QUOVADIS WP 2 A holistic approach towards quality management and classification Prof' Dr'Ing' Sabine

1
QUOVADIS WP 2 - A holistic approach towards
quality management and classificationProf.
Dr.-Ing. Sabine Flamme14. December 2006, Ispra
2
Tasks

• Development of an European database on SRF
  production according to the classification system
• Validation of CEN/TS on SRF classification and QM
  specifica-tion and classification including
  recommendations to TC 343 for the eventual
  revision of the TS before its upgrade to an
  European Standard (EN)

3
Examination of European SRF-data according to the
classification system (relates to CEN TC 343 WG 2)

• Reply
• 90 replies towards SRFs produced in 78 plants
  (from 11 nations)
• Austria (6), Belgium (9), Denmark (1), Finland
  (4), Germany (21), Italy (27), The Netherlands
  (8), Norway (3), Sweden (7), United Kingdom (4)
• (in brackets Number of replies towards SRF per
  nation))
• France Only one statement towards used tyres
  was given
• very different quality
• best case questionnaire filled in completely,
  no requests necessary
• worst case only analysis data or technical
  data available

4
Previous activities

• Check of the returns (examination of
  plausibility) and update of the questionnaires
  (including the results of analyses from external
  data bases f. e. ERFO)
• Data examination with view on plausibility
• Carry out calculations (if necessary, conversion
  from units)
• Categorisation of the data quality with respect
  to the classification parameters (single data, n
  gt 40 single data, N gt 10 and lt 40 single
  data, n lt 10 no examination possible)
• Examination of the classification and the
  classification parameters
• Examination for the class distribution(class
  code e. g. "NCV 3, Cl 2, Hg 2")

5
Summary of evaluable data
6
Mercury Compliance rules for classification

• Class code is established using median and 80th
  percentile based on
• at least the last ten validated measurements or
• at least ten validated measurements per annum
  taken at random
• Prediction method for first classification
• Using the 50 rule in case of more than ten data
  assays are available or
• Using a random generator in case of more than 40
  data assays are available
• 50 rule classification is determined by
  comparing the
• measurements results of 50 of the class limits
  (median
• and/or 80th percentile)

7
Mercury Proceeding of classification

• In most cases insufficient database for
  comparison per annum
• INFA evaluated the median and the 80th percentile
  of all given assays per SRF for the first
  classification
• Statistical evaluation when ten or more assays
  were given (comparison with the half limit
  values) 50 -Rule
• Furthermore evaluation via RND when 40 or more
  assays were given (comparison with the whole
  limit values)

8
Results classification Mercury (via statistic)
Statistical evaluation according to the whole
database per SRF and parameter
9
Comparison results classification Mercury
(statistic vs. RND)
Statistical evaluation according to the whole
database per SRF and parameter Detection limits
not always given
10
NCV and Chlorine Compliance rules for
classification and proceeding

• Compliance rules
• Comparison with the limit values of the classes
  has to be made by the mean of the values of
• at least the last ten validated measurements or
• at least ten validated measurements per annum
  taken at random
• Proceeding of classification
• In most cases insufficient database for
  comparison per annum
• INFA evaluated the mean of all given assays per
  SRF
• Statistical evaluation when ten or more assays
  were given (comparison with the whole limit
  values)
• Additional evaluation via RND when 40 or more
  assays were given (comparison with the whole
  limit values)

11
Results classification NCV and Chlorine (via
statistic)
Statistical evaluation according to the whole
database per SRF and parameter
12
Comparison results classification NCV (statistic
vs. RND)
Statistical evaluation according to the whole
database per SRF and parameter
13
Comparison results classification Chlorine
(statistic vs. RND)
Statistical evaluation according to the whole
database per SRF and parameter
14
Mercury subject to quality management system
n. s. not specified
Statistical evaluation according to the whole
database per SRF and parameter
15
Mercury subject to origin of waste, the SRF were
made from
n. s. not specified
Statistical evaluation according to the whole
database per SRF and parameter
16
Mercury subject to use
n. s. not specified co-incineration only
co-incineration, co-incineration / incineration
co-incineration as well as incineration
Statistical evaluation according to the whole
database per SRF and parameter
17
Conclusions (I)

• Necessity for analytical modification
• Analytical methods for the parameters have to be
  harmonised (necessary for standardised
  estimation)
• Mercury maximum detection limits should be
  givend
• actual state
• for each type of SRF is an adequate class
  available

18
Conclusions (II)

• Necessity to work prCEN/TS 15359 over (have to
  work out for deliverable 2.5), e. g.
• Textual modification
• for NCV and Chlorine addition of a specified
  reference to the consideration of the 95
  confidence interval in contrast to prCEN/TR
  15508 a note is missing here)
• Further validation of the TS after its revision
  and after the analytical modification should be
  done