ML20079D048

Engineering Rept, Testing & Analysis of Commercial-Grade Swing Arms in Borg-Warner Check Valves, June 1991

ML20079D048
Person / Time
Site:	Comanche Peak
Issue date:	06/21/1991
From:	Harrington C, Hopkins D, Madden F, Moehlman C TEXAS UTILITIES ELECTRIC CO. (TU ELECTRIC)
To:
Shared Package
ML20079D042	List: ML20079D041 ML20079D048 ML20079D057
References
ER-ME-057, ER-ME-057-R00, ER-ME-57, ER-ME-57-R, NUDOCS 9106270103
Download: ML20079D048 (25)

Text

. ____ _ . _ .. _ _ _ _ _ ... _ _ _. _ _ _ _ _ .

l TEXAS UTILITIES ELECTRIC DESIGN ENGIllEERIliG ORGANIZATIOli l

ENGINEERING REPORT TESTING AND ANALYSIS OF COMMERCIAL-GRADE SWING ARMS IN BORG-WARHER SWING Cl!ECK VALVES j ER-ME-057 Revision 0 l

l June 1991 J

i Prepared by: M' #" Date: 20 b k/

Daniel N. Hopkins

/

Prepared by:

c'2" d m Date: 8-26-9/'

Craig D. Harrington

/

9 Reviewed by: AM! '

<. - * " Date: ' ' W Clailide' K. Mochlman Approved by:

',w

/ /* . py)

/

ie d. Date: 6 -2/ //

, Fred W. Madden 9306270103 910621

{DR ADOCK 05000445 PDR

1 TABLE OF CONTENTS ,

j SECTION SUBJECT

2.0 BACKGROUND

ON SWING ARMS 2.1 Metallurgical Characteristics of 17-4PH 2.2 Description of the Failure i

2.3 Preliminary Evaluation of the Failure l 2.4 Probable Cause of the Failure 4 c I l 3.0 APTECH'S PRELIMINARY ANALYSIS l 3.1 Observations on Intact Swing Armr. l 3.2 Observations on the Failed Swing Arm I 3.3 Corrective and Exploratory Actions i 4.0 EVALUATION OF ALL UNIT 1 AND COMMON ARMS 4.1 Important Inspection Attributes 4.2 Basis for Acceptance Criteria l Calculated Service Stresses l Fracture Toughness critical Flaw Size and Location Other Attributes 4.3 The APTECH/TU Inspection Procedure 4.4 Results of the Inspection 4.5_ Acceptability-For-Service

! 5.0 SWRI SWING ARM TEST PROGRAM 5.1 Statistical Design 5.2 Fracture Toughness Evaluation l

5.3 Measured Flaw Sizes and Locations 5.4 Residual Stress Measurement 5.5 Characterization of Weld Repair 5.6 Miscellaneous Testing l

6.0 TU ELECTRIC'S ACCEPTABILITY-FOR-SERVICE EVALUATION l 6.1 Results adopted from APTECH Evaluations l

6.2 SwRI Evaluation Results l 6.3 Acceptability-For-Service j

7.0 CONCLUSION

S REFERENCES l

l l

ER-ME-057, Rev. O Page 3 of 25 ENGINEERING REPORT

1.0 INTRODUCTION

Prior to receiving the operating license for Comancho Peak Steam Electric Station (CPSES) Unit 1, a four-inch, nonsafety, Borg-Warner swing check valve in lake water service experienced a swing arm failure. TU Electric determined that the failure was caused by stress corrosion cracking (SCC) in a poor quality swing arm. An extensive evaluation was performed to determine whether or not this failure was an isolated event or if there was reason to question the reliability of the swing arms in other Borg-Warner check valves.

The swing arm in every Borg-Warner check valve installed in Unit 1 and Common systems was subsequently evaluated to identify and remove from service any suspect swing arms. The inspection procedure that was developed to find suspect swing arms assessed the arm's overall quality. Dimensional checks and tests of the magnetic permeability identified some of the poor quality swing arms, but the most detrimental attribute in overall swing arm quality was determined to be the presence of surface-connected flaws. Three techniques (10X visual, Liquid Penetrant (LP), and replication) were used to evaluate the surface for such flaws.

Indications discovered using these techniques were evaluated against acceptance criteria which were based on the maximum calculated value of service-induced stress and the lowest expected value for fracture toughness. In addition, the microstructure of the material was subjectively evaluated as an indicator of proper heat treatment.

Since SCC had been determined to be the cause of the observed swing arm failure, TU Electric replaced the original commercial-grade swing arms with investment cast swing arms of upgraded quality in systems that contact-lake water. Installing the investment cast swing arms provides additional margin against failure in lake water systems by reducing both the material's susceptibility to SCC and the likelihood of high residual stresses. These actions mitigated two of the three basic elements associated with SCC. The third element, a corrosive environment, is not practical to change in the case of an open system like lake water.

Results presented to the NRC in the fall of 1989 demonstrated that the swing arms were adequate for service. However, the data collected prior to licensing was insufficient to conclusively demonstrate the long-term reliability of the original swing arms that remained in other plant systems. TU Electric therefore committed to replace the remaining swing arms over three refueling cycles and as they are removed from service, to reevaluate the arms for any evidence of degradation since the original screening.

ER-ME-057, Rov. O Page 4 of 25 As an extension of the post-removal reovaluation effort and to ,

fully address the issue of long-term rollability, TU Electric contracted with Southwest Rosearch Institute (SwliI) to ovaluate the mechanical, chemical, and metallurgical charactoristics of 16 additional swing arms. Thic group of swing arms was selected to reprosent the population of original commercial-grado swing arms that are now installed in Unit 1 and common systems.

The fracture toughness and the actual dimensions of preexisting flaws were measured for the 16 arms. The results clearly demonstrato the original swing arms which remain in Unit 1 and common systems will withstand a design basis accident which results in a' check valvo slam. In addition, SwRI conducted extensive metallurgical cross sectioning that showed no evidence that service conditions have initiated cracks or induced crack growth from prooxisting flaws.

