dame: distributed engine health monitoring on the grid tom jackson university of york

DAME: Distributed Engine Health Monitoring on the Grid

Tom Jackson

University of York

http://www.escience-grid.org.uk/index.htm

http://www.epsrc.ac.uk/website/Index.aspx

SC2004 Distributed Aircraft Maintenance Environment - DAME

Contents

• Project Overview• Demonstrator Objectives• General Architecture• Data Mining Activities• Decision Support tools


Project Partners

• EPSRC Funded, £3.2 Million, 3 years, commenced Jan 2002.• UK pilot project for e-Science (£220 million programme)• 4 Universities:

– University of York, Dept of Computer Science– University of Sheffield, Dept of Automatic Control and Systems

Engineering – University of Oxford, Dept of Engineering Science– University of Leeds, School of Computing and School of

Mechanical Engineering

• Industrial Partners:– Rolls-Royce– Data Systems and Solutions– Cybula Ltd


DAME Objectives

• Building a demonstration system as proof of concept for Grid technology in the aerospace diagnostic domain.

• Three primary Grid challenges:– Management of large, distributed and heterogeneous data

repositories;– Rapid data mining and analysis of fault data;– Information management and data fusion for

diagnosis/prognosis applications;

• Other key (commercial) issues:– Remote, secure access to flight data and other operational

data and resources;– Management of distributed users and resources;– Quality of Service issues (and Service Level Agreements)


Operational Scenario

Engine flight data

Airline office

MaintenanceCentre

London AirportNew York Airport

GRID DiagnosticsCentre

Engine flight data

European data centre

Engine flight data

Airline office

MaintenanceCentre

London AirportNew York Airport

US data centre

GRID DiagnosticsCentre

Engine flight data


How Big is that Lake?

• Heathrow capped at 36 landings per hour.• If half have 4 engines and half have 2, average

aircraft carries 3 engines.• Each engine generates around 1GB of data per

flight.• 36 x 3 x 1 = 108GB raw engine data per hour.• Factor in the working day and the

rest of the world…• …Terabytes and up!

London

Tokyo

Cape Town

Mexico City


Demonstrator Objectives

The DAME demonstration system provides a diagnosis workbench (portal) which brings together a suite of analysis services via Grid technology;

– Provides access to a range of analysis tools for the engine diagnosis process

– Provides decision support and data fusion environment for diagnosis and prognosis

– Will act as central control point for automated workflows– Manages issues of distributed diagnosis team and virtual

organisations– Manages issues of security and user roles.


DAME Functional Overview

Distributed Data Repositories

‘QUICK’TM

Download

NovelData

Eng

ine

Da

ta

Ser

vice

Dat

a XTOGrid Service

AURA-GCBR

DecisionSupport

EngineModel

DataVisualisation

DAME PortalSecurity management Workflow Manager

Collaborative Working

Eng

ine

Da

ta

Eng

ine

Dat

a

AU

RA

DA

TA

Middleware Services & Data Management

Resource Broker PMC & SRBGrid Service Manager Quality of Service/SLA’s

Grid InfrastructureGT3


White Rose Grid Distribution

Leeds

Grid Middleware Services

Sheffield

Modeling & Decision Support

DAME WRGSign-on Portal

SDMDatabase

CBRAnalysis-GEngineModel-G

GT3 ServiceCBR advisor

GT3 ServiceBD25 Enginemodel wrappedas Grid Service

XTO-G

GT3 ServiceXTO plug-ins via a GridService

DataVisualiser

GT3 ServiceJchart Viewer forviewing XTO output

Workflow

Browser basedworkflow tool.Compliant withResource Broker

Resource BrokerGT2 ServiceSchedule workflowtasks on WRGresource

Oxford

Engine Data Store

Engine DataDatabase

York

Data Mining Services

AURA-GAURA-GDatabase

DataStore-GGT3 ServiceSimulates arrival &storage of QUOTEdata

Zmod Viewer

GT3 ServiceBrowser based dataviewer for zmod files

GT3 ServiceZmod datasearch facility

Collaboration tools

GT3 ServiceToolset for multiusercollaboration

WRGGT3/2

WRGGT3/2

WRGGT3/2

WRGGT3/2

DAME workbench

SecurityGT3 SecurityServiceProxy-Management

DAME GUI

GT3 ServiceBrowser based GUIto DAME services


Data Mining Activities

• Objectives have been to develop a data mining service to search fleet archives of QUICK engine data within operational time constraints.– To support diagnosis and prognosis activities– To support long term fleet predictive maintenance– Business assumption is that data is archived remotely

• Two tools have been developed:– AURA-G: Grid enabled signal search engine;– Signal Data Explorer: Interactive search GUI for signal data

• These tools are generic to any pattern matching domain, but have been tuned to RR & DS&S requirements for DAME.


