Label Presence Inspection 


in-the-Labeller Machine



Label Inspection

       Flash animation, 49 MB, ZIP

   After download click index.html 

    

          

            70 pages, 37 MB










Where to establish a correlation

The best place to control labels' application is directly in-the-Labeller Machine, thus having the possibility of an early warning about the typical issue affecting all Labeller Machines: their Consecutive Faults.   In this configuration, the photo-electric sensors have to be installed around the Labeller central carousel and after the rotary, prismatic-shaped brushes securing the labels’ adherence to the container sidewall, immediately after each labeling station.   Ideally, the label presence inspection should have to be immediately after a labelling station, so to let the Electronic Inspector feedback to the Labeller Machine automation, an earliest than possible warning signal about the fact its consecutive faults.   In the reality, this is rarely possible.   All Labeller Machines have cams part of the central carousel, also projectually defining a reduced space where the labelled bottles are oriented toward the Label Presence Inspection photoscanners, like visible in the small figure below at right side.   


Consecutive Faults

A proper orientation of the bottles should have to be imposed by the cam immediately after each label application.   Below at right side an example, showing one of these systems detecting the missing Neck-, Back- and Body-labels directly in-the-Labeller Machine, in a Carlsberg® Group Brewery.    It provides this way a vital early Stop Impulse Signal with a delay minimised to a digital I/O card part of the Labeller Machine's PLC automation.   Delay reduced to ≤2 mislabelled and rejected bottles.    A problem is however that the force exercised on the recently applied labels by the prismatic-shaped sponges securing the labels’ adherence to the container sidewall, have the potential to let them flag or remove at all many of the incorrectly applied labels.   Typically, those not sufficiently glued.   We are speaking of labels uncorrectly applied, later lost but yet detected Present by the Electronic Inspector.   Implying many non-labelled bottles later non rejected.   Because of this reason, the majority of the Labeller Machines have the cam designed to revolve to the proper inspection angle all bottles only in the out feed of the carousel, immediately before the out feed starwheel, like visible in the figure above.    When the inspection zone defined by the Labeller Machine’s cam is immediately before its outfeed starwheel, it’ll not be possible to have a truly early Stop Impulse.   In this case, on the opposite we’ll have late Stop Impulse, deriving by the Electronic Inspector's photoscanners installed immediately before the Labeller out feed starwheel.


 Body, neck and back label presence inspection analog photo sensors, installed in a Krones AG Labeller Machine.  Hold by a single bracket allowing vertical movements referred to a (metallic) half-millimetric ruler (not visible in the picture).  Orientation angles fixed by dot-like weldings.  All of the cables passing thru metal bars, rather than exposed to bottles’ hits, steam and glass splinters








Thus, meaning a delay ranging (5 ÷ 15) non-labelled rejected bottles.   An amount of True Positives and consecutively rejected bottles proportional to Labeller Machine size and Production speed.    A high price paid along the following 15 years, in terms of erroneously labelled bottles, clogging the Rejects Accumulation Table.   Then also forcing a presence of human-Operators, that in many Countries is today extremely expensive.   The fact that we are touching the subject of a proper inspection angle hints to the fact that there are also Labeller Machines whose Mechanical Designers are not fully aware of these strictly-Optoelectronic constraints.   Constraints inherent to what angle and distance they have to exist between the Electronic Inpectors’s photoscanners and the labelled bottles.   


 Body, back, neck and  all-around label presence inspections and positions.  In evidence below a counter limit for serial faults.  One key advantage of the Label Presence inspectors in-the-Labeller over the models at-the-Conveyor.   To have the possibility of an extremely early warning of consecutive faults, as an example, 2-3 steps after label application, means possibility to stop automatically the Labeller by the Inspector, preventing losses on Production and downtimes























We have memory and records of cases where the Labeller’s cam Design was clearly flawed, ignoring Optoelectronics:   

  • erroneously skewed (≠ 90º) angles from the label to the photoscanner's optic axe;
  • orientation angle correct but too-short label exposure to the photo-electric detector, thus not giving to the Electronic Inspector Time sufficient to deduce enough random variables in the sample space.    An alias of what in laymen language is not a “good measurement” or, a measurement with ample fluctuations associated to consecutive measurements of similar labels;
  • orientation angle correct only under a sensing distance greater than Photoscanner’s maximum.   Sensing distance is, as we’ll see later, a specification of the Photoscanner which cannot be modified or increased.  The labels constraints the values of reflectivity forcing the use of one or another photoscanner, so to avoid under-exposure, oscillations when operating in the interdiction area of the amplifiers, or over-exposures when operating in the amplifier’s saturation area.

False Rejects along the following 15 years' lifetime terminated to be the unavoidable outcome of a kind of too many Design incoherences, too-expensive to be fixed one time the Labeller Machine itself is commissioned.


How correlation is established

The incident photons emitted by a LED in the Photoscanner, green-coloured in the figure below at right side, interact with the label's or bottle's surface atoms, then forming secondary rays, in the example red-coloured.   These secondary photons are reflected from the surface diffusely, scattered in all directions.  Ideally, following the cosine Law of Lambert, an illuminated diffuse reflecting surface will have equal luminance from all directions which lie in the half-space adjacent to the surface.   A very small conical shaped fraction of them shall enter a photo-detector lying on side of the LED light source in the Photoscanner's plastic or metal case.


  Diffusely reflected photons’ scattering (abridged by    Renderstuff.com/2013) 



Why so many False Rejects ?

Criticity of the Label-Photoscanner state








We saw with details elsewhere in this web site, how the measurement’s steps to form a correlated object-detector state.    As an example we’ll try to determine if a bottle is labelled by mean of a common Photoscanner.    Photoscanner including a LED illuminating the container's sidewall or neck, where the label is expected to be, later receiving and processing just a fraction of the retro-reflected diffused waves.  Until switching its digital output if the retro-reflected waves’ intensity is greater than a pre-defined threshold level.    Whatever physical system, label included, is represented by its wave function or state vectors.    The physical meaning of the state vector becomes apparent when making a measurement.   Then the state of the system assumes one of the eigenstates, with probability given by the Born rule, and the result of the measurement is the corresponding eigenvalue.

  To establish a relation between a label (“Label”) and a Photoscanner (“Photoscanner”), both differentially related with the Environment (“Environment1”, Environment2”), is necessary Time.   Time to transform the previous state, in which all possible kinds of correlation of the Photoscanner coexist, in a following state in which the Photoscanner is “aware” to be correlated to a label, because having recorded eigenvalues for the eigenfunction ΦiS1 describing a cap.  Quantomechanical explanation of the measurement process, unaffected by the circularities implicit in the classic explanations.  The vertexes represent interferences


Making a further step, to definetely perceive a correlated Label-Photoscanner State, they are necessary:

  1. Timeto transform the previous state, in which all possible kinds of correlation of the Photoscanner coexist, in a following state in which the Photoscanner is “aware” to be correlated to a Label, because having recorded eigenvalues for the eigenfunction ΦiS1 describing a Label.    The correlation between the two systems, Photoscanner and Label, is progressively established during interaction and proportional to the natural logarithm (ln t) of the interaction time t.   An ideal correlation, corresponding to a maximised information of the Photoscanner about the Label, can only be reached allowing an infinite time.   The fact we cannot wait for an infinite time causes the measurements’ fluctuations, a synonimous of the spectrum of the eigenvalues, resulting in the Electronic Inspector's false positives (false rejects).   Time, for what ?    To transform the previous state, in which all possible kinds of correlation (superpositions) of the Photoscanner coexist, in a following state in which the Photoscanner is aware to be correlated to a Label, because having recorded eigenvalues for the eigenfunction ΦiS1 describing a Label.  


  1. Interaction between the systems such that the information in the marginal distribution of the object inspected is never decreased.    In a probability distribution deriving by two random variables, we remember that marginal distribution is where we are only interested in one of them.     

 

 To detect a Label they are necessary Time and a way of interaction not reducing the information in the marginal distribution of the Label.  In laymen words, an interaction not modifying the label.  Label Presence photoscanners fulfill the second condition but a limit exists on the fulfillment of the first condition.  It is the Labeller Machine’s own cycle what decides how long Time is given to the Inspector’s inspections.  The limited Time becomes relevant when inspecting the presence of translucent labels.  The red colour box evidentiates a mainly transparent label.   Its reflective area is reduced to just a dark blue low-reflectivity area.  A label like this requires the photoscanner directed at the clear text “CARLTON DRY”, its only reflective area.  As a consequence, it’ll result impossible to detect flagging or inclined labels, rather just presence




Otherwise, we’d have forced a reduction in the sample space of one of the random variables and then, we could not have any more repeatability of the measurements.  As an example, this should be the case if we’d erroneously try to use a beam of high energy neutrons, rather than LEDs’ low energy photons, to interact with the Label.   The neutrons should modify the molecular structure of the Label, modifying its eigenstates and then the eigenvalues we expected to derive by the measurement.  

What before is the scientific background explaining why also double, triple or quadruple photoscanners, and also if of the best Quality (i.e., ~260 $ each), can result insufficient to reach satisfactory results.   Satisfactory in terms of defects' rejection ratio with minimum false positives.  Be wary of proposals for systems which could later leave only you and your Company with the nightmare to have the products' store, filled with yet palletised flagging-, malpositioned- or missing-labels. 



