Chih-Chen Chang

Neural Network Modeling of a Magnetorheological Damper

The magnetorheological (MR) damper is a newly developed semiactive control device that possesses unique advantages such as low power requirement and adequately fast response rate. The device has been previously tested in a laboratory to determine its dynamic properties and characterized by a system of nonlinear differential equations. This paper presents an alternative representation of the damper in terms of a multilayer perceptron neural network. A neural network model with 6 input neurons, one output neuron and twelve neurons in the hidden layer is used to simulate the dynamic behavior of the MR damper. Training of the model is done by a Gauss-Newton based Levenberg-Marquardt method using data generated from the numerical simulation of the nonlinear differential equations. An optimal brain surgeon strategy is adopted to prune the weights and optimize the neural networks. An optimal neural network is presented that satisfactorily represents dynamic behavior of the MR damper.

Mass dampers and their optimal designs for building vibration control

The control performance of three types of mass dampers on suppressing excessive building vibration is studied and compared in this study. These mass dampers include the tuned mass damper (TMD), the tuned liquid column damper (TLCD) and the liquid column vibration absorber (LCVA). A set of generalized building-mass damper equations are established where the building is modeled as a single-degree-of-freedom system. Under a broad-band white noise excitation, a set of common formulas for the optimal properties as well as some useful design formulas for these mass dampers can be derived in closed forms for both wind and earthquake types of loading. The results show that the performance of these mass dampers depends on a parameter called efficiency index. It is shown that the larger the efficiency index the better the control performance of these mass dampers. Control efficiency among these three mass dampers is studied, compared and summarized.

Control performance of liquid column vibration absorbers

Abstract This paper reports the study of the control performance and effectiveness of a new type of passive device—liquid column vibration absorber (LCVA). The LCVA, which allows the column cross-section to be non-uniform, is a variation of the tuned liquid column damper (TLCD). The LCVA provides great versatility and architectural adaptability, since its natural frequency is determined not only by the length of the liquid column, but also the geometric configuration. in this study, an unsteady and non-uniform flow equation for the LCVA is derived based on the Lagrange equation. It is assumed that the liquid flow is uniform in the horizontal and the vertical columns, respectively. The optimal head loss coefficient for the LCVA is derived explicitly under the condition that the LCVAs frequency is tuned to that of the structure. The control effectiveness of the LVCA is numerically analyzed and compared to that of the TLCD and the tuned mass damper (TMD). It is found that if properly designed the LCVA can be as effective as or even more effective than the TLCD.

PARAMETRIC STUDY ON MULTIPLE TUNED MASS DAMPERS FOR BUFFETING CONTROL OF YANGPU BRIDGE

Abstract A study on buffeting control of the Yangpu Bridge using a multiple tuned mass damper (MTMD) system is performed in this paper. The MTMD system consists of a set of TMDs which are attached to the center region of the bridges main span and are symmetrical about the center of the main span as well as about the central line along the bridge span. It is found that the control efficiency of the MTMD system is sensitive to its frequency characteristics, namely, the central frequency ratio and the frequency bandwidth ratio. On the other hand, the damping ratio of the TMDs has less significant effects on the control efficiency. A total of seven MTMD systems with different mass ratios are designed. Each one of these seven MTMD systems can be used for the buffeting control of the Yangpu Bridge, depending on the required control efficiency and the available budget.

Structural Damage Detection Using an Iterative Neural Network

A structural damage detection method based on parameter identification using an iterative neural network (NN) technique is proposed in this study. The NN model is first trained off-line using an initial training data set that consists of assumed structural parameters as outputs and their corresponding dynamic characteristics as inputs. The structural parameters are assumed with different levels of reduction to simulate various degrees of structural damage. The concept of orthogonal array is adopted to generate the representative combinations of parameter changes, which can significantly reduce the number of training data while maintaining the data completeness. A modified back-propagation learning algorithm is proposed which can overcome possible saturation of the sigmoid function and speed up the training process. The trained NN model is used to predict the structural parameters by feeding in measured dynamic characteristics. The predicted structural parameters are then used in the FE model to calculate the dynamic characteristics. The NN model would go through a retraining process if the calculated characteristics deviate from the measured ones. The identified structural parameters are then used to infer the location and the extent of structural damages. The proposed method is verified both numerically and experimentally using a clamped-clamped T beam. The results indicate that the current approach can identify both the location and the extent of damages in the beam.

