Professional Documents
Culture Documents
Denavit Hartenberg Methods
Denavit Hartenberg Methods
difference between the input and output of the controller is at the filter frequency.
Which results in improved damping at the selected filter frequency. But in the presence of
significant delay, the deviation of phase from the expected value is proportional to the
frequency of oscillations and hence the system is prone to instability at higher frequencies
in the amplification region. There is not much freedom to modify the integer-order
control to overcome this problem. If the order of derivatives used in the control law is not
limited to the integer values, then there is a possibility of getting a better and more stable
frequency response. Fractional-order derivative shifts the harmonic signal by a pre-set
constant value which would neutralize the negative phase shift due to delay to some
extent resulting in an increased phase margin.
2. Mathematical Model
The non-dimensional equation of motion for an SDOF system with nonlinear stiffness and
damping is,
22\* MERGEFORMAT
()
The governing equation of the proposed controller and the control law is defined as
33\* MERGEFORMAT ()
variable as well as the system response. The process includes three operators, pth order
derivative, a second-order filter and a qth order derivative. So, the system performance
depends on the sum of the orders of derivatives . It can be verified from the
analytical solution. While implementing, must be set to the nearest integer to . Once the
value of q is fixed, the value of p can be obtained using . Depending on the value of
, the control strategy can be classified as Position feedback, Velocity feedback or
acceleration feedback control for q equals to 0,1 and 2 respectively.
When the system in the static condition is subjected to small fluctuations, effect of nonlinear
terms is negligible and hence nonlinear terms can be neglected. So, the linear stability
analysis of the system is sufficient to determine the static stability of the system.
44\* MERGEFORMAT ()
Applying Laplace transform, and simplifying yields the following transfer function
55\*
MERGEFORMAT ()
The system will be statically unstable if the roots of the above equation lie in the right half of
the complex plane. From the Argument principle, the difference between the number of zeros
and poles enclosed by a closed curve is given by
The characteristic equation for the above system does not have any poles so the integral over
the closed boundary gives the number of zeros in the region enclosed by the boundary. Since
the coefficients of the characteristic equation are real, the roots will be either complex
conjugates or will lie on the real axis i.e. the location of the poles will be symmetric about the
real axis. Hence stability can be determined by selecting only the first quadrant assuming that
there are no poles on the real and imaginary axes. Let the curves bounding the first quadrant
In the above expression, each integral gives the change in argument as we move from starting
point to the ending point of the curve. A curve is a real axis in plane which is mapped to
a real axis in plane. Hence there will be no change in argument as we move along the
curve .
Substituting it in 5
The function to be integrated to obtain the change in an argument along the curve is very
complicated and symbolical integration is not possible. So, the graphical method is used to
determine the stability of the system. The change in the argument of is as a point
moves on the curve in an anticlockwise direction i.e. one full rotation in the
anticlockwise direction. For a stable system, the change in the argument of across the
curve must be i.e. one full rotation in the clockwise direction. That means if
we map the curve in the plane, it must move around the origin. For a convex curve,
if the origin lies on the left side of the curve (starting from 0) i.e. inside the region formed by
the curve, then the system is stable. The equation of curve can be expressed as .
From the above expression,
For , as ,
And for , at ,
So, for , the curve starts from and moves in a clockwise direction until the
linear term becomes dominating and then start moving in the anticlockwise direction. In such
cases, the geometric approach fails to determine the stability correctly. For such cases, a
different starting point is selected as shown in Figure 2.
The Principal Argument method used above is valid only if there are no zeros or poles on the
boundary. We have considered the positive real axis as one of the boundaries of the region
and due to the presence of delay, the system may possess a positive real pole.
For , as ,
And for , at ,
That means the system will have a positive real pole only if and . To check the
presence of a positive real root in such a case, the following equation is solved for .
The filter equation is linear and hence at steady-state filter variables can be expressed in
terms of filter parameters and the system variable as follows,
Where,
Substituting the above expression for the filter variable and rearranging the terms,
66\*
MERGEFORMAT ()
Where,
Let, and be two-time scales.