  Staff in the Beverage Bottling Halls knows what a near-nightmare is a store of pallets “contaminated” by many flagging-, missing or malpositioned-Labels (    SABMiller plc, Poland, photo by T. Parker, OneRedEye, 2012)



Receiver Operating Characteristic analysis




ROC is a graphical plot adopted in the Signal Detection Theory and Statistics to illustrate the capability of a Binary Classifier system when its discrimination threshold is being varied.   Binary Classifier because all of the existing Electronic Inspectors, also named Bottling Controls or more simply Controls, are part of the great family of the Binary Classifiers.    

  ROC-curve of a Label Presence Inspection.   Imagine to process a statistically significative sample of bottles by mean of an in-the-Labeller Machine Electronic Inspector equipped with Label Presence Inspections.  In the meantime, check one-by-one all rejects attributed to all its Label Presence Inspections, like its Body-, Back- and Neck-inspections.  On the base of the Sensitivity threshold set in the photo-Detectors, they’ll arise their respective statistical distributions of the hits and of the misses, True Positives, False Positives, True Negative and False Negatives, like visible above.   If you’d repeat the same identical test changing the Detector with another with superior characteristics, you’d also see that the same Sensitivity threshold adjustment of the former test is now allowing a huge cut on Production losses, the False Positives (FP, pink colour).   The fraction of actually-labelled bottles falsely considered nonlabelled.  Production losses’ reduction without any increase of the False Negatives (FN, light blue colour), say the actually-nonlabelled bottles erroneously considered labelled and not rejected (abridged by    Jutta234/CC BY-SA 3.0/2006)


“ROC analysis allows a precise evaluation of the fraction of actually-nonlabelled bottles as a function of the fraction of actually-labelled bottles falsely considered nonlabelled”

In brief, ROC analysis is relevant in all our applications because allows: 

  • to graph the Sensitivity of our Detector, in this case, a digital photoscanner used to detect the presence of a label, as a function of the False Positives.   False Positives that in our Food and Beverage Bottling industrial field are commonly (and, unproperly) named False Rejects;
  • a precise evaluation of the fraction of actually-nonlabelled bottles (True Negatives, TN) as a function of the fraction of actually-labelled bottles falsely considered nonlabelled (False Positives, FP).



Fourier, Signal-to-Noise ratio and Rejects









The choice of the photoscanner, also named photo-electric sensor or photo-detector, occupies a central role.   It has not to contradict what the Experts of Radio Frequencies' technologies perfectly know.   Namely, that the conversion between the electromagnetic form of the energy and the electric is made in the “antenna” and that’s why this is the component where to invest a sensible part of the entire value of a detection system.     Why ?    Also, but not only, because otherwise the distortions introduced by the transducer of the electromagnetic  electric Signal, first link of the detection-amplification-switching chain, shall have further amplification.   Amplified jointly with the truly incoming Signals (see figure below at right side), until an output signal of high amplitude no more resembling the truly incoming Signal.   There is a name, the name of the great French physicist and mathematician Jean-Baptiste Joseph Fourier whose discovery since two centuries dominates the entire Mathematical-Physics and its Engineering applications. 


  Whatever periodic function however complex, may always be represented as the superposition of an average constant value and two periodic series, in cosines and sines, including infinite terms.  As an example, subtracting by the superposition all the odd-numbered terms and leaving only the even-numbered and the average value, a sinusoidal function becomes a squared-function.  A Detector subtracting and/or adding and/or modifying the amplitudes of each one Fourier's term, is introducing a distortion.  Aperiodic signals are treated pretending that they are periodic with an infinite period


It makes sense of the components:












  • existing in the Output Signal, but non-existing in the incoming Input Signal,
  • plus others, nonexisting in the Output Signal, but existing in the Input Signal.

A digital photoscanner in a Label Presence Inspector may be imagined alike an electronic component acting similarly to an antenna, but with a basic difference in its output signal.  Output signal which is not analog like those of the aerials, rather a digital on-off corresponding to: 

  • normally open (NO) and/or normally closed (NC) contact,
  • and/or PNP and/or NPN polarity.

Two of the specifications are particularly relevant during Label Presence Inspection:

  1. centre of the waveband of the Photoscanner light emission;  
  2. detection diagrams, showing the response curve.  

For the point 1. above, refer to the electromagnetic spectrum visible in the figure below, where it is visible a waveband centered in the infrared (IR).    In the following, we’ll focus some of their characteristics.   

infrared-ir-fill-level med med

 The portion of the electromagnetic spectrum named Infrared (IR)



Photoscanners’ Margin feature

“All sensors should be operated at a margin greater than 2x (x, the minimum amount of light necessary to switch).   

Detection of a low reflectivity target notoriously requires sensors with higher margin: higher Quality sensors”

The data sheet visible below is not that of the general-purpose designed ~40 years ago and present wordwide in many electronic inspectors.   Rather another general-purpose, modern and high-Quality photoscanner.   One of the most fundamental technical features of the photoscanners or photodetectors is named Margin.   “Margin” is defined as the measurement of the amount of light from the light source detected by the receiver, expressed as a multiple of the minimum amount of light level required to switch the sensor output.   Then, a margin of 2x is reached when the light level received is twice the minimum required to switch the output.    All sensors should be operated at a margin greater than 2x.   Detection of a low reflectivity target notoriously requires sensors with higher margin: higher Quality sensors. 

As an example, refer to the Data Sheet for the Photoscanner at right side.

Visibly:

  1. light emission is Infrared, centered at a wavelength of 880 nm (IR);
  2. detection diagram hints to another property (label colour and reflectivity).   Showing, as an example, that a grey colour (flat) surface at 10 cm of distance provides the top of the detected signal.    A distance which is doubled to 20 cm if the surface is white colour, then more reflective;
  3. detection diagrams are always and only referred to a flat surface an
    d not to a curved one, like a label applied to a bottle.   Implying that the amount of 
    iffused reflected electromagnetic energy is much superior to that available in our beverage packaging inspection.  
    Implying, as an example, that a grey coloured label has to be optimally detected at a distance shorter
    than 10 cm and a white label, at a distance shorter than 20 cm.




Focus on the packaging 

“Detection of a low reflectivity target notoriously requires sensors with higher margin: 

higher Quality sensors”



If the labels whose presence has to be detected are:

  • dark, and not clear colour, then less reflective, 
  • transparent or, semi-transparent, like visible in the figure below at right side,
  • curved, and not flat, then reducing the amount of electromagnetic energy retro-reflected into the phototransistor of the Photoscanner, the semiconductor device whose output signal is typically the input of a transistor acting as amplifier,

the photoscanner shall not be adapt to fulfil the Technical Guarantees for the entire Label Presence Inspector.    

 Maximum reflectance of flat labels of various colours, as a function of incident light’s wavelength  (  Omron Corporation/2006)


Today’s common transparent labels require special photoscanners.  Photoscanners rarely or never fabric-installed by the Vendor.  Frequently, labels are so transparent that several special photoscanners are necessary, so to superimpose the results of several independent inspections   

Today’s common transparent labels require special photoscanners.  Photoscanners rarely or never fabric-installed by the Vendor.  Frequently, labels are so transparent that several special photoscanners are necessary, so to superimpose the results of several independent inspections


If one of these this is the case, the Electronic Inspector shall unavoidably show:

  • False Positives, say correctly labelled bottles uncorrectly detected non-labelled and then rejected,

or: 

  • False Negatives, say truly non labelled bottles, not detected and not rejected,

and, sometimes:

  • False Positives mixed with False Negatives.






  Assessing the quality of labelled bottles in a SAB MIller Brewery.  In the distance, the rejects of the Full Bottle Inspector controlling Label Presence in-the-Labeller and Fill Level at-the-Conveyor at left side   Kompania Piwowarska SA/Tom Parker/OneRedEye/2012)


Photoscanners comparison. The Label Presence inspection most important component is the digital photoscanner, also named photo-detector, photocell, etc.   This, is also the one where frequently are concentrated designers’ efforts to spare on parts’ value.   It is frequent to observe Label Presence Inspectors paid in the range (30000-120000)$ whose most truly important component is, in a reality known to Electronic Engineers, a general-purpose photoscanner born by a design of ~40 years ago.    Also, a component which when bought in amounts of >1000 units, is paid to its Fabricants <30 $.    

To have a first contact with the key point, imagine to have just one bottle:   

labelled with a Neck Label of the today commonly adopted semi-transparent kind;   
let it revolve on a rotary carousel on a plate and its retro-reflected diffuse light be detected by two photoscanners differing in their basic characteristics and prices;   One of them a cheap general-purpose the other an expensive modern model specifically designed for semi-transparent, transparent or dark labels;
write down each outcome in the sample space Ω of the A/D converted measurements derived by the Label Presence Inspection.  Outcomes displayed by the Label Presence Inspections of different Vendors and technologies, as adimensional numbers, or distances in millimetres, or pulses;
translate the obtained gaussian profiles, til superimposing their respective peak values;
observe the spiked distribution of the modern high-Quality photoscanner, explicitly hinting to high repeatability and low false rejects to be expected in Production;
on the opposite, the cheap general-purpose photoscanner, costs eight times less than the other, but is promising Production losses to the entire Beverage Bottling Line along the following 15 years !
Then, applying a little of Arithmetic and tells us that these Design intentional choices are presently investing in the most important Label Presence Inspection component a relative paid value ranging:  

                                                 (0.1 ÷ 0.025)%   



thus affecting negatively the entire electronic inspector’s Receiver Operating Characteristic (ROC) in all critical label presence inspection applications.    ROC is a graphical plot adopted in the Signal Detection Theory and Statistics to illustrate the capability of a Binary Classifier system when its discrimination threshold is being varied.   Binary Classifier because all of the existing Electronic Inspectors, also named Bottling Controls or more simply Controls, are part of the great family of the Binary Classifiers.

