Reliability of Coke Drums

This is Part II of a two-part series on the reliability of coke drums. Part I discussed some of the causes of bulging and cracking in coke drums. Here, the effect of operation on damage will be covered, along with possible solutions for increasing drum life and decreasing the probability of an unscheduled shutdown.

Studies show that damage rates are often related to the operation. Many operators shorten their coking cycle time to increase productivity. Shorter cycle times increase productivity but operation and inspection records of coke drums show that shorter cycle times often shorten the drum life.

There are drums with over 30 years of service that have no major damage while there are others with less than 10 years of service that have bulged and cracked. These drums are essentially identical in design and construction; however, they are operated significantly different- also indicating that damage is related to operation.

Low Cycle Fatigue

Low cycle fatigue (LCF) due to cyclic thermal stresses has been identified as the main damage mechanism in coke drums. Fatigue can be divided into two phases, crack initiation and crack growth. In LCF, more than 90% of a component's life is consumed during the crack growth phase. While not appropriate for making accurate absolute predictions of LCF, life the Paris crack growth equation can be used to relate cyclic stress or strain to crack growth rates and to compare the effect of changing strain levels on life.

Field measurements combined with fatigue analysis show that a shorter cycle time often results in a shorter fatigue life. In most cases, this reduced fatigue life is not only caused by the increased number of stress cycles per year but also by the higher strains that usually result from the accelerated heating and cooling needed to achieve the shorter cycle time.

Strain Measurement

Strains and operating variables such as; process temperature, process flow, and drum pressure have been measured in several drums. These measurements have confirmed that damage is related to operation. In addition, they have shown that the increase in damage rate due to a shorter cycle is often more affected by the higher cyclic strains than by the higher cycling rate.

Figure 1 shows measured cyclic skirt strain as a function of vapor pre-heat time in a typical drum. The cyclic strain or damage decreases with a longer pre-heat duration. Statistical tests show that this effect is significant. Similar trends have been observed in other drums; however, the slope of the trend line shown in Figure 1 varies from drum to drum.