CRAIN'S
PETROPHYSICAL
HANDBOOK
c. 1978 - 2008 E. R. (Ross) Crain, P.Eng.
Rocky Mountain House, Alberta Canada T4T 2A2
403-845-2527 email us
Please be fair to the author - pay your Shareware Fee HERE
and receive a copy of Crain’s Petrophysical Handbook on CD-ROM
at no extra cost.
Updated 6 July 2005
CHAPTER
TWENTY-SIX:
DIPMETER LOGS Continue to Chapter Twenty-Seven Publication History: This Chapter was first published in 1979 as part Chapter Five of Volume Two of The Log Analysis Handbook, a series of training seminars presented by the author. Updated 1985 and 1992. The material was also published as "Dipmeter Tools and Presentations", by E. R. Crain, Canadian Well Logging Society Journal, Dec 1992. This electronic version reorganized and revised Oct 2002. CHAPTER TWENTY-SIX: DIPMETER LOGS 26.00
Introduction To This Chapter One of the major sources of information for developing exploration plays is, of course, data in existing well files. Logging tool design and data presentation have evolved dramatically over the seventy year history of logging. As a result, the log analyst will be faced with a wide variety of data quality, depending on the age of the well file. For this reason, we review the evolution of dipmeter tools, dipmeter calculation methods, and dipmeter presentation methods in considerable detail. From this, you should be able to decide if the available data will answer any of the geological questions you may wish to pose. If you plan to use existing dipmeter data for serious exploration, you must be aware of the differences and limitations of each tool and computation method. While the primary source of structural information from logs is the dipmeter, corroborative evidence from correlation to offset wells, oriented core data, and seismic data is needed to confirm or deny some analyses. There is often more than one plausible solution to the analysis of structural and stratigraphic problems. Dip
data, along with other log curves, is critical for analysis of
fractured reservoirs. This topic begins in Chapter
Twenty-Eight. Structural and stratigraphic uses of dipmeter
results begins in Chapter Thirty-One.
The most modern versions of dipmeter tools are covered in Chapter
Twenty-Seven. Resistivity microscanner tools and the most
recent software products are described there. 26.01
Evolution of the Dipmeter Concept In 1933, attempts were made to evaluate dip by analyzing resistivity anisotropy effects on a modified electrical log. The resistivity of a layer is usually lower parallel to the bed than perpendicular to it. By taking resistivity measurements with suitably arranged electrodes, the dip direction of thick, well stratified beds could be found. The dip angle had to be known from cores, and the hole direction had to be measured. This was possible using a device called an electromagnetic teleclinometer, which sent a signal up the logging cable proportional to the tool's deviation from the vertical. From this data, a crude dipmeter survey was presented. It is doubtful that any copies survive in well files. More modern data is often available in any event. The anisotropy dipmeter was supplanted in 1943 by a tool using three simultaneous spontaneous potential measurements oriented 120 degrees apart around the circumference of the logging tool. Using the same principle that three points define a plane, the tool provided sufficient data, along with bit size, deviation, and direction, and tool orientation in the hole, to calculate dip. The three points were taken as the bed boundaries defined by the SP curve from each electrode. The
balance of the data came from a photoclinometer survey taken at
stations near the top and bottom of the recorded intervals of
the dipmeter curves. A schematic example of the concept is shown
in FIGURE 26.01. The photoclinometer consisted of a magnetic compass,
a ball bearing in a graduated curved dish, and a camera that photographed
these components on demand.
This sounds simple. However, the magnitude of dip which is of interest in exploration is from a fraction of 1 degree to vertical. This poses serious constraints on tool design and data analysis. For example, a regional dip of about 50 ft/mile is equivalent to 1/2 degree dip. Local structure or drape over deeper erosional surfaces may modify this dip to flat or 1/2 degree in another direction. In some areas this is significant and could define the trapping mechanism. The dipmeter device, the recording process, and the curve correlation methods must have sufficient resolution to enable us to see this small difference. A bed dipping at 1/2 degree across a 9" borehole will be less than l/10 inch higher on one side of the hole than on the other. The displacement between curves shown on Figure 26.01 will be less than 0.1 inches if recorded at 12 inches per foot of borehole. If recorded at 5 inches per hundred feet, a normal detail logging scale, this displacement would be only 0.0004 inches on the film. As a result, scales of 60 inches per 100 feet were used. Now the 1/10 inch displacement is represented by 0.005 inches - a measurable distance on the film. Due to the relatively round shape of the SP curve at most bed boundaries, this level of resolution was not achieved with the SP dipmeter. Moreover, the tool was useless in carbonates where SP does not develop well. The only dips presented were those from major bed boundaries where dip was steep enough to be obvious. Although
the SP dipmeter was abandoned quickly in favour of three resistivity
curves, the photoclinometer survived well into the 1960's as a
directional survey tool. A sample is shown in Figure 26.02. In
addition to the photographs of the compass and deviation ball,
typed listings of computed results and a plan of the well track
were presented (Figures 26.03 and 26.04). Since the well bore
often deviated, without any help from the drilling crew, to keep
the bit perpendicular to the formation dip, the directional survey
data was sometimes used as a guide to dip.
The resistivity dipmeter used three laterolog curves instead of SP curves, mounted on the same rubber arms as were used for the SP version. Accuracy was better in hard rock areas. A sample is shown in Figure 26.05. Both SP and resistivity dipmeters were only recorded over selected intervals, chosen by observation of the other open hole logs. Only short intervals where there is lots of curve action were suitable for dip computation.
Typical computed results from the SP or resistivity dipmeter are shown in Figures 26.06 and 26.07. Although rare, examples can be found in files for wells drilled in the 1940's.
