A family of explicit algorithms for general pseudodynamic testing

A new family of explicit pseudodynamic algorithms is proposed for general pseudodynamic testing. One particular subfamily seems very promising for use in general pseudodynamic testing since the stability problem for a structure does not need to be considered. This is because this subfamily is unconditionally stable for any instantaneous stiffness softening system, linear elastic system and instantaneous stiffness hardening system that might occur in the pseudodynamic testing of a real structure. In addition, it also offers good accuracy when compared to a general second-order accurate method for both linear elastic and nonlinear systems.

Seismic retrofi tting of RC columns with RC jackets and wing walls with different structural details

An original reinforced concrete(RC) column and four strengthened specimens, two with RC jackets and two with wing walls, were tested in this study. The original column specimen was designed to comply with older(pre-1999) design standards so that the usual detailing defi ciencies in existing school buildings in Taiwan could be simulated. Two different structural details were chosen to fabricate the full-scale specimens for each retrofi tting technique. The study confi rmed that either RC jacketing or the installation of wing walls with two different structural details can effectively improve the stiffness and strength of an existing column. RC jacketing shows a better improvement in energy dissipation and ductility when compared to the columns with wing walls installed. This is because the two RC jacketed columns experienced a fl exural failure, while a shear failure was found in the two columns with the wing walls installed, and thus led to a drastic decrease of the maximum lateral strengths and ductility. Since many factors may affect the installation of a post-installed anchor, it is better to use standard hooks to replace post-installed anchors in some specifi c points when using RC jacketing or installing wing walls.

Comparison of capability of time integration methods in capturing dynamic loading

Numerical properties of the time integration method proposed by the first author of this paper in 2007 are the same as those of the constant average acceleration method （AAM） for linear elastic systems, except that the capability to capture dynamic loading was not explored. It was found that there were different quadrature equations to predict the next step displacement increment. A modified quadrature equation of this method was derived so that the equation to determine the next step displacement was numerically equivalent to the equation used in the constant AAM. It was verified that the original form of this method, in general, had a better capability to capture dynamic loadings than the constant AAM. This excellent property, in addition to computational efficiency, will help to make this method competitive with general secondorder accurate integration methods.

Nonlinear evaluations of unconditionally stable explicit algorithms

Two explicit integration algorithms with unconditional stability for linear elastic systems have been successfully developed for pseudodynamic testing. Their numerical properties in the solution of a linear elastic system have been well explored and their applications to the pseudodynamic testing of a nonlinear system have been shown to be feasible. However, their numerical properties in the solution of a nonlinear system are not apparent. Therefore, the performance of both algorithms for use in the solution of a nonlinear system has been analytically evaluated after introducing an instantaneous degree of nonlinearity. The two algorithms have roughly the same accuracy for a small value of the product of the natural frequency and step size. Meanwhile, the first algorithm is unconditionally stable when the instantaneous degree of nonlinearity is less than or equal to 1, and it becomes conditionally stable when it is greater than 1. The second algorithm is conditionally stable as the instantaneous degree of nonlinearity is less than 1/9, and becomes unstable when it is greater than 1. It can have unconditional stability for the range between 1/9 and 1. Based on these evaluations, it was concluded that the first algorithm is superior to the second one. Also, both algorithms were found to require commensurate computational efforts, which are much less than needed for the Newmark explicit method in general structural dynamic problems.

Improved formulations of the CR and KR methods for structural dynamics

Although the Chen-Ricles（CR）method and the Kolay-Ricles（KR）method have been applied to conduct pseudodynamic tests,they have both been found to have some adverse numerical properties,such as conditional stability for stiffness hardening systems and an unusual overshoot in the steady-state response of a high-frequency mode.An improved formulation for each method can be achieved by using a stability amplification factor to boost the unconditional stability range for stiffness hardening systems and a loading correction term to eliminate the unusual overshoot in the steady-state response of a high-frequency mode.The details for developing improved formulations for each method are shown in this work.

Nonlinear performance of classical damping

The performance of a classical damping matrix, constructed either from the use of initial structural properties or current structural properties, in the step-by-step solution of a nonlinear multiple degree of freedom （MDOF） system is analytically evaluated. The analytical results are confirmed by numerical examples. Consequently, some conclusions are drawn from these analytical results that might be considered as rough guidelines for practical applications. It is found that a classical damping matrix constructed from initial structural properties is adequate for practical applications, since it has approximately the same damping effect as obtained by current structural properties and is more efficient in terms of computing.

Structure-dependent difference equations for time integration

A structure-dependent explicit method with enhanced stability properties is proposed in this study. In general, the method offers unconditional stability for structural systems except those with a particular instantaneous stiffness hardening behavior. In addition, it is second-order accurate and displays no overshooting in high frequency responses. Numerical experiments reveal that the proposed method saves a substantial amount of computational effort in solving inertial problems where only the low frequency responses are of interest, when compared to a general second-order accurate integration method.

Extensions of nonlinear error propagation analysis for explicit pseudodynamic testing

Two important extensions of a technique to perform a nonlinear error propagation analysis for an explicit pseudodynamic algorithm （Chang, 2003） are presented. One extends the stability study from a given time step to a complete step-by-step integration procedure. It is analytically proven that ensuring stability conditions in each time step leads to a stable computation of the entire step-by-step integration procedure. The other extension shows that the nonlinear error propagation results, which are derived for a nonlinear single degree of freedom （SDOF） system, can be applied to a nonlinear multiple degree of freedom （MDOF） system. This application is dependent upon the determination of the natural frequencies of the system in each time step, since all the numerical properties and error propagation properties in the time step are closely related to these frequencies. The results are derived from the step degree of nonlinearity. An instantaneous degree of nonlinearity is introduced to replace the step degree of nonlinearity and is shown to be easier to use in practice. The extensions can be also applied to the results derived from a SDOF system based on the instantaneous degree of nonlinearity, and hence a time step might be appropriately chosen to perform a pseudodynamic test prior to testing.

Selection of time step for pseudodynamic testing

Although the step degree of nonlinearity has been introduced to conduct basic analysis and error propagation analysis for the pseudodynamic testing of nonlinear systems, it cannot be reliably used to select an appropriate time step before performing a pseudodynamic test. Therefore, a novel parameter of instantaneous degree of nonlinearity is introduced to monitor the stiffness change at the end of a time step, and can thus be used to evaluate numerical and error propagation properties for nonlinear systems. Based on these properties, it is possible to select an appropriate time step to conduct a pseudodynamic test in advance. This possibility is very important in pseudodynamic testing, since the use of an arbitrary time step might lead to unreliable results or even destroy the test specimen. In this paper, guidelines are proposed for choosing an appropriate time step for accurate integration of nonlinear systems. These guidelines require estimation of the maximum instantaneous degree of nonlinearity and the solution of the initial natural frequency. The Newmark explicit method is chosen for this study. All the analytical results and the guidelines proposed are thoroughly confirmed with numerical examples.