact/joint flexibility needs to be varied with contact conditions based on the
51 positions of the inner and outer races at the joint.
52
53
54
55
56
57
58 13
59
1
2
3
vx = 0
4 ..
Sf 0 18
5
.R.P C
6 . .P
R 1 C
7 Velocities at point ‘p’ on
vy = 0
8 .. supporting beam
Sf 0 1O
9 (Form supporting beam, Fig.12)
10 .R.P .C.P
v..z= 0 R 1 C
11
12 Sf 0 1[
13 .R.P .C.P
14 R 1 C
15
16 Figure 13: Bond graph model of beam support
17
18 3.5 Counterweight, main and tail shield, main and tail rotors, and DC motor
19
Fo
20 The remaining parts of the TRMS have been modeled as rigid bodies by using EJS as shown in
21
22
/% 2@J29 In the main and tail shield models and counter weight model, high stiffness pads
23 are inserted at the revolute joints so that the parts are fixed rigidly to the support beam.
rP
24 CP
..
RP
..
25 C 1 R
26
ee
Cp:C 1 R:Rp
27 mc
0
Velocities at point ‘c’ on supporting beam

28 Vx
I 18 0
29
(Form supporting beam, Fig.12)

rR
I:Jcy Cp:C 1 R:Rp

30 JcxPx
31 1Pz GY 1Py 0
CTFIB 0 CTFBI Vy
32 RP CP Cp:C 1 R:Rp
.. ..
ev
33 R 1 C
34 I:Jcz
I:mc 1 0 Vz
35 mcr g
mc mcPx ..
36
iew
Se Px
I 1O GY 1[
37 RP
..
CP
..
38 Jcx
R 1 C
CTFIB CTFBI Py
39 I 1Px 0
40 Pz
41
42
Figure 14: Bond graph model of TRMS counterweight with joint
43
44
45 In the main and tail rotor BG models (Figures 17 and 18), drag and the thrust forces due to
46 the rotor are modeled as per equations #A$J#6$ According to the relative motions between the
47 rotors and the support beam, constraints are imposed at the joints. The main rotor and tail rotors
48 can rotate only about their own axis as shown in Figure 1, with respect to supporting beam. Thus,
49 to allow the rotational motion at the revolute joint of the rotor, pad structure is removed from the
50
51
model and DC motor models are connected there. The DC motors are modeled in the same body
52 frame of reference.
53
54
55
56
57
58 14
59
1
2
3
Velocities at CG of
4
tail motor and shield
5
(To tail rotor , Fig.18)
6
Vx VyVz Px Py Pz
7 CP RP
8 .. ..
C 1 R
9
10 Cp:C 1 R:Rp
0
11 mt
Velocities at point ‘t’ on supporting beam

12 I 18 0 Vx

13 I:Jty Cp:C 1 R:Rp
JtxPx
14 0
1Pz GY 1Py CTFIB 0 CTFBI Vy
15 RP CP Cp:C 1 R:Rp
16 ..
R
..
C
1
17 I:Jtz 1 0
I:mt Vz
18 mt g
mtPx ..
19 Se
Fo
Px
20 1O GY 1[ RP CP
.. ..
21 R 1 C Py
I:mt I:Jtx CTFIB CTFBI
22
1Px
23 0
rP
Pz
24
25
26
ee
Figure 15: Bond graph model of TRMS tail rotor shield with joint
27
28
29 Velocities at CG of
rR
30 main motor and shield

31 (To main rotor, Fig.17)
32 Vx VyVz Px Py Pz
ev
33 CP
..
RP
..
34 C 1 R
35
Cp:C 1 R:Rp
36
iew
Velocities at point ‘m’ on supporting beam

37 0 Vx
18
38 Cp:C 1 R:Rp

I:mm I:Jmy
39 JmxPx
40 1Pz GY 1Py 0 CTFIB 0 CTFBI Vy
41 RP
.. CP
.. Cp:C 1 R:Rp
42 R 1 C
I:Jmz
43 I:mm 1 0 Vz
mmg
44 mmPx ..
Se
45 1O GY 1[ Px
RP CP
46 .. ..
R 1 C
47 I:mm I:Jmx CTFIB CTFBI Py
48 1Px 0
49 Pz
50
51
52
53 Figure 16: Bond graph model of TRMS main rotor shield with joint
54
55
56
57
58 15
59
1
2
3
4 Cp:C 1 R:Rp
Velocities at CG of main motor and shield

5 .m. mr
(From main motor and shield, Fig.16)

