ADVANCED DIAGNOSTICS AND PREDICTIVE ANALYTICS OF...
Transcript of ADVANCED DIAGNOSTICS AND PREDICTIVE ANALYTICS OF...
ADVANCED DIAGNOSTICS AND PREDICTIVE ANALYTICS OF VEHICLE DATA DR. JOHANN PRENNINGER
IECON 2013, Nov. 12 2013, Vienna
ADVANCED DIAGNOSTICS AND PREDICTIVE ANALYTICS OF VEHICLE DATA AT BMW
1 Contents
2 State of the Art On- and Offboard Vehicle Diagnosis
3 Analytical Challenges, the customers view
4 The FACTS process for agile Analytics at BMW
5 Conclusion
Page 2 IECON 2013, Nov. 12 2013, Vienna
1 Contents
2 State of the Art On- and Offboard Vehicle Diagnosis
3 Analytical Challenges, the customers view
4 The FACTS process for agile Analytics at BMW
5 Conclusion
Page 3 IECON 2013, Nov. 12 2013, Vienna
ADVANCED DIAGNOSTICS AND PREDICTIVE ANALYTICS OF VEHICLE DATA AT BMW
Service advisor
• Customer visit of
service station for service or repair.
• Service advisor is
picking up the car and the original statement (o-tone) of customer.
• Mechanic
performs diagnosis, reads out onboard data, failure messages etc.
• Big picture of
failure situation.
• Repair is done as
proposed and guided by diagnosis software.
• Diagnostic /
technical data is created/logged at this time.
• Data transfer of
technical and business data to the headquarter.
• Storage in
different data warehouses.
• Interpretation and
analysis of the data.
• Check and Payment of Warranty Claim.
• Starting point for
Data Mining.
DIAGNOSIS AND REPAIR. TOUCHPOINTS TO THE PROCESSES OF THE DEALERSHIP
Data ReadOut
Diagnosis & Repair
Data Transfer
Analysis & Reporting
Page 4 IECON 2013, Nov. 12 2013, Vienna
DIAGNOSIS AND REPAIR START MUCH EARLIER. CUSTOMER IMPRESSION AND FEEDBACK IS VALUABLE
Page 5
TeleServices
Connected Drive
Internet Forum, …
• The driver
“produces” data
about his driving
experience.
• He expresses himself
online about product
issues.
• He is occasionally
asked via surveys
and studies (TÜV,
IACS, JD-Powers, …)
Customer survey
TeleServices,
Connected Drive,
Internet Forum, …
• The customer
expresses & shares
his aftersales
experience online.
• He is asked
individually about
his service / repair
experience.
• He evaluates his
satisfaction.
IECON 2013, Nov. 12 2013, Vienna
Diebstahl-
warnanlage
Reifendruck-
kontrolle
Regensensor
Lichtschalt-
modul
Heizungs-
bedienteil
Heizungs-
bedienteil
(Fond)
1-Achs-
Luftfeder
Bedienzentrum
Mittelkonsole
Anhänge-
modul
Kombi-
instrument
Sicherheits- u.
Informations-
modul
Fahrzeug-
zentrum
Mensch-
Maschine-
Interface
Park Distance
Control
Digitale Diesel
/Motor
Elektronik 1
Adaptive
Cruise
Control
Schaltzentrum
Lenksäule
A-Säule links
B-Säule links
A-Säule
rechts
B-Säule links
Tür vorne
links
Sitz Fahrer
Audio System
Kontroller
Tür vorne
rechts
Sitz hinten
Aktive Roll-
Stabilisierung
Dynamische
Stabilitäts-
kontrolle
Elektronische
Getriebe-
steuerung
Sitz Beifahrer Sitzmodul
Fahrer
Sitzmodul
Beifahrer
Antennentuner Multi Media
Changer
Audio-CD
Wechsler Videomodul
Bedienzentrum
Mittelkonsole
Fond
Adaptive Light
Control
Elektronische
Dämpfer-
kontrolle
Elektromech.
Feststell-
bremse
Drehraten-
sensor
(kein SG)
Sitzmodul
Beifahrer
hinten
Powermodul
Digitale Diesel
/Motor
Elektronik 2
Schiebehebe-
dach
Chassis
Integration
Module
Verstärker
Navigations-
system
Telefon-
interface
Spracheingabe-
system
Kopfhörer-
interface
Heckklappen-
lift
Sitzmodul
Fahrer hinten
Standheizung/
Zuheizung
Controller
Wischermodul
Series
optional
Diagnose-
Zugang
Zentrales
Gateway
Modul
Car Access
System
Türmodul
Beifahrer Türmodul
Fahrertür
Türmodul
Fahrerseite
hinten
Türmodul
Beifahrerseite
hinten
ONBOARD DIAGNOSIS DELIVERS AN INTENSIVE MIX OF DATA OF A COMPLEX FAILURE-SITUATION
Page 6 IECON 2013, Nov. 12 2013, Vienna
Diebstahl-
warnanlage
Reifendruck-
kontrolle
Regensensor
Lichtschalt-
modul
Heizungs-
bedienteil
Heizungs-
bedienteil
(Fond)
1-Achs-
Luftfeder
Bedienzentrum
Mittelkonsole
Anhänge-
modul
Kombi-
instrument
Sicherheits- u.
