ML17228B226

Generic Upper Shelf Values for Linde 1092,124 & 0091 Reactor Vessel Welds, Final Rept

ML17228B226
Person / Time
Site:	Saint Lucie
Issue date:	06/30/1995
From:	ABB COMBUSTION ENGINEERING NUCLEAR FUEL (FORMERLY, ASEA BROWN BOVERI, INC.
To:
Shared Package
ML17228B225	List: ML17228B224 ML17228B226
References
CEN-622, NUDOCS 9508020111
Download: ML17228B226 (34)

Text

St. Unit 1 Dock i o. 50-335 FPL L-95-214 ENCLOSURE COMBUSTION ENGINEERING OWNERS GROUP CEN-622 Final Report GENERIC UPPER SHELF VALUES FOR LINDE 1092, 124 AND 0091 REACTOR VESSEL WELDS Task 839 Prepared for the C-E OWNERS GROUP June 1995 rst508020111 st50725 PDR ADODK 05000335 I P PDR J R, HDK ABB Combustion Engineering Nuclear Operations PQUDIjD Copyright 1995, Combustion Engineering, Inc.

LEGAL NOTICE This report was prepared as an account of work sponsored by the Combustion Engineering Owners Group and ABB Combustion Engineering.

Neither Combustion Engineering, Inc. nor any person acting on its behalf:

A. makes any warranty or representation, express or implied including the warranties of fitness for a particular purpose or merchantability, with respect'o the accuracy, completeness, or usefulness of the information contained in this report, or that the use of any information, apparatus, method, or process disclosed in this report may not infringe privately owned rights; or B. assumes any liabilities with respect to the use of, or for damages resulting from the use of, any information, apparatus, method or process disclosed in this report.

Comhastion Engineering, Xnc.

Table of Contents List of Tables List of Figures 1.0 Introduction

2.0 Background

3.0 Influence of Flux Type On Weld Deposit 4.0 Analysis of Upper Shelf Energy Data 12 4.1 Analysis of Welds For Each Flux 13 4.2 Analysis of Variance 13 5.0 Summary 8c Conclusions 14 6.0 References 16 Page 2 of 30

P I

Relative Flux Compositions 18 Comparison of The Effects of Fluxes Used for Submerged Arc Welding of RPV Weld Seams 18 Manganese/Silicon Ratio of Weld Metals by Flux Type Comparison of Bare Weld Wire & Weld Deposit 20 Chemical Analysis Charpy USE Data for Welds Using Linde 124, 21 1092 & 0091 Flux Statistical Results of Upper Shelf Energy with 23 Welds Using Linde 124, 1092 and 0091 Flux Analysis of Variance for Welds with Linde 124, 1092 & 0091 Flux Effect of Weld Flux Type on Manganese and Silicon Distribution Histogram of Data for Linde 124 26 Histogram of Data for Linde 1092 27 Histogram of Data for Linde 0091 28 Notched Box and Whisker Plots for Linde 124, 29 1092 & 0091 Flux Cumulative Distribution Function for Linde 124 & 1092 Flux 30 Page 3 of 30

GENERIC UPPER SHELF VALUES FOR LINDE 1092, 124 AND 0091 REACTOR VESSEL WELDS CEOG TASK 839

1.0 INTRODUCTION

The purpose of this task is to establish the basis for the use of generic upper shelf Charpy toughness values for Combustion Engineering (CE) fabricated reactor pressure vessel welds. Generic values are necessary to meet current USNRC requirements for welds fabricated prior to implementation of the requirements of 10 CFR 50 Appendix G (2). Upper shelf energy (USE) values were not determined explicitly for many sets of weld consumables for vessels contracted before 1973. Therefore, generic values of initial USE must be established using the set of available measurements. This report presents an evaluation of those data, and it provides both the statistical and metallurgical bases for the generic values of initial USE for the CE vessel welds.

2.0 BACKGROUND

The Charpy impact test is used to measure the toughness of reactor vessel materials.

The measurements consist of the absorbed energy (i.e., impact energy), the lateral expansion (an indication of ductility), and the fracture appearance. These provide evidence of the toughness and ductility of low alloy steel plate and weld material used in the fabrication of CE reactor pressure vessels. The impact toughness can be characterized by an "S-curve" behavior as a function of test temperature. At the lowest temperatures, the material fractures in a brittle mode with relatively little absorbed energy. This is termed the lower shelf, toughness region. At the highest temperatures, the material fractures in a ductile mode with considerable absorbed energy. This is termed the upper shelf region. Between the lower and upper shelves is the transition region in which the material exhibits a transition from brittle to ductile behavior. The fracture appearance provides a measure of the volume of material fracturing in a cleavage versus a ductile mode. In the upper shelf region, the fracture appearance is fully ductile and the absorbed energy and lateral expansion are at or near their maximums. At temperatures below the upper shelf, the fracture becomes mixed mode with an increasing amount of cleavage fracture as the lower shelf is approached.

A definition of USE for reactor vessel materials has been established 'in ASTM Standard Practice E 185-82 (1) as follows:

Page 4 of 30

"upper shelf energy level - the average energy value for all Charpy specimens (normally three) whose test temperature is above the upper end of the transition region. For specimens tested in sets of three at each test temperature, the set having the highest average may be regarded as defining the upper shelf energy."

"Charpy transition curve - a graphic representation of Chaxpy data, including absorbed energy, lateral expansion, and fracture appearance, extending over a range including the lower shelf energy (less than 5% shear),

transition region, and the upper shelf energy (greater than 95% shear)."

This definition is being used widely to compute USE for a given base metal or weldment.

