Presentation on theme: "THE BUSINESS CASE FOR IMPLEMENTING MACHINE VISION"— Presentation transcript:
1THE BUSINESS CASE FOR IMPLEMENTING MACHINE VISION
2Vision Systems International Established in 1984Consultancy concentrating on machine visionServices include:TrainingApplication related:Application engineeringSpecification writingVendor identification/evaluationMarket relatedMarket researchstrategic development and planningpartnering activitiesmarket analysis/competitive analysisdue diligenceTechnology transfer
3Introduction Electronic Imaging Where is Machine Vision Used Why Machine Vision NowMachine Vision Industry/MarketCompared to Human VisionWhy Consider Machine VisionApplicationsSystematic DeploymentWhat is Machine Vision
9Technology ReadinessUnderlying technology for machine vision has evolvedComponents developed with features required to succeed in machine vision applicationsLighting - LED - stable, long lifeCameras - solid state, progressive scan, asynchronous scan, exposure control, color, high resolutionOptics - telecentric, computer controlled zoomCompute power - PCs, DSPs, etc.Software - GUI - Windows - StandardPCI Interface, IEEE 1394
13Semiconductor Processing Market ArtworkReticle/photomask inspectionUnpatterned wafers - defectsPatterned wafers: critical dimensions, overlay registration, defectsIn early majority phase but due to changes in industry also in innovator and early adopter phases
14Semiconductor Packaging Market AlignmentOCR/OCVLead straightnessCo-planarityPackage markingsPackage cosmeticsIn late majority phase
15Electronic Market Bareboard inspection Populated PCB inspection solder pastepre-solderpost solder (optical and X-ray)In the case of bareboard systems - late majorityIn the case of populated PCB systems - early majority
16Other Market Properties that Affect Adoption GeographySegment of market: consumer electronics, telecommunications, computers, military, etc.Captive manufacturer or merchant/contract builder
17Automotive Market 3D sheet metal assembly spot welder fixture alignmentwheel alignmentpaint inspection2D - miscellaneousaftermarkethead light aimingbody alignmentMajor companies are in late majority phase; component suppliers in early majority phase
18Wood Market Yield optimizer Grade optimizer Late majority phase for yield optimizer; innovator phase for grade optimizer
19Printing Market Color registration Print registration Print inspection For first two in late majority phase; for third in early adopter phase
20Container Market Glass glassware manufacturer filler Can Plastic ClosureFor glass and can in late majority phase; for plastic in early adopter phase; for closure in early majority phase
21Pharmaceutical Market Process endvials, filled/unfilledsolid dosagesPackaging endlabel issuesIn process end in early adopter phase; in packaging in early/late majority phase
59Success Requires Senior management must foster atmosphere to encourage changesupport change agentsdemonstrate buy-in to changeencourage plant and line to take ownershipestablish realistic schedule for changes
60Success is more likelyPeople assigned are interested in new techniques and welcome changebegin with easy, non-critical applicationdefine the parameters of the project and avoid creeping expectationsselect applications not critical to labor issuesbe supportive during learning processplan for replications
61Success is more likely Obtain people involvement Avoid technology leap that is too farMake certain project is part of an overall plan
62Implementation Process Assemble task force and study production processtask force should develop understanding of what machine vision isdefine need and evaluate alternativesinvestigate - select specific applicationsassess technical feasibility and cost feasibilitywrite comprehensive specification
63Implementation Process Install and run-in.Conduct acceptance testProvide shop floor supportEvaluate system’s performance against goalsLook for another machine vision opportunity
64Implementation Process Solicit vendors with appropriate expertiseVisit vendors to review proposals, policies, expertise, QC proceduresSystematically select vendorPurchaseAcceptance test at vendorTrain all personnel Involved
65What is Machine Vision? As defined by the AIA: A system capable of acquiring one or more images, using an optical non-contact sensing device, capable of processing, analyzing and measuring various characteristics so decisions can be made.
