Using the ASME and BSI codes to predict crack growth due to
repeated thermal shock
Brian B. Kerezsi, John W.H. Price * Department of Mechanical Engineering, Monash University, P.O. Box 197, 900 Dandenong Road, East Cauleld, Vic. 3145, Australia Received 20 January 2002; revised 22 February 2002; accepted 22 February 2002 Abstract This paper examines the use of the ASME and British Standard codes to estimate the growth of cracks driven mainly by thermal shocks. Thermal shock loading of operating pressure equipment is a common occurrence, particularly in thermal power stations. The tensile stresses that are produced at the surface of a heated component exposed to a rapid thermal down shock can be high, particularly in the presence of stress concentrations. Repeated application of the thermal shocks may lead to crack initiation and crack growth. The ability to use current codes and standards to describe this type of crack growth is desirable. Unfortunately, thermal shock is a complex transient situation with highly non-linear stress distributions and environmental effects that are not well described by some codes. This paper describes attempts to use the techniques described in the ASME Boiler and Pressure Vessel code Section XI and British Standard BS7910 to predict crack growth rates derived from Monash University experiments. Areas of large conservatism in the methods currently used in industry are identied and possible alternative, less conservative approaches are suggested. If the methods are fully applied, the possibility of crack growth slowing can be captured and the replacement or repair of equipment with thermal shock cracking might be avoided. q 2002 Published by Elsevier Science Ltd. Keywords: Thermal shock; Crack growth; ASME Division VIII; British Standard BS 7910; Crack slowing and arrest 1. Introduction The aim of this paper is to compare current methods for predicting repeated thermal shock crack growth with data obtained during an experimental program conducted at Monash University. Currently, thermal shock crack growth with environmental assistance cannot be fully analysed with the tools available in the codes and standards commonly used in industry for the analysis of in-service aws. As will be shown, those processes, which are available, are very conservative. Thermal shock is common in plant involving water and steam. Down shocks often occur when low temperature uid strikes an already hot surface. Another less common situation is where there are rapid depressurisations in pressurised plant such as can be caused by sudden leaks or valve operations. Thermal shock driven cracking is one of the most common cracking phenomena observed in many types of pressure equipment such as those in electricity generating boilers, nuclear plant, steam turbine auxiliary plant and other situations. Thermal shock cracking tends to start at geometrical discontinuities [14]. Thermal shock driven cracks are often found after many years of operation when areas are inspected internally and this often creates a crisis since such cracking is very easily visible and may be regarded by inspectors as unacceptable. However, such cracking can in fact be stable and slow grow- ing and thus it may also be tolerable to return the component to service without repair. Repairs either take the form of replacement of cracked components or expensive in situ welding. Clearly, industry requires guidelines to assist the assessment of such cracking if it is found. Unlike thermal expansion stresses, thermal shock stresses are not considered in detail in design standards and similarly the peculiarities of their growth have not been considered in tness for purpose codes. Typical crack growth analysis used in industry will follow one of the commonly available tness for purpose codes: ASME Boiler and Pressure Vessel Code Section XI [5] (Appendix A), BS 7910 [6] (Section 8 and Annexe M) or API 579 [7] (Section 9 and associated appendices). In general, these codes allow for analysis of crack growth at low temperatures in selected environments (e.g. air, marine, water reactor). Temperatures relevant to thermal shock cracking are not covered, nor are the International Journal of Pressure Vessels and Piping 79 (2002) 361371 0308-0161/02/$ - see front matter q 2002 Published by Elsevier Science Ltd. PII: S0308-0161(02)00023-6 www.elsevier.com/locate/ijpvp * Corresponding author. Tel.: 161-3-9903-2868; fax: 161-3-9903-2766. E-mail address: john.price@eng.monash.edu.au (J.W.H. Price). environmental interactions that can occur. Looking for example at BS 7910, it is suggested that data for crack growth rates that is not provided by the code should be collated through testing, with replication of factors such as material, environment, loading frequency and waveform, temperature, geometry (including machining marks) and dynamic effects. When cracking is found in an industrial situation, there is no time nor is it economical to carry out such testing. For this reason, testing such as the Monash University testing is a valuable addition to the data resource. 