SIMATIC
Vision Sensor VS 110
Tips and Tricks
Edition 10
/
2002
Copyright © Siemens AG 2002 All ri ghts reserved
The reproduction, transmission or use of this document or its
contents is not permitted without express written authority.
Offenders will be liable for damages. All rights, including rights
created by patent grant or registration of a utility model or design,
are reserved.
Siemens AG
Bereich Automation and Drives
Geschaeftsgebiet Industrial Automation Systems
Postfach 48 48, D- 90 327 Nuern be rg
Disclaimer of Liability
We have checked the contents of this manual for agreement with
the hardware and software described. Since deviations cannot be
precluded entirely, we cannot guarantee full agreement. However,
the data in this manual are reviewed regularly and any necessary
corrections included in subsequent editions. Suggestions for
improvement are wel com e d.
©Siemens AG 2002
Technical data subject to change.
Siemens Aktiengesellschaft A5E00188390-01
Safety Guidelines
This manual contains notices intended to ensure pers onal safet y, as well as to prot ect the products and
connect ed equipment against damage. These notices are highlighted by t he symbols s hown below and
graded according to severity by the following texts:
!Danger
indicates that death, severe personal injury or substantial property damage will result if proper
precautions are not t aken.
!Warning
indicat es that death, s evere personal injury or substantial property dam age can result if proper
precautions are not t aken.
!Caution
indicat es that minor personal i nj ury can result if proper prec autions are not taken.
Caution
indicat es that property damage can res ult if proper precaut i ons are not taken.
Notice
draws your atte ntion to part icul a rl y import ant information on the product, handling the product, or t o a
particul ar part of the docum entat i on.
Qualified Personnel
Only qualified personnel should be allowed to install and work on this equipment. Qualified persons are
defined as persons who are authorized to commission, to ground and to tag circuits, equipment, and
syst ems in accordance with established safety practices and standards.
Correct Usage
Note the following:
!Warning
This device and its components may onl y be used for the applicat i ons described in the catalog or the
technical description, and only in connection with devic es or components f rom other manuf act urers
which have been approved or recommended by Siem ens.
This product can only funct i on correctly and safely if it is transported, st ored, s et up, and install ed
correct l y, and operated and maintained as recomm ended.
Trademarks
SIMATIC®, SI MATIC HMI® and SIMATIC NET® are registered trademarks of SIEMENS AG.
Third parties using for thei r own purposes any other names in this doc ument which refer to t radem arks might
infringe upon the rights of the trademark owners.
Vision Sensor SIMATIC VS 110 Tips and Tricks
A5E00188390-01 iii
Contents
1 Introduction 1-1
2 Optimum Settings for the SIMATIC VS 110 2-1
2.1 Optimizing the Image Window ..........................................................................2-1
2.2 Aligning the Sensor Field of View.....................................................................2-2
2.3 Avoiding Reflections..........................................................................................2-3
2.4 Aligning the Conveyor.......................................................................................2-5
2.5 Selecting Triggers and Using Them Effectively................................................2-7
2.5.1 Automatic Trigger..............................................................................................2-7
2.5.2 External Trigger.................................................................................................2-8
2.6 Obtaining a Useful Quality Limit........................................................................2-9
2.7 Setting the Y-Limit Quality Limit........................................................................2-9
2.8 Compensating Height Variations ....................................................................2-11
2.9 Using the "Info" Menu Command....................................................................2-11
2.10 Counting Objects.............................................................................................2-12
3 Mastering Borderline Applications for the SIMATIC VS 110 3-1
3.1 Parts Larger than Permitted by the Specification..............................................3-1
3.2 Deliberate Use of an Inclined Conveyor ...........................................................3-3
3.3 Parts Smaller than Permitted by the Specification............................................3-4
3.4 The Individual Test Objects of a Model are too Similar ....................................3-5
3.5 The Individual Test Objects of a Model are too Different..................................3-8
3.6 Round Objects with Asymmetrical Characteristics ...........................................3-9
3.7 Testing Endless Material...................................................................................3-9
3.8 The Maximum Number of 15 Models in Total Is not Enough..........................3-11
3.9 Precise Settin g of a "Dimension Check".........................................................3-11
Contents
Vision Sensor SIMATIC VS 110 Tips and Tricks
iv A5E00188390-01
Vision Sensor SIMATIC VS 110 Tips and Tricks
A5E00188390-01 1-1
1 Introduction
This document is intended for the user of the Vision Sensor SIMATIC VS 110. It
contains notes on installation and tips for typical applications. Some solutions for
unusual applications are also included. The aim of the document is to bring y ou a
step nearer to simple and successful solutions.
