SPWLA 54th Annual Logging Symposium, June 22-26, 2013 A NEW MULTI-FREQUENCY TRIAXIAL ARRAY INDUCTION TOOL FOR ENHANCING EVALUATION OF ANISOTROPIC FORMATIONS AND ITS FIELD TESTING Junsheng Hou, Luis Sanmartin, Dagang Wu, David Torres, and Turker Celepcikay, Halliburton Copyright 2013, held jointly by the Society of Petrophysicists and Well Log Analysts (SPWLA) and the submitting authors. This paper was prepared for presentation at the SPWLA 54th Annual Logging Symposium held in New Orleans, Louisiana, June 22-26, 2013. ABSTRACT Since its introduction to the oil industry in 2000, multicomponent induction (MCI) or triaxial induction logging has been one of the most remarkable developments of both wireline and logging while drilling (LWD) induction operations. Over the years, successful applications have proven its usefulness for characterizing varied types of anisotropic reservoirs, which are frequently overlooked or mistaken by conventional induction tools and their associated processing and interpretation approaches. It is commonly known that, around the world, significant amounts of hydrocarbon reservoirs are located in formations such as laminated sand-shale sequences and fractured/faulted formations that exhibit complicated resistivity/conductivity anisotropy. To help improve the evaluation of anisotropic formations in conventional/unconventional borehole-formation environments, the oil and gas industry has been relying much more on the use of MCI logging. To the authors’ knowledge, currently, very few MCI-type tools have been developed and commercialized. And the previous tools only provide multi-frequency but single-array measurements or multi-array measurements with no detailed information regarding the use of multiple frequencies. To meet the increasing requests in anisotropic formation evaluations, a new MCI-type tool with both multi-array and multi-frequency measurements and the associated data processing algorithms has been developed. This paper first introduces the newly developed MCI tool, with its primary feature being the use of both multiple arrays and frequencies to measure full-tensor data in vertical or deviated wells drilled with oil-based mud (OBM) or in air-filled wells. This new tool has a transmitter triad (collocated triaxial and orthogonal coils) and six sets of receiver coils. The two receivers closest to the transmitter triad are conventional induction coils, while the remaining four sets of coils are built as receiver triads. The tool operates at multiple different frequencies in the range of 12 to 84 kHz by sequentially energizing each of the coils (X, Y, and Z directions) in the transmitter triad and measuring the signals in each of the receiver coils. Hence, each of the four receiver triads measures nine signals per frequency at every logging depth. Secondly, the newly developed, fast data processing algorithm and its software system are presented, which is able to accurately recover formation horizontal resistivity (Rh), vertical resistivity (Rv), dip, and strike/azimuth from the multi-array, multi-frequency full-tensor measurements. This data processing can provide improved evaluation in difficult anisotropic formation environments. In addition to the unconventional resistivity anisotropy, dip, and strike, the tool also provides bed boundaries and conventional array-induction logs resulting from the use of identical spacing as conventional tools. Furthermore, this new processing system contains accurate calibration and temperature correction, data quality evaluation, bed-boundary estimation, horn-effect reduction based on adaptive low-pass filtering, radially one-dimensional (R1D) multistep inversion, borehole correction, and fast vertically one-dimensional (V1D) inversion combined with a zero-D (OD) inversion in homogeneous formations. These features are explained and are validated with synthetic data and with field logs obtained at the service provider’s test wells and in worldwide client wells. Several prototypes of the newly developed MCI tool were built. To date, the new MCI tool has completed its field testing in two test wells and a number of client wells around the world. Since December 2012, it has logged in at least two commercial-job wells. All field-tested and commercial applications show that this new tool can deliver wellmatched multi-array formation Rh, Rv, dip, strike/azimuth, bed boundaries, three sets of conventional induction logs (ACRt-type logs), and invasion information from the MCI processing with multiple subarrays operated at multiple frequencies in real time. This capability greatly enhances anisotropic formation evaluation in both conventional and unconventional resources. 1 SPWLA 54th Annual Logging Symposium, June 22-26, 2013 INTRODUCTION Conventional induction logging has been one of the most used well loggings for almost over half of a century because of its critical role in identifying reservoirs and computing water/oil saturation (Strickland et al., 1987; Barber et al., 1991; Beste et al., 2000; among others). However, the conventional coaxial induction can only provide formation Rh information in vertical wells because of limitations of the tool physics. Because of the sensitivity of MCI measurements to both formation resistivity anisotropy and direction, the MCI logging is capable of providing the additional information of formation Rv, dip, and strike/azimuth for the formation evaluation of conventional and unconventional reservoirs. And MCI logging has been one of the most remarkable and practical developments of wireline and LWD resistivity loggings in recent years. For many years, the oil and gas industry has been requesting this logging service to help in solving for complex formation evaluations, such as thinly laminated shale-sand reservoirs frequently overlooked by conventional induction. To the authors’ knowledge, only a few MCI-type tools for both wireline and LWD logging operations have been developed and commercialized to date. The first wireline MCI-type tool was introduced to the petroleum industry in 2000, and then its updated version was released in 2004 (Kriegshauser et al., 2000; Rabinovich et al., 2005). This tool only contains a single spacing of the x-, y-, and z-direction coils but can be operated at ten different frequencies between 20 and 220 kHz. After removal of the borehole and near-borehole effects in measured signals based on the so-called multi-frequency focusing (MFF), the processing software is finally capable of delivering formation Rh, Rv, dip, and azimuth for one subarray at every logging point. As such, its single-array inverted logs will not allow describing the invaded formation properties if this occurs. In addition, it also needs to be combined with a conventional induction tool for obtaining conventional induction logs with multiple radial depths of investigation. Rosthal et al. (2003) and Barber et al. (2004) present a multi-array triaxial induction tool, which is also used for the wireline operation. This tool consists of a triaxial transmitter, six fully triaxial receiver subarrays, and three conventional axial receiver subarrays. Its spacings include those of the conventional array induction tool, resulting to the identical conventional logs to be produced from the conventional