September 2016

Maintenance and Reliability

Identify and correct coke drum foundation bolt failures

Coke drums in many operating plants experience foundation bolt failures.

Tharakan, J., Suncor Energy Inc.

Coke drums in many operating plants experience foundation bolt failures. Industry does not have a consistent approach for the design of foundation bolts for coke drums. Older drums with diameters of up to 30 ft can be found with 1-in. ø × 24 bolts made from ASTM A307 Grade B steel. A recent trend is to design coke drums of 30-ft diameter with as many as 48 anchor bolts of 2 in. ø.

Fig. 1. Anchor chair and bolt deformation for Canadian site.
Fig. 1. Anchor chair and bolt deformation for Canadian site.

Some drums experience periodic vibrations that worsen if the anchor bolts become loose or broken. A review is provided here of coke drum bolt failures, their consequences and potential corrections.

Survey on coke drums

Data on drum foundation bolt failures and vibration were gathered from several North American and Asian facilities with coker units. Most sites had experienced bolt failures in the shear plane. At some sites, both necking and shear were observed on broken bolts.

Failure in a US refinery. The coker unit has two 26-ft diameter drums with 44 anchor bolts of 2 in. ø, conforming to ASTM A307 Grade B. The coke drums have been in operation since 1984. After 25 years of operation, one bolt was found to be broken. Several other bolts were observed to be loose.

Fig. 2. Bending, elongation and necking of the bolts.
Fig. 2. Bending, elongation and necking of the bolts.

Violent shaking of the drum structure and piping were reported during the quenching cycle, especially with the introduction of “big water”—i.e., a large flow of water following initial quenching with steam and a small volume of water. The failed bolt displayed quasi-cleavage cracking at the shear plane. The conclusion was that the failure occurred due to sudden loading.

Failure in a Middle Eastern refinery. The coker unit has four coke drums of 26-ft diameter with 24 anchor bolts of 1 in. ø, conforming to ASTM A307 Grade B. The anchor bolts were under fireproofing. After decades of operation, the drums were found in 1999 to be laterally shifted from the centerline to a maximum extent of 3.25 in.

Fig. 3. Smearing and crushing damage on the threaded ends of the bolts.
Fig. 3. Smearing and crushing damage on the threaded ends of the bolts.

The fireproofing around the foundation bolt was removed, revealing that all of the bolts on the two drums had broken. Corrosion and necking were observed on the bolts, but all of the failures were found at the shear plane. The foundation bolts on the other two drums showed preliminary stages of failure, such as corrosion and necking. These drums did not have any noticeable periodic vibration.

Failure in a Canadian facility. This failure occurred on four identical 30-ft diameter coke drums with 24 anchor bolts of 1 in. ø, conforming to ASTM A307 Grade B. The failures were noticed after eight years of operation. The failure rate increased with time, presumably due to the following reasons:

  • Two more drums were added to the system, and cycle times were shortened from 16 hr to 11 hr. Aggressive operations increased drum movements.
  • As the bolts broke, the remaining bolts experienced larger loads due to drum movements.
Fig. 4. Lateral shift of the drum and bolt failure on the shear plane.
Fig. 4. Lateral shift of the drum and bolt failure on the shear plane.

The bolts were hand-tightened only during installation. Clear evidence was seen of tensile loading of the bolts, as displayed by the necking of bolts and the bending of the compression plate (Figs. 1–3). All bolts failed in the shear plane (Fig. 4).

Drums 1 and 2 had more bolt failures than other drums (Table 1). Most bolts exhibited a combination of tensile and shear mode of failure. Bolts were bent in random directions, exhibiting no correlation with the direction of maximum vibration. The top portions of the bolts, embedded in concrete, were bent by lateral forces. Potential causes of failure included:

  • Thermal expansion of the skirt ring. If the skirt expands due to thermal expansion beyond the diametrical clearance of 0.5 in. between the bolt and the bolt holes, then the bolt will shear. The skirt was 15 ft tall; therefore, the skirt flange was always at ambient temperature, so thermal expansion of the skirt ring is not a credible cause.
  • Banana effect. The drums developed a permanent bowing of their axes during initial periods of operation with single-side inlets (i.e., the “banana effect”). Some believed that the banana effect could cause non-uniform and excessive bolt loads. Detailed analyses of the failures showed no correlation with the banana effect.
  • Vibration of the coke drum. The coke drum experienced periodic vibration in the direction perpendicular to inlets. (Note: At the time of investigation, the coke drum had dual inlets). Detailed vibration measurements included:
    • Vibration was maximized at the top of the drum in the horizontal direction, and perpendicular to the orientation of the feed inlet nozzles. The vibration peaked toward the end of the filling cycle and measured 500 mils–750 mils peak-to-peak at 1 Hz–2 Hz.
    • Using a transducer, the pressure pulses inside the drum were measured during the filling cycle. The amplitude was very small, but its frequency matched with the vibration at the top of the drum.
    • The skirt flange displayed a rocking motion of 0.08-in. peak-to-peak at a frequency of 0.3 Hz.
Fig. 5. Skirt flange vibration displacement peak-to-peak, in mils.
Fig. 5. Skirt flange vibration displacement peak-to-peak, in mils.

The deformation of the bolt and anchor chairs was approximately five times the displacement amplitude of periodic vibration at the skirt. No indication of fatigue was observed on the failed bolts. Therefore, the bolt failure was not caused by periodic vibration (Fig. 5).

  • Resonance could amplify the periodic vibration. High-amplitude periodic vibration at the skirt was not noticed. The first three natural frequencies of the full drum were 0.62 Hz, 3.88 Hz and 10.87 Hz, respectively. None of these frequencies matched the measured frequency of periodic vibration on the drum top or skirt; therefore, resonance cannot be the cause of bolt failures.
  • Thermal distortion of the shell and cone due to thermal gradient can be transmitted to the skirt flange. The skirt should remain tilted for a noticeable period if it is a static loading condition. Such observations were absent, ruling out static load due to distortion.
  • Dynamic events, such as random movements of high amplitude, sometimes occur in a coke drum. These occasional events may rock the drum, causing lateral movement:
    • Flexing of the drum due to rapid thermal gradient occurring in the shell and cone during quenching
    • An avalanche-like event caused by a large volume of coke collapsing to the bottom cone during coke cutting
    • Flashing of water into steam within the drum.
Fig. 6. Tilting of drum.
Fig. 6. Tilting of drum.

It is possible to capture occasional dynamic events only if drum movements are continuously recorded for a reasonably long period of time. Although this could not be carried out, most of the observations on the failed bolt and anchor chair lined up with the potential consequences of a dynamic drum movement.

Impulsive energy due to dynamic events. The deformation observed at the anchor chair and the bolts can be written as X = 0.4 in. (Fig. 1). Working backward from this displacement of 0.4 in., the impact energy can be calculated using the following definitions:

  • Bolt circle diameter, D = 374 in. = 31.1 ft
  • Bolt circle radius, r = 15.55 ft
  • Height of the center of gravity of the drum from foundation, H = 60 ft
  • The tilt angle θ (refer to Fig. 6) = (X ÷ D) = 0.4 ÷ 374 = 0.00107 radian
  • Tilting radius R = √(r2 + H2) = 62 ft
  • Angle α = Tan–1 (r ÷ H) = 14.53° = 0.252 radian.

The moment arm varies with tilt angle, as shown in Table 2.

The potential energy when the rotation of the drum reaches its peak, plus the work done to stretch the bolts, is the impulse energy, as explained in the following definitions:

  • Potential energy = Moment × tilt angle
  • Tilt angle θ for 0.4-in. bolt deflection from Table 2 = 0.00107 radians
  • Operating weight of the full drum, W = 6E + 06 lb
  • Moment arm = z (for a bolt stretch of 0.4 in.) = 15.5 ft (approximate)
  • Moment, M1 = Wz = 6E + 06 × 15.5 ft = 9.3E + 07 ft-lb
  • Potential energy W1 = M1 × θ = 99,510 ft-lb.

Work done in stretching the bolts can defined by the following:

  • Root area of 1-in. bolt, A = 0.606 in.2
  • Number of bolts, N = 24
  • Grip length of the bolt, L = 4 in.
  • Yield strength of bolt, y = 36,000 psi
  • Modulus of elasticity, E = 28E + 06 psi
  • Elastic extension of the bolt = L × y ÷ E = 4 × 36,000 ÷ 28E + 06 = 0.005 in.

