© I FAt: Infllrlll ;t ri lln C() n(f"o[ Prohlems in \ fa llu fatl urin g: Te{ hll o[og\. \l adrid . Spa in I~)~~'
( : (J p ~r i g hr
K:--IOWLEDGE BASED MODELLING 1I
KNOWLEDGE-BASED ORDER SPECIFIC NC DRILLING SYSTEM L. F. Pau*, D. Paus** and T. Stokka** *T fCit llira l L'lIh 'nsily of Denmark. Lvllglrl. Dfllmark **Surilir IlIlelligf111 .Halllifaclunllg. Slabekk. Soro.'ay
Abstract: The paper describes a knowledge based system which generates hole positions/types, as well as the corresponding NC macros , on custom order beams for trucks. This system exploits components data bases, CAD, and configuration procedures. It is implemented part in PROLOG, part in other languages, and uses a meta parser. The main benefits are in order customization, and reduced manufacturing times for such orders.
D
mventory
Keywords: Expert system, NC machines, Customized manufacturing, Automotive industry, Drilling, Planning.
C
descriptions of the uses of each component, and of its relations to other components
updates to production schedules at the flow control level
G
updates to NC macros ~stributed
processing and data communica-
~reviously, the generation of NC macros was carned. out manually; the operation was excluding customIzed new hol~s ?n the b.asis of his inspection of component descnptIons . This process resulted in 310 erroneous holes being drilled per day; these errors had to be corrected by tedious welding not to mention production delays and broken drills.'
The kno,:"ledl!'e ba.sed ord~r specific NC drilling system (S), IS pnmanly making use of the following information sources: Information source A Parts description database B Specification p rocedures C Fact base: use of each part and relation to other parts
To achieve the above ambitions, speecial emphasis must be put on the following techniques (Figure 1):
specification procedures, with data structures encompassing all parts description databases
F
The p~oject carried out for one company. has been foc~ssl~g on one specific aspect of the above concept, :-vhlch IS k!lowledge based order specific NC drilling m the mam beams o~board of mediumlheavy duty tru~ks. The ge.ometncal layout of holes is today avaIlable, both m drawings, and in CAD files (under a D~ssault SystemslIBM CATIA system). Custom specIfic hole layouts must be generated automatically to. J?roduce ready-to-use NC command macros. In addItion, the designer must have explanations about how the customized layouts were verified and about related production costs per beam. '
This production philosophy is quite similar to the one presiding in microelectronics over ASIC's (Application specific integrated circuits), or printed circuit boards and cell libraries in particular. "Building block" CAD files are available from the library, but are assembled and simulated jointly to achieve customized circuit or board needs (Battelle, 1987).
B
updates to CAD files
tion
The background for the project below lies in the overall ambition of several major European car and truck manufacturers, to be able to customize their products to specific client requests, while minimizing costs and inventories. In this way, the same manufacturing facilities should be able to produce customized parts , obtained by revisions to basic CAD designs but checked for geometrical and assembly compatibility, (Pau, 1986).
components description database, with attributes for all variants parts
parts conparts in
E
H
1. INTRODUCTION
A
order v~rification procedures, for ~guratlOn and checks against
Implementation IBMIMVS/ ROSAM
Language PUl
IBMlMVS/CAPRI
PU1
MS/ DOS
PROLOG 11 (see Gian nesini, 1987) FORTRAN NC command lanl!Uage PRO LOG 11 + graphics commands
E CAD files for ho les G NC macros
IBMlMVS/CATIA MS-DOS
S Knowledge based order-specific NC drillin£ system
MS·DOS
Table 1: Data and fact bases
2i5
276
L. F. Pa u , D. Paus a nd T. Stokka
This system (8) must be used interactively by the designer, and by the NC machine operator. It must give explanations to generated hole layouts, to generated macro commands, and to the use/justification of specific holes (in relation to parts assembly).
User interface: It consists of drivers to parts description data base (A), to CAD system (E), and to NC machine controls (G). It is supplemented by a menu driven windows environment, with alternative set-ups for different user groups: Designers: They get hole layouts encoded by hole types and origin (standard layout, customized variants); a mouse pointed at a hole generates an explanation for that placement and selection . The designer is also allowed to update knowledge bases, E, 8 through a simple editor. This editor is supplemented by a forward chaining inference procedure for the exploitation of all consequences of a given custom design. For each such design, a cost estimate is generated by 83.
ARCHITECI'URE OF THE KNOWLEDGE BASED SYSTEMS (S)
The architecture selected, is a hybrid knowledge based system, with interfaces to PLlI, FORTRAN, and the NC machine command language. The inference and explanation procedures are coded in PRO LOG n, with interface procedures, running on a PC/AT compatible with RS-232 links to other mainframes .
NC machine operators: They get the sequence of NC macros in one window, and in another window a script over holes when drilled as a result of the drill sequence . The operator can edit this sequence, and shift some holes to adjust to tolerances etc. This edition is supplemented by a backward chaining inference to verify given hole locations or macros, and check the tool use sequence. A drilling sequence router and optimizer is implemented, which selects the hole sequence on the basis of hole attributes and drilling criteria (e .g. minimize number of tool changes); this planner uses exploration in attributed graphs.
ii
8pecific to (8) are: a meta-parser, capable of parsing symbol strings of different types (both components descriptors, and hole geometries in the beam) ii
a knowledge representation, with fusion of all required information sources in fact or procedural form
iii
a user interface, with specific user set-ups.