USNRC regulations (10CFR50, Appendix 6) requires that the USE of all reactor vessel beltline materials be measured before plant operation and that the licensee demonstrate that each material meets specific miniinum values before and after exposure to fast neutron irradiation (2). The USE toughness requirements of Appendix G based on Charpy impact tests is 75 ft-lb minimum before and 50 ft-lb minimum after irradiation. The requirement for irradiated USE can be met based on post-irradiation surveillance program measurements of CE fabricated welds (3) or based on predictions, but the latter requires the availability of an initial USE. The practice of developing a full Charpy transition curve for all beltline welds was not implemented until after Appendix 6 was first issued in 1973. Prior to that time, the ASME Code requirement was to perform three Charpy tests at 10F and measure only absorbed energy for a given set of weld consumables (heat of weld wire and lot of weld flux).

Therefore, insufficient data exist for determining an initial USE for many welds which were deposited prior to the implementation of the Appendix G requirements. In numerous instances, full Charpy curves were developed for test plate welds and RPV surveillance program welds. Full Charpy curves are available for weld materials qualified subsequent to the issuance of Appendix G requirements. Therefore, a moderately sized database does exist which could be used to develop generic values of initial USE in order to demonstrate compliance for those cases in which the data were never generated.

The practice instituted by CE in response to Appendix G is to perform weld material certification (WMC) tests for each set of weld consumables; e.g., produce a weldment with a specific heat(s) of weld wire and lot of flux, perform a simulated post-weld heat treatment on it, machine test specimens from it, test the Charpy specimens to develop a full Charpy curve, and perform the other required tests. This practice was also followed prior to 1973 for many weld test plates and surveillance program weldments.

Therefore, the older data could serve as the certification tests (i.e., the WMC) for other welds made with the same weld wire heat and lot of flux. This approach was used in numerous cases to comply with the 10CFR50, Appendix G requirement for initial USE. In certain cases, the USE data were applied to welds fabricated with the same wire heat and using an equivalent welding process but a different lot of flux. This was Page 5 of 30

considered to be reasonable because an "equivalent" weld process would entail the use of the same types of consuinables, meaning the same type of wire and flux. A logical extension of this was to use the available data to establish generic USE values for a group of welds fabricated using an equivalent process.

CE used primarily four types of weld fiux for reactor vessel beltline weld fabrication during the period of interest (nominally 1964 to 1973). These were ARCOS B-5, Linde 1092, Linde 124 and Linde 0091..Very few values of USE are available for the ARCOS B-5 welds and, therefore, aren't considered further in this report. USE data for the Linde flux welds, however, are available to the extent that development of one or more generic values is possible. The available USE data show that the welds made using Linde 0091 flux had the highest initial USE, and welds made using Linde 80 flux (the type used by another vessel fabricator for reactor pressure vessel welds) had the lowest initial USE. The Linde 1092 and Linde 124 flux weld USE values were typically in between the Linde 0091 and Linde 80 flux welds. Furthermore, the data grouped by flux type fell within fairly distinct ranges. This distinction was not unexpected because these "neutral" flux types were known to affect the weld puddle differently during welding; i.e., each fiux varied in the extent to which constituents were added to or removed from the weld puddle.

on the need to establish generic values of initial USE for vessel weldments and

'ased the observation that USE varied relative to the type of welding flux used, this task was undertaken. Two specific aspects are addressed in the evaluation which follows. The first is to perform statistical analyses to generate best estimate values of initial USE for each flux type and to demonstrate their uniqueness. The second is to provide physical evidence in defense of using separate values of initial USE for each flux type. (Note:

The most graphic evidence for a weld flux influence on initial USE is the obvious difference between welds deposited using Linde 80 fiux versus the non-Linde 80 beltline welds deposited by CE. This fact should not be overlooked even though the evaluations which follow focus on differences between the non-Linde 80 welds.)

3.0 INFLUENCE OF FLUX TYPE ON WELD DEPOSIT Fluxes used in submerged-arc welding are granular, fusible, mineral materials containing oxides of manganese, silicon, titanium, aluminum, calcium, zirconium, and magnesium, and other compounds. Some fluxes may contain intimately mixed metallic ingredients to deoxidize the weld pool or add alloying elements to the weld deposit, or both. The flux is deposited over the welding area and is melted by the heat of the arc. In the molten condition, the flux blankets the weld metal and shields the molten weld pool from atmospheric containiiiation. Submerged-arc welding fluxes are generally chemically neutral with respect to the weld metal and must have other characteristics to enable the welding process. All fluxes are likely to produce weld metal deposits of somewhat different composition than that of the weld, wire being used. The variations in chemical composition that occur are due to chemical reactions of the flux with the welding arc, Page 6 of 30

and, if present, due to the metallic ingredients that may have been added to the flux.

Changes in welding parameters, such as arc voltage or travel speed, will change the quantity of fiux interacting with the weld which may affect the resulting chemical composition and or the mechanical properties of the weld deposit (4,5).

Submerged-arc welding fiuxes are produced in three basic forms: prefused, bonded and agglomerated. Prefused fluxes are made by mixing the various oxide ingredients, melting and solidifying, then crushing to the desired particle size. The advantages of fused fluxes are that:

1) They are extremely homogeneous in chemical composition,

2) Small particles (fines) can be removed to produce uniform particle size without affecting composition,

3) They are not hydroscopic, eliminating special storage or drying requirements,

4) Portions that are unfused after welding can be reused without affecting particle size or composition, and

5) They are suitable for high travel speeds.

The main disadvantage of fused fiuxes is that deoxidizers and eleinents for alloy additions to the weld cannot be readily incorporated in the flux due to the high temperature processing involved in fusing the flux.

