UNDERSTANDING MWD DATA ACQUISITION CAN IMPROVE LOG INTERPRETATION

Richard A. Fagin
Atlas Wireline Services
Houston

By understanding how measurement while drilling (MWD) tools acquire data and how the data are processed, engineers and geologists can better interpret MWD logs.

Wire line and MWD log data sometimes do not precisely match. If a discrepancy occurs between MWD and wire line logs run across the same interval, many log interpreters will condemn the MWD data.

Some of the differences are related to data acquisition and processing. When these anomalies occur, neither of the logs is necessarily in error. Disregarding one log or the other may not be prudent. MWD and wire line logs can differ, yet both can be correct.

Recognizing the differences and the correct data requires a better understanding of the MWD tool operational principles. Because MWD logs are becoming more widely accepted as quantitative replacements for equivalent wire line logs, the differences between logs should be analyzed logically.

Most wire line data are sampled on uniform, depth spaced intervals. Data processing and filtering match tool measurement resolution to the sample interval. Most commercial log presentation and interpretation software can only use this type of data and no unprocessed MWD data, mainly cause wire line data are lent.

Wire line log data are acquired over a very short time relative to the drilling time, and mud properties are generally constant for the duration of the logging run. Depths are measured under relatively constant cable tension with precise, purpose designed, depth encoding systems.

MWD data are sampled in time intervals, not in depth intervals, and therefore have to be converted to depth based output. MWD logs are sampled on a time basis because of the lack of direct connection between the MWD instrument and the recording system. Because mud pulse telemetry systems are relatively slow, the data received during drilling frequently have gaps or "holes."

Mud properties often vary as the well is drilled, further complicating the measurements. Each MWD logging sensor has a different relative formation exposure rime, and the time differences can be large, depending on the distance between the sensor and bit. MWD depths are measured as a function of drill pipe tally plus kelly height, but the devices used to measure the kelly height are somewhat crude. Furthermore, although some work has been done to correct measured depth for pipe stretch as a function of weight on bit, the corrections generally have not been incorporated into MWD depth systems.

Additionally, MWD logging sensors move at different speed ranges than the equivalent wire line tools, lending some question as to whether wire line presentation filters are appropriate for MWD logs.

These differences in acquisition and processing between MWD and wire line log data sometimes cause discrepancies in log readings within a field or even in a given well. Understanding these differences will help the log interpreters make better evaluations of MM and wire line log data.

TIME BASED ACQUISITION

MWD logging sensors transmit data via mud pulse telemetry or record data in downhole memory chips on a time basis. Mud telemetry data may be received as infrequently as every 25 min. This frequency depends on whether the tool sends directional survey data and on sensor configuration an@ effective telemetry rate.

Even at modest rates of penetration (ROPs) of 30 40 ft/hr, sometimes several feet of formation can be drilled between log samples. Fig. 1 shows one possible result of this low data density. The true response is shown by the top curve in the figure, and the transmitted samples are shown by the vertical bars. The second and third curves in the figure show how the sparsely sampled data could be misinterpreted depending on how the data are interpolated (dotted lines).

Any data in downhole memory are dumped to a computer when the tool is pulled, or, for some MWD tools, the data can be retrieved by downhole wet connect with wire line. Generally, these stored data are much more rapid]N, sampled than mud pulse telemetry data. The log interpreter should consider that the resolution of the final log may be limited by the ROP and the sample rate.

Fig. 2 shows the sample density and ROP for various log sample rates. Resistivity and gamma ray data frequently are sampled at 5 sec intervals. Even at these rates, obtaining four samples per foot requires an ROP less than about 180 ft/hr. (Four samples per foot is quickly becoming the wire line standard.)

Many, MWD vendors frequently extol the vertical resolution of their resistivity tools. For the log interpreter to "see" the 6 in. bed delineation often claimed for MWD resistivity tools by their manufacturers, the logs must have four or more samples per foot to satisfy basic sampling theory (Nyquist theorem).1

The varying sample rates for MWD logs, because of varying ROPs, lead to nonuniformly sampled outputs. For interpretation with standard petrophysical logging software, the data must be resampled into evenly spaced output.

FILTERING

How the data holes are filled is quite important. Fig. 3 shows several methods of resampling and shows how the output differs from the true input waveform. Fig. 3a shows an inadequately sampled block average technique, similar to the ROP curve presented on MWD and mud logs. Fig. 3b shows the same method but with a higher sample rate. Fig. 3c shows the most common methodlinear interpolation (connect the dots). In practice, the final output shown in Fig. 3c would be filtered further to reproduce the input waveform more accurately.

The drilling industry lacks a standard method of filtering resampled data. Wire line data usually are run through a "sinc" function filter to reproduce the analog equivalent waveform.1 Because the MWD sample rate changes, invariant sinc functions may not be correct.

Fig. 4 shows MWD gamma ray data (sampled every 5 sec on a wiper trip of 60 ft/hr, giving 12 samples per foot) compared with wire line data (logging speed of 1,800 ft/hr) from the same hole. The MWD data were resampled by linear interpolation but not filtered. Note on the MWD logs that these are actually two log runs presented one atop the other. The runs repeat almost perfectly.