We start by outlining some of the main points to be considered when setting up the
SIMATIC VS 110. You will see several sample images of the sensor field of view
that illustrate points you should note.
The next chapter describes applications that violate the specifications
for operating the Vision Sensor VS 110 but that can nevertheless be used for a
reliable and successful evaluation.
Introduction
Vision Sensor SIMATIC VS 110 Tips and Tricks
1-2 A5E00188390-01
Vision Sensor SIMATIC VS 110 Tips and Tricks
A5E00188390-01 2-1
2 Optimum Settings for the SIMATIC VS 110
2.1 Optimizing the Image Windo w
The following picture shows the sensor field of view of a correctly aligned sensor
head:
The following important conditions are met:
• The object being tested is located correctly in the middle of the viewing window
and does not touch the edge at any point. Only the lower edge of the object is
in contact with a black surface caused by the upper edge of the conveyor.
• There are no reflections from the conveyor or the object.
• The conveyor is aligned parallel to the lower edge of the image
• The object is sharp.
• The dimensions of the object are within the SIMATIC VS 110 specifications.
• The sensor head has been correctly adjusted using the "Settings" menu
command.
If your application meets these conditions, correct and reliable evaluation with the
Vision Sens or SI MATIC VS 110 is guaranteed.
The next sections contain further pictures of the sensor field of view that may lead
to problems in evaluation. Approaches that will avoid the problems are also
described.
Optimum Settings for the SIMATIC VS 110
Vision Sensor SIMATIC VS 110 Tips and Tricks
2-2 A5E00188390-01
2.2 Ali g ning the Sensor Field of View
In the picture shown below, the sensor head is not correctly aligned: The object
being tested in the sensor field of view (in this example a screw) is touching the
upper edge of the illuminated and therefore visible evaluation area. If you
attempted to train with this part, the error message "Error Object too far at top"
would appear.
Note
The object being tested must maintain at least the following minimum distance
from the top, left, and right edges of the field of view (in other words the visible
bright area) to allow reliable evaluation:
• 4 mm with the camera head for large objects (MLFB 6GF2 002-8AA)
• 2 mm for the camera head for small objects (MLFB 6GF2 002-8BA)
The object being tested must only touch the lower edge of the image in the sensor
field of view or a surface that is visible at the bottom of the image.
Optimum Settings for the SIMATIC VS 110
Vision Sensor SIMATIC VS 110 Tips and Tricks
A5E00188390-01 2-3
2.3 Avoiding Reflections
In the following picture, you can see reflections of the illumination unit on the
surface of the conveyor (light areas on the dark surface resulting from an oblique
view of the conveyor "from above").
These reflections may have a negative influence on the evaluation since the
SIMATIC VS 110 cannot distinguish clearly between the visible illuminated area
and fixed components in your system.
Not only reflections from fixed components (for example the conveyor system)
cause problems but also temporary reflections on the test objects themselves.
These can result in different silhouettes and therefore lead to different test results.
A temporary reflection is shown in the picture below.
Optimum Settings for the SIMATIC VS 110
Vision Sensor SIMATIC VS 110 Tips and Tricks
2-4 A5E00188390-01
One way of avoiding reflections is to mask the unused illuminated area responsible
for the reflections.
• Select the sensor field of view so that there are as few reflections as possible
(bright patches) on the test object and the conveyor. There are threaded holes
on the front of the illumination unit that can be used to fit opaque masks to the
illumination unit.
• As a rule of thumb for masking; you should attempt to leave only the minimum
distance between the upper edge of the test object and the lower edge of the
mask so that a maximum cover is achieved. You should also make sure that
you use the highest object when determining the necessary minimum distance
between the object and mask!
• You can achieve the same effect by arranging the illumination behind the
object so low that the upper edge of the illuminated area effectively takes over
the function of a mask.
The following picture shows an ideal image of a test object (on the left) when using
a mask. On the right-hand side, you can see the negative effects of a temporary
reflection: Compared with the ideal image of the test object, the test object appears
to be too flat in the middle.
Optimum Settings for the SIMATIC VS 110
Vision Sensor SIMATIC VS 110 Tips and Tricks
A5E00188390-01 2-5
Reflections can also be reduced significantly if the test object is in the upper part of
the possible sensor field of view. You can achieve this by arranging the mid axis
(optical axis) of the sensor head somewhat below the upper edge of the conveyor.
With this solution, however, you must make sure that the object being tested
remains completely visible in the image. This arrangement is only suitable for
relatively flat objects.
If you mask the illumination, you must make sure that the mask completely blocks
the infrared light being used. A translucent mask as shown in the following picture
may have very detrimental effects on reliable evaluation.
Note
For more infor mat ion on the topic of arrang ing the il lu min ation and usin g mas ks,
refer to the Vision Sensor SIMATIC VS 110 manual, page 4-5.
