Determining Capable Production Processes for New Product Ramp-up using Semantic Web Technologies Stefan Biffl1, Estefania Serral2, Roland Willmann3 Vienna University of Technology Institute of Computer Aided Automation 3Institute of Software-Technology and Interactive Systems 1Christian Doppler Laboratory „Software Engineering Integration for Flexible Automation Systems“ KU Leuven 2Department of Decision Sciences and Information Management Outline Motivation and Problem Statement State of the Art and Research Approach Cake Baking - The Reference Use Case Knowledge based Ramp-up of Products (K-RAMP) Conclusions and Next Steps Outline Motivation and Problem Statement State of the Art and Research Approach Cake Baking - The Reference Use Case Knowledge based Ramp-up of Products (K-RAMP) Conclusions and Next Steps Scope of Work Product Ramp-up Discrete Production Remote Capacity Ramp-up Motivation Yield Product Ramp-up Costs Duration Quality Difficult to Predict Weak Knowledge Exchange Time to Volume Volume on Market Time to Market First Part on Market Time Fails of ~50% of all Enterprises Ramp-up Project Motivation Knowledge Based Ramp-Up of Products Product Ramp-up K-RAMP Determine Reusable (Semi-Finished) Products and Process Segments at Target Production System Derive New Process Setup From Existing Process Setup Recommend Results Weak Knowledge Exchange For Process Qualification at Target Production System K-RAMP Shall Actively Drive the Ramp-Up of a New Product Ramp-up Project Process Reuse of Existing Knowledge – Trial 1 Unable to Reuse Existing Knowledge ? Delicious Chocolate Gloss! How do you make it? 1. 2. 3. 4. 5. 6. Temperature 4 to 5 on the hotplate Put water pot on hotplate Wait 15 minutes Put pot with chocolate in the water pot Whipping for 5 minutes … Reuse of Existing Knowledge – Trial 2 Knowledge For Making a Hot Water-Bath Exists And Can Be Reused Ahh – a hot water-bath That is clear 1. 2. 3. Delicious Chocolate Gloss! How do you make it? Make a hot water-bath Whipping the chocolate in the hot water-bath until chocolate is liquid … Outline Motivation and Problem Statement State of the Art and Research Approach Cake Baking - The Reference Use Case Knowledge based Ramp-up of Products (K-RAMP) Conclusions and Next Steps Challenge of New Product Ramp-up The Situation in Discrete Production of Many Industries is Similar But More Complex Inter-disciplinary Ramp-up Team – Product, Process, Equipment, Quality Engineering – Line Management, Corporate Management Ramp-up Teams Deal With Existing Knowledge For Reuse – Existing Semi-Finished Products (or Consumables) – Existing Processes Which Create Similar Semi-Finished Products – Development and Qualification of New Processes or New Machines Knowledge Exchange Among Team-Members is Critical Ramp-Up Not Actively Supported by Software-Tools Challenges of K-RAMP Common Generic Concept of Knowledge – Common Structure of Knowledge at Source and Target Production System – Generic, Independent of Specific Processes and Production Systems Integration of Product and Process Knowledge with Quality Measurements of Target Production System – Data Volume – Linking Semantic Knowledge and Quality Measurements – Valuable History of Quality Data Searching Within Quality Data for Opportunities for Reuse – Dynamic Setup of Data-Mining Based on Product Specification Calculation of Performance Indicators for Validation of Recommendations – Dynamic Setup of Calculations Based on Product Specification Approach Proof of Concept – Reference Product and Process (Transfer of Baking a Chocolate Cake) – Simulated Quality Measurements Interactive Tools for Planning, Specification and Verification – – – – Enterprise Architect – Planning Protégé 4.3 – Knowledge Base