Bonded fluxes are produced by mixing all of the desired proportions of ingredients and then adding silicates (e.g. water glass) to bond the ingredients together. The mixture is then crushed and screened to the desired particle size. Advantages of bonded fiuxes are:

Deoxidizers and metal alloying elements can be added since they are processed at low temperatures,

2) The flux density is lower, permitting a thicker layer of flux in the weld zone,

3) The solidified slag is readily chipped Qom the weld beads after welding.

There are two principal disadvantages of bonded fluxes. Fines cannot be removed without altering the composition of the flux. Fines are undesirable because remixing is required to distribute them which adds to the cost of welding. Bonded fiuxes are more likely to absorb moisture, thus requiring drying prior to use to avoid affecting the weld process. Moist or damp flux wiO result in porosity and cracking of the weld metal.

The third basic type is agglomerated fiux. This type is similar to the bonded flux except that ceramic binders are used in lieu of the silicates used in bonded flux. Since the ceramic binder requires high temperature processing, the use of deoxidizers and alloying elements is limited as in the fused fluxes.

Welding fluxes are also characterized by the manner in which they interact with the welding process as either "neutral," "active," or "alloy." These terms describe the flux behavior, and generally indicate the amount of manganese and silicon that transfers from Page 7 of 30

flux to weld metal (10). Neutral fluxes are defined as those which will not produce any significant change in the weld metal manganese and silicon content as a result of a large change in the welding arc voltage or arc length. Neutral fluxes are primarily used for multiple pass submerged arc welding of heavy sections, as in RPV weld seams. Neutral fluxes contain little or no deoxidizers and rely on deoxidizers in the weld wire to deoxidize the weld pool. While neutral fluxes have the ability to maintain the composition of the weld when welding conditions are changed, the composition of the weld deposit will be affected by the type of flux being used. Some oxides in the flux may break down under the heat of welding releasing oxygen which can decarburize the weld puddle, resulting in a weld deposit that may have lower carbon content than the starting electrode or wire. Other fluxes contain manganese silicate which can also decompose under heat adding manganese and silicon to the weld metal, even though they are not present as intentional alloying elements in the flux. Elements such as chromium Rom the weld wire can also be affected by the action of neutral fluxes on the weld pool. These factors have to be addressed by the welding engineers when selecting weld wire and flux combinations for submerged arc welding. The changes in chemical composition that occur are fairly consistent for a given type of neutral flux over a range of variation in welding parameters. However, the changes will vary, sometimes considerably, between different types of neutral fluxes, Active fluxes contain intentional additions of manganese or silicon or both. These are added as deoxidizers to provide increased resistance to porosity and weld cracking resulting from contamiiiation by the base materials being joined. Substantial amounts of manganese and silicon can be transferred to the weld metal by such fluxes (10). Because of this action of the flux on the welding process, cleanliness of the material and containuiants such as oxides, oils or grease on the base metals are not as critical as when using neutral fluxes which do not have strong deoxidizing action on the weld pool. The manganese and silicon contents of welds deposited with active fluxes will be affected by changes in welding parameters. Increases in manganese or silicon increase strength levels of weld metal but may lower impact properties. Because of the tight welding controls that would be required to maintain uniform composition and properties in multi-pass welds using these types of flux, active fluxes are more typically used to make single pass welds.

Alloy fluxes, as the name implies, are those which add specific alloying elements to the weld deposit. They are used with, carbon steel electrodes to make alloy steel weld deposits. The desired alloying elements are added as ingredients to the flux which then mix with the weld metal to form an alloyed deposit.

The basis for selecting the weld wire and flux type for vessel welding depended on a variety of important factors:

1) Ability of welds to meet the required strength and toughness requirements (now including the 10 CFR 50 Appendix G upper shelf energies),

2) Availability in the large quantities required, Page 8 of 30

3) Ease of detaching slag from the weld beads,

4) Ability to produce weld deposits with miiiimum number of defects and frequency of necessary repairs, and

5) The relative cost of the flux as part of the welding process.

The ability to meet the required strength and toughness requirements is affected by the chemical composition of the weld deposit and by the welding parameters. Both the filler wire and the flux composition can alter the weld metal composition, affecting weld bead morphology, microstructure and mechanical 'properties (6). However, the specific alloying content of the weld deposit is less of a factor than the flux type in influencing notch toughness. Modification of the flux composition, which removed silicon and oxygen and reduced inclusion content, was found to have a much greater influence on the toughness properties of submerged arc welds than modifying the alloy content (6).

C-E shops used neutral fluxes for submerged arc welding of reactor pressure vessel (RPV) weld seams, including ARCOS B5, Linde 1092, Linde 0091 and Linde 124.

ARCOS B5, an agglomerated type flux, was used in the fabrication of several older vessels. Its use was discontinued in the mid-1960's when difficulties making consistent welds were attributed to variations in the flux being supplied. Over a period of time, the CE shops used the three Linde fluxes which are all fused neutral type fluxes. The fluxes were selected to minimize charges to the weld deposit chemistry. There were no metallic elements intentionally added to any of these flux types. The primary purpose of the flux was to provide the shielding of the deposit required by submerged arc welding to prevent contamination of the weld. Fluxes were preferred which limited silicon, oxygen and inclusion content in the deposit to improve toughness properties.

Linde 1092 was the flux type selected after problems arose with the ARCOS B5 type flux. The changeover to Linde 1092 initially resulted in difficulty in meeting the tensile strength requirements. To address this problem and meet production schedules, nickel was added to the weld deposit. This nickel addition welding process was discontinued after the shop was able to procure nickel alloyed welding wire. Vessel welds were fabricated by both of these welding processes using Linde 1092, up until the time the Linde company discontinued production of this flux type.

