Keywords

Cycling; fast start; transient; fatigue; creep; startup; shutdown

11.1 Introduction

Today, many power plants are being forced to provide power in a dispatchable mode. It is common for a power plant to shut down and restart on a daily basis. Many users are asking how they can protect themselves against problems in the future due to this cyclic mode of operation.

In the highly competitive Heat recovery steam generator (HRSG) marketplace of 2016, suppliers are faced with the dilemma of furnishing the best equipment possible for the intended service life and at the same time remaining competitive. The user or owner on the other hand is concerned about initial cost, delivery, and reliable operation. Therefore, it is important that both suppliers and users understand factors that contribute to a good HRSG design under cyclic service conditions and know what measures can be taken to ensure that the equipment will perform as desired.

Failures from cycling can be the result of thermal expansion restraint problems. Restraint problems are normally caused by improperly designed or manufactured supports, header details, flow distribution, or any other arrangement that prevents a component from unrestrained expansion relative to another. The end result is failure within a relatively short period of time. It is very important that restraint problems be eliminated from the HRSG to ensure proper mechanical reliability. Most HRSG suppliers from experience do a good job in eliminating these types of problems. Therefore, this chapter will focus on another main cycling issue: fatigue.

Transient operation is damaging to combined cycle plants due to fatigue from temperature and pressure changes that take place during startup, shutdown, and other modes of operation. Fatigue occurs when material is subjected to cyclic or repeated stresses. The alternating stress amplitude for a cycle is the most significant factor for fatigue and not the absolute stress level. Most fatigue issues in HRSGs are considered to be low-cycle fatigue, in which some plastic strain occurs. An approximate border between high-cycle and low-cycle fatigue has been found to be 10,000 cycles. Daily starts for 30 years would equate to 11,000 cycles. A thorough knowledge of the stress and loading conditions is necessary to perform a fatigue evaluation. A life assessment in the time -independent regime is a fatigue evaluation that considers the various operating cycles and computes an estimate of unit life. It is important to recognize that failure in a life assessment is commonly the point of crack initiation.

HRSG fatigue is related to stress associated with changes in temperature and pressure. Factors that can significantly influence the alternating stress amplitude of the fatigue evaluation include construction geometry, construction details, material type, and corrosion. Creep is the continuous and time-dependent deformation of a material. At elevated temperatures, creep can become a significant engineering consideration. These factors will be discussed in more detail in the following sections. Continuous operation for 30 years would result in 263,000 hours of operation. This amount of time exceeds the 100,000-hour rupture life used as the basis for establishing allowable stress values in the ASME code. Total operating hours should be specified as part of the HRSG design criteria. There is an interaction between creep and fatigue and it may be overly conservative to specify for a unit both 11,000 cycles and 250,000 hours of operation.

11.2 Components most affected

HRSG components that are most affected by fatigue are thick wall components such as high-pressure steam drums, superheater and reheater headers, manifolds, piping, and headers with flow divider plates. Depending on the shutdown conditions, lower headers of superheaters and reheaters may be subjected to thermal shock when condensate is formed in the tubes and drains into the lower headers. Economizers or preheater inlet headers can experience a significant step change in temperature at startup. This may also warrant some review but is not typically included in HRSG life assessments. Although intermediate-pressure and low-pressure steam drums are much thinner than high-pressure drums and do not limit transient operation, they should still utilize good details of construction for cyclic operation.

Junctions of dissimilar metals at high temperatures such as austenitic steel to ferritic steel are areas of major concern and have resulted in significant failures (Ref. [1]).

11.3 Effect of pressure

Internal pressure in cylindrical and spherical shells will create a general primary membrane tension stress in boiler components. Local primary membrane stresses near nozzles and other openings can exceed the general primary membrane stress. Pressure stresses are included in the determination of the amplitude of the alternating stress during a load cycle and are included in a life assessment analysis. The rate of change of pressure is not significant other than how it influences the associated change in saturation temperature of the water.

The effect of pressure is illustrated by the following example of two different pressure cycles of a steam drum. Table 11.1 shows the result of a fatigue evaluation in which the maximum number of allowable cycles, N, are evaluated. Case 1 is a full-range pressure cycle in which the stream drum stresses are cycled from zero to a stress corresponding to 95% of the maximum allowable working pressure (MAWP), and back to zero. Case 2 cycles the stresses from 95% MAWP to 70% MAWP and back to 95% MAWP.

where:

The number of allowable cycles is directly impacted by the magnitude of the alternating stress. Using the method defined by ASME Section VIII, Div. 2 for determining allowable cycles, it can be seen that in Case 1 the magnitude of one half the alternating stress intensity gives an estimated cycle life of 33,000 cycles, whereas in Case 2, the number of allowable cycles is over 1 million. The only difference between these two operating cases is the magnitude of pressure decay between cycles. Therefore, it can be seen that larger alternating stress intensity reduces the number of allowable cycles. Hence, keeping the alternating stress intensity as small as possible is an important factor for increasing the fatigue life.