Substituting the above expression in the equation 6, and equating the coefficients of
Now, equating the coefficients of and simplifying one can get the following equations
77\
* MERGEFORMAT ()
88\*
MERGEFORMAT ()
Eliminating the secular terms from 7 and 8 and balncing the harmonics following slow flow
equations are obtained,
99\*
MERGEFORMAT ()
1010\*
MERGEFORMAT ()
1111\* MERGEFORMAT ()
1212\* MERGEFORMAT ()
At steady state,
Applying the steady state conditions in equations, 9, 10, 11 and 12
1313\* MERGEFORMAT ()
1414\* MERGEFORMAT ()
1515\*
MERGEFORMAT ()
1616\*
MERGEFORMAT ()
From above equations one can get the following polynomial, which is solved using
MATLAB and the results thus obtained are compared with the simulation.
1717\* MERGEFORMAT ()
Where,
And,
The stability of steady state solution is determined from the Jacobian of the above slow flow
equations which is given in Appendix.
Figure 4: Verification of analytical solution using MATLAB Simulink
5. Performance Optimization
5.1. Cross-Over Design
In the Crossover design, all poles of the system are placed as far as possible from the
origin which is possible only if the poles are coinciding. The crossover design ensures
that the system possesses identical eigenvalues and a single peak in frequency response.
Using first-order Pade approximant for the exponential term one can get the following
equation,
Where,
i.e.,
Let,
Substituting in 18 we get,
Equating real and imaginary parts on both sides following equations are obtained
1919\*
MERGEFORMAT ()
2020\* MERGEFORMAT
()
So, we get four equations with 6 variables including and , provided that the system
parameters are given. The other four parameters are , , and . In practice, there
will be constraints on the values of gain due to exciter dynamics as well as its saturation
limit. The lower bound on the time delay is also dependent on the hardware as well as the
efficiency of the code that is being used. Therefore, it will be appropriate to fix the gain
and time delay and get the optimized filter parameters for the selected value of .
To verify the results of the method used above. The poles of the system are calculated,
and it is observed that the complex poles are coinciding. To get the poles it is assumed
that is a rational number i.e., it can be expressed as the ratio of two integers and
(say)
The frequency response and time response of the system for optimized and unoptimized
parameters are shown in Figures 5 and 6. In an optimized system, the oscillation dies out
quickly as compared to the unoptimized system.
Figure 5: Time response of an optimized and unoptimized system subjected to impulse input
The effective damping increases with an increase in linear gain and increases as
increases, attain maximum value, and then decreases. These results are verified using
simulation results and are shown in Figure 7. The maxima shift towards the higher values
of as the gain increases. Similarly, the peak amplitude decreases with increasing ,
attains minimum value and then increases as shown in Figure 10.
Figure 6: Variation in effective damping
The effect of delay on effective damping, peak amplitude, and area under the frequency
response curve is shown in Figures 11, 12, and 13 respectively. For ,
introducing delay results in reduced effective damping. But for higher values of , there
is not much change in the effective damping. For , increasing delay reduces the
effective damping up to a certain limit of , then starts increasing. For , there
exists a value of for which the effective damping is maximum. And that value shifts
towards the right as the delay is increased.
For , both peak amplitude and the area under the frequency response curve reduce as
the delay is increased which implies that introducing delay improves the performance as
well as stability of the controller for . For lower values of in the range
, there is not much variation in the peak amplitude as delay is increased. But
for , the increased delay results in a reduced peak amplitude. For , the area
under the frequency response curve increases as the delay is increased.
Figure 9: Effect of delay on Effective Damping
Introducing nonlinear gain reduces the area under the frequency response curve for
But, the peak amplitude first reduces up to a certain limit of and then
increases as the nonlinear gain is increased.
Figure 12:Effect of nonlinear gain on Peak Amplitude
A system with a crossover design reaches the steady-state quickly and hence is effective
for systems susceptible to frequent disturbances. If the system remains undisturbed for
most of the time other designs such as , and equal peak designs can give a better
frequency response.
The combined and equal peak design is used to obtain the filter parameters for better
frequency response. The cost functions used for the optimization are as follows.
Where,
It is of the form,
At peaks,
2222\* MERGEFORMAT ()
From Equations 15 and (7) one can get the frequencies at which the slope of the
frequency response curve is zero. The amplitude of response at these frequencies can be
obtained by solving equation (6).
