Europäisches Patentamt

(19) European Patent Office

Office européen des brevets (11) EP 1 074 927 A2

(12) EUROPEAN PATENT APPLICATION
(43) Date of publication: (51) Int. Cl.7: G06F 17/50
07.02.2001 Bulletin 2001/06

(21) Application number: 00202796.9

(22) Date of filing: 07.08.2000

(84) Designated Contracting States: • Boose, John H

AT BE CH CY DE DK ES FI FR GB GR IE IT LI LU Bellevue, Washington 98008 (US)
MC NL PT SE • Boose, Molly L.
Designated Extension States: Bellevue, Washington 98008 (US)
AL LT LV MK RO SI • Post, Michael Dee
Edmond, Oklahoma 73013 (US)
(30) Priority: 05.08.1999 US 147272 P
(74) Representative:
(71) Applicant: THE BOEING COMPANY Land, Addick Adrianus Gosling et al
Seattle, Washington 98124-2207 (US) Arnold & Siedsma,
Advocaten en Octrooigemachtigden,
(72) Inventors: Sweelinckplein 1
• Baum, Lawrence Sterne 2517 GK Den Haag (NL)
Bellevue, Washington98005 (US)

(54) Intelligent wiring diagram system

(57) An intelligent wiring diagram system for auto-

matically converting electronic wiring diagrams into
intelligent wiring diagrams for display on a computer ter-
minal, the intelligent wiring diagram system primarily for
use as an interactive tool for trouble-shooting electrical
problems.
EP 1 074 927 A2

Printed by Xerox (UK) Business Services

2.16.7 (HRS)/3.6
 EP 1 074 927 A2

Description

CROSS-REFERENCE TO RELATED APPLICATIONS

5 [0001] This application is being converted to a utility application from provisional patent application Serial No.
60/147272, filed August 5, 1999.

TECHNICAL FIELD OF THE INVENTION

10 [0002] The present invention relates to electronic wiring diagrams and more particularly to a system for automati-
cally converting legacy wiring diagrams into "intelligent wiring diagrams".

BACKGROUND OF THE INVENTION

15 [0003] Mechanics use wiring diagrams in the trouble-shooting of electrical problems on an airplane. The diagrams
identify the wires and electrical components used in a particular subsystem of a plane, and also illustrate the flow of
electricity through the circuits. However, the diagrams are difficult to read, electrical connectivity of electrical compo-
nents is not always clear, and mechanics find it time consuming to manually look up component numbers in an electrical
component database or large paper-based sources when there is no available database. This situation is exacerbated
20 by the sheer volume of paper, as wiring manuals are often 5000 pages or more.
[0004] Most airplane systems have from dozens to hundreds of different potential configurations, reflecting different
modifications, capabilities and manufacturing processes. Maintenance personnel spend a significant amount of time
searching for the right configuration and consulting multiple volumes of binders to find the right component or system.
Component density on the diagrams can also impede finding the right component on a diagram. These factors compli-
25 cate the technician's job, as pressure increases to find and comprehend the necessary information quickly and accu-
rately.
[0005] There are no approaches in use commercially that solve the problem of rendering wiring diagrams in elec-
tronic data products in a way that makes the diagrams easily comprehensible or useful for mechanics. For example,
Wiring Diagrams are not yet available in a portable maintenance aid product, principally because no satisfactory solu-
30 tion could be found for adequately rendering them. Consequently, mechanics continue to deal with wiring problems by
using paper copies of the diagrams. Commercial vendors are offering tools for creating intelligent wiring diagrams, but
they require manual re-authoring of the data to identify objects and make them interactive. This re-authoring process is
extremely labor-intensive and error-prone as it requires manually creating hundreds of individual mouse-sensitive areas
for each drawing (this is called "hotspotting" or "creating hotspots") and specifying in a complex language what the sys-
35 tem should do when the user points at or clicks each hotspots. For systems such as airplanes that involve many thou-
sands of such diagrams, the cost of this manual approach is prohibitively expensive. Consequently, this invention
provides capabilities that will otherwise not be made available.

SUMMARY OF THE INVENTION

40
[0006] This invention automatically converts legacy electronic wiring diagrams into "intelligent wiring diagrams".
This significantly improves the usability of wiring diagrams when displayed on a computer and makes the trouble-shoot-
ing of electrical problems easier, faster and more reliable. Using the intelligent graphics viewer, users can quickly com-
prehend complex electrical circuitry, simulate the changes in electrical continuity and access critical component
45 information.
[0007] The present Intelligent Wiring Diagram System automatically and quickly converts wiring diagrams into elec-
tronically useful, interactive graphics. This novel approach completely eliminates the need for manual re-authoring.
Within a few seconds the software is able to accurately identify circuit continuity and nomenclature of interest and put
hundreds of hotspots on a diagram. If an expert were asked to perform the same work, it would take many days of inten-
50 sive labor to re-author and validate a single wiring diagram.
[0008] The invention consists of a set of software modules that perform the conversion and a graphics viewer that
provides the interaction. These modules are:

A. ELECTRICAL COMPONENT RECOGNIZER

55
[0009] The Electrical Component Recognizer searches through the graphical primitives in the original wiring dia-
gram and applies heuristics in order to identify components such as fuses, circuit breakers, various types of switches,
electrical connectors and relays.

2
 EP 1 074 927 A2

B. WIRENET RECOGNIZER

[0010] Using the electrical symbols found by the Electrical Component Recognizer, the Wirenet Recognizer
employs a unique and powerful Attachment Point Algorithm to discover which lines in the diagram represent individual
5 wires and to infer the electrical continuity. The algorithm works as follows:
[0011] For each electrical component there are predictable locations where wires can attach; these are the attach-
ment points. For example, in a circuit breaker wires attach at the terminal circles. Similarly, there are attachment points
for fuses, resistors, and the other electrical components, except connectors which are handled separately. Also, it is
common for designers to erase part of a wire to make space for a text string such as the wire number or a notation.
10 Consequently, for wirenet generation purposes, we consider text elements to have their own attachment points where
wires may be attached.

C. NOMENCLATURE RECOGNIZER

15 [0012] Using text matching techniques, the Nomenclature Recognizer scans the text elements in the diagram and
finds those that match wire numbers, wire bundles, grounds, splices, terminals and other equipment numbers from
databases.

