Corrosion fatigue crack growth in BS 4360 Grade 50D steel - an analysis

TWI Bulletin, September 1986

by Geoff Booth and Steven Dobbs

Geoff Booth, MA, PhD, MWeldl, is a Principal Research Engineer in the Fatigue Department. Steven Dobbs was a vacation student in the Department and is now reading engineering at Cambridge University.

Use of fitness-for-purpose concepts for assessment of defects is preferable to setting arbitrary acceptance levels, but application of the concepts requires a knowledge of the factors governing defect extension. This article, which is concerned only with fatigue, collates fatigue crack propagation data relevant to North Sea environmental conditions and recommends values of the constants required in fracture mechanics analyses of offshore structures.

There is an increasing awareness that whilst weld defect acceptance levels based on quality control are of considerable value, they are arbitrary and may be very conservative. A more rational method of defining acceptance levels is based on the concept of fitness for purpose. ^[1] Using this principle, a defect may be accepted providing that the conditions to cause failure during service are not reached. In general, a number of mechanisms of defect extension must be considered but the present article is concerned with extension by fatigue crack growth.

The rate of fatigue crack extension per cycle (da/dN) is related to the stress intensity factor range ( ΔK) by the expression:

where C and m are constants which depend upon material and environment. Equation [1] can be integrated (sometimes analytically but more often numerically) to determine the number of cycles corresponding to the growth of a fatigue crack from an initial size ( i.e. the defect size) to a final size, often determined by fracture considerations. If the design lifetime of the component or structure is less than the number of cycles thus calculated then the defect may be accepted from a fatigue standpoint. Other expressions, similar in form but more complex than [1], have also been proposed for relating crack growth rate to stress intensity factor. They are, however, more cumbersome to use.

For structural steels cycled in air, the rate of crack propagation is not sensitive to microstructure, frequency or stress ratio (R = minimum stress/maximum stress). ^[2] Work carried out previously ^[3] has enabled values of C and m to be recommended for use in conjunction with [1]. Values of C and m corresponding to the 95% upper confidence limit, mean and lower 95% confidence limit crack growth rates are given in Table 1.

Table 1 Values for C and m obtained in air ^[3]

Many fitness for purpose defect assessments are now concerned with offshore applications. In contrast to air environments, the crack propagation rate in the North Sea environment is dependent on frequency and stress ratio and may also be sensitive to steel microstructure. Hence C and m values derived for onshore applications are not appropriate. A number of results have now been published which present fatigue crack propagation data obtained under environmental conditions which simulate those experienced by a platform in the North Sea. It is the objective of the present work to collate these data and recommend C and m values for fracture mechanics analyses in offshore applications.

Data sources

Seven different sources of fatigue crack growth data obtained under conditions representative of those offshore were identified. ^[4-10] Six investigations ^[4-9] studied steel complying with BS 4360 Grade 50D and the seventh ^[10] considered steel to Euronorm 11372 grade Fe E355 KT (a similar material). The specifications of these steels are given in Table 2.

Table 2 Details of steel specifications

Crack propagation occurred exclusively in parent material and no data relating either to weld metal or HAZ were found. All tests were performed on rectangular through-thickness notched specimens. Data from the growth of semi-elliptic cracks included in some of the references were rejected because the stress intensity factor solution for this crack geometry is not as well defined. The results assessed included tests at frequencies between 0.1-1 Hz and temperatures between 5-20°C. It is accepted that these variables may play a role in determining growth rate but all these results were included to obtain a realistically sized database relevant to a range of offshore conditions.

The crack propagation results were available only in graphical form and data extraction was therefore carried out using a computer software package which converted the graphical data points into digital co-ordinate form.

Method of analysis and results

Crack propagation results were available relating to four different environmental conditions:

1. Air;
2. Free corrosion;
3. Cathodic protection (results were available at -800 and -850mV, with respect to a silver/silver chloride reference electrode);
4. Cathodic overprotection (results were available at -1000 and -1100mV, Ag/AgCl).

Figures 1 to 4 present the complete data sets for each of these four conditions.

