Perceptron Learning in Engineering Design

A model of machine learning in engineering design is presented based on the concept of self-adjustment of internal control parameters and perceptron. A perceptron is defined as a four-tuple entity which can answer either "yes" or "no" in the problem domain. The problem of structural design is cast in a form that can be described by a perceptron without hidden units. Some results from our experimentation are presented in tabular form. The paper is concluded by a comparison of perceptron and explanation-based learning.     

Microcomputer-Aided Design and Drafting of Moment-Resisting Connections in Steel Buildings

A microcomputer-aided system has been developed for interactive design and drafting of connections in steel buildings made of standard. rolled-I sections, called STEELCON. The basis of design is AISC specification. In addition to interactive design, STEELCON can displaylplot any isometric view of the connection plus three orthographic views, i.e., front, side, and top face views. The orthographic views show all the necessary details for fabrication and construction of the connections. Interactive design and drafting of two different moment resisting connections are presented. They are the shop welded and field bolted framed flange plate connection and the shop and field welded flange plate connection.     

A Knowledge-based Expert System for Design of Roof Trusses

Aknowledge-based expert system has been developed for design of roof trusses, called RTEXPERT. RTEXPERT can advise the user on the appropriate type of the roof truss, selection of the layout of the truss (such as the pitch of the truss and number of panels), and the loading. The basis of design is the American Institute of Steel Construction (AISC) specification. The truss is designed for dead, live, snow, and wind loads in accordance with the American National Standard Institute (ANSI) specification. A novel part of RTEXPERT is the automatic computation of nodal forces due to various loads. The user needs to indicate only the types of the materials used as roof covering and the location of the structure in the United States. RTEXPERT automatically generates all the nodal forces. The knowledge base and explanation facility of RTEXPERT have been developed using INSIGHT 2 + expert system shell. The mathematical computations, graphic algorithms, and data file manipulation routines have been developed in Turbo Pascal. RTEXPERT has a comprehensive graphic interface for displaying the truss configuration, cross sections, loading, and deformed shape. Information about individual members is presented through multi-window graphicstext displays. RTEXPERT can be used as an "intelligent assistant" for design of roof trusses.     

COST OPTIMIZATION OF STEEL STRUCTURES

While the weight of a steel structure is a major component of the total cost, the minimization of the cost should be the final objective for optimum use of available resources. The total cost of a steel structure includes (a) the material cost of structural members such as beams, columns, and bracings, (b) the fabrication cost including the material costs of connection elements, bolts, and electrodes and the labor cost, (c) the cost of transporting the fabricated pieces to the construction field, and (d) the erection cost including the material costs of connection elements, bolts, and electrodes and the labor cost. In this article, a chronological review of the journal articles on cost optimization of steel structures is presented. Articles on deterministic, reliability-based, and fuzzy logic-based optimization of steel structures are reviewed. Research on cost optimization can encourage the use of the optimization approach in structural steel design practice by providing a more realistic way of modeling structural steel design and resulting in additional savings compared with the weight optimization problem.     

Automatic Partitioning of Frame Structures for Concurrent Processing

An efficient 3-stage algorithm for automatic partitioning of frame structures is presented. Once the input data of the structure is fed into the computer, the domain is partitioned into a number of subdomains equal to the number of the specified processors. The underlying concept is to preprocess the input data in such a manner that subsequent analysis and optimal design steps can be carried out efficiently on multiprocessor computers. Emphasis is directed towards attaining a balance of workload among the processors, minimizing the bandwidth within each subdomain, and minimizing the number of linear equations to be solved for the interface degrees of freedom. All the vectors involved in the partitioning algorithm are integers, requiring a small storage and resulting in calculations that can be processed fast. The algorithm is specially suitable for analysis and design of large structures such as space stations. The algorithm presented is hardware independent. Our applications, however, are limited to the Encore Multimax shared memory computer. The efficiency of the algorithm is demonstrated through partitioning several truss and frame problems, for a variable number of processors.     

Microcomputer-Aided Instruction of Structural Steel Design

An innovative approach is reported for teaching structural steel design in a civil engineering undergraduate curriculum. It is based on the interactive environment and graphic capabilities of inexpensive microcomputers. Ten BASIC interactive microcomputer-aided design programs have been developed and used in a senior-level structural steel design course at the Ohio State University. The approach of interactive design and the interactive microcomputer-aided design programs make the learning of structural steel design more exciting and result in more effective training of future designers. Students can gain experience and develop insight in structural design while they are still in an academic environment.