2. Experimental data 2.1. Experimental work at Monash A thermal fatigue test rig has been purposely built for the investigation of crack initiation and growth due to repeated thermal shock loading. The experimental work, which is being conducted at Monash is unique in several ways and uses full scale test specimens which mimic plant conditions. Details of the experimental set up have been discussed elsewhere [8,9]. The tests at Monash take several months to complete as the specimens are put through thousands of cycles which take 1520 min each. The specimens are alternatively heated in a convection furnace and then cooled by sprays of cold water normally lasting for 7 s. There is pH and dissolved oxygen control on the quenching water. Approxi- mately one-dimensional conditions exist at any one crack because of use of attached thermal masses as shown in Fig. 1. A steady mechanical loading is applied to some specimens which enables the effect of primary stresses to be tested as well. The distribution of the transient peak stress intensity factors are shown in Fig. 2. The growth of the cracks is recorded every hundred or so cycles by removing the specimen from the furnace and examining it under a microscope. The one-dimensional character of the growth is conrmed by the crack front being basically straight. In addition to the test rig, a number of specimens cracked by thermal shock and records of such cracking have been collected from a number of power stations. 3. Using ASME Section XI The ASME Boiler and Pressure Vessel Code [5] is currently one of the most widely accepted standards for the analysis of cracks in pressure equipment. In particular, it is Appendix A of Section XI that deals with the analysis of aws. Developed for use in the nuclear power industry, it provides the tools required for the prediction of crack propa- gation rates. Figures A-4300-1 and 2 of Section XI provide the crack growth reference curves for translating the stress intensity factors produced at the tip of a crack into crack growth rates for carbon and low alloy ferritic steels exposed to air and water reactor environments, respectively. These reference curves also provide for the effect of load ratio R (dened as the ratio of the minimum to maximum stress intensity factors over a cycle (K min /K max )). Water reactor environments are basically A.V.T. (all volatiles water treat- ment), thus oxygen and pH are controlled at less than 10 ppb and about 7.09.0, respectively. The ASME XI code provides details of a weight function method for calculating stress intensity factors in situations of complex stress distribution such as thermal shock. The method relies on tting a polynomial stress prole as an approximation to the actual stress prole. Fig. 3 shows a typical t for the Monash University experimental work. Further details of the approach used are described in Ref. [8]. It is to be noted that ASME XI crack growth curves are produced on tensile test machines. In the case of the water reactor environment, the specimen is immersed in a treated but stagnant uid supposedly simulating the water reactor environment. The thermal shock case is not like the oscillat- ing primary load of the tensile test machine since the thermal shock stress component quickly diminishes from the surface. The environment in thermal shock also changes, either from hot to cold water, from steam to cold water or from air to cold water. In addition, the uid is also in violent motion. Consequently, the choice of which of the ASME XI crack growth reference curves to use in a thermal shock analysis is unclear. As mentioned earlier, ASME data also require a quanti- cation of the R-ratio. A problem is posed when deciding what the R-ratio is at the crack tip during thermal shock. Thermal shock stresses are highly non-linear and decrease rapidly away from the shocked surface resulting in a reduc- tion in K max with crack growth (as shown in Fig. 2). This means that for cases of non-zero primary stress and hence non-zero K min , the R-ratio is increasing as the crack grows. B.B. Kerezsi, J.W.H. Price / International Journal of Pressure Vessels and Piping 79 (2002) 361371 362 Fig. 1. Specimen design (not to scale). Because R is changing the selection of an appropriate crack growth reference curve is not clear. The ASME Boiler and Pressure Vessel Code is based on linear elastic fracture mechanics (LEFM) and relies on the stress intensity factor being suitable to describe the stress condition surrounding the crack tip. This is only the case when the bulk of the material is behaving predominantly in an elastic manner. In the case of thermal shock, the bulk of material away from the shocked face remains at stresses below yield and hence crack growth in this region may be analysed using LEFM methods. 