Creep and Creep Fracture

section epub:type=”chapter”>



Creep and Creep Fracture



21.1 Introduction


So far we have concentrated on mechanical properties at room temperature. Many structures—particularly those that are associated with energy conversion (e.g., turbines, reactors, steam, and chemical plant)—operate at much higher temperatures.


At room temperature, most metals and ceramics deform in a way that depends on stress but which, for practical purposes, is independent of time:


ɛ=f(σ)elastic/plastic solid



As the temperature is raised, loads that give no permanent deformation at room temperature cause materials to creep. Creep is slow, continuous deformation with time: the strain, instead of depending only on the stress, now depends on temperature and time as well:


ɛ=f(σ,t,T)creeping solid



It is common to refer to the former behavior as “low-temperature” behavior, and the latter as “high temperature.” But what is a “low” temperature and what is a “high” temperature? Tungsten, used for lamp filaments, has a very high melting point—well over 3000°C. Room temperature, for tungsten, is a very low temperature. If made hot enough, however, tungsten will creep—that is the reason that lamps ultimately burn out. Tungsten lamps run at about 2000°C—this, for tungsten, is a high temperature. If you examine a lamp filament that has failed, you will see that it has sagged under its own weight until the turns of the coil have touched—that is, it has deformed by creep. Figure 21.1 shows a typical example of a sagging filament.


f21-01-9780081020517
Figure 21.1 A tungsten lamp filament which has sagged under its own weight owing to creep.

Figure 21.2 and Table 21.1 give melting points for metals and ceramics and softening temperatures for polymers. Most metals and ceramics have high melting points and, because of this, they start to creep only at temperatures well above room temperature—this is why creep is a less familiar phenomenon than elastic or plastic deformation. But the metal lead, for instance, has a melting point of 600 K; room temperature, 300 K, is exactly half its absolute melting point. Room temperature for lead is a high temperature, and it creeps—as Figure 21.3 shows. And the ceramic ice melts at 0°C. Temperate glaciers (those close to 0°C) are at a temperature at which ice creeps rapidly—that is why glaciers move. Even the thickness of the Antarctic ice cap, which controls the levels of the earth’s oceans, is determined by the creep spreading of the ice at about –30°C.


f21-02-9780081020517
Figure 21.2 Melting or softening temperature.


f21-03-9780081020517
Figure 21.3 Lead pipes often creep noticeably over the years. (Source: M.F. Ashby.)

The point, then, is that the temperature at which materials start to creep depends on their melting point. As a general rule, it is found that creep starts when


T>0.3to0.4TMfor metalsT>0.4to0.5TMfor ceramics



where TM is the melting temperature in degrees kelvin. However, special alloying procedures can raise the temperature at which creep becomes a problem.


Polymers, too, creep—many of them do so at room temperature. As we said in Chapter 5, most common polymers are not crystalline and have no well-defined melting point. For them, the important temperature is the glass temperature, TG, at which the Van der Waals bonds solidify. Above this temperature, the polymer is in a leathery or rubbery state and creeps rapidly under load. Below, it becomes hard (and sometimes brittle) and, for practical purposes, no longer creeps. TG is near room temperature for most polymers, so creep is a problem.


In design against creep, we select the material and the shape that will carry the design loads, without failure, for the design life at the design temperature. The meaning of “failure” depends on the application. We distinguish four types of failure, illustrated in Figure 21.4.



  1. 1. Displacement-limited applications, in which precise dimensions or small clearances must be maintained (as in the discs and blades of gas turbines).
  2. 2. Rupture-limited applications, in which dimensional tolerance is relatively unimportant, but fracture must be avoided (as in pressure-piping).
  3. 3. Stress relaxation-limited applications in which an initial tension relaxes with time (as in the pretensioning of bolts).
  4. 4. Buckling-limited applications, in which slender columns or panels carry compressive loads (as in structural steelwork exposed to a fire).

f21-04-9780081020517
Figure 21.4 Creep is important in four classes of design: (a) displacement-limited, (b) failure-limited, (c) relaxation-limited, and (d) buckling-limited.

