Dominic McCann, David White
Sedco Forex
Paris
Gerard Rodt
Sedco Forex
Bombay
A "smart" alarm system installed on a number of offshore rigs and one land rig can detect kicks more quickly than conventional systems (Table 1).
This rapid kick detection improves rig safety because the smaller the detected influx the easier it is to control the well. The extensive computerized monitoring system helps drilling personnel detect fluid influxes and fluid losses before the changes in flow would normally be apparent.
KICK DETECTION
Kick detection systems on many rigs monitor both the active system volume and the paddle flow-out sensor and display the information as a digital readout, The digital format makes identification of trends in the data difficult. Another difficulty in interpretation arises from the display of flow-out information as a percentage of full scale and not as a calibrated value.
The limit alarms for tank levels on many conventional systems allow the driller to select only minimum and maximum threshold values. If a parameter value crosses the threshold, a warning (usually an audible alarm) alerts the driller. However, the driller can be subjected to a series of alarms at each connection or change in drilling parameters. Excessive alarms can cause the driller to lose confidence in the alarm system. Many drillers reduce the frequency of alarms by setting the threshold limits very wide, but the alarms may not trigger quickly enough.
Conventional systems for detecting an influx or loss while tripping are often very basic, such as a float and pulley system with a sliding indicator on a calibrated stick situated on the drill floor. This type of indicator may become stuck and has poor resolution. A more complex method displays the float output on a low-resolution gauge on the driller's panel. The driller has to calculate volume balances manually, typically every five to ten stands.
Because of some of the problems with conventional kick detection systems, Sedco Forex developed the automated MDS system to help drilling personnel detect a kick or loss as early as possible. This computer System monitors automatically and displays data in the best scale to detect the event. The system uses intelligent alarms to differentiate between a noisy signal and a real trend resulting from an influx or loss. 2 3
The intelligent alarm package also detects washouts and excessive well bore friction. The MDS system components include sensors, an evaluation workstation, and monitor displays. The driller, rig superintendent, and operator representative have access to six real-time color monitor displays, all referred to as the driller's display.
SENSORS
The active system volume is obtained by accurately monitoring the volume in all mud tanks on the rig. The driller uses an interactive menu to select the tanks that are part of the active system at any time. The individual tank volumes are then summed once per second to update the total active volume.
The sensors are accurate to a change in level of 20 mm, which equates to about 2.5 bbl in the larger tanks and about 0.5 bbl in the smaller tanks. The smaller tanks generally indicate the influx first because they are the first tanks downstream from the bell nipple. A typical trip tank is measured with an accuracy of about 0.25 bbl.
Flow in is computed from the pump stroke rate, which is monitored by proximity sensors on the pump gearing. System accuracy is not affected by changes in pump efficiency, which are slow relative to the time scale of a kick.
The alarm system has been tested using flow-out measurements obtained by automatically calibrating the paddle sensor and by using flow-out data from the Accu-flow sensor developed by Anadrill Schlumberger.5 Although the paddle does not give an accurate, absolute value of flow out, it has good resolution which allows it to detect changes or trends in the signal.
Bit position and string lengths are computed from a multisensor system that includes encoders at the crown block.
A fast data rate monitors use of the slips to obtain an accurate measurement of any drillstring element added or removed. 6
For ease of interpretation, data are displayed in log format, in bar charts, or in digital readouts. The driller uses a rugged keyboard to select the parameters displayed and to set the scales, limit alarms, and sensitivity of the intelligent alarms. During drilling operations, active system volume can be displayed in log format against time for clear identification of anomalies, and flow in and flow out can be displayed as colored tracks on the same log to emphasize any differences.
FLOW MONITORING
During drilling and circulating operations, the difference between flow in and flow out, which is compensated for rr heave, is computed once per second, and a statistical analysis is performed to detect trends indicating a kick or a loss. 7 This analysis assumes that the difference is zero under normal operating conditions. The Hinkley method tests this hypothesis.
The system tries to detect a statistically significant trend in the data; thus, it does not rely solely on absolute values and threshold crossing. This allows an alarm to be raised reliably at lower deviations than with a threshold or limit alarm because spikes or noise in the data are filtered (Fig. 1).
Active tank system volume is continuously monitored on a real-time log. The total volume is compared to expected volume computed from hole length and diameter, which is monitored and updated as drilling progresses. These data are statistically analyzed for trends.
When the pumps are shut down to make a connection or the pump rate is chanced for some other purpose, there is a transient change in active system volume while the surface flow lines drain or fill.
