Research Article  Open Access
Lakhdar Guenfaf, Mohamed Azira, "Generalized Minimum Variance Control for MDOF Structures under Earthquake Excitation", Journal of Control Science and Engineering, vol. 2016, Article ID 7458654, 11 pages, 2016. https://doi.org/10.1155/2016/7458654
Generalized Minimum Variance Control for MDOF Structures under Earthquake Excitation
Abstract
Control of a multidegreeoffreedom structural system under earthquake excitation is investigated in this paper. The control approach based on the Generalized Minimum Variance (GMV) algorithm is developed and presented. Our approach is a generalization to multivariable systems of the GMV strategy designed initially for singleinputsingleoutput (SISO) systems. KanaiTajimi and CloughPenzien models are used to generate the seismic excitations. Those models are calculated using the specific soil parameters. Simulation tests using a 3DOF structure are performed and show the effectiveness of the control method.
1. Introduction
During the last decades, there have been many attempts in the field of structural control to develop new algorithms according to the increasing size and flexibility of building structures. This is done especially for multidegreeoffreedom (MDOF) structures subjected to strong vibrations. A MDOF structure can be considered as a largescale system with interconnected subsystems, which are story units. Therefore, control techniques initially designed for SDOF structures cannot be directly applied to MDOF structures. The problem of interconnection between story units and the structural proprieties of the building are considered as a whole system and must be addressed. The interconnection effects can also be introduced in control law formulation or considered as external disturbances to be compensated by control actions. Another problem encountered in MDOF structural control is the number and location of actuators. In fact it seems to be more efficient to place an actuator in each story unit, but the cost of implementation may be very expensive.
Several structural models have been investigated in the literature for an active control perspective. Indeed, many researchers have used an active control strategy based on statespace model to develop and design a controller for structural vibration [1–6]. Nonlinear model has been used and presented for Stochastic Optimal Control of Structural Systems [1, 7]. Saaed Alqado et al. [8] have developed five methods for structural identification approach, specifically ARX, ARMAX, BJ, OE, and statespace models, and have implemented them for the identification process. Furthermore, the paper shows that the ARMAX and the output error models have indicated an excellent performance to predict the mathematical models of vibration’s propagation in the building. Also, it has been shown by Guenfaf and Azira [9] that the ARMAX model can give an interesting representation of the system for a digital control perspective. In this case the system is characterized as stochastic process described by a linear model subjected to external disturbances.
Vigorous researches on structural control have given more interesting solutions [10–12]; some of them have already been adopted in actual building structures. Guenfaf et al. [13] have presented GMV algorithm based on ARMAX model of SDOF structure with introducing KanaiTajimi and CloughPenzien seismic excitation within the structural model. The results show that the relative displacement of the structure and the control effort are significantly reduced. Lei et al. [14] have proposed an algorithm for the decentralized structural control of tall sheartype buildings under unknown earthquake excitation. A decentralized control algorithm based on the instantaneous optimal control scheme has been developed with limited measurements of structural absolute acceleration responses. The interconnection effect between adjacent substructures has been treated as “additional unknown disturbances” at substructural interfaces to each substructure. However, the control algorithm is based on the minimization of some cost functions such as linear quadratic regulator which requires a solution of the Riccati equation that is subjected to a boundary condition at the terminal time. That leads to a suboptimal solution. So, in such perturbated systems, the controller performances and robustness may be weak [15].
The purpose of this paper is to develop a control strategy applied to an earthquake excited 3DOF structure. This algorithm attempts to minimize a generalized cost function including variance of both output and control effort. For this purpose a specific model of the system to be controlled and its environment will be developed.
The GMV algorithm has been widely studied in the literature. It was introduced by Clarke [16, 17] as an extended version of the Minimum Variance (MV) algorithm initially developed by Åström [18].
The paper is organized as follows. Section 2 deals with the development of the dynamic model of the structure, while Section 3 deals with presentation of the MIMO ARMAX model. The GMV algorithm is introduced in Section 4. The seismic excitation models are presented in Section 5. In Section 6, simulation results show the effectiveness of the developed algorithms. Finally, some conclusions are given.
2. Dynamical Model of a MultiDegreeofFreedom Structure
The purpose of this section is to formulate dynamical equations of motion of a multidegreeoffreedom structure under seismic excitation. Figure 1(a) is a schematic representation of the structure. It is a multistory building with active tendon controllers in each story unit. In order to establish the dynamical model, the following assumptions are considered [19]:(1)Each story is supposed to be a lumped mass in the girder.(2)The two vertical axes between two adjacent floors are weightless and inextensible in the vertical direction.