To analyze these we need constitutive equations which relate the strain-rate ɛ˙si1_e or time-to-failure tf to the stress σ and temperature T.


21.2 Creep Testing and Creep Curves


Creep tests require careful temperature control. Typically, a specimen is loaded in tension, usually at constant load, inside a furnace maintained at a constant temperature, T. The extension is measured as a function of time. Figure 21.5 shows a typical set of results from such a test. Metals, polymers, and ceramics all show creep curves of this shape.


f21-05-9780081020517
Figure 21.5 Creep testing and creep curves.

Although the initial elastic and the primary creep strain cannot be neglected, they occur quickly, and they can be treated in much the way that elastic deflection is allowed for in a structure. But thereafter, the material enters steady state, or secondary creep, and the strain increases steadily with time. In designing against creep, it is usually this steady accumulation of strain with time that concerns us most.


By plotting the log of the steady creep rate, ɛ˙sssi6_e, against log σ at constant T, as shown in Figure 21.6, we can establish that


ɛ˙ss=Bσn



si7_e  (21.1)


where n, the creep exponent, usually lies between 3 and 8. This sort of creep is called “power-law” creep. (At low σ, a different régime is entered where n ≈1; we shall discuss this low-stress deviation from power-law creep in Chapter 23.)


f21-06-9780081020517
Figure 21.6 Variation of creep rate with stress.

By plotting the natural logarithm (ln) of ɛ˙sssi6_e against the reciprocal of the absolute temperature (1/T) at constant stress, as shown in Figure 21.7, we find that:


ɛ˙ss=Ce(Q/RT)



si9_e  (21.2)


f21-07-9780081020517
Figure 21.7 Variation of creep rate with temperature.

Here R is the Universal Gas Constant (8.31 J mol−1 K−1) and Q is called the Activation Energy for Creep—it has units of J mol−1. Note that the creep rate increases exponentially with temperature (Figure 21.7). An increase in temperature of 20°C can double the creep rate.


Combining these two dependences of ɛ˙sssi6_e gives


ɛ˙ss=Aσne(Q/RT)



si11_e  (21.3)


where A is the creep constant. The values of the three constants A, n, and Q charactize the creep of a material; if you know these, you can calculate the strain rate at any temperature and stress by using the last equation. They vary from material to material, and have to be found experimentally.


21.3 Creep Relaxation


At constant displacement, creep causes stresses to relax with time. Bolts in hot turbine casings must be regularly tightened. Plastic paper clips are not, in the long-term, as good as steel ones because, even at room temperature, they slowly lose their grip.


The relaxation time (arbitrarily defined as the time taken for the stress to relax to half its original value) can be calculated from the power-law creep data as follows. Consider a bolt that is tightened onto a rigid component so that the initial stress in its shank is σi. In this geometry (refer to Figure 21.4(c)) the length of the shank must remain constant—that is, the total strain in the shank ɛtot must remain constant. But creep strain ɛcr can replace elastic strain ɛel, causing the stress to relax. At any time t


ɛtot=ɛel+ɛcr



si12_e  (21.4)


But


ɛel=σ/E



and (at constant temperature)


ɛ˙cr=Bσn



Since ɛtot is constant, we can differentiate Equation (21.4) with respect to time and substitute the other two equations into it to give


1Edσdt=Bσn



si15_e  (21.6)


Integrating from σ = σi at t = 0 to σ = σ at t = t gives


1σn-11σin-1=(n1)BEt



Figure 21.8 shows how the initial elastic strain σi/E is slowly replaced by creep strain, and the stress in the bolt relaxes. If, as an example, it is a casing bolt in a large turbogenerator, it will have to be retightened at intervals to prevent steam leaking from the turbine. The time interval between retightening, tr, can be calculated by evaluating the time it takes for σ to fall to (say) one-half of its initial value. Setting σ = σi/2 and rearranging gives


tr=(2n11)(n1)BEσin-1



si17_e  (21.8)


f21-08-9780081020517
Figure 21.8 Replacement of elastic strain by creep strain with time at high temperature.