TU Electric has concluded, from these findings, that there is no technical basis for early replacomont of the original commercial-grade swing arms now remaining in CPSES Unit 1 and Common Lystems. Furthermore, the test data shows there is enough

intrinsic margin in those swing arms to justify loaving them in service for the life of the plant.

I l.._.,...__.--..-.. . , , . ~ . _ , _ , _ . , . , . . . _ . _ . . . . - , . . . _ _ . . . . _ , _ _ . . . _ _ . _ _ _ _ . . . _ _ _ . . . _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _

ER-ME-057, Hov. O Page 5 of 25

2.0 BACKGROUND

ON SWING ARMS Borg-Warner supplied TU Electric with swing check valves of various sizes for low pressuro (bolted bonnot) and high pressure (pressure seal bonnot) applications at CPSES. The original swing arms woro purchased as commercial-grado parts because the swing I arms were not considered part of the valvo pressure boundary.

Those swing arms were apparently cast at several different foundries during the early 1970's. Records that document the as-cast chemistry, NDE inspection results, weld repair  ;

procedures, and heat treat fornace time / temperature can not be found. Consequently, thoro is no documentation basis for establishing their quality.

The Borg-Warner check valves had been subjected to various maintenance and inspection activities which included the disassembly of individual valves.

2.1 Metallurgical Characteristics of 17-4PH Borg-Warner drawings show the material for the swing arms is 17-4PH (H1100), which is a high strength stainless stool. The swing arms were to be produced in accordance with the chemistry requirements of Aerospaco Materials Specification ( AMS) 5398 and heat treated to the H1100 condition per Military Specification ,

MIL-H-6875. Alloy 17-4PH is very popular for casting and is 1 I

often selected for applications where high strength, high surface hardness, and good corrosion resistance are needed. It can be rough machined with relative ease in the as-cast condition and then, by propor heat treatment, the part can achieve high strength and good corrosion r'sistance. e The nominal chemical composition for 17-4 PH is 17% chromium, 4% l nickel, and 4% copper. The chromium provides resistance to general corrosion similar to that of the 300-series stainless ,

steels. The relatively low nickel content allows the me.rtensitic structure.to form so this alloy is capable of achieving much higher levels of tensile and yield strength than the austenitic structure of the 300-series stainless steels. The copper is involved in " Precipitation Hardening (PH)," which is another strengthening mechanism. The copper coalesces in sub-microscopic zones creating local stress in the metal matrix which strengthens the alloy. However, these local variations in the chemical composition and stress level offset some of the corrosion resistanco provided by the chromium. This alloy exhibits a slight susceptibility to certain forms of local corrosion attack, such as stress corrosion cracking (SCC), but 17-4PH with an appropriate heat treatment would be expected to perform satisfactorily in the Squaw Creek Reservoir environment.

ER-ME-057, Rev. O page 6 of 25 The heat treatment for 17-4pH is relatively complex and must be followed carefu1Ay to produce the desired results. Thoro are four basic stops:

1) Casting - to form the approximato shape.

2) Solution annealing - to break up the segregation of alloying elements that occurs during solidification of the casting.

3) Quonching - to form a uniform martonsitic structure.

4) Aging heat treatment - to temper the martensitic I structure and precipitation hardon the material. )

l The proporties of this matorial can vary significantly doponding j on the aging temporagure selected for the final heat treatment j stop. The H900 (900 F aging temperaturo) heat treat condition produces the strongost combination of tempered martonsito and i strengthening precipitates (190 kai tensilo / 17g kai yield) but is more susceptible to SCC. The H1100 (1100 F aging temperaturo) heat treat condition (roforred to as overaging) is appropriate in applications where a larger margin for corrosion resistanco is desired and lower strength (140 ksi tensile / 115 kai yield) is acceptable.

2.2 Description of the Failuro on May 31, 1989, a four inch, 150 lb. class, Borg-Warner swing check valvo (1SW-048) installed in the Servico Water system at CPSES exhibited excessive backloakage. Subsequent inspection of the valvo revealed that the swing arm disk boss had fractured radially in two places and that the disk was detached. The disk stud assembly is attached to the swing arm through the disk boss.

When the swing arm and disk are properly assembled the disk will gimbal slightly in the disk boss.

At the time of the falloro, components in the Servico Water '

system had boon exposed to Squaw Creek Lake water for about savon years.

2.3 Preliminary Evaluation of the Failure The broken swing arm from ISW-048 was visually examined by nite Engineering personnel and then the parts were taken to Hurat-Metallurgical Research- Laboratory, Inc. in Euloss,-Texas for additional visual examination and photo documentation. E,oth parts of the broken swing arm woro then transferred to Stcne and Webster Engineering Corporation (SWEC), Boston for furti.sr examination and testing. Analysis of the chemical ccoposition, metallurgical structure, hardness, heat treatrant, and fracturo surface was performed. Two additional intac' <ing arms from

-e~., ~

,y -- .m. , - , . ,,.--~,,,m-_.,_r,,--r,m 3,,,my ., ,,,,%..w,,,,.. , , , _ - . , _ ,-,,m.-.me m- .m%., , , - _ . 1-~-- .. . - -_. - - .

r ER-ML-057, Rev. O Pago 7 of 25 valves 2SW-048 and 2CT-0148 were supplied to SWEC to provido a basis for comparison. The material in thoso swing arms was believed to be notallurgically cound.

SWEC observed that the fractures apparently initiated at small surfacn cracks then propagated along interdendritic formations in this material. The prosenco of the surface cracks and the interdendritic microstructure are clear indications that the casting and heat treatment woro inadequatoly controlled.

SWEC reported that hydrogen assisted cracking may have contributed to the fracturo propagation. The typical sourco for the hydrogen is a corrosion roaction, which suggests that corrosion was involved in the failure. Finally, it was observed that the fractured ends of the portion of the disk boss that had brokon off woro displaced inward. Thn magnitude of this displacement suggests that a significant residual stress had boon present in the disk boss.

l i

2.4 Probablo Cause of the Failure Stress corrosion cracking was determined to be tho most likely cause of the failure. Three proroquisito factors are involved in SCC: 1) chronic stress; 2) a corrosivo environment; and 3) a cusceptibio material. The preliminary evaluation noted that all of those factors woro present. First, thoro was chronic stress l in the disk boss based on the observed spring (displacement) in the broken part of the disk boss. Second, raw Squaw Crook lako

, water is known to contain enough chlorides to promoto SCC.

