Big Data trifft Industrie Im Internet der Bosch-Dinge und -Dienste Dr. Lothar Baum Februar 2015 1
Bosch Business Areas Automotive Technology Industrial Technology Energy and Building Technology Consumer Goods 2
Bosch on the Way to a Connected World SW enabled products 15'000 software developers (total, not limited to IoTS) 75% of all Bosch products are or soon will be controlled by embedded software IP enabled potentially IP enabled 69% of future Bosch sales will be done with products that are IP enabled IoTS 2400 developers already working on IoTS products and services 3
Use Case Examples Manufacturing Example: Predict low-mileage failures to reduce quality claim costs Healthcare Example: Predict critical patient situations to prevent hospitalizations Field Data Example: Analyze ESP logs to derive new insights and services 4
Types of Manufacturing Use Cases Example: Predict low-mileage failures to reduce quality claim costs Test Time Validate test relevance and predict test results from process data Machines Predict machine failures and maintenance requirements Warranty Costs Predict warranty problems from manufacturing test and process data Yield Identify root causes for performance problems; benchmark across lines Scrap Costs Identify process parameters influencing/creating scrap; predict scrap early in the value chain 5
Manufacturing Use Case End-of-Line Testing TG + FG (TG + FG) (TB model + FB model ) Warehouse TB + FB kpa TB model + FB model Warranty Costs Predict pressure warranty & problems pressure from manufacturing control pressure related claims test and process quality data related claim quality claim mm 6
kpa Big Data trifft Industrie Manufacturing Use Case Big Data can get messy! 23 databases >70 part types >30 different test layouts >17 bn measurements over 10 years There are tricky algorithmic problems! 3'000'000 parts shipped p.a. economic constraints: e.g.: may scrap 3 good parts for one correctly identified bad part TB model : FB model = 1:3 100 relevant claims p.a. pressure related quality claim mm 7
Healthcare Use Case 2900 patients data since 2005 Triaging Intervention measure outcomes 707'200 unique sessions Data quality is an issue! 1'600 hospitalizations Data privacy is key! 8
ESP Field Data BMW E60 (5-Series) data data collected since 2006 400'000 vehicles Every Brake Event with odometer reading, speed, braking pressure, braking duration, and many more features 9
ESP Field Data BMW E60 (5-Series) data data collected since 2006 400'000 vehicles Every Brake Event with odometer reading, speed, braking pressure, braking duration, and many more features Field data shows that linear extrapolation of ESP pump usage is incorrect 10
Summary: Big Data Challenges Technology is not the bottleneck it s data availability, quality, and understanding CRISP-DM Business Understanding Data Understanding DM question validated Need to build up Data Scientists in business units to help setting data mining goals, and help prepare the data Deployment Evaluation Data Preparation Sometimes, you don t need so Big Data but the right data and the right questions Modeling 80% of project effort! Have a solid business case before you invest hundreds of thousands of Euros into tools! 11
Thank You! Contact: Dr. Lothar Baum CR/PJ-TOP79 Robert Bosch GmbH Postfach 30 02 40 70442 Stuttgart! lothar.baum@de.bosch.com 12