Bolt loading effects on the structural integrity assessment of defects in industrial components

Abstract

Applied loads in bolted geometries of safety critical components can vary with time and operating conditions. Structural integrity and remaining life assessments of such components in aging industrial plants must consider the resultant changes in damage accumulation rates and acceptable defect sizes. Two case studies are presented that demonstrate the effect of bolt pre-load on creep and fatigue lives as well as on the acceptability assessments of defects. In the first case the sensitivity of creep damage accumulation and crack propagation rates to bolt pre-load in high temperature flanged connections are considered. Predicted results were found to compare well with actual damage quantified on a high pressure turbine loop pipe flange connection. It was shown that decreased pre-loads, in this case, leads to an increase the allowable safe defect size during assembly at room temperature. In contrast to this the second case study of corrosion fatigue cracking in a boiler water circulating pump illustrates that an increase in bolt pre-load leads to an increase in fatigue initiation life, a decrease in fatigue crack propagation rate and an increase in the acceptable defect size. Strain gauge measurements of bolt and casing strain, which correlated well with finite element calculations, indicated the necessity for close control of bolt pre-load during assembly to ensure specified levels are attained. In both cases metallurgical analysis and structural integrity assessments of cracked and excavated geometries were conducted which enabled limited continued operation of the components after which repairs and/or replacements will be implemented.

Keywords

Bolt pre-load structural integrity flange boiler water circulating pump creep fatigue

1. Introduction

Bolted connections of major components are common in industrial plant. Many of these connections are on components or systems which require high levels of integrity as failure could result in catastrophic consequences and/or injury or fatalities. In most of these cases a bolt pre-load to be applied during assembly at room temperature is specified in the design. Control of this pre-load is mostly done indirectly such as for example the arc of turn method and the torque method. If applied without the necessary caution and attention to detail significant over or under pre-load can result. This may result in a significant reduction in remaining safe operating life. Two case studies are presented. In the first case the effect of lowering pre-load on the defect tolerance in a high temperature flanged connection is investigated. Secondly, the effect on low cycle fatigue life due to the increase in tie bolt pre-load in a boiler water circulating pump bowl was investigated.

2. High temperature flanged connections

2.1. Flange design

Flanged connections on high pressure and temperature (>400 °C) main steam, auxiliary steam, turbine loop pipes and other systems are commonly used throughout industry. Some of these connections make use of the so called ‘raised face’ (RF) flange design and is usually face-to-face (i.e., no gasket). The reason for the raised face is to increase the contact pressure between the flanges and in so doing increase the pressure containment capacity. These flanges have a gap between the flange bolt faces (Fig. 1). Note that other naming conventions or similar designs, where there is a gap between flanges and an induced bending moment, may exist. Recent non-destructive-testing (NDT) inspections detected significant creep damage and cracking on some RF flanges that have been in operation for more than 200 kh [10]. Historical records indicate that cracking has been seen as early as 165 kh and 4 re-tightening cycles, which typically occur during general overhauls (GO). Creep damage and cracking can be significant and pose a threat of leaking and/or failure if left unattended. Operating temperature in this case is 535 °C.

Fig. 1.

Typical design of a ‘raised face’ flange. Showing areas where creep damage and cracking are expected as well as calculation positions.

2.2. Material of manufacture

The subject flanges were manufactured from a CrMoV steel (21CrMoV-511) and the bolts and spacers from X19CrMoNbVN11-1 material. Material properties can be found in [2].

2.3. Mechanism of damage

Bolting of the RF flange to design pre-load leads to high local stresses and cracking in the bolt-hole recesses and the flange-to-pipe radii (Fig. 2) due to bending of the flange. The pre-load in the bolts and high bending moment induced tensile stress in the flanges reduce in operation due to creep relaxation but reach a nominally constant value after approximately 20 kh. Although stress in the flanges reduce correspondingly it cycles around high enough levels for creep damage to accumulate with continued operation. Creep damage accumulation is exacerbated by repeat re-tightening, usually during general overhauls, as well as due to the multi-axial stress state (i.e., multi-axial as opposed to uni-axial) in the flange-to-pipe radii. Mult-axial stress states are known to reduce material creep ductility. Actual creep damage accumulation and creep crack growth rates are dependent on material properties which can vary significantly and hence the time at which damage and/or cracking is first detected.