Third, the material in this-particular. swing arm was susceptible l to cracking because it exhibited preexisting, casting-related, surface defects and an undesirable metallurgical structure.

l

_ ..- - , ..- . . . ~ , . . , - _ , , _ . _

,_.,_,..__.m. ,_,,_,_.,...,.e.,. , , _ , . . . - . - _ . , , , - ~ . _ . . ..,_,.,,~.._....,_v._-_,_.

- - --~- - - -.- .- - - . . - - - - . - - - - - - -

ER-ME-057, Rev. O Pago 8 of 25 3.0 APTECH'S PRELIMIN1.RY ANALYSIS Due to the complexity of the issues and the generic implications, TU Electric retained APTECH to perform an indopondent review of the failuro and the failuro analysis. They were also tasked with evaluating the likollhood of a similar failuro in the other swing arms installed in Borg-Warner check valvos throughout the plant.

APTECH's scope was then expanded to includo development of an action plan to determine the acceptability of the swing arms for service. ,

1 APTECH undertook an extensive study of the fracturo surfaces and the material in the failed swing arm. They also performed laboratory tests on material from several intact swing arms, and l calculated the loads and stressos to which the swing arm is i subjected during a design basis accident which results in a check valvo slam. This work led to an inspection proceduto that was used to evaluato the swing arms in 100% of the Borg-Warner swing check valvos installod in Unit 1 and Common systems.

The APTECH work provided additional evidence that confirmed SCC was the failuro mechanism. As previously stated, SCC typically occurs in the combined presence of a chronic stress, a susceptibio material, and a corrosivo environment.

3.1 observations on Intact Swing Arms The original commercial-grado swing arms were apparently produced as ordinary sand castings with no special quality requirements.

In sand castings, zones of porosity are common (especially in larger castings) and it is standard practico to grind out these porous areas and " repair" them with weld metal. Zones of wold repair were evident in these swing arms. Most of the wold repairs woro shallow but a few wold repairs in some of the larger arms wore through-wall in the area of the disk boss.

These wolds were determined to be sound.

The rough castings were probably machined to the approximate final dimensions before they woro heat treated to the specified H1100 condition. Other than the size of the casting thoro were no intentional-differences in the manufacturing process for the swing arms and thereforo, from the standpoint of materials the swing arms woro expected to be a homogeneous group.

Evaluation of the metallurgical structure of several intact swing arms revealed that some arms were not given an appropriate heat treatment.- This was demonstrated when the matorial from a swing arm with a poor metallurgical structure showed a significant improvement in fracturo toughness when an appropriate heat treatment was applied. Had the original heat treatment been correctly performed, little or no change in fracture toughness would be expected when the heat treatment was repeated. However, a swing arm that was not optimally heat treated may still perform adequately in service depending on the conditions in the service l environment.

1

ER-ME-057, Rev. O Page 9 of 25 Residual stress in the disk boss was evaluated by APTECH in three (3) of the intact swing arms and estimated stresses ranged from slightly compressivo to almost 20 kai tensilo. The evaluation technique involved marking two adjacent locations on the disk boss, cutting the disk boss betwoon the marks, and measuring the relativo displacement in the marks. APTECH did not expect the residual stress in a properly heat treated casting to exceed about 15 ksi. Stress corrosion failures can be strongly influenced by residual stress. APTECH concluded that "ronidual stress may be the most variable (of tho) important attrioute(s) of the investigated failuro modo."

Fracture toughness tests were conducted on the matorial from soveral of the intact arms. Charpy tests were conducted values using published the results wereCompact correlations. converted to K[8sts were lator conducted on tensile several larger arms to confirm the Charpy results. These tests established the fracture toughness values used to determino an

~

"acceptaolo" flaw size. (Soo Section 4.2, Basis for Acceptanco Criteria).

3.2 Observations on the Failed Swing Arm The swing arm disk boss in valvo ISW-048 failed radially in two places. Examination of the two sets of fracture surfaces revealed one pair of the surfaces had thicker corrosion deposits than the other. The fracture surfaces with the thickor corrosion deposit exhibited two distinct zones. The outor zone formed a narrow border of corrosion products around the cross section.

APTECH reported that the corrosion products and the surface topography soon in the outer zone were characteristic of SCC.

The inner zone was relatively free of corrosien products and the fracturo surface was characterized by interdendritic cleavage (brittle) and small areas of dimple rupturo (ductile). The other pair of fracture surfaces did not have a similar " corrosion perimeter" and were dominated by interdendritic cleavage.

l The metallurgical structure in the fulled arm gives amplo evidence to conclude that this material was susceptible to SCC.

An optical examination revealed "sovere" surface and subsurface l casting defects that are potential crack initiation sites. The

! metallurgical structure revealed strong remnants of the dendritic structure and coarse carbide particlas arranged along prior austenite grain boundaries. The relatively weak areas at dendrite interfaces provide crack propagation paths, and the local variations in chemical composition provido sites for corrosion attack. A proper heat treatment would have broken up the dondritic structure and dispersed the carbido particles thus producing a swing arm of material less susceptible to interdendritic cleavage.

l

-. _ - ~ _ - ._. - - - - --. - - - . - . - . - - - -.

ER-ME-057, Rev. O Page 10 of 25 The residual stress prior to failure was estimated based on the inward displacement (closing) observed on the free portion of the broken disk boss. Tne postulated operational loading mechanisms that could impose significant stresses capable of deforming this section of the disk boss would cause it to spring open (e.g.,

wrenching of the disk stud in the boss with one side of the boss fractured.) APTECH estimated that the residual stress in this disk boss was between 40 and 95 ksi tensile. A more precise estimate was not possible because the break did not occur under controlled conditions.

Charpy impact tests were used to measure the fracture toughness of the material in the failed arm. The Charpy results were then converted to K values using published correlations. The resulting K what would d8 e[9acturexpected for toughness values were significantly below this material.