D. XML GENERATOR
20
[0013] Extensible Markup Language (XML) provides a standardized methodology for describing information in a
document. We have created a unique Document Type Definition (DTD) for describing graphical objects and the relation-
ships between them, including specific elements for describing electrical components and electrical continuity, as well
as references to part information. The SGML Generator produces the XML markup for all of the objects discovered by
25 the three recognizers (electrical components, wirenets and nomenclature). The markup includes the geometry for each
object's hotspot and the relationships between objects and other objects and between objects and hotspots.

E. INTELLIGENT GRAPHICS VIEWER

30 [0014] Finally, the system includes a unique graphics viewer which not only renders the Wiring Diagram in its orig-
inal form, but also uses the output of the XML generator to provide visual identification of objects and user-driven inter-
activity with the diagram. Every wirenet is "hot"; when the user points at the wire with a pointer device (e.g. a mouse),
the wire and its complete continuity immediately highlight. Furthermore, the user can affect that continuity by interac-
tively changing the state of electrical components. The viewer also allows full text searching and highlighting of hits
35 found on the diagram.

BRIEF DESCRIPTION OF THE DRAWINGS

[0015]
40
Figure 1 is a diagram illustrative of the Intelligent Wiring Diagram Recognizer of the present invention showing Rec-
ognizer Top Level Flow;
Figure 2 is a diagram illustrative of 200 Store Graphical Primitives;
Figure 3 is a diagram illustrative of 300 Recognize Symbols;
45 Figure 4 is a diagram illustrative of 310 Discard shade lines;
Figure 5 is a diagram illustrative of 320 Find dashed lines;
Figure 6 is a diagram illustrative of Find Circuit Breakers;
Figure 7 is a diagram illustrative of 390 Recognize Ganged Circuit Breakers;
Figure 8 is a diagram illustrative of 400 Recognize Wire Nets;
50 Figure 9 is a diagram of 400 continued;
Figure 10 is a diagram illustrative of 500 Generate Output, 510 Generate XML and 520 Generate HTML;
Figure 11 is a diagram illustrative of 530 Generate CGM;
Figure 12 is a diagram of 530 continued;
Figure 13 is a diagram illustrative of 1000 IG Viewer;
55 Figure 14 is a diagram of 1000 continued;
Figure 15 is a further diagram of 1000 continued;
Figure 16 is a further diagram of 1000 continued;
Figure 17 is illustrative of Circuit Breaker 10;

3
 EP 1 074 927 A2

Figure 18 shows wires interrupted by labels;

Figure 19 shows wires interrupted by a "gap";
Figure 20 illustrative of recognizing a circuit breaker symbol;
Figure 21 shows wirenets beginning to sprout from a circuit breaker;
5 Figure 22 is illustrative of part of a wiring diagram after sprouting with different wirenets;
Figure 23 shows part of a wiring diagram after growing wirenets;
Figure 24 shows part of a wiring diagram after merging wirenets;
Figure 25 shows manipulating a switch with the viewer; and
Figure 26 is illustrative of database queries from the viewer.
10
DETAILED DESCRIPTION OF THE INVENTION

Intelligent Wiring Diagram System

15 [0016] The two principal components of this invention are the Wiring Diagram Recognizer and the Intelligent Wiring
Diagram Viewer:

The Intelligent Wiring Diagram Recognizer

20 [0017] The Intelligent Wiring Diagram Recognizer converts an illustration of a wiring diagram into an intelligent wir-
ing diagram by:

a) at step 200: parse the input file and store graphical primitives in separate lists of circles, circular arcs, text ele-
ments, horizontal lines, vertical lines, oblique lines, and rectangles.
25 b) at step 300, apply pattern matching rules to the graphical primitives to recognize electrical components (circuit
breakers, fuses, connectors, etc.)
c) at step 400, apply the Attachment Point algorithm to build wire networks connecting the electrical components
d) at step 500, generate output files that provide intelligent wiring diagrams in the viewing system.

30 200: Store Graphical Primitives

[0018] Step 210 provides for the variant of the present invention in which the input data is raster. (We have to date
only implemented it for vector drawing.) There are raster-to-vector conversion software packages available commer-
cially, though none, as yet, have all the functionality required for this invention. Should such software become available,
35 that would provide step 215. Everything downstream of step 215 presumes that the information is in vector format; so
either the original information should be vector or suitable raster-to-vector conversion technology is employed at step
215.
[0019] At step 220, we iterate over all the vectors looking for polygons. Each polygon is a closed figure comprised
of three or more line segments. The recognizer reasons about how the illustrations looks, not any particular way a piece
40 of arc is composed. For example, if there is a rectangle in a diagram, it might be implemented as four line elements, or
a single rectangle element, or a single polygon element. In all cases, the art will look the same. In all of these cases we
treat the art as just a set of line segments. Thus, instead of storing the polygons in a polygon list, we store each line
separately in the appropriate line list (horizontal, vertical or oblique). The details of line storage are discussed at step
260, below.
45 [0020] In step 230, we collect all the circles and store them in a list of circles, sorted according to the center point
of the circle: if the y-coordinates of the circles are different, the one with the smaller y-coordinate comes first; if they
have the same y-coordinate, the one with the smaller x-coordinate comes first, If they have the same center, the one
with the smaller radius comes first, if they have the same center and radius, they are the same circle and we ignore the
second one. For Computer Graphics Metafiles (CGM), this results in a bottom to top, left to right order.
50 [0021] In step 240, we collect circular arcs into an arc list, using the same sort algorithm, If two arcs have the same
center and radius, they are arcs from the same circle, but they are not necessarily the same. In that case we just use
the order within the input file as the sort order.
[0022] In step 250, we store text elements in a text list using the same sort order. In vector files, text has a number
of important attributes that need to be store: the x,y position of the string, the string itself the font, the character height,
55 and the character orientation. All of these attributes must be stored internally for the recognition to succeed.
[0023] In step 260, we store line segments. The orientation (horizontal, vertical or oblique) is critical to the symbol
recognition and wire net recognition algorithms, so the recognizer uses three different lists, one for each orientation.
Again, the recognizer reasons about how the art actually looks, so if there are two different line segments that are col-

4
 EP 1 074 927 A2

linear and overlap, the recognizer merges them into a single line segment. This not only reduces complexity, but also
avoids errors in wire net recognition where lines cross.

