This presentation by Bengt Lydell discussed the SKI-PIPE database evolution and background; the database structure and content; current database status; and LOCA frequency estimate conducted with the data base. Some of the specific key points from his presentation and subsequent discussion are outlined below:

• Background: The database was motivated to create a tool that would serve both PRA and the PFM/material science practitioners. It’s structured to provide information to completely define the piping systems attributes (design characteristics) and the influence functions (operating history) which govern system failure probability. By thoroughly assessing these features it is possible to determine plant specific estimates of piping system reliability.

• Structure and Content: The database covers pipe failures in commercial nuclear power plants from 1970 to the present.

• It should be stressed that SKI-PIPE only includes failures in piping systems, external to the RPV. Non-piping system failures are not included. Also, SKI-PIPE contains only passive piping failures of metallic piping.

• A pipe failure is defined in the database as any degradation that results in piping repair or replacement.

• Each record in the database is indexed. References to the original data source (e. g., LER report) and supporting information are provided. All the supporting documented is stored electronically.

• The database is organized by reliability attributes (i. e. design features such as material, dimensions) and influence factors (i. e. unique service conditions, including degradation susceptibility).

• When the original record is incomplete (such as an LER), a best effort is made to fill in database gaps by directly contacting the plant operators.

• It is noted in the database when each record consists of multiple flaws at a single component location. However, subsequent data entries are typically associated with only the largest flaw at that location.

• The database includes both surface penetrating flaws and non surface penetrating flaws (i. e., embedded flaws).

• Current Database Status:

• The database is continually being updated.

• The current OECD-sponsored OPDE project has participants from 12 nations. The first year of the three year effort is concerned with adding and validating database entries for each of the member countries from 1998 through 2001.

• Raw data is currently obtained from over 40 different sources

• Applications: Two relevant studies are the determination of LOCA frequencies for the

Barseback-1 plant and examination of IGSCC in Russian graphite moderated reactors (RBMK).

• The Barseback-1 study employed plant-specific attribute and influence functions which were comprehensively developed for all “known and credible” degradation mechanisms.

• The Beliczey and Schulz conditional rupture probability was not used in the Barseback-1 analysis. Instead a Bayesian update of a Jeffrey’s modified non-informative prior was employed.

• The database results have been compared with PFM predictions for welds in certain systems with some success.

• The RBMK studied indicated that the experience today with IGSCC in Russia is similar to US BWR IGSCC cracking experience in the late 70’s to early 80’s, before wide-spread mitigation was adopted

Discussion: The panel asked if they could get copies of the SKI-PIPE database. Karen Gott of SKI indicated that it is possible to distribute a non-proprietary version of the database. This non-proprietary version contains piping failures thru 1998.