Figure 14: Frequency response of the system with crossover design and equal peak design
Figure 15: Effect of gain on the area under the frequency response curve (Equal Peak Design)
Figure 16 Frequency response of the system for various values of filter damping
For a special case , the frequency response curve passes through two fixed points
irrespective of the filter damping as shown in the figure. The fixed point can be obtained
by finding the intersection points of frequency response curves of the system with zero
and infinite filter damping[28]. For zero filter damping, the relation between frequency
and amplitude is given by,
2323\*
MERGEFORMAT ()
Where , and
2424\*
MERGEFORMAT ()
The points of the intersection will satisfy the equation 23 and 24. From which the
following expression for amplitude can be obtained,
2525\*
MERGEFORMAT ()
From 24 and 25 one can get the frequency and amplitude of a fixed point. Let the
2626\* MERGEFORMAT ()
And
2727\* MERGEFORMAT ()
2828\* MERGEFORMAT ()
Where,
Here the first cost function ensures that the peaks are at fixed points and the second cost
function ensures that the amplitude of fixed points are equal (i.e. equal peaks)
The fractional control improves the performance for both crossover design and equal peak
design as shown in Figures 10 and 17.
6. Bounded Control
In practice the control force exerted by the actuator is limited. Once the desired control force
exceeds this limit, even the linear system shows nonlinear behaviour. In such cases, the
system may become unstable and hence suitable control law needs to be chosen such that the
control force is always bounded irrespective of the amplitude of system response.
Where, is an Algebraic saturation function.
For the sinusoidal input, describing the function of the saturation function is
equivalent to a gain obtained by the following expression,
Where,
Using the multiple time scale method, one can get the steady-state equations as
2929\*
MERGEFORMAT ()
The minimum force amplitude for which the actuator saturates can be obtained as shown
below. Since the algebraic saturation function tends to saturate at infinity, the percentage of
saturation is considered.
Substituting the above values in 29 the minimum excitation force at which actuator
saturates can be calculated.
Figure 19: effect of delay on minimum excitation force at which actuator saturates
7. Experimental setup
A cantilever beam with the first mode at 11.25 Hz, is used as a model. The experimental
setup is shown in Figures 20 and 21. The details about the instruments used are given in
Table 1. The sample time during the entire experiment was set to 1 millisecond. For the
implementation of fractional order derivative, the equivalent transfer function of 5 th order
was obtained using Crone’s method from ninteger toolbox[30]. Both the second-order
filter and filter equivalent to fractional-order derivative was discretized using the Tustin
method. For the experimental purpose, the microcontroller was used to give both
excitation and control signals. The gain of the power amplifier was set at the desired
position and that position was maintained throughout the experiments.
Figure 20 Experimental Setup
8. Experimental Results
First, the delay was set to 14 milliseconds i.e., a nondimensional delay of 0.9896. For
selected gain and delay, the optimal value of r is calculated such that the peak value of the
frequency response function is minimum. The filter parameters were obtained from the
crossover design. The experimental results comparing integer-order control and fractional
order control are shown in Figures 22, 23 and 24. From Figures 22 and 23, it can be
observed that as the gain increases the optimal value of r also increases which is in
agreement with the analytical results shown in Figure 13.
Figure 22: Comparison of performances in the frequency domain (k = -0.04)
For both integer and fractional order control, increasing the gain improves the
performance of the controller as shown in Figures 25 and 26.
From Figure 27, it can be observed that introducing the delay results in performance
deterioration for the integer-order control. Also, it results in a reduced gain margin as
shown in Figure 3. Both crossover design and equal peak design are verified
experimentally and are shown in Figures 28 and 29. By considering the nonzero
values of i.e., the nonlinear control law, higher amplitude oscillations can be
suppressed further as shown in Figure 30.
From Table 2, we can conclude that the fractional-order control is better than the
integer-order control for high values of delay. Also, fractional order control seems to
be effective for high values of gain and zero delay as shown in Figure 31. So, the
above results can be summarized as shown in Table 3.
