You are on page 1of 5

PEER-REVIEWED PIPING

Are your thermowells safe?


DAVID S. BARTRAN, DONALD R. FRIKKEN,
AND

RICHARD YEE

ABSTRACT: This report presents a new method for establishing the application limits of thermowells.
The method emphasizes use of a principal stress model and relies on finite element, dynamic analysis techniques. While most of thermowell applications in industry are not at risk, the proposed analysis shows that a surprising number are exposed to service stresses known to cause failure. Application: Mills can use the proposed method to help in the selection and application of thermowells.
dequate static pressure rating and use of corrosion resistant materials are generally sufficient to guarantee the mechanical integrity of most thermowell designs. High velocity and corrosive service applications, however, require a more comprehensive basis for design and selection. At present, manufacturers publish thermowell velocity ratings as a guide for thermowell selection. Murdocks stress model [1] and its companion, the Performance Test Code, PTC 19.3 [2], are the most commonly cited sources for these ratings. While they have been responsible for a drastic reduction in the rate of catastrophic failures [3], at least in powerhouse applications, they are not universally accepted. This lack of credibility has led to and perpetuated the use of traditional selection methods originally developed to deal with the immersion errors in filled system temperature measurements. One such rule requires that the thermowell extend into the central third of the pipe, with resulting lengths ranging from onethird to two-thirds of the pipe diameter. While this often results in thermowells that violate PTC 19.3 recommendations, they have proven to be surprisingly successful. There are no statistics dealing with the effectiveness of these methods or rates of thermowell failure. In an informal survey, 20%-30% of the applications were found to be at some level of identifiable risk (small-to-moderate resonant stresses), with 1%-3% of the applications in a high risk category (large resonant stresses comparable to or greater than the maximum allowable stress).This is followed by a very small number of applications, <0.01%, which were exposed to application stresses exceeding the yield strength of the material and at risk of catastrophic failure. These rates represent the selection method, not the actual failures, but it

is a shock to find so many thermowells exposed to stress levels normally associated with unsafe designs in piping components. In spite of this, traditional methods appear to be successful, but the high rate of misapplication sets the stage for a completely different outcome in corrosive or even slightly corrosive services. In effect, the misapplied thermowells form a pool of applications at increased risk depending upon the materials of construction and the corrosion specifics. Oddly, in sorting out the details of a stress model suited for corrosive service applications, we found that PTC 19.3 itself contributes to the pool of misapplied thermowells. This is regrettable, given that it is the de facto industry standard and as is most likely to be used as a guide in high-risk services. The most important factor in this finding is that PTC 19.3 is based on an incomplete stress analysis and it ignores the most common mode of thermowell failure, namely, bending fatigue. The following represents a brief summary of the critical issues and describes a self-consistent stress model for thermowell design and selection. It does not attempt to delve into the intricacies of the modal analysis or the finite element method used in the sample calculations.

Fluid Velocity

V
X Z (X) Y (0) =

X=L

X=0

1. Cantilever coordinates for the continuous beam model (see ref. 7).

Thermowell stresses
Under static conditions, the thermowell design is controlled by its hoop stresses. For thick walled, drilled barstock wells, these are so much greater than the design ratings of either the adjacent piping or the process connection that static pressure failure is a practical impossibility.An entirely different situation emerges under flowing conditions because of the rapid increase in secondary stresses with flow. The most important of these secondary stresses are the result of flow-induced Fig. 1 ) produced by bending (F conventional drag and vortex shedding processes.

Unlike shear failure invoked in the Murdock (and PTC 19.3) analysis, the flow-induced bending stresses are longitudinal and greatest on the outside surface of the thermowell. If the fluid forces were static, then an easily calculated windage limit would characterize the thermowell where the thermowell would simply fold over. Such failure is not likely in closed conduits and can only occur in cases where the natural modes are not excited to resonance and the fluid velocity is extremely high. Most failures occur at surprisingly common fluid velocities and are usually the result of flow-induced resonance. The velocity dependent forces acting on the thermowell are directed along the flow direction and transverse to it [4,5]. They are commonly represented as a vector force: F(x,t) = (FD + Fd Sin 2k t ) y + Fk Cos k t z . (1) On a unit length basis, the force amplitudes have the form: F = f D(x) CV2/2, where the subscript ={D,d,k} correspond to conventional (in-line) drag, oscillating (in-line) drag, and oscillating lift (transverse) forces. Averaging Eq(1) over the thermowell span and defining a