Informations-
modul
Fahrzeug-
zentrum
Mensch-
Maschine-
Interface
Park Distance
Control
Digitale Diesel
/Motor
Elektronik 1
Adaptive
Cruise
Control
Schaltzentrum
Lenksäule
A-Säule links
B-Säule links
A-Säule
rechts
B-Säule links
Tür vorne
links
Sitz Fahrer
Audio System
Kontroller
Tür vorne
rechts
Sitz hinten
Aktive Roll-
Stabilisierung
Dynamische
Stabilitäts-
kontrolle
Elektronische
Getriebe-
steuerung
Sitz Beifahrer Sitzmodul
Fahrer
Sitzmodul
Beifahrer
Antennentuner Multi Media
Changer
Audio-CD
Wechsler Videomodul
Bedienzentrum
Mittelkonsole
Fond
Adaptive Light
Control
Elektronische
Dämpfer-
kontrolle
Elektromech.
Feststell-
bremse
Drehraten-
sensor
(kein SG)
Sitzmodul
Beifahrer
hinten
Powermodul
Digitale Diesel
/Motor
Elektronik 2
Schiebehebe-
dach
Chassis
Integration
Module
Verstärker
Navigations-
system
Telefon-
interface
Spracheingabe-
system
Kopfhörer-
interface
Heckklappen-
lift
Sitzmodul
Fahrer hinten
Standheizung/
Zuheizung
Controller
Wischermodul
Series
optional
Diagnose-
Zugang
Zentrales
Gateway
Modul
Car Access
System
Türmodul
Beifahrer Türmodul
Fahrertür
Türmodul
Fahrerseite
hinten
Türmodul
Beifahrerseite
hinten
~17 Mio. Vehicles in the field ~4.000 dealerships in 90 countries, ~50.000 service people Up to 65 Electronic Control Units in a single BMW 1.000 individual, selectable Car Options >1 GByte Functional Software, 15 GByte Data onboard ~2.000 Customer relevant Software Functions ~12.000 Diagnostic Trouble Codes implemented in Onboard Diagnosis ~3.000 Metric Values in all ECUs in average per car ~ 10.500 Testmodules for all BMW series ~ 34.000 Schematics documents Up to 60.000 Diagnosis Sessions per day worldwide total size of DataWareHouse >> 30TByte
VEHICLE DIAGNOSTIC DATA TODAY. QUANTITY AND VARIETY OF ON- & OFFBOARD DIAGNOSIS
Page 7 IECON 2013, Nov. 12 2013, Vienna
1 Contents
2 State of the Art On- and Offboard Vehicle Diagnosis
3 Analytical Challenges, the customers view
4 The FACTS process for agile Analytics at BMW
5 Conclusion
Page 8 IECON 2013, Nov. 12 2013, Vienna
ADVANCED DIAGNOSTICS AND PREDICTIVE ANALYTICS OF VEHICLE DATA AT BMW
MATURIY LEVELS IN VEHICLE DATA ANALYTICS. EVEN SIMPLE QUANTIFICATIONS ARE VERY BENEFICIAL
Level 1: Count / Know
Level 2: Qualify / Compare
Level 3: Analyse / Understand
Level 4: Forecast / Predict
Level 5: Suggest / Advise
Count
Qualify
Analyse
Predict
Advice
Page 9 IECON 2013, Nov. 12 2013, Vienna
PREDICTIVE ANALYTICS CHALLENGE: FEEDBACK DATA TO REALIZE COHERENCE AND CAUSALITY
Vehicle Design
Failure Avoidance
Test Strategy
Failure Appraisal
Customer Sensitivity
Quality Function Deployment
Production Optimisation
PartsQuality + Dimensioning
Customer Understanding
Diagnostic Trouble Codes Environmental Conditions
Symptoms, Testplans
Defect Codes
Testmodules, DiagnosisCodes
LabourCodes Replacement Parts
Vehicle-Identification
Repair Advice
ECU Metrics
coherence
causality
Page 10 IECON 2013, Nov. 12 2013, Vienna
DATAMINING THE DIAGNOSIS PROCESS IN DETAIL. VISUALISATION OF THE OFFBOARD DIAGNOSIS
Page 11 IECON 2013, Nov. 12 2013, Vienna
A-PRIORI ANALYTICS TO FIND ADVANCED CUSTOMERS USAGE AND FAILURE PATTERNS
Find the causal failure patterns in vehicles with specific critical defects using a Sequence Analysis.