(a)
(b)
Figure 1(b) is a representation of the dynamic force equilibration at th story unit, which can be written aswhere is the inertial force of the mass is the damping force of th story is the elastic force of th story where relative displacement is . is the control force from the controller installed between th story and th story. Absolute displacement is defined as where is the ground motion.th story unit is characterized by its characteristic parameters. In this system, it is assumed that the structural mass, , and the elastic stiffness, , have been concentrated in floors and columns, respectively. Internal viscous damping, , is also a parameter that describes the structural behaviour.
Substituting (2), (3), (4), and (5) into (1), we obtainwhere , , , , , and represents the ground acceleration.
Equation (6) can be written in matrix form aswhere is the structural displacement vector, control vector, is the mass matrix of the structure, is the damping matrix of the structure, is the stiffness matrix of the structure, is the matrix indicating the location of actuators, and unity vector of dimension .
3. ARMAX Model of the Structure
We are interested in this section to derive the discrete ARMAX model of the MDOF structure. Consider the statespace equation of the MDOF structure which can be obtained from (7) by choosing as a state vector: where of dimension , of dimension , of dimension , of dimension , identity matrix of dimension , null matrix of dimension , and null vector of dimension .
The digital control of the system needs the knowledge of the discrete representation model. Thus, the continuous model can be written in the following form [20–22]:where is the output vector, is the input vector, is the zeromean white noise vector with covariance matrix , , , , , , for , , , shift operator defined as .
3.1. Multivariable Model
We intend in this section to derive the complete multivariable ARMAX model of the MDOF structure without neglecting the coupling between stories. We can perform this calculation using the equation of motion of the structural system. Consider (6) which can be rewritten in the following form:
Now applying the Laplace transform to this equation,
Thus, we can deduce the continuous ARMAX model given in the following:wherewith , , , and , , .
To obtain the discrete ARMAX model, we use (11) written as
Using discretization methods, we deduce the discrete form of (14):
The discrete multivariable ARMAX model is then obtained. It is given bywhere
Remark 1. We can introduce here the seismic excitation model in the derivation of the multivariable ARMAX model. By considering the seismic excitation as a filtered white noise, we follow the same calculations as done in Section 3.1, to obtain the discrete multivariable ARMAX model of the structure under a specified excitation model [23].
4. Generalized Minimum Variance Control Algorithm
The Generalized Minimum Variance (GMV) algorithm was introduced by Clarke [16, 17] to control nonminimum phase systems. It is an extension of the Minimum Variance algorithm [18] which, by choosing a certain performance criterion and a certain model of the controlled system and its perturbations, attempts to minimize the variance of the output. A generalization of the GMV algorithm for multiinputmultioutput (MIMO) systems has been proposed in the context of selftuning control [24–26]. The controlled system in this case is assumed to be described by a linear vector difference equation including a moving average of white noise.
A Generalized Minimum Variance strategy for multivariable systems will be now presented. Consider the system described by the linear vector difference equation (multivariable ARMAX model):where output vector, input vector, zeromean white noise vector with covariance , is the smallest time delay of the system [20], , , is nonsingular, and , where
We can summarize the basic assumptions on the system in the following: The number of outputs is equal to the number of inputs. is nonsingular.
The criterion to be minimized iswhere is expected value, is the vector defining the reference signal, , , and are weighing polynomial matrices of dimension each, and vectorial norm is defined as .
As in the monovariable case we have first to derive the optimal predictor of , since it is a future information. It is given by (see Appendix A)where , , and are polynomial matrices defined in Appendix A and .
Using (18) and (21) it is shown that the control strategy is given by (see Appendix B)where .
5. Mathematical Model of Earthquake Ground Motion
The earthquake ground acceleration is modeled as a uniformly modulated nonstationary random process [19, 27]:where is a deterministic nonnegative envelope function and is a stationary random process with zero mean and a KanaiTajimi power spectral density:where , are filter parameters and is the constant spectral density of the white noise. However, it can be shown that the velocity and displacement spectra, which are derived from the acceleration spectra described by (24), have strong singularities at zero frequency. These singularities can be removed by using highpass filter, as suggested by CloughPenzien [19].
Using such a second highpass filter, the KanaiTajimi spectrum is modified as follows to obtain the CloughPenzien spectrum:
A particular envelope function given in the following will be used:where , , and are parameters that should be selected appropriately to reflect the shape and duration of the earthquake ground acceleration. Numerical values of parameters are , , , , , , , and .
The KanaiTajimi and CloughPenzien ground accelerations have been simulated and are presented in Figure 2.