In 1950, better accuracy was obtained by a newly designed dipmeter utilizing three microlog resistivity pads. Now a continuous log could be made, and with very detailed resolution from the microlog pads, a fine scale dipmeter was a reality. In 1952, the microlog pads were replaced with microlaterolog pads which measured conductivity instead of resistivity. Orientation data was recorded simultaneously and continuously with a device called a poteclinometer. Poteclinometer is a contraction of the word potentiometer (a variable resistor) and inclinometer - this word sounds a lot like the earlier photoclinometer. Directional output from this device is an electrical signal instead of photographs. Data consisted of hole deviation angle, relative bearing (which describes the angle to the high side of the hole from pad number one), and the azimuth (which describes the angle between magnetic north and pad number one). This is sufficient data to orient the dip azimuth and the direction of hole deviation. The algebra is described later in this Chapter. This eliminated the need to stop the tool to take pictures with the photoclinometer. Directional surveys run with this equipment were also more accurate, but considerably more expensive. The optical comparator was also developed during this period (see next section for details of its use). This increased dip accuracy further by reducing errors in measuring the offset between traces. The computed data was presented in the same tabular and graphical fashion as previously (Figures 26.06 and 26.07), but with considerably higher frequency. However, by 1958, some hardy souls were plotting individual dips as small arrows on a graph of dip magnitude versus depth. The direction of the arrow represented the dip direction relative to a compass rose with north at the top, as in Figure 26.08. This was the precursor to the now common arrow plot, sometimes called a tadpole plot, generated by computer. Computer plotting was first seen around 1961.
The
first attempts to legitimately use detailed dip data for stratigraphic
evaluation occurred around 1955. An example of the difference
in data quality and quantity between short interval and continuous
data is shown in Figure 26.09. The raw data was recorded at 60
inches of log for 100 feet of wellbore, or 1:20 scale, shown half
size in Figure 26.10. Literally miles of this photographic paper
was developed, processed, and sifted through the optical comparator
each month. Most of it has deteriorated or been destroyed and
is not available for re-computation.
Fortunately, beginning in 1961, dipmeters were recorded on digital magnetic tape, reducing and finally eliminating the need for the detailed paper logs. The offsets between traces were derived by computer correlations, leading to a whole new language: correlation window, step length, search angle, etc.
A special "speed button" on one pad provided information to the program to compensate for minor speed differences as the tool moved up the hole. These variations created scatter in the computed results (Figure 26.11). In addition, a synthetic resistivity curve was generated from the dip curves, to be used as a correlation curve.
FIGURE 26.12: Geometry of four pad dipmeter
In 1975, secondary computer processing, called CLUSTER (Schlumberger trademark) or SHIVA (Gearhart trademark), were developed to validate the results from the standard program. Other secondary programs were developed to enhance stratigraphic features, notably GEODIP (Schlumberger trademark). These processes are described later. About 1980, three axis accelerometers and three axis magnetometers replaced the magnetic compass, relative bearing, and hole azimuth potentiometers. However the log still presented these three curves, derived now from the solid state sensors instead of the more failure prone electromechanical devices. A further refinement in 1983 created the stratigraphic high resolution dipmeter. An additional electrode set was added to each pad giving eight dip correlation curves instead of four. With this number of measurements, the results can be presented more often, as many as 10 or 20 per foot if desired, instead of the more usual 1 or 2. Better speed correction is provided by accelerometer data from sensors inside the tool. Typical raw data plot is shown in Figure 26.15. A six arm dipmeter has also been developed to meet the need for stratigraphic information, with a lower cost tool.
Three dip computation modes are available from the stratigraphic high resolution dipmeter. First is the usual pad to pad correlation, which benefits from the extra redundancy of two electrodes per pad. This is called Mean Squares Dip or MSD, and often is used for structural or regional dip analysis. The dip is a weighted average of all pad to pad dips. In strongly parallel beds, the result is very good, but in cross bedded formations with varying dip, the average dip has little significance, except to show overall direction of dip. Second is called Continuous Side by Side or CSB dip correlation using only the individual electrode pairs on each pad. Dip vectors from adjacent electrode pairs are used to define dip. CSB dips respond to short interval, low contrast changes often characteristic of internal layering in clastics, but also will respond to high contrast structural dips. It is very useful for structural dip analysis in high angle apparent dip, greater than 50 degrees. In finely bedded rocks exhibiting cross bedding, considerable detail can be shown if the correlation length and step distance are kept fairly short. Third are pad to pad correlations using a pattern recognition rather than cross-correlation system. This is called Local Dip or LOC dip and responds to non-repetitive events such as erosional surfaces or breaks in the depositional sequence. A comparison of the three modes with normal high resolution dipmeter results is shown in Figure 26.16. It is now possible to analyze data with a resolution of a few inches and compare it to core data (Figure 26.17).
In 1986, the "ultimate" dipmeter was developed, called the formation microscanner. Using an additional 27 electrodes on each of two pads of the dipmeter; each pad records 27 microresistivity curves spaced 1/10 inch apart on the borehole surface. Each pad covers a 2.8 inch wide portion of the circumference of the well bore. Several passes over the interval will often provide virtually complete coverage of the rock face. A microscanner tool with fewer (sixteen) electrodes per pad, but with four or eight imaging pads, is now available, and provides better coverage of the well bore wall than the two pad version. The electrodes are smaller, allowing for higher resolution, but are spaced to provide the same wall face coverage, about 2.5 inches per pad. In an 8 inch diameter hole, electrode coverage is about 80% and in a 6 inch hole is greater than 100%. This overcomes one of the major complaints about the FMS, namely the number of passes needed to obtain a complete image of the well bore. More detail on this tool can be found in Chapter Twenty-Seven. The
resistivity traces are translated into images based on their relative
resistivity values, in either black and white or colour. The gray
scale or colour spectrum can be stretched or squeezed in the computer
to enhance certain features, such as porosity, fractures, or shale
laminations. Images can be plotted at the same scale as the core
photographs for comparison. A sample is given in Figure 26.18.