  Cheap, obsolete general-purpose photoscanner versus modern, expensive photoscanner designed for transparent labels. What happens when exposing a statistically significant number of times one and the same labelled bottle to two different technologies’ photoscanners











We’ll deepen in the following section the main reason for so many Production losses frequently observed as a result of the Label Presence Inspection.   What follows cannot be generalised to all Vendors.    But, it touches however directly ~70 % of all the Label Presence Inspectors presently operating in the entire World.    The Label Presence inspection most important component is its digital photoscanner, also named diffuse reflective sensor, photo-detector, photoelectric sensor, etc.   Too frequently, also the component where are concentrated the efforts of some Designers to contain the grand total cost of all the Electronic Inspector’s parts.    Not many Designers but, being one of these Vendors market share as high as 40 %, the net result that the sum of all these Label Inspectors accounts in the end for ~70 %.    As an example, refer to many Label Presence Inspectors sold to Food and Beverage Packaging Companies whose price lies in the range (30000 ÷ 120000) $, or (25000 ÷ 100000) €.    You’ll discover what only Experts know, that the most important component of a Label Inspector is reduced to a “general-purpose photoscanner born by a design of ~40 years ago”.    If you want to remain astonished and proceed straight to the facts, try to Google the words:


18mm-Diffuse-Type-Adjustable-Photoelectric-Sensor-Switch            


and, in the lot of the results you’ll encounter that these transducers, the core of the Label Inspection, are components which, if bought in amounts of >100 units, are paid just 4.8 $ or, ~4 €.    To have a first contact with the key point, imagine to have just one bottle:   

  1. labelled with a Neck Label of the today commonly adopted semi-transparent kind;   
  2. let it revolve on a rotary carousel on a plate and its retro-reflected diffuse light be detected by two photoscanners differing in their basic characteristics and prices;   One of them a cheap (4.8 $/each) general-purpose the other an expensive modern model specifically designed for semi-transparent, transparent or dark labels;
  3. write down each outcome in the sample space Ω of the A/D converted measurements derived by the Label Presence Inspection.  Outcomes of measurements displayed by the Label Presence Inspections of different Vendors and technologies, as: adimensional numbers, or distances in millimetres, or pulses, etc.;
  4. translate the obtained gaussian profiles, til superimposing their respective averages; 
  5. observe the spiked distribution of the modern photoscanner (260 $), explicitly hinting to high repeatability and low false rejects to be expected in Production;
  6. on the opposite, the cheap general-purpose photoscanner, costs 68 times less than the other, but reduces the entire Beverage Bottling Line Production losses and non-labelled bottles to the Market, along the following 15 years; 
  7. evaluate the cost of an individual False Positive (false reject);
  8. evaluate the relative contribution to the Total Cost of Operation (TCO) of the less performant technology Neck Label Presence Inspection, using the other as a reference, integrating it along the expected 15 years of lifetime of that Electronic Inspector;  
  9. you’ll discover that the general-purpose photoscanner, applied in a Labeller to the detection of the modern semi-transparent, transparent or dark labels, shall cost you losses much superior to the price of a second Electronic Inspector !     Much, much more than the price you’d have paid (260 $) to eliminate this problem banally upgrading the photo-Detector technology to the Third Millennium's Designs.

Cheap, obsolete general-purpose photoscanner versus modern, expensive photoscanner designed for transparent labels.  Shown their pricing with respect to a Label Presence Inspector priced 30000 $ (25000 €).  The False Rejects you suffered until today, have a well-defined definite Root Cause. The cheap obsolete photoscanner commonly fabric-installed, literally disappears in the comparison. Weighing just 0.12 ‰ of the Label Inspector price.  Disappears in the comparison but not in the negative accountancy of a Beverage Bottling Line Production’s Losses



Making a rough back-of-the envelope calculation tells us that the present Design choices are devoting to the most important Label Presence Inspection component, just a negligible fraction of the price paid by the Bottler:

  1.  to buy the Electronic Inspector;
  2.  to operate the Electronic Inspector along its 15 years expected lifetime.

Limiting the comparison only to the precedent point 1., considering Label Electronic Inspectors valued in the range 30000 $ to 120000 $, just: 

               

               (1.6  ÷  0.4) ‱  


thus, affecting negatively the Receiver Operating Characteristic (ROC) in the most critical, prone-to-false-rejects Label Presence Inspections.   Above, graphed the comparative relative porcentages of a relatively cheap Electronic Inspector (30000 $), and of two kinds of photoscanners it may use as transducers for its Label Presence Inspections.  The modern, relatively expensive model, designed for the actual and difficult transparent and semi-transparent labels, costs <260 $.    It is the “expensive” solution.  



 In evidence the transparent self-adhesive labels.   Their reflective area reduced to just small texts, thus requiring photoscanners’ orientation at the text.  As a consequence, they’ll be rarely detected flagging or inclined labels, thus terminating to the Market.  Detection is reduced to the mere label presence or absence.  A typical case benefitted by the modern photoscanners including microprocessors, processing with fuzzy-logic just polarised light.  Detectors conceived for these kinds of labels


The cheap obsolete photoscanner commonly fabric-installed, literally disappears in the comparison.  Weighing just <0.16 ‰ in the total Label Inspector price.   Disappears in the comparison but not in the negative accountancy of a Beverage Bottling Line Production’s Losses. 


   All general-purpose cheap photoscanners, widespread and costing a few dollars, adopt circularly polarised light as interactant.   Circularly polarised is just a way to say that no filtering exists for a specific polarisation angle: all of them are used.   Now, imagine to superimpose to the clear reflective foam in these necks a relatively dark Neck Label.   Marketing Dept., and not the Bottling Dept., decides colours and shapes on base of reasons which are nearly completely out of those where the Beverage Bottling Staff may prevent a new format of label definetely so “dark”.   Labels less reflective than the foam, implying foam emulating non-existing labels.   Add a low-reflective dark label, applied in the most curved area of the bottle, to a relatively fast Labeller Machine (>40000 bph), let them checked by an intrinsically low-Quality Photoscanner of the Electronic Inspector.   The grand total shall be a maximum of Production Losses and minimum Quality.   What to do in these unfortunate cases ?    Impossible to change labels then upgrade the photo-Detectors (   SABMiller, Kompania Piwowarska SA, Poland /Tom Parker/OneRedEye/2012)



General-purpose Photoscanners’ Scope











For what kinds of labels are adapt the cheap photoscanners?    It’s a history started well before Electro-Optic components started to be applied in the Food and Beverage Industry.    In 1931 Paul Kubelka and Franz Munk created a relevant part of the Theory of Reflectance.   They developed it with paint films in mind and discovered that it also worked quite well in many circumstances also for paper.   The Theory works best for optically thick materials, those where >50 % of light is reflected and <20 % is transmitted.   Comprehensibly, the modern semi-transparent, transparent and very dark labels are exactly those where <50 % of the light is reflected and >20 % transmitted or absorbed.   These two German scientists however saw their theory was not completely rendering the diffused reflection of the: 

  • dyed papers,
  • very dark, unbleached papers,

when light absorption reaches a high level.   And one more constrain of their Theory is that the absorbing and scattering media have to be uniformly distributed through the label. 

Diffused light general-purpose photoscanner limits.  General-purpose Photoscanners’ Scope





















For what kinds of labels are adapt the cheap photoscanners ?    It’s a history started well before Electro-Optic components started to be applied in the Food and Beverage Industry.    In 1931 Paul Kubelka and Franz Munk created a relevant part of the Theory of Reflectance.   They developed it with paint films in mind and discovered that it also worked quite well in many circumstances also for paper.   The Theory works best for optically thick materials, those where >50 % of light is reflected and <20 % is transmitted.   Comprehensibly, the modern semi-transparent, transparent and very dark labels are exactly those where <50 % of the light is reflected and >20 % transmitted or absorbed.   These two German scientists however saw their theory was not completely rendering the diffused reflection of the: 

dyed papers,
very dark, unbleached papers,
when light absorption reaches a high level.   And one more constrain of their Theory is that the absorbing and scattering media have to be uniformly distributed through the label.  The Theory was not created for the kinds of labels we name “critical”:  

very dark and absorbing, nearly never used in 1931, during an epoch of human visual Label Presence inspection, 
plastic transparent self-adhesive labels today so common.
When the general-purpose photoscanners, today priced 4.8 $/each were conceived, over 40 years ago at the start of the Electro Optics industrial applications like the LEDs, Kubelka-Munk's Theory of Reflectance was correctly representing also all then existing paper labels.   Poliethilene and PVC transparent or semi-transparent self-adhesive labels, today so common and probably tomorrow the only kind to be used, started to enter the Beverage Bottling and Labelling decades later. Label inspection with Polarised light



