and Reasoning (Pellet) Magnum Opus, Version 4.6.3 – Association Rule Mining MiniTab 17 – Calculation of Performance Indicators Manual Execution of the Process Validation of Results with Respect to Reusability Iterative Improvement of Knowledge Base and Reasoning Example Use Case – Knowledge Transfer Between Bakeries Initial State Product Structure and Specifications of Product To Transfer Resulting State Structure and Specifications of Existing Products Structure and Specification of Existing Products and Transferred Product K-RAMP Process Process Setup Quality Data of Product To Transfer Original Bakery Process Setup Quality Data of Existing Baking Processes Process Setup Quality Data of Existing and new Baking Processes Target Bakery K-RAMP Partial Models Characteristics Product :Characteristic :Specification :SpecificationRange :PhysicalUnit :ProductClass :Product :ProducedPart :QualificationState Process :ProcessClass :ProcessSegment :ProcessJob :ProcessSetup QualityControl :AssociationSearchSetup :AssociationRule :AggregationFunction :CapabilityIndex Specialization of K-RAMP K-RAMP General Model General Conceptual and Instance Models Baking Branch-Specific Model Common Product Classes, Specifications, Characteristics, Process Classes Location-Specific Models Individual Process Segments, Products, Produced Parts and their Measurements Source-Bakery Target-Bakery Outline Motivation and Problem Statement State of the Art and Research Approach Cake Baking - The Reference Use Case Knowledge based Ramp-up of Products (K-RAMP) Conclusions and Next Steps Transfer of Chokolate Cake Baking Product Class / Products Requirements How Shall the Product Function? Requirement-1: Reflective, Dark Brown, Crispy Chocolate Surface With Bitter-Sweet Taste Requirement-2: Soft, Succulent, Dark Brown, Fruity Apricot-Chocolate Taste Product Classes support Requirements Variation of Satisfaction of Requirements due to Individual Specifications of Products Transfer of Chokolate Cake Baking Product Specification Requirements Is There Already a Product of Type “Gloss” IsWhich ThereMeets Already a Product Class Product of Type “Gloss” Specification Ranges Specification to be Adjusted in Order Viscosity: … Flow Value: … to Meet Specification Ranges How Shall the Product Function? Requirement-1.1: Dark brown Chocolate Surface With Bitter-Sweet Taste Requirement-1.2: Reflective surface Requirement-1.3: Crispy surface Requirement Specification Reflectivity: … Uniformity: … Transfer of Chokolate Cake Baking Process Specification Product Specification Product Class Specification Viscosity: … Flow Value: … How Is the Product Created? Is there a Process of Type “Provide Gloss” Is There Process of Type WhichaCreates “Provide Gloss” “Gloss” Within Which Could be Adjusted Expected Specification InRanges Order to Meet Specified Ranges Process-Class-1.1: Provide Bitter Chocolate Process-Class-1.2: Provide Gloss Process Setup Water/Sugar Ratio: … Temperature: … Duration: … Outline Motivation and Problem Statement State of the Art and Research Approach Cake Baking - The Reference Use Case Knowledge based Ramp-up of Products (K-RAMP) Conclusions and Next Steps K-RAMP – Recommendation of Preliminary Process Plan Product Hierarchy Requirements Process Hierarchy Chocolate Cake Chocolate Cake Provider Chocolate Cake Chocolate Sponge Dough Chocolate Gloss Chocolate Gloss Provider Gloss Coating Chocolate Gloss Chocolate Sponge Dough Provider Chocolate Sponge Dough Chocolate Gloss requires Bitter Chocolate Recommended Process Plan for Target Production System Bitter Chocolate Provider Chocolate Gloss Provider Gloss Provider Chocolate Sponge Dough Provider Bitter Chocolate Gloss Bitter Chocolate Provider Gloss Provider provides Some