300: Recognize Symbols

5
[0024] Graphics recognition must be tolerant of imperfections in the art due to human error or other problems. If the
endpoints of two line segments are within a few pixels of each other, they appear to the human eye to be touching and
the recognizer should consider them to be coincident even though they are actually not. Thus, in all of the recognition
algorithms, any reference to two objects touching or having the same endpoint or similar notions means that the objects
10 a close enough .
[0025] Symbol recognition consists of a sequence of individual recognizers that search through the above lists look-
ing for combinations of graphical primitives that go together to make up the desired symbol. In order to avoid ambiguity,
the recognizer looks for the more complex symbols first. For example, small terminal circles are part of a number of
symbols such as circuit breakers or fuses. However, such circles can be symbols by themselves, representing isolated
15 terminals in a circuit. Thus, if the recognizer looked for terminal circles before circuit breakers, the circuit breaker finder
would fail, since its terminals would have already been incorrectly recognizer as isolated terminals.
[0026] Because the recognizers scan the various lists of primitives repeatedly it is important (for both efficiency and
to avoid recognition errors) to eliminate from consideration as many primitives as possible that serve some other pur-
pose in the drawing. For example, sometimes circles are filled with many shading lines to make them appear gray.
20 These shade lines are not essential for symbol recognition. Those superfluous lines are discraded in step 310. In addi-
tion, small line segments that are close together and collinear comprise dashed lines. Dashed lines serve a number of
purposes in wiring diagrams such as linking off-sheet references to objects or for ganged components (see below). In
step 320, we find dashed lines. We iterate over horizontal line segments. When we find one that is short enough, we
see if there is another short one with the same y-coordinate to its right, close enough for a reasonable gap in a dashed
25 line. If we find one, we again look to its right for anpother line segment. If this results in a list of at least three line seg-
ments, we instantiate a dashed-line object and populate it with all the line segments in that list. We find vertical dashed
lines in the analogous fashion, starting at the bottom and scanning up. (Oblique dashed lines play no role in our recog-
nition algorithms.)
[0027] We illustrate symbol recognition with step 330, the circuit breaker recognizer; other recognizers are defined
30 similarly. See Error! Reference source not found. The recognizer iterates over the circular arc list looking for arcs of
approximately 60 with a reasonably sized radius and oriented either horizontally or vertically. It then looks for terminal
circles at each endpoint; that is, circles of the right size such that the endpoints in question lie on the boundary of the
circle. Circuit breakers may have a pair of small line segments representing a handle drawn at the mid-point of the arc.
If the two terminal circles are found, the recognizer has found a circuit breaker. It instantiates a circuit breaker data
35 structure which includes all the primitives (arc, circles, lines) that comprise the symbol and this object is stored in the
circuit breaker list. The primitives are maked as being used in a symbol, so that they are no longer considered for other
symbol recognition.
[0028] After all the individual symbols are recognized, the recognizer then looks for ganged symbols . These are
two or more symbols joined together by a dashed line that represent componets that are physically joined together so
40 that they operate as one unit. For exmple, a ganged circuit breaker consists of two or more breakers with a single han-
dle. When an operator pulls or pushes the handle all of the breakers are either pulled or pushed simulataneously. The
recognition of ganged circuit breakersis done in step 390. For each dashed line, the recognizer iterates over the list of
breakers looking for any that the dashed line intersects. If two or more such breakers are found, the recognizer instan-
tiates a ganged breaker with the list of the breakers that intersected the line.
45 [0029] At the end of the symbol recognition process, the system has in hand various lists of electrical components
and all of the art that comprises those components is marked as inSymbol and will hence be ignored in the next step,
Wire Net recognition.

400: Wire Net Recognition

50
[0030] Wires flow from symbol to symbol joining the symbols at predictable locations we call attachment points . For
example, wires are attached to circuit breakers at the two terminal circles. Complicating the wire net detection problem
is the fact that the line segments that represent the wires are usually interrupted by pieces of text for the wire labels and
other gaps (See Error! Reference source not found. and Error! Reference source not found.):
55 [0031] Another problem is that connectors are unpredictably sized and have no predictable attachment points;
wires can be attached at any point along the edge of a connector.
[0032] At step 410, we handle the wire label text. We iterate over all the text elements and compute a reasonable
sized bounding box. For each such box, we find all the similarly oriented lines that have an endpoint inside the box. If

5
 EP 1 074 927 A2

we find any, we instantiate a wire net with those lines and set that wire net's attachment point set to the other endpoints
of the lines.
[0033] At step 420, we handle the gap text. We iterate over the text, as in 410, and look for pairs of line segments
that would join if extended across the space defined by the text. If we find any such pairs, we instantiate another wire
5 net for each pair and put both line segments in the net and initialize the nets attachment point set to the outermost end-
points of the lines.
[0034] At step 430, we iterate over each electrical symbol, excluding connectors. Each symbol has a know set of
attachment points. For each such point we instantiate a wire net with that point as the initial attachment point. For exam-
ple, a circuit breaker has two attachment points. Thus two wire nets would be created and each would have one attach-
10 ment point initially.
[0035] At the end of step 430, then, the systems had a large number of instantiated wire nets, each with an initial
(possible empty) set of lines and a set of attachment points.
[0036] At step 440, we iterate over every wire net. For each wire net, we iterate over all the unused lines. If either
of the line's endpoints is the close enough to an attachment point, the line is added to the wire net, its other endpoint is
15 added to that wire net's attachment points and the line is marked as inSymbol . We repeat this process until we no
longer find any line to add to any wire net. At this point, if any wire net still has an empty line list, it is eliminated; this is
what happens to many of the wire nets that were instantiated in step 410. Initially, the lines we place in a wire net are
those attached to the symbols. After that, additional lines that are added represent where the wires makes a 90-degree
turn or attaches at an oblique angle; this is because we merged collinear line segments back in step 260.
20 [0037] At step 450, we merge wire nets. We do a double iteration comparing every pair of attachment points from
every pair of wire nets. If the two points are close enough, the wire nets are merged. If they are not close enough, we
then check if they are on opposite sides of some connector. In that case, the wire nets are merged as well. To merge
two wire nets, we copy the lines and attachment points of the second net into the first and eliminate the second one.
We keep performing this iteration until no more merges occur. At this point, every remaining wire net is complete.
25
500: Generate Output

[0038] Step 510 generates an Extensible Markup Language (XML) representation of the recognized intelligent
graphics (IG) objects. For each IG object, we write out an XML hierarchy of elements and attributes that completely
30 define the intelligence extracted from the graphic. This XML representation is used to generate the CGM Version 4 file
used by the IG Viewer Control.
[0039] In generating the XML file, we first iterate over the wire nets writing out the elements and attributes that are
defined for wire nets. We use the element object with the role=WIRENET and an hslist (hot spot list) element containing
a hotspot for each line segment.
35 [0040] For example, the XML representation for a wire net with two line segments might be:

40

45

[0041] Second, we iterate over all of the different electrical components writing out their markup. We use an object
with role =EC_COMP . We use the name attribute for the type of electrical component. Each component has one or
50 more states , an init_state , the state in which it is drawn. Each state is an object in its own right (role =EC_STATE ). We
use draw and erase attributes to hold actual graphical elements that should be drawn or erased when the component
enters that state. This is key to good visual effect during simulation. There is also a magtext attribute, which is a short
description of the state. The hslist for the component describes the rectangular extent of the hotspot on the component.
[0042] For example, the XML representation for a circuit breaker might be:
55

6
 EP 1 074 927 A2

10

15

20

25

30 [0043] Third, we iterate over all of the external references (i.e. textual references to other graphics).
[0044] An example external reference (to diagram 24-54-02) looks like:

35

40

45

[0045] Fourth, we iterate over the list of electrical component labels (i.e. textual labels that identify individual elec-
trical components contained on the graphic).
[0046] An example component label (for wire bundle W0513) looks like:
50

55

7
 EP 1 074 927 A2

10

[0047] Fifth, we write out XML markup for all of the text found on the graphic. This enables text search in the viewer
15 application.
[0048] An example of this is:

20

25

[0049] The resulting XML file is a concise and parse-able representation of the intelligence discovered during the
recognition step. The XML file is used to generate the CGM Version 4 file, and is also useful for Quality Assurance and
Testing purposes.
30 [0050] In step 520, we generate a Hypertext Markup Language (HTML) representation of the IG objects. The
HTML file produced is the Web Page used by Internet Explorer to render the Intelligent Graphic inside the IG Viewer
Control. The HTML file contains event-handling instructions for hotspots.
[0051] Two types of events are handled for hotspots: enter events and click events. An enter event occurs when the
user moves the mouse over a hotspot region. A click event occurs when the user presses and releases the mouse but-
35 ton while over a hotspot region. This is how the system knows when to display a new graphic or link to a different appli-
cation. In generating the HTML file, we first write out an HTML header. Next, we iterate over the hotspots generating the
code for enter events and then the code for click events. Finally, we write the HTML footer.
[0052] The resulting HTML file enables the following features:

40 • Dynamically displaying additional information about hotspots

• Linking to other graphics
• Linking to other applications (e.g. a database information system)

[0053] In step 530, we generate a CGM Version 4 file using the XML file and the original CGM file as inputs. CGM
45 Version 4 allows both the visual aspects of a graphic and the user-interface behaviors (such as automatic highlighting
and magnification) to be encoded in a single file. This CGM4 file is the only static input needed by the IG Viewer Control.
[0054] To build the CGM4 file the software performs several steps. Initially, the XML and source CGM files are
opened and parsed. The content needed to generate the CGM4 file is stored internally in memory. The CGM4 Header
is written out. Next, all text in the XML content that is identified as either an external reference or an electrical compo-
50 nent label (i.e. a hotspot), is written in color to provide the user with visual feedback as to what text is selectable (hot).
Then, viewports and hotspots for the wire nets and electrical components are added to the CGM4 file. The standard
viewports and hotspots are written. The electrical component, hotspot and text layers are written. And finally, the CGM4
footer is written out.
[0055] The resulting CGM4 file provides the IG Viewer Control with the information needed to support the following
55 features:

• Wire net highlighting

• Automatic text magnification

8
 EP 1 074 927 A2

• Full text searching

• Electrical component state changing (e.g. turning a switch on and off)

1000: Intelligent Graphics Viewer

5
[0056] The CGM4 file contains hotspot information for:

• Each wire net segment

• Each electrical component
10
[0057] A wire net is a collection of directly connected wire segments.

1010: Build Internal Structures For Cgm4 File

15 [0058] IG Viewer control calls the ISO View engine with the CGM4 file and it is read and displayed. The ISOView
engine is then queried so that the hotspot information from the CGM4 file can be retrieved and built into internal data
structures for each wire segment and each electrical component. Next, we iterate over the list of electrical components
adding component state, connection and drawing information. The state, connection and drawing information are initial-
ized for each component.
20 [0059] For each component, build its state information. For example, for a fuse, add two states, open and closed.
Each state contains a list of wire net segments that the component connects when in that state. The component infor-
mation also contains the initial component state. For example, fuses are usually drawn closed, so that would be the ini-
tial state. When in this state it connects wire nets on either side. When the fuse is open it connects no wire nets.
[0060] For each component, build its connection information. Build the wire net connection list based on the initial
25 state. For example, a closed fuse may connect two or more wire nets.
[0061] For each component, build its drawing information. Turn on or off graphic layers corresponding to the com-
ponent's current state. For example, if a fuse is open, hide the original fuse art by drawing a "white out" layer of the
closed fuse on the drawing, then turn on a color layer showing the open fuse.

30 1020: Set The Current States Of Each Electrical Component

[0062] Set all the electrical component state information based on the current state. For each wire net segment, ini-
tialize the internal data structure. Build each segment's internal data structure based on information from the CGM4 file.

35 1030: Merge Wire Nets

[0063] Cross reference all the wire nets that are connected in the current state. For instance, if wire net WN1 is con-
nected to a fuse which connects WN1 and WN2 in the current state, and WN2 is connected to a switch that connects
WN2 and WN3 in the current state, connect WN1 to both WN2 and WN3, connect WN2 to both WN1 and WN3, and
40 connect WN3 to both WN1 and WN2. WN1 and WN2 are immediately adjacent to the fuse; WN1 is remotely connected
to WN3.
[0064] Build immediately adjacent wire net connections for each component. Set each component's connections to
an empty list. Add immediately adjacent connections to each component for the current state.

45 1040: Merge Remotely Connected Wire Nets

[0065] Initialize a list of wire nets yet to be merged to this component's list of immediately adjacent connections.
Each wire net is associated with a final connections list. Initialize this list with the set of wire nets yet to be merged.
[0066] As we trace each set of connections we add new wire nets to the list of wire nets yet to be merge. We con-
50 tinue this process until the list of wire nets yet to be merged is empty. Take the next element of the list of wire nets yet
to be merged. Search all the other unchecked components to see if this wire net is on its list of immediately adjacent
connections. If the wire net is on this component's list of immediately adjacent connections, add the wire net to the final
connections list. Add each wire net in this component's list of immediately adjacent connections to the list of wire nets
yet to be merged, if it is not already there. Remove the current wire net from the list of wire nets yet to be merged. Add
55 the final connections list to this component.

