Information Notice No. IN 90-79: Failures of Main Steam Isolation Check Valves Resulting in Disc Separation

                                UNITED STATES
                        NUCLEAR REGULATORY COMMISSION
                    OFFICE OF NUCLEAR REACTOR REGULATION
                           WASHINGTON, D.C.  20555

                              December 20, 1990


Information Notice No. IN 90-79:  FAILURES OF MAIN STEAM ISOLATION CHECK 
                                      VALVES RESULTING IN DISC SEPARATION 


Addressees:

All holders of operating licenses or construction permits for nuclear power 
reactors.

Purpose:

This information notice is intended to alert addressees to potential problems 
involving the design and location of main steam isolation check valves 
(MSCVs) that could result in disc separation as a consequence of fatigue 
failure of the post (or stud).  It is expected that recipients will review 
the information for applicability to their facilities and consider actions, 
as appropriate, to avoid similar problems.  However, suggestions contained in 
this information notice do not constitute NRC requirements; therefore, no 
specific action or written response is required.

Description of Circumstances:

On October 8, 1990, the Tennessee Valley Authority (the licensee) noted that 
three of four MSCVs had failed at the Sequoyah Nuclear Power Plant, Unit 1.  
Each failure involved the disc separating from the swing arm.  In 1982, the 
disc on one of the same Sequoyah MSCVs had become disconnected, and in 1983, 
the disc on a main steam isolation valve (MSIV) at the Joseph M. Farley 
Nuclear Plant, Unit 2 (effectively the same design as the Sequoyah MSCVs) 
became disconnected, too.  (See "Discussion" below.) 

In March 1990, before the recent MSCV failures, the licensee inspected an 
MSCV at Sequoyah Unit 1 in response to Electric Power Research Institute 
(EPRI) guidance (see below).  The licensee found that this MSCV was severely 
worn where the surface of the swing arm collar met the surface of the disc.  
The valve post that connects the disc to the swing arm was worn at the 
unthreaded portion that meets the swing arm collar; and both washers, 
originally located where the threaded and unthreaded portions of the post 
meet, were missing.  As a result of these findings, the licensee inspected 
the other three MSCVs and found that two of them had similar levels of 
degradation.  To correct the problem, the licensee, in consultation with the 
valve manufacturer, modified the posts of all three affected MSCVs by 
machining down the unthreaded portion of the posts, adding weld buildup using 
high-hardness filler material, and remachining the posts to their original 
diameter.  They also redistributed the stresses at the post-to-disc 
connections by reducing the thread torque preload on the post from 2800 to 
2400 foot-pounds. 


9012140086 
.

                                                            IN 90-79 
                                                            December 20, 1990
                                                            Page 2 of 4 


In September and October 1990, personnel at Sequoyah Unit 1 heard loud noises 
coming from the steam piping.  A radiograph revealed that an MSCV disc had 
separated from its swing arm.  Further investigation revealed that the discs 
in two other MSCVs had also separated from their respective arms.  The 
licensee found that the detached disc in one MSCV had lodged against its 
valve stop, and the detached discs from the other MSCVs had traveled several 
hundred feet through the piping and had lodged just upstream of the turbine 
throttle valves.  All valve posts had broken at a machined fillet where the 
portion threaded into the disc joins the unthreaded portion that passes 
through the swing arm collar (see Figure 1).  In addition, an inspection of 
the MSCVs in Unit 2 revealed cracks similar to those that led to the 1982 
failure at Unit 1.  The Unit 2 valves had not received the weld buildup 
modification noted above. 

Discussion:

The MSCVs at Sequoyah Unit 1 were manufactured by Atwood & Morrill Co., Inc.  
These valves are 32-inch, articulated, swing check valves with a single-post 
design and are located just downstream of the main steam isolation valves 
(MSIVs).  All of the MSCVs that experienced disc post failures are actually 
welded to their corresponding MSIVs; the MSCV with the disc still intact was 
located 2.5 pipe diameters downstream of its corresponding MSIV and was not 
severely worn.  The NRC staff believes that although the valve location may 
have made the MSCVs susceptible to high-cycle, low-stress fatigue failure, 
the weld buildup modification in 1990 appears to have so accelerated the rate 
of fatigue wear that the posts failed within seven months.  

EPRI studied the mechanisms that cause check valves to degrade and the 
methods to prevent this degradation.  The results of this study are 
documented in EPRI NP-5479, "Application Guidelines for Check Valves in 
Nuclear Power Plants," January 1988.  The EPRI study indicates that placing 
check valves just downstream of a source of turbulence could cause failure 
and is "deserving of priority attention."  Flow disturbances just upstream of 
a check valve can cause discs to flutter and subject valve internals to 
cyclic loads and premature failure.  