Terminology of Quality Management (Semi-Finished) Product is Qualified if FPY > Target FirstPassYield (1 Count( AllParts) ) *100% Count(GoodParts) Specification Limits Good Parts are All Parts Within the Requested Specification Limits (Specification Range) Population of 99.9989% within Specification Limits is equivalent to Process Capability Index (cpk) of 1.43 is equivalent to Defective Parts 11.7 ppm If Specification Range of Product-1 is Within the Specification Range of Product-2 (LSL1 LSL2 and USL1 USL2) Then All Good Parts of Product-1 Satisfy Product-2 Specification Range as Well Is There Some Systematic Process Setup to Achieve an Appropriate Yield for Tighter Specification Limits of Product-1 (a New Product) LSL2 LSL1 USL1 USL2 K-RAMP Process Overview Reusable Existing Semi-Finished Products Recommend SemiFinished Product yes Recommend Process Setup no groups left some groups left no Flag Produced Parts of Super Class If Within Specification Range Clustering of Flagged Produced Parts no clusters Recommend Need For New Process Development Drop Groups Where Yield Below Threshold clusters Gather Produced Parts in Groups Based on Clusters Calculate Yield Per Group With Respect to New Product Outline Motivation and Problem Statement State of the Art and Research Approach Cake Baking - The Reference Use Case Knowledge based Ramp-up of Products (K-RAMP) Conclusions and Next Steps Conclusion Automated Determining of Reusable Products and Processes Possible For Simple Use Cases External Applications Required (Rule Mining, Statistics) Hybrid Architecture OWL2 and SWRL Seems Not Sufficient for Reasoning JENA Rework of Ontology Next Steps Addressing Reasoning for Closed-World Questions – Non-Qualified-Product \ Qualified-Product Declaration of new Classes or Restrictions Through Reasoning – If there is an Association Rule X with hasContext {Y, Z} then there is a Restriction X with onProperty hasContext value {Y, Z} More Complex Use Cases – Sequence of Production Process Segments Consideration of Regression Functions Between Process Specification and Produced Parts’ Measurements Detailed Concept For User Interaction http://www.informatik.tuwien.ac.at/english Roland Willmann – roland.willmann@tuwien.ac.at K-RAMP Find Produced Parts Within Specification Range Initial Situation Product Class Chocolate is Qualified at Target Bakery Product Class Bitter Chocolate is Non-Qualified at Target Bakery Bitter Chocolate Specification Range Tighter Than Chocolate’s Produced Parts of Chocolate are Available Assumption Produced Parts Hold Values for Each Characteristic A Subset of Produced Parts of Chocolate are in Specification Ranges of Bitter Chocolate Actions to be Performed If There is Some Non-Qualified Product Class X And There is Some Qualified Super-Class of X Then Declare Restriction of Produced Part Instances Within Specification Range of X. K-RAMP Find Systematic Process Setup Initial Situation Existing Produced Parts Within Specification Range of Bitter Chocolate Assumption Produced Parts Refer Process Setup (a.k.a. Process Context) Actions to be Performed Query Produced Parts of Chocolate Within Specification Range of Bitter Chocolate Good Parts Query Produced Parts of Chocolate Totality Query Classes of Referenced Process Context Left Hand Side (LHS) of Association Rules Query Asserted Association Rule Setup Right Hand Side (RHS) of Association Rules Good Parts IN Totality – Good Parts OUT K-RAMP Find Systematic Process Setup Resulting Association Rules for New Product Class MT-01 & O-01 & S-03 IN [Coverage=0.043 (11); Support=0.031 (8); Strength estimate=0.659; Lift estimate=2.31; Leverage=0.0190 (4.9); p=0.0850] S-01 & MT-01 & O-01 IN [Coverage=0.051 (13); Support=0.031 (8); Strength estimate=0.571; Lift