2.4 Ali gning the Conve yor
When training the background (in other words the fixed components of your
system), the device records the area in which the objects being tested are
expected for future evaluation. When training the background, a distinction is made
between fixed and moving objects. On completion of the background training, the
SIMATIC VS 110 calculates the largest "free" rectangle in the sensor field of view
(in other words the rectangle in which no fixed objects were located during the
background training). The sides of this rectangular evaluation area are aligned
parallel to the edges of the image window.
The size of the evaluation area obtained during training can be checked in the run
mode using the INFO menu command. Based on the length and height in
millimeters, it is possible to gauge whether the background training achieved a
useful result.
Optimum Settings for the SIMATIC VS 110
Vision Sensor SIMATIC VS 110 Tips and Tricks
2-6 A5E00188390-01
It is therefore important that the upper edge of the conveyor is aligned parallel to
the lower edge of the image window to achieve reliable results with the SIMATIC
VS 110. If this condition is not met as shown in the picture below (on the left), the
device will evaluate in an invalid or at least unnecessarily reduced area as shown
in the picture below on the right.
A
rea cannot be e va l uated
and will be ignored!
Evaluation area
Conveyor
Tool Holder Transport or Grab Unit
Conveyors of all shapes and sizes without a horizontal top edge, such as tool
holders or grab units may lead to a reduction of the area available for evaluation.
The area that is cut off in the picture below is, for example, not evaluated.
Evaluation area
All the areas that are not within the trained evaluation window are treated by the
device as "don't care" regions during evaluation and they are ignored. If, due to
unfavorable mounting of the camera or "indented" arrangements on a tool holder,
this area is too large, the quality and reliability of the evaluation will suffer.
Optimum Settings for the SIMATIC VS 110
Vision Sensor SIMATIC VS 110 Tips and Tricks
A5E00188390-01 2-7
Note
Particularly when using grab units, you should make sure that only the relevant
evaluation areas of the object exist in the field of view and not parts of the grab
unit. If this proves to be impossible due to the mechanical properties of the system,
the grab unit should be moved when training the background to avoid creating
additional "don't care" areas for the evaluation. When training the good parts, you
must, however, then make sure that all the components of the grab unit visible in
the picture and in contact with the actual test object are also evaluated and
therefore directly influence the quality of the result!
"Fine Tuning" of the Qualtiy Limit
The evaluation of the test object; in other words the comparison of the test object
with the trained good part is based on the match between the expected/reference
surface of the object and the surface of the object actually detected. A percentage
is used to determine the quality ("Q-Limit").
Due to this relationship, the qualtiy limit can be set by "Fine Tuning" (in other words
in extremely small steps of absolute surface deviation) if the object visible in the
picture has a small surface (within the permitted specifications).
This does, however, only relate to the setting of the qualtiy limit. This has no effect
whatsoever on the basic evaluation accuracy of the device related to the resolution
of the sensor chip and the algorithmic picture preprocessing. For this reason, you
should always attempt to display the characteristics of interest of the object being
tested as large as possible and the irrelevant characteristics should be visible as
little as possib le in the pic t ure.
2.5 Selecting Tri ggers and Using Them Effectivel y
2.5.1 Automatic Trigger
If you use the automatic trigger (the device permanently records pictures and
decides automatically whether and when a new part is present for evaluation), the
parts in the field of view of the sensor must only move horizontally from left to right
or right to left. The conveyor speed must be constant so that the objects can be
manipulated after the test at the right position (for example at an ejector).
If the speed changes during operation, the objects will be selected and
manipulated either too early or too late by the SIMATIC VS 110. Your system is no
longer functioning correctly.
While the objects are moving through the sensor field of view one of the two
silhouettes is visible depending on the position of the object in the image field. With
test objects that extend a long way in the z direction this effect can lead to incorrect
test results in the automatic trigger mode since the visible surface changes
depending on the perspective. In this case, an external trigger should be used.
Optimum Settings for the SIMATIC VS 110
Vision Sensor SIMATIC VS 110 Tips and Tricks
2-8 A5E00188390-01
Note
To achieve a reliable and error-tolerant evaluation, the Vision Sensor SIMATIC
VS 110 also returns a NOK signal if the image obtained cannot be evaluated (for
example due to an image or illumination disturbance). It is therefore absolutely
necessary in the autotrigger mode to use the evaluation signal of the VS 110
directly. Linking the quality signal of the SIMATIC VS 110 and an initiator signal at
the ejector position using a shift register in the additional controller should be
avoided!
2.5.2 External Trigger
The use of a precise external trigger (for example a precise light barrier) always
improves the maximum achievable precision of the evaluation. The test object is
always recorded at the same position and therefore always has the same distortion
resulting from the perspective.