11.4 Change in temperature

Temperature difference in a component creates stress. This section describes the factors influencing temperature-related stress associated with transient operation. The temperature difference can be a through-thickness temperature gradient or a temperature difference can result between adjacent parts of different thickness or operating conditions. When a metal component such as a steam drum is exposed to changing fluid boundary conditions, a through-thickness variation in temperature will result. This variation in temperature will generate thermal stresses that must be accounted for in the determination of alternating stress amplitude. The internal surface is the surface that experiences the changing boundary conditions. As fluid temperature at the inside surface increases during startup, the material expands and is subjected to compression stress. As fluid temperature at the inside surface decreases during shutdown, the material contracts and is subjected to tension stress.

The maximum temperature variation in a metal component is a function of the metal thickness. thermal diffusivity (β), rate of heat transfer from a fluid to a surface, fluid temperature, and initial metal temperature. The thicker the component, the greater the temperature variation will be. The influence of the fluid boundary conditions impacts the temperature profile in the metal. A high heat transfer rate to a metal surface would result from flowing or boiling water or condensing steam compared to lower heat transfer rates associated with flowing steam or stagnant conditions. Components in contact with steam would lag in temperature response due to a temperature change more than a surface in contact with water.

For a component exposed to a step change in temperature, the maximum temperature difference is equal to the temperature step regardless of the component thickness. Fig. 11.1 shows the mean and cold face temperature response for a 100°F hot face step change in temperature for different geometries. There is some geometric influence as shown by the flat plate temperatures in comparison to the temperatures for a shell-type configuration. The hot face associated with the curves in Fig. 11.1 is the shell inside surface. A spherical geometry such as a drum hemispherical head would have an even different temperature response.

In many cases, ramp rates are defined for transient conditions where the temperature boundary conditions change with time. Fig. 11.2 shows the temperature response for the flat plate where the hot face temperature is ramped at 10°F/minute up from 220–320°F. While the time frame to achieve a certain mean temperature is extended from that shown in Fig. 11.1, the maximum hot face to mean temperature difference is significantly reduced. Fig. 11.3 shows the temperature response for a 5°F/minute ramp rate. Comparing the maximum hot face to mean temperature difference between Figs. 11.2 and 11.3 shows that this temperature difference decreases as the ramp rate decreases.

For a temperature ramp condition, the maximum temperature difference of the hot face to mean depends upon the total absolute temperature change for a component. The mean through-thickness rate of temperature change will approach that of the boundary condition ramp rate for significant periods of temperature change. A maximum value of this difference will be reached if the overall temperature change is great enough. This maximum temperature difference is:

$\mathrm{\Delta }T=\frac{\mathrm{Rate}\times {\mathrm{thk}}^2}{3\times \beta }$ (11.1)

(11.1)

For radial geometries, the temperature difference needs to be multiplied by a factor based upon the shell dimensions and can be approximated by

$\mathrm{\Delta }{T}_{\mathrm{radial}}=\mathrm{\Delta }T*{\left(\frac{\mathrm{OD}}{\mathrm{ID}}\right)}^{0.506}$ (11.2)

(11.2)

A more exact shell relationship is given by Taler et al. (Ref. [2]). Figs. 11.2 and 11.3 show that the maximum temperature difference exists at the end of the temperature ramp. The asymptotic value for Fig. 11.2 is 53.3°F. The asymptotic value for Fig. 11.3 is 26.7°F.

Fig. 11.4 shows how the maximum surface temperature difference of a component starting at 220°F will vary depending upon the ultimate temperature of the hot face. Small changes in steady state operating temperatures can be ramped quickly or even instantaneously without generating damaging temperature differences. Larger changes in operating temperatures must be ramped more slowly to limit the temperature difference. Fig. 11.4 also shows the strong effect of component thickness.

Surface temperatures change as a result of a change in a convective boundary condition and changing internal fluid temperature. The magnitude of the convective heat transfer coefficient to the surface will affect the metal temperature profile. The lower the heat transfer rate, the greater the temperature difference between the bulk fluid and surface temperatures. Fig. 11.5 shows the hot face and mean temperatures for a ramp rate of 10°F/minute for two different convective heat transfer coefficients (btu/h ft² F). The maximum temperature difference is not significantly different but there is a significant metal temperature lag from the fluid temperature for the lower convective rate.