The evidence suggests that the direct cause of the failure was a stress corrosion crack initiated at a surface flaw. As the stress corrosion crack grew it became a more significant stress concentrator, and it also decreased the cross section of the disk boss. Eventually, when the swing arm was subjected to a typical operational load, the concentrated stress caused a predominately brittle failure in the remaining ligament in the disk boss.

Several factors contributed to this failure. Poor control of casting quality resulted in surface defects that provided crack initiation sites. Inadequate heat treatment failed to eliminate the as-cast interdendritic structure which typically exhibits poor fracture toughness and poor corrosion resistance.

These factors resulted in a more susceptible material condition.

The large residual stress provided the driving force for SCC.

Although there is no conclusive evidence that identifies the cause of this high residual stress, indentions that may be evidence of prior mechanical deformation were observed on the disk boss.

A weld repair was present on the disk boss of the failed swing arm. Areas of porosity can be seen in the casting along the margirs of this weld, but the fracture faces do not intersect the area of weld repair or its heat affected zone at any point. Weld repair was considered as a possible source of residual stress.

However, stresses caused by welding would be expected to result in an opening displacement of the broken disk Loss due to weld shrinkage rather than the closing displacement actually observed.

The evidence therefore suggests that the weld repair observed in this boss did not contribute to the failure.

l l-

ER-ME-057, Rev. O Page 11 of 25 3.3 Corrective and Exploratory Actions As a result of the evaluations performed by APTECH and others, the following actions were taken

- All arms in the lake water environment of the Service Water system [six per unit) were replaced with arms of upgraded quality which are expected to provide additional margin against stress corrosion attackt

- The critical attributes were-determined which would conservatively identify the arms with the lowest margin to failure and appropriate acceptance critoria were established for those attributes;

- An inspection program was developed which employed established nondestructive inspection techniques with resolution capabilities commensuratl with the acceptance criterlat

- All remaining arms in Unit 1 and Common systems were inspected and only arms that met the acceptance criteria were returned to service.

These actions were intended to minimize or eliminate the potential for such a failure to recur by addressing as many of the contributing factors as was practical. Changing the lake water environment is difficult at best and detecting the presence of residual stresses in the disk boss does not lend itself to field-applied nondestructive methods. However, by screening the arms using appropriate nondestructive techniques, TU Electric has been able to remove from service any arms exhibiting casting or microstructural features that are indicative of an increased susceptibility to corrosive attack. g The swing arms of upgraded quality were produced in accordance with ASTM Specification A-747, Cb7CU-1 using the investment casting process with a comprehe*', Ave QA program applied through final shipment. Each upgraded arm received a surface examination by liquid penetrant or magnetic particle techniques and was radiographed. Full traceability was maintained and hardness and tensile tests were performed for each-heat treatment furnaco load of each heat of material. Although these arms are fundamentally the same as the original commercial-grade arms, the upgraded quality provides greater margin against the corrosive attack experienced in the failed arm. Installing those arms-in the lake water service applications was deemed to be an appropriate enhancement to assure reliability.

I ER-ME-057, Rev. O Page 12 of 25  !

4.0 EVALUATION OF ALL UNIT 1 AND COMMON SWING ARMS APTECH worked in cooperation with site Engineering to identify the important inspection attributes, define appropriate acceptance criteria, and develop an inspection procedure. The inspection procedure was incorporated into the disposition of NCR 89-7476 and was used to screen all of the swing arms installed in Unit 1 and Common systems Swing arms which failed any part of the screening critoria were removed from service.

Following completion of the screening process, APTECH reviewed all of the results obtained from the inspections, analysis, and materials testing in order to make a determination regarding the acceptability of the remaining original swing arms for continued se rvice .

4.1 Important Inspection Attributes l

The important inspection attributes for the original arms were selected based on the observations and conclusions of the swing ,

arm destructive tests. These attributes were determined to be '

the presence and size of surface-connected flaws, the adequacy of heat treatment, and overall manufacturing quality.

Both fracture toughness and susceptibility to SCC can be severely degraded by a poor heat treatment in 17-4 PH and cracks which exceed certain critical dimensions. Inspection of the microstructure of the material was included as an indicator as to whether the heat treatment history of the swing arm appeared appropriate for 17-4PH in the H1100 condition.

Dimensional and other acceptance criteria related to the overall manufacturing quality were also established to ensure that all aspects of the swing arms were considered in the evaluation of their acceptability for service.

4.2 Basis for Acceptance Criteria Fracture mechanics may be used to determine the potential for growth of a preexisting flaw given the material's fracture toughness and the stress in the vicinity of the flaw. Fracture toughness is a material property (like tensile strength) which may be affected by chemical environment, strain rate, or temperature. Stress in the area of the flaw is the combination of residual stress and stress from service-induced loads.

Flaws can be expected to occur in these swing arms but most flaws are harmless and remain harmless throughout the life of the part.

By determining the lower bound of fracture toughness and the upper bound on stress (along with its lecation), it is possible-to determine the size and location of the " critical" flaw. This

" critical" flaw may then be used as a benchmark for evaluating known or distavered flaws.

ER-ME-057, Rev. O Page 13 of 25 4.2.1 Calculated Service Stresses APTECll performed finito element analyses to evaluate the intensity and distribution of bending stress in a swing arm subjected to the check valve slam resulting from a sudden pipe break. Other less severe operational loadings were also evaluated. The bending force is a function of the angular velocity (inertia) of the swing arm at the time of impact of the valve disk on the valve seat. A pipe break-induced valve slam produces the maximum operational bending stress that can be imposed on a swing arm and is considered to be the design basis accident for the purpose of this analysis.

The finite element n.odel provided the surface stresses that act on surface-connected flaws. The surface-connected flaws were considered the most important in the analysis because in bending, stresses are greatest-at the surface and decline rapidly at deeper locations in the part. Eo significant stress would be expected near the center of the swing arm since bending forces are balanced at the neutral axis.

A conservative estimate of residual stress was then added to arrive at a combined stress value as the basis for determination of an acceptable flaw size.

4.2.2 Fracture Toughness Fracture toughness was measured using Charpy V-notch tests with material from five intact swing arms from Units 1 and 2. These arms were used to aid in development of the inspection procedure and were expected to be representative metallurgically of the population of swing arms installed in the plant at the time of the failure.

