A cellular manufacturing system consists of several work cells in which parts are processed under machines. Identification of parts and machines in each work cell, known as cell formation, is a major step in design of a cellular manufacturing system. This thesis presents a method for cell formation. The method uses clustering approach to identify work cells in three steps. First, production information represented in an incidence matrix is evaluated for finding the coupling relationships between parts and machines. In second step the gathered coupling information in the first step is used to reorder the incidence matrix rows and columns, and create a tree diagram. Using the tree diagram, work cells and parts and machines in each work cell are identified in the third step. Performance of the method is evaluated by solving two types of cell formation problem.  The results indicate that the method can produce solutions as good as other methodologies. In comparison to clustering cell formation methodologies, the method has a flexible solution procedure that simultaneously groups parts and machines, and it does not need predetermined production information for executing cell formation.