70Relevance of Pixels Pixels 512 X 512 1/4M 1300 X 1200 1.4M AP Wire Photo M35 mm color film M
71Steps to Take When Buying a Machine Vision System
72Steps to Take When Buying a Machine Vision System Identifying Machine Vision OpportunitiesAssess Application FeasibilityUnderstand the ApplicationUnderstand the VendorsResponsive ProposalsSystematic Buy-off ProcedureMistakes in Buying AutomationProject Justification
74Identifying Machine Vision Opportunities Lowest value addedExpensive fixturingLengthy set up times100% inspection required to sort bad partsHazardous environmentContaminantsCapital expansionOperator limitations
75Profile of Good Machine Vision Opportunity Perceived valueCost justifiableRecurring concernCan do something about itStraight forwardTechnically feasible
76Profile of Good Machine Vision Opportunity User friendly potentialDedicated lineLong line lifeOperation championManagement commitment
77Global Competition Requires Higher manufacturing productivityIncreased demandHigher product qualityBetter customer serviceFlexible manufacturingGreater return on manufacturing assetsChanging standards of manufacturing performance
78Computer Aided Inspection Provides traceability - recordsStatistical data base - isolate production problemsReal time machine correction/adaptive controlAutomatic QC data collection and analysisRemove drudgery of humans
79Hidden Costs Machine Vision Can Help Lost business because product not produced on timeShipment of wrong productsExcess inventoryIdle labor because parts are not availableDoing a job overLoss of valuable information
80Machine Vision and Factory Automation Data driven automationMachine vision = data !
81Statistics Measurements Parts recognized Classification Types of defectsTrend analysisPerformance assessmentRecord keepingProcess Control
82Successful Application Requires Comprehensive understanding of needsProper application processGood equipment and performance specificationsComprehensive understanding of machine vision system capability
83Steps to Take When Buying a Machine Vision Machine Vision is in Widespread UseBest Justification is Process ControlInfrastructureResources: AIA and MVA
84How To Select Machine Vision Equipment Understand the technologyAssess application feasibilityUnderstand the applicationUnderstand the vendorsResponsive proposalsSystematic buy-off procedureApplications in pharmaceuticals
88Steps to Take When Buying a Machine Vision System Assess FeasibilityBasis rests with size of a pixel/FOVMVA slide ruleTypical system handles 500 pixelsFunction of generic application:verificationgaugingpart locationflaw detectionOCR/OCV/pattern recognition
89Verification Function of contrast - real or artificial high contrast - feature should cover3 X 3 pixel arealow contrast - feature should cover more pixels
90Gauging 500 marks on a ruler = resolution subpixel interpolation - factor of 4 to 10requirements driven by tolerancerules of thumb:repeatability: 1/10th of toleranceaccuracy: 1/10th to 1/20th of tolerancesum of accuracy + repeatability = 1/3 tolerance
91GaugingDiscrimination - smallest change in dimension detectable with measuring instrumentDiscrimination = sub-pixel resolutionRepeatability = +/- DiscriminationAccuracy - determined by measurement of calibration standard = Discrimination
92Gauging - Example 2” part and 2” FOV tolerance: +/ ”, total range 0.010”repeatability: 1/10 X 0.010” or 0.001”discrimination/accuracy: 1/20 X 0.010” or ”with 500 X 500 pixel camera, resolution = 0.004”with sub-pixel resolution 1/10, discrimination = ” = accuracy, sorepeatability is = ”
93Part Location Analogous to gauging Can expect to achieve sub-pixel resolution: repeatability and accuracy
94Flaw Detection Contrast! Contrast! Contrast! Detection Vs. ClassificationDetection: High Contrast, normalized background (no pattern), can detect a flaw that covers 3 X 3 pixelsClassification: flaw should cover 25 X 25 pixels
95OCR/OCV Stroke width - 3 pixels wide Character should cover 25 X 25 pixelsSpacing between characters - 2 pixelsSingle font style - boldResult % read rate effectiveness
96Linear Array Image Capture pixelsScanning rates up to KHzSpeed should be well regulatedResolution in direction of travel function of speed and sampling rate of camera
98Questions to Ask Before Proceeding From: Wayne Chaneski, Taking Steps to Improve Your Process, Modern Machine Shop, February, 2000.Can we eliminate the process?Does each step in the process add value?Is the step a duplication?Is the step a built in correction now unnecessary?
