Innovative Computer Technologies for Integrated Design of Steel Structures

这项研究的目的是利用几种新兴的计算机技术开发用于钢结构集成设计的计算机模型。它们包括面向对象编程(OOP)、遗传算法、神经网络计算和并发处理。该研究包括a)使用OOP范例开发用于钢结构集成设计的黑板体系结构，b)使用增强的实体关系模型来管理在钢结构集成设计中遇到的大量输入、中间和输出数据的面向对象的数据库管理系统，c)基于并行版本图的面向对象的版本管理模型，D)用于钢结构详细设计的增广拉格朗日遗传算法，e)用于训练多层前馈神经网络的自适应共轭梯度集成设计机器学习算法，f)将遗传算法与误差反向传播多层神经网络相结合的混合学习算法，G)使用共享存储机器(Cray YMP 8/864)和分布式多计算机(Transputer)的多处理能力对高层建筑等大型钢结构进行综合/优化设计的并行算法，以及h)混合神经网络/遗传算法的并行算法。

The objective of this research is to develop computer models for integrated design of steel structures using several emerging computer technologies. They include object-oriented programming (OOP), genetic algorithms, neural network computing, and concurrent processing. The research includes development of a) a blackboard architecture for integrated design of steel structures using the OOP paradigm, b) an object-oriented database management system using an enhanced entity relationship model for managing a multitude of input, intermediate, and output data encountered in integrated design of steel structures, c) an object-oriented version management model based on the parallel version graph, d) an augmented Lagrangian genetic algorithm for detailed design of steel structures, e) an adaptive conjugate gradient integrated design machine learning algorithm for training of multilayer feedforward neural networks, f) a hybrid learning algorithm by integrating genetic algorithm with error backpropagation multilayer neural networks, g) concurrent algorithms for integrated/optimum design of large steel structures such as highrise buildings using the multiprocessing capabilities of shared memory machines (Cray YMP 8/864) and distributed multicomputers (transputers), and h) concurrent algorithms for the hybrid neural networks/genetic algorithm.