To continue detection during such a change, the system compares the actual instantaneous volume with the volume predicted by a model based on parameters which describe the flow system (Fig. 2).
In this example of actual connections, the model predictions closely follow the actual tank volume. If trends in the measured volume exceed the model prediction, an alarm warns of the influx. The system returns to steady state monitoring after the transient stabilizes.
Ideally, there will be no transfer of mud to or from the active system during drilling operations. Any change in the active system mud volume may result in a transient which changes the maximum volume. This situation must be taken into consideration for subsequent calculations of pit volume.
One method is to include the tank from which the transfer is made as part of the active system. However, the larger the volume or surface area on which the system must detect changes, the less sensitive the detection will be. A second method is to use the system's interactive reset when a mud transfer occurs-this prevents an alarm during the transfer of mud. After the monitor resumes normal operation, the system parameters will automatically reset to the new measured total volume and model parameters.
TRIPPING
During tripping operations, the system automatically monitors the tripped drillstring length and the volume change in the trip tank. These data and the string linear volume are then processed to give early warning of influx or loss.
The driller's display shows two volume balance curves on the same track in real time. One curve indicates the difference between the estimated string linear volume and the expected value, summed on a standby-stand basis. The other change shows the cumulative difference over the entire trip. This amount is the difference between the volume of tools and equipment removed from the hole and the volume change in the trip tank.
If the system detects consistent deviations of the measured string linear volume from the expected value, it sends an alarm.
An important function of the system is to give a good estimate of any influx or loss detected for possible later use in well control. The system allows the driller to update the expected string linear volume with the true measured value.
This can be a critical factor because experience has shown that the difference between the two can reach 8% as a result of wear or other factors.
If the string remains idle for more than 3 min and no mud is being transferred (for example, during a singleshot survey), the system monitors trip tank volume and sends an alarm for any consistent change in level. When tripping resumes, the system automatically switches back to monitoring string linear volume.
TEST TRIALS
Commissioning tests were performed on the Sedco 600 semisubmersible in conjunction with Total Indonesia. These tests, which took place on a well just after the surface casing had been set, used the cementing pump to simulate an influx by pumping mud from an isolated tank down the choke line into the well. Since the cementing unit was not assigned as one of the working pumps in the system, its flow rate was now added to the flow in.
To simulate drilling into an overpressured zone or gas expansion in the annulus, the cement pump rate was slowly ramped from 13 gpm to 66 gpm over a period of about 15 min. The simulated kick started at about 18:14, and the slowly increasing flow-out signal can be seen clearly (Fig. 3). There was also indication of a small increase in the active volume at this stage, shown on an expanded scale of only 12.5 bbl.
The delta flow influx alarm sounded at approximately 18:24, when the difference in flow was about 26 gpm and the total influx volume was approximately 3 bbl. About 4 min later an alarm was raised on the active system volume when the total influx volume was about 5 bbl. At this time both the delta flow influx alarm and the active tank influx alarm were active and were flashed alternately on the driller's display.
To test the model on pit level transients during connections, mud was circulated at a rate comparable to that used while drilling. The flow rate dropped to zero for a few minutes to simulate a connection, and then it ramped back up to simulate resumed drilling. This procedure was repeated several times to allow the tank monitoring software to "learn" the characteristics of the transients on the active system volume. During the fourth simulated connection, an influx was introduced into the system by increasing the cement pump rate from 0 to 132 gpm in about 5 min.
The tank volume followed the model closely during the first three connections. However, when the measured tank volume produced a trend that increased over the modeled transient during the fourth connection, the driller's display signaled an alarm. The delta flow influx alarm detected the flow-out reading increase with no corresponding flow in. Total influx volume was about 6.3 bbl when the alarms were triggered.
INFLUX WHILE DRILLING
On a floating rig, the heaving motion causes volume changes in the riser and surges in the flow out along the order of several hundred gallons per minute, masking any changes from an influx. To compensate for these fluctuations, the MDS system monitors the riser motion and the slip joint.
A model corrects for the heave and significantly improves the detection sensitivity to around 50 gpm.7 For pit levels, the average is taken from several sensors to eliminate errors because of mud level changes across the tank resulting from rig pitch and roll. Early detection of an influx on a semisubmersible is important to prevent large amounts of gas from entering the riser.