Linde 0091 flux was the next type to be qualified and used by the CE shops. Problems with the consistent performance of the flux, similar to those experienced with the ARCOS B5, also occurred with Linde 1092 and 0091. This led to the qualification of Linde 124 to obtain better consistency. Experience showed that, although higher toughness properties were obtained with the Linde 0091 flux, fewer weld indications were encountered with the Linde 124 flux which reduced the required number of weld repairs.

Many of the indications reported from radiographic examinations of Linde 0091 welds turned out to be "ghost indications" (i.e., not real defects). However, the presence of the indications still necessitated actions which impacted the shop production schedule. The improved performance of Linde 124, coupled with the fact that it was readily available and less expensive, resulted in its use for welding of many vessel components (11).

Page 9 of 30

One measure of the effectiveness of the flux in maintaining the cleanliness of the weld deposit is the Basicity Index (Bl). One formula for estimating BI from the flux composition (cited in Reference 6) is given as:

GaQ +Mrna + + pp +~a~Mi +

'A(AlgOq +TiOg + ZrOQ 'iOq+

In general, it has been found that the higher the basicity index of the flux, the cleaner the weld deposit will be (fewer nonmetallic inclusions). Oxygen recovery by the weld metal has been reported to be a function of BI (12). However, the index alone is not the sole indicator of oxygen potential. The flux components which tend to form suboxides in the presence of the weld plasma are in the denominator of the formula. When the fraction of these flux components is small, the BI becomes large even though the resultant volume of suboxides (and, therefore, inclusions) is small. The basicity index can be considered as a measure of the tendency of flux components to dissociate into suboxides and oxygen in the weld plasma, which can contaminate the weld pool. It has also been reported that high-flux basicity will reduce weld metal oxygen levels with a resultant increase in toughness properties (6,7). The relative flux compositions for Linde 80, Linde 124 and Linde 0091 are shown in Table 1 (8). The BI values for the ARCOS B5 and Linde fiuxes used in RPV fabrication are shown in Table 2, along with qualitative comparisons of the effect of weld fiux type on several weld parameters. The Linde 80 flux type shown in Table 2 was not used by the C-E shops for RPV welding. Linde 80 was used by other vessel manufacturers and is included in Table 2 just for purposes of comparison to the Linde flux types used by C-E. As seen in Table 2, the basicity index, as determined from the relative composition and the formula above shows a uniform trend of increasing BI value for the flux types with improved upper shelf energy.

Another factor influencing the toughness properties of submerged arc welds is the oxygen content of the welds. The oxygen level of weld metals is influenced by many factors such as the deoxidizers present (silicon, manganese, aluminum, etc.), welding process and welding conditions. CaF> and TiOz additions were found to be beneficial in improving weld metal toughness. The Linde 80 and 124 fluxes are very similar, except for the increased fluoride'content of Linde 124. Linde 124 welds exhibit significantly better upper shelf properties than Linde 80 deposits. Oxygen contents were lower and silicate inclusions were fewer and smaller in welds using fluxes modified with such elements. Oxygen content in the weld metal affects the weld metal transformation behavior, influencing the resulting microstructure of the weld and thus, influencing the toughness properties (6). One study reported that the weld metal oxygen content dropped from 900 ppm to 300 ppm for a basicity index change from 0.5 to 1.5 and then tended to remain constant with increasing flux basicity (7).

Page 10 of 30

Manganese and silicon are other elements that can vary in the weld metal as a function of the type of flux being used. High silica slags may result in increased silicon deposits.

Manganese in the weld wire will tend to be oxidized by the welding process. The resulting composition of the weld will, therefore, tend to have lower manganese content than the wire. Depending on the composition of the flux being used, manganese can also be picked up by the weld metal from the molten flux to replace some or aH of the manganese being oxidized. Manganese can be restored to the weld pool either by additions of metallic manganese to the flux or additions of manganese oxide (MnO).

Manganese-silicates can break down in the heat of the weld introducing manganese and silicon to the weld pool. Free oxygen generated by these reactions can also enter the molten weld pool, contributing to the oxygen content and resulting loss of toughness described above. These slag/metal reactions may also significantly affect the resultant dissolved oxygen and inclusion content which influence the mechanical properties as described above (9). Increases in manganese content can increase the yield and tensile strength of weld metal but generally, does not have a significant effect on the toughness properties. Manganese in the weld metal helps to promote an acicular microstructure in the weld metal which in turn produces good cleavage resistance.

The primary role of silicon in the welding process is in the deoxidation of the weld pool.

Silicon can also reduce the inclusion volume of the weld metal when it is kept low in proportion to the manganese content. Increasing the manganese/silicon ratio helps reduce the volume of inclusions which can have beneficial effects on tensile strength and resistance to microvoid coalescence (9).