9
 EP 1 074 927 A2

1060: Link Wire Net Hotspot Segments With Same Element Identifier

[0067] Connect and cross-reference all wire net segments in the same wire net. Each segment in the same wire
net contains a related identifier produced by the recognizer and file builder. Initialize each wire net segment's internal
5 link structure. Search through the list of other wire net segments. If another wire net segment has the same identifier
add each one to the other's wire net segment cross-reference links.

1080: Pass Control To User Event Loop

10 [0068] We are through building wire net structures for the current state of electrical components. Pass control to the
user event loop.

1100: User Events

15 [0069] The viewer is an event-driven application which responds to various events, typically an end-user action.

1110: User Mouse Action: Move Over Wire Net Segment

[0070] The mouse passes into a wire net segment. ISOView raises a mouse event containing identifiers that enable
20 the viewer to find the associated wire net segment in the internal data structures. Highlight all the segments in the wire
net associated with this segment. Highlight all the segments in all the other wire nets found on adjacent components'
final connection lists. Display any magnified text associated with this segment's wire net.

1120: User Mouse Action: Move Outside Of Wire Net Segment

25
[0071] The mouse passes out of a wire net segment. ISOView raises a mouse event containing identifiers that ena-
ble the viewer to find the associated wire net segment in the internal data structures.
[0072] Remove highlighting from all the segments in the wire net associated with this segment. Remove highlight-
ing from all the segments in all the other wire nets found on adjacent components' final connection lists. Hide any mag-
30 nified text associated with this segment's wire net. Wire nets may be marked as selected if a user clicks the mouse
while over the wire net. Each selected wire net is displayed in a different color. Selecting a wire net makes it stand out
from other nearby wire nets that are not selected. Highlighting a wire net temporarily overrides the selection color, if this
wire net is marked as selected then re-display it with its selection color.

35 1130: User Mouse Action: Click On Wire Net Segment

[0073] Wire nets may be marked as selected if a user clicks the mouse while over the wire net. Each selected wire
net is displayed in a different color. Selecting a wire net makes it stand out from other nearby wire nets that are not
selected. Clicking on a wire net segment toggles the selection state for the segment's wire net and for any remotely con-
40 nected wire nets as well.
[0074] If this segment's wire net is already in a selected state, turn off the selection color. We do this by turning off
the selection state for all the segments in this segment's wire net and for any segments in remotely connected wire nets
as well. Re-display the wire net segments without the selection color. Remember that all these segments are not cur-
rently selected (so the next time, the selection color will get turned on).
45 [0075] If this segment's wire net is not already in a selected state, turn on the selection color. We do this by turning
on the selection state for all the segments in this segment's wire net and for any segments in remotely connected wire
nets as well. Generate a selection color that is not currently being used. Re-display the wire net segments with the
selection color. Remember that all these segments are currently selected.

50 1160: User Menu Command: Select All Wire Nets

[0076] Sometimes the user would like to show every wire net in a different color. This could help trace circuits in the
wiring diagram or help developers verify the behaviors of the recognizer, file build, and viewer.
[0077] For each wire net in the internal list follow the procedure for turning on the selection color (step 1143).
55 Remember that all these segments are currently selected. Generate a selection color that is not currently being used.
Highlight all the segments in this wire net and any wire nets connected in the current state.

10
 EP 1 074 927 A2

1170: User Menu Command: Reset Selected Wire Nets

[0078] Turn off (de-select) any selected wire nets. Follow the procedure for turning off the selection color for each
selected wire net (step 1132).
5
1180: User Menu Command: Select An Electrical Component

[0079] The user may change the state of an electrical component by clicking on it with the mouse. Each component
has a state list. For example, a fuse has a list of two states, closed and open. Some components have more than two
10 states. For example, a rotary switch may have three or more states.
[0080] Select the next state in the component's state list. If we have reached the end of the list, start again at the
beginning. Build the connection list based on the new state (step 1014). Draw this component based on the new state
(step 1015). Re-merge the wire nets based on the new state (step 1030).

15 1190: User Menu Command: Exit

[0081] Exit the viewer.

SPECIFIC EXAMPLE OF SYSTEM FUNCTION

20
INVENTION DESCRIPTION

[0082] The Intelligent Wiring Diagram System automatically and quickly converts wiring diagrams into electronically
useful, interactive graphics. This novel approach completely eliminates the need for manual re-authoring. Within a few
25 seconds the software is able to accurately identify circuit continuity and nomenclature of interest and put hundreds of
hotspots on a diagram. If an expert were asked to perform the same work, it would take could easily take many days of
intensive labor to re-author and validate a single wiring diagram.
[0083] The invention consists of a set of software modules that perform the conversion and a graphics viewer that
provides the interaction. These modules are:
30
A. ELECTRICAL COMPONENT RECOGNIZER

[0084] The Electrical Component Recognizer searches through the graphical primitives in the original wiring dia-
gram and applies heuristics in order to identify components such as fuses, circuit breakers, various types of switches,
35 electrical connectors and relays.
[0085] Example (Figure 20): To identify a circuit breaker, the module first looks for a circular arc whose physical
characteristics satisfy a pre-defined set of pattern rules. For each such arc, it looks for terminal circles and an optional
handle, each of whose physical characteristics also must satisfy a pre-defined set of pattern rules. Moreover, identifica-
tion of these graphic primitives as electrical components enables circuit simulation by the viewer when coupled with the
40 established behaviors of the components.

