Graham C. Archer

A new object-oriented finite element analysis program architecture

Abstract This paper presents a new architecture for finite element analysis software, developed using object-oriented design. The resulting system is capable of the modeling and simulation of structural behavior, including the consideration of nonlinear static and dynamic effects. An innovation of the system design is the creation of several classes of objects that separate the analysis tasks from the details of the finite element model. This separation leads to flexible, extensible code. The overall system design and a prototype implementation are presented.

A DISTRIBUTED OBJECT-ORIENTED FINITE-ELEMENT ANALYSIS PROGRAM ARCHITECTURE

This article presents a distributed object-oriented design for a nonlinear finite-element analysis using the message-passing paradigm and a single-program, multiple-data scheme. The architecture is an extension of an existing sequential object-oriented architecture. The design recognizes the costly communication startup time penalty by attempting to minimize the frequency of communications. This is facilitated by distributing not only the elements in the model but also their associated nodes and mapping between the degrees of freedom and the analytical equations of equilibrium. The proposed object design was implemented and tested on a nonlinear static pushover analysis of three moment-resisting frames.

A technique for the reduction of dynamic degrees of freedom

This paper presents a variation on the component mode technique for the dynamic substructuring of large-scale structural analysis of building and bridge frames. The principal innovation of the proposed method of dynamic reduction is that the resulting mass matrix of the reduced substructures remains diagonal. As in the component mode technique, the reduction is accomplished by transforming the degrees of freedom in the substructure using boundary shapes and internal shapes. The diagonal mass matrix is achieved by orthogonalization of the boundary shapes to the internal shapes, and a selective row-by-row summation of the mass matrix into the diagonal entry (where off-diagonal terms are unavoidable). To aid in recovering the accuracy of the rigid-body inertias that is lost in the diagonalization process, additional pseudo-rigid-body-mode shapes are proposed.

Mid-Panel Cracking of Portland Cement Concrete Pavements in Indiana

Portland cement concrete slabs are a common form of highway pavement in Indiana. As a result of their widespread use, the economic impact of their maintenance and life span is therefore tremendous. In Indiana, these types of pavements have been experiencing premature random transverse mid-panel cracking. This phenomenon has been observed under a variety of environmental and traffic conditions. Also, it has been found that the cracking occurs in conjunction with the opening of the lanes to traffic, within months of construction. In this project, a research synthesis was carried out to determine exactly what the current state of knowledge on random transverse mid-panel cracking of portland cement concrete pavements is. This was accomplished by means of a comprehensive literature review of published works and through a survey of other state departments of transportation. In addition, preliminary finite element analyses were carried out to help further understand and confirm the findings from the research synthesis.

Exploration of Structural Stability: A Computer Implementation

In this paper, the education of structural stability is viewed from the perspective of a range of analysis techniques emphasizing the exploratory nature of not only instability phenomenon, but also engineering problems, in general. The purpose of this research was to develop a computational‐based educational platform upon which to perform “experimental” tests. Three aspects of this work are discussed: the philosophy of the platform, the design concepts behind its implementation, and the programme itself.

Distributed Object-Oriented Finite Element Analysis using Data Structures

This paper presents a distributed object-oriented design for a nonlinear finite element analysis using the message-passing paradigm and a single-program-multipledata scheme. In recognition of the costly communication startup time-penalty of distributed processing, the design uses data structures to minimize communications frequency. The proposed object design was implemented and tested on the nonlinear static push-over analysis of a 20-story moment-resisting frame. 1.0 Introduction As the science of computers evolves, so does the design of structural engineering analysis software. In the field of computer science, the 1990’s saw the rise of object-oriented programming, OOP (Rumbaugh 1991), standardized data structures, and parallel & distributed processing. Structural engineers are now rising to the challenge. At present, the C++ language and the Standard Template Library STL (Stroustrup 1997) have become the standards of OOP and data structure libraries, respectively. Likewise, MPI (Message Passing Interface Forum 1994) has become a reasonable communication standard for parallel & distributed processing. Object-oriented finite element analysis (OOFEA) programs consist of collections of abstract data types, describing sub-assemblages, elements, nodes, degrees of freedom, etc., and numerical data objects such as sparse matrices and vectors. OOFEA objects contain collections of both types of data in a complex hierarchical structure. These disparate objects are logically stored in container objects (standardized data structures) such as lists. Since parallel & distributed processing demands the transfers of data amongst various processes, a necessary requirement for the integration parallel processing into OOFEA programs is the ability to easily pass these complex objects amongst the processes. This paper presents a distributed architecture for an FEA program based on message passing. The architectural style employed is termed single-programmultiple-data (SPMD); i.e. the same program runs simultaneously on all processes, using different data. The distributed architecture is built on an existing sequential object-oriented finite element analysis program, OOPSE (Archer 1996, 1999). It is assumed that the program will execute on a distributed network of single-processor machines as is often available in structural engineering offices -- namely a network of Intel/NT workstations. Parallel and distributed processing is not a new concept. There exist libraries of excellent parallelized numerical solvers (for example SCALAPACK, Choi 1996). In addition, many excellent algorithms for parallel finite element analyses have been proposed (Sotelino 1990, Modak 1997). The purpose of the current research is to describe and implement a distributed program architecture that is sufficiently flexible to utilize the existing body of research and also to allow the investigation new solvers and algorithms.

Determination of Building Modal Parameters Using Low-Level Excitation

Over the past few years, the authors have performed forced vibration testing on a variety of low-rise building structures. In this paper, the procedure to experimentally determine the modal parameters (natural frequency, mode shape, and damping) of building structures through forced vibration testing is described. In addition, a comparison of the experimental results to computational predictions of the building response is made to highlight the accuracy of the computational modeling. The experimental setup consists of a small linear shaking device that gently (below human perception) vibrates the building structure and highly sensitive (but inexpensive) hardware and software to collect and process the results. Great care must be taken to properly position the equipment, filter out the extraneous data, and process and interpret the data. Interpretation of the data is complicated by the excitation of all modes when the building is subject to forced vibrations. A procedure for determining the mode shapes directly from the forced vibration testing is presented and compared to the experimental results.