6 I 18 0 Vx
I:Jmry
7 Cp:C 1 R:Rp
JmrxPx
8 0 CTFBI Vy
1Pz 1Py CTFIB 0
9 GY
1
RP CP Cp:C R:Rp
10 .. ..
R 1 C
11 R:Cmv I:Jmrz 1 0 Vz
I:mmr
12 mmr g
..
13 m
. .mr
mmrPx SE: Cmt Pz Pz
Se
14 I 1O GY 1[ Px
R
15 .C.P . .P
C 1 R Py
CTFIB CTFBI
16 .J.mrx 1Px
17 I 0 Pz
18 0
19
I:Lm
Fo
20
Km Q
21 .. Vm
..
22 1 GY 1 Se
23
rP
24 DC motor R:Rm
25
Figure 17: Bond graph model of TRMS main rotor with joint
26
ee
27
28 The 0 junction (equal effort and flow sum junction) at the left of the model in Figure 17
29 indicates that the main rotor angular velocity is the sum of the main motor speed and yaw rate of
rR
30 the support beam (same as shield). The basic property of the bond graph (equal effort at 0
31 junction) also ensures that the motor torque is applied on the rotor and equal reaction torque in
32 opposite direction is simultaneously applied on the support beam. Likewise, the action and
ev
33
34
reaction torques at the tail rotor due to the DC motor are modeled at the top-right corner of
35 Figure 18.
36 The unique causal structure of bond graphs and properties of the junction structure ensure
iew
37 that proper physics is represented in the models. Moreover, the force/moment balance equations
38 and kinematic constraints are simultaneously satisfied by the model due to the power conserving
39 properties of the bond graph junction structure.
40
The integrated BG model is developed by assembling the sub-models as described in the
41
42 word bond graph (Figure 9). The causality assignment imposes a computation structure and the
43 equations of motion are then derived from the model in the state-space form through well-known
44 algorithms which have been implemented in several software. These equations of motion are
45 then solved (simulated) with the parameters given in Table 1 and the results obtained are
46 22
presented in section 4. In BG software, the equations of motion are written in an optimized
47
48 form as C++ language functions which are then compiled to a dynamic link library (DLL) with
14
49 further optimization by Microsoft or Borland compiler. The DLL is linked to the simulator
50 module of the software at runtime. This speeds up the numerical solution.
51
52
53
54
55
56
57
58 16
59
1
2
3
I:Lt
4
Kt Q V
5 .. . .t
0 1 GY 1 Se
6
7 ..Cp ..Rp DC motor
8 C 1 R R:Rt
9 Cp:C 1 R:Rp
10 0
Velocities at CG of tail motor and shield

11
(From tail motor and shield, Fig.15)

