| The practical applications of learning systems are still limited due to the lack of a methodology for their use, particularly in the area of design knowledge acquisition. As the result of this project, such methodology has been proposed, and it is based on the results of experiments with several commercial and experimental learning systems. The experiments were conducted in the area of conceptual design of wind bracing systems in steel skeleton structures of tall buildings.; The proposed methodology includes methods for the preparation of examples, for the conducting and monitoring the learning process, and for the verification of knowledge produced as the result of learning. It is based on the concept of a multistage learning process, in which all available examples are divided into several groups, and the learning is gradually performed using examples from individual groups and their combinations. Its progress is monitored using various empirical error rates and proposed monitoring learning parameters as well as their learning curves.; The learning processes were conducted using two collections of examples: expert-prepared and generated by SODA, a computer program for the analysis, design, and optimization of steel structures. Their results demonstrate that learning systems can be used for engineering design knowledge acquisition. In the reported learning processes, the best results in terms of error rates were obtained for the computer-generated examples. In this case, the overall error rates are within range 3.9%-7.6%, i.e., if the produced knowledge is used in a knowledge-based decision support system, it will produce advice regarding unknown cases which will be correct in 96.1%-92.4% of cases.; The proposed methodology was developed using structural design examples, and the feasibility of inductive learning in knowledge acquisition was clearly demonstrated only in this area. However, both the methodology and feasibility results should be useful in the other areas of engineering, although in this case more research might be necessary. |