Figure 29 Comparison of performances with Crossover design and Equal Peak design(r = 2.5)
Figure 30 Effect of Nonlinear gain on frequency response (r = 2)
9. Conclusion
In this paper, a generalized model of an SDOF system with nonlinear stiffness and
damping is considered. To control the vibrations due to external harmonic excitation, a
fractional-order control law is proposed. The approximate solution of the system of
equations is obtained from the method of multiple scales. The static stability of the
system is checked using the Argument principle. Optimization of filter parameters is done
by using the crossover design and Equal peak design. The performance of bounded
control is studied, it shows trends similar to the unbounded control. The effect of the gain
corresponding to the nonlinear term in control law is studied. Increasing this gain results
in the suppression of high amplitude oscillations near resonance without affecting the low
amplitude oscillations at other frequencies. Analytical results are verified with both
numerical simulation and experiments.
From the analytical solution and with the supporting experimental results we can
conclude that the fractional-order control gives better performance for the higher value of
delay and gain. This implies that it can be used in more complicated network control
strategies for the vibration control of multiple modes of the system, where more
processing time is required unlike the simpler AVC of the SDOF system. It can be also
used in applications where low-cost controllers with less computation power are used.
Appendix
[1] M.J.B. Rogers, K. Hrovat, M.E. Moskowitz, Effects of exercise equipment on the microgravity
environment, Advances in Space Research. 24 (1999) 1283–1287. https://doi.org/10.1016/S0273-
1177(99)00734-6.
[2] M.M. Jovanović, A.M. Simonović, N.D. Zorić, N.S. Lukić, S.N. Stupar, S.S. Ilić, Experimental
studies on active vibration control of a smart composite beam using a PID controller, Smart Materials
and Structures. 22 (2013). https://doi.org/10.1088/0964-1726/22/11/115038.
[3] S. Kumar, R. Srivastava, R.K. Srivastava, ACTIVE VIBRATION CONTROL OF SMART PIEZO
CANTILEVER BEAM USING PID CONTROLLER, n.d. http://www.ijret.org.
[4] S.M. Khot, N.P. Yelve, R. Tomar, S. Desai, S. Vittal, Active vibration control of cantilever beam by
using PID based output feedback controller, JVC/Journal of Vibration and Control. 18 (2012) 366–
372. https://doi.org/10.1177/1077546311406307.
[5] E. Eshraqi, M. Shahravi, M. Azimi, Consideration of spillover effect in active vibration suppression
of a smart composite plate using piezoelectric elements, Latin American Journal of Solids and
Structures. 13 (2016) 2343–2356. https://doi.org/10.1590/1679-78253089.
[6] S. Nima Mahmoodi, M. Ahmadian, Active vibration control with modified positive position feedback,
Journal of Dynamic Systems, Measurement and Control, Transactions of the ASME. 131 (2009) 1–8.
https://doi.org/10.1115/1.3089565.
[7] E. Omidi, R. Mccarty, S.N. Mahmoodi, Implementation of modified positive velocity feedback
controller for active vibration control in smart structures, 9057 (2014) 1–11.
https://doi.org/10.1117/12.2044478.
[8] S.N. Mahmoodi, M. Ahmadian, Modified acceleration feedback for active vibration control of
aerospace structures, Smart Materials and Structures. 19 (2010). https://doi.org/10.1088/0964-
1726/19/6/065015.
[9] E. Omidi, S.N. Mahmoodi, W.S. Shepard, Multi positive feedback control method for active vibration
suppression in flexible structures, Mechatronics. 33 (2016) 23–33.
https://doi.org/10.1016/j.mechatronics.2015.12.003.
[10] E. Omidi, N. Mahmoodi, Hybrid Positive Feedback Control for Active Vibration Attenuation of
Flexible Structures, IEEE/ASME Transactions on Mechatronics. 20 (2015) 1790–1797.
https://doi.org/10.1109/TMECH.2014.2354599.
[11] Y. Wu, W. Zhang, X. Meng, Y. Su, Compensated positive position feedback for active control of
piezoelectric structures, Journal of Intelligent Material Systems and Structures. 29 (2018) 397–410.
https://doi.org/10.1177/1045389X17708045.
[12] M. Sayed, M. Kamel, 1:2 and 1:3 internal resonance active absorber for non-linear vibrating system,
Applied Mathematical Modelling. 36 (2012) 310–332. https://doi.org/10.1016/j.apm.2011.05.057.
[13] A. Sarkar, J. Mondal, S. Chatterjee, Controlling self-excited vibration using positive position
feedback with time-delay, Journal of the Brazilian Society of Mechanical Sciences and Engineering.