The primary use of the tool is for identification of irregular features, such as vugs and fractures, for accurate sand counts in thin bedded zones, and for identifying stratigraphic features. If sufficient rock face is imaged, dips can be found by digitizing the bedding planes visible on the microscanner image, or by automatic computation using all valid image traces. Note that a planar, dipping, bedding plane will trace a sine wave on a circumferential image, such as those made by the microscanner or a borehole televiewer. The dips found by FMS dip processing are superior to CSB or LOC dips because a larger number of resistivity traces can be used in the calculation. They can be computed automatically and displayed on the FMS image. In addition, calculated dips can be edited or removed, and new bed boundary correlations picked with a mouse on an interactive CRT image. Thus dips that pass or fail preconceived processing criteria can be deleted or added as the analyst desires. An example of this technique is shown as a case history in Chapter Seven. A microscanner has about 10 times the spatial resolution of a televiewer and 500 times the amplitude resolution, due to the difference in contrast between the resistivity and acoustic impedance ranges measured by the respective tools. Schlumberger introduced a dipmeter for use in nonconductive mud systems in 1988. It uses micro induction resistivity measurements instead of the usual electrical resistivity pads. A knife blade electrode, or scratcher pad, version is available from several suppliers In 1989, a 4 arm focused acoustic dipmeter was introduced by Atlas Wireline, with a resolution of about 1 cm.
Manual correlation and computation methods were developed first and there are several different methods of doing the work. The dipmeter curves must first be correlated; this may be done by slipping a print of a log under the film used to make the print and measuring the depth displacement between peaks and valleys on the curves. Pad number one is used as a reference to measure displacements to each of the other curves. Another method of curve correlation uses an optical comparator, a system of mirrors and lenses which allow the user to optically lay one curve over another and shift it up and down. The amount of shift is measured mechanically on a dial and is recorded as the displacement. After these correlations have been made, the azimuth of the number one electrode, the borehole deviation angle, the relative bearing, and the borehole diameter from the calipers are recorded. This information, plus the depth, is necessary to compute the dip angle and dip direction of a point referenced to magnetic north. Because true dip is referenced to true north, we must also account for magnetic declination of the region. Mathematical formulas to solve this geometric puzzle are given later in this Chapter. The manual calculation of dip magnitude and direction with the above information was made in several ways: by using a calculator and trigonometric tables, a scientific programmable calculator (after 1970) with trig functions, a mathematically derived physical computing device (in other words, an analog computer), or stereographic nets, the latter being the most common manual method used in the past. A very small amount of hand calculator work is still done today. Another method of dipmeter computation utilized manual correlation and computer reduction of the data. This type of processing was originally developed to minimize turnaround time and to allow the tedious, time consuming computation and plotting of results to be performed by a digital computer. This may still be done today for re-computation of continuous dipmeters recorded on paper, or on 7 track digital tapes (which are unreadable by most modern computers) for which the paper records are still available. The most recently developed system of computation is computer correlation and calculation from data on digital magnetic tape. The data from the magnetic tape is entered into a digital computer and processed. In the correlation program, the digital information representing the dipmeter curves is stored in memory and the data from one trace is compared to the other traces to determine the vertical displacement between the traces. After these displacements are calculated, the tool orientation information is used to compute the actual formation dips. The standard correlation process is performed by a mathematical function called cross-correlation, in which the offset distance between events on two curves are found. The distance between the center and the maximum amplitude on the correlagram indicates the displacement between the two curves. The offsets for all curve pairs are then adjusted to obtain the offsets relative to the center of the correlation interval. More exotic forms of correlation, some based on pattern recognition theory, are used in the newer programs. The length of the portion of the curve being correlated is called the correlation interval, correlation length, or correlation window. Correlation interval is usually between one and four feet, but can be smaller or larger. The correlation is calculated at regular intervals along the log. The distance between correlations is called the step distance and is usually 1/2 to 1/4 of the correlation interval. One dip value is calculated at the center of each correlation window, and the dip value is plotted at each step distance.
FIGURE 26.19A: Dipmeter computation definitions The number of dips computed from computer processed logs can be any density required for a particular purpose. For structural analysis, normal densities range from one computation every one or two feet to one computation every ten feet. In those instances where additional information is required, such as for stratigraphic analysis, points as close as every few inches can be computed. The usual way to describe these parameters is in the form CORR x STEP x ANGLE. For example a 4 x 1 x 45 process uses a 4 foot correlation, a 1 foot step, with a 45 degree search angle. The recommended defaults for dipmeter processing are: Low angle structural dip: 4 x 2 x 45 eg: normal or reverse faults, folds High angle structural dip: 8 x 4 x 80 eg: overthrust faults, recumbent folds Sand body stratigraphic dip: 2 x 1 x 30 eg: beach, bar, channel, drape Complex stratigraphic dip: 1 x 0.5 x 30 eg: submarine fan, scree slope, turbidite A
fourth parameter is sometimes used to indicate that the program
can search farther up the curve if no correlation is found. This
is shown as: The effect of a shorter correlation interval is shown in of Figure 26.19B, where only regional dip is found in the long interval case, and stratigraphic dip is superimposed on the regional when a short interval is used.
The problem with dip determination by cross-correlation is that it does average all dips found in the correlation interval. If both structural and stratigraphic dips are present, the average may not reflect either of them correctly, regardless of the correlation interval. Regional dip is therefore usually chosen in a nearby shale or bedded carbonate thick enough to give an accurate result, without interference from stratigraphic events. Many dipmeters have been computed with inappropriate parameters and could be improved by re-computation with a better choice of values. The defaults shown above are just starting points. In particular, parameters for steeply deviated holes may need considerable experimentation and variation throughout the hole.
In manual correlations, one could correlate three pairs, determining three displacements. For perfect closure, the algebraic sum of the displacements must be zero. Usually, because of the inaccuracy of the optical comparator, a small closure error existed. This error could then be distributed among the three displacements as a small correction before final determination of the dip. In practice, this was an onerous task, and two pairs were often picked with no attempt to determine closure error. In automatic correlations, two kinds of closure errors can occur: small ones due to minor variations in shape between the three curves, and large errors. Small errors are handled as for manual computation. When a large error exists, it is because at least one of the correlations is in error - the same geological event is not being picked on all three pairs. In manual correlation, a large error was usually fixed by re-picking one of the correlated curves. For an automatic computation, we have to choose between three possible computable dips, only one of which may be correct. There are no strong mathematical rules to choose the correct dip. If closure error is large, the usual procedure is to compute no result and display no dip arrow. The three arm tool is also vulnerable to adverse hole conditions. If one curve degenerates, for instance when one pad fails to make a good contact with the borehole wall, the computation of dip cannot be made at all. This happens often in deviated holes or in out-of-round holes, resulting in more intervals with no result.
