Embed Size (px)
Citation preview
1
Big Data Analytics for Fleet Asset Management
Karl Reichard2, Jeffrey C. Banks1 The Applied Research Laboratory The Pennsylvania State University
1Originally Presented at: Sapphire Now + ASUG Annual Conference May 14-16, 2013 by Jeffery Banks Phone: 814-863-3859 Email: [email protected]
2Phone: 814-863-781 Email: [email protected]
2
Global Fleet Readiness View
3
BCT Location and High Level Readiness View
4
Individual Vehicle Engine Coolant Temperature History
5
Current Architecture
This automated analysis is currently a batch process and need to be continuous.
6
Objective: To organize vehicles into paths of potential component failure or maintenance actions, based on comparing real-time data trends of test vehicle data to reference vehicle data.
Drive Train Failure
Engine Failure
Operational Data
Predictive Analytics
Analysis Process: Does the operational data match a pattern for a known failure?
X
Days
7
Brigade Combat Teams
8
Data Warehouse
Parametric and Metadata Database (in memory)
Data Ingest Process
Data Source (e.g. Vehicle
Data)
ABCD Files ABCD
Files ABCD Files
Staging Directory
ABCD Files ABCD
Files ABCD Files
‘Real Time’ Analytics User Interface Web Service
Analysis and Recommendation
Results
Future Architecture
Information Customers: Unit Maintainers, Life Cycle Managers, etc.
9
Prediction = Cost Savings
• Failure prediction is achievable. – Prediction capability enables cost savings,
enhanced mobility and increased readiness for DoD.
• This technique requires enormous amounts of data.
• Prediction capability is enabled by near real time analytics at the enterprise level.