When they were developed, around 40 years ago, what are today considered by too many Vendors as “standard” photo-electric sensors to be devoted to label presence inspection, polarised light was yet a standard.   But a standard for Military and Scientific applications, and not in the Civil industrial applications like those of the Food and Beverage Industry.    Why polarised light ?     The answer lies in the solution of a key problem.   One affecting all imaging, included the human visual imaging, is that most sources of light are classified as incoherent and unpolarized (or only “partially polarized”) because they consist of a random mixture of waves having different spatial characteristics, frequencies, phases, and polarization states.    These superimpose themselves in the photo-detectors and the effect are phenomena of destructive interference.    As an example, imagine to illuminate a photo-detector with a huge amount of radian energy, i.e. 1 joule, later discovering that the Signal shall be later represented by only a small fraction of a non-negligible amount of energy which entered the detector.  Above a common plastic self-adhesive tape as seen by one and the same optic:















left side:    what a 4.8 $ “standard” photoscanner detects.  Its LED illuminates the label with circularly (360º) polarised light.   Also the reflected diffused light is not polarised.   The effect is that the plastic tape nearly disappears yet when over 1 million of pixels are used to detect it.
right side:    what a 260 $ modern high-Quality photoscanner detects.  Its LED illuminates the label with light polarised in just one specific angle.   Also the reflected diffused light reaches the phototransistor after passing through a polariser filter oriented the same angle as the LED emitted light.   The effect is that the plastic tape is distinctly visible.   In the meantime, the light blue background, equivalent to Noise in whatever Signal-to-Noise relation ratio, disappears.    A further contribution to the immediate and cheap reduction of the false rejects of whatever Label Inspector.









The Theory was not created for the kinds of labels we name “critical”:  

  • very dark and absorbing, nearly never used in 1931, during an epoch of human visual Label Presence inspection, 
  • plastic transparent self-adhesive labels today so common.
Photoscanner. Adding a stability LED to a circuit like this, results in a medium-quality general-purpose photoscanner.  Medium quality mainly because of the fact that at least this circuit includes a Schmitt’s Trigger output amplifier, with an hysteresis allowing to enhance the stability of the detector’s output.  Q1 is the phototransistor acting as Detector.  The most important component, being that accomplishing the transduction from the electromagnetic to the electric form of the energy.  RV1 is the trimmer used to manually adjust the Sensitivity threshold

When the general-purpose photoscanners, today priced 4.8 $/each were conceived, over 40 years ago at the start of the Electro Optics industrial applications like the LEDs, Kubelka-Munk's Theory of Reflectance was correctly representing also all then existing paper labels.   Poliethilene and PVC transparent or semi-transparent self-adhesive labels, today so common and probably tomorrow the only kind to be used, started to enter the Beverage Bottling and Labelling decades later.    


 Adding a stability LED to a circuit like this, results in a medium-quality general-purpose photoscanner.  Medium quality mainly because of the fact that at least this circuit includes a Schmitt’s Trigger output amplifier, with an hysteresis allowing to enhance the stability of the detector’s output.  Q1 is the phototransistor acting as Detector.  The most important component, being that accomplishing the transduction from the electromagnetic to the electric form of the energy.  RV1 is the trimmer used to manually adjust the Sensitivity threshold




Polarised Light Label Inspection










When they were developed, around 40 years ago, what are today considered by too many Vendors as “standard” photo-electric sensors to be devoted to label presence inspection, polarised light was yet a standard.  But a standard for Military and Scientific applications, and not in the Civil industrial applications like those of the Food and Beverage Industry.  Why polarised light?  The answer lies in the solution of a key problem. One affecting all imaging, included the human visual imaging, is that most sources of light are classified as incoherent and unpolarized (or only “partially polarized”) because they consist of a random mixture of waves having different spatial characteristics, frequencies, phases, and polarization states.  These superimpose themselves in the photo-detectors and the effect are phenomena of destructive interference.  As an example, imagine to illuminate a photo-detector with a huge amount of radian energy, i.e. 1 joule, later discovering that the Signal shall be later represented by only a small fraction of a non-negligible amount of energy which entered the detector.

Above a common plastic self-adhesive tape as seen by one and the same optic:

  Label Presence Inspection without (left) and with (right) photoscanners adopting polarised light.  At right side, the adoption of polarised light shows the impressive increase on the Signal-to-Noise relation for a common transparent self-adhesive plastic tape








  • left side:    what a 4.8 $ “standard” photoscanner detects.  Its LED illuminates the label with circularly (360º) polarised light.   Also the reflected diffused light is not polarised.   The effect is that the plastic tape nearly disappears yet when over 1 million of pixels are used to detect it.
  • right side:    what a 260 $ modern high-Quality photoscanner detects.  Its LED illuminates the label with light polarised in just one specific angle.   Also the reflected diffused light reaches the phototransistor after passing through a polariser filter oriented the same angle as the LED emitted light.   The effect is that the plastic tape is distinctly visible.   In the meantime, the light blue background, equivalent to Noise in whatever Signal-to-Noise relation ratio, disappears.    A further contribution to the immediate and cheap reduction of the false rejects of whatever Label Inspector.

 The impressive increase of the Signal-to-Noise ratio felt when circularly polarised light is emitted  through a polariser filter and later detected again filtered by a second polariser, depends by a non-classic quantum feature of the photon named “spin”.  In the example depicted here, shown left-handed clockwise circularly polarised light propagating along the z-axis.  Integrating several of these waves in a photo-detector let them interfere destructively.  Meaning less photons detected with the same spin angle, each one of them contributing to form the Signal


Flagging-Labels


An Electronic Inspector with photoscanners in-the-Labeller Machine may be usefully equipped with additional label inspections in the Conveyor, preventing flagging labels from reaching the Market.  More, Labeller’s brushes may remove labels yet applied and detected applied by the photosensors.  Thus, to have a final check out of the Labeller at-the-Conveyor, results cheap and advantageous. 

 






    

  Label inspection counters for an electronic inspector with inspections in-the-Labeller and redundants at-the-Conveyor


A different, extremely cheaper approach in the figure below.   Here, four additional independent label inspections check the wrap-around label position by four different places.   Photoscanners’ independence being the key to satisfactory performances, without extra-costs, operative complexity and expensive spare parts, always associated to camera-systems.


 Flagging label inspection, by mean of four independent inspections, adopting the cheapest general-purpose photo-scanners.  Their outfeeds take advantage of the Probability Multiplication Theorem.  Application on PET bottles labelled with semi-transparent labels







Such a design profits of the implications of the Probability Multiplication theorem. Applied  to an Electronic Inspector (a Binary Classifier) composed by: 

  • a set of N inspections, identified by the suffix-i, each one having:                             0 < detection ratio < 1;
  • whose individual rejects converge to a single Rejector; 

the probability that all the N inspections fail to detect a single defect, is reduced to:




Now, imagine to have a first inspection in-the-Labeller Machine, and four redundant successive of that same label at-the-Conveyor.   The probability P1,2,3,4,5  of joint occurrence of five events 1, 2, 3, 4, 5, any collection of outcomes of an experiment:

  • whose respective probabilities are P1P2P3, P4 and P5;
  • 0 < P< 1;
  • following normal (gaussian) distribution;
  • reciprocally independent;

is the product of their individual probabilities:


                            P1, 2, 3, 4, 5    =    P1  *  P2  *  P3  *  P*  P5


Finally, after considering that each one of the N inspections has:  


                                                                             0 < detection ratio < 1




it results immediately evident an impressive reduction on the amount of defective containers falsely considered “correct” (False Negatives), proceeding to the Market.  



Case Study


5 Binary Classifiers in a single Label Presence Inspector








As an example, we’ll consider the case with = 5 (five) inspections, all of them featuring the same individual False Positive (false reject) ratio FPi < 0.01 %.    Five independent Binary Classifiers, independently acting on a single rejector, whose individual:

  • accuracies; 
  • reproducibilities; 
  • specificities; 
  • sensitivities;

are such that the same property of an object is classified with Probability of False Negatives, say the undetected nor rejected but truly “defective” bottles:


                                                                  10.00 %


Defects’ Detection Ratio




The joint probability representing the Risk R that the five homogeneous systems (inspecting different projections of the same physical property, in different sample spaces of one and the same state space) simultaneously fail to recognise one and the same defect, shall be limited to:

                  R1, 2, 3, 4, 5    =    Πi=1 Ri   =   0.001 %

  Meaning:




















  • that no more than 1 (one) non-labelled bottle in a row of 100000 non-labelled consecutive bottles out feeding by the Labeller Machine, shall not be detected defective by the Electronic Inspector, thus passing to the Market,
  • detection ratio >99.998 %. 


False Rejects Ratio

We can look also on the other side, to the expected average value of the False Positives (false rejects) deriving by the superposition of the random variables measured by a Binary Classifier.    We have immediately different possible approaches, corresponding to our own prudential choice.   In this context, the term “prudential” is related to the level of confidence we attribute to the classification enacted by each individual Binary Classifiers.   For the value expected by the sum (or, superposition) of N independent random variables, following the confidence we attribute to the outcomes of our measurements, we’ll have at least three qualitative kinds of estimation:

  1.  most prudential 
  2.  prudential 
  3.  less prudential.