42 (2020). https://doi.org/10.1007/s40430-020-02544-7.
[14] A. Sarkar, J. Mondal, S. Chatterjee, Controlling self-excited vibration using acceleration feedback
with time-delay, International Journal of Dynamics and Control. 7 (2019) 1521–1531.
https://doi.org/10.1007/s40435-019-00577-y.
[15] J. Mondal, S. Chatterjee, Controlling self-excited vibration of a nonlinear beam by nonlinear resonant
velocity feedback with time-delay, International Journal of Non-Linear Mechanics. 131 (2021).
https://doi.org/10.1016/j.ijnonlinmec.2021.103684.
[16] S. Chatterjee, Vibration control by recursive time-delayed acceleration feedback, Journal of Sound
and Vibration. 317 (2008) 67–90. https://doi.org/10.1016/j.jsv.2008.03.020.
[17] C.A. Monje, B.M. Vinagre, V. Feliu, Y.Q. Chen, Tuning and auto-tuning of fractional order
controllers for industry applications, Control Engineering Practice. 16 (2008) 798–812.
https://doi.org/10.1016/j.conengprac.2007.08.006.
[18] S.H. HosseinNia, I. Tejado, B.M. Vinagre, Fractional-order reset control: Application to a
servomotor, Mechatronics. 23 (2013) 781–788. https://doi.org/10.1016/j.mechatronics.2013.03.005.
[19] S.H. Hosseinnia, I. Tejado, D. Torres, B.M. Vinagre, V. Feliu, A General Form for Reset Control
Including Fractional Order Dynamics, n.d.
[20] S.H. Hosseinnia, I. Tejado, V. Milanés, J. Villagrá, B.M. Vinagre, Experimental application of hybrid
fractional-order adaptive cruise control at low speed, IEEE Transactions on Control Systems
Technology. 22 (2014) 2329–2336. https://doi.org/10.1109/TCST.2014.2308837.
[21] S. Kapoulea, C. Psychalinos, A.S. Elwakil, S.H. HosseinNia, Realizations of fractional-order PID
loop-shaping controller for mechatronic applications, Integration. 80 (2021) 5–12.
https://doi.org/10.1016/j.vlsi.2021.04.009.
[22] L. Marinangeli, F. Alijani, S.H. HosseinNia, Fractional-order positive position feedback compensator
for active vibration control of a smart composite plate, Journal of Sound and Vibration. 412 (2018) 1–
16. https://doi.org/10.1016/j.jsv.2017.09.009.
[23] D. Feliu-Talegon, A. San-Millan, V. Feliu-Batlle, Fractional-order integral resonant control of
collocated smart structures, Control Engineering Practice. 56 (2016) 210–223.
https://doi.org/10.1016/j.conengprac.2016.07.016.
[24] H. Butler, C. de Hoon, Fractional-order filters for active damping in a lithographic tool, Control
Engineering Practice. 21 (2013) 413–419. https://doi.org/10.1016/j.conengprac.2012.12.011.
[25] E. Omidi, S. Nima Mahmoodi, Vibration control of collocated smart structures using H ∞ modified
positive position and velocity feedback, JVC/Journal of Vibration and Control. 22 (2016) 2434–2442.
https://doi.org/10.1177/1077546314548471.
[26] M.P. Bayon de Noyer, S. v. Hanagud, Comparison of H2 optimized design and cross-over point
design for acceleration feedback control, Collection of Technical Papers -
AIAA/ASME/ASCE/AHS/ASC Structures, Structural Dynamics and Materials Conference. 4 (1998)
3250–3258. https://doi.org/10.2514/6.1998-2091.
[27] S. Krenk, J. Høgsberg, Equal modal damping design for a family of resonant vibration control
formats, JVC/Journal of Vibration and Control. 19 (2013) 1294–1315.
https://doi.org/10.1177/1077546312446796.
[28] A.K. Mallik, S. Chatterjee, Principles of Passive and Active Vibration Control, Affiliated East-West
Press Private Limited, NEW DELHI, 2014.
[29] B.B. Alagoz, Hurwitz stability analysis of fractional order LTI systems according to principal
characteristic equations, ISA Transactions. 70 (2017) 7–15.
https://doi.org/10.1016/j.isatra.2017.06.005.
[30] Duarte Valério, ninteger, (2022).