The manganese/silicon ratios of weld deposits made using the three Linde fluxes are shown in Table 3. Figure 1 shows the manganese and silicon distributions for Linde 0091, 1092 and 124 fluxes. Comparison of the distributions reveals very similar composition between Linde 1092 and 0091 but higher manganese and significantly higher silicon in the Linde 124 weld deposits. Table 4 shows comparisons of the bare wire and weld deposit chemical analyses using two different flux types with the same heat of weld wire. Four cases (A, B, C and D) are given in which two or three flux types were used to deposit welds with the same wire heat. In general, differences in weld deposit analysis can be attributed to the flux types used. Comparisons in Table 4 confirm that for a given heat of weld wire, 'significant differences in deposit chemistry occur depending on the type of flux used. The most notably differences are for manganese and silicon as described above. For example, Weld Wire Heat A welded with Linde 0091 and 124 fluxes show more silicon is picked up and more manganese is recovered in the weld with Linde 124 when compared to the Linde 0091 deposit. Weld wire heat B, in Table 4, provides a comparison of three Linde fluxes. There is an increasing level of manganese recovered in the weld deposit as the flux basicity index (see Table 2) increases. The silicon level is nominally the same for 0091 and 1092 in the weld. Linde 124 flux resulted in a higher silicon content than the other two fluxes. Chemical'nalyses for wire heats C and D compare the composition of Linde 0091 and Linde 1092 welds with the composition from Linde 80 flux using the same heats of weld wire. Significantly higher manganese and silicon contents resulted from the use of Linde 80 flux. As shown in Page 11 of 30

Table 2, the effects of the Linde 80 flux on weld deposit tend to produce the lowest USE values of any of the four Linde fluxes considered.

Other elements, in particular phosphorus, sulfur, nitrogen and other impurities, could have significant effects on weld metal toughness, but no clear trends were observed for any of these elements as a function of the type of flux.

In summary, Linde 80, 1092, 124 and 0091 are all neutral fluxes, but they differ in the degree of neutrality. This difference is proportional to the basicity index; the higher the index, the cleaner the weld and the lower the oxygen content of the weld. Cleanliness and low oxygen contents tend to produce higher USE results. Chemical analysis results for weld deposits support the literature findings regarding variability of manganese and silicon by flux type. Therefore, since Linde 1092, 0091 and 124 differ in basicity index, the welds deposited using these different flux types can be expected to differ in cleanliness, oxygen content, manganese and silicon content, and general microstructure.

4.0 ANALYSISOF UPPER SHELF ENERGY DATA The purpose of this section is to establish best estimates of the USE for welds made with three different Linde weld fluxes. In, addition, the statistical uniqueness of the USE database for the three flux type welds is to be demonstrated. The applicable Charpy USE data (13) for welds using Linde 1092, 0091 and 124 fluxes are compiled in Table 5. The analysis presented below establishes the mean and standard deviation of the USE for each type of weld. Analysis of Variance was used to demonstrate independence of the weld types. The tests used were the F Test (16), Kruskal-Wallis Test (17), and Komogorov-Smirnov Test (15).

10 CFR 50, Appendix G (2) currently requires that reactor vessel beltline materials must exhibit an initial USE of at least 75 ft-lb. For many vessels, compliance with this requirement cannot be established because many welds lack USE measurements.

Generically-based USE values could be used in these cases to enable compliance.

Currently, the use of generic mean initial values for other properties in lieu of specific material data is acceptable to the NRC. For example, generic values for initial RTNor may be used for RT~ analysis ifmeasured values are not available. Therefore, there is a high degree of confidence that the use of generic USE values will be acceptable to the NRC as long as it can be demonstrated that both physical and statistical evidence exists to support unique USE values for each flux type. The objective of this analysis is to establish the statistical basis. The physical evidence was discussed in the previous section.

To determine generic USE values for these three weld flux types, ABB reviewed the original fabrication records maintained in our Chattanooga facility. The available Charpy data contained in both the ABB database and in the Power Reactor Embrittlement Database, Version 2 (18) were collected.

Page 12 of 30

Table 5 gives the Charpy USE data for the welds for each flux type. In Section 4.1, the three flux types are analyzed assuming that they are three different populations. The analysis of variance that justifies treating the three flux types as separate populations is provided in Section 4.2.

4.1 Figures 2 through 4 give the histograms for the Charpy data for the three treatments of welds. Using a normal distribution, mean, variance, standard deviation, upper quartile, and lower quartile values are given in Table 6. Figure 5 gives a notched box and whisker plot for the three weld types. The whiskers represent the first quartiles. The box represents 50% of the data. The center line of the box is the median value and the ends of the notch represent the 95%

confidence interval of the median. For Linde 0091 flux welds, the data points outside the whiskers represent outliers. Welds with Linde 0091 flux are shown to have significantly higher USE than welds with the other two fluxes.

The mean, sum and variances for the three fluxes given in Table 6 were calculated with both EXCEL (Version 5.0, Ref. 14) and STATGRAPHIC PLUS (Version 6, Ref. 15). Consistent results were obtained from both codes.

4.2 A one way analysis of variance was performed to demonstrate that the flux types (treatments) produced independent populations. The most common test is the F Test (16). Variations of the samples about the mean for each treatment are compared with variations of the values between the treatments. The null hypothesis states that ifthe treatments are equal then each treatment has the same mean and variance and the variance between treatments is zero. The F Test compares the ratio of the mean variation between the treatments with the mean variation within the treatments. This ratio is compared with the ratio predicted in a F distribution for the appropriate degrees of freedom and confidence level.

Table 7 gives the results of this analysis and concludes that the three flux types are independent.

The F Test was performed with both EXCEL and STATGRAPHICS. The equations for the F Test were taken from Ref. 16 and programmed into EXCEL.

In a separate calculation the initial data were analyzed using STATGRAPHICS.

Both analyses produced the same results for sums, treatinent means, grand means, and F ratios.

Figure 5 shows that welds with Linde 124 and Linde 1092 flux are closer to each other than welds with Linde 0091 flux. A separate F Test was performed for these two flux types and concluded that they were independent. In addition, a Page 13 of 30

Kolmogorov-Smirnov Test of these flux types was performed. These results are given in Table 4. The Kolmogorov-Smirnov Test compares the cumulative distribution functions (CDF) of two types (see Figure 6) and looks for the maximum vertical distance between the CDFs. Again, the test concludes that the USE of the Linde 124 and Linde 1092 flux type welds are independent. Separate two group analysis of the welds with Linde 0091 flux and the other fluxes was not performed because Linde 0091 was shown to have significantly higher USE (Figure 5).