  1. Imagine that we’d decide to adopt the most prudential definition of sum of N independent random variables, an Electronic Inspector (a Binary Classifier) composed by a set of N inspections whose individual rejects Ri converge to a single Rejector (or, ejector), each one identified by the suffix-i.     In this case, the probability that each one of the N homogeneous inspections simultaneously and erroneously classify “defective” (False Positive) their physical measurements of a container  has an upper bound which is their sum: 

          and, for i = 1, 2, …, 5 we’ll have:

                                   FPexpected   0.05 %    





  1. Adopting the prudential definition of superposition of N independent random variables, we’ll have a correspondingly best estimation of the expectation value for the total False Positives (later false rejects) of the entire system of Binary Classifiers. Concident with the euclidean norm:

          which let us estimate the Total False Positives ratio as:

                                 FPexpected   =  0.022 %  





  1. Adopting the less prudential definition of superposition of N independent random variables, we’ll have a correspondingly estimation of the expectation value for the total False Positives (false rejects) of the entire system of Binary Classifiers, as:

         implying an estimation of the Total False Positives ratio:

                                 FPexpected   =  0.01 %  


Conclusions











Well before to mislabel 100,000 consecutive bottles, the Rejects Accumulation Table of the Electronic Inspector, should be fully occupied by rejects. The sensor of the Conveyor Control Cabinet managing this part of the Beverage Bottling Line, set in the Rejects Accumulation Table should control an automatic stop of the Labeller Machine. The straightforward application of the frequentist definition of Probability, silently assuming an infinite capability to accumulate rejected bottles, is countered by a reality where the:

  1. limited accumulation space, typically (80 - 300) standstill bottles, on the Rejects' Accumulation Table,
  2. existence of a sensor to detect that status and controlling Labeller Machine stop.

Then, it is better to get out of that classic definition of Probability and modernise our idea of expectation of the future Events to the relatively modern Bayesian viewpoint.   The expected Risk that a single non-labelled bottle pass to the Market, when consecutively inspected by five independent gaussian measurement systems, results still much lower than 0.001 %.    


Non-adherent Seals Check





Some of the products include labels adherent to the top of the closures, whose function is to evidence that the closure has not been opened.   Not all of the seals are correctly applied and, due to the particular geometry, uneasy to identify in-the-Machine. To evidence also slightly non-adherent seals, it is possible to increase their visibility by mean of compressed air.  The air blow forces the seals to rise enough to enter in the photo scanner light beam.



Locating and inspecting 


in-the-Labeller Machine















Typically, the Label Presence Inspectors having synchronisation sensors and inspections in-the-Labeller Machine, have their main cabinet installed immediately at the outfeed of the Labeller machine.   There is an excellent reason for this positioning: to limit the amount of non-labelled bottles each one time there is a serial fault created by one or more of the Labeller’s labelling stations.   The Electronic Inspector tracks the defective non-labelled container mixed in the flow of the others, by mean of a serie of inductive or optic sensor set in the Labeller Machine's carousel and at-the-Conveyor outfeeding the Labeller’s outfeed star wheel.    In-the-Labeller Machine the role of basic reference for all distances, there measured in units of Labeller’s Pitch, is played by a virtual Trigger.    It is obtained by the falling edge of the Machine Cycle sensor, all times a Container Presence sensor signal is associated, in-the-Labeller Machine until the Transfer Point.    As a matter of fact “How many Labeller Machine steps” is a distance value which can be used also out of the Labeller.    Be used  in the initial meters (however <2.0 m) of the outfeeding Conveyor, if this is straight and does not present any cause for bottles’ sliding.   One of such parameters exists for each one inspection.


Labeller Synchronisation sensor

In the Labeller it exists a synchronization sensor indicating the rotary plate no. 1, named Labeller Synchronisation sensor.  


 

  Machine Cycle (green) and Container Presence sensor signals (orange), as seen by an Electronic Inspector in a Labeller running at 60000 bottles-per hour.  Visibly, the Containrer Presence sensor signal is a short one.  A Container shall be considered “Present” when the signal’s duration (low phase, zero volt, equivalent to a logic “ 1” in this NPN logic) shall result at least 2 ms long.  The scale of the image is expressed in millimetres, because of the presence of an Encoder.  Then, they are millimeters of Distance equivalent to seconds of Time 



Machine Cycle sensor

























Then a Machine Cycle sensor, providing a pulse each single pitch of rotation of the Labeller’s carousel.   Filler Synchronisation, Machine Cycle and Container Presence are a reference for all the Labeller, as seen by the viewpoint of the electronic inspector.    


Container Presence sensor

A third one, named Container Presence sensor, is a true inspection and not a synchronization sensor.   It detects the mere presence of a bottle entered in the Inspector’s Shifting-Register.   Machine Cycle and Labeller Synchronisation are signals always surely infeeding to the Inspector as a consequence of the Labeller's revolution.   But, a Labeller can rotate also empty, and that’s why a Container Presence sensor is necessary.   Container Presence sensor is vital to discriminate two cases (or, states) otherwise impossible to separate.   Alternative cases when no label has been detected applied, and:  

  1. a bottle is present;
  2. no bottle is present.


Transfer Point

The number of steps from the inspection to the Transfer Point is programmed.    On the Transfer Point, it terminates the Machine Cycle and it starts the shifting-register area: as yet seen, there lies a virtual trigger.   Transfer point is the point where the Labeler (or Filler) outfeed starwheel starts to leave the container on the outfeeding Conveyor, part of the inspector shifting-register.    More precisely, it corresponds to the vertical projection, onto the conveyor surface, of the center of a container lying in a position a few millimeters before it starts to be released to the outfeeding conveyor.   How-many millimeters are “a few” ?     

An amount x:                                     x < 2      

preventing the otherwise progressively increasing False Triggers, accumulated due to wearing of the Labeller Machine starwheel vans and of its axial fixing and bearings.  Four typical label inspections in the Labeler carousel are represented below as blue-, violet-, red-, light blue-coloured signals, respectively corresponding to Back-, Body-, Neck- and Foil-inspections:




 Back label signals, referred to the falling edge of the Machine Cycle sensor, adopted as point zero to measure distances along each one individual step. The multiple blue colour signals indicate the superposition of several negatively affecting causes: oscillations, nonlinearities, reflections, insufficient or excessive retro-diffusion by the label, labels too dark or too clear, inspection during bottle rotation, etc. 





  Body label signals, referred to the falling edge of the Machine Cycle sensor, adopted as point zero to measure distances along each one individual step.   Visibly, Body label signal has a low phase (Signal “1” in this NPN logic) asting too long.  Its falling edge erroneously starting before, rather than after, Machine Cycle sensor falling edge.  Thus indicating superposition of several negatively affecting causes: excessive gain of the Photoscanner, Photoscanner’s nonlinearity, excessive retro-diffusion by the label, labels too clear, etc. 






 Neck label signals, referred to the falling edge of the Machine Cycle sensor, adopted as point zero to measure distances along each individual step










  Foil label signals, referred to the falling edge of the Machine Cycle sensor, adopted as point zero to measure distances along each individual step






In this Labeller they appear erroneously installed white-coloured brushes.  Erroneously, because set after the area devoted to Foil Presence inspection.  Then, as visible, the brushes were removing lots of aluminium foils glued with not enough glue.  Defective bottles undetected and terminating to the Market. Problems solved by mean of an uncommon modification of the Electronic Inspector original design.  We moved the Foil Presence inspection photo scanner to Labeller's outfeed star wheel, then after the brushes



 In this Labeller they appear erroneously installed white-coloured brushes.  Erroneously, because set after the area devoted to Foil Presence inspection.  Then, as visible, the same brushes were removing lots of aluminium foils glued with not enough glue.  Defective bottles undetected and terminating to the Market.  Problems solved by mean of an uncommon modification of the Electronic Inspector original design.  We moved the Foil Presence inspection photo scanner to Labeller's outfeed star wheel, then after the brushes



Tracking in-the-Labeller & at-the-Conveyor 


Why Containers' Triggering is a vital subject





“It makes no sense to speak of the “defective status of a container” in presence of ambiguity about “what container” is defective”











“Pharmaceutical Industry, ...with few exceptions discards as dangerous the rejection over an external Conveyor of the containers detected “defective” in-the-Machine… both detection and rejection have to happen in-the-Machine





To let the Reader of these notes fully understand why all of the Triggering subject is developed in so many pages in this web site, there is to remark that:

  1. the identity of an object is its most fundamental information.   It makes no sense to speak of the “defective status of a container” in presence of ambiguity about “what container” is defective.   In other words, containers’ Tracking, is much more important than whatever inspection, say whatever measurement of one of the physical properties of these objects;
  2. when comparing Food and Beverage Packaging Industry with the Pharmaceutical Packaging Industry, we discover that the former with few exceptions discards as dangerous the rejection over an external Conveyor of the containers detected “defective” in-the-Machine.    Following Pharma safety standards and customs, detection and rejection have to happen both in-the-Machine.  The rejection has not to happen out of its Shifting-Register, one truly satisfying the requisites of rigidity and impossibility of the containers’ sliding out of the cells them attributed.   Pharma Industry refrains to operate like Food and Beverage Packaging Industry operates following the OEMs and Vendors’ design choices. 