(a) KanaiTajimi model
(b) CloughPenzien model
6. Simulation Results
Simulation tests are performed using a 3DOF structure with structural parameters [28] Kg, N/m, and Ns/m, Kg, N/m, and Ns/m, Kg, N/m, and Ns/m.
An active tendon controller is installed in every story unit and the angle of incline of the tendons with respect to the floor is 25°. Thus, the control force vector from the controllers is . Thus, we can suppose that the force is applied at the top of each story and assumed to be activated externally by an independent power supply.
To implement the multivariable GMV algorithm, a sampling period is used. Parameters of the multivariable ARMAX model of the structure are given in the following:
Ponderation polynomials are
is the output signal filter transfer function given with set to unity to ensure offsetfree control. For process control, if the process output has a large overshoot to setpoint and load changes, the transfer function can be used to penalize this overshoot. The output signal filter influences both the reference tracking and the disturbance rejection properties. When it is applied, the control can be slower, the overshoots can be damped, and the changes in the manipulated variable can be suppressed.
is the reference weighting filter. Higher weighting leads to faster control (with more overshoot) and to higher control signal increments. On the other hand, Smaller weighting leads to slower control (with less overshoot) and to smaller changes in the manipulated signals. The reference signal filter does not influence the disturbance rejection or a plant/model mismatch. When it is applied, the control can be slower, overshoots can be damped, and the changes in the manipulated variable can be suppressed.
is the weighting of the control increments. Higher weighting leads to slower control (with less overshoot) and to smaller changes in the manipulated variable. However, smaller weighting leads to faster control (with more overshoot) and to bigger changes in the manipulated variable.
Table 1 gives the output variance for each story.

Responses of the structure to KanaiTajimi and CloughPenzien excitation models are shown in Figures 3 and 4, respectively.
(a)
(b)
(c)
(a)
(b)
(c)
Generalized Minimum of Variance (GMV) control strategy for MDOF structure using KanaiTajimi and CloughPenzien seismic models is presented in Figures 3 and 4. GMV algorithm is based on the minimization of a quadratic cost function, and the solution of the Diophantine equation is required. We compare the displacement responses without control and the ones controlled by the GMV algorithm. The control parameters are chosen to give the optimal performances according to the effect of each parameter discussed (, , and ). The results show a considerable reduction of the structural vibration under the applied seismic load. The proposed control strategy has proved the considerable efficiency for the mitigation of the structural response under seismic ground motion. The ARMAX model presents an interesting form to formulate a MDOF structural model for a control perspective in the earthquake engineering problem.
In the last decade, several works have shown the effectiveness of the active control strategies in the earthquake engineering problem. Guenfaf and Allaoua [29] have investigated a flux oriented control of Active Mass Damper (AMD) induction motor for buildings under seismic ground motion. The LQR control algorithm has been used for displacement reduction of the structure. The simulation results show that the torque supplied by the induction motor pursuits its reference provided by the LQR control algorithm. Significant reductions of 65% and 78%, respectively, in the roof acceleration and roof displacement have been reached. Xu et al. [30] have proposed a realtime Active Mass Damper (AMD) subsystem testing method, which has been applied to verify the performance of an innovative AMD control system driven by a linear motor. Yamamoto and Sone [31] have presented the feasibility, stability, and reliability of the integrated AMD subsystem, including hardware and software. The main purpose of an AMD that was installed in a highrise building in Tokyo is to maintain human comfort in two horizontal directions under the strong earthquakes. The control performance of the AMD was confirmed. The AMD controlled the building acceleration such that it was 60% of its maximum derivation during the Tohoku earthquake 2011. Ikeda [32] has presented the practical applications and verification of the active and the semiactive control strategy for the structural control problem. Practical applications of active and semiactive vibration control of buildings in Japan are reported, and the installation tendency is described in relation to research phases of structural control. Control verification methods are introduced to link structural control with monitoring and system identification. The state of the art shows that the use of structural control and its related technologies have been expanding in the last two decades since the late Professor Kobori suggested the importance of active structural control in the civil engineering field in the mid1980s.
7. Conclusions
This paper investigated GMV control technique applied to earthquake excited MDOF structures. Our control approach was developed and presented based on the ARMAX model calculation. The multivariable GMV have demonstrated good performances and the structural responses are significantly reduced. The control strategy used in this paper is an extension to the SDOF case of the classical GMV algorithm. It is designed for the structure as a multivariable system.
The multivariable GMV algorithm is an efficient control technique with a wellestablished theoretical development. However, it has the disadvantage of the great amount of calculations and matrix manipulation especially in the case of tall buildings with a large number of degrees of freedom. Nevertheless, it takes into account the coupling terms between different storey buildings, and this gives more realistic results due to the nonlinear nature of this coupling terms.