It is also advisable to use the external trigger particularly with test objects whose
dimensions or transport speed approaches the limits of the permitted specifications
for the SIMATIC VS 110 you are using.
In the setup mode, the SIMATIC VS 110 returns a live image of the sensor field of
view via the serial port. To set the optimum trigger position and arrangement of the
sensor head during setup, the SIMATIC VS 110 shows the corresponding image in
the stop mode for 4 seconds after each trigger signal (in later versions for 60
seconds). On completion of this time, the device returns to the live image mode.
If the time in which the triggered image is displayed is not adequate, it is possible
to display the triggered image permanently until the next trigger signal is received.
To do this, follow the steps outlined below:
1. Select the "Train" menu command from the main menu and then select a
model. Confirm with OK.
2. "Train background" is displayed on the device.
IMPORTANT: Do not confirm this display with OK but remain in this menu item
(before training the background). Otherwise, you would delete the previously
trained model stored at this memory location!
3. As of the next incoming trigger signal, only the triggered pictures are sent over
the serial interface. You can now set the optimum trigger position and sensor
head position.
4. When you are finished, remember to close the menu with "ESC" otherwise you
will trigger a new training sequence.
Optimum Settings for the SIMATIC VS 110
Vision Sensor SIMATIC VS 110 Tips and Tricks
A5E00188390-01 2-9
2.6 Obtaining a Useful Quality Limit
After training the background and the good parts in view A and B, the SIMATIC
VS 110 expects a number ("Samples") of good parts in both views to calculate the
useful quality limit for this particular application.
Please make sure that during the calculation of the quality limit only good parts are
fed through since the evaluation would otherwise classify the bad parts as OK.
The number of samples can be modified using the system settings. The following
notes will m ake your selec tion eas ier:
• Using a low number of samples (approximately 10) means that you can
complete the training run quickly. This low number of samples should only be
used when the parts are fed through extremely precisely and only have very
slight variations in their surface appearance.
• With an increased number of samples (at least 50), you can achieve a more
reliable evaluation. This is particularly useful for objects that can vary
considerably from part to part. The "Screw Test" example illustrates this point.
- When screws with a hexagon head are fed through, the head of the screw
can occur at any position relative to the longitudinal axis in the image.
Depending on its position, it has a higher or flatter contour in the image.
- Depending on the position of a screw relative to its longitudinal axis, the
thread will app ear dif ferently in the im age.
2.7 Setting the Y-Limit Quality Limit
The use of this parameter ("Y-Limit enable") is important when two objects of the
same shape can differ in height as illustrated by the two bolts of different lengths in
the picture below.
The evaluation of a test object compared with the trained reference test object is
made exclusively based on the fundamental surface match if the "Y-Limit disable"
parameter is set and does not take into account the difference height of an object
compared to the trained reference part. This ensures that differences in height of
the test objects, for example due to unevenness in the conveyor belt does not have
a negative influence on the evaluation.
Optimum Settings for the SIMATIC VS 110
Vision Sensor SIMATIC VS 110 Tips and Tricks
2-10 A5E00188390-01
If, however, it is necessary to evaluate the height of the object as a quality
characteristic (evaluation of the height of a part with an identical silhouette), the
"Y-Limit enable" parameter should be set. In this case, the permitted height
variation of the object is checked first and only then the permitted surface variation.
The basic check of the surface variation cannot be deselected.
To be able to evaluate the different lengths of the two bolts in the picture above,
the two bolts should be trained in view A/longer bolt and view B/shorter bolt with
"Y-Limit enable" set. On the one hand, this allows the two bolts to be distinguished
from each other (quality signal OK_A for long bolts, OK_B for short bolts) and on
the other hand, bad parts (damaged bolts, foreign parts) can be sorted out based
on the NOK quality signal.
If you had set the parameter "Y-Limit disable" for the two bolts discussed above
during training, you may receive the "Identical Objects" error message. The reason
for this is the fact that the two bolts differ only in their length but not in their basic
shape. When training the objects in view A and B, the device would detect them as
being identical. This means they cannot be distinguished and therefore not trained.
Note
• In applications in which higher and flatter but otherwise practically identical
objects nee d to be disti ngu is he d, you sh ou ld alwa y s selec t the "Y- Lim it en ab le "
parameter and train the objects in the A and B views. The parts are then
distinguished as OK_A and OK_B and test objects differing from these two
trained objects are classified as NOK.
• In applications in which the required object must be distinguished in the image
from similar objects with an identical shape but different height, select "Y-Limit
disable" and train the object only in the A view. All test objects that do not
match this part will then be classified as NOK.