I:mtr 1] 0 Vx
12 Cp:C 1 R:Rp
13 JtrxPx
I:Jtry
14 1Pz GY 1Py CTFIB 0 CTFBI Vy
15 R:Ctv Cp:C 1 R:Rp
16 I:Jtrz
R: Ctd |Py |
1 0 Vz
17 mtr g
18 mtrPx ..
Se
19 I:mtr
1O GY 1[ Cp Rp
Px
Fo
20 Ctt |Py| Py .. ..
C 1 R
CTFBI Py
21 Se
I:mtr
CTFBI
22 I:Jtrx 1Px 0
Pz
23
rP
24
25 Figure 18: Bond graph model of TRMS tail rotor
26
ee
27
28
4. Results and discussion
29 The BG model is simulated by using the parameters given in Table 1. The mass and geometric
rR
30
31
parameters are taken from references8,12, TRMS manual and measurements.
32
Table 1: TRMS parameters
ev
33
34 lt = 0.282 m Jty = 1.209×10-4 kg.m2 Ctv = 2×10-5 N.m.s/rad
35 lm = 0.246 m -4
Jtz = 6.457×10 kg.m 2 Cmtp =1.2×10-5 N.s2/rad2
36
iew
lcb = 0.2182 m -7
Jbx = 14.2×10 kg.m 2 Cmtn= 6.8×10-6 N.s2/rad2
37
38
lb = 0.018 m Jby = 7.066×10-4 kg.m2 Cttp= 4×10-6 N.s2/rad2
rms = 0.155 m -4
Jbz = 7.066×10 kg.m 2 Cttn = 2.48×10-6 N.s2/rad2
39
40 rts = 0.1 m Jcx = 3.22×10-4 kg.m2 Cdm = 2.4×10-7 N.m.s2/rad2
41 mtr = 0.01426 kg -4
Jcy= 3.22×10 kg.m 2 Cdt =1.08×10-8 N.m.s2/rad2
42 mmr= 0.0385 kg Jcz = 3.51×10-6 kg.m2 Ram = 8 `
43 mcb= 0.0686 kg Jmrx = 2.02×10-6 kg.m2 Rat = 8 `
44 -4 2
mt= 0.3358 kg Jmry = 2.52×10 kg.m Lam = 0.86 mH
45
46 mm = 0.4199 kg Jmrz = 2.54×10-4 kg.m2 Lat = 0.86 mH
47 mb= 0.01782 kg Jtrx = 3.9×10-7 kg.m2 Kam = 0.0202 N.m/A
48 Jmx = 2.679×10-3 kg.m2 -5
Jtry = 3.39×10 kg.m 2 Kat = 0.0202 N.m/A
49 Jmy = 2.679×10-3 kg.m2 Jtrz =3.37×10-5 kg.m2 r; = 0.0048 N.m.s/rad
50 Jmz = 5.226×10-3 kg.m2 -5
Cmv = 3.8×10 Nms/rad r< = 0.022 N.m.s/rad
51 Jtx= 6.457×10-4 kg.m2 Cp = 105 N/m or 10 N.m/rad Rp =1 N.s/m or 0.01 N.m.s/rad
52
xp = JK K29 m yp = JK KA m zp = 0 m
53
54 xc = JK K29 m yc = 0 m zc = 0.2182 m
55 xm = 0.264 m ym = 0 m z m= 0 m
56 xt = JK 8<@ m yt = 0 m zt = 0 m
57
58 17
59
1
2
3 The responses from both the models for various input voltages (open-loop system) are
4
5
compared with the test data. The cross coupling effect is observed in Figures19 through 22 when
6 only main rotor is driven. The responses for the constant input voltage of 1.2V applied to the
7 main DC motor are compared in Figures 19 and 20. In these responses, it can be observed that
8 the yaw angle also gets affected along with the pitch angle. The yaw angle reaches a steady state
9 as shown in Figure 20, but the actual system exhibits continued oscillations about the mean
10 position.
11
12
13
0.6
Test data Bond graph
Pitch angle (rad)
14 0.5
15
0.4
16
17 0.3
18
0.2
19
Fo
20 0.1
21 0.0
22 0 10 20 30 40 50 60
23
rP
24 Time (s)
25 Figure19: Beam pitch angle response for input amplitude of 1.2 V to the main motor
26
ee
27 1.4
28
1.2
Yaw angle (rad)
29
rR
30 1.0
31 0.8
Test data
32 0.6 Bond graph
ev
33
0.4
34
35 0.2
36
iew
0.0
37 0 10 20 30 40 50 60
38 Time (s)
39
40 Figure 20: Beam yaw angle response corresponding to input amplitude of 1.2 V supplied to the
41 main motor
42
43 0.6 Test data Bondgraph
44
0.4
Pitch angle (rad)
45
46 0.2
47
0.0
48
49 -0.2
50 -0.4
51
52 -0.6
53 0 10 20 30 40 50 60 70 80 90 100
54 Time (s)
55 Figure 21: Beam pitch angle response corresponding to sinusoidal input with a frequency of 0.16
56 Hz and amplitude of 1.5 V only supplied to the main motor
57
58 18
59
1
2
3 1.0
4 Test data Bondgraph
5
Yaw angle (rad)
0.5
6
7
8 0.0
9
10 -0.5
11
12 -1.0
13 0 10 20 30 40 50 60 70 80 90 100
14 Time (s)
15 Figure 22: Beam yaw angle response corresponding to sinusoidal input with a frequency of
16 0.16Hz and amplitude of 1.5 V supplied to the main motor.
17
18
19 Similarly, the responses to the sinusoidal input voltage of magnitude 1.5 V with a
frequency of 0.16Hz given to the main motor are shown in Figures 21 and 22.
Fo
20
21 In the next experiment, the yaw angle is varied by giving a sinusoidal input of 0.6 V
22 amplitude and 0.08 Hz frequency to the tail motor and the responses for the same are shown in
23
rP
Figure 23. The pitch angle response for the input is not plotted as it is insignificant. This is due to
24 small frequency and amplitude given to the tail motor which produces negligible cross coupling.
25
26
It can be observed from Fig.24 that the pitch angle response from the model is similar to that
ee
27 of the test data when low frequency input volatage is supplied to the main motor. As shown in
28 Fig. 25, the fundmental frequency and amplitude of the pitch angle is close to the test data when
29 moderate frquency input signal supplied to both the motors. From the Figs.26 and 28, it is clear
rR
30 that the yaw angle response of the model also shows good match with the test data. As shown in
31 Figs. 27 and 28, the amplitudes of the pitch and yaw angle deviations are low when the input
32 voltage frequency is high. This happens due to slow mechanical response time.
ev
33
34
35 1.8
36
iew
Yaw angle (rad)
1.2
37
38 0.6
39 0.0
40
41
-0.6
42 -1.2
43
-1.8
44 0 10 20 30 40 50 60 70 80 90 100
45
46 Time (S)
47 Figure 23: Beam yaw angle response corresponding to sinusoidal input with a frequency of
48 0.08Hz and amplitude of 0.6 V supplied to the tail motor.
49
50
51
In this article, the objective is to develop a complete physics based model without
52 introduction of ad-hoc elements which cannot be explained from the physics of the system.
53 Therefore, the small errors between the experimental and simulated results are acceptable. It is
54 noteworthy to mention here that the developed model will be used for implementation of model-
55 based control. If the model and the system responses are exact then the controller robustness
56
57
58 19
59
1
2
3 against the modeling uncertainties and external disturbances cannot be ascertained. As an
4
5
example, in a data-driven model such as an artificial neural network (ANN), over fitting can give
6 a very small training error but the model becomes less robust and may give a large testing error.
7 A good robust controller should be able to overwhelm the small modeling errors.
8 0.4
Test data
9 0.3
Pitch angle (rad)
Bondgraph
10 0.2
11
12 0.1
13 0.0
14 -0.1
15 -0.2
16
17 -0.3
0 10 20 30 40 50 60 70 80 90 100
18
19
Time (s)
Figure 24:Beam pitch angle response corresponding to sinusoidal input with a frequency of 0.02
Fo
20
21 Hz and amplitude of 1.0 V supplied to the main motor.
22
23
rP
0.4
Test data Bondgraph
24 0.3
Pitch angle (rad)
25
26 0.2
ee
27 0.1
28 0.0
29
rR
-0.1
30
31 -0.2
32 -0.3
ev
33 0 10 20 30 40 50 60 70 80 90 100
34 Time (s)
35 +
36
iew
Figure 25: Beam pitch angle response corresponding to sine input with a frequency of 0.05 Hz to
37
38
the main motor (amplitude of 0.8 V) and to the tail motor (amplitude of 0.5 V).
39
40
41 2.0
Test Bond graph
42 1.5
Yaw angle (rad)
43 1.0
44 0.5
45 0.0
46
-0.5
47
48 -1.0
49 -1.5
50 -2.0
51 0 10 20 30 40 50 60 70 80 90 100
52 Time (s)
53
Figure 26: Beam yaw angle response corresponding to sine input with a frequency of 0.05 Hz to
54
55 the main motor (amplitude of 0.8 V) and to the tail motor (amplitude of 0.5 V).
56
57
58 20
59
1
2
3 0.06
4 Test data Bond graph
5 0.04
Pitch angle (rad)
6 0.02
7
8 0.00
9 -0.02
10
11
-0.04
12 -0.06
13 0 10 20 30 40 50 60 70 80
14 Time (s)
15
16 Figure 27: Beam pitch angle response corresponding to sine input with a frequency of 1 Hz and
17 amplitude of 2.5 V to both the main and the tail motor.
18
19 0.3
Fo
20
0.2
Yaw angle (rad)
21
22 0.1
23
rP
0.0
24
25 -0.1
26
ee
-0.2
27
28 -0.3
0 10 20 30 40 50 60 70 80
29
rR
30 Time (s)
31 Figure 28: Beam yaw angle corresponding to sine input with a frequency of 1 Hz and amplitude