Appendix
A. Derivation of the Optimal Predictor for MIMO Systems
We first introduce the polynomial matrix equation:wherewhere and ,where and .
There exist nonunique matrices and verifyingwhere .
We introduce the polynomial matrix :From (12), (A.3), (A.4), (A.5), and (A.6) we demonstrate thatwhere and .
Equation (18) can be written asMultiplying (A.8) by and after some calculations using (A.7) we obtainWe remark that is the future information. Thus, the optimal predictor of is given by
B. Derivation of the GMV Control for MIMO Systems
Using (A.10), the criterion to be minimized (20) becomesThe term is uncorrelated with , , and for . Thus, can be written asThe minimum of is found by forming the gradient of with respect to :orwhere .
Substituting (A.10) into (B.4), we obtain the control strategy:where .
Competing Interests
The authors declare that there are no competing interests regarding the publication of this paper.
References
 G. Mei, A. Kareem, and J. C. Kantor, “Model predictive control of structures under earthquakes using acceleration feedback,” Journal of Engineering Mechanics, vol. 128, no. 5, pp. 574–585, 2002. View at: Publisher Site  Google Scholar
 S. Purohit and N. K. Chandiramani, “Optimal static output feedback control of a building using an MR damper,” Structural Control and Health Monitoring, vol. 18, no. 8, pp. 852–868, 2011. View at: Publisher Site  Google Scholar
 A. Shafieezadeh, K. Ryan, and Y. Chen, “Fractional order filter enhanced LQR for seismic protection of civil structures,” Journal of Computational and Nonlinear Dynamics, vol. 3, no. 2, Article ID 021404, 2008. View at: Publisher Site  Google Scholar
 S.G. Wang, “LQGα control and its simulations for structural benchmark problems against winds and earthquakes,” in Proceedings of the 42nd IEEE Conference on Decision and Control, pp. 6572–6577, December 2003. View at: Publisher Site  Google Scholar
 L. Guenfaf and S. Allaoua, “Linear quadratic AMD controller with an observer for buildings under seismic ground motion,” in Proceedings of the 13th International Conference on Control, Automation and Systems (ICCAS '13), pp. 1559–1564, IEEE, Gwangju, Republic of Korea, October 2013. View at: Publisher Site  Google Scholar
 L. Guenfaf and S. Allaoua, “Optimal choice of actuator for AMD linear quadratic controller,” Journal of Electrical and Control Engineering, vol. 3, no. 3, pp. 31–38, 2013. View at: Google Scholar
 L. Guenfaf, S. Allaoua, M. Djebiri, M. S. Boucherit, and F. Boudjema, “Generalized minimum variance gain scheduling controller for nonlinear structural systems under seismic ground motion,” Journal of Electrical Systems, vol. 8, no. 4, pp. 472–480, 2012. View at: Google Scholar
 T. E. Saaed Alqado, G. Nikolakopoulos, and J.E. Jonasson, “Comfort level identification for irregular multistorey building,” Automation in Construction C, vol. 50, pp. 40–49, 2015. View at: Publisher Site  Google Scholar
 L. Guenfaf and M. Azira, “ARMAX model calculation for structures under seismic ground motion including soilstructure interaction,” Journal of Electrical and Control Engineering, vol. 4, no. 1, pp. 8–15, 2013. View at: Google Scholar
 S. Paul, W. Yu, and X. Li, “Recent advances in bidirectional modeling and structural control,” Shock and Vibration, vol. 2016, Article ID 6275307, 17 pages, 2016. View at: Publisher Site  Google Scholar
 H. Qian, H. Li, G. Song, and W. Guo, “Recentering shape memory alloy passive damper for structural vibration control,” Mathematical Problems in Engineering, vol. 2013, Article ID 963530, 13 pages, 2013. View at: Publisher Site  Google Scholar
 Y. Si, H. R. Karimi, and H. Gao, “Modeling and parameter analysis of the OC3hywind floating wind turbine with a tuned mass damper in nacelle,” Journal of Applied Mathematics, vol. 2013, Article ID 679071, 10 pages, 2013. View at: Publisher Site  Google Scholar
 L. Guenfaf, M. Djebiri, M. S. Boucherit, and F. Boudjema, “Generalized minimum variance control for buildings under seismic ground motion,” Earthquake Engineering and Structural Dynamics, vol. 30, no. 7, pp. 945–960, 2001. View at: Publisher Site  Google Scholar
 Y. Lei, D.T. Wu, and S.Z. Lin, “Integration of decentralized structural control and the identification of unknown inputs for tall shear building models under unknown earthquake excitation,” Engineering Structures, vol. 52, pp. 306–316, 2013. View at: Publisher Site  Google Scholar
 M. Athans, “The role and use of the stochastic linearquadraticGaussian problem in control system design,” IEEE Transactions on Automatic Control, vol. AC16, pp. 529–552, 1971. View at: Google Scholar  MathSciNet
 D. W. Clarke, “Introduction to self tuning controller,” in Self Tuning and Adaptive Control: Theory and Application, vol. 15 of IEE Control Engineering Series, Peter Pregrinus, 1981. View at: Google Scholar
 D. W. Clarke, “Selftuning control of nonminimumphase systems,” Automatica, vol. 20, no. 5, pp. 501–517, 1984. View at: Publisher Site  Google Scholar  MathSciNet
 K. J. Åström, Introduction to Stochastic Control Theory, vol. 70 of Mathematics in Science and Engineering, Academic Press, New York, NY, USA, 1970. View at: MathSciNet