B. WIRENET RECOGNIZER

[0086] Using the electrical symbols found by the Electrical Component Recognizer, the Wirenet Recognizer
45 employs a unique and powerful Attachment Point Algorithm to discover which lines in the diagram represent individual
wires and to infer the electrical continuity. The algorithm works as follows:
[0087] For each electrical component there are predictable locations where wires can attach; these are the attach-
ment points . For example, in a circuit breaker wires attach at the terminal circles. Similarly, there are attachment points
for fuses, resistors, and the other electrical components, except connectors which are handled separately . Also, it is
50 common for designers to erase part of a wire to make space for a text string such as the wire number or a notation.
Consequently, for wirenet generation purposes, we consider text elements to have their own attachment points where
wires may be attached.
[0088] The Attachment Point Algorithm consists of three major steps:

55 B1 SPROUTING WIRENETS

[0089] For each attachment point, we search for line segments which end at the point. These are the sprouts for
our wirenets. After sprouting, therefore, each symbol now has a list of wirenets attached to it. Each wirenet consists of

11
 EP 1 074 927 A2

a list of line segments and a list of attachment points, which are the set of points where additional line segments can
join that particular wire net.
[0090] Example (Figure 21): Consider a circuit breaker CB1 with terminal circles TC1 and TC2. The sprouting oper-
ation will cause the creation of two wirenets, WN1 and WN2 with attachment points (p1} and {p2} where p1 is the known
5 point on circle TC1 where a wire may be attached and similarly for p2 on TC2.
[0091] Sprouting finds line L1 that has p1 as an endpoint and lines L2 and L3 which both have p2 as an endpoint.
This results in wirenets WN1 and WN2:

CB1: wirenets: {WN1, WN2}

10 WN1: lines: {L1}
attachment points: {p1, p3}
WN2: lines: {L2, L3}
attachment points: {p2, p4, p5)
where p3 is the other endpoint of L1 and p4 and p5 are the other endpoints of L2 and L3 respectively.
15
[0092] Figure 22 depicts part of a wiring diagram after all the wirenets have sprouted:

B2 GROWING WIRENETS

20 [0093] After all the sprouting is done, the wirenet recognizer iterates over all the wirenets looking for any new lines
that end at any of the attachment points. When a line is added to a wire net, its other endpoint is added to the attach-
ment point list for that wirenet. (Once the Wirenet Recognizer knows that no more lines end at a particular attachment
point, that attachment point can be discarded to avoid needlessly reconsidering it). This process of growing wirenets
continues until no more line segments can be found to add to any wirenet.
25 [0094] Figure 23 depicts the state of the diagram after the growing process is complete. At this point, the algorithm
has found every wire segment, but the wirenets are still fragmented.

B3 MERGING WIRENETS

30 [0095] The next step is to merge wirenets. Each pair of wirenets is compared. If they share an attachment point,
they are merged; that is all of the lines and attachment points of the second are copied to the first and the second dis-
carded.

SPECIAL HANDLING OF CONNECTORS

35
[0096] In addition, two wirenets are merged if they each have an attachment point on opposite sides of the same
connector directly opposite each other, derived by identical x or y coordinates. This is how continuity through a connec-
tor is achieved.
[0097] As each merge occurs, 2 wirenets become 1 wirenet. At the end of this step, we have a greatly reduced set
40 of wirenets each of which captures a full continuity. Figure 24 shows the final results of the Attachment Point Algorithm.

B. NOMENCLATURE RECOGNIZER

[0098] Using text matching techniques, the Nomenclature Recognizer scans the text elements in the diagram and
45 finds those that match wire numbers, wire bundles, grounds, splices, terminals and other equipment numbers from
databases.

XML GENERATOR

50 [0099] Extensible Markup Language (XML) provides a standardized methodology for describing information in a
document. We have created a unique Document Type Definition (DTD) for describing graphical objects and the relation-
ships between them, including specific elements for describing electrical components and electrical continuity, as well
as references to part information. The XML Generator produces the XML markup for all of the objects discovered by the
three recognizers (electrical components, wirenets and nomenclature). The markup includes the geometry for each
55 object's hotspot and the relationships between objects and other objects and between objects and hotspots.
[0100] For example, consider a switch to which the system has assigned the unique identifier SWITCH1 . In order
to make the switch interactive, it is linked to an HSLIST object which defines the rectangular hotspot. The switch also
has two states , UP and DOWN, each of which is captured in an explicit object which the systems names S1-UP and

12
 EP 1 074 927 A2

S1-DOWN , respectively. The init-state attribute informs the viewer which of those states is the one depicted in the illus-
tration as drawn by the illustrator. Each state has four important attributes:

connects : the value of this attribute is the list of wirenets that are connected when the component is in this state
5 draw: the value of this attribute is a set of drawing instructions that tell the viewer what art to draw when the com-
ponent is in this state
erase: the value of this attribute is a set of drawing instructions that tell the viewer what art to erase when the com-
ponent is in this state
magtext: the value of this attribute is a string the viewer should display when the user points at the object and it is
10 in this state.

[0101] There are three wirenets conected to the switch, WN1 , WN2 and WN3 . When the switch is up it connects
WN1 to WN2; when it is down it connects WN1 to WN3. Each wirenet has its own HSLIST describing the x-y coordi-
nates of all the line segments that comprise that wirenet.
15 [0102] The SGML representation for this switch might look like:

The switch itself:

20

25

30

35

The three wirenets connected to the switch

40

45

50

INTELLIGENT GRAPHICS VIEWER

[0103] Finally, we have invented a unique graphics viewer which not only renders the Wiring Diagram in its original
55 form, but also uses the output of the XML generator to provide visual identification of objects and user-driven interac-
tivity with the diagram. Every wirenet is "hot"; when the user points at the wire with a pointer device (e.g. a mouse), the
wire and its complete continuity immediately highlight. Furthermore, the user can affect that continuity by interactively
changing the state of electrical components. The viewer also allows fill text searching and highlighting of hits found on

13
 EP 1 074 927 A2

the diagram.
[0104] Example (Figure 25): Suppose the diagram depicts a switch with wirenet W1 attached on the right at the
switch pivot point and wirenets W2 and W3 attached on the left at the upper and lower switch terminals; and suppose
that the diagram shows the switch in the UP position. Then when the user points at W1, all of W1 and W2 will highlight,
5 since the switch connects them. In addition, the user can "click" the switch to change its state to the DOWN position.
When the user does this, the viewer will automatically redraw the switch in the down position and change the continuity,
so that pointing at W1 will now highlight W1 and W3, but not W2. By change the states of various components, the user
can easily see how electrical continuity is affected.
[0105] In addition, using the hotspots identified by the Nomenclature Recognizer, the viewer supports database
10 queries. By clicking a wire number or connector number, the user can quickly access part information by hyperlinking
to other documents or other wiring diagrams. See Figure 26.
[0106] Alternative Embodiment: If the wiring diagrams are originally authored using a Computer-Aided Design
(CAD) tool then much of the intelligence may already be captured in the original CAD dataset. In that case the Symbol
Recognizer and Wirenet Recognizer would be replaced by a CAD Dataset Interpreter which would extract symbol and
15 wirenet information directly from the CAD dataset. Note, however, that this embodiment has significantly less generality
because, for much of the legacy wiring diagrams, reliable CAD datasets are unavailable.

