The Analysis and Prediction of Steel Bar’s Mechanical Properties by Intelligent System

• Main Authors: Ying-tsung Chen, 陳盈璁
• Other Authors: Rey-Chue Hwang
• Format: Others
• Language: zh-TW
• Published: 2009
• Online Access: http://ndltd.ncl.edu.tw/handle/61193491655215772750

碩士 === 義守大學 === 電機工程學系碩士班 === 97 === The steel bar is the necessary and important material widely used in many engineering constructions, including business building, bridge, and road. Its quality is highly related to the safeties of construction and human’s life. In fact, the degree of withstanding earthquake of the construction is closely linked with the quality of steel bar. Thus, many countries have set the standard of the bar’s quality. Basically, the disqualified steel bars are not allowed to sell and must be melted and reproduced. Any failed steel bar will certainly increase the cost of the steel manufacturing company. Therefore, how to make a good control in the manufacturing process of steel bars becomes a very important issue for the manufacturers. It’s also the aim of this research. Usually, in the rolled process of steel bar, the relative control parameters, such as size, rolling speed and hydraulic pump and water segment, are mainly determined by the technician with full experiences in accordance with the compositions of billet. However, the compositions of billet include too many chemical elements. Some of them are even unknown, especially when the sources of metal scrap came from different countries. Such a simple way for setting the control parameters based on human’s experiences easily makes the steel bar produced be disqualified. In other words, it also implies that the cost of steel company will be increased with no doubt. Recently, due to the powerful learning and adaptive capabilities, neural network has been widely applied into engineering and business areas. Through a simple training, neural network can automatically develop the complex and nonlinear relationships between input and output pairs of training data provided. Such a well-trained network then can be used to perform a specific work. In this research, the mechanical property estimator of rolled steel bar by using neural network was studied and developed. Such an estimator is expected to help the technician to set the related control parameters for the rolling process of steel bars.