Optimum Settings for the SIMATIC VS 110
Vision Sensor SIMATIC VS 110 Tips and Tricks
A5E00188390-01 2-11
2.8 Compensating Height Vari ati ons
Depending on the conveyor you are using you may experience differences in
height due to unevenness of the conveyor. These are differences that are not
caused by any variation in the height of the object but are caused in some way by
the conveyor. One possible cause could be, for example, a seam in a belt.
Variations in height of this type are recognized in the image recorded by the sensor
head as an enlargement or reduction in size of the test object; it is not possible to
identify the exact cause (conveyor or bad part). There are, however ways of
impr oving the ev al uat ion :
• Increasing the precision of the conveyor (this generally involves more work but
achieves the best results)
• Using a large number of "Samples" to eliminate the "Accidental" variations from
the evaluation.
• Training the largest sample (highest permitted test object).
Explanation: If the test object is raised by the conveyor (enlarged in the image
field), the SIMATIC VS 110 evaluates the additional visible surface as an
invalid enlargement of the test object. If, however, the test object is lying in a
depression in the conveyor belt where it is transported, the SIMATIC VS 110
detects this and does not evaluate it (within the permitted tolerances) as an
error. For this reason, the highest possible test object should be trained.
2.9 Using the "Info" Menu Command
With many simple applications, it is possible to commission the Vision Sensor
SIMATIC VS 110 without attaching it to a PC for visualization as long as you
adhere to the installation instructions. The operator is, however, left with a certain
subjective uncertainty as to whether the training run really took place successfully.
To be able to check that the training was successful, the operator can use the
"Info" menu command in the RUN menu. Here, the current dimensions of the
usable evaluation window as length x width and the test objects in height for view A
and B and width for view A and B. This display is always in millimeters so that the
operator can interpret them directly. The evaluation area ("Background") should not
change from model to model if the conveyor itself has not changed. The actual
dimensions of the test object can be obtained directly by measuring the object and
can then be compared with the information provided by the device (± 1 mm
tolerance).
Optimum Settings for the SIMATIC VS 110
Vision Sensor SIMATIC VS 110 Tips and Tricks
2-12 A5E00188390-01
2.10 Counting Objects
The SIMATIC VS 110 itself is not capable of counting a specific number of objects
and then outputting an appropriate digital signal. To be able to implement
applications in which a specific number of parts, tablets etc. must be counted prior
to packing, a separate controller, such as a SIMATIC S7-200 can be used.
The use of an external controller is also always useful when more complex
handling is required for bad parts such as using a Pick-and-Place unit.
Vision Sensor SIMATIC VS 110 Tips and Tricks
A5E00188390-01 3-1
3 Mastering Borderline Applications fo r th e
SIM ATIC VS 110
As a basic rule: The closer your application comes to the limits specified for
successful use of the SIMATIC VS 110 (for example in terms of part dimensions,
distance between parts), the more important precise (reproducible) feed of the test
objects and reduction of the feed rate become.
The use of an external trigger also improves the evaluation when the application is
close to the specified limits. By using a precise sensor, the part being checked is
then always recorded at the same point in the image window and effects such as
distortion caused by perspective are reduced significantly.
3.1 Parts Larger than Permitted by the Specification
If you test a part that has dimensions exceeding the specified limits, the message
"Error object too large" appears when you attempt to train the part. To be able to
test such parts with the SIMATIC VS 110, you can use the following tricks:
If you need to check the entire object, you can make the image of the object
smaller by increasing the distance between the sensor head and object. The object
"appears" smaller. The image is, however, not as sharp and the achievable
evaluation quality is reduced. You must also make sure that the object is fully
visible without touching the edge in the active area of the illumination. This solution
is only possible when you do not need a detailed check and the part you are
testing is only slightly larger than the permitted specification.
If you do not need to check the complete object but only a part of it, you can try the
following approach: Select an arrangement of the sensor head/illumination and test
object so that only the part of the object relevant for the test is visible in the image
field. Make sure that the area of interest does not touch the right/left/top edge in
the sensor field of view. The part of the test object that is of no interest must only
extend beyond the bottom of the sensor field of view.
Mastering Borderline Applications for the SIMATIC VS 110
Vision Sensor SIMATIC VS 110 Tips and Tricks
3-2 A5E00188390-01
Special situation: "Extra long" objects
If you have objects that are too long according to the specification (the object
would extend beyond the sensor field of view to the left and/or right during
horizontal transport, for example long bolts or shafts), two tests are nevertheless
possible:
• Type distinction/position detection based on characteristics at one end of the
object.