10

TAPPI JOURNAL APRIL 2002

PIPING

coherent vortex shedding rate on a tip basis results in a conservative estimate of the excitation forces and resultant bending stresses. The vortex shedding rate is given by k=2 S V/D.The Strouhal number is S ~ 0.22, and the force coefficients: CD, Ck ~1 and Cd ~ 0.1, are all constants. Note that the in-line force component inherently oscillates at twice the vortex shedding rate. In the slender beam, small amplitude limit, the in-line and transverse responses are independent and can be solved separately using conventional analysis methods. The two responses have identical mode structures and critical frequencies, so the solutions only differ in their respective excitations. Since the in-line force cycles at twice the rate of the transverse force, the combined response exhibits two velocity criticals for each bending mode: one in-line and the other transverse. The in-line critical (or resonance) occurs at half the velocity of the transverse resonance. As a rule, transverse resonance dominates thermowell failures in steam and gas services while in-line resonance, with its lowered velocity requirement, dominates those in liquids. In dense fluids, namely steam above 1500 psig (10 MPa), and in liquids, both possibilities must be considered. At least one thermowell failure [6], albeit an extreme case, occurred as a result of extending Murdocks stress model to liquid services. Significantly, these same concerns apply to its use in high pressure steam, where the PTC is supposed to provide credible design advice. The longitudinal bending stresses are greatest at the base of the thermowell and are distributed about the neutral axis along the lines of an eccentrically loaded beam, the peak stress is given by: S = Sl SD + Sd Sin 2kt + Sk Cos kt. (2) The equation uses a positive sign for the peak stress on the downstream side of the beam and a negative sign for the peak stress on the upstream side of the thermowell. The longitudinal stress components, all taken as positive quantiities, are: Sl (static pressure), SD (drag), Sd (oscillating drag), and Sk (oscillating lift), which for a static pressure, Po, are given by [7]:

1,000,000

LONGITUDINAL STRESS (psi)

100,000

Static MAWS =19300 psi (133.0 MPa) Fatigue Allowable =9650 psi (66.5 MPa) 10,000

1000 Total Stress Dynamic Stress 100 1 10

316 SS Straight 50 PSIG Acid / 400F B31.3 Design Basis Velocity = 10.5 ft/s(3.2 m/s) Length =16"(154 mm) O.D. = 0.875"( 10.4 mm) Bore = 0.260"(6.3 mm) Tip Thickness=0.25"(6.3 mm) 100

VELOCITY (ft/s)

2. An application where sensor failure occurred as a result of excessive acceleration.

Sl = PoD2/(D2-d2) SD = D

(3) (4)

Sd = d ci [ 1 - (2 / i )2 + 4 j i / i ]-1 (5) Sk = k ci [ 1 - ( / i)2 + 2 j i / i]-1 (6) These equations are not as imposing as they seem; the summations represent a net magnification ratio for the respective in-line and transverse responses.The individual modes are characterized by the modal amplitudes, ci, natural frequencies, i, and relative damping, i. Only the lower order modes are important, so the summations are quite manageable. The zero-frequency, stress amplitudes, , are proportional to the aerodynamic pressure. For straight shank thermowells have the form = fCD2L2V2/8I. As a minimum, even without full knowledge of the natural frequencies of the thermowell, the static windage stress is bounded by ~D and the peak resonant stress is bounded by ~D/2. In the absence of actual data, a relative damping of ~ 0.01 can be assumed.

stress model, but limits consideration to the peak shear stress on the interior surface of the thermowell.This requires that failures originate at a point completely isolated from the process interface and cannot account for localized stress reversals that figure so importantly in known failures. As described earlier, the longitudinal bending stresses control the design at high velocities, with localized tensile failure occurring long before the shear limit is reached. For design, the sum of the static stresses caused by pressure and drag should not exceed the maximum allowable MAWS): working stress of the material (M Sl + SD MAWS (7)

