Author: Gonzalez-Berges, M.
Paper Title Page
TUCPA04 Model Learning Algorithms for Anomaly Detection in CERN Control Systems 265
 
  • F.M. Tilaro, B. Bradu, M. Gonzalez-Berges, F. Varela
    CERN, Geneva, Switzerland
  • M. Roshchin
    Siemens AG, Corporate Technology, München, Germany
  
  At CERN there are over 600 dif­fer­ent in­dus­trial con­trol sys­tems with mil­lions of de­ployed sen­sors and ac­tu­a­tors and their mon­i­tor­ing rep­re­sents a chal­leng­ing and com­plex task. This paper de­scribes three dif­fer­ent math­e­mat­i­cal ap­proaches that have been de­signed and de­vel­oped to de­tect anom­alies in CERN con­trol sys­tems. Specif­i­cally, one of these al­go­rithms is purely based on ex­pert knowl­edge while the other two mine his­tor­i­cal data to cre­ate a sim­ple model of the sys­tem, which is then used to de­tect anom­alies. The meth­ods pre­sented can be cat­e­go­rized as dy­namic un­su­per­vised anom­aly de­tec­tion; "dy­namic" since the be­hav­iour of the sys­tem is chang­ing in time, "un­su­per­vised" be­cause they pre­dict faults with­out ref­er­ence to prior events. Con­sis­tent de­vi­a­tions from the his­tor­i­cal evo­lu­tion can be seen as warn­ing signs of a pos­si­ble fu­ture anom­aly that sys­tem ex­perts or op­er­a­tors need to check. The paper also pre­sents some re­sults, ob­tained from the analy­sis of the LHC Cryo­genic sys­tem. Fi­nally the paper briefly de­scribes the de­ploy­ment of Spark and Hadoop into the CERN en­vi­ron­ment to deal with huge datasets and to spread the com­pu­ta­tional load of the analy­sis across mul­ti­ple nodes.  
slides icon Slides TUCPA04 [1.965 MB]  
DOI • reference for this paper ※ https://doi.org/10.18429/JACoW-ICALEPCS2017-TUCPA04  
Export • reference for this paper using ※ BibTeX, ※ LaTeX, ※ Text/Word, ※ RIS, ※ EndNote (xml)  
 
TUPHA034 SCADA Statistics Monitoring Using the Elastic Stack (Elasticsearch, Logstash, Kibana) 451
 
  • J.A.G. Hamilton, M. Gonzalez-Berges, B. Schofield, J-C. Tournier
    CERN, Geneva, Switzerland
  
  The In­dus­trial Con­trols and Safety sys­tems group at CERN, in col­lab­o­ra­tion with other groups, has de­vel­oped and cur­rently main­tains around 200 con­trols ap­pli­ca­tions that in­clude do­mains such as LHC mag­net pro­tec­tion, cryo­gen­ics and elec­tri­cal net­work su­per­vi­sion sys­tems. Mil­lions of value changes and alarms from many de­vices are archived to a cen­tralised Or­a­cle data­base but it is not easy to ob­tain high-level sta­tis­tics from such an archive. A sys­tem based on the Elas­tic Stack has been im­ple­mented in order to pro­vide easy ac­cess to these sta­tis­tics. This sys­tem pro­vides ag­gre­gated sta­tis­tics based on the num­ber of value changes and alarms, clas­si­fied ac­cord­ing to sev­eral cri­te­ria such as time, ap­pli­ca­tion do­main, sys­tem and de­vice. The sys­tem can be used, for ex­am­ple, to de­tect ab­nor­mal sit­u­a­tions and alarm mis­con­fig­u­ra­tion. In ad­di­tion to these sta­tis­tics each ap­pli­ca­tion gen­er­ates text-based log files which are parsed, col­lected and dis­played using the Elas­tic Stack to pro­vide cen­tralised ac­cess to all the ap­pli­ca­tion logs. Fur­ther work will ex­plore the pos­si­bil­i­ties of com­bin­ing the sta­tis­tics and logs to bet­ter un­der­stand the be­hav­iour of CERN's con­trols ap­pli­ca­tions.  
poster icon Poster TUPHA034 [5.094 MB]  
DOI • reference for this paper ※ https://doi.org/10.18429/JACoW-ICALEPCS2017-TUPHA034  
Export • reference for this paper using ※ BibTeX, ※ LaTeX, ※ Text/Word, ※ RIS, ※ EndNote (xml)  
 
TUPHA035 Data Analytics Reporting Tool for CERN SCADA Systems 456
 
  • P.J. Seweryn, M. Gonzalez-Berges, B. Schofield, F.M. Tilaro
    CERN, Geneva, Switzerland
  
