Predictive Analytics in Quality Early Warning Systems Smarter with Analytics SC Lim Growth Markets Leader Integrated Supply Chain Engineering
Agenda: IBM Point of View Maintenance & Quality Management IBM Predictive Maintenance & Quality (PMQ) Deeper Dive: Quality Early Warning System(QEWS)
What is Predictive Maintenance? Utilizing analytics to predict asset failure Maintenance Maturity Model Reactive Maintenance (machine fails, then fix) Preventative Maintenance (based on manufacturers schedules, time, or operational observations) Conditionbased Maintenance (based on monitoring to assess condition of assets) Predictive Maintenance (based on usage and wear characteristics to predict failure) Source: Gartner
For Quality problems, Time is Money It s vital to detect maintenance/quality problems quickly Product Design Engineering & Pre-production Manufacturing & Production After sales & Service 4 Objective: early insight reduces cost Quality Point of Sale Even a small delay in detecting problems can have big costs: reworked or scrapped product asset downtime/utilization recall of defective product higher costs in warranty claims loss of customer satisfaction delayed product shipments Maintenance Cost of Quality Industry Recalls Battery $450m+ Chip $400m+ Auto $5b+ 2012 IBM Corporation
For Maintenance/Quality Problems not if but when and how big is the impact No company is immune to the challenges that maintenance & quality pose Increased complexity in Design & Processes Growing reliance on Suppliers Speed to Market Cost Challenges Customer Demands A Predictive Maintenance & Quality Management System using Analytics is needed to take us to a new level.
IBM Journey with Analytics SQMS PQMS Prescriptive - What should we do about it? Automated Alerts, Subtle changes Predictive Maintenance & Quality (PMQ) - Quality Early Warning System (QEWS) Supplier Risk Management 6 Pre Stop / Stop Ship BoM Assist explore PCN SPL Qpex Offspec Checklist QIN TQA Clear Quest Oceans of Data Big Data Predictive - What will happen? Forecasts, Models Field Data Critical Parts Management Tool Hotline Descriptive - What happened? Query / Drill Down, Reporting, Dashboards 2012 IBM Corporation
The IBM solution: IBM Predictive Maintenance and Quality Reduce costs Improve asset productivity Increase process efficiency
Predictive Maintenance and Quality provides several key features Real-time Capabilities Big Data, Predictive Analytics, Business Intelligence Quick and Accurate Decisioning Easy Data Integration Open Architecture QEWS - Advanced Quality Algorithms Accelerated Time to Value
What is IBM Quality Early Warning System? An implementation of IBM PMQ software solutions which uses proprietary IBM technology to detect and prioritize quality problems earlier and more definitively than can be done using traditional techniques of statistical process control. QEWS answer the critical question: Has anything change(subtle) enough to require action? 9 ISC Engineering
The value of QEWS Lower Costs in Manufacturing reduced rework reduced scrap in Distribution fewer recalls in Warranty fewer claims Improved Productivity in Manufacturing / Purchased Product reduced rework reduced scrap increased capacity utilization more on-time shipments higher assurance of delivery of quality products in Engineering more product coverage, and more effective coverage with existing engineering resources prioritization of the most pressing issues single effective process for the entire enterprise Improved Brand Value improved brand image protect high-stakes, high volume product launches higher top-line growth and customer retention 10 ISC Engineering
Where can QEWS be applied? At any point where a test, measurement, or inspection is made: In the supply chain: suppliers final test incoming inspection of raw materials incoming inspection of procured components In manufacturing: at individual production operations at final product test In product field performance warranty claims 11 ISC Engineering
QEWS Analysis Capabilities Module Examples Typical Applications Attribute data failure rates yields sort categories Monitoring quality of components procured from suppliers Monitoring quality of manufacturing operations via inline tests Monitoring quality of manufacturing operations via final product test Parametric data Reliability data machine controls process controls dimensional measurements electrical measurements warranty claims stress tests Monitoring stability of manufacturing process operations Detecting shifts in product performance Monitoring quality of product in customer use environments Detecting product wearout 12 ISC Engineering