Additionally, the high cycle dynamic stress components should not exceed the fatigue allowable stresses: Sd + Sk K MAWS (8)

Design criteria
The Murdock stress model includes conventional drag and oscillating lift in the

where K is the high-cycle, stress-reduction factor for the given application and materials. This new criteria is more restrictive than that used previously [7], but is recommended for those cases where cyclical stresses dominate and corrosion is present. The allowable tensile MAWS) for common materials stresses (M
VOL. 1: NO. 2 TAPPI JOURNAL

11

PIPING

100,000

Sample calculations
The following four examples demonstrate the proposed method and illustrate the sort of applications where the thermowell is at risk of failure.Three of these involve cases where failures actually occurred, with the fourth being an example taken from the current PTC 19.3. All four illustrate how the proposed model can be used to identify high risk applications using information available in the design stage. The longitudinal stresses are calculated by a combination of finite element and modal analysis techniques [9]. The mechanical properties of Type 316 stainless steel are used in the calculations, except in the PTC example, where Type 321 stainless steel is assumed.The respective figures indicate dimensions, service conditions, and temperatures. The maximum allowable static stress is taken from B31.3 (Process Piping) for the first three examples, while B31.1 (Power Piping) is used in the fourth, and the fatigue allowable stresses are assumed to be one-half the allowable static tensile stress. Fig. 2) involves the The first example (F failure of a platinum resistance element due to acceleration stresses of soldered extension wire connections. This failure could not be fully explained at the time, but the model calculations clearly show that the thermowell was operated just below the onset of the oscillating drag condition, with a sustained tip acceleration >5 gs. Use of a metal sheathed sensor with high temperature joining techniques eliminated additional failures, but raised questions about suitability of resistance sensors in high velocity services. Fig. 3), was in a The second example, (F liquid sodium cooling loop [6] where thermowell failure led to a coolant spill and subsequent fire.The reported conditions associated with this failure coincide with the in-line resonance condition, with stress levels approaching the allowable stress of the material.The calculated tip acceleration (not shown) is ~150 gs at the in-line resonance condition and compares with the measured tip acceleration of ~260 gs derived from water tests. When the maximum allowable stress is derated to allow for liquid metal embrittlement, it is obvious that this thermowell is predisposed to failure. The

LONGITUDINAL STRESS (psi)

Static MAWS =19300 psi (133.0 MPa) Fatigue Allowable =9650 psi (66.5 MPa)

10,000

1000

Total Stress

Cyclic Stress

316 SS Straight 50 PSIG Liquid Sodium/400F B31.3 Design Basis Velocity = 17.1 ft/s(5.2 m/s) Length =6.1"(155 mm) O.D. = 0.409"( 10.4mm) Bore = 0.150"(3.8 mm) Tip Thickness=0.125"(3 mm) 100

100 1 10

VELOCITY (ft/s)

3. An application where thermowell cracking resulted in a loss of process containment, and a subsequent fire [6].
100,000

LONGITUDINAL STRESS (psi)

Static MAWS =19300 psi (133.0 MPa) Fatigue Allowable =9650 psi (66.5 MPa) 10,000

1000 316 SS Straight 150 PSIG Steam / 400F Velocity = 229 ft/s(70 m/s) Length =7"(178 mm) Root O.D. = 0.965"( 24.5 mm) Bore = 0.260"(6.6 mm) Tip Thickness=0.25"(6.3 mm) 100 1000

Total Stress Dynamic Stress

100 10

VELOCITY (ft/s)

4. A moderate pressure steam application where the thermowell is misapplied.

are tabulated in the piping codes [8].You can use the piping code rules to estimate the fatigue allowable stress in the absence of actual data. The maximum allowable tensile stress is used to set the static pressure rating of the thermowell although these stresses are actually compressive. This is in line with current industry practice and is a

matter of agreed convention. For reference, the hoop stress Sh = 2P0D2/( D2-d2) controls the pressure rating of the shank, while the peak radial stress Sr ~ 3Po(d/)2 /16 controls that of the tip. The thermowell dimensions are outside diameter, D, bore d, tip thickness and overall length L.