99Questions to Ask Before Proceeding From: Wayne Chaneski, Taking Steps to Improve Your Process, Modern Machine Shop, February, 2000.Can we change the process?How can the operation be changed?Can different equipment make the operation better?Can we use less costly options?Can the operation be done faster?Can the operation frequency be reduced?
100Questions to Ask Before Proceeding From: Wayne Chaneski, Taking Steps to Improve Your Process, Modern Machine Shop, February, 2000.Can we rearrange the process?Is it now the most efficient?Would changing the layout eliminate handling, transport, etc.?Is the operation sequenced properly?Can operations be performed in parallel?
101Questions to Ask Before Proceeding From: Wayne Chaneski, Taking Steps to Improve Your Process, Modern Machine Shop, February, 2000.Can we combine steps in the process?Can suppliers perform some operations more effectively?Can customers perform some operations more effectively?
102Questions to Ask Before Proceeding From: Wayne Chaneski, Taking Steps to Improve Your Process, Modern Machine Shop, February, 2000.Can we simplify the process?Are we taking full advantage of tooling/fixturing/material handling to assist operators?Are instructions easy to understand?Has all the information been made available, including “helpful hints” and “pitfalls” inherent in the process?
103Questions to Ask Before Proceeding From: Wayne Chaneski, Taking Steps to Improve Your Process, Modern Machine Shop, February, 2000.Can we imagine the perfect process?What would it be?How much time would it take?Would we get a better yield rate?Could we ship products to our customers faster?How different is the perfect process?Can perfect process be documented - flow chart, etc.?
104General Defect prevention is better than the cure! Study application site personally!Consider vision to enhance people!Expect productivity to decline!
105Steps to Take When Buying a Machine Vision System application issues:generic applicationvariables: part, presentation, etc.material handlingoperator interfacemachine interfacesenvironmental issuessystem reliability/availabilitymiscellaneous: documentation, warranty, training, software, spares, serviceacceptance test/buy off procedureresponsibilities
106Tools Job descriptions Present specifications Part drawings Floor space drawingsSamplesPhotos/videosPersonnel
107Steps to Take When Buying a Machine Vision System Write functional specificationUse “Machine Vision Requirements Checklist” - available from MVA - forces examination of:production processjustification issuesapplication issues
108System Spec Defines “what” system is and “how” system will work involves examination of implementation detailsprogramming standardsstylecontrol methods
109System Specification The spec is not what the customer wants! Creeping expectations!Variables - Gotchas!
110System Specification Adhere to factory standards Adhere to engineering standardsUse conventional jargon for part descriptions and to describe the processUse existing frames of reference to develop acceptance test
111Before RFP Prepare preliminary conceptual design Develop schedule - be realisticAssess costDetermine technical and cost feasibility
112Developing Functional Requirements What does the system do?What specific function do you want the MV value adder to do?What goals do you expect to achieve with MV?Will the MV system be for a retrofit or next generation product?
113Developing Functional Requirements Defines “what” system is and “how” system will workinvolves examination of implementation detailsprogramming standardsstylecontrol methods
114Developing Functional Requirements Does the application involve:One object at a timeMultiple objectsHow many different objectsWhat are the part numbers?Is it a batch operation or continuous dedicated process?What are the changeover times and frequency of changeovers?
115Developing Functional Requirements What are the skill levels involved in changeover?How is function currently being performed?Can new variations to the part be expected? What might they be?Where do parts come from? What is material handling surrounding MV?
116Developing Functional Requirements Can rejected parts be repaired?Where do pass and fail objects go?When does the project have to be completed?How many shifts is the equipment used?If machine vision fails, what is the option?
117Developing Functional Requirements How many MV systems will be required annually?What are the consequences of a failed MV sequence?What are the consequences of a false reject?