induction measurements. The wellsite software corrects for the borehole effect in the measured data and provides the similar formation parameters of resistivity anisotropy, dip, and azimuth for multiple subarrays plus conventional array induction logs (Wang et al., 2006; Wu et al., 2010). However, it only operates two frequencies of 13 and 26 kHz, and therefore it will be challenging to deliver accurate inverted logs in high-resistivity formation environments. A good review of the latest wireline MCI tools and their applications can be found in Zhang et al. (2007) and Rabinovich et al. (2007). In addition, Bittar et al. (2011) present a new LWD triaxial induction tool with multiple subarrays containing both titled transmitter and receiver sensors that operate multiple frequencies. This tool can provide formation resistivity anisotropy and dip in vertical and deviated wells by data inversion. With significantly increasing amounts of oil and gas reserves located in formations that exhibit complicated electrical anisotropy, the petroleum industry is relying much more on the use of MCI logging for identifying and quantifying different types of anisotropic reservoirs. To meet these applications to enhance anisotropic formation evaluations, a new MCI-type tool and the associated data-processing algorithms (and software system) have been developed. This new MCI tool includes multiple triaxial subarrays with sharing of a common triaxial transmitter. It operates multiple frequencies in the range of 12 to 84 kHz. As such, it can acquire multi-array (R-signal and Xsignal) full-tensor measurements at multiple frequencies. Compared to previous tools, this new tool not only has the capability of acquiring both multi-array and multi-frequency measurements, but also has a wider frequency operation range to target high-resistivity formation evaluation for multiple triaxial subarrays. At the same time, the developed real-time data processing algorithms and the software system are used to perform the borehole correction in measured data and recover the formation Rh, Rv, dip, and azimuth/strike for multiple arrays at multiple frequencies in addition to formation bed positions as well as conventional array induction logs. In the following sections, the tool configuration and its response functions are described. Then, the data process algorithms and the workflow are discussed. Finally, the model-data validation is performed and the field testing results are presented. All model and field examples demonstrate the capability of this new tool to deliver accurate formation Rh, Rv, dip, and strike/azimuth as well as conventional induction logs. The addition of formation anisotropy and direction can help reduce evaluation uncertainty and so provide accurate reserve estimates in complicated anisotropic reservoirs. 2 SPWLA 554th Annual Logging Sympo osium, June 2 22-26, 2013 TOOL L CONFIGUR RATION AND D GEOMETR RICAL FACTO ORS The newly developeed MCI tool em mploys a triad of transmitterr coils (T) and six sets of recceiver coils ( R (1) through T transmitter triad is locatedd at the top of all receiver cooils. The schem matic layout R ( 6) ) spaced along the tool axis. The is pressented in Fig. 1. The transmiitter triad has three t collocatedd and orthogonnal coils ( Tx , Ty , and Tz ) inn the X, Y, and Z directions. The T two sets of receiver coils c ( R (5) andd R (6) ) closest to the transm mitter triad arre the two conven ntional coaxial induction sub barrays, while the remainingg four sets of the receiver ccoils are built aas receiver triads ( R (1) through R ( 4) ). Each recceiver triad is a proprietary aarrangement off six coils: threee orthogonal (X, Y, and m m b nd Rz ) and annother three coiils for a buckinng receiver triaad ( Rx , Rby , Z) coils for a triad of main receivers ( Rx , Rym , an b odel for a threee-triad subarrayy is representedd diagrammatically in the rigght panel of and Rz ). The equivaalent dipole mo Fig. 1. Both the maain and buckin ng coils are co ollocated, and the bucking ccoils are wound to minimizee the direct coupliing signals and d other spuriou us coupling beetween transmiitters and receiivers. Each off the coils in thhe triads is paralleel to the corressponding ones of the other triiads. The tool operates at muultiple frequenccies (e.g., 12, 336, 60, and 84 kH Hz). These freq quencies ensure the tool has the adequate signal level inn the full rangee of the targett formation resistiv vities/conductiivities. In addiition, this new tool uses the identical spaciings to the connventional induuction tool (Xiao et al., 2006), and it also hass all its subarrrays on one sidde, which resuults in a minim mized tool lenggth without sacrifiicing the multip ple targeted rad dial depths of investigation i (ee.g., 10, 20, 300, 60, and 90 innches). Fig. 1 Diagram of thee new MCI too ol configuration and the equivaalent dipole moodel for a three--triad subarray. Here, ( xt , yt , z t ) represents thee tool/measurem ment coordinate system, Lm is tthe spacing betw ween transmitteer and main receeiver triads, and Lb is the spacing between b transmitter and buckin ng receiver triadds. See the text for explanationns of other param meters. The to ool operates by y sequentially energizing eacch of the coilss of three orthogonal directioons (X, Y, andd Z) in the transm mitter triad and d measuring th he signals in each of the recceiver coils. Hence, each onee of the multipple triaxial subarrrays produces nine induction n voltage sign nals per frequeency at every logging depthh. Then, meassured ninecompo onent voltages are collectiveely written as a 3 3 tensorr (or matrix) ffor multiple trriaxial arrays ooperated at multip ple frequencies: V ( i , j ) ( z t ) V IJ( i , j ) ( 3 3 ) , I , J x / X , y / Y , z / Z , i 1, 2, ..., and N , j 1, 2 , ..., and M .................................. (1) (i, j ) (i, j ) ( z t ) reppresents the measured where V m nine--component vooltage tensor aalong logging depth z t , couupling VIJ t J-direction receiver and eexcited by thee I-direction traansmitter for a given i-th denotees the voltage measured by the subarrray operated at the j-th frequeency, N is the total t number off the triaxial arrrays (N = 4 foor the new MCII tool), and 3 SPWLA 54th Annual Logging Symposium, June 22-26, 2013 M is the total number of the operating frequencies (M = 4 for the new MCI tool). After the amplified and digitized process through electronic circuits and firmware, and then calibration and temperature correction (see Appendices A and B for more details), the induction voltages measured on all receivers are converted/calibrated into apparent conductivities with a linear transformation. Also, the apparent conductivities are symbolically expressed as a 3 3 tensor for multiple triaxial arrays operated at multiple frequencies: a(i , j ) (i , j ) (i , j ) (i , j ) xx xy xz (yxi , j ) (yyi , j ) (yzi , j ) IJ(i , j ) (i , j ) (i , j ) zx zy zz(i , j ) (33) , .............................................................................................. (2a) or, another notation often used for the apparent conductivity tensor is a(i , j ) XX (i, j) YX (i, j) ZX (i, j) XY (i, j) YY (i, j) ZY (i, j) XZ (i, j) YZ (i, j) IJ (i, j) ZZ (i, j) ( 