Experimental values for n, A, and Q for the material of the bolt thus enable us to decide how often the bolt will need retightening. Note that overtightening the bolt does not help because tr decreases rapidly as σi increases.


21.4 Creep Damage and Creep Fracture


During creep, damage accumulates in the form of internal cavities. The damage first appears at the start of the tertiary stage of the creep curve and grows at an increasing rate thereafter. The shape of the tertiary stage of the creep curve (refer to Figure 21.5) reflects this: as the cavities grow, the section of the sample decreases, and (at constant load) the stress goes up. Since ɛ˙σn,si18_e the creep rate goes up even faster than the stress does (Figure 21.9).


f21-09-9780081020517
Figure 21.9 Creep damage.

It is not surprising—since creep causes creep fracture—that the time-to-failure, tf is described by a constitutive equation which looks very like that for creep itself:


tf=Aσme+(Q/RT)



si19_e


Here A′, m, and Q are the creep-failure constants, determined in the same way as those for creep (the exponents have the opposite sign because tf is a time whereas ɛ˙sssi6_e is a rate).


In many high-strength alloys this creep damage appears early in life and leads to failure after small creep strains (as little as 1%). In high-temperature design it is important to make sure:



  •  That the creep strain ɛcr during the design life is acceptable
  •  That the creep ductility ɛf,crsi21_e (strain to failure) is adequate to cope with the acceptable creep strain
  •  That the time-to-failure, tf, at the design loads and temperatures is longer (by a suitable safety factor) than the design life

Times-to-failure are normally presented as creep-rupture diagrams (Figure 21.10). Their application is obvious: if you know the stress and temperature you can read off the life; if you wish to design for a certain life at a certain temperature, you can read off the design stress.


f21-10-9780081020517
Figure 21.10 Creep-rupture diagram.

21.5 Creep-Resistant Materials


From what we have said so far it should be obvious that the first requirement that we should look for in choosing materials that are resistant to creep is that they should have high melting (or softening) temperatures. If the material can then be used at less than 0.3 of its melting temperature creep will not be a problem. If it has to be used above this temperature, various alloying procedures can be used to increase creep resistance. To understand these, we need to know more about the mechanisms of creep—the subject of the next two chapters.


Worked Example


The photograph on the next page shows an easy experiment you can rig up at home to demonstrate creep—and even measure the creep exponent, n. Take a length of wooden curtain pole or broom handle (mine was 28.5 mm diameter—1⅛”), and wrap electronic solder wire (mine was 1.60 mm diameter—116si22_e”) around the pole to make a coil that has no gaps between the turns. I wound 36 turns of wire on to the pole. Then hook the end of the coil over the end of the pole as shown, and put the axis of the pole vertical (hold the lower end in a vice or clamp). Do the experiment indoors in the warm, and within a few minutes the coil will start to open out. After about 15 minutes it will look like the one in the photograph. Obviously the solder (an alloy of lead and tin) has experienced considerable creep deformation. This is exactly what you would expect—electronic solder melts at 183°C (456 K), so a room temperature of 20°C (293 K) is 0.64 TM—well into the régime where creep will occur.


You can see from the photograph that the coils have opened out much more at the top (where the wire supports the maximum load) than they have toward the bottom (where the wire supports a much smaller load). This tells you that the creep rate is increasing with stress, as it should do. But how fast does it increase? The value of the creep exponent n can easily be found as follows.


f21-11-9780081020517

Starting from the top of the coil, measure the positions of the crests of the wire (e.g., 1, 18, 34, 49, 63 mm). List the differences between these measurements which is the pitch of each turn (e.g., 17, 16, 15, 14 mm) and correlate these pitches with the number of turns of wire they support (e.g., 36, 35, 34, 33). The opening-out of each turn is found by subtracting the wire diameter (1.6 mm) from the pitch, giving 15.5, 14.5, 13.5, 12.5 mm, and so on (rounded to the nearest 0.5 mm).