3.1. Crack growth analysis Fig. 4 shows experimental crack growth data plotted as cyclic crack growth vs. DK as calculated using the ASME weight function method for cracks growing in an LEFM region. Superimposed on the gure are the Paris Law reference growth laws suggested in the ASME Boiler and Pressure Vessel Code for air and water reactor environ- ments. It is immediately visible that the water environment curves of the ASME code provide very conservative estimates of crack growth when compared to the experimen- tal results, regardless of R-ratio or environment. In fact in some cases, a factor of 10 conservatism on growth rate is observed. Alternatively, the ASME air curves seem to lie much closer to the bulk of the test data, particularly those generated at low R-ratios. The conservatism shown by the ASME code can be partially attributed to the fact that the crack growth reference curves are based on the upper bounds of their own testing data. Comparing actual experimental crack growth to that predicted by using the ASME Paris Law reference curves is completed by integrating the appropriate reference curves along the desired crack length N 1 C Za f a i DK 2m da 1 where N is the number of cycles to grow a crack from length a i to a f , and C and m are the Paris Law coefcients obtained from the ASME code itself. Fig. 5(a) and (b) show crack length against number of thermal shock cycles for growing an established crack that is 5 mm deep. Experimental test data are shown as points and predicted data are shown as solid curves. From these curves, it seems that the ASME code is successfully predicting the trends of the experimental crack growth. As expected, predictions using the water reference curves are fairly conservative. Even with an assumption of R 0; predictions based on the water curves still over-estimate actual experimental crack growth rates. In Ref. [9] it was reported that during repeated thermal shock crack growth, any environmental assistance was high- est when R-ratios were above 0.3. This is reected in the results shown here. Growth curves based on the ASME water reference curve seem to predict the enhanced growth B.B. Kerezsi, J.W.H. Price / International Journal of Pressure Vessels and Piping 79 (2002) 361371 363 Fig. 3. Fitting a polynomial to the actual (measured and calculated) stress prole. Fig. 2. Maximum stress intensity factor proles during 7 s shock from 370 8C, with and without 90 MPa primary load. Note that the combined stress intensity factors are modied by plastic zone correction factors in this presentation. of the test data with R , 0:5 (Fig. 5(a)), while the growth curves based on the air reference curves provide better predictions for all the data where R , 0:3 (Fig. 5(a) and (b)). 4. Using BS 7910 The British Standard BS 7910 [6] has wide applications. In our case, the aws being investigated are not normally associated with welds unless the welds are at a point of stress concentration. The focus of the analysis in this section will be on the crack growth analysis used in BS 7910 Section 8. 4.1. Generation of stress intensity factor proles The rst step in the BS 7910 crack growth analysis requires the simplication of the stress prole in the mate- rial into two components, namely the membrane and bend- ing stresses. In the case of a highly non-linear stress prole, as is the case with thermal shock, correctly following the code procedure can result in a very conservative result. Fig. 6 shows a typical stress prole generated during the thermal shock of the specimens used in Section 2 of this work. Also included is the suggested BS 7910 `linearisation' prole. This rather conservative approach (over-estimation of stress through the component thickness) is required due to the fact that uctuating thermal stresses are treated as primary (as opposed to sary) loads in a fatigue assessment (BS 7910: 1999, Section 6.4.3). The results of this stress linearisation are combined with the equations provided in BS 7910, Annex M to calculate DK for a number of complex aw geometries including provisions for nite component size. Crack growth rates are then predicted using Paris Law type equations. Deriva- tives of the Paris Law equations are provided to take into account environment (air and marine environments) and load ratio R. Unfortunately, the data are provided with a number of restrictions on their use. Included in these restric- tions are that the air data be used only for non-aggressive environments at temperatures less than 100 8C and that the marine data only be used for environments of less than 20 8C. No suggestion is given for where data can be sourced for components operating outside of these limits. 