Both full-size and sub-size Charpy V-notch specimens were tested.

The-Charpy data was converted to K values using published correlations and the results rangefcfrom 13 to >100 kai/in. The fracture toughness value for the failed swing arm was 20 ksi/in and values over 100 ksi/in occurred in arms that had a predominately austenitic structure.

Fracture toughness testing is potentially strain rate sensitive.

The vary high strain rate of the Charpy test is extremely severe and tends to produce low toughness values. There is also uncertainty in the empirical correlation of Charpy and K values. Compact tensile tests were later run on material from L three large swing arms from Unit 1 and Unit 2. The relatively slow strain rate in compact tension tests performed under standard' conditions tends to result in higher toughness values.

This s?ow strain rate may not adequately represent conditions postulated for the design basis accident, and therefore a factor was applied to correct for strain rate.

_ _ _ . _ _m -

ER-ME-057, Rev. O Page 14 of 25 The very high strain rate of the Charpy test provided a conservative measure of the fracture toughness for the l development of acceptance criteria. The resulting data was also compared to relevant results found in the literature for similar material as a validity check. I l

4.2.3 Critical Flaw Size and Location l Surface-connected cracks were considered the most likel, type of flaw to cause a swing arm failure. Using estimates of the material's fracture toughness and the anticipated maximum combined stress, Linear Eiastic Fracture Hochanics (LEFM) analytical methods were then used to determine the maximum flaw size which would be acceptable. This postulated flaw was also further evaluated as to depth, orientation, and shape. The results were then used to establish an appropriate set of acceptance criteria for flaws in the remaining swing arms.

4.2.4 Other Attributes Heat treatment acceptability was determined to be adequately represented by the extent of austenite-to-martensite transformation within the material. Dimensional requirements for minimum wall evaluations were established on the basis of the finite element evaluation of the swing arms performed by APTECH.

4.3 The APTECH/TU Inspection Procedure The inspection procedure developed was detailed within the

" Exploratory" disposition of NCR 89-7476. The inspection effort was_ focused on surface-connected flaws since such flaws were found to be crack initiators in the observed failure and subsequent analyses supported this conclusion. Additionally, the surface is where the maximum bending stress and the relatively aggressive lake water environment coexist. Three different techniques were selected to afford a range of resolution from relatively-large to extremely tight flaws. A 10X visual inspection of all accessible surfaces was intended to reveal larger flaws and the general quality and condition of the ,

casting. Wet Fluorescent Liquid Penetrant (LP) testing was used l to locate and size surface-connected flaws. Replication of specific areas was used to detect common features such as tight or debris-filled flaws, assess metallurgical structure, and provide a record of these findings.

Although the most critical swing arm quality attributes were related to the condition of the surface, additional attributes were evaluated to screen for unacceptable swing a ms.

Dimensional checks, particularly in the area of tne disk boss, were used to to verify adequate manufacturing practicos.

ER-ME-057, Rev. O Page 15 of 25 Two tests were used to determine the adequacy of the heat treatment. First, the magnetic permeability was measured and if it was too low (i.e., the magnet would not stick), it was assumed that the martensite phase had not formed and the arm was rejected. This test was performed but was not formally described in the NCR. Second, a metallurgist would subjectively interpret the replicas and determine whether of not the microstructure appeared to be appropriate for 17-4PH in the 01100 condition.

The swing arms were inspected with the disk assembly attached.

The disk stud retainer nut is welded in place and the risk of mechanical damage and weld-induced sensitization of the disk stud as a result of disassembly / reassembly was judged to be significant. In addition, APTECH determined that inspection results from the approximately 70% of the accessible surface could be extrapolated to the surface hidden by the disk assembly and leave only a 2-4% risk of missing a flaw.

4.4 Results of the Inspection The swing arm in every Borg Warner check valve installed in Unit 1 and Common systems was inspected using the procedure described in the disposition of NCR 89-7476. The swing arms that passed this screening were determined to have ample margin against failure and were returned to service. Swing arms that failed the screening may have performed satisfactorily in service but were conservatively replaced either with an arm which had passed the screening or with a new arm of upgraded quality.

4.5 Acceptability-for-Service Following the examinations and testing of the failed swing arm (1SW-048), it was concluded that this particular casting was of extremely poor quality. This was a major factor in a unique set of circumstances that led to the observed failure. The 100%

screening performed on the swing arms in all Borg Warner swing check valves in Unit 1 and Common systems demonstrated that this combination of undesirable attributes was unique to the failed arm. While the screening process resulted in several arms being rejected for various reasons, none. exhibited the combination of factors that resulted in the failure of the 1SW-048 swing arm.

The screening was intended to identify swing arms with attributes indicative of a reduced margin to failure and ensured that only swing arms with ample margin to failure were reinstalled.

APTECH's acceptability-for-service evaluation considered the collective results of the failure analysis of the failed arm, the destructive examination of a number of intact arms, literature surveys for reference materials properties, the finite element stress calculations, and the screening program. In addition, a detailed analysis of the operational requirements for each valve

ER-ME-057, Rev. O Page 16 of 25 was also performed to assure that applicable operational requirements were appropriately considered, llaving both determined the critical characteristics required of an acceptable ,

arm and inspected all of the subject arms in accordance with  :

those requirements, APTECH concluded that the swing arns returned to service could be be expected to perform rollably. TU Electric concurred in this assessment.

These efforts had however, produced insufficient data to positively conclude that the remaining arms exhibited suf ficient margin against to corrosive attack and that the arms were therefore suitable for long-term service. Consequently, TU Electric committed to early replacement of these remaining arms coupled with post-removal reinspection of the arms for evidence of degradation since the initial screening.

l i

l

\

l L . . ..- , . _._ _ ___ _ _ _ _ . . _ _ - _ - - . _ _ . - _ _ - _ . _ _ _ _ _ - _ - ._ _ _ _ . .