Why is IBM s QEWS unique? Uses advanced analytics to provide earlier identification of Quality Issues Highlights trend not typically visible with few false alarms A proven enterprise-level solution: Easily handles hundreds of thousands of data points Automated warnings to stakeholders for rapid action Reduces cost of quality and cost of warranty Empowers Engineering Staff: Automatically prioritizes problems; Built-in tools to help jump-start problem-solving powerful and fast data drill-down 13 ISC Engineering
Why is early problem detection difficult? Traditional quality monitoring systems use Statistical Process Control: like the QWERTY keyboard: not optimal, but very difficult to change If set up for early detection, it creates many false alarms - staff don t know when the system is crying wolf vs. when there s a real quality problem 14 ISC Engineering
SPC can t provide useful early detection of subtle problems Real problem Credible results are late Credibility For subtle problems, SPC has a poor trade-off between timeliness and statistical confidence (i.e., credibility) False alarm Early results are not credible Late Timeliness of detection Early 15 ISC Engineering
SPC charts: Which ones require attention? 16 ISC Engineering
QEWS makes changes visible 17 ISC Engineering
QEWS Demonstration Results: Earlier Detection 1 st QEWS alert This chart shows QEWS analysis results for a set of quality data. The x-axis is aligned in time to the chart below. QEWS alerts when the cumulative evidence crosses above the horizontal threshold line (in black.) 1 st SPC alert This chart shows SPC analysis results for the same set of data as above. The x-axis is aligned in time to the chart above. SPC alerts when a point falls outside the control limits (at the extreme right-hand side of the chart.) 8 weeks QEWS alerts 8 weeks earlier than SPC. 18 ISC Engineering
QEWS Demonstration Results: Definitive Detection 1 st QEWS alert mounting evidence continuing alerts For this set of data, QEWS alerts, then stays in alert mode (above the horizontal black threshold line.) The positive slope of the cumulative evidence line indicates the quality problem is getting worse. From the first alert onward, QEWS presents a clear message that action should be taken. 1 st SPC alert 2 nd SPC alert no SPC alerts For the same set of data as above, SPC alerts once, then does not alert again until many points later. Many engineers would dismiss the first alert as an anomaly, and not take action until the second alert. 19 ISC Engineering
Based on a proven architecture created from best-of-breed technologies End user reports, dashboards, drill downs Predictive analytics Telematics, manufacturing execution systems, existing databases, distributed control systems Decision management Analytic data store High-volume streaming data Business intelligence (Prebuilt data schema for storing quality, select machine and production data, and configuration) Integration bus (Prebuilt data schema for storing quality, select machine and production data, and configuration) Enterprise asset management systems Advanced analytics powered by IBM SPSS and IBM Cognos Data integration provided by IBM Websphere Message Broker and IBM Infosphere Master Data Management Collaborative Edition, which feeds a prebuilt, DB2-based data schema Analytics Solution Foundation provides data orchestration Process Integration with Maximo automatic work order generation Includes data models, message flows, reports, dashboards, business rules, adapters and KPIs
IBM Predictive Maintenance and Quality analyzes data from multiple sources and provides recommended actions
IBM Implementation PMQ/QEWS Detail page View History View Trend data View AVT data View QEWS data Key Algorithms Threshold setting Target Optimization Prioritization View Data Detail View Comments Create Ticket 22 ISC Engineering
Summary: Predictive Maintenance and Quality enables better business outcomes Monitor, maintain and optimize assets for better availability, utilization and performance Predict asset failure and identify poor quality parts earlier & more definitely to better optimize operations and supply chain processes Reduce guesswork and incorporate experiential knowledge during the decision-making process Includes foundational models, dashboards, reports and source connectors