4.2. Crack growth reference curves Fig. 7(a) and (b) shows the experimental data plotted as cyclic crack growth vs. DK calculated using the accurate weight function approach (as opposed to using BS 7910 linearisation). Superimposed on these gures are the crack growth reference curves that are presented in BS 7910 for dry and marine environments, respectively. Each set of curves from BS 7910 consists of three individual curves. The rst is based on mean test data, the second on mean test data plus two standard deviation (2SD) data and the third `simple' curve is based on a conservative analysis. Initial observations of Fig. 7(a) and (b) shows that the dry curves do a good job of bounding the experimental data obtained at low R-ratios when environmental inuence is small. Similarly, the marine curves do a good job of bound- ing the experimental data obtained at high R-ratios when environmental inuence is large. The following is a list of suggested guidelines for using the BS 7910 growth curves in predicting repeated thermal shock crack growth in a water environment. They are based on comparisons with the experimental data shown in Fig. 7(a) and (b) only. 1. For a slightly conservative analysis at all points, the simple BS 7910 data curves can be used for crack growth prediction. The marine curves would be used for high B.B. Kerezsi, J.W.H. Price / International Journal of Pressure Vessels and Piping 79 (2002) 361371 364 Fig. 4. Comparison of experimental crack growth data with prediction curves of ASME Boiler and Pressure Vessel Code, Section XI (1998), Appendix A. R-ratio data and dry curves for low R-ratio data. Trends indicate, however, that these curves could under-estimate crack growth rates at DK values greater than those used in the experimental work. 2. For slightly more accurate predictions, the mean 12SD data from the marine and dry curve sets could be used. Use of the curves outside of the experimental data region is not suggested until more data can be collected. 3. The use of the mean curves for crack growth prediction is not recommended, as they sometimes under-estimate observed crack growth rates. 4.3. Comparison with experimental data Fig. 8(a) and (b) shows crack length vs. number of cycles for a starting crack length of 5 mm. Experimental crack B.B. Kerezsi, J.W.H. Price / International Journal of Pressure Vessels and Piping 79 (2002) 361371 365 Fig. 6. Stress linearisation for fatigue analysis in BS 7910 [6] for crack length a in plate of width W. Fig. 5. Crack length vs. number of cycles. Comparison of actual data with ASME XI [5] predictions for a 7 s thermal shock. For (a) shocks from 370 8C while for (b) shocks from 350 8C. Solid points refer to test data taken with dissolved oxygen (D.O.) of 8 ppm, hollow points for data with D.O. of 2 ppm. growth data are compared with those generated using the BS 7910 technique of stress linearisation to calculate DK and the mean data Paris Law reference curves. The conservatism generated by the BS 7910 analyses is very large. According to the analysis, crack growth due to the repeated thermal shocks will accelerate into the test specimen until failure (by fracture or plastic collapse) occurs. As observed experimentally, however, this is far from the case with slow crack growth tending towards deceleration and possible crack arrest observed in all experimental cases. The large amount of conservatism shown by Fig. 8(a) and (b) is not due to variations in the Paris Law reference equations. Instead, it is due to inaccuracies developed by applying the linear approximation to the thermal shock stress prole as shown in Fig. 6. 4.3.1. A modied approach In addition to providing stress linearisation proles for fatigue analysis, BS 7910 [6] also provides similar, yet less conservative, methods for use in fracture analysis. An B.B. Kerezsi, J.W.H. Price / International Journal of Pressure Vessels and Piping 79 (2002) 361371 366 Fig. 7. (a) Comparison of experimental crack growth data with dry environment prediction curves of BS 7910 [6]. BS 7910 curves for R , 0:5 are shown for: mean data, mean data plus two standard deviations (2SD), and simplied growth data. (b) Comparison of experimental crack growth data with marine environment prediction curves of BS 7910 (no cathodic protection). BS 7910 curves for R , 0:5 are shown for: mean data, mean data plus two standard deviations (2SD), and simplied growth data. example of applying these methods to the case of thermal shock stress is given in Fig. 9. Note that using this method, the stress prole along the crack length is modelled much more accurately. The process of modelling stresses only along the crack length is also used in the ASME Boiler and Pressure Vessel Code [5]. The use of this form of stress prole linearisation results in the following comparison of observed and predicted crack growth shown in Fig. 10(a) and (b). It is clear that the models now more closely t the observed data. However, crack deceleration is still not adequately predicted. Rather, predicted crack growth appears to continue at a constant rate until failure. The disadvantage of this method is that the stress prole must be re-calculated for each increment of crack growth. Fortunately, automation of such a procedure is fairly simple using modern computa- tional software. B.B. Kerezsi, J.W.H. Price / International