12

TAPPI JOURNAL APRIL 2002

PIPING

100,000

CONCLUSIONS
PTC 19.3 Example 321 SS Tapered Well 2400 PSIG Steam-1050F B31.1 Design Basis Velocity = 300 ft/s (91.4 m/s) Length =4.5"(114 mm) Root O.D. = 1.125"( 28.6 mm) Tip O.D. = 0.9375"( 23.8 mm) Bore = 0.575"(14.6 mm) Tip Thickness=0.1875"(5 mm) Static MAWS =9200 psi (63.4 MPa)

10,000

Fatigue Allowable =4600 psi (31.7 MPa) 1000 Total Stress Dynamic Stress

100 10

100

1000

VELOCITY (ft/s)

5. A thermowell design predisposed to failure and based on PTC 19.3 [2].

authors recommended use of a tapered thermowell to correct the design. This modification increases the resonant frequency (the principle benefit) and reduces vortex shedding coherence, with a net reduction in the root stresses to much safer levels. Fig. 4), was a modThe third example (F erate pressure steam application at 150 psig (~1.0 Mpa) where the thermowell was reportedly damaged in service. Unfortunately, this example is based on a second hand report (a common problem with thermowell failures); however, it demonstrates the sort of applications encountered in operating plants. The most significant aspect of this application is the presence of a high stress, oscillating lift condition at ~200 ft/s. This is just below the maximum normal flow condition and would be cause for concern if corrosion were present. It is also characterized by a lower stress, oscillating drag condition at ~100 ft/s that does not present a particular application risk in this service. The presence of a high stress critical in the normal range of flows usually requires an alternative thermowell selection. The preferred method is to reduce thermowell length and shift all resonances well above the normal range of flows. This is easily accomplished with off-the-shelf designs, but in this case, the

7 in. design was actually replaced with a longer one. While the proposed stress model permits any design change that shifts the high stress criticals outside the normal range of flows, moving them to lower velocities is normally advisable only when the flow conditions are well defined and a reliable stress prediction method is used. The fourth example is a high-pressure steam application used in PTC 19.3 to demonstrate its design method. Figure 5 shows the dynamic response, according the new calculation. It is characterized by an oscillating drag critical at ~290 ft/s (88 m/s) and oscillating lift critical at ~575 ft/s (175 m/s), both of which produce resonant stresses in excess of the MAWS of the material.The peak stress criteria limits the maximum service velocity to ~275 ft/s (85 m/s). The critical avoidance requires that the service velocity be restricted to less than 80% of the flow critical or 232 ft/s (70 m/s)! All of these are considerably below the PTC rating for this design of ~392 ft/s (120 m/s).To meet the stated service requirements when both thermowell stress and sensor acceleration are taken into account, thermowell length should be restricted to 2 in. (50 mm), again considerably less than the 7 in. (180 mm) permitted for this design under the PTC rules.

The application limits of thermowells are easily established.The hoop stress in the shank and the radial stresses of the tip determine the pressure rating, although process connection design usually governs the overall assembly. The localized bending stresses produced by conventional drag and vortex shedding forces determine the velocity rating.A principal stress theory of failure offers the simplest means for identifying the elastic limits of the thermowell when used in conjunction with allowable stress tables contained in the relevant piping code, typically B31.1 or B31.3.The proposed model is generally more conservative than the PTC rating method and it has proven successful in explaining nearly a dozen thermowell failures. Use of finite element and dynamic analysis methods accuracy in predicting thermowell ratings are not an absolute requirement. For design and selection, simplified spreadsheet calculations are frequently more than adequate, as long a principle stress model is used and the dynamic analysis represents spanwise excitation of the thermowell. In pulp and paper mills, the over-all risk of thermowell misapplication is moderate, because the greatest mass flows and highest velocities are in noncorrosive, medium pressure steam applications. While common thermowell designs tolerate a good deal of abuse in such services, it would be wrong to characterize any thermowell application as safe without at least, a cursory check of the fluid velocity, windage, and peak resonant stresses. TJ