Therefore, a 0.4-in. deflection (θ = 0.00107) will cause the bolt to yield. Assume that the average resistance of the bolt during elastic-plastic stretch is equal to the yield strength:

  • Bolting force, F = y × A = 36,000 × 0.606 = 21,816 lb
  • Moment that stretches the bolt, M2 = D × N × F ÷ 4
  • M2 = 31.1 × 24 × 21,816 ÷ 4 = 4.07E + 06 ft-lb
  • Work done for stretching bolts W2 = M2 × θ = 4,355 ft-lb
  • Energy of impact (total work done) = W1 + W2 = 99,510 + 4,355 = 103,865 ft-lb.

Are the bolts capable of absorbing the impact energy? Consider 2.5-in. ø × 24 bolts of high strength (y = 100,000 psi) with a grip length of 12 in.:

  • Tensile area of the bolt = 4.4 in.2
  • Elastic extension of bolt, X = (y/E) grip length = 0.0432 in.
  • Tilt angle θ from Table 2 = 0.00012 radian
  • Moment arm from Table 2 = 15.51 ft
  • Restoring moment, M1 = 15.51 × 6E + 06 = 9.3E + 07 ft-lb
  • Bolting force, F = y × A = 100,000 × 4.4 = 4.4E + 05 lb
  • Moment, M2 = D × N × F ÷ 4 = 31.1 × 24 × 4.4E + 05 ÷ 4 = 8.21E + 07 ft-lb
  • Total opposing moment = M1 + M2 = 9.3E + 07 + 8.21E + 07 = 1.75E + 08 ft-lb
  • Work done = 1.75E + 0.08 × θ = 1.75E + 08 × 0.00012 = 21,000 ft-lb; 21,000 ft-lb < 103,865 ft-lb.

The work done is much smaller than the energy of the impact. Therefore, the bolts will be stretched beyond yield. The situation can be improved slightly by increasing the grip length of the bolts, but the required grip length to retain elastic stretch sufficient to absorb the energy of the impact is excessive.

Solution with Belleville washers

A design having 24 × 1.5-in. ø bolts with 14 Belleville washers stacked in series, at each bolt, will allow a cumulative extension that is equal to the number of washers multiplied by the deflection of each. The maximum deflection of each Belleville washer is 0.076 in. at a full flattening load of 32,000 lb. The Belleville washers will be preloaded to 16,000 lb by deflecting half the load during assembly:

  • Remaining deflection for the Belleville washers = 0.076 × 14 ÷ 2 = 0.532 in.
  • From Table 2, corresponding tilt angle = 0.00142 radian
  • Moment arm from Table 2 = 15.43 ft
  • Restoring moment due to gravity, M1 = 6E + 06 × 15.43 ft = 9.26E + 07 ft-lb
  • Average resistance of Belleville washers, F = (16,000 + 32,000) ÷ 2 = 24,000 lb
  • Moment due to Belleville washers, M2 = D × N × F ÷ 4 = 31.1 × 24 × 24,000 ÷ 4 = 4.48E + 06 ft-lb
  • Total moment, M1 + M2 = 9.7E + 07 ft-lb
  • Work done = Moment × θ = 9.7E + 07 × 0.00142 = 137,740 ft-lb.

This result is greater than the energy of impact (103,865 ft-lb). Therefore, using Belleville washers for absorbing the energy of impact is a feasible solution to the problem. The bolt load will remain between 16,000 lb and 32,000 lb. The design can accommodate a deflection of 0.532 in., which is greater than the maximum observed deflection of 0.4 in.:

  • Root area of 1.5-in. ø bolts = 1.405 in.2
  • Yield strength of the bolts proposed = 100,000 psi
  • Tensile strength of one bolt = 1.405 × 100,000 = 140,500 lb
  • Stress ratio in tension = 32,000 lb ÷ 140,500 = 0.23.

Shear load on the bolt

Fig. 7. Rocking velocity.
Fig. 7. Rocking velocity.

The use of Belleville washers allows the energy of the impact to be absorbed by drum tilting within limits. An unwanted outcome of this motion is a tendency for lateral skidding on the foundation.