33) , ...................................................................................... (2b) where a(i , j ) is referred to as the MCI (R- or X-signal) apparent conductivity tensor in the tool/measurement (i, j) coordinate system ( xt , yt , z t ) and IJ (or IJ Consequently, for example, when I , J = x/X, (yyi, j) (or YY (i, j) ), and when I , J = z/Z, (i, j ) ) are the measured-conductivity couplings of a(i , j ) . IJ(i, j ) is the direct coupling xx(i, j) (or XX (i, j) ), when I , IJ(i, j ) is zz(i , j ) (or (i, j) J = y/Y, IJ is ZZ (i, j) ), which are the traditional multi-array induction measurements. Hence, the total should be the 2M×9 signals (M×9 R-signal and M×9 X-signal data) per triaxial subarray for every log point. However, for two axial subarrays, even though they are capable of acquiring signals of (i, j) (i, j) (i, j) (i, j ) zx , zy , and zz , the tool only acquires zz couplings, so tensor a(i , j ) reduces to a scalar, and therefore (i, j) only 2M signals (M R-signal and M X-signal data) are present per co-axial subarray. Additionally, IJ = (i, j) VIJ(i, j ) K IJ(i, j) , where K IJ(i, j) are the calibration factors (or gain) of the coupling IJ determined by the calibration experiments or tool constants determined by the analytical equations. For the accurate calibration of the new MCI tool, an extension of the calibration method of conventional array induction tools is proposed and implemented (for details, please refer to Appendix A). As stated in the previous discussion, the primary features of this newly developed MCI tool are the use of both multiple arrays and frequencies to measure full-tensor data in vertical or deviated wells filled with air or OBM. Moreover, extensive numerical simulations already showed multi-array and multi-frequency tensor measurements contain the sensitivity to formation anisotropy, dip, and direction. In addition to the induction measurements mentioned, the new MCI tool also acquires the sonde temperatures for its temperature-effect correction. This tool is usually run with a directional package so that true formation dip and strike/azimuth can be found, and it also is run with a multi-arm caliper to find the borehole size and relative position of the tool in the hole. In the previous sections, the tool configuration was presented, which shows the capability of measuring conductivity tensors at multiple subarrays operated at multiple frequencies. Next, the tool’s two important geometrical factors are discussed — vertical geometrical factor (VGF) and integrated radial geometrical factor (IRGF). These factors describe the tool’s vertical resolution and estimate the depth of investigation (DOI), respectively. The VGFs of MCI measurements can be obtained using the following equations for multiple arrays at multiple frequencies: 4 SPWLA 54th Annual Logging Symposium, June 22-26, 2013 (i , j ) VGFIJ(i , j ) ( z, ) g IJ ( z, , , ) dd , I , J x / X , y / Y , z / Z , i 1, 2, ..., and N , j 1, 2 , ..., and M ..... (3a) 0 where g IJ( i , j ) ( z , , , ) is the two-dimensional (2D) Born response function for a three-coil array in cylindrical coordinates at ( , , z ) . It can be determined by simply summing the 2D Born response functions for two 2-coil arrays; the 2D Born response function for a two-coil array can be found in Barber et al. (2004). Theoretically, the result should show a VGF tensor with nine VGFs for the given (i, j) but only VGF xx( i , j ) ( z , ) , VGF yy( i , j ) ( z , ) , and (i , j ) VGF zz( i , j ) ( z , ) are non-zero and VGF xx( i , j ) ( z , ) = VGF yy ( z , ) . Fig. 2 presents the VGFs for a ZZ coupling (top) of six arrays (A1 through A6) and a XX (or YY) coupling (middle) of four triaxial arrays (A1 through A4) at 12 kHz; the background conductivity shown is 1 mS/m. Following Gianzero and Gao (2004), the so-called combo VGFs can be defined by the expression (i , j ) VGF zzxxyy ( z , ) a ( i , j ) VGF zz( i , j ) ( z , ) b ( i , j ) VGF xx( i , j ) ( z , ) c ( i , j ) VGF yy( i , j ) ( z , ) ......................................... (3b) (i , j ) where VGF zzxxyy ( z , ) are referred to as the combo VGFs from the combination of VGF xx( i , j ) ( z , ) , VGF yy( i , j ) ( z , ) , and VGF zz( i , j ) ( z , ) , which are shown in the bottom panel of Fig. 2. The undetermined parameters a (i, j ) , b (i , j ) , and c (i , j ) can be chosen so as to have the optimized vertical characteristics of the combo VGFs. For instance, in Fig. 2, a (i, j ) = 1.5, b (i , j ) = c (i , j ) = -0.5. By comparison, it can be seen that the long-tail effects in the profiles of VGF xx( i , j ) (or (i , j ) VGF yy( i , j ) ) are sharply reduced in the profiles of VGF zzxxyy ( z , ) . Hence, this combination can lead to the enhanced resolution of the combined logs, which offers a reduction of undesirable effects, such as the shoulder-bed effect. The IRGF can be used to evaluate the cumulative contribution of the enclosed measurement volume to the overall measurement and to estimate the DOI for an induction array, which is normally defined to be the depth at which the IRGF equals 0.5. The IRGF is defined as (i , j ) IRGFIJ(i , j ) ( , ) g IJ ( z, , ' , ) dzd ' d ............................................................................................ (4) 0 In the same way as above, we can verify that only IRGF xx( i , j ) ( , ) , IRGF yy( i , j ) ( , ) , and IRGF zz( i , j ) ( , ) are nonzero and IRGF xx( i , j ) ( , ) = IRGF yy( i , j ) ( , ) . Fig. 3 shows the IRGFs of a ZZ coupling (top) for six arrays (A1 through A6) and a XX or YY coupling (middle) for four triaxial arrays (A1 through A4) at 12 kHz; the background (i , j ) conductivity shown is 1 mS/m. The combo IRGFs IRGF zzxxyy ( , ) are presented in the bottom of Fig. 3. From IRGF (i, j ) ( , zz ) of Fig. 3, the DOIs are found for all six ZZ arrays (A1, A2, A3, A4, A5, and A6): 113, 69, 40, 23, 14, and 9 inches. From IRGF xx( i , j ) ( , ) (or IRGF yy( i , j ) ( , ) ) of Fig. 3, the DOIs are found for all XX/YY arrays (A1, A2, A3, and A4): >150, 119, 68, and 40 inches. Note that, because of the large negative contribution near the hole, the DOIs for all four XX/YY arrays are deceptively deeper than the traditional ZZ arrays. From the (i , j ) IRGF zzxxyy ( , ) in Fig. 3, the DOIs are found for all four combo arrays (A1, A2, A3, and A4): 73, 45, 26, and 15 inches. The DOIs of the combo arrays are less than those of the ZZ arrays, but this could be a small disadvantage compared to the greatly enhanced vertical resolution and much reduced shoulder-bed and borehole effects provided by the combined logs. Keep in mind that all DOIs are functions of both frequency and background conductivity. For example, if the operating frequency is higher and the background conductivity is fixed, then the tool’s DOI will become shallower. Hence, an indirect way to correctly estimate the DOI for the XX/YY arrays is available. 