دانلود فایل مقاله
The Artificial Neural Network is one of the branches of Artificial Intelligence that is inspired by human nervous system. In neural networks, data processing is done through an interconnected network of little processors that act in parallel with each other. To solve rigid connection design, in welded metal structure building that has numerous stages, these networks were used because of their high speed processing and to show whether or not design speed is affected by processing speed. After writing network using MATLAB software and introducing all of design formulas, initially, 80% of data was used for network training and after the network got familiar with the pattern, by using 20% remaining data, the network was tested. Full analysis of design of some options, such as welding electrode material and weld size based on tenth subject of National Building Regulations, should be selected in designing, but since the choice of these variables is done mainly through the maximum 3 different types, and also they are constant in much of the designs, so it was not used in this article. All of connections should be considered rigid and web-flange connection should be full penetration diffusion welding. The thickness of reinforcing fillet welds on each side of web was considered thicker than 8 mm. the used neural network was perceptron network (MLP) with the back-propagation algorithm (BP), respectively.