Six curve pair correlations can be attempted between four curves. The adjacent curve pair displacements are designated respectively as h12, h23, h34, and h41, and the diagonal displacements as h13 and h24. These six displacements can in turn be paired in thirteen different ways to provide thirteen dip evaluations for the same level. For the six arm dipmeter, 15 pairs are possible, leading to additional redundancy. The result from each combination is referred to as a dip determination. In recent practice, however, only four or five correlations are made, leading to a maximum of eight possible dip determinations per level. This reduces computer time. Four
arm closure error (Ec) is given by the algebraic sum of the four
adjacent curve displacements: For perfect closure, Ec = 0. Three arm closure error can also be computed on a four arm or six arm dipmeter. In this case, closure error is given by the algebraic sum of two adjacent curve displacements and their associated diagonal displacement. This error is distributed around the displacements in proportion to the amount of each displacement.
When
four arm closure error is zero, planarity error (Ep) is defined
as: For perfect planarity, Ep = 0. Similar equations exist for the six arm dipmeter. If
closure error is zero and planarity is not zero, then several
things may be possible. One is that the bedding may not be planar,
such as in the case of festoon current bedding or aeolian dune
surfaces. Other possibilities are lack of pad contact with the
hole wall and possible miscorrelations. The latter are, The flow chart in Figure 26.20 shows the complex logic involved in Schlumberger's high resolution dipmeter program. It handles the closure and planarity problems in numerous ways, based on the number and quality of correlations found. The output listing from this program is shown in Figure 26.21. Notice that some of the logic choices are coded on the listing and others on the arrow plot by use of alternate symbols, indicated on the bottom of Figure 26.21.
Dips can also be coded and presented in such a way as to indicate the fact that they are non-planar. This would help an analyst interpret the bedding, as shown in the example in Figure 26.22, which was processed using Gearhart's OMNIDIP program.
26.07
Determining Dip By Clustering and Pooling None of these approaches used any geological knowledge or any sophisticated statistical aids in the solution. The cluster approach for dip selection was developed by Schlumberger to help eliminate the problem of closure and planarity errors. The CLUSTER program name is a registered trademark of Schlumberger. The CLUSTER program does no curve correlation; it operates on output data from an existing dipmeter program. The best reference is “Cluster - A Method for Selecting Most Probable Dip Results”, V. Hepp and A. Dumestre, SPE Paper 5543, 19726. The CLUSTER method assumes that correlations are valid if they repeat when the correlation window is moved over a small step distance. If a dominant anomaly exists, it controls the correlation on at least two adjacent dip computations, and it follows that the dominant anomaly defines the same dip value for as long as it is included inside the correlation window. The scattergram of points shown on Figure 26.23 presents an illustrative plot of all the dips computed from all the retained displacement pairs of ten computation levels. Each dip is plotted at a location on the plot defined by its magnitude and azimuth, and coded to represent a weight indicating the quality of the correlation. There is a great deal of scatter, indicating the noisy nature of the correlated curves. However, two concentrations of points of greater consistency, marked Cluster 1 and Cluster 2, are present. Redundant dip results thus allow us to choose groups of dips which show some stability throughout the zone and to choose the displacement combinations which contribute dips to the group. Since Cluster 1 represents the greatest concentration of dips, it should be nearest to the dip defined by the dominant anomaly. If no displacement pair contributes to Cluster 1, then perhaps a contribution is made to Cluster 2 and this, also, should be a valid dip, even though the indication of consistency is not as strong. Failing this, the displacement information must be regarded as meaningless. For such levels no results will be printed on the CLUSTER output listing. In the example of Figure 26.23, ten levels were grouped together from an arbitrarily selected interval. In the actual clustering procedure an attempt is made to group levels together in a meaningful fashion into short intervals or zones. Zoning is achieved by testing the stability of successive adjacent curve displacements in the input listing.
The test for stability checks the displacement value in the next level upwards to see if it is similar to the current one. If this test is satisfied, over several consecutive levels in at least two contiguous adjacent curve displacement columns, the zone is stable. Zones that do not satisfy these criteria are called open zones. The two types of zones are merely a convenient way to break up the interval for clustering. Both kinds of zones can provide meaningful dips, depending on the quality of the correlations. Zoning is a preliminary sorting procedure. Both stable and open zones are subsequently treated in the same fashion. Zone length can vary from one to fourteen consecutive displacements. No indication of the zoning used is shown in the output arrow plots or the standard output listing. The correlation coefficient measured along with the displacement correlation is an important criterion of the quality and is not ignored in the choice of good correlations. To account for this, the dip points placed on the scattergram are weighted according to a coefficient called the level weight. A greater weight raises the contribution of retained dip determinations and enhances their chances of being selected as candidates for clustering. If the quality of the correlation reported for the level by the source dipmeter program is good, the contribution to the level weight is 3, if fair, it is 2, if poor, it is 1. If the level shows four arm closure (a double asterisk on the original listing), weighting is doubled. Thus, the level weight varies from 1 (poor) to 6 (excellent). Clusters thus identify the probable ranges of dips for the zone. The program returns to each dip level in turn and retains only those dip determinations which fall within one of the clusters. If one is found in the highest ranked cluster, it is retained, and if there are two or more, their vector average is retained. If none are found, the program can expand the area included in the cluster. If cluster expansion fails, the cluster of next lower rank is checked. It may happen that no contribution is found from a level to any of the defined clusters, in which case this level is considered to have no result. Similarly, if no clusters are found at all within the zone, no result is shown on the output listing. This occurs when the data are so poor that no meaningful displacement combinations can be made. Since clustering only uses data from a previously applied dipmeter program, it cannot find new correlations and it cannot find dips where none were found on the original. It may be possible to obtain new results in "no result" intervals by reprocessing the original dipmeter with new parameters. A typical set of input data to CLUSTER is shown in Figure 26.24 and output for the same interval is shown in Figure 26.25.