5 SPWLA 54th Annual Logging Symposium, June 22-26, 2013 Fig. 2 VGFs for six ZZ-coupling arrays (A1 through A6) in the top, four XX or YY couplings (middle), and their combo (bottom) arrays (A1 through A4) at 12 kHz; the background conductivity is 1 mS/m. The undesirable complicated responses near the origin are clearly observed in the VGF profiles of the four XX- or YY-coupling arrays. Fig. 3 IRGFs for six ZZ-coupling arrays (A1 through A6) in the top, four XX or YY couplings (middle), and their combo (bottom) arrays (A1 through A4) at 12 kHz; the background conductivity is 1 mS/m. DATA PROCESSING AND WORKFLOW It is commonly known that the primary objective of MCI logging is fast and accurate delivery of formation anisotropy, dip, and azimuth/strike from multi-array tensor measurements. Because of the complication of measured data, as mentioned in the literature, they cannot be interpreted using “eyeball” inspection. Therefore, the associated data processing algorithm and related software system must be developed to obtain the above mentioned formation properties requested for evaluation of complex formations. The workflow for the newly developed data processing algorithms and software is shown in Fig. 4. It generally consists of two parts—a down-hole part and an up-hole part. The MCI down-hole tool acquires the voltage signals from all triaxial subarrays and the data from all of the conventional short-spacing axial arrays at the multiple operating frequencies. These acquired data are digitized and sent back to the up-hole part as its inputs. The advanced-processing software in the surface system removes noise, stacks, and samples in depth for depth alignment, compensates for temperature effects on the sonde and electronics, and converts the measured voltages that miss the unit into apparent conductivity signals using the so-called calibration. Then, the data are processed to produce formation anisotropy, relative dip, azimuth, borehole-effect corrected (BHC) logs as well as the bed boundaries, the conventional array induction logs, and invasion parameters. All of these answer products can be available in real-time, and they are also processed for speed correction. Next, 6 SPWLA 54th Annual Logging Symposium, June 22-26, 2013 some important parts in the whole processing system are introduced. ZZ-Coupling Processing. The MCI ZZ-coupling data measured on the z-directional receivers when the z-directional transmitter is fired, in conjunction with the data from the short arrays, are processed using the conventional array induction processing algorithms to obtain the same logs as the conventional array tools. The ZZ-coupling data processing includes skin-effect correction, borehole correction, 2D software focusing, and ZZ-R1D inversion to produce three sets of conventional resistivity logs in addition to invasion parameters (depth of invasion and flushedzone [Rxo] and virgin-zone resistivity [Rt]). For the detailed algorithms and application results, see Xiao et al. (2006). Adaptive Low-Pass Filtering. For the purpose of adaptive low-pass filtering, the Kaiser window is used as the lowpass filter function because it is a nearly optimal window function. The Kaiser window is defined by the following equation: K w I0 ( , m ) 1 ( 2m 1) 2 M I 0 ( ) ,0 m M ; K w ( , m ) 0 for other cases ....................................... (5) where I0 is the zero-order modified Bessel function of the first kind, parameter α is an arbitrary real number that determines the shape of the low-pass window, and the integer M is the length of the window. The larger the value of |α|, the narrower the window becomes. Conversely, for larger |α|, the width of the main lobe increases in its Fourier transform, while the side lobes decrease in amplitude. Thus, this parameter controls the tradeoff between main-lobe width and side-lobe area. For a large α, its shape tends toward a Gaussian window. For a given M, the Kaiser window is totally defined by the parameter α. Hence, for the purpose of reducing high-frequency noise, α is determined based on the data noise level (or uncertainty); for the purpose of reducing the horn effect in some couplings of the MCI tensor, α is determined based on both the data uncertainty and the distance between the current logging point and the bed boundaries. Bed-Boundary Estimation. Numerical simulations show that the coefficients of variation (CV) of different combined MCI signals, such as 3ZZ-(XX+YY)/2, 2ZZ-XX, or 2ZZ-YY, have good sensitivity to bed boundaries, so they can be used to estimate the position of boundaries. First, the CVs of the selected combined MCI signals are computed within a predetermined depth window (here, the window length should be great or equal to the vertical resolution of the MCI tool). The CV is defined as a function of log depth, and the depth of the window center is assigned to the computed CVs. Then, a maxima larger than a predetermined threshold value is searched for, and tentative boundaries are placed at these depths. If the thickness of two boundaries is less than the vertical resolution of the tool, then an average of the two boundaries is used as the new boundary and the two original ones are removed. R1D Inversion and BHC. The R1D inversion is based on a R1D model implemented by splitting the inversion problem of one high-dimension unknown vector into several lower-dimension ones, based on their sensitivity to different components of measured conductivity tensors for different subarrays operated at different frequencies. This reduction in dimensionality makes the non-linear inversion considerably faster, more reliable, and robust. Another critical element for this accurate and fast inversion is the use of the prebuilt lookup table/library as the forward engine to replace the time-consuming three-dimensional (3D) simulation in the inversion. To remove the MCI borehole effects, the conductivity tensor of one of the triads, usually the shorter arrays (e.g., A3 or A4), must be first used to estimate the formation-, borehole-, and tool-position parameters (Rh, Rv or Rvh, borehole size, tool eccentricity and its azimuth, dip, and azimuth/strike) by using the R1D inversion. This inversion assumes no invasion and no shoulder-bed effects. It uses a prebuilt library (or BHC library) that includes the response of the tool to the various formation-borehole parameters. After these parameters are obtained, the algorithm calculates and removes the borehole effects for all triads and for all frequencies. More details on the algorithms of the R1D inversion and MCI BHC are described in Hou et al. (2012). V1D Inversion. The V1D inversion is also a model-based inversion processing method. It assumes the formation is 7 S SPWLA 54th Annual A Loggin ng Symposium m, June 22-26,, 2013 oof a vertically transverse-iso otropic (TI) an nisotropic layeered structure. The newly ddeveloped V1D D inversion is a rrigorous and faast approach fo or determining the horizontal and vertical reesistivities, bedd-boundary poositions, dip, annd aazimuth/strike angles from BH HC MCI log data. d In its firstt step, the azim muth/strike anggle is determineed by rotating to aan equivalent formation f mod del with a zerro azimuth/striike angle. Theen, a variance//CV-based meethod is used to eestimate the initial position of o the bed boun ndaries to be used u in the invversion. Next, an OD inversion algorithm is ssolved in an infinitely homog geneous mediu um to obtain a good initial guuess model to bbe used as inittial inputs in thhe V V1D layered in nversion algorrithm. Finally, the V1D layeered inversionn problem is trransformed intto a constraineed nnonlinear miniimization prob blem that can be solved ussing the regullarized