statistics:MSA procedure on survival tests

Der Titel is kacke aber ... so sei es: Messsystemanalyse zur Bewertung von Reproduzierbarkeit und Wiederholbarkeit bei Lebensdaueruntersuchungen (Übertragung und Anwendung)

actually it is about a MSA (measurement system analysis) on ATL (accelerated life testing)

when

  • done
  • 30.09.2013 Vorschläge eingereicht
  • 30.10.2013 go for: MSA-procedure on survival analysis tests
  • 15.11.2013 check in Publicity handout (title / short intro / 6s experience / picture)
  • next
  • 30.12.2013 check in handout
  • 10.01.2013 check in presentation (finalized)
  • 07.03.2014 calm down and do it

what - short description

Die Eigenschaften von Lebensdauertestverfahren erschweren die Bewertung von Reproduzierbarkeit und Wiederholbarkeit insofern, dass weder dasselbe Teil wiederholt geprüft werden kann, noch Normteile bekannter Lebensdauer zur Verfügung stehen und zusätzlich auch Spezifikationsgrenzen selten hinreichend definierbar sind. Der Vortrag zeigt anhand eines Beispiels, wie die Messsystemanalyse auf einen derartigen Test angewendet wurde, und somit zu einer quantifizierten Aussage über die Vertrauenswürdigkeit der Ergebnisse geführt hat.

presentation

intro

  • background
  • test strategies on reliability tests (table: strategy | informative value | phase of PLM| pro | con | example)
    • censored in time -- delivery / distribution function
    • censored in events --
    • sudden death
    • success run -- only if distribution function is known / available
    • accelerated testing
    • (?)
  • 6s - Measure phase to be closed with clear data on: Vertrauenswürdigkeit der Daten - ggf. MSA plus Sigma Level
  • no such procedure available for survival analysis tests (especially accelerated testing)
  • short description of precise problem / GB project

main

  • Datenaufbereitung
  • initial situation
  • change counting
  • presentation to survival analysis
  • choice of function (histogram to distribution function - model building- referring to additive/minitab road show talk)
  • options (lean usage / when & how & which)
  • pro / con Weibull 3param
  • choice of parameter(s) to analyse
  • list of params (shape, scale, MTTF, probability at 20, percentiles @ minimum error)
  • proposal-definition on which parameter to use for which problem / test / product
  • best practice
  • definition of sample types: product-properties to choose for MSA (as different as possible, or what / when) ?
  • definition of sample size (samples per curve & curves needed) - is shitty complex => proposals
    • visualize/ sim data - items per distribution vs. variation of (chosen) parameter
    • calc sample size = number of curves (=repetitions) (?)

closing

  • validity of procedure (sum up)
  • outlook & going deep
  • need for MSA on SAT (pro / con / limits / open items -> to be discussed )
  • need for "SPC" on MSA (generally) - track repetitions of MSA test (results) over years / repeatability of MSA
  • concept on categorization (MSA per product -- note: Trabbi vs. Jaguar on same motor test bench) ... in what(?) we trust ...

how & why - todo

diff * survival analysis tests (SAT) * Parametric Survival Models.pdf * [http://cran.r-project.org/web/views/Survival.html] * reliability test * ATL (accelerated life testing) * success run = zero failure = bogey test?


Todo