PROCESSING AND INTERPRETATION OF FMI IMAGES By Shri V.L.N. Avadhani

Basic Theory of Electrical Imaging

Fullbore Formation Micro Imager (FMI*) Measurement and Interpretation Theory

The FMI* is the latest generation electrical imaging device and belongs to the family of imaging services provided by the MAXIS 500* system with its digital telemetry capability. It is an extension of dipmeter technology in which scanning electrodes arranged in 24 electrode per pad/flap arrays (of four pads and four flaps) are used to provide a high spatial sampling of formation microresistivity in both the vertical and azimuthal directions on the borehole surface.A GPIT (General Purpose Inclinometry tool) is also run along with this service to calculate hole azimuth and relative bearing in order to orient the measurements with respect to north and the borehole. These two-dimensional microresistivity data are then mapped to gray scale or color to produce core-like borehole wall image that allows fine scale geological features to be described with a very good vertical resolution of about 0.2inch (5mm).

Tool Specifications (FMI) : This tool was developed by Schlumberger in 1991 as an improvement on the FMS4 (4 Pad-tool) developed earlier in 1987. The advantage was better borehole coverage - about 80 % in 8 1/2 holes. It is a semi-active focusing device (no direct focusing electrodes as in a laterolog) , so the response cannot be output directly as resistivity but is relatively proportional to the conductivity of the formation. A 16 Khz voltage is applied across each button with a return to the cartridge and an EMEX current at the button is measured at the same frequency. The voltage is increased automatically against resistive formations and lowered against conductive formation to ensure activity on the individual microconductivity curves. Electrodes: There are in total 192 electrodes distributed on four pads and four flaps. Resolution: The high resolution component of response gives button resolution ~ button size (5mm). The low resolution component of response is similar to LLS or SFL i.e., in front of thick homogeneous or thick thinly laminated formations, the response is close to LLS and SFL. Depth of investigation varies according to the resistivity of the formation. Against resistive beds the tool reads deeper than against conductive beds. The tool has a very large dynamic range - from less than 0.1ohm-m to more than 10,000ohm-m. The maximum logging speed is 1600 ft/hr (500m/hr), but outside zones of interest, it can be run at 3200 ft/hr (1000 m/hr) to acquire SHDT* dipmeter data only.

Image Processing : The field log prints show the individual microresistivity curves , the variable EMEX voltage, cable tension, hole size calipers,borehole azimuth,borehole deviation, relative bearing (Pad1 to top of hole angle) ,pad 1 azimuth and a preliminary equalized image. A moving window averaging algorithm is applied to the accelerometer and magnetometer data as they are recorded and the tool responses are compared to the Earths gravitational and magnetic field. The objective of data processing chain (at Wellsite or GeoFrame*) is to represent the conductivity changes as a variable density gray or color scale image.The major steps are: Speed Correction: This is necessary as the downhole tool motion does not correspond exactly to the surface cable movement due to the cable being elastic.When tool speed is higher than cable speed at wellhead, an interval of formation will be compressed i.e., a shorter log for a longer formation interval. If tool speed is lower than the cable speed at wellhead ,a stretching occurs.This will cause an error on the dip computation as well as bed thickness. On the image a saw-tooth pattern may occur.There are two methods available for speed correction a) Z-axis accelerometer (GPIT) based and b) image based. These two corrections remove most of the acceleration effects on the images prior to further processing.

Equalisation: This is done to correct the effects of electronics, non perfect pad contact and slightly different measurement sensitivities of electrodes.Without this step stripes may appear on the resulting images, due to different responses on individual buttons. The response is equalised by gain-and-offset operation over a sufficiently long interval so as not to remove significant formation features. Resistivity Scaling: This is done by using an LLS or SFL log to scale button response and finds use in obtaining sand counts , fracture aperture estimation and high resolution resistivity studies. This is a type of calibration. Normalisation: This is done to optimise the colour spectrum of the images. Normalisation can be either Static or Dynamic.

