This paper presents an expert system approach which, given a product model, automatically generates assembly sequences for robotic assembly system. The assembly sequences thus obtained basically characterizes key features of an assembly system required for the product assembly; assembly line layout, assembly cell task planning, fixture complexity, assembly operation difficulty, and unit assembly cost. This approach consists of three hierarchical modules; (1) product modeling, (2) precedence constraint inference, and (3) assembly sequence inference. The given product is modeled in terms of connections between two connecting parts, using human intelligence excellent in recognizing topological relations. Each connection is modeled in terms of connecting methods in 6 directions in the form of 2X3 contact matrix and 2X3 fit matrix. Based on the contact matrices of all connections, three orthodirectional contact level graphs are drawn to show contact level of each part along three assembly axes X, Y and Z.
A precedence constraint for each connection is inferred through a graphical method which is based on the path-finding algorithm in the contact level graphs. The precedence constraint for part to be assembled with a base assembly is inferred as union set of precedence constraints of all connections between the part and the base assembly.
The assembly sequence has been inferred in two ways and discussed comparing to each other: (1) linear plan, (2) subassembly extraction plan. The linear plan is to assemble one part at each assembly operation and is inferred in the forms of feasible sequence and stable sequence. The feasible sequences are inferred only considering the precedence constraints, wheras the stable sequences are inferred, in addition to precedence constraints, calculating instability of base assemblies when their postures are required to change. The subassembly extraction plan is to extract stable subassemblies based on the stable fastening connections and, regarding them as the same as the other single parts, is inferred in the forms of feasible and stable sequences, respectively.
Finally, in oreder to provide the optimal plan for the robotic assembly automation, the optimal sequence is inferred considering the design criteria of the robotic assembly system such as fixture complexity, robot dexterity, and the number of turning devices.
This approach has some advantages such as no requirement of geometric data and geometric reasoning taking excessive time and efforts, automatic generation of precedence constraints by graphical algorithms taking short time, and automatic generation of feasible, stable and optimal assembly sequences on user menu choice, reflecting the various practical conditions of robotic assembly system. To verify the proposed algorithms, the assembly sequences of relay product, alternator assembly, etc. have been obtained in computer and on precedence diagram and compared.