estimate=2.00; Leverage=0.0168 (4.3); p=0.213] O-01 & S-02 & MT-02 IN [Coverage=0.051 (13); Support=0.027 (7); Strength estimate=0.505; Lift estimate=1.77; Leverage=0.0129 (3.3); p=0.267] S-01 & O-01 & MT-03 IN [Coverage=0.066 (17); Support=0.035 (9); Strength estimate=0.504; Lift estimate=1.77; Leverage=0.0162 (4.2); p=0.235] O-01 & S-02 & MT-03 IN [Coverage=0.059 (15); Support=0.031 (8); Strength estimate=0.504; Lift estimate=1.77; Leverage=0.0145 (3.7); p=0.262] Action to Perform For Each Association Rule X Declare Restriction of Produced Part Instances Based on LHS (a.k.a. Process Context) of X K-RAMP Calculating Initial Capability Indexes and Yields MT-01 & O-01 & S-03 S-01 & MT-01 & O-01 O-01 & S-02 & MT-02 S-01 & O-01 & MT-03 O-01 & S-02 & MT-03 K-RAMP Calculating Initial Capability Indexes and Yields Action to be Performed YieldTotal: 53% MT-01 & O-01 & S-03 YieldTotal: 77% S-01 & MT-01 & O-01 YieldTotal: 38% O-01 & S-02 & MT-02 YieldTotal: 60% S-01 & O-01 & MT-03 YieldTotal: 56% O-01 & S-02 & MT-03 YieldCacao: 99% For Each Restriction X Caused by an Association Rule Calculate YieldTotal With Specification Range of Product Class Which is Associated With Association Rule YieldSugar: 78% K-RAMP Recommendation for Ramp-Up Team YieldTotal: 53% MT-01 & O-01 & S-03 YieldTotal: 77% S-01 & MT-01 & O-01 YieldTotal: 38% O-01 & S-02 & MT-02 YieldTotal: 60% S-01 & O-01 & MT-03 YieldTotal: 56% O-01 & S-02 & MT-03 Action to be Performed Query Process Setup Related to Restriction X Caused by an Association Rule Where YieldTotal > YieldThreshold And Recommend it as Initial Process Setup For Qualification of New Derived Process K-RAMP Find Produced Parts Within Specification Range At Target-Bakery :Chocolate is Qualified :BitterChocolate is Non-Qualified :Chocolate’s Specification Ranges Envelop Bitter Chocolate’s Specification Ranges :ProducedPart rdfs:subClassOf :ProducedPartChocolate :ProducedPartBitterChocolate Some Produced Parts of Chocolate are in Specification Ranges of Bitter Chocolate :BitterChocolate rdf:type rdf:type ratioCacao: [40, 70]% ratioSugar: [0, 15]% rdf:type rdfs:subClassOf rdf:type :ProductClass Knowledgebase and Reasoning Protégé 4.3 rdf:type :Chocolate ratioCacao: [10, 95]% ratioSugar: [0, 27]% K-RAMP Find Systematic Process Setup :ProducedPart Association Rule Mining Based on :AssociationSearchSetup Instances External Tool Magnum Opus, Version 4.6.3 rdfs:subClassOf :ProducedPartChocolate :onTotality :BitterChocolate ratioCacao: [40, 70]% ratioSugar: [0, 15]% :onProductClass :AssociationSearchSetup rdfs:subClassOf :Chocolate ratioCacao: [10, 95]% ratioSugar: [0, 27]% minLift: 1.0 minSupport: 0.02 minStrength: 0.5 minCoverage: 0.02 :onLHS :onGoodParts :ProducedPartBitterChocolate K-RAMP Find Systematic Process Setup :AssociationRule MT-01 & O-01 & S-03 -> satisfies [Coverage=0.043 (11); Support=0.031 (8); Strength estimate=0.659; Lift estimate=2.31; Leverage=0.0190 (4.9); p=0.0850] :ProducedPart rdfs:subClassOf S-01 & MT-01 & O-01 -> satisfies [Coverage=0.051 (13); Support=0.031 (8); Strength estimate=0.571; Lift estimate=2.00; Leverage=0.0168 (4.3); p=0.213] O-01 & S-02 & MT-02 -> satisfies [Coverage=0.051 (13); Support=0.027 (7); Strength estimate=0.505; Lift estimate=1.77; Leverage=0.0129 (3.3); p=0.267] S-01 & O-01 & MT-03 -> satisfies [Coverage=0.066 (17); Support=0.035 (9); Strength estimate=0.504; Lift estimate=1.77; Leverage=0.0162 (4.2); p=0.235] O-01 & S-02 & MT-03 -> satisfies [Coverage=0.059 (15); Support=0.031 (8); Strength estimate=0.504; Lift estimate=1.77; Leverage=0.0145 (3.7); p=0.262] :hasSetup :AssociationSearchSetup minLift: 1.0 minSupport: 0.02 minStrength: 0.5 minCoverage: 0.02 rdfs:subClassOf :ProducedPartChocolate :ProducedPartBitterChocolate