32 of 2.5 V to both the main and the tail motor.
ev
33
34
35
These results indicate that when the yaw angle is small, there is a very good match
36 between the test-bed and simulated results, as in Figure 28. For large yaw angles (above 0.6 rad),
iew
37 the non-linear effects become predominant, as shown in the step response given in Figures 19, 20
38 and elsewhere. When the frequency and amplitude of sinusoidal excitation are increased, the
39 peak accelerations (proportional to square of input frequency) of the rigid bodies increase. This
40 causes significantly large inertial forces and thus, more deviations in the model response in
41
comparison to the experiments. This is evident from the results shown in Figures 24 and 25.
42
43
Also, equal change in tail motor voltage produces less effect as compared to same change in
44 main motor voltage due to the non-linear gains shown in Figures 4 and 5.
45 Next, a proportional-integral-derivative (PID) control is implemented on the model and
46 the TRMS plant to control both the pitch and yaw angles. A schematic diagram of the PID
47 controller is shown in Figure 29, where Kp, Ki, and Kd indicate the proportional, integral and
48 derivative gains of the controller, respectively. The time constant, C, in the controller transfer
49
function block is related to the cut-off frequency. The same PID controller (Fig. 29) is used as
50
51 the TRMS setup default controller and hence, a BG model for the same controller structure is
52 developed. In the derivative part of the PID model, the time constant C avoids derivatives of
53 high-frequency components which can introduce increased measurement noise sensitivity. The
54 default TRMS controller parameter values for the pitch angle and yaw angle stabilization are
55 given in Table 2 and the same values are used in the model.
56
57
58 21
59
1
2
3
4
5 e(t) Ki s Um(t), Ut(t) BG Model/
Kp Kd
6 des , des + s C*s 1 TRMS
-
7
8 t , t
9
10
11 Figure 29: Block diagram of PID control of the model.
12
13 Table 2: PID controller parameters for the TRMS
14 Pitch angle controller gains Yaw angle controller gains
15
16 Kp Ki Kd Kp Ki Kd
17 5 8 10 4 3 5
18
19 The desired trajectory is specified as 0 rad yaw angle and 0.3sin(0.2t) rad pitch angle,
Fo
20 which have to be simultaneously tracked. Figures 30 and 31 show that the responses from the
21
BG model and the physical system track the desired trajectory with reasonable accuracy. The
22
23 yaw angle deviates from the desired position (0 rad) due to the strong cross coupling present
rP
24 between the rotors, as shown in Fig. 31. Further tuning of the controller gain values through least
25 square error minimization, especially to reduce the yaw angle disturbance, did not yield
26 significant improvement.
ee
27
28 0.4 Reference Test data Bond graph
29
Pitch angle (rad)
rR
0.3
30
31
0.2
32 0.1
0.0
ev
33
34 -0.1
35 -0.2
36
iew
-0.3
37 0 10 20 30 40 50 60 70 80 90 100
38
39 Time (s)
40 Figure 30: Closed loop pitch angle response of TRMS for reference sinusoidal pitching.
41 0.20
42 Reference Test data Bond graph
0.15
Yaw angle (rad)
43
44 0.10
45 0.05
46 0.00
47
48
-0.05
49 -0.10
50 -0.15
51 0 10 20 30 40 50 60 70 80 90 100
52 Time (s)
53
54 Figure 31: Closed loop yaw angle response of TRMS for reference sinusoidal pitching..
55
56
57
58 22
59
1
2
3 Next, the closed loop responses of the controller for set-points 0.5 rad pitch angle (i.e.
4
5
constant pitch) and 0 rad yaw angle are plotted in Figures 32 and 33. Both the BG model and the
6 TRMS plant are found to be tracking the set-points with sufficient accuracy after about 10s.
7 0.8
Reference Test data Bond graph
8 0.7
Pitch angle (rad)
9 0.6
10 0.5
11 0.4
12 0.3
13 0.2
14
0.1
15
16
0.0
0 5 10 15 20 25 30 35 40 45 50
17
18 Time (s)
19 Figure 32: Closed loop pitch angle response of TRMS for constant reference pitch angle.
Fo
20
21
22 0.2 Reference
Yaw angle (rad)
23 Test data
rP
24 Bond graph
0.1
25
26
ee
27 0.0
28
29 -0.1
rR
30
31 0 5 10 15 20 25 30 35 40 45 50
32
ev
33 Time (s)
34 Figure 33: Closed loop yaw angle response of TRMS for constant reference pitch angle.
35
36
iew
5. Conclusions
37
38
39 In this work, a dynamic model of the TRMS by using bond graphs (BG) has been developed. In
40 the developed BG model, various parts of the TRMS have been modeled separately and then
41 integrated. The sub-models developed in the bond graph are assembled with appropriate
42 constraints between the rigid bodies. Different ways of assembly can yield different other system
43 configurations such as helicopter, tricopter and quadcopter models. Aerodynamic propulsive
44
forces, drag forces, viscous damping forces, and nonlinear stiffness due to the flat cable are
45
46 considered in the developed models. The responses are compared with that of the test-bed and it
47 is found that the model is in good agreement. It is also observed that the pitch angle response is
48 unaffected by the nonlinear nature of the flat cable. The open loop responses of the model and
49 the real TRMS test bed are similar for various input voltages. A PID controller is then
50 implemented on the model as well as the real TRMS test bed and closed loop step responses
51 from the model and the test bed show simultaneous pitch and yaw angle set-point tracking.
52
53
The following conclusions can be drawn from the present study.
54
55 In BG method, the complete system is modeled explicitly with the controller and electrical
56 sub-system (DC motor). The BG model is open ended and can be converted to a block diagram
57
58 23
59
1
2
3 form or direct simulation code from the equations of state and those can be solved with various
4
5
numerical integration procedures. The equations derived from the bond graph model can be
6 viewed by the modeler to detect any modeling errors and also to optimize the simulation code.
7 Bond graph software performs code level optimization of equations for faster simulation.
8
The BG model is developed in a modular form with sub-system models for individual parts.
9
10 Those sub-system models can be assembled differently with appropriate constraints to model
11 other similar systems such as helicopter, tricopter, quadcopter and other flying objects with full
12 six-degrees-of-freedom (roll, pitch, yaw, surge, sway and heave).
13
14
Model-based real-time control requires interfacing with the hardware. MATLAB Simulink
15 already has such in-built interfacing facility. Simulink Coder (formerly Real-Time Workshop)
16 generates and executes C and C++ code from Simulink models. On the other hand, bond graph
17 software generates C++ code in which hardware interface needs to be separately coded.
18
19 However, a reduced order BG model can be developed for faster response without
Fo
20 compromising with the essential system dynamics, which is useful for BG model based
21 trajectory control. Reduced order models are developed through the use of energy-based metrics
22 called junction activity index. The parts of the model which do not show sufficient energetic
23
rP
interaction (power flow through bonds) with the rest of the system can be removed from the
24 44
25 model. Energy-based model order reduction is not possible with CAD based software, where
26 other approaches like singular value decomposition may be used on linearized models to identify
ee
27 the subsystems with least significant dynamics.

