Control System Data Analysis Future Vision Author Axel

Control System Data Analysis Future Vision Author: Axel Voitier CERN EN-ICE

Control Systems § Monitoring & Control problem is solved § Our toolbox: Industrial tools + In-house Frameworks Openlab Workshop – 20 November 2013 2

Control Systems § Monitoring & Control problem is solved Our toolbox: Industrial tools + In-house Frameworks § Architecture § MOON (Monitoring) Supervision layer TN DIM/CMW Process layer PLCs OPC High Voltage Fieldbus Field layer Sensors & Actuators Openlab Workshop – 20 November 2013 3

Control Systems § Monitoring & Control problem is solved Our toolbox: Industrial tools + In-house frameworks § Architecture § Five major installations § Application Win. CC OA Systems Parameters (Million) ALICE 100 3 ATLAS 130 12 CMS 90 10 LHCb 160 10 Accelerator Complex 120 10 § Tons of data acquired & stored § Although, not much processing yet Openlab Workshop – 20 November 2013 4

Approach • Difficult to impose a data analysis tool • Many equipment groups (Cryo, Gas, Vacuum, Machine Protection, etc. ) • Need for a flexible framework • That takes care of common problems • To offer analysis as a service to our users Openlab Workshop – 20 November 2013 5

Integrated with our control systems Data Analysis. Framework Supervision layer Process layer MOON (Monitoring) Data Processing Modules FFT Neural Network (R) Machine Learning CEP (Java) Patterns PLCs (Lab. View) (Watch. CAT) TN DIM/CMW OPC High Voltage Fieldbus Field layer Data collection & Sensors feedback & Actuators Openlab Workshop – 20 November 2013 6

Scalable Data Analysis. Framework Data Processing Modules FFT Neural Network (R) Machine Learning CEP (Java) Patterns (Lab. View) (Watch. CAT) Data collection & feedback Supervision layer MOON (Monitoring) TN DIM/CMW Openlab Workshop – 20 November 2013 OPC 7

Reliable Data Analysis. Framework Data Processing Modules Backup Data Analysis Preventive Alarm Processing FFT Neural Network (R) Machine Learning CEP (Java) Patterns (Lab. View) (Watch. CAT) Data collection & feedback Supervision layer MOON (Monitoring) TN DIM/CMW Openlab Workshop – 20 November 2013 OPC 8

Visualisation Data Analysis. Framework Data Processing Modules Backup Data Analysis Preventive Alarm Processing FFT Neural Network (R) Machine Learning CEP (Java) Visualisation Patterns (Lab. View) (Watch. CAT) Data collection & feedback Supervision layer MOON (Monitoring) TN DIM/CMW Openlab Workshop – 20 November 2013 OPC 9

Analysis flow written by the domain expert Data Analysis. Framework Data Processing Modules Backup Data Analysis Preventive Alarm Processing FFT Neural Network (R) Machine Learning CEP (Java) Expert Visualisation Patterns (Lab. View) (Watch. CAT) Data collection & feedback Supervision layer MOON (Monitoring) TN DIM/CMW Openlab Workshop – 20 November 2013 OPC 10

This is our vision Data Analysis. Framework Data Processing Modules Backup Data Analysis Preventive Alarm Processing FFT Neural Network (R) Machine Learning CEP (Java) Expert Visualisation Patterns (Lab. View) (Watch. CAT) Data collection & feedback Supervision layer MOON (Monitoring) TN DIM/CMW Openlab Workshop – 20 November 2013 OPC 12

Introducing ELVis by Siemens § Configurable analysis flow by user § + It can use custom analysis software § High scalability of analysis processes § From laptop to multi-node cluster § Stream based data processing engine: Storm § No. SQL data storage engine § Web-based visualisation interface § § HTML 5, Data pushed by Web. Sockets Desktop and mobile devices Openlab Workshop – 20 November 2013 13

ELVis: General View CT RTC BAM COM-RU Openlab Workshop – 20 November 2013 14

Integration of ELVis at CERN CT RTC BAM COM-RU Openlab Workshop – 20 November 2013 15

Complicated Data Processing schemes CT RTC BAM COM-RU Openlab Workshop – 20 November 2013 16

ELVis: Web-based Visualisation CT RTC BAM COM-RU Openlab Workshop – 20 November 2013 17

Portfolio of framework § Watch. CAT (Siemens) Semi-automatic discovery of “symptoms” in collections of unstructured logs § Algorithm-driven, human-validated § Next: Event processing § § Drools Fusion (Red Hat) Open-source rules engine § Knowledge-based, temporal reasoning § § ELVis (Siemens) Scalable scaffolding for custom real-time analysis § Visualisation of analysis results § Openlab Workshop – 20 November 2013 18

Conclusion § We have many use cases and users § Various kinds of analysis to perform § We need a flexible analysis environment § Integrated with our monitoring and control environments § Current focus Alarm threshold learning with Drools Fusion § Evaluating ELVis § Watch. CAT for event processing § Openlab Workshop – 20 November 2013 19

Any Questions Thank you for attending! Openlab Workshop – 20 November 2013 20