Then plot log (opening out) versus log (number of turns supported), e.g., log 15.5 versus log 36, log 14.5 versus log 35, etc. The full graph is shown in the following diagram. Over most of the range, there is an excellent fit to a straight line with a slope of 2.1, so n = 2.1. Only toward the bottom of the coil do the data points go off, but this is because it is very difficult to make accurate measurements of pitch from the photograph when the turns have opened out so little.


f21-12-9780081020517

Examples




  1. 21.1 A cylindrical tube in a chemical plant is subjected to an excess internal pressure of 6 MN m−2, which leads to a circumferential stress in the tube wall. The tube wall is required to withstand this stress at a temperature of 510°C for 9 years. A designer has specified tubes of 40 mm bore and 2 mm wall thickness made from a stainless alloy of iron with 15% by weight of chromium. The manufacturer’s specification for this alloy gives the following information.

    Over the present ranges of stress and temperature the alloy can be considered to creep according to the equation

    ɛ˙=Aσ5eQ/RT



    where A and Q are constants, R is the universal gas constant, and T is the absolute temperature. Given that failure is imminent at a creep strain of 0.01 for the present alloy, comment on the safety of the design.
  2. 21.2 An alloy tie bar in a chemical plant has been designed to withstand a stress, σ, of 25 MN m−2 at 620°C. Creep tests carried out on specimens of the alloy under these conditions indicated a steady-state creep rate, ɛ˙si1_e, of 3.1 × 10−12 s−1. In service it was found that, for 30% of the running time, the stress and temperature increased to 30 MN m−2 and 650°C. Calculate the average creep rate under service conditions. It may be assumed that the alloy creeps according to the equation

    ɛ˙=Aσ5eQ/RT



    where A and Q are constants, R is the universal gas constant, and T is the absolute temperature. Q has a value of 160 kJ mol−1.
  3. 21.3 The window glass in old buildings often has an uneven surface, with features that look like flow marks. The common explanation is that the glass has suffered creep deformation over the years under its own weight. Explain why this scenario is complete rubbish (why do you think the glass does appear to have “flow marks”?).
  4. 21.4 Why do bolted joints in pressure vessels operating at high temperature have to be retorqued at regular intervals?
  5. 21.5 Why are creep–rupture diagrams useful when specifying materials for high-temperature service?
  6. 21.6 Structural steelwork in high-rise buildings is generally protected against fire by a thick coating of refractory material even though a fire would not melt the steel. Explain.
  7. 21.7 Aluminum alloy (TM = 920 K) pistons in large diesel engines are subjected to working temperatures of around 150°C (423 K). When the engine is started up, thermal gradients in the pistons produce thermal strains (coefficient of thermal expansion × temperature difference) of the order of 0.3%. Young’s modulus for the alloy at 150°C is 74 GN m−2, so assuming linear elastic material behavior the thermal stress is calculated to be 220 MN m−2 (74,000 × 0.003). The tensile ductility (failure strain) of the alloy is 0.7% at 150°C, so the pistons should not fail (and in fact they don’t). However, the tensile strength of the alloy is 170 MN m−2 at 150°C, so the linear elastic stress analysis indicates that the pistons should fail! Explain the reason for this apparent discrepancy.
  8. 21.8 What are the requirements when designing against creep in high-temperature applications?
  9. 21.9 Why does electronic solder creep rapidly at room temperature?
  10. 21.10 Rotary furnaces are large tubes of creep-resistant metal (typically 2 m diameter, 10 m long). The furnace is supported at each end by a ring, which sits on a pair of rollers. The rollers are driven by an electric motor, and they rotate the tube slowly about its axis. The outside of the tube is heated by gas burners. When warming the furnace up, or cooling it down again, it is vital that it is kept rotating. What is the reason for this?