Levenb nberg-Marquarddt minimizatioon m method (Van den d Bos, 2007)). To speed up the V1D layeered inversion, the Jacobian/ssensitivity mattrix is computeed uusing forward differences on nly once at thee beginning off the inversionn process. Durring subsequennt iterations, thhe JJacobian matrix x is updated by y the Broyden’’s method. Forr real logging iinversion proceessing, a large number of bedds m must be inverted, which will result in a laarge V1D layeered inversion problem withh many unknowns. Too manny uunknowns in an a inversion model m will inccrease computaation complexxity and deteriiorate inversioon accuracy annd eefficiency. To avoid this, a so-called layeer-sliding inveersion method is introducedd: the original large inversioon pproblem is equ uivalently and efficiently so olved by subseequently inverrting a set of small V1D laayered inversioon pproblems and combining c theiir results simu ultaneously. Mo oreover, the paarallel-program mming implem mentations of thhe tiime-consuming operations further f lead to o the reduced computation ttime in the enntire inversion.. Its applicatioon rresults of synth hetic data can be b found in Hou u et al. (2012). F Fig. 4 New MC CI data processin ng workflow in ncluding down-hole and up-hoole parts (modiffied from Hou eet al., 2012). Thhe ddown-hole part is for the dataa acquisition, an nd the up-hole part is for deliivering the ansswer products thhrough data prrepprocessing and advanced a proceessing. B Before its appllication to field d data, the dev veloped data processing p worrkflow associat ated with softw ware system waas tested on a nu umber of MCI synthetic datta sets, which were generateed by using a fast and accuurate 3D finitteddifference (FD)) forward apprroach developeed by Hou et al. (2011). Fig. 5 shows one bborehole-formaation model thhat w was used to tesst the processin ng workflow an nd its software system describbed above. Thhis testing model consisted off a 550-degree deviiated borehole filled with OBM surroundeed by a nine-llayer TI formaation. This moodel assumes nno innvasion of borrehole mud fluiid into the form mation. The borrehole diameteer was 8 inchess, and the mud resistivity (Rm m) w was 1000 ohm--m (or OBM). The MCI tool logs through th he hole at the eeccentricity of 10% and an ecccentricity anggle oof 30 degrees. As this was a fully 3D mod del, the synthettic logs only ccould be generaated using a fuully 3D forwarrd ccode. The synthetic logs werre assumed to be b acquired with w the new M MCI tool (multiiple operating frequencies annd ffour triaxial an nd two axial arrrays). All otheer remaining model m parameteers are shown in Fig. 5. Thee simulated MC CI loog data were contaminated c with w a 5% pseu udo-random errror. All contam minated syntheetic log data weere the inputs oof thhe data processing system fo or the adaptive low-pass filterring, horn reduuction, verticall-resolution enhhancement, beddbboundary detecction, R1D inveersion, BHC, and a other proceessing. F Fig. 6 presentss the recovered d model param meters of Rh, Rv, dip, and strike logs forr four triaxial subarrays from innversion of th he contaminated MCI data att 12 kHz. For the purpose oof comparison with the recovvered formatioon 8 8 SPWLA 554th Annual Logging Sympo osium, June 2 22-26, 2013 param meters, the correesponding truee model param meters are also sshown in Fig. 6. The true annd recovered R Rh, Rv, dip, and strrike logs are presented in Traacks 1 through h 4 (from left too right). The trrue model paraameters are shoown by the dashed d-dot lines in magenta, m whilee the inverted ones o for four trriaxial subarrayys (A1, A2, A33, and A4) aree plotted by the solid lines in diffferent colors (b blue, green, read, and cyan). After comparrisons with the true ones, it w was noticed he true and inv verted strike, Rh, R and Rv as well as the truue and recoverred dip, exceppt the isotropicc formation that th section ns, agree very y well. After the comparisson among thhe inverted reesults for diffferent frequencies, good consisstency was alsso found. The good agreem ment between the inverted aand true form mation parametters clearly demon nstrates that th his R1D inverssion algorithm m performed veery well, evenn for the contaaminated data with a 5% random m noise level. This confirmss the validity and a robustnesss of the R1D iinversion approoach. On the oother hand, one sh hould be awaree that accurate recovered r dip results r of the R R1D inversion are only possibble in formatioon intervals with reesistivity aniso otropy as there is no sensitivitty of the MCI m measurements to dip in isotroopic formationns. Fig. 5 Schematic diaagram of a fully y 3D model used for the data--processing validation. This 3D D model consissts of a 50hole filled with OBM O and surro ounded by a horrizontal nine-layyer formation w without invasion.. Each layer degreee deviated boreh is charracterized by Rh h, anisotropic rattio (Rvh = Rv/R Rh), and top andd bottom layer-bboundary locatioons. Layers 1, 3, 5, 7, and 9 are TI and have the saame Rh and Rvh h of 2 ohm-m an nd 5; the remainning layers are isotropic and havve the same Rh of 20 ohmol logs through an 8-inch boreehole and is deccentralized insidde the hole witth eccentricity oof 10% and m. Thee new MCI too eccentrricity angle of 30 degrees. Heree, ( x f , y f , z f ) designates the fformation beddding-plane coorddinate system. Fig. 6 Recovered Rh, Rv, dip, and sttrike for four triiaxial subarrayss (A1, A2, A3, aand A4) from pprocessing of coontaminated MCI data d with a 5% error at 12 kHzz are presented.. They are show wn with solid liines in Tracks 1, 2, 3, and 4 (ffrom left to right), respectively. For F comparison purposes, p the tru ue model param meters are show wn as magenta daashed lines for every track. t Rh and Rv logs l are plotted in a log scale, and the dip andd strike logs are plotted in a linnear scale. Noticce that good Also, the consisttency exists amo ong the four-array results, but th he recovered Rhh and strike are more accurate. 9 SPWLA 54th Annual Logging Symposium, June 22-26, 2013 After the R1D inversion discussed previously, the BHC is performed to the synthetic MCI logs at multiple frequencies. Fig. 7 shows the BHC-corrected MCI nine-component logs of six subarrays at a frequency of 12 kHz for the model shown in Fig. 5. The true formation horizontal and vertical conductivities (Ch and Cv) are only included in this figure for reference. For the validation purpose of the corrected logs, the MCI logs are also computed without the borehole effect in the same layered formation without a hole, as shown in Fig. 5 with an EM semi-analytical solution in V1D layered TI formations (Zhong et al., 2006). The computed results are shown as the dashed lines in Fig. 7. By comparison, it can be observed that all corrected components have a good match with the corresponding simulated V1D results, with exception of the logging intervals around the boundaries. These discrepancies could be primarily caused by the BHC model, which does not account for shoulder-bed effects. Finally, the BHC data are used as inputs for the ensuing further processing, such as for the V1D inversion and ZZ array processing without BHC for producing ACRt-type logs described previously. Fig. 7 BHC-corrected