118Developing Functional Requirements Describe the applicationGenerically, does the application involveGaugingAssembly verificationFlaw inspectionPattern recognition
119Developing Functional Requirements If GaugingWhat are the tightest tolerances?What is the accuracy design goal?What is the repeatability design goal?Are there reference features?What are calibration requirements?
120Developing Functional Requirements If assembly verificationDimensions of assemblyIs it presence/absenceOrientation verificationWhat is the smallest piece to be verified and dimensions of that piece?Is part correctness also required?
121Developing Functional Requirements If flaw inspectionDescribe flaw typesWhat is the smallest size flaw?Does the flaw affect surface geometry?Does the flaw affect surface reflectance?Is it more of a stain?Is classification of flaws required?
122Developing Functional Requirements If location analysisWhat is the design goal for accuracy?For repeatability?What is the area over which the “find” is required?Will angular as well as translation correction be required?Will scale change?Describe calibration requirements
123Developing Functional Requirements If pattern recognitionWhat is the size of the pattern?Describe difference between patterns?Is there a background pattern?Does pattern involve color? Geometry?Number of different patterns?Is objective to identify? To sort?
124Developing Functional Requirements If specifically OCR/OCVFixed font? Variable font? What is font?What is the height of the characters?What is the stroke width?What is spacing between and around characters?How many characters in string? How many lines? Color of print?Describe background – color, “busyness”
125Developing Functional Requirements Object dealing withWhat is material?What is finish (texture) like? Dull, glossy, specular?Is surface finish the same on all surfaces?For all part numbers? Production runs?Any platings, coatings, films, paints?Markings?
127Developing Functional Requirements Object dealing with –Possibility of warping, shrinking, bending, etc?Any change in appearance over time?Any markings?General appearance variables?Sensitivity to light?
128Developing Functional Requirements Material handlingPresent handling or being considered?Production rates? Currently? Future?Parts static? Moving continuously? Speed?If indexedHow long stationary?Total in-dwell-out time?Settling time?Acceleration?
129Developing Functional Requirements Material handlingMaximum positional variations – translation, rotation?More than one stable state?Volume envelope for MV?Any restrictions or obstructions?What triggers action?What is result of MV?
137Good RFP Describes project in detail Describes operation’s business Reviews why the project is being solicitedReviews schedule
138RFP Should Request Schedule Training Service Warranty Software ownershipDocumentationInstallation support
139Steps to Take When Buying a Machine Vision System Identifying VendorsAIA - DirectoryMVA - DirectoryOpto*Sense databaseVendor type:image processing boardgeneral purpose machine vision systemapplication specific machine vision systemsystem integrator
147Steps to Take When Buying a Machine Vision System Evaluate vendors systematicallyUse Decision Matrix technique to assess proposalsVisit the “best” vendors to assess:application engineering skillsquality control proceduressoftware practicestraining materialsdocumentationpoliciesreferences
149Proposals Should Include Review of implications of variables:stagingimage processingimage analysisImplication of organization/lack of organization of partsTime budget to demonstrate confidence throughput can be met
150Proposals Should Include Position/temperature error budgetsInterfacing issues:peoplemachine/lineMiscellaneous issues:enclosures start up/changeoverbattery back up maintenancediagnostics calibrationreports
151Proposals Should Include Exceptions to the specResponsibilities:installationspecifically what is required for system to be successfulAcceptance testingvalidation procedurechallenge set
152Proposals Should Include Policy reviewtraininginstallationwarrantyfield servicesparessoftware upgradesdocumentation
158Sizing up Vendors - Differentiators How long before a service call is made or phone support is obtainedAre software upgrades included in the price? Is upgrade notification automatic?What is the company’s annual sales revenue in the specific product/application
159Finalize Evaluations Make sure vendor understands Visit most responsive vendorsVisit up and running installations
160Vendor Decision Previous work Quality of work Reputation Ability to meet scheduleUnderstanding of your business and application
161Reference Checks Quality of work Ability to meet schedule Policies SupportWould they do it over !!!
162Systems No system should be more complicated than it need to be! Good application engineering is critical!Contrast, Contrast, Contrast!Staging is important, if not more important than image processing algorithms!