 R. W. Clough and J. Penzien, Dynamics of Structures, McGrawHill, New York, NY, USA, 1993.
 L. Dugard, G. C. Goodwin, and X. Xianya, “The role of the interactor matrix in multivariable stochastic adaptive control,” Automatica, vol. 20, no. 5, pp. 701–709, 1984. View at: Publisher Site  Google Scholar  Zentralblatt MATH
 M.H. Liu and W. Lin, “Pole assignment selftuning controller for robotic manipulators,” International Journal of Control, vol. 46, no. 4, pp. 1307–1317, 1987. View at: Publisher Site  Google Scholar
 M.H. Liu and W. Lin, “Multivariable selftuning control with decoupling for robotic manipulator,” IEE Proceedings D: Control Theory and Applications, vol. 135, no. 1, pp. 43–48, 1988. View at: Publisher Site  Google Scholar
 A. Mohamed and G. Lakhdar, “GMV Control algorithm for structures under recorded strong earthquakes including soil characteristics,” in Proceedings of the 2015 15th International Conference on Control, Automation and Systems (ICCAS '15), pp. 1687–1692, Busan, Republic of Korea, October 2015. View at: Publisher Site  Google Scholar
 U. Borison, “Selftuning regulators for a class of multivariable systems,” Automatica, vol. 15, no. 2, pp. 209–215, 1979. View at: Publisher Site  Google Scholar  Zentralblatt MATH
 H. N. Koivo, “A multivariable selftuning controller,” Automatica, vol. 16, no. 4, pp. 351–366, 1980. View at: Publisher Site  Google Scholar  MathSciNet
 S.J. Lang, X.Y. Gu, and T.Y. Chai, “A Multivariable generalized selftuning controller with decoupling design,” IEEE Transactions on Automatic Control, vol. AC31, no. 5, pp. 474–477, 1986. View at: Google Scholar
 J. N. Yang, A. Akbarpour, and P. Ghaemmaghami, “New optimal control algorithms for structural control,” Journal of Engineering Mechanics, vol. 113, no. 9, pp. 1369–1386, 1987. View at: Publisher Site  Google Scholar
 A. G. Chassiakos and S. F. Masri, “Modelling unknown structural systems through the use of neural networks,” Earthquake Engineering and Structural Dynamics, vol. 25, no. 2, pp. 117–128, 1996. View at: Google Scholar
 L. Guenfaf and S. Allaoua, “Flux oriented control of Active Mass Damper induction motor for buildings under seismic excitation,” in Proceedings of the 15th International Conference on Control, Automation and Systems (ICCAS '15), pp. 1676–1681, BEXCO, Busan, Korea, October 2015. View at: Publisher Site  Google Scholar
 H. Xu, C. Zhang, H. Li, and J. Ou, “Realtime hybrid simulation approach for performance validation of structural active control systems: a linear motor actuator based active mass driver case study,” Structural Control and Health Monitoring, vol. 21, no. 4, pp. 574–589, 2014. View at: Publisher Site  Google Scholar
 M. Yamamoto and T. Sone, “Behavior of active mass damper (AMD) installed in highrise building during 2011 earthquake off Pacific coast of Tohoku and verification of regenerating system of AMD based on monitoring,” Structural Control and Health Monitoring, vol. 21, no. 4, pp. 634–647, 2014. View at: Publisher Site  Google Scholar
 Y. Ikeda, “Active and semiactive vibration control of buildings in Japan—practical applications and verification,” Structural Control and Health Monitoring, vol. 16, no. 78, pp. 703–723, 2009. View at: Publisher Site  Google Scholar
Copyright
Copyright © 2016 Lakhdar Guenfaf and Mohamed Azira. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.