MCI nine-component logs of six arrays (A1 through A6) at a frequency of 12 kHz for the synthetic logs of the 3D model shown in Fig. 5. The BHC-corrected logs are shown as solid lines in different colors (blue, red, magenta, black, cyan, and yellow). The synthetic logs produced by a V1D code are plotted in the dashed lines. Notice that good agreement exists between the corrected and V1D logs. Here, Ch and Cv denote the true formation horizontal and vertical conductivity shown as the green and red dot-dashed lines for every panel. Only the ZZ couplings are plotted in a log scale. FIELD-TESTING EXAMPLE The new MCI tool was designed and several prototypes of the new tool were built. To date, the new MCI tool has completed its field testing in two test wells and a number of client wells. Since December 2012, it has logged in at least two commercial-job wells. All application examples demonstrate that the targeted tool capability to provide accurate formation anisotropy (Rh and Rv) and direction information for enhancing complicated formation evaluations has been achieved. This evaluation involved different borehole-formation environments, such as low- and high-resistivity formations, large and small holes, and low- and high-dip formations. Because of the limitation of acquired data release from the operators, only one example is presented to illustrate the tool’s field testing applications. The application results in one test well from Texas can be found in Hou et al. (2012). The MCI data for this example was acquired in a well from South America drilled using OBM and an 8.5-inch bit. This well’s deviation angle was between 40 to 50 degrees, and its azimuth angle was between 150 and 170 degrees. Aside from the MCI induction data for multiple arrays and multiple frequencies, the borehole imaging data, the directional data, the six-arm caliper data, neutron, and density data were also acquired and were available for this same well. The borehole imaging data were interpreted to determine the formation dip and azimuth angles. The directional data were used to determine the final true/structural dip, azimuth angles of the MCI, and imaging relative angles. Fig. 8 shows the measurements of three direct couplings (ZZ, XX, and YY) and the combo (XX+YY)/2 of XX and YY of three subarrays of A2, A3, and A4 for a 500-ft interval of this well and their comparisons between the main (M) and repeat (R) runs. As expected, for longer spacing arrays in the OBM well, the ZZ-coupling data from both 10 SPWLA 54th Annual Logging Symposium, June 22-26, 2013 acquisitions perfectly overlay, but the virtual differences of the two horizontal direct couplings of XX and YY can be observed for three arrays between the two runs because of their sensitivities to the tool position and direction in the hole, especially for short array A4. Because of the function of the combo (XX+YY)/2 of XX and YY to reduce the tool-position effect, perfect agreement is observed for the combo (XX+YY)/2 data for the two runs, as shown in the rightmost track of Fig. 8. All shown good data repeatability indicates the good data quality of the MCI measurements. In summary, both ZZ and the combo signals can be used to evaluate the data quality by the comparisons between the M and R runs. Following the evaluation of raw data quality, the real-time data processing software previously discussed is run to produce the conventional array induction logs (or ACRt-type logs) for the main run over the same interval as Fig. 8 and to recover formation Rh, Rv, dip, strike/azimuth logs, and bed boundaries from the measured MCI data for four triaxial subarrays (A1, A2, A3, and A4) at multiple frequencies (e.g., 12 and 36 kHz). Fig. 9 shows the three sets of conventional array induction logs and Rxo, Rt, and depth of invasion logs from the ZZ R1D inversion. These logs show very little resistivity changes over the whole interval, and the resistivity values are between 1and 5 ohm-m in most parts. All resistivity logs stack together, which shows there almost is no invasion in the profile. Fig. 10 shows the recovered four-array formation Rh, Rv, dip, and strike/azimuth logs from the 12-kHz MCI measurement plus the two conventional logs (R10 and R90) for the purpose of comparison with the inverted Rh logs. The Rh, Rv, dip, and strike/azimuth logs are plotted in Tracks 1, 2, 3, and 4 (from left to right) in Fig. 10. It can be first noticed that the four-array inverted results are fairly consistent with each other. The inverted Rh logs correlate fairly well with the ZZ measurements, and the conventional array induction logs, such as R10 and R90. As expected, the Rh logs are less than or equal to the conventional logs, especially in high-dip sections (e.g., the zone between 050 and 250 ft) or low-dip ones, respectively. However, the inverted Rv logs clearly indicate there is formation resistivity anisotropy for the whole interval, and these Rv logs show relatively obvious resistivity change. Moreover, good correlation can be observed between the inverted Rv and XX/YY logs shown in Fig. 8. So both Rh and Rv can provide more accurate Rsand (sand resistivity) and saturation. Multi-array inverted logs can further reduce their uncertainty. The recovered dip logs present relative dips that vary from approximately 30 to 60 degrees and recovered azimuths that vary between 100 to 180 degrees. Again, good correlation can be observed for the recovered dips and azimuths from MCI and direction data as well as those of borehole imaging data processing. In addition, for the validation of the recovered Rh, Rv, dip, and azimuth, the inverted results from multi-frequency measured data and different runs (M and R) were compared, and good consistency was found among them. Fig. 8 Comparison of three direct couplings (ZZ, XX, and YY) and the combo (XX+YY)/2 of XX and YY between the M and R runs for a 500-ft interval between 000 to 500 ft in a well from South America. Tracks 1, 2, and 3 (from left to right) present the comparisons of the three direct couplings (ZZ, XX, and YY) for three triaxial subarrays of A2, A3, and A4 at 12 kHz. The rightmost track shows the comparisons of the combo (XX+YY)/2 of XX and YY over the same interval. Also, A2M, A3M, and A4M denote the M run measurements for 3 arrays of A2, A3, and A4, and A2R, A3R, and A4R denote the R run measurements for the same three arrays. 