28
29 The BG model developed in this article starts with full six-degrees-of-freedom sub-models
rR
30 for rigid bodies and then some degrees of freedom are implicitly constrained for the TRMS. This
31 has been done purposefully so that the same sub-models can be used to develop other system
32 models. On the other hand, direct or explicit constraint implementation at the beginning can
ev
33 provide a lean or reduced order model. The same reduced order model can as well be derived
34 from the full model by removing all junctions and elements with almost zero power transfer. For
35
36
example, the no support bar roll assumption may be used to remove the 1-junctions for the roll
iew
37 angular velocity from all the EJS and then the corresponding GY elements in those EJS can be
38 removed. The EATF and pads used for implementing the zero-roll constraint may then be
39 removed to obtain a simpler model structure specific to the TRMS application.
40
41 Once the BG models are available in template or sub-model form, they can be iconized as
14
42 objects called capsules and users with little knowledge in bond graph modeling can use those
43 sub-models for other system models such as for helicopter, tricopter and quadcopter.
44
45 In view of the afore-mentioned discussions, model order reduction, real-time controller
46 development, fault diagnosis and fault tolerant control, etc. can be considered for future
47 extension of this work and its derivative problems in the field of surveillance drone systems.
48
49 Funding Statement
50
51 This research received no specific funding from any source.
52
53 Declaration of Conflicting Interests
54
55 The Author(s) declare(s) that there is no conflict of interest.
56
57
58 24
59
1
2
3 References
4
5 [1] Rahideh A, Shaheed MH and Huijberts HJC. Dynamic modelling of a TRMS using
6 analytical and empirical approaches. Control Engineering Practice, 2008; 16(3): 241–259.
7
8 [2] Toha SF and Tokhi MO. ANFIS modelling of a twin rotor system using particle swarm
9 optimisation and RLS. In: 9th International Conference on Cyberntic Intelligent Systems
10 2010, pp. 1–6. IEEE.
11
12 [3] Ahmad SM, Chipperfield AJ and Tokhi MO. Dynamic modelling and open-loop control of
13 a twin rotor multi-input multi-output system. Proc IMechE, Part I: J. Systems and Control
14 Engineering, 2002;216(6): 477–496.
15
16 [4] Ahmad SM, Shaheed MH, Chipperfield AJ and Tokhi MO. Non-linear modelling of a one-
17 degree-of-freedom twin-rotor multi-input multi-output system using radial basis function
18 networks. Proc IMechE, Part G: J Aerospace Engineering, 2002; 216(4):197-208.
19
[5] Rahideh A, Safavi AA and Shaheed MH. NN-based modelling of a 2DOF TRMS using
Fo
20
21 RPROP learning algorithm. In: European Control Conference (ECC), 2007; pp. 2648-2654.
22
[6] Toha SF and Tokhi MO. Real-coded genetic algorithm for parametric modelling of a
23
rP
24 TRMS. In: Congress on Evolutionary Computation, 2009; pp. 2022-2028. IEEE.

