Case study

99.5% of the non-labelled bottles 

sent to the Market

design relevance med hr-2

  Non labelled and flagging-labelled bottles, manually rejected at the outfeed of a Labeller Machine. Manually rather than automatically rejected by the inspection system, because of a design-error in its configuration, amount of tracking Triggers photocells and Encoders. These amounted to one-half of those necessary to assure bottles’ tracking in the conveyors out of the Labeller.  “Correct Tracking” related to the Technical Guarantees provided to the Beverage Bottler for the non-labelled and flagging labelled False Negatives (truly defective non-rejected bottles) 

The Technical Guarantees all Vendors provide to their Customers implies limits to their duties. What is into those limits is the area where, on the opposite, the Customer can exercises its rights.  Imagine an Electronic Inspector born configured with label presence photoscanners in-the-Labeller-Machine. Label presence sensing in-the-Labeller-Machine allows an early knowledge of the fact that one of the labelling stations ceased to accomplish its function. One of the most common kinds of temporary malfunctions which may affect whatever Labeller Machine.  Early knowledge means the possibility to react automatically:

  • stopping the Labeller, after a preset amount of consecutive defects is counted;
  • attract the attention of the Labeller Operator, showing the correct diagnostic alarm.

The Electronic Inspector was guaranteed providing:

  • 99.5 % detection and rejection of the non-labelled bottles;
  • total false rejects, whatever their cause, < 0.01 %.

The visible practical after the Labeller and Inspector startup is what visible in the figure above.  Each one time one of the labelling stations ceased to apply the labels, these continued unrejected and undetected toward the Market. Only manually separated, keeping a couple of Operators doing what that design was not capable to do.   In the present case, to exclude all thinkable alternative scenarios, we:

  1. cared ramp-up and ramp-down acceleration and deceleration phases in the Labeller machine and its following single-lane Conveyors.  Fine-adjusted until the Labeller speed had maximum deviation with respect to the outfeed Conveyor, equivalent to ~10 mm.  What for a 2.5 m/s fast, 60000 bottles-per-hour, Labeller Machine is not banal to do.  The issue ameliorated only in minimal extent, say that before 195 missing-label bottles of 200 were not rejected and now 180 of 200 missing-label bottles continued not being rejected, going to the Market.  We are speaking of an in-the-Machine design based over Shifting-Register;  
  2. levelled ~15 meters of conveyors, part of the Electronic Inspector’s Shifting-Register, until < 0.1º, reducing bottles’ sliding;
  3. straightened ~30 meters of conveyors' side guides, so to be vertical until < 0.1º, reducing bottles’ sliding;
  4. quadruplicated the amount of lubricant poured over the 15 meters of conveyors, to reduce the dispersion of the bottles' sliding distribution; 
  5. changed all pre-existing “polizene” plastic guides, assuring the stainless steel belts’ movement, with new models of higher quality and, most important, same identical thickness, reducing bottles’ sliding;
  6. fine-adjusted all laser Trigger photocells beams gain parameters, reducing systematic errors;
  7. cleaned all of the optics (photocells and mirrors) of the original Trigger tracking system, reducing systematic errors;
  8. fine-adjusted all tracking Trigger distance parameters until < 1 mm, reducing systematic errors;
  9. checked one-by-one the wiring, from each tracking Trigger photocell until the CPU, so to be sure it was exempt by false-contacts introducing systematic errors;
  10. checked one-by-one the wiring, from each tracking Trigger photocell until the CPU, so to be sure no wiring error could have been made, then introducing systematic errors.

Readers may imagine other factors which could create that macroscopically effect amounting to 99.5 % of the non-labelled bottles sent to the Market.  Most important is to understand that yet these ten independent “causes” listed above may also be simultaneously present, and amounts to total:

  10 !   =  10 * 9 *…..* 2 * 1   =   3 628 800  combinations

Amount which starts to let us feel the full meaning of the words of the physicists when describing a “superposition of a multitude of terms”.  We kept like eleventh cause what since the start was suggested by our own intuition. Readers may consider incoherent such strategy, one studying only in the last exactly what seems the most promising cause.  In the reality, it is nearly mandatory to act this way because:

  • our own intuition (the Root Cause Analyst intuition) may fail to recognise the correct tree-like structure, then attributing an excessive measure (aka, probability) to the “cause” intuitively conceived;
  • Vendors (and, OEMs) if not compelled by facts, are slow to recognise the mere possibility of Design error.   

Then, it was necessary to test all thinkable alternative scenarios.  The eleventh cause, the one which revealed itself having the greatest measure, is object of the following sections of these Case Study, where the Shifting-Registers basics are focused.

Focus on the Shifting-Register

A basic rule, valid for whatever Shifting-Register, says that each one cell can be occupied by a single element and that the element may slide on adiacent cells but only until less than one-half of the cell size.  Size which can be conveniently expressed in millimetres.  The Electronic Inspector’s inner diagnostic menus still showing deviations of the bottles’ positions until 300 mm.  The outfeed starwheel of the Labeller featuring 116 mm pitch, implying maximum deviation along all of the serie of following Conveyors limited to (116/2) mm, say 58 mm.   And that's why so expensive Electronic Inspector could not do its job: 

                              observed deviation  >>  limit deviation to allow tracking

in our case:                                     

                                               300 mm  >>  58 mm

What could be still causing such a generalised tracking error ?  The Shifting-Register did not terminated in the conveyor immediately out of the Labeller.  To slow down the bottles and to allow the Labeller's bypass (for another kind of container) the Electronic Inspector was on the third successive conveyor after the Labeller outfeed star wheel.  And, visibly, the entire system had a single Encoder, in the area where both the Inspector’s main cabinet and rejector were installed. Meaning two cross-overs without Encoders nor Trigger photocells.  

Shifting-Registers are physical 

Knowingly, galileian kinematics implicit in the Shifting-Registers requires exactly the opposite than “holes”. Areas where the elements (containers) have unmeasured and then non-univoquely defined speed.  Visibly, the Electronic Inspector had no problem to detect the missing labels but was later loosing their identity along the serie of conveyors.  Exactly the case where an element is ambigously attributed to several leaves (or, sheets) at once.  One time informed, with plenty of technical details and evidences, the Inspector's Vendor expressed its disbelief in the mere possibility that its Designers could have made such a design error.    But, a specific counter, named False Triggers, part of the Inspector cleared the reality: still 90 % of the containers' identities synchronized in the Labeller, were “lost” along the successive journey until the main Inspector cabinet.  Lost because containers arrived too early or too late front of the following Trigger.  Too early or too late, as seen from the point of view of the most basic Law of Kinematics.  “Too early” representing containers apparently too fast, and “too late” representing containers apparently too slow.  Typical outcome of a situation in which tracking is not assured by a flawed Triggering, where more than one container may correspond to a single cell of a Shifting-Register.    To get rid of this pernicious problem it was necessary to:

  1.  add two Encoders, to the conveyors where were missing;
  2.  to double the amount of tracking Trigger photocells, 

reducing this way the amount of False Triggers to < 0.05 %, allowing rejection of ~ 99.95 % of the defectively labelled bottles.  Root Cause Analysis revealed itself fundamental to convince the Electronic Inspector’s Vendor of the reality that the cause of the issue was internal and implicit effect of to its Design.  

Links to other Case Studies:

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