A PRACTICAL APPLICATION OF GRAPHICS RECOGNITION AND DELIVERY TECHNOLOGIES TO SUPPORT

HIGHLY INTERACTIVE WIRING DIAGRAM INFORMATION TO TECHNICIANS TO SIGNIFICANTLY IMPROVE AIR-
20 PLANE MAINTENANCE

[0107] Consider this scenario: Suppose there is a problem reported with the N265 airspeed indicator. First, we
must consult the diagram equipment list to find out what diagram N265 is on. After scanning through microfiche cards,
we print the wiring diagram and take it to the airplane. We have to manually scan the printout looking for N265. Tests
25 reveal that there is no power at pin 22 of the indicator. Using a highlighter to trace the circuitry in the diagram, we next
test a transformer, then a circuit breaker, still finding no power at specific points in the circuit. At this point, we must fol-
low an off-sheet reference to another diagram (meaning another visit to the microfiche) and continue the trouble shoot-
ing. Eventually, we discover that the C1315 circuit breaker is faulty and we need to order a spare. This means another
trip back to the maintenance shack to look up the circuit breaker in the equipment list to obtain a part number and ven-
30 dor information.
[0108] Our goal is to build an integrated wiring information system that solves all of the difficulties inherent in this
scenario. We enter the problem component, N265, into the system, which responds with part information including a
hyperlink to the correct diagram. When we click on that link, the system automatically finds N265 on the diagram and
zooms in on the area of interest in the diagram. Again we test the component and find the problem at pin 22. We only
35 need to click on the wire at that pin in the diagram and the full continuity is highlighted. Doing the same test, we work
our way to the off-sheet reference, click on it and the system fetched the next diagram and again automatically zooms
in on the area of interest. We find the problem circuit breaker and click on it to obtain the needed part number and ven-
dor information.
[0109] In order to deliver such a system on a large scale, we have developed graphic recognition technology that
40 can automatically convert wiring diagrams into these interactive objects.

RECOGNITION

[0110] There are two major recognition challenges in this work: recognizing electrical symbols, such as fuses, cir-
45 cuit breakers and connectors, and recognizing the networks of wires that flow between them. This work involves vector
wiring diagrams, not raster, so our recognition algorithms involve reasoning about the relationships between graphical
primitives (lines, circles, circular arcs, and text elements) rather than reasoning at the pixel level. In order to reliably
process thousands of legacy images, created by different authors over many years, the algorithms must be tolerant of
slight inaccuracies and not depend on specific choices of primitives when more than one implementation exists. For
50 example:

a) We classify a line as horizontal if it's slope is less than .05 and vertical if the reciprocal of it's slope is less than
.05.
b) When comparing coordinates, two x(or y) values are considered equal if their difference is small enough . The
55 exact tolerance depends on the scale of the diagram's coordinate system.
c) We treat a polygon as if it were simply a set of line segments no different than any of the other line segments in
the file. Thus if a recognition step needs to find a rectangle, it always looks for four line segments that satisfy the
appropriate constraints. This avoids problems where one illustrator used a rectangle element and another actually

14
 EP 1 074 927 A2

drew four lines.

d) if two line segments are collinear and touch or overlap, we merge them into a single, longer line segment,
because to the human eye there does appear to be a single line. This is especially important when we are recog-
nizing wire nets, as it obviates problems were two wires cross orthogonally.
5
SYMBOL RECOGNITION

[0111] Wiring diagrams use a variety of standard symbols to represent electrical components such as circuit break-
ers, resistors, fuses and connectors. For each such symbol, we have developed a feature-based recognizer that
10 searches for patterns in the graphical primitives that characterize that symbol.
[0112] A simple example of a symbol recognizer is that for a circuit breaker:

15

20

25
FIGURE 27

[0113] Other symbol recognizers work similarly, though they can vary considerably in complexity. Connectors
present unique problems because they can have arbitrary length and be complete or partial (Figure 28), but they still
30 obey a set of predictable constrains in terms of the graphical primitives which comprise them.
[0114] The symbol recognizers are parameterized by a number of tolerance values (such as the radius of a terminal
circle or the angle span of a circular arc). By experimentally refining those parameters, we are able to tune the recog-
nizers to achieve over 99% accuracy on full manual sets.
[0115] In addition to the basic symbology, we recognize ganged symbols, i.e. two or more symbols joined with a
35 dashed line which represents physically connected components which move together (Figure 29). In order to do this we
must first perform dashed-line recognition. Again this is a matter of defining constrains that define what it means to be
a dashed line and iterating over line segments looking for those that are small enough, collinear and close enough
together to satisfy those constraints.

40 WIRE NET RECOGNITION

[0116] Having found the electrical symbols in the diagram, the next step is to determine which of the remaining line
segments comprise the various wires and to infer the wiring continuity , that is how the wires connect various compo-
nents together. We have devise an attachment point algorithm to perform this task. This algorithm is based on the fol-
45 lowing principles:

a) Wires connect to symbols at predicable point on the symbols; for example, wires connect to a circuit breaker at
one of the two terminal circles.
b) wires change direction (from vertical to horizontal or vice versa) only at 'L' shaped intersections, not at 'T' or '+'
50 shaped intersections.
c) The lines for a wire are often broken by test labels or other text were the amount of white space can vary consid-
erably.
d) Two wires that are collinear and terminate on opposite sides of a connector represent the same continuity.

55 [0117] We represent wire nets with two lists:

a) a list of line segments that comprise the net and

b) a list of attachment points.

15
 EP 1 074 927 A2

[0118] Recognition proceeds in four phases:

a) Seeding: we iterate over all electrical symbols (excluding connectors). For each one we instantiate one or more
wire nets. Each such wire net will initially have an empty line list and an attachment point list with one or more of
5 the predictable points. A circuit breaker, for example, will generate two wire nets, one for each of the terminal cir-
cles. On the other hand, a splice will have two attachment points, but only one wire net will be instantiated and both
of the attachment points will belong to that wire net, because a splice does not have multiple states.
We also iterate over the text elements and look for line segments that intrude into an appropriately sized
bounding box. If we find any, we instantiate an additional wire net. In this case, the net's line list is initialized with
10 those lines and the attachment point list is initialized with the other endpoints of the lines.
b) Propagation: We iterate over each attachment point in each wire net looking for new line segments that end at
that point. If we find such a line, we add it to the net's line list and add its other endpoint to the attachment point list.
c) Merging: Finally we iterate over every pair of wire nets to see if they should be merged. Merging means setting
the line list of the first to the union of the two separate line lists and similarly for the attachment point list. There are
15 two conditions that indicate a merge:

a) if the nets share an attachment point or

b) if there are attachment points p1 in net1 and p2 in net2 and a connector C such that p1 and p2 are directly
opposite each other. In that way, we achieve wire net continuity through connectors.
20 d) Text Analysis: We use layout logic to group text elements into paragraphs. In addition we use text-pattern
matching rules to identify references and equipment nomenclature.