163Customer Software and hardware should be transparent! Tinkering should be discouraged!Should not specify equipment, rather function!Samples furnished should be representative of all variables expected!Training is critical!A little knowledge is dangerous!
164Vision Company Room lighting is a No - No! Vision company should have all disciplines required!Beware of “Piece of cake!”Look for relevant experience!Verify quality practices!Verify policies: training, documentation, etc.!
166Application Engineering Material handlingMust avoid jamming regardless of deformities!Murphy’s Law - If it can go wrong, it will!LightingLighting is not a constant!Never use software to compensate for poor lighting!Shrouds are cheaper than software fixes!
167Application Engineering OpticsThere are limits to resolution!Nothing exceeds the speed of light!Image resolutionNyquist’s theorem does apply!More resolution means more compute power!A pixel is not a fixed size! - Magnification issues
168Steps to Take When Buying a Machine Vision System Write an acceptance test plan/buy-off procedureDifferent for:attribute inspection system - based on Thorndyke Chart to arrive at sample size - to test for both escapes and false rejectsgauging/location analysis - repeatability/accuracy performance at upper limit, nominal and lower limit of tolerance
169Acceptance Testing Includes evaluation of operator interface basic operationcalibrationaccuracy & repeatabilitythroughputsensitivitymaintainabilityavailability
170Acceptance Testing Test at system level Test at other than nominal Test failure modesTest everything in system specDon’t put anything into spec that can not be tested!!!
171Buy-off at Supplier Simulate external equipment Generate reports Run through all screen functionsSimulate alarms and failure modesPower up/down system and components
172Steps to Take When Buying a Machine Vision System Using the Thorndyke charte.g.. for 0 defects95% confidence400 PPM (reliability)from chart np - 3.0n = 3/400 x 10 -6n = 7,500for every factor: color, finish, size, etc.
176Steps to Take When Buying a Machine Vision System Create a challenge to verify performance
177Working With The System In The Factory Should not deteriorate production speed!Ideally, avoid having to re-engineer the manufacturing process to accommodate machine vision!System should have the capacity to be reconfigured!
178Training Basic principles of operation Normal operating procedures screen functionspower up/downreportsAlarm conditions and recovery procedures
179Training Back-up procedures Normal and emergency maintenance Calibration
181Mistakes in Buying Automation 1. No equipment specification2. Requesting quotes before visiting prospective suppliers3. Incorrect cost estimate4. Insufficient in-house machine support5. No input from production people
182Mistakes in Buying Automation 6. Poor communication with vendor7. Acceptance of inadequate equipment8. Failure to supply latest drawings and parts with specifications9. Failure to design for automation10. Using the wrong technologyper E. Martin, Lanco/NuTec, Assembly March 96
183Reasons Why Automation Fails Per Automation Research Corp. StudyUnclear or false expectations regarding what is to take place and the results that are to be achievedLack of commitment by user managementOver dependence on technical solutions
184Reasons Why Automation Fails Lack of acceptance by the user organizationPoor project managementNot properly taking into account the human resources issues
185SI Difficulties With Users Inadequate specificationsLack of technical knowledgeNo management commitmentInternal policiesSeparating needs from wantsInability to take over systemChanges in midstream
186SI Difficulties With Users No one person in chargeTight project constraintsLack of communicationPrice constraintsInability to take risksManpower shortagesRigid specifications
188Benefits of Machine Vision Scrap reductionScrap disposal costsReworkInventory reduction associated with reworkAvoiding value addedImproving machine uptime - capital productivityAvoiding return and warranty costsImproving customer satisfaction
189Project Justification Tangible benefits:increase productivityreduce scrapreduce rework time/inventoryavoid adding value to scrapavoid product returns - warranty issuesavoid liability issuesavoid field service
190Project Justification Tangible benefits:avoid freight costs on returnsavoid equipment breakdowns/improve machine uptimeimprove product fabrication cycle and impact on inventorysave indirect labor costsave floor space to store rework inventory
191Project Justification Tangible benefits:training/labor/turnover/recruiting costsout of cycle costs due to schedule upsetswaste disposal costscosts of overruns to compensate for yieldpersonnel/payroll costs per employee:average worker’s compensationaverage educational grant per employeetooling/fixturing savings
192Project Justification Intangible benefitsimprove quality - consistency of qualitypredictability of qualityinformation automationflexibilitypeople effectiveness/limitationssample inspection only monitors system errors, not random errors
193Project Justification Intangible benefits:process controlenvironmentconsumer/government pressure“eyes” for automationexpansion needsseasonality
194Project Justification Because some things appear to be intangible does not mean they have zero value !!!In final analysis, justification of technology is a management issue - not an accounting issue !!!!