• Under some circumstances, type distinction/position detection due to the
different lengths of the objects
To be able to use the functionality of the vision sensor for this application, refer to
the following graphic:
Evaluation area of the camer a
this is t ur ned through 90°
clockwise
Left edge
Right edg e
Conveyor Conveyor
The sensor is mounted turned through 90° so that the object to be tested extends
beyond the bottom of the image.
Note
Make sure that you always check the end of the test object that has at least one
important characteristic that allows the type to be distinguished.
A gap must be created in the conveyor system so that the end of the object to be
evaluated is freely visible in front of the illumination unit. To allow the SIMATIC
VS 110 to function reliably, the largest possible free area is helpful, however for
problem-free transportation of the objects (wherever possible without interruption
and limit stop), a mechanical compromise is necessary here. The minimum length
of the gap should, however, be at least 25 mm (for sensor head type
6GF1 002-8AA) or approximately 20 mm (for sensor head type 6GF1 002-8BA).
Based on experience, the test objects must then be at least 2.5 to 3 times longer
than this gap.
Mastering Borderline Applications for the SIMATIC VS 110
Vision Sensor SIMATIC VS 110 Tips and Tricks
A5E00188390-01 3-3
The exact position of the test object must be signaled to the SIMATIC VS 110 by
an external trigger (light barrier, etc.). If you also want the length of the object to be
included in the evaluation, the other end of the object (the end that is not tested by
the SIMATIC VS 110) must be used as the trigger point. The trigger signal must
be a rising edge. The layout of the conveyor therefore only shows the "upper"
(actually the front or back) edge of the test object, the major condition of the
specification for the SIMATIC VS 110 (test object touching the edge only at the
bottom) is therefore adhered to.
By training the test object (in the graphic above this is a shaft with a turned,
tapered end) in view A (shaft in correct position) and view B (shaft the wrong way
round), both the position of the test object can be detected and an incorrect type
(different shape of the shaft end, different diameter, possibly also a different
length) can be detected. If the trigger operates extremely precisely and you select
"Y-Limit enable" and set a comparatively low tolerance, the correct length of the
test object can also be evaluated relatively accurately!
3.2 Deliberate Use of an Inclined Conveyor
The solution above for long objects has several mechanical disadvantages. The
most obvious disadvantage is the need to open the conveyor belt so that it is not
visible in the sensor field of view. This solution is suitable only for long and
mechanically robust objects, it cannot be used for objects that are "slightly too
long". Here, a different approach is necessary.
By skewing the sensor head through approximately 15° to 30° compared with the
conveyor system, the test object appears inclined in the image. When training the
background, however, it is always the largest possible rectangle parallel to the
main axes of the sensor head that is selected. For the evaluation, this means that
only the upper triangle of the test object is evaluated as shown in the picture below.
Mastering Borderline Applications for the SIMATIC VS 110
Vision Sensor SIMATIC VS 110 Tips and Tricks
3-4 A5E00188390-01
Note the following restrictions with this arrangement:
• This application is only genuinely useful for position checks or type distinction.
Testing for faults or anomalies is not possible since only a smaller section of
the test object is visible.
• This application must be externally triggered. The requirements in terms of
reproducibility of the trigger position are high (laser light barrier, etc.).
• In particular for comparatively flat objects, the trigger position should be
selected so that the test object is located close to the side margin (while still
maintaining the minimum distance to the margin).
• The angle of inclination should be selected so that the distinguishing
characteristic on which the position detection/type distinction is based is
adequately captured in the image and so that at the same time no part of the
object is visible any longer at the other edge of the image field; in other
words, the edge contact due to the excess length is below the visible area.
• On the illumination unit, you should fit a mask at an angle, so that in the actual
image field of the sensor, a horizontal lower edge is produced and so that
reproducible training of the background is forced over the entire width of the
sensor field of view.
3.3 Parts Smaller than Permitted by the Specification
If you test a part with dimensions less than those permitted by the specification, the
message "Error object too small" appears on the display. To be able to test such
parts with the SIMATIC VS 110, you can use the approach described below:
• A precise external trigger should be used to ensure that the object is always
tested at the same position and that distortion resulting from the different
perspectives at different positions in the image (and consequently different
perspectives in the view of the test object) does not affect the evaluation.
• In particular with round, highly reflective objects (with a smooth, polished
surface), masks should be fitted to mask unnecessary parts of the illuminated
area. This causes an apparent sharpening of the object edges and maximizes
the visible surface of the test object. See also Section 2.3.
• By reducing the distance between the sensor head and object, the object
appears larger in the image. Although it appears larger, the image is, however,
not as sharp and the achievable evaluation quality is reduced. This solution is
only possible when you do not need a detailed check and the part you are
testing is only slightly smaller than the permitted specification.