The through-thickness temperature profile for a ramp condition of 10°F/min is shown in Fig. 11.6. It should be noted that the mean temperature is not located at the midpoint of the plate because of the nonlinear profile.

At the beginning of a temperature ramp as shown in Fig. 11.4, there is very little hot surface to mean temperature difference. An initial step change followed by a temperature ramp is a way to speed up a large change in temperature without exceeding the maximum temperature difference from Eqs. (11.2) and (11.3). Fig. 11.7 shows the temperature difference for a case where there is an initial step change in temperature of 50°F followed by a rate of 10°F/min. The initial step change would decrease the startup time 5 minutes in this case without increasing the maximum temperature difference.

Holding HRSG operating conditions to a specific ramp rate is often not practical. There can also be confusion as to how a ramp limitation is applied, i.e., as an instantaneous limit or as an overall temperature/time limit. The varying boundary conditions and through-thickness temperature differences for components due to changing flow, temperature, and pressure for various cycles need to be quantified by a transient thermohydraulic network model. The thermohydraulic model results are used to determine surface heat transfer rates and fluid temperatures. It is then necessary to then predict through wall temperature differences from these boundary conditions. EPRI (Ref. [3]) recommends that for a life assessment analysis “[a] one-dimensional dynamic thermo-hydraulic network model shall be used to develop detailed characteristics of steam pressure, temperature and mass flow.” Fig. 11.8 shows a typical high-pressure superheater outlet condition for a cold startup. The HRSG response for a startup cannot be equated to a simple ramp rate.

11.5 Materials

Material properties not only vary between different materials but also as a function of temperature. Low-alloy steels are used in HRSG construction.

If we combine Eqs. (11.1) and (11.2) we see that

$\mathrm{Startup}\mathrm{Rate}\alpha \frac{\beta }{E\times \alpha }$ (11.3)

(11.3)

In Fig. 11.9, the property group β/(Eα) is illustrated as a function of material type and temperature. A higher operating temperature will have a lower associated ramp rate for a given material. Different materials also have different allowable stresses, which impact component thickness. Fatigue curves can also be material specific depending upon the design code used for analysis. All of these factors make it difficult to directly compare different material types.

Table 11.2 shows example life assessment results using EN-12952 (Ref. [4]) methodology for quick-starting HRSG high-pressure drums comparing two different drum materials and a smaller drum diameter. The SA-302b material, which is a higher strength material, is advantageous for this application and it shows a conventionally designed steam drum is capable of significant cycles for quick-start applications. It does show that the SA-516–70 material is not adequate for these specific conditions in that the total life percentage exceeds 100%. The table also shows the relative fatigue life consumption of the different types of cycles. The assumed start conditions for the different types of cycles are a major assumption in life assessments. The starting pressure would be the saturation pressure consistent with the start temperature.

Table 11.2 also shows the effect of changing drum diameter where smaller drums result in less stress. Reduced diameter drum concepts can thus be another option for quick-start HRSGs (Ref. [12]). The shutdown maximum temperature difference in the table is small. Faster cool-down rates would impact the total life percentages.

11.6 Construction details

Design and fabrication details for HRSG components must be suitable for flexible operation. Construction and weld details for drum and header attachments have an impact on the stress and fatigue life as they affect component thickness and have different stress concentration factors. Better details come at an increased cost.

The EN-12952–3 (Ref. [4]) method considers surface finish and differences in the nozzle construction detail such as a set-on, stick-through or extruded type. There is also a difference if a drum nozzle is on a cylindrical shell section or a hemispherical section such as a drum head. Integrally reinforced nozzles should be used instead of nozzles with nonintegral type reinforcement. Contouring of nozzles and blending of nozzle welds also helps minimize peak stresses as compared to nozzles with sharp corners or sharp welds.

EN 12952–3 distinguishes between partial penetration weld versus full penetration weld details with a significant penalty associated with partial penetration welds.

Tube-to-header connections can utilize a tube stub detail, such as that shown in Fig. 11.10, which can be considered for reinforcement of the opening in the header. This detail can reduce the required header thickness as compared to other methods, such as ligament efficiency, and provide a better transition for thermal stress.

Changes in materials within components such as in superheaters and reheaters also create stress because of different material properties. Material transitions should be made away from points of fixity, such as at the tube-to-stub welds instead of at a tube-to-header weld. This will minimize the stresses resulting from the incompatibility of the material properties.