195Project Justification Data required:How many pieces are produced per month per line?How many production lines make the piece?What is the current inspection time per piece? (minutes/piece)What is the inspection labor rate? ($/hr including benefits)
196Project Justification Data required:How many rejects per month (%)?What is the value of a reject - $ -?What is the value of the raw material in the piece - $ -?What percent of the rejects are reworked per month?What is the average rework time/piece (minutes/piece)?
197Project Justification Data required:What is the monthly warranty cost - $? - includes costs of field service, field returns, repairs, shipments to and from plant, paperwork, etc.Product liability costs per month - $? - includes liability claims, lawyer fees, insurance, paperwork, etc.
198Project Justification Data required:What percent of the rejects are scrapped per month? - the difference between the number of rejects per month and the number of rejects reworked per month and returned to inventoryWhat are the monthly waste disposal costs due to the scrapped pieces?
199Project Justification Data required:What are the scrap and rework inventory costs per month? - eg. Calculate based on average number of units scrapped and in inventory per month multiplied by the value (cost) of the piece divided by 10 (factor that assumes any such unit will only be in inventory an average of two days)How many shifts does the line operate?
200Project Justification Data required:Total hours operating per shift?Hours worked per month/shift/person? - paid hoursNumber of units sold per month? Average selling price of the piece? - not cost Indirect (supervisory) labor rate ($/hr with benefits)?What is the profit per piece produced? ($)
201Project Justification Data required:Current cost of money? Prime rate + 1%?If sample inspection, hours per month for specific piece?
202Project Justification Calculated values:annual direct cost of inspection per piece = inspection labor rate X hours worked per month/shift X number of shifts line operates X 12annual indirect cost of inspection per piece = indirect labor rate X hours worked per month/shift/person X number of shifts X 12
203Project Justification Calculated values:cost of rejects scrapped = percent rejects/ month X value of a reject X % pf rejects scrapped/ month X number of pieces produced/month X 12cost of rework = percent of pieces reworked/month X number of pieces produced per month X rework time X rework labor rate X 12
204Project Justification Calculated values:warranty costs = monthly warranty costs X 12liability costs = monthly product liability costs X 12scrap disposal costs = monthly cost X 12scrap and rework inventory costs = monthly X 12training costs - based on turnover experience
205Project Justification Assigning values:value of reliable data = sum of annual direct and indirect labor costs X 0.05value of improved customer satisfaction = average selling price of the piece X number of units sold per month X 12 X 0.001
206Project Justification Assigning values:percent uptime line improvement anticipated - an estimated valuevalue due to gain in line uptime = cost of machine vision system X number of systems required X 0.05
207Project Justification Costscost of machine vision system (or systems)launch costs (training, etc.) - estimate 10% of machine vision system costsannual service contract - estimate 10% of machine vision system costs
208Project Justification Costs:opportunity cost - function of the cost of money = cost of the machine vision system X number of systems + launch costs + annual service contract X number of systems X current cost of money
209Project Justification Costs:total equipment costs = cost of machine vision systems X number of systems + launch costs + annual service contracts X number of systems + opportunity costaverage annual cost over four years = total equipment costs/4
210Project Justification Return on investment = (average annual savings/total equipment costs) X 100Payback (years) = total equipment costs/(average annual savings + average annual costs with machine vision)
211Average Payback Period by Company Size Per Automation Research Corp.
212Average Payback Period in Years by Industry Per Automation Research Corp.