Mastering Borderline Applications for the SIMATIC VS 110
Vision Sensor SIMATIC VS 110 Tips and Tricks
A5E00188390-01 3-5
3.4 The Individual Test Objects of a Model are too Similar
It is possible that you want to test objects and distinguish them reliably from each
other although they only have slight differences in their silhouettes when you use
the normal arrangement of the Vision Sensor SIMATIC VS 110 according to the
installation instructions. These test objects appear very similar and, in the worst
case, even identical.
It may nevertheless be possible to distinguish them using the approaches outlined
below:
• When training the objects in view A and B (in other words, either two different
objects in the same position or one object in two positions) you should only use
objects with the original dimensions (i.e. test objects with zero dimensional
toleranc e) wh en train ing the orig in al vie ws (tr a in view A, train vie w B). W hen
training the qualtiy limit, on the other hand, all possible deviations up to the
permitted smallest and largest dimensions should be used. The advantage of
this is that when you train vie w A and B, the larg est poss ib le dif f erenc es
between the objects result.
• With very small test objects, the automatically calculated qualtiy limit may need
to be corrected manually. Due to the algorithm, the automatically calculated
value for the Q-limit is very seldom at values higher than 98.1%. However, to
allow minimal deviations in smaller objects to be detected, this correction may
be useful.
Mastering Borderline Applications for the SIMATIC VS 110
Vision Sensor SIMATIC VS 110 Tips and Tricks
3-6 A5E00188390-01
• If two test objects are very similar, it is usually helpful to include and evaluate
only the relevant characteristics in the image. This is illustrated by the following
four images:
The two original views of the bolt reveal only minimal differences (small
chamfer at one end of the bolt). With a well-positioned mask on the illumination
unit, only the upper part of the bolt is seen, the relative difference between
chamfer left and chamfer right (relative to the remaining visible surface of the
bolt) is now greater. With this arrangement, the minimum required height of the
test object in the image as explained in the manual of the SIMATIC VS 110
must be kept to. This effect could, of course, be greatly increased if the bolt is
fed through standing with only the end of the bolt being visualized.
Mastering Borderline Applications for the SIMATIC VS 110
Vision Sensor SIMATIC VS 110 Tips and Tricks
A5E00188390-01 3-7
• This effect is much greater in the following images:
By masking the unimportant parts, only the upper clamp is taken into account
in this example, the base with the large thread is excluded from the surface
comparison. The upper clamp can now be evaluated with greater precision and
sensitivity.
• With a skillfully modified arrangement of the sensor head and illumination unit,
it may well be possible to achieve a clear distinction between parts that differ
from each other as objects in space, but that would have no significant
differences when viewed with an arrangement according to the SIMATIC VS
110 specification and could therefore not be tested. This trick involves, in
particular, deep objects or objects with a significant difference in contour along
the depth axis. By inclining the sensor head to the transport axis (depending on
the application and mechanical setup, approximately 30° to 60°), it is also
possible to include the depth axis of the test object in the evaluation.
Depending on the application, it may then be necessary to incline the
illumination unit as well.
The basic arrangement is shown in the following diagram:
The parts would be identical in the
backlight with the normal installation
of the sensor .
The inclined angle of the sensor
makes recognition possible.
Conveyor
Illumination
Sensor head
Mastering Borderline Applications for the SIMATIC VS 110
Vision Sensor SIMATIC VS 110 Tips and Tricks
3-8 A5E00188390-01
This method does, however, only allow a rough evaluation due to the
considerable perspective distortion; a precise comparison with the guaranteed
evaluation accuracy according to the specification is therefore not possible.
3.5 The Individual Test Objects of a Model are too Different
If the variation of the objects of a single model is too high, internal NOK
classifications may result during training of the qualtiy limit; in other words, these
"mavericks" are ignored during the automatic calculation of the qualtiy limit
because the device assumes bad parts among the good batch. This is a process
that takes place within the device and is not to be confused with the NOK signal
that is always triggered during training! In this case, it is advisable to train the test
object both in view A and view B (with the objects in the same position!) to be able
to cover the possible or necessary variation in the evaluation. Outputs OK_A and
OK_B must then have the same significance.
The two pictures below provide an example. The difference in shape (with or
without hole) of the individual objects of this type it too great to achieve acceptance
during training of the qualtiy limit simply by feeding through an adequately high
number of samples. Here, the approach outlined above is necessary.
Mastering Borderline Applications for the SIMATIC VS 110
Vision Sensor SIMATIC VS 110 Tips and Tricks
A5E00188390-01 3-9
3.6 Round Objects with Asymmetrical Characteristics
When feeding through round objects (bolts, screws, shafts etc.) with a through hole
in them or flattened on one side, you may encounter problems during evaluation.