ER-ME-057, Rev. O Page 17 of 25 5.0 SWRI SWING ARM TEST PROGRAM TU Electric initiated a program with SWRI to address the issues of service-induced flaw growth and the stability of existing flaws in the event of a design basis accident. A representative sample-of swing arms that was large enough to provide t. h!gh degree of statistical confidence was destructively e:.at irsd to determine the fracture toughness and establish the cito of preexisting flaws. In addition, certain other che-ica ,

physical, and mechanical properties and material cor.ditt cus ,are evaluated to fully characterize the population of sing unas.

These results were then combined with the previous ntress analysic by - APTECH to evaluate the long-term acceptability-for-service of the remaining swing arms.

5.1 Statistical Design A total population of 56 original commercial-grade swing an.is remain in Unit 1 and Common systems. Each of thanc arms has been inspected and accepted in accordance with the inspection process described in NCR 89-7476. This current set of swing arms is therefore considered to be fundamentally different from the original set of vendor-supplicd swing arms because the arms which did not meet the screening acceptance criteria were removed from service and are no longer part of the population.

In order to obtain a representative sample for this population,

.several assumptions and constraints were imposed. These included ,

the following:

- Minimize hardware and operational impacts on Unit 1;

- Unit 2 arms may be considered to be representative of-the Unit 1 population if they have been inspected and accepted in accordance with the NCR requirements;

- Swing arm size is not a metallurgically significant variable, however a representative distribution of sizes will be included;

- All previously screened arms which have seen subsequent service shall be re-screened for evidence of flaw I

initiation / propagation occurring since the initial screening;

- If any readily available arms have seen service, attempt to include arms from a representative set of system l environments:

- A " target" of approximately 15 arms was established.

l t

ER-ME-057, Rev. O Page 18 of 25 The distribution of sizes and represented systems of the remaining installed original commercial-grade swing arms was reviewed and a list was developed of the desired number of swing arms of each size and where appropriate, the preferred systems.

If arms of the desired sizes or systems were not readily available from Unit 1 or Common systems, then the arms were obtained from Unit 2.

A sample set of 16 swing arms was finally selected with the '

following distribution:

Nominal Valve Size Number of (inches) Swing Arms 16 2 10 1 8 2 6 3 4 4 3 4 This distribution included valves from the following systems:

Containment spray; component Cooling; Auxiliary Feedwater; Feedwater; Chilled Water; and Domineralized Water.

Systems with original commercial-grade swing arms installed but

- not represented in the sample included:

Service Air; Instrument Air; CVCS/ Boron Recycle; Main Steam; l Spent Fuel Cooling; Heating and Ventilating; and Waste i Processing.

Other than the requirement that the arms selected must have been accepted in accordance with the NCR screening process, and the emphasis on arms which had seen service, no metallurgically significant variables were intentionally employed in selecting the sample arms for testing.

5.2 Fracture Toughness Evaluation Fracture toughness for the sixteen swing arms was measured using compact tension specimens. Each specimen was fabricated and tested in accordance with the requirements of ASTM E399 (Ky ) and The material was expected to exhibit l

ASTM sufficiently E813 (JIc).

nigh toughness characteristics that testing using l

the more complex J test was selected over the simpler K methodology. A vaIOe for K Ic was then calculated from thfc results.

Tests were run with duplicatg specimens from two large-arms at both room temperature and 40 F to evaluate the effects at the lowest expected service temperature. Based on negligible differences in the results at the two temperatures, all subsequent tests were conducted at room temperature.

L

,---.-,,y--+- -

--ere-a, y- t-,--,,,y-,% , -Niw-w,

. g mm -w,'u m t-'--w- $' wW9 e ,y y ,-+u-wa-=-.wwm--p---u=?' -m---my w-=,t t ' 7"e9'

ER-ME-057, Rev. O Page 19 of 25 Specimens were removed from the shank portion of the swing arms at various locations depending on the arm size, the number of specimens needed, and the locations of flaw indications targeted

3 for detailed evaluation.

Three sizes of compact tension specimens were employed to accommodate the range of sizes of swing arms. The larger two are specimen sizes commonly used for fracture toughness testing. The third was a special subsize comp 1 rension specimen designed, fabricated, and tested to obtai dAr ;t measurements of K from the 3- and 4-inch arms. The c 'r dimeneionsofthisshEcimen conformed to the standet1 propos ans described in the ASTM Standards which control the dimensions of the larger standard specimens. In order to validate the results obtained from the subsize specimens, the broken halves from two of the large specimens were machined to the subsize diuensions and tested with comparable results.

The fracture toughness measurements from these sixteen swing arms established the statistical distribution of the fracturn toughness for this population of arms. Based on these results, we can confidently establish the lower bound of toughness for the swing arms remaining in service. This lower bound was determined to be 44 ks1/in based on 95% confidence that 95% of the population exceeds this lower bound. The median value (best estimate) of fracture toughness was 74 ksi/in. These values confirm the fracture toughness results reported by APTECH.

5.3 Measured Flaw Sizes and Locations In order to quantify and characterize flaws in the sample set of swing arms, a methodical evaluation program was developed combining both destructive and nondestructive techniques. The ~

initial steps in this systematic approach involved repeating the 10X visual and the wet fluorescent dye penetrant exam of the screening process. In addition, all sixteen arms were subjected to radiographic examination.

These steps allowed both surface and subsurface indications to be identified as to location and " ranked" according to size or extent. Subsequent sectioning activities were then planned to evaluate in detail both flaws judged to be " typical" and those

" ranked" as the largest or most extensive.

The location for material removal from within each arm for other testing, such as for compact tension specimens, was carefully selected to avoid whenever possible, any indications or flaws which might be selected for further evaluation.

SWRI was given complete freedom to select and characterize the flaws present in the sixteen arms to the extent necessary to confidently report the most limiting conditions which might be expected in the total population of arms. Sectioning of the arm

ER-ME-057, Rev. O Page 20 of 25 was carefully planned to expose indications in sequential planes such that any flaws could be microscopically examined and characterized with regard to their form and size. In addition to the sectioning of selected indications, a methodical approach was established to section the area at the juncture of the disk boss and the shank of the swing arm. This step was to determine whether " tight" cracks or other flaws not detectable by either the screening process techniques or the SwRI NDE vere present.

This systematic approach assured that the results were comprehensive and represented the entire arm.