A-priori datamining methods with SystemML™ in a BigData environment compute results near realtime.
Numerous „events“ during a vehicle‘s
Lifecycle (production, diagnosis, warranty, etc)
Millions of historic datasets of the BMW & Mini
fleet. Billions of individual attributes available.
© 2012 BMW & IBM Corporation
All chains of events for cars with a specific defect
All chains of events for all cars.
Page 12 IECON 2013, Nov. 12 2013, Vienna
1 Contents
2 State of the Art On- and Offboard Vehicle Diagnosis
3 Analytical Challenges, the customers view
4 The FACTS process for agile Analytics at BMW
5 Conclusion
Page 13 Advanced Automotive Diagnostic Systems, 22.-24.04.2013
ADVANCED DIAGNOSTICS AND PREDICTIVE ANALYTICS OF VEHICLE DATA AT BMW
THE FACTS PROCESS AT BMW. FIELDDATA ANALYSIS FOR CUSTOMER SATISFACTION
Analytical Helpdesk (Tickets)
Data
assignment
Datasources
and interfaces
Support
of
professionals
(OnePager / project
description )
Vehicle
file
market
radar FBM-Viewer
Apps for internal
R&D experts (actually more than 120
Apps available)
Analytic-
Dashboards
and Applications
Tele- service
DMS DWH Parts
FBM
...
Internal Customer (Requirements)
Self-Service
... ... ... ... ... ... ... ... ... ...
... ... ... ... ... ... ... ... ... ... Repeat Visit
Repeat Repair Monitoring TÜV-Targets
Service-
history
Monitoring
Analysis of
Diagosis &
Flashupdates
... ... ... ... ... ... ... ... ... ...
... ... ... ... ... ... ... ... ... ... Satisfaction
[%] ?
What
helps … ?
Qualtiy after
Warranty ?
Field
penetration?
Problems in
Repair?
CoC
Making the data available group wide.
Experience in DataMining
Page 14 IECON 2013, Nov. 12 2013, Vienna
THE AGILE FACTS PROCESS. CRISP DATA MINING AND SCRUM COMBINED
BMW-
Data warehouse
1
Request /
Question
Evaluation
4
Data
Preparation
Calculation /
Modelling
3
Business
Understanding
Data
Understanding
2
Typical FACTSProcess Steps:
1 Question of Department X: Automatic
Detection of an important vehicle
annoyance in the field.
2 Balance of request: How can the question
be answered with specific data available
3 Initial concept (FACTS) and realization of
data load, modeling. Iterative („Scrum“)
enhancements with enquirer.
4 Presentation of Results, evaluation of
Business Case and definition of
Deployment.
5
Deployment in FACTS App-Store as
Webservice for self-service Realtime-
Analysis (defined users with specific
access rights).
FACTS Process - CRISP & SCRUM:
Deployment 5
Page 15 IECON 2013, Nov. 12 2013, Vienna
PREDICTIVE ANALYTICS TOOLS CLASSICAL DATAMINNG TOOLS FOR EXPERTS ONLY.
Page 16 IECON 2013, Nov. 12 2013, Vienna
THE TOOLCHAIN FOR DATA ANALYTICS. INTERACTIVE TOOLSET TO DISCOVER THE ISSUES
Page 17 IECON 2013, Nov. 12 2013, Vienna
THE BMW FACTS ANALYTICAL APP STORE. PREDICTIVE ANALYTICS AS SELF SERVICE PROCESS
Welcome to BMW FACTS Analytical APP Store
IECON 2013, Nov. 12 2013, Vienna
1 Contents
2 State of the Art On- and Offboard Vehicle Diagnosis
3 Analytical Challenges, the customers view
4 The FACTS process for agile Analytics at BMW
5 Conclusion
Page 19 IECON 2013, Nov. 12 2013, Vienna
ADVANCED DIAGNOSTICS AND PREDICTIVE ANALYTICS OF VEHICLE DATA AT BMW
Understanding the vehicle quality and realizing the customer impression in the field is a complex multidimensional task.
On- and off-board diagnosis play an essential role, but need interpretation by the engineers and developers.
FACTS delivers a process to answer many questions by combining the data available.
Predictive analytics is provided as a set of analytical apps, encapsulated and simplified to deliver results as an interactive self service process.
By enabling intensified usage of feedback from the field we understand and improve customer satisfaction.
FACTS DELIVERS INTERACTIVE, USER AND CUSTOMER CENTERED BUSINESS ANALYTICS AS SELF SERVICE
Page 20 IECON 2013, Nov. 12 2013, Vienna