Sliding Conditions

One time a bottle is released by the Labeller out feed starwheel to the outfeeding Conveyor, all of the information regarding the bottle contained in its accompanying data sheet, is received by one or more Tracking Triggers.   Just one Trigger, in this case named Trigger 1, if the Electronic Inspector main cabinet and related Rejector are close to the Labeller Machine out feed starwheel's Transfer Point.   How much close ?     No more than 2.0 m from the Transfer Point.   A distance which could be 1.0 m longer, if the fastest container format expected to be treated is a relatively slow one, e.g. <0.8 m/s.   At least six conditions make a container prone to slide, during its release by the Machine starwheel and along its journey through the Conveyor.   Each one of them is sufficient to let a container slide that much (>1/2 container diameter) which is enough to clear its identity in the Inspector Shifting-Register:

  1. distances >2.0 m,
  2. curves between the Inspector main cabinet and the Rejector, 
  3. speed is >0.8 m/s, 
  4. light containers (<0.8 kg),
  5. equipments (e.g., air blowers, twists, etc.) whose action perturbates containers’ motion, exists along the Conveyor, interposed between Machine’s Transfer Point and following tracking Trigger,
  6. Conveyors’ cross-overs,

If any or more of these reasons conditions the installation’s layout, a serie of Triggers becomes vital to assure the conservation of the original data written in the container’s accompanying data sheet.    


Machine-Conveyors-Inspector Synchronisation

















Whatever the case, then also for relatively slow speed Labellers (<1.0 m/s), it is vital to synchronise the ramp-up and ramp-down phases of the Labeller and of the out feeding Conveyor, so that their speed difference results always <0.1 %.    0.1 % is a value knowingly close to the nonlinearity limit of the commercial Frequency Converters.  Say the equipments regulating the speed of the motors in the Labeller and in the Conveyance systems.    As an example, 0.1 % of divergence between bottle in-the-Labeller and same bottle at-the-Conveyor, at a speed of 1.5 m/s are 1.5 mm.    1.5 mm may appear a small divergence.    But it is the maximum which can be allowed to be, keeping apart the today rare cases where containers’ speed is relatively slow (<0.8 m/s).     Containers’ speed, being related to the containers’ kinetic energy following a quadratic law, enhances the slidings’ occurrences and then their negative effects.   


False Triggers and Data Sheets

The Trigger controlling the Rejector, say the last one immediately before the Rejector, shall be exposed to count False Triggers, corresponding to bottles whose identity is unknown because lost.   False Triggers which, in your own full interest, have to be absolutely be rejected by the Inspector, also if part of them corresponds to correctly labelled bottles.  How many of them has to be expected corresponding to correctly labelled bottles, then False Rejects ?    A quite precise answer is given by the Inspector Counter menu for False Triggers, visible below as 0.01 %.    Exactly that shall be the portion of the total False Rejects due to Falsely Triggered bottles.     The mere existence of the Container Presence sensor shall force this rejection.    Do not forget what seen above about the Transfer Point of the Labeller Machine's out feed starwheel.    

  False Triggers Reject Definition.  Similar menus for the Tracking Triggers exist whatever be the Vendor of the electronic inspector, however with differences in the colours, logos and motives.   The icon indicating a deviation out of the Production flow (a rejection) has to look like in the red box “enabled”.   This way, the falsely triggered bottles shall be rejected.  In a Label Presence Inspector in-the-Labeller, whatever different setting you could encounter is a deliberate decision to send to the Market also bottles detected defective because not labelled.  Have you requested or authorised this ?     











We refer to the necessary synchronisation of the Transfer (or, release) Point better than possible and with a positional error <2 mm.    A difference between Labeller and of the out feeding Conveyor >10 mm, say a divergence >0.67 %, implies well visible false rejects caused by falsely Triggered bottles.    The only case where it’s admitted a divergence > 0.1 %, is during the Labeller’s emergency stops.    Special and not frequent cases when to reject some correctly labelled bottles is the best action we can take.  Containers’ tracking starts under the control of a virtual Trigger based in-the-Labeller Machine and has to continue over the following Conveyors, until after the Rejector and the Reject Verification Trigger set immediately after the Rejector.    A problem today widespread is that the relative porcentage of the Electronic Inspectors in-the-Machine (Label, Fill level or Cap Inspectors) really operating like in their Order and their Design, is reduced to minimal porcentages, never seen along past two decades.  A synchronisation respectful of the contractual superior limit given to the False Rejects (False Positives) and to the defective and not rejected containers (True Positives), whatever their nature, implies a limit also on the False Triggers.   



Oscillographic Testing

What explained before is a malfunction commonly known to the Electronics Engineers engaged with high frequency measurements of digital electronic cards, their setup and Quality Control.  They use on a daily base oscilloscopes like the one visible on side.  By mean of these instruments they are also controlled eventual malfunctions of the digital circuits.  In these circuits are treated nearly rectangular waveforms subject to well-defined phase relations.  An erroneous phase relation can cause whatever thinkable negative effect, depending on the kind of technical, medical, aerospatial, scientific or military application.  The romboidal grey coloured area in the centre of the yellow oscillograms visible on side is meant as an off-limit zone.  No yellow colour signal has never to pass through it, because it’d mean a bold violation of the phase-relation between rigorously time-ordered actions.  

   The romboidal grey coloured area in the centre of the yellow oscillograms visible on side is meant as an off-limits  zone.  No yellow colour signal has never to pass through it, because it’d mean a bold violation of the phase-relation between different time-ordered actions

Exactly what a False Trigger Signal typically is.  If you have availability of an oscilloscope with memory, you are in the best conditions to verify on your own counting them one-by-one, the reality of the situation.   It may be considered acceptable that <0.5 %, one bottle each two hundred, is referred to a Trigger Signal invading the grey-coloured off-limit zone.    In some cases, when auditing systems allegedely commissioned, we discovered that over 99.9% of the bottles was in the reality not-synchronised at all.   Meaning that 70 % of the total price paid for those equipments was, in the reality, not operating.   Do not accept any “statistically significative only” pseudo-explanation for what is known us Electronic Engineer to be a rough evidence of erroneous setup or malfunction.   



False Triggering Causes















Falsely Triggered container, is a PET- or glass-bottle, a can, keg, crate or case, which lost its “identity” due to several causes and between them:

  1. outfeed too early or too late by the precedent Labeller-, Filler-, Capper-, Closer-, Seamer-Machine:
    1.  typically indicating a commissioning error in the synchronization Machine-Inspector, 
    2.  rarely due to a hardware fault of one of the detectors,
  2. slided forward, anticipating its arrival front of the following Trigger,
  3. slided backward, posticipating its arrival front of the following Trigger,
  4. manually 
    1. introduced in the flow of containers,
    2. removed out of the flow of containers,
  5. arrived so inclined front of the following Trigger, that it terminated to occupy a precedent or following cell in the Electronic Inspector’s Shifting-Register,
  6. flagging label or foil, emulating a container neck arrival front of the following Trigger,
  7. defective Tamper Evident Ring of plastic caps, emulating a container neck arrival front of the following Trigger,
  8. broken splinters, whose passage is registered front of a Trigger.

“Identity” originally attributed in the data sheet, later accompanying each one of them along their journey through the Electronic Inspector’s “Machines” and “Shifting-Registers”.   



Expensive In-the-Machine Inspectors…, 


(silently) downgraded to Cheap Standalone


  Do  you know how is it possible to have the behaviour described by the figures above ?  At right side, the same identical bug as left side, whose size is 60 times smaller.  Label, Container Presence, Torque, Bottle Burst or Asepticity inspections, when accomplished in-the-Machines require special Commissioning cares.  If the good-practices and rules are not observed, you’ll see a strange behaviour.  Small defects detected and rejected, the very big too detected but strangely and visibly not rejected.  In the meantime, …little or nothing of false rejects.  Kind of operation hinting to an illusory in-the-Machine electronic inspector whose Shifting-Register is losing containers 







Actions to hide never commissioned synchronisation of Inspector, Labeller and Conveyor, also implying a permanent impossibility for the Bottler to fulfill the triadic quest for: 

  1. minimum losses, 
  2. minimum  downtimes, 
  3. maximum rejection of defects and safety”

















“...service technician’s skills are part of the final added-value of the Electronic Inspector, because he is the one who shape the future of the equipment.   

World’s most expensive and higher Quality Electronic Inspector, left in the hands of some of the worst technicians, always perform much worse than World’s cheaper and lower Quality Electronic Inspector, left in the hands of some of the average-skilled technicians of  the World”

The Electronic Inspectors, whatever their Vendor, are built in two alternative configurations:

  1. standalone,
  2. in-the-Machine.


Standalone configuration

Standalone configuration is by far the cheapest.  2-5 times cheaper than the in-the Machine.   It does not include any connection, sensors, inspections, detectors nor logic relation with a precedent Labeller-, Filler-, Closer- or Seamer-Machine.   Its: 

  • tasks are limited to handle a single Shifting-Register, rarely a couple of them;
  • hardware is reduced to the minimum, typically one Trigger sensor over which the unique Shifting-Register is logically built, inspections and a rejector.       
  • fitting, wiring, starting and commissioning are accomplished in a limited time and, not being particularly complex tasks, they do not require the typically few available top-skilled service technicians.   