Drum thickness is greatly affected by drum diameter. Specification of large drum storage volume (large retention time) for high-pressure drums increases the drum diameter impacting the HRSG life. Drum water storage volume should be minimized but must accommodate shrink and swell conditions of the drum level during transients. Usually transient operation has a larger impact on intermediate-pressure drums than high-pressure drums and therefore appropriate sizing of the intermediate-pressure drum is important for cycling. Newer HRSG configurations for high-pressure drums focus on reducing drum diameter.

11.7 Corrosion

Corrosion due to nonideal water chemistry conditions can accelerate damage from fatigue. If the stress level in a component is great enough to cause the protective magnetite layer to crack, corrosion fatigue can occur. Fig. 11.11 shows corrosion fatigue in an evaporator tube.

To preclude magnetite cracking, the component stress should be limited. EN-12952–3 [Ref] limits water-touched surface principal compressive stress to less than 600 MPa (87,000 psi, 0.3% strain) and principal tensile stress to less than 200 MPa (29,000 psi, 0.1% strain). It is assumed that the magnetite layer forms during normal operating conditions so that there is no stress in the layer at those conditions. EPRI (Ref. [5]) recommends that the life fraction be limited to 0.1 if these oxide stress level limits are exceeded. The life fraction is defined as the computed fatigue life divided by the specified fatigue life.

Thermal fatigue is a special type of corrosion fatigue that occurs due to rapid cooling of a hot surface (Ref. [6]). This condition can exist in lower superheater and reheater headers during shutdown when steam condenses inside tubes and flows into the lower header. Oxide layers will crack due to stress in steam-touched surfaces as well. Any exposed base metal will oxidize in operation and the cycle would continue to be repeated when a unit is shut down.

Cycling has an effect on possible exfoliation of oxide from superheater and reheater tubes. Differences in temperature between the oxide and base metal will result in interfacial stresses that can cause to oxide to crack. As the oxide layer increases to a critical thickness, it will spall from the tube surface. Exfoliation occurs in both ferritic and austenitic steels.

11.8 Creep

Creep is the continuous and time-dependent deformation of a material. For low-alloy chrome steels used in HRSG construction, creep per ASME Code (Ref. [7]) becomes a significant engineering consideration at temperatures greater than 900°F. For carbon steels, creep may become a consideration at temperatures as low as 700°F. ASME Code allowable stresses are, in part, limited by the stress to produce creep rupture at the end of 100,000 hours. If the intended HRSG operating hours at temperature are greater than this, the hours must be taken into account in the HRSG design by derating the allowable stress. Elevated temperatures at this level are experienced in superheaters, reheaters, and associated piping. Creep is also a function of the specific material composition, and the amount of time at the elevated temperature. The time/temperature relationship for SA-213-T22 is illustrated by a Larson–Miller curve for stress rupture (Ref. [8]) in Fig. 11.12. Each material type will have different Larson–Miller parameters (P):

$T_{\mathrm{R}}(C+\log \theta )=P$

The combination of creep and fatigue can result in synergistic damage. Evaluation of creep–fatigue combined damage is complex and needs to be evaluated for components exposed to cyclic loading conditions in the creep range.

11.9 HRSG operation

A fatigue evaluation can be made in great detail but any evaluation is based upon numerous assumptions. It does not guarantee a design will function for given number of cycles. How equipment is operated and what construction details have been provided is far more important than how much analytical work has been performed.

11.10 Life assessments

A life assessment evaluation is an analytical method to compute a creep and fatigue life for a component. The computed fatigue life divided by the specified fatigue life is defined as the life fraction.

11.11 National Fire Protection Association purge credit

The National Fire Protection Association (NFPA) in 2011 implemented purge credit criteria to exempt the need for a gas turbine exhaust purge at startup thus shortening the time for startup. This was later revised in 2015 (Ref. [10]). This practice utilizes added hardware, interlocks, and controls to avoid the need for the startup purge time. It also minimizes the amount of condensate generated in superheaters and reheaters during warm and hot start conditions. There are two methods for gaseous fuels and three methods for liquid fuels. The gaseous fuels options are a valve proving method for a credit up to eight days and a pressurized pipe section method, which is for an indefinite period. For liquid fuels, there is a valve proving method for a credit up to eight days, a pressurized pipe section method for an indefinite period and a liquid level monitoring method for an indefinite period. In all methods, there are triple block and double bleed valves required for the fuel. For pressurized pipe section methods, there is an additional requirement for a pressurized gas purge between the last two block valves. The supply air also requires double block and bleed valves. For the liquid level monitoring method, a vent and level detection is included between the second and last block valves.

11.12 Miscellaneous cycling considerations

Fatigue is not the only concern associated with cycling. Many components are affected by cycling and therefore there is a need for more focused operation and increased maintenance of units that are cycled. This section addresses miscellaneous topics related to cycling.