25 [7] Toha SF, Latiff IA, Mohamad M and Tokhi MO. Parametric Modelling of a TRMS Using
26
ee
Dynamic Spread Factor Particle Swarm Optimisation. In: 11th International Conference on
27
28
Computer Modelling and Simulation, 2009 Mar 25; pp. 95-100.IEEE.
29 [8] Rahideh A and. Shaheed MH. Mathematical dynamic modelling of a twin-rotor multiple
rR
30 input-multiple output system. Proc. IMechE, Part I: J. Systems and Control Engg,
31
2007;221(1): 89–101.
32
ev
33 [9] Moness M and Mostafa AM. An algorithm for parameter estimation of twin-rotor multi-
34 input multi-output system using trust region optimization methods. Proc IMechE, Part I: J
35
Systems and Control Engineering, 2013; 227(5): 435-450.
36
iew
37 [10] Rotondo D, Nejjari F and Puig V, Quasi-LPV modeling, identification and control of a
38 twin rotor MIMO system. Control Engineering Practice, 2013;21(6): 829-846.
39
40 [11] Mullhaupt P, Srinivasan B, Levine J and Bonvin D. Control of the Toycopter Using a Flat
41 Approximation. IEEE Trans. Control Systems Technology, 2008;16(5): 882–896.
42
43 [12] Tastemirov A, Lecchini-Visintini A and Morales-Viviescas R M. Complete dynamic
44 model of the Twin Rotor MIMO System (TRMS) with experimental validation. Control
45 Engineering Practice, 2017;66(8): 89–98.
46
47 [13] Karnopp DC, Margolis DL and Rosenberg RC. System dynamics: modelling, simulation,
48 and control of mechatronic systems. John Wiley & Sons; 2012 Mar 7
49
[14] Mukherjee A, Karmakar R and Samantaray AK. Bond Graph in Modeling,Simulation and
50
51 Fault Identification. CRC Press, FL,2006.
52 [15] Karnopp DC. Lagrange’s Equations for Complex Bond Graph Systems. Journal of
53 Dynamic Systems, Measurement, and Control, 1977; 99(4): 300.
54
55 [16] Gawthrop PJ and Bevan GP. Bond-graph modeling. IEEE Control Systems Magazine,
56 2007; 27(2): 24–45.
57
58 25
59
1
2
3 [17] Engja H. Bond graph model of a reciprocating compressor. J. Franklin Institute,
4
5
1985;319(1–2): 115–124.
6 [18] Drozdz W and Pacejka HB. Development and validation of a bond graph handling model
7 of an automobile. Journal of the Franklin Institute, 1991;328(5–6): 941–957.
8
9 [19] Pathak PM, Mukherjee A and Dasgupta A. Impedance Control of Space Robots Using
10 Passive Degrees of Freedom in Controller Domain. Journal of Dynamic Systems,
11 Measurement, and Control, 2005;127(4): 564.
12
13 [20] Bera TK, Samantaray AK and Karmakar R. Bond graph modeling of planar prismatic
14 joints. Mechanism and Machine Theory, 2012;49: 2–20.
15
16 [21] Banerjee N, Saha AK, Karmakar R and Bhattacharyya R. Bond graph modeling of a
17 railway truck on curved track. Simulation Modeling Practice and Theory, 2009; 17(1): 22-
18 34.
19
[22] Borutzky W. Bond Graph Modelling of Engineering Systems: Theory, Applications and
Fo
20
21 Software Support, Springer, 2011.
22
[23] Mishra N and Vaz A. Bond graph modeling of a 3-joint string-tube actuated finger
23
rP
24 prosthesis, Mechanism and Machine Theory, 2017;117: 1-20.