11 SPWLA 54th Annual Logging Symposium, June 22-26, 2013 Fig. 9 Conventional array induction logs and ZZ-R1D inversion logs are presented for the M run over the same interval as Fig. 8. Here, RO, RT, and RF shown in the three leftmost tracks are the 1-, 2-, and 4-ft array induction logs with the DOIs of 90, 60, 30, 20, and 10 inches. Rxo and Rt logs from the ZZ R1D inversion are shown in Track 2. The rightmost track shows the results of the inverted depth of invasion. Fig. 10 Four sets of recovered formation Rh, Rv, dip, and strike/azimuth logs for four triaxial subarrays (A1, A2, A3, and A4) are presented for the M run over the same interval as Fig. 8. They are plotted in Tracks 1, 2, 3, and 4 (from left to right) and represented by the solid lines. In every track, the recovered results for four arrays are given by the blue, green, read, and cyan colors. Here, in the leftmost track, R10 and R90, shown by the magenta and yellow dashed lines, are the two 2-ft conventional array induction logs (or ACRt-type logs) with the DOIs of 10 and 90 inches, respectively. Notice that the fourarray inverted parameters show good consistency with each other. CONCLUSIONS The new MCI tool targeting OBM-well operations has been successfully developed and field-tested. This new tool is a multi-array triaxial induction tool. It operates multiple frequencies in the range of 12 to 84 kHz. The associated fast and accurate data processing algorithms and their software system also have been developed and implemented. The accurate and fast data processing system can recover answer products of formation Rh and Rv, dip, and strike/azimuth, along with the bed-boundary position, the three sets of conventional array induction logs (ACRt-type logs), and the invasion parameters. To date, the new MCI tool has completed field testing in two test wells and a number of client wells. And it has logged in at least two commercial wells since December 2012. The field-testing and commercial application results, 12 SPWLA 54th Annual Logging Symposium, June 22-26, 2013 and the related model data tests, demonstrate the capability of this new tool to deliver accurate multi-array formation Rh and Rv, dip, and strike/azimuth, as well as the bed-boundary position and the conventional array induction resistivity logs, which can provide enhanced reserve estimates in complicated anisotropic reservoirs, such as lowresistivity or fractured/faulted formations. NOMENCLATURE MCI LWD OBM BHC Rh Rv Rvh Dip Strike Ch Cv R1D OD 1D V1D 3D Multicomponent induction Logging while drilling Oil-based mud Borehole-effect correction Horizontal resistivity Vertical resistivity Anisotropic ratio (Rv/Rh) Formation relative dip Formation strike/azimuth Horizontal conductivity Vertical conductivity Radially one-dimensional Zero-dimensional One-dimensional Vertically one-dimensional Three-dimensional ACKNOWLEDEGEMENTS The authors thank the oil companies for consent to release their field data. The authors also thank Halliburton for permission to publish this paper. Special thanks go to EM-physics colleagues Drs. Burkay Donderici, Turker Celepcikay, Baris Guner, and Yumei Tang for their contributions related to the issues of tool calibration, MCI library, as well as V1D inversion, and also John Purba and James Wang of Halliburton for providing valuable support concerning field data processing and real-time software development. REFERENCES Barber, T., Anderson, B., Abubakar, A., Broussard, T., Chen, K., Davydycheva, S., Druskin, V., Habashy, T., Homan, D., Minerbo, G., Rosthal, R., Schlein, R., and Wang, H., 2004, Determining formation resistivity anisotropy in the presence of invasion: Paper SPE 90526 presented at the SPE 80th Annual Technical Conference and Exhibition, Houston, Texas, USA, 26–29 September. Barber, T. and Rosthal, R., 1991, Using a multiarray induction tool to achieve high-resolution logs with minimum environmental effects: Paper SPE 22725 presented at the SPE 66th Annual Technical Conference and Exhibition, Dallas, Texas, USA, 6–9 October. Beste, R., Hagiwara, T., King, G., Strickland, R., and Merchant, G., 2000, A new high resolution array induction tool: Paper presented at the SPWLA 41st Annual Logging Symposium, Dallas, Texas, USA, 4–7 June. Bittar, M., Althoff, G., Beste, R., Li, S., and Wu, H., 2011, Field testing of a new LWD triaxial sensor for anisotropy and dip measurement in vertical and deviated wells: Paper presented at the SPWLA 52nd Annual Logging Symposium, Colorado Springs, Colorado, USA, 14–18 May. Gianzero, S. and Gao, L., 2004, Method of combing vertical and horizontal magnetic dipole induction logs for reduced shoulder and borehole effects: US Patent No. 6,819,112 B2. Hou, J., Bittar, M., Wu, D., Sanmartin, L., and Guner, B., 2011, New scattered potential finite-difference method 13 SPWLA 54th Annual Logging Symposium, June 22-26, 2013 with anisotropic background to simulate multicomponent induction logs: Paper presented at PIERS 2011, Suzhou, China, 12–16 September. Hou, J., Sanmartin, L., Wu, D., Celepcikay, T., and Torres, D., 2012, Real-time borehole correction for a new multicomponent array induction logging tool in OBM wells: Paper presented at the SPWLA 53rd Annual Logging Symposium, Cartagena, Colombia, June 16–20. Kriegshauser, B., Fanini, O., Forgang, S., Itskovich, G., Rabinovich, M., Tabarovsky, L., Yu, L., Epov, M., and Horst, J.v.d., 2000, A new multicomponent induction logging tool to resolve anisotropic formations: Paper presented at the SPWLA 41st Annual Logging Symposium, Dallas, Texas, USA, 4–7 June. Rabinovich, M., Tabarovsky, L., Corley, B., and Horst, J., 2005, Processing multicomponent induction data for formation dip and azimuth in anisotropic formations: Paper presented at the SPWLA 46th Annual Logging Symposium, New Orleans, Louisiana, USA, 26–29 June. Rabinovich, M., Gonfalini, M., Rocque, T., Corley, B., Georgi, D., Tabarovsky, L., and Epov, M., 2007, Multicomponent induction logging: 10 years after: Paper presented at the SPWLA 48th Annual Logging Symposium, Austin, Texas, USA, 3–6 June. Rosthal, R., Barber, T., Bonner, S., Chen, K.C., Davydycheva, S., and Hazen, G., 2003, Field test results of an experimental fully triaxial induction tool: Paper presented at the SPWLA 44th Annual Logging Symposium, Galveston,. Texas, USA, 22–25 June. Strickland, R., Sinclair, P., Harber, J., and DeBrecht, J., 1987, Introduction to the high resolution induction tool: Paper presented at the SPWLA 28th Annual Logging Symposium, London, England, 29 June–2 July. Van den Bos, A., 2007, Parameter estimation for scientists and engineers: Wiley-Interscience. Wang, H., Barber, T., Morriss, C., Rosthal, R., Hayden, R., and Markley, M., 2006, Determining anisotropic formation resistivity at any relative dip using a multiarray triaxial induction tool: Paper SPE 103113 presented at the SPE 82nd Annual Technical Conference and Exhibition, San Antonio, Texas, USA, 24–27 September. Wu, P., Wang, G.L., and Barber, T., 2010, Efficient hierarchical processing and interpretation of triaxial induction data in formations with changing dip: Paper SPE 135442 presented at the SPE 86th Annual Technical Conference and Exhibition, Florence, Italy, 19–22 September. Xiao, J., Buchanan, J., Bittar, M., Davis, E., Sanmartin, L., Hu, G., Zannoni, S., Morys, M., and Liu, W., 2006, A new asymmetrical array induction logging tool: Paper SPE 101930 presented at the SPE 82nd Annual Technical Conference and Exhibition, San Antonio, Texas, USA, 24–27 September. Zhang, Z., Akinsanmi, O., Ha, K., Bourgeois, T., Jock, S., Blumhagen, C., and Stromberg, S., 2007, Triaxial induction logging – an operator’s perspective: Paper presented at the SPWLA 48th Annual Logging Symposium, Austin, Texas, USA, 3–6 June. Zhong, L., Shen, L., Li, S., Liu, R., Bittar, M., and Hu, G., 2006, Simulation of tri-axial induction logging tools in layered anisotropic dipping formations: Paper presented at the SEG International Exhibition and 76th Annual Meeting, New Orleans, Louisiana, USA, 1–6 October. ABOUT THE AUTHORS Junsheng Hou received his BS and MS degrees in applied geophysics from Wuhan College of Geosciences, China, in 1984 and China University of Geosciences at