The process of dip retrieval that has just been described systematically attempts to provide one dip for each correlation window. However, the basic idea of the method is that consecutive correlation intervals must overlap, in order that dominant anomalies can affect the clustering process. As a result, it is quite usual that the same dip is repeated twice when the overlap between consecutive levels is 50 percent of the correlation length, or four times when the overlap is 75 percent. Users of dipmeter surveys should train themselves to recognize doublets or quadruplets as representing a single anomaly. However, it would be nice if the computer would do the same and represent it by a single dip result, at the midpoint between the depths of the two or four component levels. This is accomplished by pooling clustered dip results. Pooling
consists of testing the results from successive levels, up to
a number of levels called the pooling constant and controlling
whether their angular dispersion does not exceed a fixed value,
called the pooling angle. If the test is satisfied, the component
dips are replaced by their vector sum, the pooled vector. Its
dip magnitude and azimuth are converted to geographic coordinates
and printed out at the mean depth, together with other data about
the computation. The sample in Figure 26.26 can be compared to
the un-pooled results in Figure 26.25.
Two separate output files are created: one for the clustered data and one for clustered and pooled data. Thus, in reality, two different dipmeters are created from the same data, using different rules in their analysis. Figure 26.27 (left side) shows an arrow plot for clustered and pooled results. The arrows with black circles represent high quality ratings. Usually a blackened circle corresponds to pooled results; however, it is possible that a non-pooled result from a high quality level could plot as a blackened circle.
Pooled results are generally plotted on 1 or 2 inch per 100 feet depth scale. This can be done since there are fewer arrows to plot. Thus, one use of pooling is to provide a dip record on a depth scale commonly used for correlation. Usually, structural analysis is all that can be accomplished with this plot. The arrow plot represents dip magnitude and azimuth from the output listing at their proper depth. However, it does not represent the effect of uncertainties, as represented by the dispersion of dip values and their directions in the original data. The fan plot is a method to present this knowledge as the quality indicator instead of the more usual open or filled circles. A sample is shown on the right side of Figure 26.26. In the fan plot presentation, a small circle surrounds the center value of dip magnitude. A small line segment extends on both sides from a lower to a higher dip magnitude value, essentially indicating an error bar. In similar fashion, a fan extends from a lower to a higher dip azimuth value. These values are determined from the combination of the pooled dip magnitudes and azimuths and the angular dispersion parameters. They encompass all values within one standard deviation from the mean. The length of the fan represents the number of dips used in the statistic. Thus, it is probable that the true dip is contained inside the possible values within the fan, both in magnitude and azimuth. The same value of the angular dispersion parameter may correspond to a nearly closed fan at high values of dip to a wide open fan near zero dip magnitude. When angular dispersion exceeds dip magnitude, the azimuth value cannot be specified with any kind of certainty and no fan is drawn.
The following description was paraphrased from “An Approach to Detailed Dip Determination Using Correlation by Pattern Recognition”, P. Vincent et al, SPE Paper 6823, 1977. One of the objectives of GEODIP is to overcome the rigidity of the fixed correlation interval procedure and provide a density of information more closely related to the geological detail seen on cores. There was also the feeling that the dipmeter raw data contained more information than was actually being used, even by the improved processing achieved with clustering and pooling. After all, the electrodes had a resolution of 0.2 inches and often one or two foot data was being presented. Many features, such as peaks and valleys, are identifiable by eye from curve to curve on the dipmeter. These features have various thicknesses (from one inch to several feet), amplitudes, and shapes. Each feature may be considered to be the signature of a geological event in the depositional sequence. Moreover, the dip of the bedding is not necessarily constant, and may sometimes vary rapidly. The method of correlation by pattern recognition is best adapted to automatically detect these curve features, to recognize them from curve to curve, and to derive dips for the boundaries of each individual feature. Different curve features of the same type are often very similar and easy to confuse. The human correlator avoids this ambiguity by constant eye movements to confirm or invalidate hypothetical correlations. In so doing, the correlator implicitly, often unconsciously, applies some logic rules which are integrated into the perception process. In the GEODIP method, equivalents of such rules and safeguards are included, as far as they have been recognized, in the program logic. Programs of this type have been called expert systems, or knowledge based systems, because they contain the rules of experienced analysts. The method is constructed around a basic law justified by geological conditions of deposition, the rule of non-crossing correlations. This rule states that the layers are deposited one over another, so that they can wedge out but they cannot cross. The consequence is that if Event A appears above Event B on one curve, it cannot appear below B on another one. This rule induces a certain interdependence between all of the correlations. In this method, the correlations are not viewed as independent realities, but as parts of a more general structure having internal organization and rules. Where only two curves are considered, it is a simple matter to recognize crossover correlations and disregard them. But when more than two curves are involved, as in Figure 26.28, complex logic is required within the computer program to perceive that the correlation (A1, A2), is inconsistent with the correlations (B1, B3) and (C2, C3). Actually, it is the set of the three correlations which is, as a whole, inconsistent. It cannot be inferred, from what is shown, which one is incorrect.
The goal of the computer logic is to select the largest set of curve to curve correlations that does not include any crossovers or implied crossovers. To meet this goal, a branch of modern mathematics called the theory of partially ordered sets has been applied to the description and consistency checking of sets of correlations between curves. While this theory is necessary to properly implement on a computer the rule of non-crossing correlations, an understanding of the mathematics is not needed to appreciate what it achieves. The
method of correlation by pattern recognition is composed of two
main phases: In phase one, each curve is analyzed individually with reference to a catalog of standard patterns or types of curve elements, such as peaks, troughs, spikes, and steps, and is decomposed into a sequence of such elements. At the end of the feature extraction phase, the curves are replaced by their description in terms of elements. Each
element is associated with one or two boundaries which give the
position of the element on the initial curve as well as a pattern
vector, which is a series of numbers characterizing the shape
of the element. The pattern vector for a peak contains a description
of its: Other features have their own unique list of parameters in their pattern vector. In
the correlation phase, the method tries to successively match
elements of one curve to similar elements of the others. The objective
is to recognize the same geological event as it appears on different
curves. The basic criterion is the comparison of pattern vectors.
