The Data Science behind Predictive Maintenance for Connected Vehicles
- 1. The Data Science behind Predictive Maintenance in Connected Vehicles Esther Vasiete Srivatsan Ramanujam Pivotal Data Science Data Engineers Guild - Meetup June-21, 2016
- 2. Picture credit (from L to R): http://www.techlicious.com/blog/ericsson-mobility-report-internet-connected-devices/ http://www.mdpi.com/1424-8220/14/10/19260/htm http://www.thehindubusinessline.com/info-tech/other-gadgets/care-for-a-connected-car/article5777444.ece Devices are Increasingly Connected
- 4. How can these connected devices in our home be smart enough to make daily life easier?
- 5. How does this… …become this? By recognizing this And by processing this Sensors + Other Unstructured Data
- 6. How can we know a tree has fallen on a power line before the residents complain?
- 7. How can we use data to help prevent accidents like the Macondo Disaster ?
- 8. Gene Sequencing Smart Grids COST TO SEQUENCE ONE GENOME HAS FALLEN FROM $100M IN 2001 TO $10K IN 2011 TO $1K IN 2014 READING SMART METERS EVERY 15 MINUTES IS 3000X MORE DATA INTENSIVE Stock Market Social Media FACEBOOK UPLOADS 250 MILLION PHOTOS EACH DAY In all industries billions of data points represent opportunities for the Internet of Things Oil Exploration Video Surveillance OIL RIGS GENERATE 25000 DATA POINTS PER SECOND Medical Imaging Mobile Sensors
- 9. To realize this opportunity requires the right tools and techniques Problem Formulation Modeling Step Data StepApps Step
- 11. Data Lake Ingest Business Levers Dashboard/App PL/X Modeling• Data cleaning • Data Exploration • Feature Engineering Model Validation Feedback loop for continuous model improvement Driver and Vehicle Meta Data Data Ingestion Platform ✔ ✔ ✔ ✔ ✔✔ ✔ ✔ Data to Apps
- 12. Data Science Use-cases for connected cars 1
- 13. Data Science Use-Cases 1 ● Predictive Car Maintenance ‒ More accurately predict part failure ‒ Optimize part repair and replacement schedule ● Leveraging Driving Behaviour ‒ Useful to differentiate insurance pricing based on driving style ‒ Optimize car design ● Improving GPS Systems ‒ Establish baseline for traffic congestion ‒ Create more meaningful metrics for routing ‒ Infer public transportation effects on traffic ‒ Predict how long incidents would take to clear ● Predictive Power for Assistance Systems ‒ Optimize fuel efficiency ‒ Predict the future state of a car in the next 2 minutes (starts, stops, emergency braking) ● Traffic Light Assistance ‒ Signal timing of traffic lights ‒ Crowd sourcing of traffic signals ‒ Optimize traffic light patterns to reduce congestion
- 14. Preventive Maintenance for Connected Vehicles 1
- 15. On-Board Diagnostics Diagnostic Trouble Codes (DTC) Unscheduled repairs AB1029 – Power steering pump replacement CT3408 – Wheel alignment
- 16. Solving the preventive maintenance problem Automakers Customer Satisfaction Auto Repairs
- 17. Data Sources for Predictive Maintenance VIN Timestamp DTC Code Odometer Speed Acceleration Engine Temperature Engine Torque GPS Coordinates etc. VIN Date vehicle in Date vehicle out Repair code Parts replaced Warranty claims Repair Comments Vehicle Sensor Data Vehicle Repairs Data
- 18. Predicting Job Type from Diagnostic Trouble Codes (DTCs) Time Job Type: Transmission Job Type: Transmission Engine Job Type: Regular check DTC: B DTC: B, P, C DTC: U DTC: B DTC: B DTC: B, P, C, U DTC: P, B, U DTC: P DTC: B DTC: B,P DTC: B,P Can the DTCs observed here predict this Job Type? Can the DTCs observed here predict this Job Type? Can the DTCs observed here predict this Job Type?
- 19. Predicting Job Type: a multi-class classification problem DF 12 10 DF 12 15 DF 29 80 AB 10 29 AB 16 22 AB 16 25 AB 86 22 CT 34 02 CT 34 08 CT 35 60 CT 24 09 Vehicle Features
- 20. Two-stage Hierarchical Classification Framework Vehicle Features DF 12 10 DF 12 15 DF 29 80 AB 10 29 AB 16 22 AB 16 25 AB 86 22 CT 34 02 CT 34 08 CT 35 60 CT 24 09
- 21. Model Parallelism One or more job on the same day Multi-labeling problem One-vs-rest classifiers built in parallel 1 0 0 1 0 1 0 Class 1 Class 2 Class 3 One-vs-Rest Classification Red vs. Non Red On Segment 1 Green vs. Non Green On Segment 2 Blue vs. Non Blue On Segment N
- 22. • Predictive maintenance problems are challenging because DTC signals are not always symptomatic of an ensuing repair. • Given the hierarchical nature of repair codes, we built a two stage hierarchical classification framework comprising a top- down cascade of classifiers. • Major system jobs can be predicted earlier to the repair date. Key Takeaways
- 23. Reference Architecture %%publish model info. / Microservices (Spring Boot) /load_model /score_model Spring Cloud Data Flow vehicle data (streaming) connector exploratory data analysis & model training Rabbit/Kafka source training (offline) scoring (online) / web or mobile app dashboard