AURA-G Service & PMC

• AURA is a highly scalable, binary associate memory• DAME has developed a distributed Grid enabled version of this

technology;• Provides capability to mine huge volumes (Terabytes) of

complex signal data with sub second response times.• AURA-G is managed by a DAME middleware service called

Pattern Match Controller:– Provides mechanisms for distributing and managing search

queries

• A virtual data archiving system, Storage Request Broker(SRB) provides the middleware for data cataloguing and archiving.


AURA Integration & Deployment

• AURA is a collection of processes; data adaptors, search-engine and back-check. It wraps around an existing data storage system;

Data

Dat

a A

dapt

orS

tore

Se

arc

h

Inp

ut

pa

tte

rn

Candidate Engine(Back check)

Indexer

Output pattern

AURASearchEngine

Results

binary

Store & Search

Store &

Search

Indexes or Data

ResultStore

AURA Search Process

Correlation Matrix Memory (CMM).

• Scalable high-performance.

• Family of generic techniques.

• Wide range of data types.

• Large volumes of data.

• Find exact and near-matches.


Distributed Pattern Search

• Complex time-series pattern matching process driven from a visualisation front end.

Engineer selects region of interest

A search is launched across the fleet data archives using the Grid enabled AURA search engine


Distributed Pattern Search, 2

• All matched pattern records are retrieved from the fleet archives and ranked according to similarity

Matches can be viewed


Pattern match Control

• The SDE interfacesto the PMC middleware;

• PMC provides:– Distributed search– Interface to data archive

system (SRB or other)– Scalablity– Robustness

Pattern Match Control (PMC)‘master’ node

Pattern Matching Service

Search()Store() ReturnResults()

NodeSearch()

Signal DataExplorer

Storage Service

Extractor

PMC

PMC

Other PMCnodes

ReturnResults()

SearchConstraints

Builder

Raw Data

SRB

Engine Data

Tracked Orders

Store()

Scope

Constraints

Search()

Stored Data

Raw Engine Data

GetResults()

Engine Data


PMC cont

• PMC architecture has been developed on business premise of remote data.

• E.g. Airports act as data repositories for Engine health data

• SRB provides hugely scalablevirtual catalogue& index system

Services•Pattern

Match Control

•Data Extractor

•Pattern Match Engines

DAME Airport Node - Gatwick

Services

DAME Airport Node - Munich

DAME Airport Node - JFK MCAT

Signal Data Explorer

Services

Tracked Orders

SRB

CatalogueServices

•Pattern Match Control

•Data Extractor

•Pattern Match Engines

DAME Airport Node - Heathrow


Decision Support tools

• Objectives are to maximise use of available data assets to support business processes:– Diagnosis & Prognosis– Fleet predictive maintenance– Workflow management– Domain knowledge capture

• Achieved through integration of diverse data assets through Grid enabled Case Based Reasoning tools.

• Grid provides standardised interface to legacy systems and data servers.


CBR Maintenance Analysis in DAME

• User driven CBR processing

• Automated CBR processingvia Workflow Manager


CBR Tools

• Results of CBR search & rankResults of CBR search & rank

• View a case and it’s attachmentsView a case and it’s attachments


Summary

• DAME is demonstrating the potential of Grid-based diagnostics for health-monitoring applications;

• Grid paradigm offers:– Scalable solutions for management of distributed data– Secure access to services and data– Role based access and virtual collaboration on fault data– Compute resource on demand for high performance tasks,

such as signal processing and modelling

• DAME has shown how data growth problems could potentially be addressed within scope of distributed data assets.

dame: distributed engine health monitoring on the grid tom jackson university of york

Documents

data fusion environment

gb of data

operational data

average aircraft

gb raw engine data

grid technology

dept of automatic control

diagnosis workbench