DCU Coke Drum Skirt Failure
Delayed Coker Unit equipment was introduced in the 1930's but by the late 1950's,
several problems became sufficiently pressing that the first of several API surveys was
undertaken to understand the damage mechanism and prescribe remedies. The analytical tools were
limited but correlations were introduced; some were quite insightful. The Unit
Quench Factor or UQF was one such measure that has actual foundation in the physics of coke
drum damage and reliability. Unfortunately, conclusions were misinterpreted and motivated designers
to use increasingly higher strength materials such as SA 387 22 (2¼ Cr - 1Mo) and
even SA 387 Grade 21 (3Cr - 1Mo).
Equally perplexing is the misunderstanding by vessel designers of basic engineering
principles impacting drum component reliability and, specifically for this discussion, skirt cracking.
Classic Coke Drum Skirt Cracking Failure
Photo 1 provides illustration of the typical OD cracking location observed in coke drums.
In "recent times", "improvements" were implemented for the purpose of increasing
reliability but, instead, have led to premature failure. Drum designs previously demonstrating 50 years of
trouble-free skirt weld integrity currently appear to experience cone-side cracking in less than 10 years.
Also, surprisingly, cracking is initiating at the toe of the skirt cone side ID joint weld;
these are newer drums with purported enhanced skirt-to-drum dimensional and weld details.
Recent Coke Drum Skirt / Cone Cracking Failure
In Figure 1, an FEA was completed on a drum which had experienced "pre-mature" cracking; this
cracking was extensive and entered into the cone from the toe of the 2" radius skirt ID weld.
The generous contour and more robust skirt dimensions had been intended to reduce discontinuity
stresses but inadvertently caused substantially increased stress & strain concentrations by virtue of the
disproportional material distribution. In effect, the susceptible crack initiation relocated to the cone.
The strain at the toe of the crotch is approximately 7,600 µstrain;
the consequential predicted fatigue life is some 1,550 operational cycles or only some 5 years of operation.
This nominally matched actual operation; the joint, in essence, was inadvertently designed to fail.
|
Photo 1 Classic OD Crack at Skirt to Shell Joint in Repair
|
|
Figure 1 Relocated Crack Initiation Point
|
|
Proposed Remedy for Damaged Skirt-to-Drum Joint
Consequently, short term and long term reliability and safety concerns were raised. The decision was made to
replace the drums for a design incorporating a completely new drum-to-skirt detail. As an interim mitigation,
a decision was made to add multiple longitudinal brackets to the upper portion of the skirt
to provide an independent support path should the cone or skirt weld crack through prior to a turnaround scheduled 4 years out.
An FEA of the proposed "fix" using these brackets spanning the skirt attachment point indicates
that the high strain location on the cone incurs further strain
concentration since the brackets further stiffen the joint. This is a "rookie" design mistake when
designing for displacement controlled loads. In this specific case, the strain increases to approximately
12,800 µstrain; and the fatigue life is reduced to some 1¼ years. Activation of specific fracture mechanics
processes, easily determined, may accelerate crack propagation. Hence, the proposed "fix" exacerbates the situation to the extent
an uplanned shutdown is likely well prior to the planned turnaround.
This uncertainty prompts a risk assessment explicitly addressing personnel safety for the contingent event of uncontrolled
release of vessel contents.
As a point of interest, modelling of the "classic" detail indicates an apparent crack initiation life at some 20 years by
pushing the location of high strain into wrought skirt material. Avoiding weld residual stresses and taking advantage of
fracture mechanics mechanisms contribute to substantially reducing the velocity of crack propagation .
The major advantage, as demonstrated by successful operation of older units, is that cracking associated with joint discontinuity
is precluded from occurring in the cone or progressing from skirt into the shell. Astute efforts during the design stage
can avoid selecting "fatigue incompatible" coker skirt designs. Designers, thus, need to also be alert to specific deficient
design guidance in some industry standard practice documents such as API TR 934 G.
A followup screening analysis of several industry recommended drum-to-skirt joint designs suggests that there are significant
differences in performance. In extreme situations, owners have mistakenly replaced drums in the belief that prematurely
failed skirt joints were an indicator of balance-of-drum integrity.
The details for implementing design and design check methodologies for DCU coke drums are covered in detail in our course
John Aumuller, P. Eng., Ph. D.
|
|
|