If the position of these objects can change around the longitudinal axis so that this
asymmetrical characteristic is more or less visible as a contour in the silhouette
due to the rotational position, two different applications must be distinguished:
• The task of the SIMATIC VS 110 should be the detection of the correct
rotational position. In this case, only view A should be learnt, if a second
distinguishing position needs to be detected, view B must, of course, also be
trained. When training the qualtiy limit, the parts must also be fed through in
this distinguishing position so that the device can evaluate precisely later.
• The task of the SIMATIC VS 110 should be simply to recognize the correct
type of part fed through. In this case, depending on the rotational position, the
variable contours will cause problems in precise evaluation. Here, a large
number of samples should be used for automatic training of the qualtiy limit
and the parts fed through in all possible positions during training. The
restrictions (as described in the previous section) must, of course, not be
ignored.
3.7 Testing Endless Material
Basically endless material (railing, material for punching etc.) cannot be tested
according to the specification of the SIMATIC VS 110 since such material would
lead to invalid contact with the right and left edges of the image. Under certain
conditions, however, it may nevertheless be possible to check the correct punching
of materials.
The following conditions must be met:
• It must be possible to arrange a rectangular frame around the detail to be
checked that touches the test object only at the bottom edge.
• The object being checked must be positive, in other words, you must check
what is produced by the punch and not the remnants of punching (i.e. not the
hole left by punching).
• The object to be checked must not be too "fine"; in other words, the individual
details must be significantly larger than the minimum qualtiy limit.
• The application (as usual with punch material) must be triggered externally.
Mastering Borderline Applications for the SIMATIC VS 110
Vision Sensor SIMATIC VS 110 Tips and Tricks
3-10 A5E00188390-01
The two pictures below illustrate such an object. It is possible to arrange a
rectangle around the contact stud that touches the test object only at the lower
edge. You could, for example, create this frame with a suitable mask on the
illumination unit. If, however, as in this example, you only want to check the
parallelism of the contact studs, you can significantly raise the precision of the
evaluation. In this case, you select an image section so that only the ends of the
two contact studs are visible (left image). Changes compared with the trained
reference image (right image) then count for a significantly higher percentage in
the evaluation.
Note
If you are testing an object similar to the one in the picture above, make sure that
you choose an external trigger. Since the two contact studs have an identical
shape, if it was trigg er ed aut omat i call y, the SIM ATIC VS 110 would ass ume two
individual, identical parts and evaluate each stud separately. All the separate,
unconnected parts in the sensor field of view are interpreted as belonging together
only when you use external triggering!
Mastering Borderline Applications for the SIMATIC VS 110
Vision Sensor SIMATIC VS 110 Tips and Tricks
A5E00188390-01 3-11
3.8 The Maximum Number of 15 Model s in Total Is not
Enough
If you need to save more models and if each type only needs to be checked in one
position (only in view A), view B can be interpreted as a second storage level on
the device. In this case, part 1 is trained as view A under model 1, part 2 as view B
also under model 1 etc. In this way, instead of 15 models, it is possible to store up
to 30 different models in non-volatile memory on the device.
Two restrictions do, however, apply here:
• To train two different types in view A and view B, they must be similar parts.
Parts with considerable differences cannot be stored under the same model.
• To evaluate the quality signals OK_A and OK_B, an additional control is
necessary. This control must know which type is currently being tested and be
able to interpret the other quality signal OK_x as NOK, so that parts of type 2
are not classified as OK when testing parts of type 1.
3.9 Precise Setting of a "Dimension Check"
SIMATIC VS 110 cannot measure, its mode of operation is to compare.
Nevertheless, to set the comparison to a precise tolerance in terms of dimensions,
the following method is helpful:
• A good part within the permitted tolerance is trained as view A.
• To inform the device of the fluctuations in tolerance that can be accepted
during evaluation, the largest and smallest samples of this type are fed through
alternately when training the qualtiy limit automatically.
• SIMATIC VS 110 adapts itself to the variation of the good parts during training
of the qualtiy limit, in this case to the largest and smallest permitted parts. All
parts within this tolerance are evaluated as 100% matches during later
evaluation.
• By keeping the tolerance of the Q-limit qualtiy limit tight (for example 98%), if a
slight violation of the tolerance occurs during the evaluation, a NOK reaction
from the SIMATIC VS 110 can be forced.
Note
This method aiming at high evaluation precision can only be used to any practical
purpose when the test objects are fed through very precisely and an external
trigger is used . This rema ins a compar a ti ve met hod based on the comp ar is on of
the contour surface of the test object before the backlight with the trained reference
surface. There is no actual measurement of the test object even with this
approach.
Mastering Borderline Applications for the SIMATIC VS 110
Vision Sensor SIMATIC VS 110 Tips and Tricks
3-12 A5E00188390-01