Journal of Pressure Vessels and Piping 79 (2002) 361371 367 Fig. 8. Crack length vs. number of cycles. Comparison of actual data with BS 7910 [6] predictions for a 7 s thermal shock. For (a) shocks are from 370 8C while for (b) shocks are from 350 8C. Solid points refer to data taken with D.O. of 8 ppm, hollow points for data with D.O. of 2 ppm. Fig. 9. Modied stress linearisation for fatigue analysis in BS 7910 [6] for crack length a in plate of width W. 5. Combining ASME and BS 7910 approaches While the stress intensity factor prole for repeated thermal shock is accurately modelled by the ASME code, the generation of accurate crack growth data from this can be difcult and confusing. Alternatively, the British Standard method provides inaccurate solutions for generat- ing repeated thermal shock stress intensity factor proles, although it does recommend a series of Paris Law reference curves that are relatively accurate and easy to apply. Conse- quently, a combination of the two methods may provide a better alternative than the application of either alone. Fig. 11(a) and (b) shows the results of a combined analy- sis. Crack length vs. number of cycles is plotted again, this time using the ASME weight function technique combined with BS 7910 crack growth reference curves (mean 12SD data). For the case of high R-ratios (and hence high environmental interaction) the marine reference curves provide good crack growth estimations. Similarly, while R-ratios are low (as is the environmental inuence), the dry reference curves provide a good growth estimation. 6. Discussion In service, when a cracked component is discovered, a fracture mechanics analysis can be performed to see whether the component should be immediately replaced or if it is safe to return to service. If the component is not in danger of failure, but the crack is predicted to continue growing, a schedule for inspection of the component is required if the component is returned to service. Monitoring of the crack growth needs to be made to ensure removal before failure. The schedule is based on the predicted B.B. Kerezsi, J.W.H. Price / International Journal of Pressure Vessels and Piping 79 (2002) 361371 368 Fig. 10. Crack length vs. number of cycles. Comparison of actual data with modied BS 7910 [6] predictions for a 7 s thermal shock. For (a) shocks are from 370 8C while for (b) shocks are from 350 8C. Solid points refer to data taken with D.O. of 8 ppm, hollow points for data with D.O. of 2 ppm. crack growth rate, which in the case of a component exposed to repeated thermal shock would be one of the methods outlined in Sections 24. As can be seen for repeated thermal shock loadings in the cases examined in this paper, the crack growth is in fact steadily slowing down to the point of nearly arrest- ing. In this case, the component may never need to be repaired. To place the various levels of conservatism generated by the techniques outlined in Sections 35 into perspective, the predicted number of cycles required for a crack to grow to a predetermined length are compared in Table 1. The crack length at which failure in a component will occur can be conservatively estimated using a level one failure assessment diagram (FAD) from BS 7910. The crack length at which failure is expected in the test specimens has been calculated as approximately 23 mm. Estimates for the number of thermal shock cycles required for a crack to grow to a length of 23 mm from an initial aw size (crack depth) of 4.8 mm are shown in Table 1 for each of the methods outlined in this paper. As shown in Table 1, methods 14, based on conservative linear approximations of the thermal shock stress proles, all predict crack growth to a length suitable for specimen failure in a relatively small number of cycles. Depending on the number of thermal shock cycles, a component was expected to face in service, using this analysis could easily (and most probably) lead to a recommendation of compo- nent replacement. Alternatively, methods 57 are based on more accurate polynomial stress prole approximations and weight function techniques. The results of these solutions indicate that the specimen should have essentially an innite life. This means the specimen is aw tolerant and should be suited to continuous operation. Inspection of in-service crack growth in the specimen could be carried out at the B.B. Kerezsi, J.W.H. Price / International Journal of Pressure Vessels and Piping 79 (2002) 361371 369 Fig. 11. Crack length vs. number of cycles. Comparison of actual data with combined ASME XI [5] and BS 7910 [6] approach for a 7 s thermal shock. For (a) shocks are from 370 8C while for (b) shocks are from 350 8C. Solid points refer to data taken with dissolved oxygen (D.O.) of 8 ppm, hollow points for data with D.O. of 2 ppm. time of the next planned shutdown, with no additional down-time expenses. Based on a balance of simplicity of application and accuracy, method 6 could be deemed the most successful. This is because it has shown sufcient accuracy to avoid the incorrect conclusion that the component should be replaced and is still based on data taken from a single Paris Law reference curve, irrespective of the R-ratio. 