Fig. 2.

Cracking in bolt recess and flange-to-pipe radius.

2.4. Bolt loading

The initial bolt force (F) and bolt stress (𝜎) at room temperature can be calculated from: $\begin{eqnarray}\displaystyle F\text{} & =\text{} & \displaystyle \frac{{\Delta}\text{LAE}}{L}\end{eqnarray}$ (1) $\begin{eqnarray} \displaystyle {\sigma}\text{} & =\text{} & \displaystyle {\epsilon}E=\frac{{\Delta}L}{L}E \end{eqnarray}$ (2) where ΔL is the elongation, L is the original length, A is the sectional area and E is the elastic modulus. When the assembly is heated to the operating temperature the elastic moduli reduce. In addition, the differential thermal expansion between the bolt, spacers and flange result in an increase in bolt elongation relative to the flange material. The combined effect results in a slight bolt load and stress reduction. A power curve was fitted to creep relaxation data (with 0.2% initial strain) for the bolt material (X19CrMoNbVN11-1) at 535 °C obtained from [2] and is shown in Fig. 3. The equation for the power curve is: $\begin{eqnarray}\displaystyle {\sigma}_{t}=456.24t^{-0.146} & & \displaystyle\end{eqnarray}$ (3) with t the operating hours and the stress 𝜎_t in MPa. For a calculated bolt pre-load stress of 238 MPa the creep relaxation and the resultant reduction in bolt force was calculated using Eqs (1) to (3) above. A reduction in bolt pre-load, at room temperature, was considered to reduce the flange stress and the risk of plastic deformation (due to reduction in yield strength with temperature) during the initial warming to full operating temperature. Defect tolerance at room temperature during re-tightening is also expected to increase. It was found by finite element analysis that a reduction in bolt pre-load does not have a significant effect on seal face pressure (Fig. 4) after creep relaxation.

Fig. 3.

Reduction in bolt load due to creep relaxation.

Fig. 4.

Effect of bolt preload on seal face pressure.

2.5. Multi-axial stress conditions

It is well known that the state of stress in a component influences the creep rupture life whilst creep rupture properties are typically obtained from uniaxial creep tests. To account for the multi-axial stress state Othman and Hayhurst [8] developed a formulation to calculate a representative stress (𝜎_rep) which can then be used with the uniaxial creep rupture data to predict rupture life in the component: $\begin{eqnarray} \displaystyle {\sigma}_{rep}={\alpha}{\sigma}_{1}+(1-{\alpha}-{\beta}){\sigma}_{eqv}+3{\beta}{\sigma}_{m} & & \displaystyle \end{eqnarray}$ (4) where the sum of 𝛼 and 𝛽 range between 0 and 1. This suggests that maximum principal (𝜎₁), equivalent (𝜎_eqv) and/or hydrostatic (𝜎_m) stress (or some combination of these) may determine material creep rupture behaviour. These two material parameters have to be determined by multi-axial creep testing (e.g., notched bar creep tests). However, a number of tests have indicated that the hydrostatic component can be neglected for most materials (e.g. [3,5]). Furthermore, it has also been shown in a number of cases that equivalent stress is a reasonable predictor of rupture life (e.g. [3,6]). As such the representative stress was taken as the equivalent stress i.e., 𝛼, 𝛽 = 0. However, for areas with compressive 2nd and 3rd principal stresses the resultant equivalent stress may be overly conservative. In these cases, the 1st principal stress was used for the determination of the creep rupture time (i.e. uniaxial conditions are assumed). Multi-axial stresses are also known to reduce material creep ductility and time to rupture.