To find these correlations, a coefficient is computed which is
a measurement of the likeness between any two elements, using
the following equation: Where: Low values for L mean a high degree of likeness. Then, the procedure attempts successive correlations according to a built in order of precedence: large troughs, then large peaks, then medium troughs,... The program retains already accepted higher precedence correlations in order to forbid crossing them in further attempts with correlations of lower rank. When two elements are considered to be a match, the corresponding upper and/or lower boundaries are then correlated. The resulting dips are computed from the displacements measured between these correlated boundaries and not those measured between the elements themselves. At the beginning of the correlation phase, an initial search angle, corresponding usually to the highest value of expected dip magnitude, is imposed. The initial search distance is computed from the input search angle, the orientation parameters, and the diameters measured by the tool at the particular level. As correlations are made and accepted, the search distances are modified, as necessary, to avoid crossing correlations. It may happen that no large element can be correlated with any large element of the same type on the search curve. To handle these cases in following passes, requirements are relaxed, for instance, by authorizing the correlation of a large element of the base curve with a medium element of the same type on the search curve. On the other hand, the correlation of unlike elements, such as peaks with troughs, is forbidden. Thus, the correlation phase proceeds by successive passes, searching first for the most obvious correlations, those having the lowest likeness coefficients. Each time a correlation is retained, it is memorized in order to limit subsequent search lengths for correlations with higher likeness coefficients. Pattern recognition correlation is also used in determining the velocity correction, allowing almost inch-by-inch detection of speed variations. Figure 26.29 shows the graphic presentation made by automatic plotter. Because of the large number of dip results found, a depth scale of 1/40 (30 in. per 100 ft.) or 1/24 (50 in. per 100 ft.) is used instead of the usual 1/240 or 1/200 scales. This uncommon depth scale is better adapted for the high resolution available for very thin beds. The semi-horizontal lines connecting the traces represent the correlation of element boundaries. Figure 26.30 shows a typical listing from this program.
With GEODIP there is no quality rating of the dip determination. The visual display of the curves and the correlations enable analysts to decide for themselves about the reliability of the correlations according to the character of the curves. Comparison to core data is one way to check the validity of the results of stratigraphic analysis. Figure 26.31 illustrates one such comparison.
1.
MSD Dips (Mean Squares) There is no vertical continuity logic or clustering routine in the MSD computation, and each level is autonomously processed. The clustering is thus replaced by an analysis of the local scattering of the displacements. This method benefits from the ample redundancy available from 28 displacements, while two would be enough to define a dip, reducing the possibility of producing random dips or noise correlations. 2.
CSB Dips (Continuous Side-by-Side) With only four side-by-side correlations, the only cross check available is to verify that, for a planar bed, the displacements obtained from opposite pairs of curves (eg. 1-1A and 3-3A) should be equal in value and opposite in sign. This occurs if closure error is zero. If this is the case, any combination of these displacements yields the same dip and any orthogonal pair is used to produce the dip at that depth. If this is not the case, a window is opened around the level under examination, and the vertical continuity of the displacements within the window is checked. The orthogonal pair showing the smoothest continuity within the window is selected for dip computation. Whether a good quality dip (full arrowhead), a low quality dip (open arrowhead), or no dip is output, is a function of the quality of the side-by-side correlations established and of the vertical continuity of the displacements. 3.
LOC Dips (Local Derivative) To be retained as a LOC dip, an event has to be recognized on at least 7 of the 8 curves, while the GEODIP logic requires only 3 out of the 4 curves. Thus the LOC dip logic is more demanding than the GEODIP logic, which explains why generally fewer LOC dips are obtained than GEODIP results on comparison runs. The LOC dips are further refined by a cross correlation made on a 12" interval, while GEODIP results are computed directly from the spot events on the curves. This cross correlation involves the eight curves and includes a repetitive best fit and rejection logic as in the MSD computation, with similar criteria for quality coding (full or open arrowhead). A measurement of the planarity is derived from each of the possible dip planes at any level. The retained value corresponds to the surface which best approximates the set of these planes. By convention, a perfectly planar surface has a planarity of 100. Some events are recognized on only some of the dip curves. In this case, the available correlations are traced across the applicable curves, with an optional notation of "F" (Fracture) or of "P/L" (Pebble/Lens) for single pad events or two/three pad events, respectively. These interpretations, however, are not to be considered as certain, but rather as possible. Due to their origin (pad-to-pad correlations), the LOC dips have meaningful lateral significance. If structural dip is present, it will normally be seen by the LOC dips rather than by the CSB dips. Generally the statistical agreement between the LOC and the MSD dips can be expected to be quite good. DUALDIP is the computer program which produces the standard SHDT result presentation. This includes CSB and LOC dips, the eight dip curves, the synthetic resistivity and gamma ray curves, calipers and hole drift data. The depth scale is usually 1/40, and as an option the MSD dips can be added to this output. A sample was shown earlier as Figure 26.16 and is repeated here as Figure 26.32.