A)Static normalisation : A fixed set of color classes,obtained fromthe entire data set is applied to the image. It does not enhance all details but different parts of the image can be compared on a consistent basis. In a static image the colour will relate to the resistivity of the formation opposite the electrode. B)Dynamic normalisation: Sets of classes are determined using a sliding window placed along the image, and interpolated to give a continuously varying set of data classes. This is a technique of image enhancement . Dip information in conductive and resistive intervals may be obtained . A default of 1.0 m window is generally sufficient. The resulting image has a more or less uniform intensity.

Image Display: The lighter the shade the more resistive the formation.The left-hand edge of the azimuthal image track is usually North in vertical wells. For horizontal or inclined wells the top of hole (TOH) presentation is used as a reference. A green line marks the pad1 image on the plots.

Interpretation Techniques: There are various workstation based FMI/FMS image and SHDT interpretation software modules. Under GeoFrame* on Sun Workstations (UNIX based) the following programs are used for interactive dip picking and interpretation. BorDip* : This is used to do automatic Mean Squared dip (MSD) and Continuous Side by Side (CSB) computations. The answer products show structural and stratigraphic dip trends which can be further interpreted. BorView* : This is used for interactive (hand-pick) dip and fracture analysis on images. Dip sinusoids are used to fit planar features on the images. These dips are more accurate for bed boundaries picked in a sequence. Cross-bed analysis and fault identification can be done. Also fractures can be picked and mapped with apertures calculated using conductivity mapping techniques on scaled images. Stratigraphic interpretation on images of features like slumps, unconformities and clay drapes can be done. Also upward coarsening or fining sequences can be identified on the images with juxtaposition of conventional logs. A typical BorView* presentation can be seen in Fig. 1. Stereonet* : This provides a Wulf or Schmidt net based polar plot of dip data and can be very useful in paleogeologic reconstuction.

(N.B. A good way to check automatic dips (MSD & CSB) from SHDT data is to generate a SHDT Pseudo Image and do hand picking of sinusoids on that.)

Fig-1: A typical BorView* presentation with headers,dips and conventional logs. The boundaries (green tadpoles and sinusoids are hand-picked and are found dipping at low angles eastwards ,Well X , India.

The following two pages show some typical examples of image display and dip interpretation using interactive (hand-picked) dips using BorView* and Steronet* analysis.

Fig-2: FMI* Static and 1m-Dynamic normalised and scaled images ; Interval 2190.5-2192.5 , Well Y Hole size: 12 1/4. Interpretation results:

This interval shows a probable minor fault (subparallel to a main growth fault below -- not covered in this picture) at 2191.5m dipping at 40deg southeasterly. The fault is picked on the high angled conductive sinusoidal event showing bedding displacement.

Fig-3: Stereonet (Wulf) projection structural interpretation , Interval 2197m to 2291m , Well Y

Interpretation results:

The interval 2197 to 2291m was analysed in BorView* Stereonet . The bed boundaries (green dips) are seen to have a predominantly NNW-SSE trend with low dip scatter. The pole of the great circle passing through the beds has a N232deg azimuth . The faults (pink dips) are trending NNW-SSE as well indicating probable growth faulting and antithetic faulting as a result of slumping during delta front progradation.(from SE to NW). The low dip scatter may be indicative of a low energy, marginal marine environment of deposition.

Theory :

Automatic Dip Computation

There are currently two computation algorithms available on GeoFrame BorDip* (MSD and CSB)

MSD : This is a Mean Square Dip computation method . For SHDT* data at any depth level there are 28 possible cross correlograms obtained from 8 button response curves. This method keeps a fixed window on one curve and a moving window on the other.. Then correlograms are obtained by calculating correlation coefficients between the windowed intervals as a function of their relative displacements. A correlation interval length , step distance and search angle have to be defined in this method.

Fig-4: Four pad SHDT* showing Mean Squared (MSD) dip configuration of buttons.