Increase of critical flutter wind speed of long-span bridges using tuned mass dampers

Abstract In this paper increasing the critical flutter wind speed of long-span bridges by using tuned mass dampers (TMDs) is theoretically and experimentally studied. Equations governing the motions of a bridge with TMDs are established. The Routh-Hurwitz stability criterion is used to study the aerodynamic instability of the bridge based on the characteristic equation of the system of the bridge and TMDs. A sectional model wind-tunnel test on the Tiger Gate Bridge, a suspension bridge with a steel box deck and a center span of 888 m, is carried out to confirm the numerical results. Some new findings from the test and the calculation are presented.

Unified dynamic absorber design formulas for wind-induced vibration control of tall buildings

The objective of this paper is to establish some unified design formulas for various kinds of passive dynamic absorber for wind-induced vibration control of tall buildings. A total of five different passive dynamic absorbers (TMD, TLCD, LCVA, C-TLD and R-TLD) are considered in this study. Firstly, unified equations of motion for the building–absorber system under wind-induced excitation are presented. A set of unified formulas for the optimal properties and the equivalent damping ratios for these five dynamic absorbers are then derived analytically. The Shanghai Central Plaza, which is a thirty-nine-story reinforced-concrete building, is used as an example to demonstrate the procedure and to verify the accuracy of the unified approach. The results show that these unified formulas provide direct performance evaluation and comparison between the five dynamic absorbers for the control of wind-induced vibration of tall buildings.

Three-Dimensional Structural Translation and Rotation Measurement Using Monocular Videogrammetry

Measuring displacement for large-scale structures has always been an important yet challenging task. In most applications, it is not feasible to provide a stationary platform at the location where its displacements need to be measured. Recently, image-based technique for three-dimensional (3D) displacement measurement has been developed and proven to be applicable to civil engineering structures. Most of these developments, however, use two or more cameras and require sophisticated calibration using a total station. In this paper, we present a single-camera approach that can simultaneously measure both 3D translation and rotation of a planar target attached on a structure. The intrinsic parameters of the camera are first obtained using a planar calibration board arbitrarily positioned around the target location. The obtained intrinsic parameters establish the relationship between the 3D camera coordinates and the two-dimensional image coordinates. These parameters can then be used to extract the rotation and translation of the planar target using recorded image sequence. The proposed technique is illustrated using two laboratory tests and one field test. Results show that the proposed monocular videogrammetric technique is a simple and effective alternative method to measure 3D translation and rotation for civil engineering structures. It should be noted that the proposed technique cannot measure translation along the direction perpendicular to the image plane. Hence, proper caution should be taken when placing target and camera.

Adaptive neural networks for model updating of structures

A model updating methodology based on an adaptive neural network (NN) model is proposed in this study. The NN model has a feedforward architecture and is first trained off-line using some training data that are obtained from finite-element analyses and contain modal parameters as inputs and structural parameters as outputs. To reduce the number of training data while maintaining the data completeness, the variation of structural parameters is arranged using an orthogonal array. This NN model is then adaptively retrained on-line during the model updating process in order to eliminate the difference between the measured and the predicted modal parameters. A modified back-propagation algorithm is developed, in which the learning rate is dynamically adjusted once every few iterations. A jump factor is introduced to overcome the numerical difficulty caused by the saturation of the sigmoid function in order to improve the convergence performance of the NN model. The current adaptive NN updating procedure is applied to a suspension bridge model and verified both numerically and experimentally. The results indicate that by adaptively training the NN model and iteratively adjusting the structural parameters, it is possible to reduce the differences between the measured and the predicted frequencies from a maximum of 17% to 7% for the first eight vertical modes.

Structural Degradation Monitoring using Covariance-driven Wavelet Packet Signature

A covariance-driven wavelet packet signature extraction method is proposed for monitoring of structural degradation under random ambient excitation. Advantageous features of this method are it only requires measuring response at one location of the structure and no structural mathematical model is needed. Assuming that the excitation forces are stationary Gaussian white noises, the covariance of measured response is calculated and wavelet packet decomposed. A damage indicator is then formulated to extract discriminate information from the constructed wavelet packet component energies. A threshold for a damage alarm is established according to the statistical properties and the one-sided confidence limit of the damage indicator from successive measurement. For demonstration, a numerical study is performed on the health monitoring of the benchmark steel-frame scale structure proposed by the IASC-ASCE Structural Health Monitoring Task Group. Four damage cases, involving different levels of damage severity at various locations, are introduced. The results show that the health condition of the frame model can be accurately monitored by the proposed method. It is also seen that the proposed method is quite insensitive to measurement noise. The monitoring results are virtually unaffected even when the response is contaminated with measurement noise of the same magnitude as that of the signal.