The flaw identification and characterization efforts are summarized as follows:

- 10X visual inspection, and dye penetrant tests for surface-connected flaws revealed a few scattered pits which were determined to be blunt and were therefore unlikely to grow;

- Radiography revealed that a few of the swing arms (mostly large swing arms) had areas of internal porosity which were metallurgically sectioned and found to be either shrinkage cavities or clusters of small voids associated with the solidification of the casting;

- Routine metallurgical sectioning of the area connecting the disk boss to the shank of the swing arms revealed a few small fissures or cracks. Subsequent analysis determined that one of these previously undetected cracks was the most limiting actual surface-connected flaw found during this testing.

The consequential flaws identified by these examinations were therefore of two distinct types: 1) small fissures or cracks, and

2) shrinkage cavities or clusters of small shrinkage voids.

SwRI characterized the five most limiting flaws, including the shrinkage cavities and void clusters, as surface-connected, remi-elliptical, sharp-edge cracks whose dimensions totally enveloped the volume of the flaw. The flaws were assumed to be located at the point of highest stress and oriented in the least favorab.le direction in the stress field.

This approach represents the flaws in a conservative manner. In the case of a crack, the semi-elliptical c5 ape typically will cover a larger area than the actual flaw ' the case of shrinkage cavities and void clusters, the ussumptions of curface-connected, sharp-edged flaws, and an enveloping volume are extremely conservative. The stresses on the arm are not likely to propagate cracks in the area of cavities and void clusters since these flaws are rounded and not effective as stress concentrators. Furthermore, these flavs are exposed to the lower internal stresses of the arm rather than the higher stresses which occur on the surface.

ER-ME-057, Rev. O Page 21 of 25 Using a value of 60 ksi (from the APTECH analysis) for the tensile stress at the outer fiber due to bending, SwRI calculated stress intensity factors (K ) ranging from 18.6 to 31.1 ksi/in for the five limiting flaws 5.4 Residual Stress Measurement In order to estimate any residual stress present in the disk boss of each swing arm, small punch marks were made on the face of the boss and a saw cut was made between the marks. Any resulting displacement was then measured. The maximum opening displacement was 0.003-inch and the maximum closure was 0.004-inch. Based on a first approximation of the residual stress, ten of the sixteen arms exhibited stresses less than 1.0 ksi. Of the remaining six arms, the maximum tensile stress was 3.6 ksi and the maximum compressive stress was 3.8 ksi.

5.5 Characterization of Weld Repair Weld repairs were known to exist in the commercial-grade swing arms. In order to better evaluate the impact of these repairs on the acceptability-for-service of the arms, each arm was etched to reveal the presence and extent of repaired areas, and specific repair areas were selected for sectioning and metallurgical examination. Chemical analysis and the response to the etchant indicate that Type 308 austenitic stainless steel filler material was used in these repairs. The microstructures of the heat affected zone (HAZ) associated with several repair areas indicate that some were properly heat treated following the repair and others were not. In all repairs sectioned, no specific defects such as voids, lack-of-fusion, or underbead cracking were observed.

The distribution of repair areas included the shank of the arm and the disk boss. In most cases, especially in the shank, the repairs were shallow. However, there were instances of through-wall repairs in the disk boss area. No evidence of residual stress induced by weld repair was observed in the disk boss.

5.6 Miscellaneous Testing SwRI also obtained chemical analyses and performed hardness measurements for each arm. To verify tensile properties, four tensile specimens were machined from arms with sufficient dvailable material and tested. The results demonstrated that variations exist in the composition and materials properties of the arms, but the fundamental assumption that this sample represents an essentially homogeneous metallurgical population was validated.

1

ER-ME-057, Rev. O Page 22 of 25 6.0 TU ELECTRIC'S ACCEPTADILITY-FOR-SERVICE EVALUATION TU Electric has reviewed the APTECH evaluation and extracted portions of their work for use in this current assessment of the long-term reliability of the swing arms. The results obtained from SwRI have been combined with APTECH's work to develop a positive conclusion regarding the acceptability-for-service of the original, commercial-grade swing arms.

6.1 Results Adopted from APTECH Evaluations Several of the topics covered in the APTECH report apply to the long-term evaluation of acceptability-for-service and were adopted including:

- The finite element model analyses that provides the intensity and distribution of surface stress experienced during the design basis accident and more typical operational evolutions;

- The strain rate correction factor that is applied to the fracture toughness test results to more conservatively represent the strain rate expected in the design basis accident;

- The results of the field inspections for the swing-arms which established the population of swing arms that was represented in the sample evaluated in the SwRI work.

Although no attempt has been made to refine the results and assumptions reported by APTECH, TU Electric believes that there is considerable inherent conservatism in the assumed angular velocity when the valv clams shut during the design basis accident. In the finite element analysis, this angular velocity produces a maximum bending stress on the order of 60 ksi. This bending stress value was provided to SwRI as a basis for determining the stress intensity values associated with the observed flaws.

6.2 SWRI Evaluation Results The SwRI work characterized both certain material properties and the bounding flaws in a representative sample of the population of swing arms that are now installed in Unit 1 and Common systems. The fracture toughness results obtained by SWRI provide a statistically significant basis for determining the expected lower bound of fracture toughness for the population of swing arms now installed in Unit 1 and Common systems. These results are in excellent numerical agreement with the relevant APTECH results.

I 1

ER-ME-057, Rev. O Page 23 of 25 In order to conservatively account for the effects on fracture toughness which result from the difference in strain rate between the test and actual service conditions, the lowest expected K was further reduced. APTECHdeterminedthatfortheextremelf',

high strain ratos conservatively assumed to result from the valve slam associated with a pipe break, the dynamic fracture toughness, K would be 80.5% of K . This results in a lowest expected K ID 3

• 8 "*

Flaw size, shape and location were determined with extensive sectioning and metallurgical examination. All surfacer of the swing arm were available for inspection and there were no limits imposed on SwRI that could prejudice their flew characterization efforts. The maximum stress intensity factor determined from these flaws was 31.1 ksi/in.