In systhesis: when configured to operate Standalone, all Inspectors imply an installation task that all of the Optoelectronics' service technicians fulfil.    


In-the-Machine configuration

The Electronic Inspectors with locating positions and/or inspections in-the-Machine (Labeller-, Filler, Closer-, Seamer-Machine) are complex and their:   

  • tasks extended to handle several Shifting-Registers, typically two minimum;
  • hardware is maximised, several Trigger sensors (inductive and optic) over which the various Shifting-Registers are logically built, many inspections and frequently more than one Rejector.       
  • their complexity is reflected in their pricing, always a factor 2-5 more expensive than the corresponding models operating standalone
  • years of practice, trainings, deep and true comprehension of the Shifting-Register's concept, are pre-requisite to an accomplished Commissioning.   Commissioning honouring the Contract between an OEM or Vendor and a Food and Beverage Bottling and Packaging Company, also respectful of a row of legal requirements imposing, as an example, that the Machine Safeties cannot be left systematically disabled to the unaware Customers.   


Human-factor does matter

Commissionings can only be named and underwritten that way in a worksheet whose siognature is asked you when: 

  • honouring the Contract between an OEM or Vendor and a Food and Beverage Bottling and Packaging Company,
  • respectful of a row of legal requirements imposing, as an example, that the Machine Safeties cannot be left systematically disabled to the unaware Customers. 

We are suggesting the Readers that the service technician’s skills are part of the final added-value of the Electronic Inspector, because he is the one who shape the future of the equipment.   Unavoidably, World’s most expensive and higher Quality Electronic Inspector, if left in the hands of some of the worst technicians, always performs much worse than World’s cheaper and lower Quality Electronic Inspector, when left in the hands of some of the average-skilled technicians of  the World.    


Standalone (left) versus In-the-Machine (right) configuration.  At left side depicted two Standalone-configured Full Container Inspectors, each of them processing ~46000 cans-per-hour.  Their Binary Classification task reduced to a single deviation of the red-coloured Rejects out of the green-coloured Production flow.  All Containers entering each Inspector and going to be attributed an identity in the unique Shifting-Register by the unique Trigger, here named and orange-coloured as Production.  At right side, the In-the-Machine inspection system receives 4 additional digital inputs by 2 Machines, here a Filler and a Seamer, which in general may also be Capper, Closer, Labeller or Blowformer.  Also an additional outside blue-coloured for the Advanced Sampling.   A much deeper comparison of the two configurations, one only known to the Vendors, regards the installation (mechanical, fitting, wiring, start-up, commissioning, training) timing.   However difficult may this result to be imagined, the in-the-Machine configuration at right side needs >16 man*days when the Standalone at left side, needs <4 man*days.   Meaning that the complexity underlying the In-the-Machine delivery to the Bottler is >4 times bigger than for a relatively easy Standalone inspector.   Figures representing the full can inspection systems assuring the Quality of a 92000 cans-per-hour Canning LineStandalone (left) versus In-the-Machine (right) configuration.  At left side depicted two Standalone-configured Full Container Inspectors, each of them processing ~46000 cans-per-hour.  Their Binary Classification task reduced to a single deviation of the red-coloured Rejects out of the green-coloured Production flow.  All Containers entering each Inspector and going to be attributed an identity in the unique Shifting-Register by the unique Trigger, here named and orange-coloured as Production.  At right side, the In-the-Machine inspection system receives 4 additional digital inputs by 2 Machines, here a Filler and a Seamer, which in general may also be Capper, Closer, Labeller or Blowformer.  Also an additional outside blue-coloured for the Advanced Sampling.   A much deeper comparison of the two configurations, one only known to the Vendors, regards the installation (mechanical, fitting, wiring, start-up, commissioning, training) timing.   However difficult may this result to be imagined, the in-the-Machine configuration at right side needs >16 man*days when the Standalone at left side, needs <4 man*days.   Meaning that the complexity underlying the In-the-Machine delivery to the Bottler is >4 times bigger than for a relatively easy Standalone inspector.   Figures representing the full can inspection systems assuring the Quality of a 92000 cans-per-hour Canning Line





  Standalone (left) versus In-the-Machine (right) configuration.  At left side depicted two Standalone-configured Full Container Inspectors, each of them processing ~46000 cans-per-hour.  Their Binary Classification task reduced to a single deviation of the red-coloured Rejects out of the green-coloured Production flow.  All Containers entering each Inspector and going to be attributed an identity in the unique Shifting-Register by the unique Trigger, here named and orange-coloured as Production.  At right side, the In-the-Machine inspection system receives 4 additional digital inputs by 2 Machines, here a Filler and a Seamer, which in general may also be Capper, Closer, Labeller or Blowformer.  Also an additional outside blue-coloured for the Advanced Sampling.   A much deeper comparison of the two configurations, one only known to the Vendors, regards the installation (mechanical, fitting, wiring, start-up, commissioning, training) timing.   However difficult may this result to be imagined, the in-the-Machine configuration at right side needs >16 man*days when the Standalone at left side, needs <4 man*days.   Meaning that the complexity underlying the In-the-Machine delivery to the Bottler is >4 times bigger than for a relatively easy Standalone inspector.   Figures representing the full can inspection systems assuring the Quality of a 92000 cans-per-hour Canning Line


Standalone versus In-The-Machine 

Commissioning Pricing



















“…possibility that the in-the-Machine Electronic Inspector you acquired is in-the-Machine only in its expensive mechanics, hardware and firmware.    

But that later was not Commissioned in-the-Machine rather Standalone”










How to predict what should be the pricing of an Electronic Inspector’s  installation, say its  fitting, wiring, start-up, Commissioning and Operators’ Training?   Examing a posteriori many hundredths of different installations of Electronic Inspectors in Beverage Bottling Lines:

  • where the above named installation activities had really been made, and made the only way the laws let them be valid in your country, because conformal to Safety rules and other technical good practices it is possible to see that a relation really exist,
  • in different countries,
  • part of Projects of Bottlers in competition (example, Coca-Cola vs. PepsiCo, AB InBev and SABMiller), whose Quality standards are different,
  • configured with all of the existing spectrum of inspections,
  • processing all existing kinds of containers (glass-, PET-, PRB-bottles, metal cans, crates and cases, etc.),
  • from a wide range of productions, from 8000 til 92000 containers-per-hour,

regression analysis shows an approximate but straightforward relation deriving be the comparison of the prices of the equipments, when the installation activities are not included. The installation pricing factor has empiric origin, nor we’d imagine what another way could be imagined to observe tendential relations which can only be obtained by linear regression studies a posteriori. As an example, if a Full Bottle Inspector costs 8000 € , ~10000 $ (weighs net 18 kg) is mainly because that amount is proportioned to the sum of the values of its mechanical, pneumatical, optoelectronics and documentation components. Another kind of Full Bottle Inspector of that same Vendor, priced ten times more, i.e., 80000 €, ~100000 $ (weighs 138 kg) shall need installation activities extended approximately to a duration >4 times longer than 8000 € model, needing just 3 days.    Superior models cost ten times more than the basic because they sport many more functions and optionals to: 

  • install (Rejectors, Triggers, cables, sensors, etc.), 
  • fit, 
  • wire, 
  • start, 
  • commission, 
  • train the staff.

Basic questions submitted to you own critical thinking:

  1. an Electronic Inspector needing >12 man*day of activities in your site, has been (strangely) delivered you after just 5 man*day ?    
  2. What let you feel comfortable with the idea that its configuration is what you paid for ?      What techniques you used to test it ?   
  3. Maybe some test bottles had been passed one hundred times before to decide ?   If this is the case, you have to know this is the kind of inspection a cheap Standalone Electronic Inspector does.   An expensive in-the-Labeller Machine is paid to do many more additional useful things (like, advanced sampling or inspection in-the-Labeller) than just inspection at-the-Conveyor where the test bottles have been introduced.
  4. Would you order a new car without to be informed before about its configuration, motor and performances ?    We don’t think so, then it makes sense the following question:
  5. Have you received a precise listing of the functions assured you (Electronic Inspector’s Technical Guarantees) by the OEM or Vendor ?     If not, why not ?



In-the-Machine Inspectors reduced to Standalone

trigger1-copy-2_medtrigger2_med

Containers’ Tracking Triggers 1, 2, 3, 4, 5.   [Similar menus for the Tracking Triggers exist whatever the Vendor of the Electronic Inspector, however with differences in their motives, colours or logos].   

Setup of “additional containers not accepted” and “loss of containers not accepted” has mandatorily to be identical to that in a red colour box.  Be wary of setups you may discover presenting “additional containers accepted” and/or “loss of containers accepted”.    In a Label Presence Inspector in-the-Labeller, they'd correspond to a deliberate decision to send to the Market also bottles yet detected defective because not labelled.  Have you ever requested or authorised in a written way to leave disabled Machine Safeties and/or to counter Vendor's Technical Dept. written design recommendations like these ?    


trigger3_med

Food and Beverage Companies managers’ triadic quest for: 

  1. minimum losses, 
  2. minimum  downtimes, 
  3. maximum rejection of defects and safety. 
trigger4-2_med

implies that all Quality Control equipments have to be at least as performant as they were declared by the respective Vendors in the Technical Guarantees part of that Offer which had been accepted by the Bottler. But, the Electronic Inspectors’ Shifting-Register is the most complex kind thinkable.  It joins rigid and banal Shifting-Registers, like those of the rotary machines (Filler, Labellers, Closers, Cappers, Seamers) to the anelastic deformable Shifting-Registers (Conveyor’s belts) where containers can freely move themselves, rather than be firmly kept blocked in the Shifting-Register’s cells.   

