Modified semiactive control with MR dampers for partially observed systems

• Published in: Engineering structures 2019-07, Vol.191, p.129-147
• Main Authors: Bhaiya, Vishisht, Shrimali, M.K., Bharti, S.D., Datta, T.K.
• Format: Article
• Language: English
• Subjects: Acceleration; Active control; Algorithms; Building frames; Clipped optimal control; Control algorithms; Control stability; Control theory; Covariance; Earthquake dampers; Earthquakes; Experimentation; Filters; Ground motion; Internet; Kalman filter; Kalman filters; Linear Quadratic Gaussian (LQG); MR damper; Noise; Noise generation; Noise measurement; Real time; Seismic activity; Seismic engineering; Seismic response; Seismic stability; Semiactive control; Semiactive damping; Sensitivity analysis; Sliding mode control; State estimation; State variable; Theoretical analysis; White noise
• ISSN: 0141-0296; 1873-7323
• DOI: 10.1016/j.engstruct.2019.04.063

•Direct online implementation.•Meeting the requirement of Gaussian white noise for optimum state estimate using Kalman filter.•Helps in selection of covariance matrices of noise and excitation for state estimation for site specific excitations.•The full non linearity of the MR damper with the structure is retained in the analysis. A modified semiactive control scheme with the MR damper for partially observed system is presented. The proposed control scheme augments the state variables by two filter variables and passes a white noise through the filters to obtain the desired seismic excitation to the structure. The two filters are incorporated at the base of the structure, making the input excitation to the structure-filter system a white noise. Thus, a more theoretical rigor is incorporated in the use of the Kalman filter for the state estimation as both the excitation and measurement noise should be ideally white, if the full state of the system is to be derived from the measured states using the Kalman filter. Further, using the results of a sensitivity analysis, the proposed algorithm fixes the covariances of the excitation and noise, that are provided as inputs to the Kalman filter. This is done in order to avoid any numerical instability of the control algorithm. Semi active control of a ten story building frame fitted with three MR dampers under earthquake ground motion is taken as an illustrative example. Theoretically obtained results of the proposed algorithm are compared with those of the conventional algorithm in which the ground motion is directly provided as input to the structure without the use of filters. Further, the use of the developed control algorithm in real time application which requires a trained ANN to generate compatible white noise signal from the online measurement of the ground motion, is described. The generated white noise signal (in real time) is provided as input to the proposed algorithm. The online application of the control algorithm is validated by a numerical experimentation in which three different types of specified time histories of ground acceleration are assumed as the expected future ground accelerations, which are measured online. The results of the numerical experiment are compared with those obtained from the theoretical analysis. It is shown that the scheme of the online application of the proposed algorithm performs satisfactorily. Further, it is shown that the proposed control algorithm may become unstable