The Kruskal-Wallis Test (17) (also called the H Test) for the three flux types was also performed. This test compares the sum of the ranking of the samples for the treatments instead of the means. It, therefore, is less sensitive to the effects of outliers. It is a non-parametric method that does not require normally distributed samples. The Significance level (0) was less than the confidence interval (0.05) and, therefore, the null hypothesis is rejected and the three flux types are assumed to be independent.

5.0

SUMMARY

R CONCLUSIONS CE reactor welds were fabricated primarily with four types of welding flux, Linde 1092, Linde 0091, Linde 124 and ARCOS BS. ARCOS B5 data were too limited in number for statistical analysis. CE did not employ Linde 80 flux for reactor vessel beltline welds.

2. Following the issuance of 10CFR50, Appendix G, welds were tested to determine the upper shelf energy (USE) toughness. Prior to the issuance of Appendix G, USE measurements were not routinely obtained. This necessitated a generic approach to establishing USE.

3. The statistical analysis results are given in Table 6 and summarized below:

LiM~Z Lin~991 Mean USE, ft-lb 102.3 112.4 150.3 Standard Deviation 9.4 8.1 25.3

4. The analysis of variance concluded that the populations of USE values for the three weld flux types were independent.

5. The metallurgical analysis of five different flux types concluded that each was unique and, therefore, could result in different properties.

Tables 1 through 4 and Figure 1 describe these differences which are Page 14 of 30

summarized as follows. Each flux type is neutral, but they vary in the degree of neutrality (passivity). The upper shelf energy is related to the basicity of the weld flux:

L why Lmh~ LiadaJ99Z Lin~991 Mean USE, ft-lb 69.7(Ref. 19) 102.3 112.4 150.3 Basicity Index 1.09 1.25 (not 1.47 available)

The correlation is likely to be a function of the relative cleanliness of the weld deposit. The higher the Basicity Index, the cleaner the weld deposit (in terms of, e.g., inclusion content) can be assumed to be.

Variations in alloy content should have a smaller influence on the USE of the weld deposit. Certain elements such as manganese and silicon which vary consistently as a function of flux type should, therefore, have only a small effect on USE.

6. Generic values of initial upper shelf energy toughness based on flux type are clearly justified. The different flux types have unique metallurgical effects in the weld puddle and in the resultant weld deposit. The resultant upper shelf energy varies in proportion to the basicity index of the welding flux. The USE values for each flux type are found to represent independent populations based on analysis of variance.

Therefore, generic values of USE based on flux type are valid for use with CE fabricated vessel weldments.

7. The following conservative values of initial USE are recommended for use in satisfying 10 CFR 50, Appendix G requirements:

Elux TZW Linde 124 83.5 Linde 1092 96.2 Linde 0091 99.7 The preceding are the mean USE minus two standard deviations based on the statistical analyses described in Section 4.0.

Page 15 of 30

6.0 REFERENCES

1) "Standard Practice for Conducting Surveillance Tests for Light-Water Cooled Nuclear Power Reactor Vessels", ASTM Designation E185-82, American Society for Testing and Material.

2) "Fracture Toughness Requirements", Title 10, Code of Federal Regulations, Part 50, Appendix G.

3) "Evaluation of Low Charpy Upper Shelf Energy Materials", CEOG Report CEN-604, Revision 1, August 1993.

4) V W , American Society for Metals, Metals Park, Ohio, 1977.

5) SFA 5.23 (ANSVAWS A 5.23), Appendix A, "Guide to AWS Specification for Low Alloy Steel Electrodes and Fluxes for Submerged Arc Welding," ASME Boiler & Pressure Vessel Code,Section II, Part C, American Society for Mechanical Engineers, New York, NY, 1992.

6) J. E. Indocochea and D. L. Olson, "Relationship of Weld-Metal Microstructure and Penetration to Weld-Metal Oxygen Content," J. Materials for Energy Systems, Vol. 5, No. 3, December 1983, pp. 139-148.

7) T. W. Eagar, Welding Journal, Vol. 57, 1978, p. 76s.

8) L-TEC Submerged Arc Welding Fluxes, Material Safety Data Sheet, ESAB Group, Inc. Form 7960, Dated Mar. 31, 1993.

9) K. E. Dorschu, "Factors Affecting Weld Metal Properties in Carbon & Low Alloy Pressure Vessel Steels," Welding Research Council Bulletin No. 231, WRC, New York, NY, October 1977.

10) G. D. Uttrachi, "Selecting Fluxes for Submerged Arc Welding: What You Should Know About Them," Welding Design & Fabrication, Vol. 78, Feb. 1978, pp. 78-79.

J. D. Varsik, "Evaluation of the Effect of Weld Flux on Irradiation Sensitivity,"

Combustion Engineering, Inc. Report MCM-79-180, December 1979.

12) C. S. Chai and T. W. Eagar, "Predictions of Weld-Metal Composition During Flux-Shielded Welding," J. Materials for Energy Systems, Vol. 5, No. 3, December 1983, pp. 160-164.

Page 16 of 30

13. C.D. Stewart, "Generic Upper Shelf Energy for CEOG Vessel Welds," ABB Report PENG-95-042, March 21, 1995.

14. EXCEL Version 5.0, Microsoft Corp.

15. STATGRAPHICS Plus, Statistical Graphics System By Statistical Graphics Corporation, Version 6, 1992.

16. , Spiegel, M.R., McGraw-Hill Book Company, 1975.

17. , Freund, J,h.E., Prentice-Hall, Inc., 1984.