The Cocks–Ashby relation [4] is one method to calculate the ductility factor (DF), i.e. the ratio of the multi-axial ductility to the uniaxial ductility as a function of the tri-axiality factor (TF) which is given as: $\begin{eqnarray}\displaystyle TF_{i}=\frac{{\sigma}_{1,i}+{\sigma}_{2,i}+{\sigma}_{3,i}}{3{\sigma}_{eqv,i}} & & \displaystyle\end{eqnarray}$ (5) TF was limited to 0.333 (uniaxial tensile stress) and 2, which correlates to a creep ductility reduction of ∼90%. The Cocks–Ashby ductility factor (DF) is given as: $\begin{eqnarray} \displaystyle DF_{i}=\frac{\sinh \left[\left(\frac{2}{3}\right)\left(\frac{n-0.5}{n+0.5}\right)\right]}{\sinh \left[(2)\left(\frac{n-0.5}{n+0.5}\right)TF_{i}\right]}=\frac{{\varepsilon}_{f,i}^{\ast }}{{\varepsilon}_{f,i}} & & \displaystyle \end{eqnarray}$ (6) where ${\varepsilon}_{f,i}^{\ast }$ is the reduced creep ductility due to tri-axiality and 𝜀_{f, i} is the uniaxial creep ductility. The Cocks–Ashby time to rupture reduction factor for multi-axial stress conditions are given as [4]: $\begin{eqnarray} \displaystyle \mathit{TR}_{i}=\frac{1}{\sinh \left[(2)\left(\frac{n-0.5}{n+0.5}\right)TF_{i}\right]}=\frac{t_{f,i}^{\ast }}{t_{f,i}} & & \displaystyle \end{eqnarray}$ (7) where $t_{f,i}^{\ast }$ is the time to rupture under multi-axial stress conditions and t _{f, i} the time to rupture under uniaxial stress conditions. Note that this mathematical fit to the multi-axial creep behaviour is accurate for high values of TF (≥0.833). Beyond this point the TR factor increases (incorrectly) beyond 1. The stress ratio is defined as the ratio of the tensile multi-axial stress (T) and the uniaxial stress (𝜎_a) such that: $\begin{eqnarray}\displaystyle {\sigma}_{1}\text{} & =\text{} & \displaystyle {\sigma}_{a}+T\end{eqnarray}$ (8) $\begin{eqnarray} \displaystyle {\sigma}_{2}\text{} & =\text{} & \displaystyle {\sigma}_{3}=T \end{eqnarray}$ (9) Substitution in Eq. (5) results in: $\begin{eqnarray} \displaystyle TF_{i}=\frac{1}{3}+\frac{T}{{\sigma}_{a}} & & \displaystyle \end{eqnarray}$ (10) where T∕𝜎_a is the stress ratio. For uniaxial conditions T is zero and TF reduces to 1/3. Equation (7) was used to calculate TR but it was limited to a maximum of 1. DF and TR were found not to change significantly with time. TF, DF and TR for a test case are shown in Fig. 5. The flange creep damage fraction accumulated over the operational history was calculated based on the relaxed equivalent or 1st principal stresses calculated using a finite element model and a time-based creep damage accumulation approach: $\begin{eqnarray} \displaystyle LF_{t,i}=\frac{t_{i}}{t_{f,i}^{\ast }} & & \displaystyle \end{eqnarray}$ (11) where t _i is the interval time length and $t_{f,i}^{\ast }$ is the time to rupture for the stress interval accounting for the multi-axial stress state. The accumulated life fraction (LF _{t, i}) after each time step was then calculated by simple summation of the preceding interval life fractions. A life fraction of 1 denotes crack initiation. The accumulation of the damage fraction with operating time, and re-tightening cycles, for three positions (as per Figs 1 and 5) were calculated.

Fig. 5.

Tri-axiality factor (TF) (left), ductility factor (DF) (centre), and time to rupture reduction factor (TR) (right) for a test case.

Position 1 is between two bolt holes on the external surface, position 2 is a volumetric position below the bolt-hole recess and position 3 is on the internal diameter in-line with the bolt-hole recess.

Although the effective stress at position 2 is not the highest the creep damage fraction after 250 kh is the highest relative to the other two positions. This is due to the tri-axial state of stress in this position. Due to the higher tightening stress the creep damage fraction at position 2 is also higher than at position 3. Note that only the damage fraction for position 1 is affected significantly by re-tightenings (Fig. 6).

