| Collapses of super high-rise buildings would result in catastrophic personal casualty and property loss, thus it is necessary to study the cataclysmic law of the structures under strong earthquakes. Inelastic dynamic time-history analysis is one of the most important approaches for studying the behavior of super high-rise buildings under strong earthquakes. Through credible inelastic dynamic time-history analyses, the collapse margin, evolution of damage, failure path and collapse mechanism of structure can be revealed. However, finite element model of super high-rise buildings have a large number of elements and nodes, which leads to inefficiency when performing inelastic time-history analysis. Although computer hardware is updated contemporarily and parallel computing becomes more and more popular, the computing time is still too long to be tolerable, especially in the case of incremental dynamic analysis. Therefore, it is urgent to develop some more efficient algorithms to speed up the analysis. Base on the aforementioned concerns, the following work has been done in this paper.(1) A new finite element framework focusing on inelastic time-history analysis of super high-rise buildings is established. The framework, which is programed with C++ object-oriented language, is built based on absorbing the advantages of two popular open-source finite element program named FEAPpv and OpenSees. Basic materials, elements and algorithms for inelastic time-history analysis of super high-rise buildings are implemented in the framework. Through comparing the analytical results of the program with results from experiments and other programs, the results of the program is found to be accurate. Besides, the program has better computational efficiency and memory usage than OpenSees.(2) State transformation procedures (STP) are proposed to speed up the process of state determination of elements. Based on the characteristic that nonlinear behavior tends to develop in a few portion of components in building structures, a method which combines linear constitutive and nonlinear constitutive is devised to speed up state determination of fiber beam-column elements and multi-layered shell elements. The procedures is further accelerated with shared memory parallel computing by OpenMP programming language. Numerical tests validate that the efficiency of STP/PSTP would decrease as ground motion intensity increase, and the speedup of PSTP is from 8.09 to 14.7.(3) The most appropriate linear algebraic equations solver for super high-rise buildings is selected baseed comparative study. Several solvers, including direct and iterative parallel solvers based on CPUs or GPUs, are compiled and implemented in the program. Among these solvers, the parallel supernodal Cholesky factorization solver (PF) in CHOLMOD combined with multi-threaded OpenBLAS is most efficient. The computing time of PF is found to be proportional to the number of elements, and the process of matrix factorization costs more than 95% computing time of the process of solution.(4) Inexact Newton-Cholesky iterative algorithm (INC) is proposed to speed up the process of the solution of nonlinear equilibrium equations. The matrices after factorization of the Jacobian are repeatedly used to obtain the solution which match the inexact Newton condition. Thus to avoid unnecessary factorization of the Jacobian when the tangent stiffness matrix remains constant or changes slightly. Though numerical validations, it is found that INC only require several a few times of factorization of the Jacobian during inelastic time-history analysis. The parameters, α and β, in the formula of calculating forcing terms, are suggested to be 0.5 and 0.01, respectively. The convergence tolerance is suggested to be from 0.001 to 0.1.(5) Comprehensive validations on four super high-rise buildings are performed using the program with INC, PF and PSTP (IPP) implemented. It is found that state determinations of elements and solution of the system linear equations take most of the computing time of inelastic time-history analysis. As the scale of structure increase, computing time of solution of system linear equations increases from shorter than that of state determination of elements to longer. With the acceleration of IPP, computing time of inelastic time-history analysis become significantly shorter while little error can be observed. The speedup of IPP ranges from 28.7 to 38.9.(6) Dynamic seismic full-process analysis on two frame-core tube mega tall building is performed. The collapse margin, damage development, failure path and collapse mechanism of the two structures are analyzed. Through the study, the dynamic full-process of a frame-core tube mega tall building is depicted herein as:the initiation of the failure of link beams of shear walls results in the failure of limbs of shear walls spanning over different stories; accompanying the continuing action of ground motion, the increase of failed components induces the growth of overturning moment shouldered by outer frame; and at last, structural global failure is triggered together with the overturning failure of outer frame. An attempt is made to improve these two structures with two optimization plans, one reinforcing link beams of the shear walls and the other strengthening the mega columns of the outer frame. The finding is that reinforcing link beams could achieve a better improvement and a larger economy benefit. |
PDF Full Text Request