NASA's Jet Propulsion Laboratory, Pasadena, California
A system of algorithms is being developed to automate the planning of sequences of manufacturing or field operations in which components are put together into assemblies. The system is based partly on the concept of recursive decomposition of an assembly into subassemblies. To guide the generation of a preferred assembly plan, the system incorporates design-for-assembly (DFA) analysis, in which the designs of subassemblies and components are analyzed for their effects on the feasibility and cost of the assembly sequence; the results of DFA analysis can also be used to modify the designs to optimize them with respect to assembly considerations.
The planning system takes account of special processes (for examples, cleaning, testing, and labeling), which must occur during the assembly, and handles nonreversible as well as reversible assembly tasks through backward assembly planning. To increase the planning efficiency, the system avoids the analysis of decompositions that do not correspond to feasible assembly tasks: this is achieved at each stage of the sequence by grouping and merging those parts that cannot be decomposed at this stage because of the requirements of special processes and the constraints imposed by the feasibility or infeasibility of the affected interconnections between parts.
The system proceeds as follows: First, the special processes involved in making the product assembly are represented by a symbolic tree or set of trees called a "special process forest" (see Figure 1) and are incorporated into the backward assembly planning via a grouping principle. Given the special process forest, the grouping principle governs the identification of those parts that should be grouped together in a subassembly at the current stage of backward assembly planning, to enable the special processes to be carried out properly. In addition to the grouping principle, there is a merging principle, according to which those parts that are not decomposable at the current stage of backward assembly planning because of the infeasibility of interconnection are merged. Together, these two principles help to reduce the complexity of the space of alternative sequences that has to be searched. The system then proceeds to introduce criteria of stability, directionality, assembly pose, manipulability, process planning, and parallelism, and to quantify these criteria for use in selection of the best subassemblies in backward assembly planning. Most significantly, these criteria are evaluated with a direct connection to the cost of assembly on the basis of (1) the identification of the number of holding devices to stabilize assembly operations, (2) the derivation of the number of reorientations required during mating operations, (3) the determination of the best assembly poses for individual subassemblies generated during planning, and (4) the estimation of the effect of the manipulability of parts and subassemblies on the cost of mating.
Next, a globally optimal plan is found by an algorithm called "AO*," which searches a symbolic tree (see Figure 2) that represents alternative sequences of decomposition of the assembly into its components. The tree contains AND and OR nodes. The decisions at the OR nodes are made with the help of a cost function and a heuristic function defined in terms of the criteria mentioned above. In the process of searching for an optimal assembly plan, DFA analysis is performed for each assembly operation on the basis of the detailed evaluation of these criteria. The result is summarized into a DFA analysis table.
This work was done by Sukhan Lee of JPL/Caltech for NASA. More details can be found in RBackward Assembly Planning with DFA Analysis,S Computer- Aided Mechanical Assembly Planning, edited by Homen de Mello and Sukhan Lee, Kluwer Academic Publishers, Norwell MA., pp. 341-381.
Point of Contact:
Sukhan Lee
Mail Stop 198-219
Jet Propulsion Laboratory
4800 Oak Grove Drive
Pasadena, CA 91109
818-354-2013
Sukhan.Lee@jpl.nasa.gov
Telerobotics Program
page
Please email the site webmaster
with any comments, criticisms or corrections
for this page.
Maintained by: Dave Lavery
Last updated: May 10, 1996