Fig. 6.

Stress history and calculated creep damage fraction.

3. Fatigue cracking in a boiler water circulating pump

3.1. Background

Inspection of a boiler water circulating pump (BWCP) by magnetic particle (MT) testing and phased array (PAUT) inspection indicated significant cracking (full circumferential) from an internal radius position (Fig. 7). PAUT indicated intermittent defects up to a maximum depth of approximately 33 mm (in a wall thickness of ∼60 mm). Material replications were evaluated under a light microscope and cracking was found to be transgranular, multi-facetted, branched with blunt tips and oxide filled (Fig. 8). It was concluded that the damage mechanism was low cycle corrosion fatigue. All defects were carefully excavated up to a maximum depth of 24 mm at an angle of approximately 32° [11].

Fig. 7.

Multiple linear indications detected on the inner radius of the BWCP.

Fig. 8.

Photomicrograph showing the defects at 50× magnification. (5% Nital etch.)

3.2. Design and operating conditions

The BWCP bowl is manufactured from 15NiCuMoNb5 (WB36) material. As the operating temperature is ∼355 ^°C a sample was extracted and a Charpy ISO-V impact test was conducted to check for temper embrittlement. An impact energy significantly higher (233 J) than the minimum specification, i.e. 40 J [12], was obtained which can be correlated to approximately 92 MPa $\surd \text{m}$ using the Sailors–Corten correlation [9]. Relevant loads experienced by the BWCP bowl are primarily the internal pressure of 20.8 MPa (maximum) and loading due to the bolts attaching the electric motor to the pump BWCP bowl flange (see general layout Fig. 9). The BWCP is only used intermittently during starting or stopping of a boiler but may accumulate from 6 to 15 pump start/stop cycles per single boiler start-up.

Fig. 9.

Typical layout of BWCP complete with motor and heat exchanger [7].

3.3. Assembly and operating stress

Tightening of the assembly tie bolts are affected by heating of the bolts and turning to a pre-set revolution to obtain a design bolt elongation of ∼2 mm which implies a bolt load of ∼700 kN. To ensure correct installation two high temperature strain gauges were fitted to a tie bolt as well as two positions on the BWCP bowl using standard foil type rosette strain gauges. Initial results indicated an incorrect installation (low tie bolt load) and the pump was re-assembled to a tie bolt load of 656 kN. A cyclic symmetric finite element model containing a sector of the BWCP bowl, heat barrier, motor casing and one tie bolt was developed. Calculated stress results for the assembly condition (i.e. bolt load only, 656 kN) at the two bowl strain gauge positions compared fairly well with measured results which gives confidence in applying the FE model for further analyses (Fig. 10).

Fig. 10.

FE (lines) and measured (single point data) stress response at strain gauge position 1 (POS1) and position 2 (POS2) for the pre-load case (i.e. 0 MPa internal pressure).

3.4. Structural integrity assessment

Stress intensities for a range of crack depths were calculated by directly modelling the cracks in the FE software (Ansys 17.1) for full load conditions i.e., internal pressure and bolt load. Based on a conservative material toughness of 66 MPa $\surd \text{m}$ obtained by applying a safety factor as per BS 7910:2005 [1] the allowable crack depths for tie bolt loads of 700 (design), 656 (actual) and 603 kN (14% lower than design) were calculated as 31, 28.5 and 25 mm respectively. For the case of a 24 mm deep excavation and a lower bolt load of 603 kN this implies that a re-initiated crack up to 1 mm only is tolerable. For a higher bolt load, e.g., 700 kN (design), this is significantly larger (Fig. 11).

Fig. 11.

Stress intensity increase as a function of crack depth.

Fig. 12.

Crack propagation from an assumed 25 mm initial crack depth, 15 pump starts per unit start and 10% higher loads.