The design includes a 2.5-in. ø × 6-in. sleeve at each bolt to increase the shear area to compensate for the uncertainty on the number of bolts sharing the shear load in the tilted condition. The standard design calculation assumes that 50% of the bolts share the shear load.

The rocking drum lands on the foundation and may bounce back. This calculation does not assume bounceback. The inertia of the rotating mass causes a horizontal force in the rocking direction prior to touching the foundation. Assume that rocking causes the center of the gravity of the drum to lift by h = 0.4 (Fig. 7).

When the drum lands on the foundation:

  • The vertical velocity Vv = √2gh = √2 × 384 × 0.4 = 17.5 in./sec
  • Vv = Vcos(90 – α) = Vsinα
  • Tangential velocity, V = Vv/sinα
  • Horizontal velocity, Vh = Vsin(90 – α) = Vv/tanα = 17.5/tan(0.253) = 17.5 ÷ 0.259 = 67.5 in./sec

Assume that Vh changes from 67.5 in/sec to 0 in 1 sec:

  • Acceleration, a = (67.5 – 0) ÷ 1 = 67.5 in./sec2
  • Weight of the vessel with contents, W = 6E + 06 lb
  • Acceleration due to gravity, g = 384 in./sec2
  • Inertia force in horizontal direction = W × a/g = 0.175W = 6E + 06 × 0.175 = 1.05E + 06 lb-force

The horizontal force is calculated similar to seismic design, with a seismic factor of 0.175 for this example:

  • Shear strength of the sleeve/bolt = 50,000 psi
  • Shear area of bolting assembly = 0.785 × 2.52 × 24 ÷ 2 = 58.9 in.2
  • Shear stress induced in bolting assembly = 1.05E + 06 ÷ 58.9 = 17,826 psi
  • Stress ratio of bolt in shear = 17,826 ÷ 50,000 = 0.35.

Combined mode of failure of the bolt

Tension and shear modes complement each other. The summation of stress ratio for tension and stress ratio for shear should be less than 1; i.e., the combined stress ratio = 0.23 + 0.35 = 0.58.

The bearing stress in the grout encasing the bolts can be described as follows:

  • Horizontal force = 1.05E + 06 lb-force
  • Bearing area of bolts = 2.5 × 6 × 24 ÷ 2 = 180 in.2
  • Allowable bearing stress of concrete, fc = 5,000 psi
  • Allowable bearing stress of grout surrounding bolt = 7,000 psi
  • Calculated bearing stress = 1.05E + 06 ÷ 180 = 5,833 psi
  • Calculated bearing stress of 5,833 psi < allowable bearing stress of the grout.
Fig. 8. Modified anchoring system.
Fig. 8. Modified anchoring system.

Modified design details

The modified design of the coke drum anchor bolt system used at the Canadian facility in anchor bolt restoration work is shown in Figs. 8 and 9. The new bolts were installed between the existing bolts, which were either broken or ineffective. The reanchoring work was completed while the drums were in operation. The basic features of the anchoring included the following:

  • 24 × 1.5-in. ø bolts
  • 2.5-in. ø × 6-in. long sleeve
  • 14 Belleville washers stacked in series on each bolt.

Takeaway

Fig. 9. New anchor bolt installation details.
Fig. 9. New anchor bolt installation details.

Several noticeable improvements were observed after the installation of the new anchor bolts: The periodic rocking motion of the drums stopped; the amplitude of horizontal vibration at the top of the drum was reduced by half; and no bolt failures have occurred in the last four years of operation after replacement.

Based on this work, the author recommends:

  • If the coke drum is undergoing periodic vibration, then the bolts cannot be left loose. Bolt preload is essential to lower the equipment vibration.
  • The energy released during significant impulse-type events that cause coke drums to rock and move cannot be absorbed by conventional bolting designs. Belleville washers allow drums that experience significant impulse loads to move and dissipate energy without damaging the anchoring.
  • The use of a sleeve increases the effective shear area of the bolt and bearing area in concrete. HP   

The Author

Related Articles

From the Archive

Comments

Comments

{{ error }}
{{ comment.comment.Name }} • {{ comment.timeAgo }}
{{ comment.comment.Text }}