Page 145 - Mechatronics for Safety, Security and Dependability in a New Era

Ch27-I044963.fm Page 129 Monday, August 7, 2006 11:26 AM
Page 129
11:26 AM
Ch27-I044963.fm
Monday, August 7,2006
129
129
ASSEMBLY SEQUENCE PLANNING
USING K-NEAREST-NEIGHBOR RULE

1
2
T. Murayama , T. Eguchi , and F. Oba 2
'Division of Oral Health Engineering, Faculty of Dentistry, Hiroshima University
1-2-3 Kasumi, Minami-ku, Hiroshima, 734-8553, Japan
pt. of Mechanical Systems Engineering, Hiroshima Univer
1-4-1 Kagamiyama, Higashi-hiroshima, 739-8527, Japan

ABSTRACT

This paper describes an approach to the efficient planning of assembly sequences. K-nearest-neighbor
rule reduces the search space for the assembly sequences by using sample data on products, of which
assembly sequences are known. Additional sample data are made from the assembly sequences
generated by this approach. As the assembly sequence planning and the addition of the sample data
are executed more times, the assembly sequences can be generated more efficiently. Some
experiments are carried out to show: the effectiveness and efficiency of the approach; and the
superiority of the k-nearest-neighbor rule over the heuristics that were used in our previous work.

KEY WORDS

Assembly Sequences, Assembly Planning, K-nearest-neighbor Rule, CAPP, CAD/CAM

INTRODUCTION

Recently, many research efforts have been made to plan assembly sequences automatically and
efficiently. Most of the existing approaches generate a disassembly sequence by identifying a part or
subassembly to be removed from a product repeatedly, and then generate an assembly sequence by
reversing the disassembly sequence (Lambert, 2003.) In order to identify a part or subassembly to be
removed, the approaches test which parts and/or subassemblies can be removed from the product. The
tests for all the parts and/or subassemblies are computationally very expensive, especially in the case
that paths to remove them are searched for at the tests. Therefore some of the approaches focus on
reducing the number of the tests. Bourjault (1984) proposed superset and subset rules that can avoid
the unnecessary tests; however, the number of the remainder (i.e., the necessary tests) is still large
especially for the products composed of many parts. Subassembly extraction (Lee & Yi, 1993) and
heuristics (Murayama & Oba, 1993) are effective to reduce the number of the tests further.