  This paper de­scribes the con­cept of a generic data an­a­lyt­ics re­port­ing tool for SCADA (Su­per­vi­sory Con­trol and Data Ac­qui­si­tion) sys­tems at CERN. The tool is a re­sponse to a grow­ing de­mand for smart so­lu­tions in the su­per­vi­sion and analy­sis of con­trol sys­tems data. Large scale data an­a­lyt­ics is a rapidly ad­vanc­ing field, but sim­ply per­form­ing the analy­sis is not enough; the re­sults must be made avail­able to the ap­pro­pri­ate users (for ex­am­ple op­er­a­tors and process en­gi­neers). The tool can re­port data an­a­lyt­ics for ob­jects such as valves and PID con­trollers di­rectly into the SCADA sys­tems used for op­er­a­tions. More com­plex analy­ses in­volv­ing process in­ter­con­nec­tions (such as cor­re­la­tion analy­sis based on ma­chine learn­ing) can also be dis­played. A pilot pro­ject is being de­vel­oped for the WinCC Open Ar­chi­tec­ture (WinCC OA) SCADA sys­tem using Hadoop for stor­age. The re­port­ing tool ob­tains the meta­data and analy­sis re­sults from Hadoop using Im­pala, but can eas­ily be switched to any data­base sys­tem that sup­ports SQL stan­dards.  
poster icon Poster TUPHA035 [1.016 MB]  
DOI • reference for this paper ※ https://doi.org/10.18429/JACoW-ICALEPCS2017-TUPHA035  
Export • reference for this paper using ※ BibTeX, ※ LaTeX, ※ Text/Word, ※ RIS, ※ EndNote (xml)  
 
THPHA021 Large-Scale Upgrade Campaigns of SCADA Systems at CERN - Organisation, Tools and Lessons Learned 1384
 
  • R. Kulaga, J.A.R. Arroyo Garcia, M. Boccioli, E. Genuardi, P. Golonka, M. Gonzalez-Berges, J-C. Tournier, F. Varela
    CERN, Geneva, Switzerland
  
  The paper de­scribes plan­ning and ex­e­cu­tion of large-scale main­te­nance cam­paigns of SCADA sys­tems for CERN ac­cel­er­a­tor and tech­ni­cal in­fra­struc­ture. These ac­tiv­i­ties, re­quired to keep up with the pace of de­vel­op­ment of the con­trolled sys­tems and rapid evo­lu­tion of soft­ware, are con­strained by many fac­tors, such as avail­abil­ity for op­er­a­tion and planned in­ter­ven­tions on equip­ment. Ex­pe­ri­ence gath­ered through­out the past ten years of main­te­nance cam­paigns for the SCADA Ap­pli­ca­tions Ser­vice at CERN, cov­er­ing over 230 sys­tems dis­trib­uted across al­most 120 servers, is pre­sented. Fur­ther im­prove­ments for the pro­ce­dures and tools are pro­posed to adapt to the in­creas­ing num­ber of ap­pli­ca­tions in the ser­vice and re­duce main­te­nance ef­fort and re­quired down­time.  
poster icon Poster THPHA021 [1.262 MB]  
DOI • reference for this paper ※ https://doi.org/10.18429/JACoW-ICALEPCS2017-THPHA021  
Export • reference for this paper using ※ BibTeX, ※ LaTeX, ※ Text/Word, ※ RIS, ※ EndNote (xml)  
 
THPHA037 Future Archiver for CERN SCADA Systems 1442
 
  • P. Golonka, M. Gonzalez-Berges, J. Guzik, R. Kulaga
    CERN, Geneva, Switzerland
  
  Funding: Presented work is conducted in collaboration with ETM/Siemens in the scope of the CERN openlab project
The paper pre­sents the con­cept of a mod­u­lar and scal­able archiver (his­to­rian) for SCADA sys­tems at CERN. By sep­a­rat­ing con­cerns of archiv­ing from specifics of data-stor­age sys­tems at a high ab­strac­tion level, using a clean and open in­ter­face, it will be pos­si­ble to in­te­grate var­i­ous data han­dling tech­nolo­gies with­out a big ef­fort. The fron­tend part, re­spon­si­ble for busi­ness logic, will com­mu­ni­cate with one or mul­ti­ple back­ends, which in turn would im­ple­ment data store and query func­tion­al­ity em­ploy­ing tra­di­tional re­la­tional data­bases as well as mod­ern NOSQL and big data so­lu­tions, open­ing doors to ad­vanced data an­a­lyt­ics and match­ing the grow­ing per­for­mance re­quire­ments for data stor­age. 		 
poster icon Poster THPHA037 [7.294 MB]  
DOI • reference for this paper ※ https://doi.org/10.18429/JACoW-ICALEPCS2017-THPHA037  
Export • reference for this paper using ※ BibTeX, ※ LaTeX, ※ Text/Word, ※ RIS, ※ EndNote (xml)