actionable insight in semiconductor manufacturing, from ... · pdf fileactionable insight in...

Click here to load reader

Post on 20-Feb-2018

215 views

Category:

Documents

1 download

Embed Size (px)

TRANSCRIPT

  • Actionable insight in semiconductor manufacturing, from big data to Cognitive Computing

    Christophe Begue, beguec@us.ibm.com July 2016

    Copyright 2016 IBM

  • GPS

    Semiconductor firms see significant opportuni2es for Big Data to op2mize the way they execute across func2ons

    Product Development & Manufacturing compress design, development & manufacturing lead time and improve yield and asset utilization

    Marketing & Sales ... design and execute more effective marketing with optimized product assortments, affinities and pricing

    Supply Chain & Distribution ... optimize inventory and assets and deliver a reduction in supply chain and distribution costs with single view product

    Market Research & Product Ideation

    ... align product concepts with consumer desires, improve new product ideas, and new

    product launch effectiveness for IoT

    Procurement & Vendor Management ... embed insight into business processes from Manufacturer to Distributor to Customer to Consumer

    Finance ...grow revenue and improve margins with greater business performance insight, and improved forecasting and planning

    External Data

    Massive Internal Data

    Field and Warranty Management ... collect field data from connected devices, understand part behavior, predict failures, reduce warranty cost

    Copyright 2016 IBM

  • Traditional Fab Infrastructure

    MES

    Messaging Bus

    Recipes Auto mation Data

    collect

    FDC Reports SPC APC YMS

    Current Scope

    Advanced Capabilities

    Stop bad product or tool

    Wafer / Lot Analysis Single parameter trends

    Uni Multivariate

    Tool Sensor Diagnostics Predictive Maintenance

    Virtual Metrology

    Data Mining

    Enhanced Data Mining Predictive Modeling

    Automated Models

    Big Data

    Many fabs struggle to move from current capabili2es to advanced predic2ve techniques for yield and asset management

    Copyright 2016 IBM

  • Manufacturing, Product data in context of supply chain, consumers, quality, external events

    Responding dynamically to multi dimensional and time sensitive situations

    Applying machine learning techniques to interconnected physical devices data, patterns and trends

    Marrying instrumented and interconnected data with unstructured data

    Yet the growth of sensor and IoT data calls for a different approach

    Copyright 2016 IBM

  • Why is Cogni2ve necessary? Data from billions of interactions between machines, devices and people is massive, complex

    and variable.

    Pre-defined programs arent able to analyze it. Traditional systems can't make sense all the IoT data combined with unstructured data.

    A cognitive system makes sense of all types of data, it works across data sources and decides

    which patterns and relationships matter.

    It uses machine learning and advanced processing to organize the data and generate insights. It evolves and improve through learned self-correction and adaptation.

    ANALYTICS COGNITIVEINFORMATION KnowledgeDATASENSORS

    Copyright 2016 IBM

  • There are three capabili2es that differen2ate cogni2ve systems from tradi2onal programmed compu2ng systems

    Copyright 2016 IBM

  • Today Cognitive Manufacturing

    Constraints in the past Data: Structured /historical data (e.g. wafer maps, process tools, etc) Analysis: Reactive / Preventive Maintenance Production Planning: Daily Cycle with managing demand

    New Capabilities in the future Data: Structured /Unstructured / Stream (e.g. Adaptive Test) Analysis: Conditional to predictive Maintenance, Yield Analysis Production Planning: Real Time WIP planning and routing

    Based on real-time prediction model and sensor data to diagnose the risk of EQP. 1 min

    Based on Inspection data to discover the EQP with unstable performance. 6hrs

    Engineer was informed to shutdown the EQP and started the process of trouble shooting. 4hrs

    Products had been hold until Engineer recover EQP

    Influenced product can't be reworked

    Produce 30 lots to cover the lost of scrap, high priority product type was delay to deliver. 8hrs

    EQP auto-shutdown and Analyzer Engine send root-cause ranking list for Engineer to fix the problem. 10 mins

    Optimizer Engine to calculate the best dispatch scheduling, send decision to control center, enable backup EQP. 5 mins

    Integrate information of influenced products to Analyzer Engine to generate revised recipe and revised route for next operation. 2 mins

    high priority product type on time

    Interconnec2vity, real 2me analysis, structured process and unstructured context data and machine learning are the essen2al elements of Cogni2ve Manufacturing

    Copyright 2016 IBM

  • Connect & Configure

    Develop & Visualize

    Analyze & Predict

    Become Cognitive

    Instrument your equipment/assets to collect data

    Gather already existing data

    Visualize your data in meaningful dashboards

    Start to see patterns

    Gain insights from the data

    Predictive models

    Add context data Bring in cognitive

    components Machine learning Understand,

    Reason, Learn Asset needs to be connected, outfitted with sensor or data gathered

    Use analytical models to predict equipment failures and

    Use the platform to quickly build dashboards for data visualization

    Use speech, video, image to diagnose complex problems

    STAIRCASE TO COGNITIVE MANUFACTURING

    Copyright 2016 IBM

View more