The power law, [1], is valid only for intermediate values of ΔK. At low values of ΔK the growth rate diverges from the power law line and becomes asymptotic to the vertical axis, corresponding to a threshold, ΔK _t , below which crack propagation does not occur. This is particularly evident in Fig.1. In addition, at high values of ΔK the growth rate accelerates above the power law line when static modes of fracture occur as the material fracture toughness is approached. Subsequent analyses of the four data sets were therefore restricted to ΔK in the range 300-1800 N/mm ^3/2 as this represents the region in which [i] is valid. Under corrosion fatigue conditions, stress ratio influences crack growth rate and therefore two levels of mean stress defined by 0 < R ≤ 0.1 and R = 0.5 were identified for each environmental condition.

Each of the eight sets of data thus defined was analysed using the method of least squares to obtain the best fit straight line assuming an expression of the form given in [1 ].

In addition, 95% confidence limits for the data were also calculated, assuming that these corresponded to straight lines with the same gradient ( i.e. m) as the mean line.

Figures 5-8 show the modified data sets and the regression lines for 0 < R ≤ 0.1 and Fig.9-12 Show identical information for R = 0.5. Values of C and m are tabulated in Tables 3 and 4 for the two mean stress conditions.

Table 3 Values of C and m for 0<R ≤ 0.1

Table 4 Values of C and m for R=0.5

Discussion

Figure 5 shows that for data obtained in air with 0 < R ≤ 0.1 fatigue crack growth rates in BS 4360 Grade 50D steel are similar to the growth rates reported for a wide range of structural steel, HAZ and weld metal microstructures obtained with R = 0. ³ At R = 0.5, however, the growth rate in BS 4360 Grade 50D steel was more rapid than that previously reported, as illustrated in Fig.9. The reason for this apparent effect of stress ratio in BS 4360 Grade 50D steel is not known.

Many defect assessments for design and inspection purposes require the upper bound to fatigue crack growth data to be used. For such assessments, it is recommended that values of C and m corresponding to the upper 95% confidence limits, given in Tables 2 and 3, are used. However, it must be recognised that there is considerable scatter in the data, e.g. Fig.6, and it is extremely unlikely that a fatigue crack would grow at this upper bound growth rate over the whole interval of ΔK. The upper limit growth rate is typically a factor of two to four greater than the mean and this can be compared with the factor of four recommended in ref.^[1] . To predict average behaviour, as opposed to design studies, the mean values of C and m ( Tables 3 and 4) are appropriate.

A summary of the mean lines derived for all four environments for 0 < R ≤ 0.1 is given in Fig.13.

Freely corroding and cathodically protected conditions result in similar crack growth rates and correspond to an acceleration factor with respect to the mean growth rate in air of x5 at low values of ΔK reducing to x2 at high ΔK values. The cathodic overprotection case leads to an anticlockwise rotation of the mean line, corresponding to an acceleration factor (with respect to air) of x2 at low ΔK values and x4 at high ΔKs.

The effect of a sea water environment on fatigue threshold ( ΔK _t ) is not clear. In one investigation ^[9] performed at 0< R ≤ 0.1, freely corroding conditions reduced the in-air threshold by approximately 20% whereas cathodic protection restored the in-air value. In contrast, other work ^[11] reports an increase in threshold under freely corroding conditions over that in air. For design purposes, the relationship

has been proposed ^[1] for fatigue loading in air. As far as the authors are aware, no values of ΔK _t in sea water have been reported which are less than those given by [2]. Therefore, it is suggested that [2] is used to estimate ΔK _t for design purposes for all environments. An alternative approach would be to suggest assuming ΔK _t = 60 N/mm ^3/2 for all sea water environments, as no values less than this have been reported, as far as the authors are aware. It is clear that this area requires further review as more data become available.

In conclusion

A review of fatigue crack growth data relating to BS 4360 Grade 50D steel in sea water has been carried out. A statistical analysis of the data has been performed and values of C and m in the equation

have been derived for the mean and upper and lower 95% confidence limits for four environmental conditions and two mean stress levels. These values of C and m are proposed for use in defect assessment calculations concerned with offshore applications.

References