MWD companies frequently filter out the jaggedness of their data to make the curves appear more like wire line curves. Fig. 4 shows how that filtering may not be accurate in that some necessary detail may be lost.

The repeatability of the small details suggests that they are valid data and not just "noise." Ultimately, the MWD companies should design adaptive filters to match the output resolution to the input data density (sample interval). To make the interpretation better, a data density curve should be presented (in samples per foot) for each unique time sample rate in the MWD tool string. These curves are easy to compute from the raw data samples and the depth between samples.

Engineers and geologists who use MWD log data should ask their MWD representatives for information on the company's particular resampling and filtering process.

The raw data should be recorded so that future technology in resampling and filtering can be applied to current MWD data. Raw data can be converted to log information standard (LIS) or similar format by assigning the nearest level spacing frame depth to the log values recorded in time. Any data holes should be left as is and should be assigned null values (such as 999.25

With 3 in., level spacing output files, the amount of round off in the depth approximation will be much less than the total systematic errors in the depth measuring system. Therefore, the round off in depth approximation can be ignored.

DEPTH CONTROL

The reasons for macroscopic differences in measured depth between wire line and MWD have been well documented in the drilling literature. 2 The more subtle differences lead to stretch and shrink discrepancies in apparent bed thickness shown on wire line and MWD logs. The nuclear sensors in particular are susceptible to this type of error.

Because of statistical variations nuclear log readings, even in a fixed environment, most nuclear logging tools must accumulate event counts for some period of time specific to each nuclear sensor. In wire line tools, this period is usually the amount of time during which the tool moves 1 ft, or about 2 8 sec. Because of the MWD tool's thicker walls and rotational motion, they typically need to accumulate counts for 5 30 sec to make a single measurement.

In high ROP zones, the MWD sensors may move a significant distance during a sample interval, shifting apparent bed boundaries and thicknesses relative to what the wire line shows. If the well bore is sticky, cable stretch could contribute to similar errors for wire fine logs. Wire fine logging companies are making increased use of downhole tool accelerometer data to correct for this type of velocity induced depth error. Fig. 5 shows an example of this kind of depth mismatch. The MWD and wire line logs were run in the same well. The depth matching was based on the gamma ray curves. Note the occasional depth offset in the neutron and density curves.

This type of depth error is normally not troublesome. However, when a log interpreter is trying to match casing collar logs for perforation depth control, picking the open hole tie in curve may be difficult.

If both MWD and wire line resistivity logs are available, then the wire line shallow focused log should be correlated to the MWD resistivity log. If only MWD logs were run in open hole, then a gamma ray/neutron collar log should be used for tie in to the MWD log. The MWD resistivity log should be correlated to the neutron log, with both gamma ray logs as a backup reference.

Cased hole stretch and shrink on the nuclear logs are much less of a problem because of the generally lower friction on the logging tool inside casing; those correlation log bed boundaries usually can be trusted. When available, resistivity data are preferred because they are not statistical, thereby, avoiding bed boundary shifts from event accumulation. The use of wire line log data for depth control in high angle, high bed dip holes should be restricted to the shallow curve to avoid bed boundary displacement because of the apparent dip.

ENVIRONMENTAL VARIABLES

During a wire line log run, most of the important borehole properties, such as mud, hole size, and fluid influx are static and can be considered constant for the duration of the log run.

The temperature can vary in long stretches of open hole, and the effects are well documented in logging texts.

In the MWD environment, all of these properties, and even the bit size, mal, change. Variations in log values are likely to result and quite frequently appear as step like shifts (as bit size changes the borehole correction, for example). The best way to avoid interpretation trouble is to document all changes in drilling conditions both on the log at the depth of occurrence and on the header.

For wire line logs, the formation exposure time since original penetration by the bit is essentially constant during the log run, because the run time represents only a small portion of the drilling time.

With MWD, however, formation exposures can vary from only a few minutes to a day, depending on ROP, bit to sensor distance, and whether a pipe trip was made. To eliminate confusion, MWD logs should have a time since drilled, or formation exposure time, curve recorded for each unique sensor measurement point in the MWD string. This curve can be computed from the time/depth record made during logging.

This point is important because many companies log a formation with the deeper sensors at the end of a bit run yet log with sensors higher up in the drillstring after a bit trip. Most importantly, subsequent passes of a sensor through the same depth interval, even very, short sections, should be recorded to the raw data file as a measured after drilling (MAD) curve. These data from MAD runs should never be mathematically processed with MWD data in the raw data file.

QUALITY CONTROL

The presentation of all of the data suggested in this article would clutter a standard 8.25 in. log. To get the most out of MWD, the industry might have to consider using a 16.75 in., four track detailed log. On this type of log, all of the quality indicators such as data density, ROP, weight on bit, torque, and exposure time can be presented in a useful fashion without compromising the readability of the main log. Fig. 6 is an example of a log using this format.

REFERENCES

Anstey, N.A., Wiggles: A Graphical Introduction to Signal Theory, IHRDC, Boston, 1987, Ch. 8.
Kirkman, M., and Seim, P., "Depth Measurements With Wireline and MWD Logs," oral presentations to Houston chapter of the Society of Professional Well Log Analysts, Apr. 28, 1992, published by Baker Hughes Inc.