The backstop design may be another possible contributor to the MSCV failures 
at Sequoyah Unit 1.  The configuration is such that a single backstop makes 
contact with the disc edge.  The swing arm is beveled where it contacts the 
disc, permitting oscillation of the disc that is further exaggerated by a 
40-mil clearance between the post and swing arm collar.  This arrangement 
allows a maximum resultant disc oscillation of up to approximately 0.25 inch 
at the rim of the disc.  In the case of Sequoyah Unit 1, the effects of this 
design were magnified by having the post built up with filler material; this 
modification concentrated the stresses on the post where it meets the disc 
and accelerated the failure of the post.  Atwood & Morrill recommended, as 
corrective action, modifying Sequoyah's single-backstop MSCV design to a 
three-point backstop design that may help limit lateral disc oscillation.  
The NRC staff has identified that other Atwood & Morrill and Schutte & 
Koerting swing check valves with single posts may be of this design and may, 
therefore, be susceptible to this mode of failure.  Other valves with a 
similar design may also be affected.  

.

                                                            IN 90-79 
                                                            December 20, 1990
                                                            Page 3 of 4 


In previous instances, MSCV discs have separated from their respective swing 
arms when the post(s) failed before approaching its design fatigue life.  For 
instance, the previous Sequoyah Unit 1 failure occurred in 1982 when the post 
came unscrewed from its disc when a tack weld connecting the post to the disc 
failed.  At Farley Unit 2, in 1983, an Atwood & Morrill, 32-inch MSIV disc 
separated when the posts failed.  Although neither of these failures involved 
the 1990 weld buildup modification performed at Sequoyah Unit 1, both cases 
involved the articulated check valve design that helps ensure proper 
disc-to-seat alignment, but also allows flexure of the disc during flow 
conditions while the valve is on the open backstop.  This flexure, caused by 
flow disturbances, exposed the posts of both the Sequoyah Unit 1 and Farley 
Unit 2 valves to cyclic loads and, in the case of Farley Unit 2, resulted in 
excessive wear and eventual fatigue failure of the posts.  

It appears that any check valve with disc flexure is susceptible to cyclic 
fatigue failure of the posts.  Such features as single-post design, backstop 
contact with the disc edge instead of the disc nut, and valve location near a 
source of turbulence, increase the susceptibility of a check valve to cyclic 
fatigue failure.  The result may be that the post unscrews from the disc, 
unscrews from the nut, or simply fails; and the disc separates.  An example 
of such a failure is a main feedwater regulator check valve (manufactured by 
Pacific Valve Company) disc separation at the San Onofre Nuclear Generating 
Station, Unit 1, in 1985, when the post failed because of high flow velocity 
and valve installation near a source of turbulence.  (Other recent examples 
of failures of check valve internals are discussed in Information Notice 
90-03, "Malfunction of Borg-Warner Bolted Bonnet Check Valves Caused by 
Failure of the Swing Arm", dated January 23, 1990, Information Notice 89-62, 
"Malfunction of Borg-Warner Pressure Seal Bonnet Check Valves Caused by 
Vertical Misalignment of Disk", dated August 31, 1989, NRC Bulletin 
No. 89-02, "Stress Corrosion Cracking of High-Hardness Type 410 Stainless 
Steel Internal Preloaded Bolting in Anchor Darling Model S350W Swing Check 
Valves or Valves of Similar Design", dated July 19, 1989, and Information 
Notice 83-54, "Common Mode Failure of Main Steam Isolation Nonreturn Check 
Valves", dated August 11, 1983.) 

Among the corrective actions taken by the industry in response to these check 
valve failures are:  tack welding the post to the disc, better disc nut 
locking devices, interference threads for the post, multiple-backstop 
designs, and multiple-post designs.  Despite these actions, check valves 
continue to fail prematurely.  Most recently, a potential problem was found 
at Sequoyah Unit 2 during their 1990 refueling outage, where an MSCV had 
cracks in its post and disc near the post-to-disc weld similar to those that 
led to the 1982 post-to-disc weld failure at Unit 1.  These cracks occurred 
despite the corrective actions--i.e., reweld of the disc and increased torque 
preload on the post--taken by the licensee in response to the 1982 post 
failure.  

Check valves are used widely in the nuclear industry and can play a key role 
in the safe operation of the plant.  EPRI has taken steps to improve the 
longevity of these valves by providing guidance for improving check valve 
reliability.  The NRC staff also has developed an action plan (NUREG-1352) 
aimed at improving check valve reliability.  An underlying concern with check 
valve reliability appears to 
.

                                                            IN 90-79 
                                                            December 20, 1990
                                                            Page 4 of 4 


be that certain features, such as valve installation location and backstop 
design, may increase the susceptibility of check valve internals to premature
cyclic fatigue failure. 

This information notice requires no specific action or written response.  If 
you have any questions about this matter, please contact one of the technical 
contacts listed below or the appropriate NRR project manager.




                              Charles E. Rossi, Director
                              Division of Operational Events Assessment
                              Office of Nuclear Reactor Regulation


Technical Contacts:  Angie P. Young, NRR
                     (301) 492-1167

                     Yun-Seng Huang, NRR
                     (301) 492-1417

                     Edward Girard, RII
                     (404) 841-4186


Attachments:  
1.  Figure 1.  Articulated, Swing Check Valve with Single Backstop
2.  List of Recently Issued NRC Information Notices
.ENDEND