Journal of Engineering Mechanics

The role of uncertainty analysis in the development of computer knowledge bases for civil engineering systems is discussed. It is argued that the term “expert system” should be replaced by “advice system” to avoid misunderstandings between the designers and users of such systems. An engineer is an expert to the extent that he is a responsible decision maker. The computer can only advise him; it cannot replace him. A new inference method of direct use in advice systems is presented. The method combines features of the theory of fuzzy sets, Shafer's theory of evidence and probability theory. The method follows the Horn clause logic representation of Prolog. Unlike probability theory, it does not rely upon a closed‐world form of knowledge representation, and can allow the modeling of total uncertainty. The Prolog form of negation by failure is not used. The method involves the use of necessary and possibility measures for the truth or dependability of a proposition. Examples are given of its use and a small model knowledge base for an advice system is described.

A finite‐difference analysis was employed to calculate nonlinear hydrodynamic pressures on vertical and inclined faces of dams. Two idealized ground accelerations and an actual earthquake excitation were used as the prescribed boundary condition. This dynamic condition and the free surface profile were cast in a dimensionless form of the basic equations, making the moving boundary problem simpler in computational analysis. The effect of water compressibility could be augmented by the nonlinear effect. Comparison of the results between inviscid and viscous analyses reflected not only the insignificant viscous effect but also the potential numerical noise associated with the viscous solution.

A numerical implementation of the “bounding surface plasticity” model for cohesive soils is described. An adaptive multistep integration procedure is used in conjunction with local iteration and radial return. Emphasis is placed on the development of a robust algorithm. The incorporation of the material model into finite element codes for stress‐consolidation problems in geotechnical engineering is described. Example analyses are given for one element problem, an idealized footing problem and two centrifugal simulations of field problems.

A method suggested for the analysis of thin‐walled beams is extended into the case of large displacements. Large displacement kinematic relations and a step‐by‐step integration method are used. Curvature‐bending moment curves can be tracted by using a program capable of treating arbitrary cross sections and some new encouraging results are produced.

A method is presented to find nonstationary random seismic excitations with a constraint on mean square value such that the response variance of a given linear system is maximized. It is also possible to incorporate the dominant input frequency into the analysis. The excitation is taken to be the product of a deterministic enveloping function and a zero mean Gaussian stationary random process. The power spectral density function of this process is determined such that the response variance is maximized. Numerical results are presented for a single‐degree system and an earth embankment modeled as shear beam.

An infinitely long elastic beam rests under gravity on a tensionless elastic foundation and is acted upon by a pair of moving loads. This action results in the creation of lift‐off regions between the beam and the foundation that effect the character of the response. The solution includes the determination of the location, magnitude, and extent of the lift‐off region(s) as well as the beam deflection at any point. Sample calculations for parameters relevant in railroad systems show that the effect of lift‐off is substantial. Also included in this analysis is the effect of constrained thermal expansion.

Wind‐tunnel models of long‐span bridges have been exploited in various forms since the late 1930s, and the state of the art has progressed in a parallel fashion. With the rapid increase in the number of new cable‐stayed designs worldwide, new occasions to review the aeroelastic theory and practice of bridge wind‐tunnel modeling have arisen. In particular, the new cable‐stayed bridges have distinct vibration modal forms that are a direct reflection of the design philosophy behind this type of structure. The new modal forms present new challenges to modeling and its interpretation. In this context, the present paper offers a critical examination of the mechanisms involved in the windtunnel modeling of bridges. The method employed is based on flutter derivatives obtained experimentally from deck section models.

Although significant progress on the modeling of concrete behavior has been made in recent years, no unified treatment of the existing mathematical models from which a comprehensive three‐dimensional constitutive model can be formulated, has been attempted. This paper summarizes our recent efforts in developing a relatively comprehensive and sophisticated model for progressive failure analysis of concrete structures. The main features of the proposed short‐term, time‐independent constitutive model include the most sophisticated Willam‐Warnke five‐parameter or Hsieh‐Ting‐Chen four‐parameter failure surface, the nonuniform hardening rule, the nonassociated flow rule with changing dilatancy factor, linear softening for post‐cracking in tension, and multiaxial softening for post‐failure in compression. The present model accounts for the hydrostatic pressure sensitivity and Lode angle dependence behavior of concrete, not only in its strength criterion, but also in its hardening characteristics. The model is suitable for any stress range (tension, compression, or mixed), covers the whole deformation process (elastic, hardening plastic, and softening). Good agreement with a wide range of experimental data has generally been observed. The proposed model has been coded in a nonlinear finite element program available for concrete structural analysis. A numerical example is also given.

Green's functions are key elements in boundary element formulations. This paper describes the implementation of an approximate, layer‐stiffness‐based Green's function in a boundary element analysis of harmonically loaded pile groups embedded in a nonhomogeneous medium. Two separate problems are addressed: the optimum layer discretization appropriate for the analysis of pile foundations and the depth of soil stratum required for simulating a half‐space. Validity is established by comparisons with previously published static and dynamic solutions. The implementation described allows the boundary element formulation to be extended to applications that have hitherto proved intractable.

A frequency domain iterative method is proposed to solve the non‐linear dynamic equation of motion encountered in the dynamic analysis of offshore structures under regular and random waves. The nonlinearities due to the relative velocity‐squared drag term and variable submergence are considered and treated using an approximate Newton‐Raphson method. In order to facilitate the computation, the formulation makes use of the normal mode theory. The resulting iterative scheme, which introduces a modal hydrodynamic damping on both sides of a decoupled equation of motion in modal coordinates, is shown to be highly efficient for the analysis of fixed base offshore structures. The numerical study shows that only 3–5 iterations are required to obtain the converged solutions for hydrodynamic loadings on the structure produced by random and regular waves. The computational time is marginally greater than that required for the linearized methods which in certain cases may introduce considerable error in the results.

The estimation of frequency-dependent correlation functions for pairs of accelerograms is important in the analysis and modeling of the space-time variation of earthquake ground motions. Two methods of estimating the frequency-dependent correlation functions, the conventional method and the ``moving window'' approach, are reviewed. It is shown that unless certain precautions are taken the moving window approach can sometimes yield estimates that are heavily biased and hence meaningless.