25 [24] Dauphin-Tanguy G, Rahmani A and Sueur C. Bond graph aided design of controlled
26
ee
systems. Simulation Practice and Theory, 1999;7 (5–6): 493–513.

27
28 [25] Margolis D and Shim T. A bond graph model incorporating sensors, actuators, and vehicle
29 dynamics for developing controllers for vehicle safety. J. Franklin Institute, 2001; 338: 21–
rR
30 34.
31
32 [26] Merzouki R, Medjaher K, Djeziri MA and Ould-Bouamama B. Backlash fault detection in
ev
33 mechatronic system. Mechatronics, 2007;17(6): 299–310.

34
35 [27] Vijay P, Samantaray AK and Mukherjee A. A bond graph model-based evaluation of a
36
iew
control scheme to improve the dynamic performance of a solid oxide fuel cell,
37 Mechatronics, 2009; 19(4): 489–502.
38
39 [28] Sharma K, Bera TK. Modelling of autonomic regulation in cardiovascular system based on
40 baroreflex mechanism and overwhelming controller. Proc. IMechE Part I: Journal of
41 Systems and Control Engineering. 2017 Apr; 231(4):299-311.
42
43 [29] Gor MM, Pathak PM, Samantaray AK, Yang JM, Kwak SW. Fault-tolerant control of a
44 compliant legged quadruped robot for free swinging failure. Proc. IMechE Part I: Journal
45 of Systems and Control Engineering. 2018 Feb; 232(2):161-77.
46
47 [30] Dalla VK, Pathak PM. Power-optimized motion planning of reconfigured redundant space
48 robot. Proc. IMechE Part I: Part I: Journal of Systems and Control Engineering. 2018 Nov
49 23:0959651818814133.
50
51 [31] Dalla VK, Pathak PM. Curve-constrained collision-free trajectory control of hyper-
52 redundant planar space robot. Proc. IMechE Part I: Journal of Systems and Control
53 Engineering. 2017 Apr; 231(4):282-98.
54
55
[32] Borutzky W. Fault indicators and unique mode-dependent state equations from a fixed-
56 causality diagnostic bond graph of linear models with ideal switches. Proc. IMechE Part I:
57
58 26
59
1
2
3 Journal of Systems and Control Engineering. 2018 Jul; 232(6):695-708.
4
5 [33] Samantaray AK and Ghoshal SK. Sensitivity bond graph approach to multiple fault
6 isolation through parameter estimation. Proc. IMechE Part I: J Systems and Control
7 Engineering, 2007;221(4): 577–587.
8
9 [34] Prakash O, Samantaray AK and Bhattacharyya R. Model-Based Diagnosis of Multiple
10 Faults in Hybrid Dynamical Systems With Dynamically Updated Parameters. IEEE
11 Transactions on Systems, Man, and Cybernetics: Systems, 2017; 1–20, In Press, doi:
12
10.1109/TSMC.2017.2710143.
13
14 [35] Prakash O, Samantaray AK and Bhattacharyya R, Model-based multi-component adaptive
15 prognosis for hybrid dynamical systems. Control Engineering Practice, 2018; 72:1–18.
16
17 [36] Loureiro R, Merzouki R and Ould Bouamama B. Extension of the bond graph causality
18 inversion method for fault detection and isolation, Mechatronics, 2014; 24(8):1042–1049.
19
[37] Li T, Zhang Y, Gordon BW. Passive and active nonlinear fault-tolerant control of a
Fo
20
21 quadrotor unmanned aerial vehicle based on the sliding mode control technique. Proc.
22 IMechE Part I: Journal of Systems and Control Engineering. 2013 Jan;227(1):12-23.
23
rP
24 [38] Chen F, Zhang K, Wang Z, Tao G, Jiang B. Trajectory tracking of a quadrotor with
25 unknown parameters and its fault-tolerant control via sliding mode fault observer.
26 Proceedings of the Institution of Mechanical Engineers, Part I: Journal of Systems and
ee
27 Control Engineering. 2015 Apr; 229(4):279-92.

28
29 [39] Li C, Yang X, Xiao B. Adaptive attitude tracking control of a 3-degrees-of-freedom
rR
30 experimental helicopter with actuator dead-zone. Proc. IMechE Part I: Journal of Systems
31 and Control Engineering. 2019 Jan; 233(1):91-9.
32
[40] Faris F, Moussaoui A, Djamel B, Mohammed T. Design and real-time implementation of a
ev
33
34 decentralized sliding mode controller for twin rotor multi-input multi-output system. Proc.
35 IMechE Part I: Journal of Systems and Control Engineering. 2017 Jan; 231(1):3-13.
36
iew
37 [41] Pandey SK, Dey J, Banerjee S. Design of robust proportional–integral–derivative controller

38 for generalized decoupled twin rotor multi-input-multi-output system with actuator non-
39 linearity. Proc. IMechE Part I: Journal of Systems and Control Engineering. 2018 Aug;
40 232(8):971-82.
41
42 [42] Zeid A and Chung CH. Bond graph modeling of multibody systems: a library of three-
43 dimensional joints. Journal of the Franklin Institute, 1992;329(4): 605–636.
44
45
[43] Merzouki R, Samantaray AK, Pathak PM and Ould Bouamama B. Intelligent Mechatronic
46 Systems. London: Springer London, 2013.
47 [44] Louca LS. Power Conserving Bond Graph Based Modal Representations and Model
48
Reduction of Lumped Parameter Systems. J. Dynamic Systems, Measurement and Control,
49
50 2014; 136(6): 061007.
51
52
53
54
55
56
57
58 27
59
1
2
3 Appendix-1
4
5 Nomenclature
6 , ,
7
Roll, pitch and yaw z-y-x Euler angles
8 des , des Desired pitch and yaw angles
9
10
( , , ) Angular velocities in body-fixed frames, same as x , y , z
11 m , t Angular speed of main and tail rotors (blades) in stator fixed reference frame
12
, , Angular velocities in body-fixed frames
13 x y z
14 e Pitch and yaw error signal vector

15 g Acceleration due to gravity
16
17 lt Length of tail part of the support beam
18 lm Length of main part of the support beam
19 lcb Length of the counterweight beam
Fo
20
21 lb Distance between counterweight and pivot
22 mtr Tail rotor mass
23
rP
mmr Main rotor mass

24
25
mcb Counterweight mass
26 mt Tail shield and dc motor mass
ee
27 mm Main shield and dc motor mass

28
29
mb Counterweight beam mass
rR
30 rms Main shield radius

31 rts Tail shield radius
32
r; Viscous damping coefficient on pitch
ev
33
34 r Viscous damping coefficient on yaw
35 ( , , ) Velocities in body-fixed frames, same as (Vx,Vy,Vz)
36
iew
( , , ) Accelerations in body-fixed frames about in x,y,z axes