Knowledge and fact bases: The information sources, A,B,C,D,E,G lead to the following knowledge or fact bases, with each their specific data structures: AI : Parts codes with variants and text descriptors of their uses; these are of the type nTm8 where n is the parts number string, T the type, m the variant code, and 8 the text descriptor string.
PARSER
Meta Parser
This meta parser (see Table 2) is used to analyze and evaluate all strings in AI , Cl , El, and GI for different selections of the terminal vocabularies. Example
Cl: Description string of the use of each hole, and of the other parts mounted on the beam via this hole
For example, in the case of a four hole layout DI, this one could be described by the string:
a)
El: Nominal hole layout of a beam section, with coordinates, diameter, and type/profile of each hole GI : Data structures for all NC drilling macros, written in meta-assembler syntax
layout = a(tl,dl) + (b-c) (t2,d2) + 3*e(t3,d3) where:
81: Lists of alternative hole types, for a given coordinate range on the beam; variants to the same due to differences in drilling tools and profiles for the NC machine 82: Exclusion rules for eliminating tentative hole placements and/or type, in view of: placement procedures, of minimal distance to edge or other holes, etc. 83: Cost evaluation procedures, in view of an optimized drilling path; the COSts account for set-up time, drilling time, traveValignment time, and tool changes
Knowledge representation: This is essentially in form of logic predicates (coded in PROLOG), operating on lists generated by the dedicated parsers, and with attributes attached to the list header. The parser describing hole geometry allows for portability of beam geometries into knowledge bases of type 8 1/2/3.
a,b,c,e:
are given vectors in two dimensions, of lengths (x,y) projected according to a coordinate system on the beam
(ti,di ):
type and diameter of i-th hole generated by the string to the left
+,-
vector operations
3*:
scale factor
The parser then returns the structured decomposition of the layout, and the way in which these elements relate in terms of displacements of NC commands (+,-). b)
In the case of a numerical calculation, the operator would enter a formula to be evaluated in string form : cost =
HOLE(t,d)*5+RANGE(d)*256+ 8ETUP
~c
Drilling Syste m
where :
CONCLUSION
HOLE(t,d): drilling cost per hole of type t and diameter d RANGE(d):number of tools, i.e . of hole diameters, in a given layout SETUP:
277
set-up cost
The parser would then return in graph form, the way in which the functions operate on different data sets, and also the calculated value of the expression.
Such a knowledge based order specific NC drilling system (S) is of strategic importance for the truck/car manufacturer for which the project is carried out, in view of the following new capabilities, made possible by the applications of artificial intelligence (Krakauer, 1987):
•
bring to the NC machine shop floor integrated design and manufacturing data
•
reduce the errors due to end pro due t customization control costs due to customized layouts generate NC macro sequences
Terminal vocabulary
Application
Parts codes
Hole uses
Hole layout
Drilling
increase the dialogue between designer and manufacturing engineers
Knowledge base
Al
Cl
E1JE2
GlJG2
Elements in vocabulary
Separators
Integers (n ,m) Words (T) Strings of words (S)
Spaces
Parts codes
Separator (.) and parenthesis 0
Vectors (x,y) giving transitions betweenhole positions
Numerical calculations
with type and diameter attributes
Operations on vectors: +: addition -: sub straction scaling 0 : paren thesis
Drilling macro
Separator
..
numbers or names
S3
Numbers
reduce bottle-necks around a truck component (here the main beam) of paramount importance to the total manufacturing cycle reduce handling of drawings for custom designs, and thus adjustment time
BmLIOGRAPHY
()
macros
•
Functions
Table 2: Knowledge bases and the oon-esponding vocabularies as used by the parser
1.
Battelle Institute, Artificial intelligence in manufacturing, GenevalColumbus (1987)
2.
Giannesini, F ., Kanoui, H., Pasero, R., Van Caneghem, M., PROLOG, Addison-Wesley Publ. (1987), or Inter Editions, Paris (1985)
3.
Krakauer, J., (Ed.) Smart manufacturing with artificial intelligence, SME Publications, Dearbom, MI, (1987)
4.
Pau, L.F., Artificial intelligence in economics and management, North Holland Publ., (1986)
L. F. Pau. D. Paus a nd T. Stokka
27 8
Meta-parser in PROLOG
TableS:
enter (p)
--t
read (p) expression (e,
word (a,
--t;
expression (e,
sum (e,
--t
product (p,
--t
product (p,
draw-tree (e);
primary (f,
--t
primary (n,
--t
word (n,
primary (e,
--t
word ("C',
rest-of-sum (p,e ,
word (0, ,e,
--t
rest-of-sum (e,e,
rest-of-product (f,f,
--t
integer (n);
word (O,
DEFINITIONS OF OPERATIONS:
Op-add, op-mul in relation to symbols +,-, *; or text: add, sub, mul
READING INPUT STRINGS for parsing READ STRING ATTRIBUTES from external knowledge bases by drivers INTEGRATE OTHER KNOWLEDGE BASES EVALUATE EXPRESSION, AND OUTPUT RESULT including parsing tree and colouring of beam layout
" C: Drilling S\"ste m
279
CUSTOM ORDER
l ~ USER INTERFACE
DESIGNER
+--+
NC MACHINE OPERATOR
KNOWLEDGE BASED DRILLING SYSTEM S1: Hole types S2: Exclusion rules S3: Cost evaluation
i NOMINAL HOLE LAYOUT (El)
DRILLING MACROS (Gl)
HOLE USE AND RELATIONS
PARTS CODES (Al)
NEW HOLE LAYOUT (E2)
Figure 1:
GRAPHICS/ EXPLANATIONS
NEW MACROS (G2)
Knowledge based order specific NC drilling system (S).