A Unified Approach to Drillstem-Failure Prevention (includes associated papers 26205 and 26213 )

Abstract

Summary. Practical drillstem-failure prevention seems to be based moreon individual knowledge of specific failures than on any standard model. To present a broader approach to drillstem-failure prevention, this paperdraws on failure-prediction models, failure-investigation results, drillpipe finite-element analysis, and the API/IADC Drillstring Database toconstruct a unified approach to failure prevention. Introduction Drillstem failures, even such routine failures as drillpipe washouts, can contribute significantly to the cost to drill today's wells. These costs grow exponentially when the failure results in fishing operations, and in extreme cases, failures can even cause well-control problems. In a1985 study, McNalley reported that 45% of deepwell drilling problems were related to drillstem failures. Moyer and Dale concluded that drillstem separations occurred in one in seven wells and cost an average of $106,000 each. For such routine failures as drillpipe washouts, the failure often is accepted as "part of the business." The offending components arereplaced and operations are resumed. If the cause of failure is unusual, analysis sometimes is performed. Results are reported and recommendationsare made to prevent similar failures. These failures seem to be handledcase-by-case, however, without an overall approach to prevention. Thispaper summarizes results from analysis of 76 drillstem failures and drawson data from a recently presented report on the API/IADC Drillstring Database. This paper presents a unified approach to decrease drillstemfailures on the basis of these results. Failure Analysis Results Table 1 summarizes results from analysis of 76 individual drillstem failures that occurred in a variety of drilling conditions in the U.S., Dutch North Sea, east Africa, and Central America. No attempt is made to break down the failures by area or drilling conditions. The 76 failures are simply those that were sent to us for evaluation from 1987 through 1990. The data are broken down by cause and component (Table 1). Because these data are based on failures submitted for analysis, they are weighted toward the more unusual, higher-cost failures. This is evident when comparing the percent of string separations (46 %) in the sample data with the results inthe API/IADC database. The database, which included 1,785 records of drillpipe [not bottomhole assembly (BHA)] failures, recorded that string separations (twistoffs) accounted for more than 5% of all drillpipefailures. While they are not representative of all failures, the 76 failures Table 1reports probably can be considered representative of those submitted for analysis; on the basis of these data several significant observations can be made.Tension and torsion failures combined accounted for only 13% of thetotal failures reported.Fatigue was the primary cause of 50 of 76 failures (65%) andcontributed significantly to 10 other failures. Fatigue directlycaused 43% of the twistoffs.Low material-fracture toughness was the primary cause of six of thetotal failures (8%). Each of these failures had a significant fatiguecomponent as well. When torsion and tension failures were examined individually, in each casewhere loads could be determined, these loads exceeded the values recommended in widely recognized design practices. Therefore, little that is new can be said about prevention of torsion or tension failures on the basis of the cases presented here. Drillstem design practice focuses on tension and torsion, which probably explains why these mechanisms accounted for so small a portion of total failures. Fatigue- and material-related failures are another matter; not only didthey contribute the lion's share (73%) of total failures, but they also accounted for more than 60% of the twistoffs. Unfortunately, predicting fatigue behavior is far more difficult than predicting drillstring behavior under torsion and tension loading. And while API RP 7G offers someguidelines for fatigue in drillstring design, we believe that thecontinuing high proportion of fatigue failures provides evidence thatfatigue prevention procedures still can be improved. Fatigue Mechanism Fatigue damage and failure occur because the drillstem is loaded cyclically during rotation. The process occurs in three stages, as Fig. 1 diagrams. Stage 1-Initiation. Fatigue damage causes a microscopic crack that grows with repeated stress cycles (Points A to B). Stage 2-Growth. At Point B, the crack has enlarged to the point where itacts as its own stress concentrator and continues to grow with each stresscycle (Points B to D). Stage 3-Failure. Final, sudden fracture of the remaining cross sectionoccurs when the crack reaches critical size (Point D). Critical crack sizewill vary with part geometry, loading conditions, and material toughness. Point C represents the smallest crack size detectable by inspection. Fatigue behavior for a given material can be represented by an S/N curve (Fig. 2) or a crack-growth-rate curve (Fig. 3). On the S/N curve, stress amplitude is plotted against total cycles to failure. For steels in a noncorrosive environment, a stress amplitude exists below which fatigue damage will not occur, even after an infinite number of cycles. Thisstress amplitude is called the endurance limit, SL, for that material. Ina corrosive environment, however, no endurance limit may exist, so fatiguedamage can occur even at low stress amplitudes. The fracture-mechanics representation of fatigue-crack growth usually isdisplayed on a curve similar to that shown in Fig. 3. Here, a crack or crack-like flaw is assumed to exist already, and crack-growth rate per cycle, da/dN, is plotted against stress intensity range, . Fatigue cracks can grow in one of three ways, depending on stress intensity range, as Fig. 3 shows. In Region 1 (low stress intensities) little or no crack growth occurs with each cycle. As stress intensity increases, the crack behavior enters Region 2, where it grows in a stable manner. As stress intensity approaches the critical level, crack growth becomesunstable, and failure by rapid fracture is expected (Region 3). Becausethe value of stress intensity includes the effect of geometry (e.g., anelliptical fatigue crack on a drillpipe tube) the concept of a criticalcrack size presents itself. That is, for a given loading, geometry, andmaterial fracture toughness, a critical crack size exists above which rapidcatastrophic failure can be expected by either brittle fracture or grossplastic deformation. SPEDE P. 254⁁