Structural interpretation is done using the MSD dips. Due to the logic used, namely cross correlation made using long intervals, the MSD dips are the ones likely to represent laterally significant and vertically consistent geological events. For optimum use of the MSD dips, a reduced scale (1/200) plot is normally produced. This plot is also the single SHDT product when no fine scale studies are contemplated. The prime objective of the SHDT tool design is to improve the ability to provide reliable answers to sedimentary interpretation problems. While the rules of interpretation remain essentially the same as in HDT interpretation, there are additional possibilities. Among the information that can be retrieved by visual analysis of the dip curves, reconstructed resistivity, and dip arrows are: - type of lithology (shale, sand, conglomerate) from the shape and likeness of the curves. - fining upwards, coarsening upwards sequences. This is done by analyzing the resistivity variations across the sequence, either with the dip curves themselves or with the synthetic resistivity curve. Other open hole logs, such as the gamma ray (combinable with SHDT), are useful here. Care should be exercised using the resistivity, however, since fluid saturations have to be accounted for when inferring grain size variations from resistivity gradients. - homogeneous bodies (no apparent bedding) as opposed to finely striated, laminated bodies. - parallel vs nonparallel bedding. This is especially important in sandstones, and has found recent applications to the study of turbidites. - correlation lines: some correlations involve the eight resistivity curves, some do not. The most appropriate interpretation (pebble, lens, fracture, other) will be made on the basis of the dip curves (conductive or resistive anomaly, number of pads involved, etc.). - fractures: open fractures will show as isolated conductive spikes which may or may not correlate with similar spikes on other dip curves. Some
of the important uses of the CSB dips are: - determination of the direction of sediment transport, a corollary to the above. This is especially interesting in severe cases of cross bedding, when the only dips produced by long interval correlations generally correspond to those of the individual sedimentary units, seen at their interfaces, and not to the actual current bedding surfaces. - conventional sedimentary interpretation (red, blue patterns, direction of sand body thickening, etc.). All of this can be done on an almost microscopic scale. CSB dips are also very useful, and often better than MSD dips, in high angle apparent dips, when longer correlation intervals are used. LOC dips can be used to study such features as: - nonparallel bedding, for example, when the upper and the lower boundaries of thin beds do not have the same dip. In cases of poor planarity, the event recognition logic will be too tight for a LOC dip to be produced, and the MSD curves may then provide the answer. This is particularly important if this bed is to be found in another well, or when looking for the direction of updip or downdip thickening. - cross bedding: the LOC dips will see the interfaces between the individual sedimentary units, when apparent. This dip may not coincide with the angle and direction of deposition in cross bedded formations (eg. tabular bedding, foreset beds). - turbulence of deposition, when causing non-planarity of bedding. The
MSD dips are normally not used for sedimentary studies, being
the result of an averaging of the dip curve anomalies over the
length of the correlation interval. They are usually presented
on the DUALDIP plots, however, for structural reference. The vertical
(depth) scale used for stratigraphic work makes it difficult to
see structural patterns in the MSD data. 26.10
In Conclusion Although dipmeter analysis can be ambiguous, sufficient geological constraints, local knowledge, and experience serve to improve skills and speed analysis. Modern computer processing, in particular dip removed arrow plots and stick plots, are essential ingredients. Image processing techniques, while relatively new, have proven useful because of their visual impact. However, the analysis of structure and stratigraphy from dipmeter data still depends on the basics: dip angle, dip direction, and a plausible model that fits the data.
2. Define correlation window, step length, search angle, closure, and planarity. What uses are made of this information? (15 marks) 3.
Describe cross correlation, pattern recognition, clustering, and
pooling as 4. What kinds of presentations are used for dipmeter data? (20 marks) 5. Give a brief chronology of dipmeter history from the 1930's to the present. (20 marks) 6. Explain the differences between MSD, CSB, LOC, and FMS dip calculations. What is each used for? (20 marks) 7. Find dip direction and azimuth from the following log segments. Draw a stick diagram for each. (25 marks)
8. Find dip direction and azimuth from the following log segments. Draw a stick diagram for each. (25 marks)
26.12:
Bibliography for Chapter Twenty-Six 2: The poteclinometer and the microfocused devices; Bricaud,J.M., Poupon,A.; 5th World Petroleum Congress, 9 p., 1959 3: Schlumberger continuous dipmeter; Perez,A.A., Hartsell,W.H., Gilreath,J.A.; Series #8, 29 p., 1960 4: The continuous recording of dipmeter surveys; Thibodaux,J.B.; Pan Geo Atlas Corporation Service Report, 5 p., 1963 5: The high resolution dipmeter tool; Allaud,L.A., Ringot, J.; Society of Professional Well Log Analysts: The Log Analyst, p. 3-11, 1969 6:
The high resolution dipmeter reveals dip related borehole and
formation characteristics; Cox,J.W.; Society of Professional Well