18. "PR-EDB: Power Reactor Embrittlement Database, Version 2", Oak Ridge National Laboratory Report ORNL/TM-10328/R2 (bKJREG/CR-4816), January 1994.

19. A.S. Heller and A.L. Lowe, Jr., "Correlations for Predicting the Effects of Neutron Radiation on Linde 80 Submerged-'Arc Welds", Babcock & Wilcox report BAW-1803, January 1983.

Page 17 of 30

TABLE 1 RELATIVEFLUX COMPOSITIONS FLUORIDES Cao (e.g. CaFg Mg0 MnO FeO Si02 A1203 TiOq

<50 <15 <2 <10

<10 <2 15

<7 <15 <2 15 COMPARISON OF THE EFFECTS OF FLUXED USED FOR SUBMERGED ARC WELDING OF RPV WELD SEAMS RELATIVEPROPERTIES TYPE BASICITY BASICITY OXYGEN OXYGEN & UPPER OF OF INDEX POTENTIAL SILICON SHELF FLUX SLAG IIBItl OF SLAG CONTENT ENERGY LINDE 80 LOWEST 1.09 HIGHEST HIGHEST LOWEST LINDE 124 ARCOS BS LINDE 1092 LINDE 0091 HIGHEST 1.47 LOWEST LOWEST HIGHEST Linde 80 flux not used in the fabrication of C-E RPV welds.

Linde 1092 A ARCOS B5 Fluxes are no longer produced, therefore relative compositions could not be obtained to calculate the BI.

The sequence of these two flux types in the relative order of properties shown has been estimated qualitatively.

Page 18 of 30

TABLE 3 MANGANESE/SILICONRATIO OF WELD METALS BY FLUX TYPE Sample Size Average Ratio Mn/Si Standard Deviation 7.11 1092 6.56 2.58 108 3.08 0.72 Page 19 of 30

TABLE 4 COMPARISON OF BARE WELD WIRE R WELD DEPOSIT CHEMICALANALYSES WIRE HEAT Si Mo Ni BARE WIRE 1.91 0.07 0.021 0.016 0.54 N/A 0.08 0091 1.24 0. 17 0.017 0.020 0.49 0.01 0.09 0.35 0.019 0.019 0.47 0.01 0.13 WIRE HEAT Si Mo BARE WIRE 1.94 0.05 0.012 0.010 0.55 N/A 0.12 1.29 0.33 0.011 0.016 0.52 N/A 0.32 1.12 0.18 0.009 0.013 0.51 N/A 0.30 1092 1.27 0.14 0.010 0.015 0.49 N/A 0.27 WIRE, FLUX TYPE HEAT Si Mo Ni BARE WIRE 1.86 0.02 0.025 0.015 0.53 N/A 0.10 C

1.12 0. 14 0.015 0.014 0.53 N/A 0.27 1.40 0.52 0.018 0.013 0.52 N/A 0.29 WIRE HEAT Si Mo Ni BARE WIRE 1.91 0.05 0.010 0.012 0.57 0.05 0.09

1. 15 "

0.21 0.012 0.021 0.56 N/A N/A 1.43 0.50 0.012 0.018 0.52 N/A N/A Page 20 of 30

TABLE 5 CHARPY USE DATAFOR WELDS USING LINDE 124, 1092 dh 0091 FLUX

=='.'USE=== ,".':;USE,"", '""'i,'.USI;-';'.:

j'Values'.-". .:-.-Valuesl"; Nol=::.--'.;-,USE-'.;g F=.,:Values".b

= -"'Heat::5o.;=;: -:; ;==(ft-.lb)..=': =:=."=-.Beat:-'No',"='-'-"'-,". ,"-""(ft'31)' &".".:',Heat

'.;:-:,Values';.";-.,:.(ft-':,lb)';-..:

~He'at'No.,"-'-'. ="=(fthm b).::-". :-'-"-:-'Heat;No,":-."';,;.

90069 125.7 90154 102.3 1P3571 126.0 83637 237.3 87603 143.3 90067 124.3 90209 100.7 12008/21935 121.0 4P6519 224.0 90077 143.3 89827 118.3 91762 100.3 W5214 120.0 90067 215,0 90130 143.3 83637 116.7 89476 100.0 W5214 118.0 90211 212,0 90209 142.7 4P7927 116.3 90128 99.3 3277 118.0 4P6052 200.3 33A277 142.7 4P7927 115.7 3P7317 98.3 305424 112.0 90077 187.0 89022 141.3 90077 115.7 LP2P8374 98.3 W5214 112.0 4P 6524 177.3 83646 141.0 90132 115.0 90157 98.0 12008/20291 112.0 89833 =

171.7 3P7150 139.3 89833 113.3 5P7388 97.7 13253 111.0 90128 167.3 90132 139.0 90071 112.7 90069 97.6 1P3571 105.0 88114 166.3 SP7388 139.0 90159 112.7 3P7246 97.3 39B196 104.0 90069 161.7 83640 137.7 90077 112.3 4P7869 97.3 305414 104.0 90154 160.0 10137 137.0 89408 110.7 3P7150 97.0 27204 98.0 33A277 160.0 90159 136.0 3P8013 110.0 5P9028 96.7 V89476 158.0 89024 135.7 3P7802 109.7 90146 96,7 83650 154.0 3P8349 135.0 4P8632 109.3 651A708 96.3 90132 153.7 83653 134.3 89828 109.3 89828 96.0 87003 153.0 90136/10137 132,0 LPSP9744 109.0 89833 95.7 87005 152.0 87000 131.7 Page 21 of 30

M W W W W W Table 5 (cont'd)

USERS.:;.