Fig. 4 shows a field example of a kick detected by the MDS system on a floating rig in the North Sea. These are part of the logs seen by the driller on his display and later reproduced on the system workstation. The well was being drilled at 12,580 ft true vertical depth with a 95/8-in. casing shoe at 12,450 ft. While drilling ahead, an influx alarm was raised on the system after only a 4-bbl gain, and the Well was then checked for flow.
The check was positive and the well was shut in with a total gain of less than 8 bbl. The flow out signal in Fig. 4 increased at 10:05 as the influx crew but remained below its threshold of 25 gpm. This small influx allowed straight-forward well control operations with minimal time loss. The quick detection reduced the risk of differential sticking.
An influx detected on the active system volume during drilling operations on another Sedco Forex rig occurred during bottoms-up circulation after a drill stem test.
The control plan was to circulate 5,560 strokes, but the alarm was raised after 3,998 strokes. Circulation started at about 15:00 and flow in and flow out remained essentially constant. However, at approximately 15:14, active volume increased slowly, and an alarm was raised at 15:37 after an increase of about 4 bbl. Three minutes after the alarm was raised, the well was shut in and circulation was continued through the choke.
This illustrates the importance of monitoring both active system volume and the delta how. Very slow influxes that may not be seen on the delta flow can be caught on the active volume, and rapid influxes are seen more quickly on the delta flow than on the active volume." This is true for both oil based and water-based muds.10
On another rig an influx was detected on the delta flow while drilling out a casing shoe (Fig. 5). During the initial stages, flow in and flow out were very closely matched. At about 15:01, flow out started to increase slowly. It peaked at roughly 21 gpm above flow in after approximately 3 min, which corresponded to the bottoms-up time. The total gas reading on the driller's display showed that the increase in flow was from small amounts of gas.
Flow out then dropped gradually but never returned to its original value. At 15:09, flow out increased very rapidly, a delta flow influx alarm sounded, and the driller reacted quickly to shut in the Well. The delta flow alarm was raised before there was a noticeable increase in active system volume-the increase Was less than 1 bbl. The Well was shut in with 1,970 psi on the casing and a total influx volume of about 9.4 bbl. The quick detection minimized the volume of influx into the well, thereby making the kill procedure much easier.
REFERENCES
- McCann, D.P., White, D.B., Marais, L., and Rodt, G.M., "Improved Rig Safety by Rapid and Automated Kick Detection," SPE 21995, presented at the 1991 SPE/IADC Drilling Conference, Amsterdam.
- Chevallier, J., Quetier, F., and Marshall, D., "Technical Drilling Data Acquisition and Processing With an Integrated Computer System," SPE/IADC 13494, presented at the 1988; SPE/IADC Drilling Conference, New Orleans.
- Peltier, B., "Computer Monitoring of Surface Parameters while Tripping," SPE/IADC 16056, presented at the 198- SPE/IADC Drilling Conference, New Orleans.
- McCann, D.P., Arnold, M., Burgess, T.M., and Curwen, S.C. "Early Automatic Detection of Drillstring Washouts Reduces the Number of Fishing Jobs," SPE/IADC 19933, presented at the 1990 SPE/IADC Drilling Conference, Houston.
- Orban, J.J., and Zanker, K.J., "Accurate Flow-Out Measurements for Kick Detection, Actual Response to Controlled Gas Influxes," SPE/IADC 17229, presented at the 1988 SPE/IADC Drilling Conference, Dallas.
- Unsworth, M.I., Burgess, T.M., and Kerbart, Y., "How an improved Measurement and Smart Processing Can Help the Driller Improve Efficiency," SPE/IADC 19965, presented at the 1990 SPE/IADC Drilling Conference, Houston.
- Jardine, S.I., McCann, D.P., White, D.B., and Blake, A.J., "An Improved Kick Detection System for Floating Rigs," SPE 23133, presented at the 1991 Offshore Europe Conference, Aberdeen.
- Hinkley, D.Y., "Inference About the Change-Point From Cumulative Sum-Tests," Biometrika 42, No. 6, pp. 1897-1908., 1971.
- Speers, J.M., and Gehrig, G.F., "Delta Flow: An Accurate, Reliable System for Detecting Kicks and Loss of Circulation During Drilling," SPE/IADC paper No. 13496, presented at the 1985 SPE/IADC Drilling Conference, New Orleans.
- White, D.B., and Walton, I.C., "A Computer Model for Kicks in Water and Oil-Based muds," SPE 19975, presented at the 1990 SPE Annual Meeting.
Copyright 1992 Oil & Gas Journal. All Rights Reserved.
