BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Memento EPFL//
BEGIN:VEVENT
SUMMARY:Process monitoring\, using methods from system identification and 
 chemometrics
DTSTART:20100226T101500
DTSTAMP:20260407T102816Z
UID:c3af758a74948c71140fbb9df5b738148161e1bbb91aeaeda2348cc7
CATEGORIES:Conferences - Seminars
DESCRIPTION:Prof. R.Ergon - Telemark University College\, Norway\nOn-line 
 measurements of product qualities (primary outputs y) from industrial proc
 esses \n(chemical plants\, food industry\, etc.) are often not feasible. I
 nstead\, samples are taken at \nmore or less regular intervals and brought
  to the laboratory for costly and time-consuming \nanalyses. There is thus
  a need for primary output estimation at a high sampling rate\, based \non
  known inputs u and secondary process measurements z (flows\, temperatures
 \, etc.). In my \ntalk I will discuss several related aspects of this: \nF
 or dynamical systems\, we may use Kalman filtering and system identificati
 on methods\, also \nwhen the primary samples are obtained at a very low an
 d irregular sampling rate. Spectral \nsecondary measurements (NIR\, acoust
 ics\, etc.) must then be compressed into principal \ncomponents\, using pr
 incipal component or least squares regression (PCR/PLSR). The \nestimates 
 of y may also be used in\, e.g.\, Smidt controller feedback structures.\nP
 CR and PLSR may also be used directly for primary output estimation\, usin
 g statistical limits \nfor normal process operation. The Hotelling's T2 st
 atistic is then used to see if new z samples \nare acceptably close to the
  normal operating point within the projection space\, while the \nsquared 
 prediction error SPE or Q statistic is used to detect abnormal deviations 
 outside of \nthe projection space. For this purpose new samples are split 
 into z = zmodel + e\, and for \nPCR this is unproblematic. The correspondi
 ng splitting in PLSR is much discussed at the time\, \nand the issue is al
 so complicated by sampling errors in y\, which are often larger than error
 s \nin z.\nThe splitting of z in PLSR\, i.e. the definition of zmodel  and
  e\, also affects the score-loading \ncorrespondence\, and is thus of inte
 rest for fault diagnosis methods.  
LOCATION:ME C2 405 
STATUS:CONFIRMED
END:VEVENT
END:VCALENDAR