7. Conclusions In loading with repeated thermal shock loadings there is a signicant possibility that the crack growth will actually slow down after a certain stage of growth. It is important to capture this possibility when developing responses to the discovery of thermal shock cracks. This paper has provided a comparison of existing design and analysis code techniques that are commonly applied for crack growth prediction with actual experimental test data for a at plate specimen exposed to repeated thermal shock. The following code specic conclusions can be made: ASME Boiler and Pressure Vessel Code The polynomial stress prole approximation and weight function technique produce an accurate model of the crack length vs. number of cycles growth curve for a at plate specimen exposed to repeated thermal shock. The selection of the provided crack growth reference curves appropriate for converting stress intensity factors to crack growth rates is not immediately obvious when environmental interaction occurs. BS 7910 The stress linearisation techniques used for determin- ing stress intensity factor proles are too conservative for use with the highly non-linear stresses generated during a thermal shock. The crack growth reference curves for a marine envir- onment provide a good approximation of observed experimental crack growths when environmental assistance is present (i.e. when R . 0:3). The crack growth reference curves for a dry environ- ment provide a good approximation of observed experimental crack growth rates when environmental assistance is small or not present (i.e. with R , 0:3). It is observed that a combined technique using the ASME approach for calculating the stress intensity factor prole and the simple crack growth reference curves from BS 7910 provides a good balance between accuracy and simpli- city when estimating crack growth due to repeated thermal shock. Acknowledgements This work has been completed with the assistance of an Australian Research Council grant with contributions from HRL Technology Ltd, Optima Energy, Western Power, Pacic Power and the Electric Power Research Institute of USA. The Monash University Postgraduate Publications Award has also provided assistance. References [1] Dooley RB, McNaughton WP. Boiler tube failures: theory and prac- tice. Palo Alto, USA: Electric Power Research Institute, 1996. [2] Ng HW, Lee CK. Remaining life of a vessel containing an internal corner crack under repeated thermal shock. Proc Ins Mech Engrs 1997;211(E):2159. [3] Hayashi M. High cycle thermal fatigue crack initiation and growth behaviour in a semi-innite plate model. J Press Vess Technol, ASME 2001;123:3059. [4] Yagawa G, Ishihara K. Cleavage and ductile thermal shock fractures of corner cracked nozzles. J Press Vess Technol, ASME 1989;111:2417. [5] American Society of Mechanical Engineers. ASME boiler and pressure B.B. Kerezsi, J.W.H. Price / International Journal of Pressure Vessels and Piping 79 (2002) 361371 370 Table 1 Predicted number of cycles required for a crack to grow from 4.8 to 23 mm in an experimental specimen when exposed to repeated 7 s thermal shocks from 370 8C combined with a 90 MPa steady state primary load Method Description Stress approximation Crack growth prediction curve used Number of cycles 1 BS 7910 analysis Linear across component BS 7910, simple (8.2.3.4), marine curve 350 2 BS 7910 analysis Linear across component BS 7910, mean 12SD, (Table 5) marine curve 2770 3 Modied BS 7910 analysis Linear along crack face BS 7910, simple (8.2.3.4), marine curve 5930 4 Modied BS 7910 analysis Linear along crack face BS 7910, mean 12SD, (Table 5) marine curve 9710 5 ASME XI analysis Polynomial along crack face ASME, water reactor environment curves 6.955 10 8 6 Combined ASME and BS 7910 analysis Polynomial along crack face BS 7910, simple (8.2.3.4), marine curve 3.317 10 8 7 Combined ASME and BS 7910 analysis Polynomial along crack face BS 7910, mean 12SD, (Table 5) marine curve 3.317 10 9 vessel code. Section XI, rules for inservice inspection of nuclear power plant components. New York: ASME, 1998. [6] British Standards. BS 7910: 1999guide on methods for assessing the acceptability of aws in fusion welded structures. London: BSI, 1999. [7] American Petroleum Institute. API 579: recommended practice for tness for service. Washington: API Publishing Services, 1999. [8] Kerezsi BB, Kotousov A, Price JWH. Experimental apparatus for ther- mal shock fatigue investigations. Int J Press Vess Piping 2000;77(7):42534. [9] Kerezsi BB, Price JWH, Kotousov A. Features of fatigue crack growth due to repeated thermal shock. Fatigue Fract Engng Mater Struct Feb, in press. B.B. Kerezsi, J.W.H. Price / International Journal of Pressure Vessels and Piping 79 (2002) 361371 371