-" -; ::":USE.". USE:: ": ": """:"'-:: "USE

":";":.Wir'e'.,, :.Valu'es";: ;:;'::; Wire';:'.".;::,', ".':Values," 'Wire".'.'"":,.:..:,'.,  ;",.;-'Wir'e".::::i,",::':,:,

.'. Heat No

:'.. 'ft-lb)'", ,':,"..Heat.'.No ",,, '.;::

(ft-.lb)".:.-:,';::;;:; ':'.Valiies'-,;.'e'at:,No'.":'.,',.::,;:;,.',':(ft-'lb)

':;;  :, Heat';:No.'.'..':::'.:: ',.:..',:(ft-'Ib). '.::';:,":;,:,',.Heat'No;:::.',,::,-:;.,.::.: (ft-lb).",:

3P7246 108.0 3P7317 94.0 88112 149.7 3P8344 131.0 4P7869 107.7 651 A708 94.0 90146 149.7 87003 129.3 5P7388 107.7 90149 94.0 89828 149.0 2P8354 127.3 5P8866 107.7 3P8013 93.7 89024 148.3 4P6519 127.0 83653 106.7 4P7869 93.7 90211 148.3 30502 122.7 4P 6524 106.3 90144 93.0 88114 147.0 30502 118.3 83646 106.0 4P7656 91.3 90071 147.0 30502 106.3 89833 105.3 5P8866 91.3 4P5 174 146.7 2P8354 104.7 3P7802 104.7 F69025 91.0 90128 146.7 3P8013 104.7 87005 90.3 90209 146.0 4P7869 104.7 E56906 89.3 90149 145.3 30502 104.3 69025 88.6 89827 145.0 89828 104.3 4P7656 88.0 90136 145.0 3P7246 103.7 91762 88.0 83642 144.7 3P7317 103.0 3P7150 86.0 90157 144.0 3P7317 102.7 90211 82.7 5P5622 144.0 Page 22 of 30

TABLE 6 STATISTICALRESULTS OF UPPER SHELF ENERGY WITH WELDS USING LINDE 124, 1092 4 0091 FLUXED Linde Flux Type Variable: 1092 0091 Sample size 68. 60.

Mean USE, ft-lb 102.3 112.4 150.3 Mean - 2 std. dev. 83.5 96.2 99.7 Median USE, ft-lb 102.5 112. 145.

Variance 88.5 65.4 638.

Standard deviation, ft-lb 9.41 8.09 25.3 Range, ft-lb 43. 28. 132.6 Lower quartile, ft-lb 95.85 105. 137.4 Lower 95% Conf. Mean, ft-lb 97.9 102.3 147.1 Upper 95% Conf. Mean, ft-lb 106.7 122.4 153.5 Upper quartile, ft-lb, 109.15 118. 153.8 Coef. of variation, ft-lb 9.19 7.20 16.81 Page 23 of 30

TABLE 7 ANALYSIS OF VAlUANCEFOR WELDS WITH LINDE 124, 1092 & 0091 FLUXES W

Grand Mean, ft-lb 123.7 Variation between treatments 75159.

Variation within treatment 44383.

Total Variation 119543.

F (Calculated) 116.8 F (Table, 95% Conf.) 3.07 Result: F(Calc.) > F(Table), Null Hypothesis is rejected.

v v Estimated overall stat. DN 0.516 Significance Level (SL) 0.00601 Confidence Interval (Cl) 0.05 Result: SL ( CI, therefore Null Hypothesis is rejected.

Variation between treatments 1103.

Variation within treatment 6712.

Total Variation 7815.

F (Calculated) 13.0 F (Table, 95% Conf.) 4.0 Result: F(Calc.) > F(Table), Null Hypothesis is rejected.

-W Test Statistic 100.0 Significance Level (SL) 0.0 Confidence Interval (CI) 0.05 Result: SL ( CI, therefore Null Hypothesis is rejected.

Page 24 of 30

I I

0<

eO tt 0

~<C 0~1~A

~CC~4)t)1

'O

~ ~ ~ ~

~ ~

11 (4

(4

FIGURE 2 HISTOGRAM OF DATAFOR LINDE 124 FLUX 20 z0 10 1 U l l

l

\

1 r ~r 60 80 100 120 160 UPPER SHELF ENERGY (ft-Ib)

Page 26 of 30

FIGURE 3 HISTOGRAM OF DATAFOR LINDE 1092 FLUX 0

Z Q

0 60 80 100 120 160 UPPER SHELF ENERGY (ft-Ib)

Page 27 of 30

FIGURE 4 HISTOGRAM OF DATAFOR LINDE 0091 FLUX 80 110 170 200 230 260 UPPER SHELF ENERGY (ft-Ib)

Page 28 of 30

FIGURE 5 NOTCHED BOX AND VVHISKER PLOTS FOR LINDE 124, 1092 AND 0091 FLUX 250 225 J3 200 0-(9 175 K

LU z

LU LL 150 Uj I

CC 125 ill 0

0 100 75 50 LINDE 1092 LINDE 124 LINDE 0091 Key to Plot TYPE OF LINDE FLUX Upper Quartile Mean Median Lower Quartile 0 Outhers Page 29 of 30

FIGURE 6 CUMULATIVEDISTRIBUTION FUNCTION FOR LINDE 124 AND 1092 FLUX 0.9 ---- ~-- LINBE $ 24-LINDE 1092 0.8 I

I 0.7 ~

p 0 0.6 0.5 I I

I 0.4 0.3 I I

0.2 L 0.1 0

80 90 100 110 120 130 UPPER SHELF ENERGY (ft-Ib)

Page 30 of 30