Have you seen macroscopic defects passing through the Inspector, not detected ?    

We refer ourselves to bottles:

trigger5-2_med
  • semi-empty,
  • without closure,
  • without all of the labels.

Missing rejections made strange by the fact that, simultaneously extremely smaller defects are correctly detected and rejected, as an example:

  • 2 mm underfilled bottles,
  • closure inclined or too high for a few millimetres,
  • bottles rejected because one of the labels is misplaced (but present) for 1-3 mm.




















Commissioning Good Practices and Rules

Each one of the cases depicted before hint to a parameterisation (or, setup) violating basic rules of Commissioning of the Inspector’s Shifting-Register and of the positions of the Inspector’s detectors in-the-Machines.    

Written rules diffused by Vendors’ Technical Departments.   Rules clearly defining what is the most basic “inspection” in whatever Electronic Inspector: the constant control and cross-check of each container’s identity in the Shifting-Register.   If the identity, the identification of an object is ambigous, how and why could we consider it non-defective and ship it to a Customer ?    This can be obtained forcing determinate “bits” to do what Vendors’ Technical Dept. and rationality do not want absolutely be done, nor should ever be done when owning the necessary know-how. 



Small defects correctly rejected and some extremely 

big defects are not ?


Foreign object 60 times smaller in a bottle base. Have you repeatedly observed that an extremely small defect detected and rejected and that, on the opposite, extremely bigger defects also detected but were later strangely not rejected ?   And, simultaneously, no false reject of a correct bottle happens ?   Then, it is highly probable that the expensive in-the-Machine (Labeller, Filler, Closer, Capper, Seamer) electronic inspector you acquired, had been parameterised forcing it to operate as a cheap Standalone model

If the scenario described above in bold letters corresponds to what is being observed, then the problem affecting the Quality of your Company does not derive by missing detection, because of inspections’ sensitivity not adequate to the defects.    

Banally, a bottle nearly always rejected because:

  • its plastic cap is 1 mm too high, 
  • one of the labels is 2 mm misplaced, 
  • 1 mm underfilled, 

implies that completely missing caps or labels and under- or over-filling are detected well over 3-standard deviations.   Then, they’d had to be rejected if…, if the Inspector's Shifting-Register has not undue programmed parameterisations forcing that outcome when the container loose its identity.   If cases like these have yet been and continue to be observed in your Beverage Bottling Line, then there is a high possibility that the in-the-Machine Electronic Inspector you acquired is in-the-Machine only in its expensive mechanics, hardware and firmware.  

But that later was not Commissioned in-the-Machine.  Rather, as it’d be a much cheaper model operating Standalone at-the-Conveyor.     


 











Defect copy. Have you repeatedly observed that an extremely small defect detected and rejected and that, on the opposite, extremely bigger defects also detected but were later strangely not rejected ?   And, simultaneously, no false reject of a correct bottle happens ?   Then, it is highly probable that the expensive in-the-Machine (Labeller, Filler, Closer, Capper, Seamer) electronic inspector you acquired, had been parameterised forcing it to operate as a cheap Standalone model

   Have you repeatedly observed that an extremely small defect detected and rejected and that, on the opposite, extremely bigger defects also detected but were later strangely not rejected ?   And, simultaneously, no false reject of a correct bottle happens ?   Then, it is highly probable that the expensive in-the-Machine (Labeller, Filler, Closer, Capper, Seamer) electronic inspector you acquired, had been parameterised forcing it to operate as a cheap Standalone model


We have records of too many Inspectors in-the-Machine only in their expensive hardware, encountered forced to operate Standalone and not in-the-Machine.


     To downgrade an expensive in-the-Machine (left-side) electronic inspector in a cheap and basic Standalone (left-side), is much easier than to commission the in-the-Machine.  Easy to do and difficult for the Bottler to perceive


Systematic Actions   

                   Actions in some cases, extended to the Inspector's hardware, so to let a green colour traffic light on keep calm the dubious Bottlers, switching artificially off the red colour traffic light that the Electronic Inspector’s inner diagnostics was in the meantime trying to show

~70 of the past >100 electronic inspectors in-the-Machine in their hardware and firmware, have been and are presently being discovered and recorded on the opposite operating Standalone, because of a well defined pattern of actions:

  • targeting the False Triggers Reject Definition, a particularly important Machine Safety of the Electronic Inspector, so not to reject the containers without identity, including those defective; 
  • targeting the operational definition of the Triggers (1, 2, 3, 4, 5).   All containers “lost” or “additional” forced to be considered “correct”.   The opposite than what an expensive Electronic Inspector in-the-Machine has to do;
  • targeting the Reject Verification Trigger, leaving it disabled (e.g., marked like Trigger 5 in the menus above at right side).   Reject Verification Trigger which is the most important Machine Safety;
  • aimed at hiding the never commissioned synchronisation of the system Inspector + Labeller + Conveyor, also implying a permanent impossibility for the Bottler to fulfill his quest for: 
    1. minimum losses, 
    2. minimum downtimes, 
    3. maximum rejection of defects and safety.
  • extended to the Inspector's hardware in some cases, so to let a green colour traffic light keep calm the dubious Bottlers, forcing artificially off the red colour traffic light.  That permanent alarm status that the Electronic Inspector’s inner diagnostics should have showed if those jumps (or, bridges) should not have been made.

Above, at right side, a sequence of menus translating in a practical real case of Inspector in-the-Machine, how it looks like the correct sequence of Tracking Triggers.   It is a basic and fundamental set of instructions: there are additionals.








 Reject Verification Trigger is the most important Machine Safety.  Systematically targeted also the Reject Verification Trigger, leaving it disabled.  In the menus above at right side it appears marked as Trigger 5




































Take extreme care that the definitions evidenced in the red boxes have to appear: 

  • “additional containers not accepted”  
  • “loss of containers not accepted”.     

No other setup is conformal to the Technical Department of the Electronic Inspectors’ Vendors and for obvious reasons.   Different setups, like: 

  • “additional containers accepted”  
  • “lost containers accepted”

strictly and only mean that the bottles (or, cans) who have lost their own identity in the Inspector’s Shifting-Register, have to be considered “correct bottles”.   Apply this to an Electronic Inspector with Label Presence inspections in-the-Labeller Machine.    It means to send straight to the Market all bottles without labels, also without any one label, if these bottles have lost their identity.   To loose the identity, it is enough to slide one-half of the bottle diameter and/or, to be in the Shifting-Register of an erroneously commissioned Electronic Inspector.   What surely happen very frequently in an Inspector when misrespecting any of the recommendations listed above in the section: “Locating and inspecting in-the-Labeller” and others.


Root Causes

  1. Vendor’s Service Technician.   If the Service Technician is not capable to commission the in-the-Machine inspector existing at least in its hardware and firmware, several way-outs known us Experts remain open to let the Inspector “looks-like it is operating in-the-Machine” when in the reality it is not.  

         We recorded >70 of these occurrences along >100 in-the-Machine inspectors’.  

  1. Maintenance Dept. Staff.   Alternatively, the observed permanent anomaly could have origin in an erroneous reprogramming of a correct parameterisation, by someone part of the Bottling Maintenance Dept. Staff.   However someone owning an access level high enough to have the possibility to change these parameters. 

         We recorded <5 of these occurrences along >100 in-the-Machine inspectors.   

  1. Production Operators.   Just a possibility.  Never observed nor one case.
  2. Random, spontaneous changes of the parameterisation.   They are excluded.   These “errors” are not randomly affecting all of hudredths or sometimes thousands of parameters, rather always and only a few very well definite. If it’d be just bad luck, we’d be forced to admit that, after 2004 “mr. Bad Luck” uses very powerful glasses when targeting exclusively the in-the-Machine Electronic Inspectors to downgrade them to Standalone.   

A precise, factual and evidential attribution of the Root Cause of these undue parameterisations, can be easily obtained after controlling the historic memory of the Data Saving operations in each Electronic Inspector. Simply checking the setup of these decisive parameters on the date when Vendor passed to Bottling Company the equipment declaring it complete and corresponding to the Purchase Order. 

wine-101 med


Links to the other pages:



Contact





















This website has no affiliation with, endorsement, sponsorship, or support of Heuft Systemtechnik GmbH, MingJia Packaging Inspection Tech Co., Pressco Technology Inc., miho Inspektionsysteme GmbH, Krones AG, KHS GmbH, Bbull Technology, Industrial Dynamics Co., FT System srl, Cognex Co., ICS Inex Inspection Systems, Mettler-Toledo Inc., Logics & Controls srl, Symplex Vision Systems GmbH, Teledyne Dalsa Inc., Microscan Systems Inc., Andor Technology plc, Newton Research Labs Inc., Basler AG, Datalogic SpA, Sidel AG, Matrox Electronics Systems Ltd.


                                                                                                            Copyright Graphene Limited 2013-2019