In case of low apparent dips , nearly all data points within the correlation interval are considered for computation of correlation. As the apparent dip increases , less and less points enter into correlation. In areas where high dips or high apparent dips (due to borehole deviation) are expected , an initial displacement can be entered by use of a focusing plane . This focusing plane may be fixed, normal to tool axis or a plane defined from a previous dip calculation. Generally , in 10-15 deg dip regions a correlation length of 1.0m , step of 0.5m and search angle of 60 deg is found to be suitable for MSD dip computations.. The button to button displacements are computed and a minimum mean squared criteria is used to best fit a plane through them. MSD is generally used for structural dip computation.

CSB: This is a vector correlation program which correlates only two curves of each pad (in SHDT*) and does not correlate between pads. This yields 4 dip vectors , one at each pad position. This employs a Continuous Side by Side button correlation and is a variant of the MSD to pick high and variable dip.For a four arm tool 4 side-by-side dips are obtained from the four pairs of adjacent pads..CSB dip is a three- dimensional vector result of two adjacent pad dip vectors.

Fig-5: Four pad SHDT* showing Side - by - Side (CSB) configuration of dip buttons.

The CSB computation makes use of correlations between the two dip buttons on each pad . There will be a great similarity between the microresistivity curves recorded in each pad due to the small horizontal spacing. Each pair of curves is cross correlated using short correlation intervals (0.5m or less ) to produce a vector parallel to the dip plane . A similar vector from an adjacent pad combines to define a dip plane. Generally in high angled cross-bedded sandstones it is suitable to use a Correlation length of 0.4 to 0.5m , step of 0.2 to 0.25m and search angle of 45 to 50deg. Increasing the search angle beyond 60deg often results in finding fewer dips. This is because the search interval increases as the tangent of the search angle. A large search angle will result in a very long search interval which will in turn increase the likelihood of finding false correlations.A default of 60 deg is a good compromise. Very small correlation length , step and angle may result in mirror image dips when the search approaches the true bed dips or if dip magnitude is equal to borehole deviation and their azimuth is 180 deg from borehole drift. The CSB program is responsive to fine bedding structure of the formation, making it particularly effective for defining stratigraphic features (like cross-beds).

Scaling of Electrical Images:

Scaling is a process whereby the microresistivity button current from FMS* or FMI* is transformed to give a measurement resembling the formation conductivity, as measured by a true resistivity tool.

Theory : To make a resistivity measurement , a logging tool creates a voltage between a pair of electrodes and measures a current flowing between them . Provided the current path is controlled (usually with focusing electrodes in laterologs) , the conductance measured (current/applied voltage) can be converted to a measure of the average conductivity or resistivity of the formation through which the current flows. The laterologs (LLD,LLS and SFL) are examples of such measurements which by virtue of their different electrode configurations focus the current path to give the formation resistivity at different depths of investigations.

In contrast , microresistivity buttons on the various imaging tools (FMS/FMI) and the SHDT provide: a) A button current which is not focussed. b) Button current which is not calibrated.

Essentially, the process of calibration or scaling (using FMSCAL or BorScale*on GeoFrame* ) involves : Calculating an average current from the uncalibrated FMS button response and cross-plotting it against an LLS derived theoretical current. Then we best fit the cluster with a piecewise linear equation to get a synthetic resistivity response (SRES) for the FMS measrement. This is mandatory if we want to take up Fracture aperture or sand count studies in the future or running BorTex./SPOT porosity analysis.

Fig-6: Procedure of Scaling an Average Current from FMS/FMI* with an LLS/HLLS or SFL derived theoretical current. The Output is a Synthetic Resistivity curve SRES.SCA (green curve on the right) which matches well with the LLS (blue). This example is from the FMS at Well # Charada-1, Baroda.

Fig-7: A comparison of FMS* raw image (left) and equalised and scaled image(right). The scaled image gives a consistent presentation . Also on the fourth track is a playback of the LLS resistivity curve and the BorScale* output synthetic resistivity curve (SRES). The good agreement was achieved after scaling . Blue and cyan sinusoids and tadpoles are hand-picked dips of fractures and unconformable surfaces. They are clearly visible from scaled images. This example is from Well # Charada-1 , Baroda.