Beijing (CUGB) in 1988, respectively. He obtained his PhD degree in geophysical/geological exploration from Northeastern University, China, in 1993. Dr. Hou began his career as a field geophysicist with Chinese Henan Geophysical Exploration in 1984. Junsheng joined Halliburton in 2005 and currently is working as a scientific advisor in Halliburton’s Houston technology center. Dr. Hou has published extensively on the 14 SPWLA 54th Annual Logging Symposium, June 22-26, 2013 topics of resistivity logging, formation evaluation, and surface geophysics. His current research focuses primarily on 3D numerical simulation, data processing, inversion, log interpretation, and tool development of multicomponent induction logging. Email: Junsheng.Hou@Halliburton.com Luis Sanmartin is a senior technical professional leader in the sensor physics group at Halliburton in Houston. He received his PhD degree in physics from the University of Illinois at Urbana-Champaign in 1998 and joined Halliburton after graduation. His primary interest is in design and development of electromagnetic instruments. Dr. Sanmartin is a member of SPWLA, IEEE, and SPE. Email: Luis.Sanmartin@Halliburton.com Dagang Wu is a principal scientist at Halliburton in Houston. He received his PhD degree in electrical engineering from the University of Houston in 2006. His current interest is in development and processing of electromagnetic/resistivity logging tools. Dr. Wu is a member of SPWLA. Email: Dagang.Wu@Halliburton.com David Torres is the project manager for the multicomponent induction tool and is responsible for overseeing the development, testing, and marketing of the tool. For more than 20 years, Torres has worked in various management and R&D positions in the oil industry. He began his career in 1980 with Gearhart Industries in Brazil. He moved to Texas and held several positions within the organization. He holds several patents and has authored various papers in signal and image processing. Torres holds an EE degree and a MSEE degree from the University of Madison, Wisconsin. Email: David.Torres@Halliburton.com F. Turker Celepcikay holds a BS degree in electrical and electronics engineering from Bilkent University, Ankara, Turkey (2005) and a PhD degree in electrical engineering from the University of Houston, USA (2010). Celepcikay has worked as a senior scientist in the formation evaluation group in Halliburton Energy Services in Houston. His current interest is in development and improvement of electromagnetic/resistivity logging tools. APPENDIX A: CALIBRATION OF NEW MCI TOOL The calibration method of the new MCI tool is similar to the calibration routine of conventional array induction tools, such as ACRt tool. The important difference is that the new MCI tool has three transmitter orientations and three receiver orientations that give rise to nine-coupling signals for each triaxial subarray at each frequency. In the MCI calibration, the seven components of XX, XZ, YY, YZ, ZX, ZY, and ZZ are first calibrated. Then, from the measurements of the seven experimental values, the gains for the remaining two components (XY and YX) can be derived. After the gains have been found, the offsets can be evaluated by lifting the tool from the surface to a height of 20 ft, in a so- called “air-hang” experiment. The offsets are evaluated as: offset S air _ hang G ...........................................................................................................................................(A-1) where offset is the offset, G is the gain, and S air _ hang is the signal received at the “air-hang” step. Finally, with gains and offsets evaluated, the apparent conductivities for every component at every frequency are obtained from the following equation: a S G offset ...............................................................................................................................................(A-2) or a ( S S air _ hang ) G .......................................................................................................................................(A-3) 15 SPWLA 54th Annual Logging Symposium, June 22-26, 2013 where a is the calibrated apparent conductivity, S is the received signal, and offset is the additive constant, also called the sonde error. On a rigorous calibration formulation, the calibration accuracy depends on the compensation for the temperature effect, the earth ground effect, and the suppression of the random noise. For details of how to incorporate these corrections into a rigorous calibration scheme, see Xiao et al. (2006). APPENDIX B: TEMPERATURE CORRECTION OF NEW MCI TOOL A typical logging tool consists of a sonde section and an electronics section. Temperature causes performance drift for the electronic section and produces response changes for the sonde. The sonde response change with temperature is referred to as the temperature effect. The temperature effect is a function of the temperature and temperature distribution inside the sonde. The temperature effect comes from thermal expansion of the sonde and the physical property variation of the sonde materials with temperature. Modern electronics can self-compensate the performance drift by periodic performance of self-calibration. In this appendix, methods are described for effectively monitoring the temperature inside the sonde and for compensating for the sonde response change. The temperature effect of induction tools primarily comes from three sources: (1) the conductivity of the feed pipe that shields the wires, which varies as temperature changes; (2) thermal expansion, which changes the distances between coils, the radii of the coils, and the radius of the feed pipe; and (3) the resistance of the coils, which varies as temperature changes. The coil radius expansion is controlled by the temperature at the coils. The temperature effect from the feed pipe depends on the temperature of the feed pipe. The axial expansion is even more difficult to evaluate. This heterogeneous temperature distribution complicates the temperature correction with two questions: (1) how to effectively measure the temperature and temperature distribution of the tool body, and (2) how to accurately correct for the temperature effect of multiple sources. These questions were answered by developing a thermodynamic analysis of the heat flow within the tool for the conventional array induction tools. The same analysis is applicable to the new MCI tool. The analysis shows that two temperature sensors are sufficient to achieve accurate correction. The two sensors are placed spaced apart along the pipe. The temperature effect of the sonde is evaluated in an oven, and the drift is fit to a cubic polynomial approximation in three variables: temperature of Sensor A, time derivative of temperature in Sensor A, and temperature difference between Sensors A and B. The precise expression of the cubic polynomial is: T a0 a1 (TA ) a 2 (TA ) 2 a3 (TA ) 3 b1 (TAB ) b3 (TAB ) 3 ............................ (B-1) T T c1 ( A ) c3 ( A ) 3 , t t where the coefficients a 0 , a1 , a 2 , a 3 , b1, b3 , and c1, c3 are obtained by optimal fitting to the heat run test of each of the measurements of the MCI tool. For more details on the temperature correction, see Xiao et al. (2006). 16