LONGITUDINAL STRESS (psi)

LITERATURE CITED
1. Murdock, J.W., ASME Journal of Engineering for Power, Oct. 1959, pp. 403-416. 2. ASME Performance Test Code, PTC 19.3, 1974. 3. Love, T.J., Power (Design and Equipment Application Section), Feb. 1961, pp. 80-81. 4. ASME Boiler and Pressure Vessel Code, Article N-1300. 5. Blevins, R. D., Flow-induced vibration, Krieger Publishing, Malarbar, Florida, 1994. 6. Ogura K., Morishita, M., Yamaguchi, A., ASME Trans. PVP-Vol.363, 1998, pp. 109-117.

VOL. 1: NO. 2 TAPPI JOURNAL

13

PIPING

7. Bartran, D.S., Frikken, D.R., Kinsey, J.M., Schappelle, R., Yee, R., ISA Trans., 39, 2000, pp. 133-142. 8. ASME Code for Pressure Piping, B31. 9. Thompson, W.T., Dahleh, M.D., Theory of vibration, Prentice-Hall, 5th Edition, 1998.

Received: July 14, 2001 Revised: October 13, 2001 Accepted: October 14, 2001

This paper is also published on TAPPIs web site (www.tappi.org) and summarized in the April Solutions! for People, Processes and Paper magazine (Vol. 85 No. 4).

INSIGHTS FROM THE AUTHORS What began as a routine attempt to develop velocity ratings for corrosive services based on the accepted standard for steam services turned into an unexpected challenge. In the process of sorting out the details of the calculation and updating the allowable stress data, we uncovered a number of inconsistencies we could not answer within the framework of the existing standard. This led to a complete restatement of how thermowell application limits are set. Most previous work focuses on the primary membrane stresses normally used to define the pressure rating of the thermowell, and virtually ignores the localized secondary stresses that result from flow induced bending. The only exception to this was some early work (1961) by Tom J. Love at the University of Oklahoma. The primary stresses correctly determine the collapse strength (or rupture) of the thermowell, while the secondary stress offer a more credible description of the stresses characterizing the actual pattern of failure. The most difficult part of this effort was putting aside cherished notions about what constitutes an acceptable thermowell selection and learning to look upon the thermowell as a mechanical device with competing thermal design requirements. This has been one of the most challenging efforts that we, as practicing engineers, have dealt with in quite some time. The greatest surprise was recognizing and accepting that thermowells can and do fail, and that such a great number of thermowells are at risk of failure simply for having been selected without

regard to their actual design limitations. Of course, as Murphys Law would have it, only those thermowells in severe or high consequence applications are the ones that ever seem to fail. Safety and process integrity are important issues in all paper mills. We all have gained a greater appreciation for the mechanical limitations of thermowells and no longer take any application for granted. Readers can easily reduce the proposed method to practical estimates capable of providing credible advice in all of the cases we have considered. The tip design is far more critical to personnel safety than one might expect and the practice of using rods to clear test wells of debris before taking measurements needs to be approached cautiously. As a next step, we would like to see wider appreciation of thermowells as critical elements in the mechanical integrity of the process system and an improvement in current industry standards for their design and application. David S. Bartran
Bartran and Yee are with Monsanto Enviro-Chem Systems; Frikken is with Solutia Inc., Chesterfield, Missouri; contact Bartran by email at david.s.bartran@monsanto.com.

Bartran

Frikken

Yee

Subscribe to the New

JOURNAL

TAPPI members can subscribe to the printed TAPPI JOURNAL for US$ 100. The Journal of Pulp and Paper Science (JPPS) is also available by subscription to TAPPI members for US$ 100, or US$ 175 for both TAPPI JOURNAL and JPPS. To subscribe, contact TAPPIs Member Connection Center: Fax 770-446-6947, Phone: 1 800 332-8686 (USA), 1 800 446-9431 (Canada), or +1 770 446-1400; email: memberconnection@tappi.org; or visit www.tappi.org.

14

TAPPI JOURNAL APRIL 2002

You might also like