Log Analysts 11th 7: Continuous dipmeter; WEC; 17 p., 1971 8: HDT troubles; Schlumberger; Interoffice memo, 35 p., 1977 9: Slim hole dipmeter; Wroot,R.W.; 7th Formation Evaluation Symposium of Canadian Well Logging Society, 12 p., 1979 10: Dipmeter validity in deviated boreholes; Fitzgerald,D.D., Theriot,J.C., York,P.L.; Society of Professional Well Log Analysts: The Log Analyst, p. 8-18, 1980 11: Advances in diplog instrumentation; Johnson,W.M.,Jr., Angehrn,J.; 8th Formation Evaluation Symposium of Canadian Well Logging Society, 13 p., 1981 12: Stratigraphic high resolution dipmeter tool; Schlumberger; Manual, 23 p., 1983 13: The six arm dipmeter a new tool for detailed reservoir description; Goetz,J.F.; 10th Canadian Well Logging Society, 29 p., 1985 14: Micro-induction sensor for the oil based mud dipmeter; Kleinberg,R.L., Chew,W.C., Chow,E.Y., Clark,B., Griffin,D.D.; 62nd Annual Technical Conference of Society of Petroleum Engineers, p. 189-201, 1987 15: The oil based mud dipmeter tool; Dumont,A., Kubacsi,M., Chardac,J.L.; Society of Professional Well Log Analysts 28th Annual Logging Symposium, 15 p.,1987 16: Measuring RXO and dip in oil based mud with the six arm dipmeter; Chemali,R., Su,S.M., Goetz,J.F.; Society of Professional Well Log Analysts 30th Annual Logging Symposium, 25 p., 1989 Dipmeter
Processing 2: Supplementary computer programs for dipmeter analysis; Matthews,R.R., Mooney,T.D., Haynie,R.B., Albright,J.C.; Society of Professional Well Log Analysts, 19 p., 1965 3: An accurate method of low angle dip calculation; Schoonover,L.G.; Society of Professional Well Log Analysts 14th Annual Logging Symposium, 15 p., 1973 4:
Computer recognition of diplog patterns a tool for stratigraphic
analysis; Schoonover,L.G.; Society of Professional Well Log Analysts
15th Annual Logging 5: Cluster: a method for selecting the most probable dip results from dipmeter surveys; Hepp,V., Dumestre,A.C.; 50th Annual Technical Meeting of Society of Petroleum Engineers of American In stitute of Mining Metallurgical Engineers, SPE 5543, 1975 6: Geodip: an approach to detailed dip determination using correlation by pattern recognition; Vincent,Ph., Gartner,J.E., Attali,G.; 52nd Annual Technical Meeting of Society of Petroleum Engineers of American Institute of Mining Metallurgical Engineers, SPE 6823, 1977 7: True vertical depth, true vertical thickness and true stratigraphic thickness logs; Holt,O.R., Schoonover,L.G., Wichmann,P.A.; Society of Professional Well Log Analysts 18th Annual Logging Symposium, 19 p., 1977 8: The log analyst and the programmable pocket calculator; Bateman,R.M., Konen,C.E.; Society of Professional Well Log Analysts: The Log Analyst, p. 3-9, 1978 9:
A simplified true vertical thickness (TVT) calculation using a
programmable 10: Formation dip determination - an artificial intelligence approach; Kerzner,M.G.; Society of Professional Well Log Analysts: The Log Analyst, p. 10-22, 1983 11: SHIVA processing: the integration of fundamental geological principles with dipmeter computations; Enderlin,M.B., Epps,D.S., Yuratich,M.A.; Society of Professional Well Log Analysts: The Log Analyst, 22 p., 1985 12: Effect of tool rotation on the computation of dip; Waid,C.C.; Society of Professional Well Log Analysts 28th Annual Logging Symposium, 22 p., 1987 13: A rule based approach to dipmeter processing; Kerzner,M.G.; 63rd Annual Technical Conference of Society of Petroleum Engineers, p. 239-251, 1988 14: Dipvue analysis package; Atlas Wireline; Brochure, 1 p, 1988 15: The effects of noise on interval correlation methods; part 1: accuracy and precision; Waid,C.C., Faraguna,J.K., Easton,S.B.; 63rd Annual Technical Conference of Society of Petroleum Engineers, p. 227-238, 1988 Directional
Surveys 2: Surwel gyroscopic directional surveys; Sperry Sun; Catalogue, 20 p., 1970 3: Computing accurate directional surveys; Blythe,E.J.,Jr.; World Oil, p. 25-28, 1975 4: Model gives accurate wellbore displacement; Guillory,C.,Jr.; The Oil and Gas Journal, p. 138-141, 1975 5: Gyro technics; Slusarchuk,L.M.; Schlumberger, 53 p, 1976 6: Directional survey calculation; Craig,J.T.,Jr., Randall,B.V.; Petroleum Engineer, p. 38-54, 1976 7: Surface recording gyroscope system saves rig time; Guillory,C.,Jr.; The Oil and Gas Journal, p. 186-192, 1978 8: On site computer helps speed directional survey analysis; Kipcheff,J.T., Honeybourne,J.W.G., Penny,S.J.; The Oil and Gas Journal, p. 79-86, 1978 9: Survey procedures prepared for Panarctic Oils Ltd; Sperry Sun, 13 p., 1979 10: Directional survey calculation methods compared and programmed; Callas,N.P., Novak,P.C., Henderson,J.R.; The Oil and Gas Journal, p. 53-58, 1979 11: Programs enhance directional drilling; VanDusen,M.D., Stephens,H.D.; World Oil, p. 103-106, 1979 12: Graphs give thickness in deviated wells; Travis,R.B.; The Oil and Gas Journal, p. 87-94, 1979 13: How to avoid gyro misruns; Plite,J., Bount,S.; The Oil and Gas Journal, p.109-117, 1980 14: Directional drilling technology strives for speed and accuracy; Enenback,J.H.; Petroleum Engineer International, p. 124-132, 1980 15: Directional drilling information is refined by computer at SDI facilities; Ibbotson,G.; The Journal of Canadian Petroleum Technology, p. 35-36, 1981 16: Randomly simulated borehole tests accuracy of directional survey methods; Fitchard,E.E.; The Oil and Gas Journal, p. 140-150, 1981 17: Use of magnetic ranging logging tool to direct Amoco et al Steep A7-28-66-7W26M drilling; Duguid,A.T.; 33rd Annual Technical Meeting of Petroleum Society of Canadian Institute of Mining, 13 p., 1982 18: Use well logs to find proximity of relief wells to blowouts; Baldwin,W.F.; World Oil, p. 67-69, 1983
|
FIGURE
26.01: Photoclinometer for recording dipmeter data
The
geometry of a four pad device is shown schematically in Figure
26.12 and the arrangement of tool components in Figure 26.13.
Typical raw data curves and an answer plot are found in Figure
26.14. 
In
order to determine how far up and down each adjacent curve the
correlation is to be performed, a search angle is defined. In
moderate structural dip the search angle is usually 45 degrees,
but if expected dips are low, the angle can be reduced to eliminate
noise, or spurious dips caused by erratic wiggles on the curves.
Some computer programs use a search length instead of a search
angle. In steep dips, a higher search angle is required. These
terms are illustrated in Figure 26.19A.
ABOUT THE AUTHOR 
E.
R. (Ross) Crain, P.Eng. is a Consulting Petrophysicist and a Professional
Engineer with over 35 years of experience in reservoir description,
petrophysical analysis, and management. He has been a specialist
in the integration of well log analysis and petrophysics with
geophysical, geological, engineering, and simulation phases of
oil and gas exploration and exploitation, with widespread Canadian
and Overseas experience.