6.3 Acceptability-For-Service TU Electric has evaluated the technical issues associated with the acceptability of the original commercial-grade swing arms and the results of the extensive testing, inspection, and analytical work undertaken to resolve those issues. Several specific areas of particular importance to an overall determination of the arm's acceptability-for-service are summarized in the following paragraphs.

The ability of a swing arm to withstand the stress of a violent valve slam such as that induced by the occurrence of a design basis accident, can be demonstrated through the application of the analytical methods of fracture mechanics. The worst caso stress intensity factor of 31.1 knifin determined by SwRI may be directly compared to the lowest expected dynamic fracture toughness of 35.5 ksi/in. The fracture toughness exceeds the stress intensity by almost 15%. This demonstrates that despite significant conservatisms in each area of the analysis, the swing arms would be expected to exhibit sufficient margin against fracture in the remaining valve installations at CPSES.

Weld repairs were observed in roughly half of the swing arms tested at SwRI but there is no evidence that the weld repairs degraded the reliability of the swing arms in any way. The higher stressed area of the shank only exhibited shallow weld repairs while the lower stressed boss area contained some through-wall repairs. Those repairs which received a proper heat treatment following welding would be expected to exhibit acceptable properties in the heat affected zone characteristic of 17-4PH. Repairs that did not receive proper post-repair heat treatment would have locally affected the properties by applying an overaging or resolutionizing heat treatment to the heat affected zone. Either of these effects has a tendency to produce a heat affected zone which exnibits higher toughness and lower strength than the base metal.

l

ER-ME-057, Rev. O Page 24 of 25 The 308 SS filler metal properties reflect significantly lower strength than 17-4PH but also significantly higher toughness.

Since the calculated maximum surface stresses in the disk boss do not excoad the yield strength for 308 SS, it can be concluded that the filler material is sufficiently strong to perform satisfactorily in service. Therefore, the through-wall weld repair observed in the disk boss will not result in degradation in the performance of the arm.

SwRI-observed that a martensitic microstructure dominated the base metal of every arm as evidenced by the macroetching response of the exterior surface and as a feature of the metallurgical sections used in the flaw evaluation phase. No consequential variations in the basic martensitic structure were observed by SwRI within a given swing arm. From this it may be concluded that no significant differences in properties, such as fracture toughness, would be expected at different locations in the arm.

Reliability of the commercial-grade swing arms is further assured since two of the three factors associated with SCC have been mitigated. The commercial-grade swing arms have.all been removed from systems in contact with lake water and replaced with arms with grtater margin against SCC. Furthermore, the negligible residual stresses reported by SwRI in all of the sixteen arms they examined c?early indicate that damaging levels of chronic stress are not likely in the swing arms remaining in the plant.

Based on these factors it is reasonable to conclude that SCC will not be a problem for the remaining commercial-grade swing arms.

Extensive metallurgical sectioning was performed at SWRI to check for flaws that the nondestructive methods failed to detect. All sixteen of the cwing arms were examined for evidence of cracks and crack propagation. The valves containing the swing arms have been installed for a number of years and while some of these valves were exposed to actual service conditions, others may have seen no service at all. However, all of the flaws observed were associated with fabrication processes as opposed to being service-induced, and there was no evidence of in-service growth of any flaw. This indicates that service-induced degradation is not occurring.

On the basis of all of the evidence presented, TU Electric has concluded that the commercial-grade swing arms remaining in service in Unit 1 and Common systems at CPSES will perform reliably for the life of the plant. '

ER-ME-057, Rev. O Page 25 of 25

7.0 CONCLUSION

S Since the initial discovery of the failed swing arm in the Service Water system in May of 1989, TU Electric has devoted significant resources to determine the cause of the failure and '

ensure that such a failure doesn't reoccur in any of these check valves. Each of the original commercial-grade swing arms remaining in Unit 1 and Common systems has been subjected to and passed a detailed screening inspection. The inspection techniques employed were selected to specifically identify those arms with attributes indicative of a reduced margin to failure.

Every arm that failed this screening inspection has been replaceu with either another that did pass the screening or more likely,

-with a new swing arm of upgraded quality.

Recently, additional metallurgical testing and evaluation was completed. The results have demonstrated that the earlier decision to return these arms to service was technically correct.

Additionally, the results clearly demonstrate that there is a sound technical basis for allowing them to remain in service for the life of the plant.

REFERENCES:

1. Hurst Metallurgical Research - Laboratory, Inc. , "A Preliminary Non-Destructive Metallurgical Failure Analysis of a Check Valve Swing Arm," June 22, 1969

2. Stone & Webster Engineering Corporation, P.O. 665-71882,

" Evaluation of Failure and/or Casting Flaws on the Disc Swing Arm of Valve 1SW-048," June 23, 1989

3. Stone & Webster Engineering Corporation, P.O. 665-71882,

" Swing Arms-Evaluation of Failure (1SW-048) and Material Condition (2SW-048 and 2CT-0148)," July 7, 1989

4. APTECH Engineering Services, Inc., " Failure Analysis and Service Suitability of Check Valve Swing Arms," December 1989, (CPSES VL-4090)

5. APTECH Engineering Services, Inc., Letter Report

" Examination of Replicas from Borg-Warner Check Valve Swing Arms from Comanche Peak," January 14, 1991, (CPSES VL-5164)

6. Southwest Research Institute, " Analysis of Check Valve Swing Arms," March 1991, SwRI Project No. 06-3893-100, (CPSES VL-5792)

rll il1 '

gl7 ANALVSIS OF CHECK VALVE SWING ARMS ll:

Task 1

• fll Evaluation' of Commercial Swing Arms

. Prepared by -

l: H. C. Burghard, Jr. -

M. L. Bartlett

-lC cg

FINAL REPORT-fE. SwRI Project-No. 06-3893-100-ll to Ll TU Electric Co.

Comanche Peak S.E.S.

,: g ' P.O. Box- 1002 '

Glen Roso,: TX 76043 -

Dg L!' MarchL199_1 l

n l .

L tRo -

SOUTHWEST RESEARCH INSTITUTE Lg y "g ~

S AN ANTONIO HOUSTON

.W: g A ,/J DETROIT W AS HIN GTON, DC

- .- - - - . - . - - - - - - -