BROADEN Business Resource Optimisation for Aftermarket and Design on Engineering Networks Prof. Jim Austin Computer Science, University of York UK CEO Cybula Ltd. Outline Bit of History What was the data and processing challenge 2 Data Management Data analysis Middleware architecture The industry problem: Aerospace The science problem: Neuroscience Scaling to other industries NESC 2008 History 3 Started with DAME Distributed Aircraft Maintenance Environment Worked with Rolls-Royce to develop ideas and learn about the Grid. Completed in 2003 NESC 2008 Partners 4 Rolls-Royce plc. EDS Defence Ltd. Lost Wax Media Ltd. Concurrent Thinking Ltd. Oxford BioSignals Ltd. Cybula Ltd. Computing Laboratory, University of Oxford Department of Computer Science, University of York Informatics Research Institute, School of Computing, University of Leeds Department of Automatic Control and Systems Engineering, University of Sheffield NESC 2008 History BROADEN started at the end of DAME Ran for 3 years, from Jan 2005. DTI funded Four objectives: (1) large scale and CFD simulation and visualisation (2) very large scale agent-based modelling of aftermarket processes incorporating logistics and the supply chain (3) integrated diagnostic and prognostic tools for health monitoring across development test, production pass-off and in-service aero engines (4) the formulation of a strategy to transfer proven Grid technology to production LANs 5 NESC 2008 History BROADEN 6 Took DAME into Rolls Royce Focussed on deploying DAME technology Built a Rolls Royce Grid Tested technology on test bed systems NESC 2008 The generic challenges Challenge Data Asset or System Data Store Knowledge • Massive data collection happening everywhere • Focus of TSB last call • BUT Need to get knowledge from the data 8 NESC 2008 Challenge Europe China US One Data Warehouse Japan 9 NESC 2008 Rolls-Royce Engine flight data Engine flight data London Airport New York Airport Airline office GRID Diagnostics Centre Maintenance Centre US data centre European data centre 10 NESC 2008 Challenge 11 Single data store is a processing bottleneck Data changes too fast to communicate it and keep up to date Networks too slow to transfer the data The system has a central point of failure Large processing needed on centralised store NESC 2008 Data & processing management Local nodes PME PMC PME PMC PME PMC PMC - Pattern Match Controller architecture : manages processing PME - Pattern Match Engine : Manages pattern match 12 NESC 2008 PMC and PME Do not move the data - move the processing 13 That’s much lower effort More reliability Scalable Networks need not be high bandwidth Leverages available local computation NESC 2008 Data analysis The Signal Data Explorer SDE Data feed Asset Data silos ! Event detector Basic SDE data process 15 NESC 2008 SDE Allows a user to: 16 Set up triggers on multiple, complex, real time data feeds. Find examples of events that are seen but unknown. Visualise and ‘explore’ the data NESC 2008 SDE 17 Signal Search and analysis Finds examples like previous NESC 2008 SDE 18 NESC 2008 SDE 19 NESC 2008 SDE 20 NESC 2008 SDE 21 NESC 2008 SDE 22 NESC 2008 SDE principles Focussed on 23 High performance search using neural networks and other methods Very fast visualisation Working with the user to solve the problem through ‘helpful’ tool sets, such as CBR etc. NESC 2008 Link to other tools Leeds developed CBR under DAME and BROADEN Quick toolset developed by Oxford/Oxford Biosignals 24 (most valuable patent in RR last year!) Re-deployed in other areas of RR and elsewhere SDE-CBR essential for full operation CBR-Matlab and other services in the pipeline within CARMEN NESC 2008 Portal 25 NESC 2008 Enterprise system Enterprise system 27 Looked at service bus’s for integrating diverse services (York) Looked at portal designs that were flexible (now applied in CARMEN) Workflow manager, a resource broker and Service Level Agreement (SLA) manager (Leeds) NESC 2008 WorkBench Workflow Manager (or PXE) Resource Broker Other Services This diagram represents an architecture with a distributed QUICK Feature Provider and XTO sharing a common Distributed Controller with the Pattern Match Service (see notes) Enterprise Service Bus (ESB) CBR Viewer CBR Engine PCT Analysis Visualiser PCT Analysis Visualiser CBR KnowledgeBase PCT Viewer SDE (Limited IRRSIS) QUICK DataViewer GRID Factory Distributed Node 0 ? Distributed Controller PMS Service (XTO) Distributed Node 1 Distributed Node 2 QUICK Feature Provider XTO Centralised Remote Storage Request Broker (SRB) Master node 1 (with MCAT) Database for other data (e.g. RDF, extractions, etc.) New Engine Data 28 PMS Search Database Database for Meta Data Zmod Database Data Orchestrator NESC 2008 Deployment in Rolls Royce Progress 30 Working with a large company is a long slow process Tool is on the desktop within Rolls Royce RR Grid used for CFD etc. Currently looking at blade tip flutter Looking at new deployments with Cybula. NESC 2008 Science Neuroscience: CARMEN Science 32 BROADEN tackled Industry, CARMEN is aimed at Science. Data is often not shared from experiments Individuals duplicate many expensive experiments Should share both data and methods Distributed data and service repositories needed to support this NESC 2008 CARMEN Tackles the scientific application of SDE Builds on the York’s teams 6 years of experience in building these applications. 33 Tom Jackson, Martyn Fletcher, Bojian Liang and Mark Jessop NESC 2008 Freeflow Use of intelligence to improve traffic flow Freeflow 35 A TSB/DfT/EPSRC project, £4.5M Looking at using the SDE tool to look at traffic data from road networks Initial trails in York (Hull Road), London (Hyde Park Corner) and Kent. York trial later this year. NESC 2008 Objectives 36 To use SDE to allow users to explore traffic data Flows (from loops), phases of traffic lights etc. User can then set up event triggers, which are down loaded to the Pattern Match Engine for action NESC 2008 37 NESC 2008 38 NESC 2008 Scaling to other sectors Other sectors though Cybula Stand alone SDE tool (out of the box). Medical Rail industry Pipeline monitoring – for leaks Power Generation 40 Track and carriage monitoring Oil and Gas Visualisation of MEG data (York Neuro-imaging) Analysis of renal data Monitoring of generation equipment NESC 2008 Thanks 41 DAME, BROADEN, Freeflow and CARMEN teams Support of EPSRC, DfT, DTI and industrial collaborators NESC 2008