Answers




  1. 21.1 The main steps in this question are as follows. (i) Add 273 to each temperature, to convert it to K. (ii) Get the reciprocal of each temperature, 1/T (it will be in units of K−1). (iii) Get the natural log (ln) of each creep rate. (iv) Plot ln (creep rate) versus 1/T, as in Figure 21.7. The data points will fall on a straight line. (v) Extrapolate this line to 1/(783 K), corresponding to 510 °C. (vi) This gives ln (creep rate) = –20.90, or creep rate itself = 8.3 × 10–10 s−1. (vii) Hoop stress in tube wall σ = pr/t, giving σ = 60 MN m−2. (viii) Reduce strain rate at 510 °C for stress, multiplying 8.3 × 10−10 s−1 by the factor (60/200)5. This gives a strain rate of 2 × 10−12 s−1. The creep strain after 9 years is then 0.00057, so the design is safe.
  2. 21.2 There are three steps in this question. (i) Find the strain rate at the same (design) temperature of 620 °C, but at the elevated stress of 30 MN m−2. This means multiplying 3.1 × 10–12 s−1 by the factor (30/25)5, giving a strain rate of 7.71 × 10−12 s−1. (ii) Correct this strain rate to a temperature of 650 °C. This means using the natural log (ln) form of the creep equation

    lnɛ.=lnAσ5QRT



    si26_e


    (note 5 is now a constant), writing it out twice, and subtracting

    lnɛ.1lnɛ.2=QR1T11T2



    to give ɛ.2si28_e = 15.5 × 10−12 s−1. When putting numbers into this equation, it is critical to input temperatures in K, and to input Q values in J mol−1 (not kJ mol−1). This is because R = 8.314 J mol−1 K−1, and Q/RT must be dimensionless (it is an exponent).
    (iii) Work out the average creep rate, from

    ɛ.av=ɛ.1×70100+ɛ.2×30100



    This gives an average creep rate of 6.82 × 10−12 s−1.
  3. 21.3 From Table 21.1, the softening temperature of soda glass is in the range 700–900 K. The operating temperature of window glass is rarely more than 293 K, so T/Ts is at most 0.42. Ceramics only begin to creep when T/TM > 0.4 (see Section 21.1), and then only under a large stress (far greater than the self-weight stress). Thus, the flow marks cannot possibly be due to creep. In fact, they come from the process used to make panes of glass in the past, which consisted of blowing a large “bubble” (or boule) of molten glass, then cutting-open the boule and flattening it. Deforming the molten glass produced the flow marks. An “urban myth” busted!
  4. 21.4 See Section 21.3.
  5. 21.5 See Section 21.4 and Figure 21.10.
  6. 21.6 A major fire would increase the temperature of the steelwork to the point at which it would creep under the applied loads, and the subsequent deformation could trigger the collapse of the building. This is why the World Trade Center towers collapsed on 9/11. The vertical load-bearing columns suffered creep buckling (see Figure 21.4(d)).
  7. 21.7 The linear elastic stress analysis ignores the fact that the material will deform by plastic deformation and creep, which will keep the actual stress well below the tensile strength. For this reason, the correct failure criterion is the one based on strain, not stress.
  8. 21.8 See Section 21.4, next-to-last paragraph.
  9. 21.9 See Worked Example, paragraph 1.
  10. 21.10 If it were not kept turning, the tube would sag by creep deformation, and end up permanently bent. If the tube is kept turning, every part of the metal experiences repeated cycles of alternating tension and compression. The positive and negative creep strains produced by these stress cycles cancel out over time.

Aug 9, 2021 | Posted by in General Engineer | Comments Off on Creep and Creep Fracture
Premium Wordpress Themes by UFO Themes