A model is proposed to understand the structuring of social networks in a fixed setting such as, for example,inside a university. The friendship formation is based on the frequency of encounters and mutual interest. The model shows distinctive single-scale behavior and reproduces accurately the measurable experimental quantities such as clustering coefficients, degree distribution, degree correlation, and friendship distribution. The model produces self-organized community structures and can be described as a network of densely interconnected networks. For the friendships, we find that the mutual interest is the dominant factor, which optimizes the network and that the number of encounters determines the statistically relevant distributions.