Primary loading driving fatigue initiation and propagation was assumed to be due to the internal pressure cycles. Assuming an initial crack depth of 1 mm at the base of the excavated geometry, i.e., an initial crack depth of 25 mm, the number of unit starts allowed to reach the critical crack depth was calculated. The number of pump cycles (stop/starts) per unit stop/start was reported to be 6. To account for additional un-reported cycles and other possible transient effects this value was increased to 15. A 3rd order polynomial fit to the 3D finite element calculated stress intensities as a function of the crack depth was done and the resultant equation is: $\begin{eqnarray}\displaystyle K_{a}=5.684\times 10^{-4}a^{3}-2.0766\times 10^{-2}a^{2}+1.3203a+27.917, & & \displaystyle\end{eqnarray}$ (12) where K _a is in MPa $\surd \text{m}$ and a, the crack depth, is in mm. This equation is used to calculate the maximum stress intensity per cycle as a function of the crack depth. The minimum stress intensity per cycle was taken to be zero as it was shown that the crack will in fact close due the compressive stress in the radius caused by the pre-load of the studs. The cyclic stress intensity is thus given directly by K _a and the R-ratio in this case is 0. As a conservative approach the propagation rate per cycle is calculated using Paris law constants recommended in BS7910:2005 [1] for freely corroding steel in a marine environment (mean curve, R < 0.5). Combining the allowable crack depth and crack propagation rate calculations the allowable number of boiler start/stops were calculated for the 700, 656 and 603 kN bolt load cases as 149, 83 and 0 cycles respectively (Fig. 12). This clearly demonstrates the advantageous effect of a higher tie bolt pre-load on remaining useful life.

4. Conclusions

The effect of over and under pre-loading of bolts used in major structural connections were demonstrated at the hand of two case studies. In the first case it was shown that the allowable crack size during re-tightening at room temperature increases with a reduction in pre-load. In the second case it was shown that an increase in the tie bolt loading of a BWCP bowl to motor connection is advantageous in terms of allowable crack depth as well as fatigue life. It is thus of upmost importance that the tightening of these connections be controlled to obtain the desired design pre-loads to manage component life expectancy.

Footnotes

Acknowledgements

The authors wish to acknowledge inputs from Dr Mark Newby, Mr Rohit Bhagwandas and Mr Henrico van Rooyen for assistance with strain gauging. Eskom is thanked for the opportunity to report and present this work.

Conflict of interest

None to report.

References

BSI, Guide to methods for assessing the acceptability of flaws in metallic structures (BS 7910:2005), BSI Standards Publication, 2007.

BSI, Steels and nickel alloys for fasteners with specified elevated and/or low temperature properties (BS EN 10269: 1999), 2009.

Chang

Lan

and Li

, Research on representative stress and fracture ductility of P92 steel under multiaxial creep, Eng. Fail. Anal. (2015). doi:10.1016/j.engfailanal.2015.09.011.

Cocks

and Ashby

, Intergranular fracture during power law creep under multiaxial stresses, Met. Sci. (1980), 395–402.

Goodall

I.W.

and Skelton

R.P.

, The importance of multiaxial stress in creep deformation and rupture, Fatigue Fract. Eng. Mater. Struct. 27 (2004), 267–272.

Goyal

Laha

Das

C.R.

Selvi

S.P.

and Mathew

M.D.

, Finite element analysis of uniaxial and multiaxial state of stress on creep rupture behaviour of 2.25 Cr–1 Mo steel, Mater. Sci. Eng. A 563 (2013), 68–77. doi:10.1016/j.msea.2012.11.038.

KSB, https://www.ksb.com/centrifugal-pump-lexicon/circulating-pump/192060/.

Othman

and Hayhurst

, Determination of large strain multi-axial creep rupture criteria using notched-bar data, Int. J. Damage Mech. 2 (1993), 16–52.

Phaal

and Brown

, Correlations between fracture toughness and charpy impact energy (TWI), 1994.

10.

Scheepers

, Confidential Eskom Report, 2016.

11.

Scheepers

Van der Meer

and Bhagwandas

, Confidential Eskom Report, 2016.

12.

Valourec & Mannesman, The WB 36 Book for Feedwater Systems.