XML REPRESENTATION

25 [0119] The output of the recognizer is an Extensible Markup Language (XML) representation of the various objects.
[0120] For example, the XML representation for a wire net with two line segments might be:

30

35
[0121] The XML representation for a circuit breaker might be:

40

45

[0122] Note that the circuit breaker has two states, open and closed and that when it is closed it connects two wire
50 net objects. When it is open, it connects nothing. Also, we can encode explicit draw/erase instruction in a state, so that
when the object enters that state the viewer can indicate the new state by changing what the user sees.
[0123] Similarly, we generate XML markup for references, paragraphs and equipment nomenclature. One impor-
tant benefit of this aspect is that we are able to generate discrepancy reports by comparing the text that is actually in
the diagram with wiring databases that specify which equipment should be depicted. This quality assurance tool is
55 resulting in significantly better quality in the data The Boeing Company delivers.
[0124] A separate application parses the XML representation and populates the original CGM file with application
program structures that make the wiring diagram work interactively in our viewer.

16
 EP 1 074 927 A2

INTELLIGENT GRAPHICS VIEWER

[0125] Our intelligent graphics viewer is based on the IsoView ActiveX control from Itedo Corporation. We have
developed our own control, IGView, which contains Iso View. Iso View renders the CGM files, highlights hotspots and
5 exposes events to IGView. IGView responds to those events to provide powerful trouble shooting tools to the user.
When the user points at a wire with the mouse cursor, the continuity for that wire is highlighted. IGView computes that
continuity dynamically based on the states of all the electrical components (Figure 25). The user changes the state of
a component by clicking on its hotspot. For example, by clicking a switch, the user can toggle its state between down
and up. I response, IGView will use the draw/erase instruction to make the switch visually move back and forth between
10 the two positions. In addition, the continuity changes in accordance with the switch position. In the same manner, the
user can simulate pulling a circuit breaker or blowing a fuse.
[0126] Of course, the viewer supports hyperlinking to other diagrams from reference text. In addition, because we
recognize equipment nomenclature, the user can click on an equipment number to dynamically query databases
regarding spare parts, vendors, and other critical information. In addition, because we catalog every paragraph, the
15 viewer supports full text search and will automatically zoom in on the area of interest when the user links to a diagram
from the database screen.

CONCLUSION

20 [0127] We have described a recognition algorithm for vector wiring diagrams that recognizes electrical symbols and
induces wiring continuity. Because we are dealing with vector data and because there is considerable consistency in
the production of these diagrams, we are able to achieve very high recognition accuracy.
[0128] We have also described a practical application of the results of this recognition, an intelligent wiring informa-
tion system which provides enhanced trouble shooting capabilities: interactive wiring continuity display, component sim-
25 ulation, full text search, hyperlinking and database query functionality. This system also provides and excellent training
tool.
[0129] We have hereinabove described the nature of electrical troubleshooting, how wiring diagrams are used, and
the problems inherent in simply displaying them online without a layer of intelligence. We then discussed our solution
to this problem, the Wiring Information Navigation System. This entails the use of symbol recognition technology for
30 finding electrical components in vector illustrations, our attachment point algorithm for finding the layout of wires in the
diagrams, the encoding of recognition results within the illustration and our viewing system which allows users to inter-
actively model circuit continuity and query information systems.

Claims
35
1. A method for converting legacy wiring diagrams into intelligent wiring diagrams comprising:

utilizing an intelligent graphics viewer for comprehending complex electrical circuitry;

utilizing said intelligent graphics viewer to simulate changes in electrical continuity; and further
40 utilizing said intelligent graphics viewer to access critical component information.

2. An intelligent wiring diagram system comprising in combination:

an intelligent component recognizer for searching through graphical primitives in a wiring diagram and applying
45 heuristics in order to identify components;
a wirenet recognizer for determining:
which lines in the diagram represent individual wires and to infer the electrical continuity;
a nomenclature recognizer:
Extensible markup language (XML) generator for producing the XML markup of all objects discovered by a plu-
50 rality of recognizers, said XML markup including the geometry for each object hotspot and the relationships
between objects and between objects and hotspots, and,
an output device comprising and intelligent graphics viewer for providing the wiring diagram in its original form,
said intelligent graphics viewer further utilizing the output of the XML generator for providing visual identifica-
tion of objects and user-driven interactively with said wiring diagrams.
55
3. In combination:

an XML generator;

17
 EP 1 074 927 A2

an intelligent graphics viewer for illustrating a wiring diagram in original format

An XML representation for wirenets, electrical components and their states; and,
said intelligent graphics viewer further providing visual identification of objects and user-driven interactivity with
said wiring diagrams.
5
4. In combination:

an electrical component recognizer for searching through the graphical primitives in a wiring diagram and
applying heuristics in order to identify components; and,
10 a wirenet recognizer for utilizing the electrical symbols provided by said electrical component recognizer and
determining which lines in the wiring diagram represent individual wires conferring electrical continuity.

15

20

25

30

35

40

45

50

55

18
EP 1 074 927 A2

19
EP 1 074 927 A2

20
EP 1 074 927 A2

21
EP 1 074 927 A2

22
EP 1 074 927 A2

23
EP 1 074 927 A2

24
EP 1 074 927 A2

25
EP 1 074 927 A2

26
EP 1 074 927 A2

27
EP 1 074 927 A2

28
EP 1 074 927 A2

29
EP 1 074 927 A2

30
EP 1 074 927 A2

31
EP 1 074 927 A2

32
EP 1 074 927 A2

33
EP 1 074 927 A2

34
EP 1 074 927 A2

35
EP 1 074 927 A2

36
EP 1 074 927 A2

37
EP 1 074 927 A2

38
EP 1 074 927 A2

39