37
38 (xc,yc,zc) Position of counter weight CG with respect to support beam CG
39 ( , , ) Velocities of the CG in body-fixed reference frame
40 (xm,ym,zm) Main rotor’s motor and shield CG position with respect to support beam CG
41
42 (xp,yp,zp) Position of pivot point with respect to support beam CG
43 (xt,yt,zt) Position of tail rotor’s motor and shield CG with respect to support beam CG
44 C Time constant of filter used to take derivative of error signal in PID controller
45
46
Cmv Coefficient of viscous damping on main rotor
47 Ctv Coefficient of viscous damping on tail rotor
48 Cmtp Thrust force coefficient of main rotor for positive spin
49
50
Cmtn Thrust force coefficient of main rotor for negative spin
51 Cttp Thrust force coefficient of tail rotor for positive spin
52 Cttn Thrust force coefficient of tail rotor for negative spin
53
Cdm Drag force coefficient of main rotor
54
55 Cdt Drag force coefficient of tail rotor
56 Cp Pad stiffness
57
58 28
59
1
2
3
(Fx,Fy,Fz) External or constraint forces in body fixed frame in x,y,z axes
4
5 I x ,I y ,I z Moment of inertia about principal axes (body-fixed x,y,z axes)
6
Jbx Supporting beam mass moment of inertia about x-axis
7
8 Jby Supporting beam mass moment of inertia about y-axis
9 Jbz Supporting beam mass moment of inertia about z-axis
10 Jcx Counterweight mass moment of inertia about x-axis
11
12 Jcy Counterweight mass moment of inertia about y-axis
13 Jcz Counterweight mass moment of inertia about z-axis
14 Jmx Main shield and DC motor mass moment of inertia about x-axis
15
16 Jmy Main shield and DC motor mass moment of inertia about y-axis
17 Jmz Main shield and DC motor mass moment of inertia about z-axis
18
Jmrx Main rotor mass moment of inertia about x-axis
19
Fo
20 Jmry Main rotor mass moment of inertia about y-axis
21 Jmrz Main rotor mass moment of inertia about z-axis
22
23 Jtx Tail shield and DC motor mass moment of inertia about x-axis
rP
24 Jty Tail shield and DC motor mass moment of inertia about y-axis
25 Jtz Tail shield and DC motor mass moment of inertia about z-axis
26
ee
27 Jtrx Tail rotor mass moment of inertia about x-axis

28 Jtry Tail rotor mass moment of inertia about y-axis
29 Jtrz Tail rotor mass moment of inertia about z-axis
rR
30
31 Kam. Torque coefficient of main DC motor
32 Kat. Torque coefficient of tail DC motor
ev
33 (Kp,Kd,Ki) PID controller’s proportional, derivative and integral gains

34
35
Kyaw Nonlinear cable stiffness opposing support beam yaw
36 Lam Armature inductance of main motor
iew
37 Lat Armature inductance of tail motor

38
39
(Mx,My,Mz) External or constraint moments in body fixed frame in x,y,z axes
40 Ram Armature resistance of main motor
41 Rat Armature resistance of tail motor
42
RP Pad damping coefficient
43
44 (Um,Ut) Input signals to main and tail rotor motor drivers
45 (Vm,Vt) Voltages applied to main and tail rotor terminals
46 (Vx,Vy,Vz) Velocities in body-fixed frames in x,y,z axes
47
48 (X,Y,Z) Positions in inertial reference frame, subscripts indicate the point
49 ( , ,!) Velocities in inertial frame in X,Y, Z axes, subscripts indicate the point
50
51
52
53
54
55
56
57
58 29
59

Bond Graph Paper A K Samantary

Uploaded by

Copyright:

Available Formats

Bond Graph Paper A K Samantary

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Bond Graph Paper A K Samantary

Uploaded by

Copyright:

Available Formats

Journal of Systems and Control Engineering

Journal: Part I: Journal of Systems and Control Engineering

Manuscript Type: Original article

Date Submitted by the

Complete List of Authors: Srinivasarao, Gopisetti ; Birla Institute of Technology, Department of

dynamic modelling, mechanical control systems, system simulation,

The dynamics of a Twin Rotor Multi-Input Multi-Output (MIMO) System

Figure 3: Block diagram of interface gain or amplifier function

14 Electronic balancing machine

24 The nonlinear stiffness is experimentally measured by using electronic balancing machine

30 1.4 radians and thereafter it increases.

(To main motor and shield, Fig.16)

Velocities at point ‘m’ on

27 singularity will not arise.

Velocities at point ‘c’ on supporting beam

(Form supporting beam, Fig.12)

I:Jcy Cp:C 1 R:Rp

Velocities at point ‘t’ on supporting beam

(Form supporting beam, Fig.12)

30 main motor and shield

Velocities at point ‘m’ on supporting beam

(Form supporting beam, Fig.12)

Velocities at CG of main motor and shield

(From main motor and shield, Fig.16)

Velocities at CG of tail motor and shield

(From tail motor and shield, Fig.15)

27 the subsystems with least significant dynamics.

24 TRMS. In: Congress on Evolutionary Computation, 2009; pp. 2022-2028. IEEE.

24 prosthesis, Mechanism and Machine Theory, 2017;117: 1-20.

systems. Simulation Practice and Theory, 1999;7 (5–6): 493–513.

33 mechatronic system. Mechatronics, 2007;17(6): 299–310.

27 Control Engineering. 2015 Apr; 229(4):279-92.

37 [41] Pandey SK, Dey J, Banerjee S. Design of robust proportional–integral–derivative controller

14 e Pitch and yaw error signal vector

mmr Main rotor mass

27 mm Main shield and dc motor mass

30 rms Main shield radius

( , , ) Accelerations in body-fixed frames about in x,y,z axes

27 Jtrx Tail rotor mass moment of inertia about x-axis

33 (Kp,Kd,Ki) PID controller’s proportional, derivative and integral gains

37 Lat Armature inductance of tail motor

You might also like

Pfad - The Proxy pFad of © 2024 Garber Painting. All rights reserved.