PETROLEUM EXPLORATION AND DEVELOPMENT

Volume 39, Issue 5, October 2012

Online English edition of the Chinese language journal

Cite this article as: PETROL. EXPLOR. DEVELOP., 2012, 39(5): 591–596. RESEARCH PAPER

Fluid identification method based on 2D

diffusion-relaxation nuclear magnetic resonance (NMR)
HU Falong1,2,*, ZHOU Cancan1,2, LI Chaoliu1,2, XU Hongjun1,2, ZHOU Fengming3, SI Zhaowei3
1. PetroChina Research Institute of Petroleum Exploration and Development, Beijing 100083, China;
2. PetroChina Research Laboratory of Well Logging, Beijing 100083, China;
3. PetroChina Jidong Oilfield Company, Tangshan 063004, China

Abstract: Based on current acquisition modes of MRIL-Prime NMR logging tool, 2D NMR signals could be obtained by the combina-
tion of logging data from different modes, then the fluid properties in complicated reservoirs could be distinguished by 2D diffu-
sion-relaxation NMR logging data distribution of pore fluids, generated by multi-echotrain joint inversion. In comparison with 1D NMR
logging, this method could increase fluid information in diffusion regime, separate oil, gas and water signals in 2D space and enhance the
identification capacity of fluid properties from NMR logging. The 2D NMR logging in the multi-echowave interval was applied in the oil
pays in Well A and the water layers in Well B in the Nanpu Sag by MRIL-Prime tool, and the interpretation matches the well testing re-
sult. It indicates that 2D NMR logging has advantages on the identification of light oil, and fluids in macropore reservoirs than 1D NMR
logging.

Key words: 2D NMR logging; acquisition mode; joint multi-echotrain inversion; diffusion coefficient

Introduction nessed large-scale application in foreign oil fields, but domes-

With the continuous development of the nuclear magnetic
resonance (NMR) techniques, NMR logging has become one
of the important geophysical logging methods, and plays an
important role in fluid typing, pore structure evaluation,
bound water saturation calculation and so on[1]. Because of the
T2 spectrum overlap of oil, gas and water in the pore of reser-
voir, some fluid typing methods proposed previously, such as
differential spectrum method (DSM), shifted spectrum
method (SSM), time domain analysis (TDA), diffusion analy-
sis (DIFAN) and enhanced diffusion method (EDM), need to
meet some certain prerequisites, which restrict these methods
in practical applications[2–5]. Sun and Hürlimann successfully
applied two-dimensional (2D) NMR technique in spectros-
copy to the logging field, which broadened the application
range of NMR logging [6,7], and achieved good results in pore
fluid property identification. Based on this method, Schlum-
berger and Baker Hughes have designed and manufactured
advanced NMR logging tools, and designed the corresponding
2D NMR logging acquisition modes and data processing
methods according to the tools’ characteristics respectively.
This technique solved the problem of 1D NMR logging signal
overlapping in the oil-water layer identification. It has wit-
tic research has just begun. Xie Ranhong and Xiao Lizhi con-
ducted some researches on 2D NMR numerical simulation
and experiments, and verified the validity of the method [8,9].
NMR logging tools used in China are mostly MRIL-Prime
tool of Halliburton, of which the acquisition, processing and
evaluation methods are entirely based on 1D NMR technology.
However, there are only two 2D NMR logging equipment
manufacturers across the world, and they just provide services
and don’t sell the tools. Can we change the measurement
modes of the existing 1D NMR tools and use innovative
processing approach to achieve the measurement effect of 2D
NMR? Based on the existing acquisition modes of
MRIL-Prime tool, this paper uses multiple measurements with
different modes of logging to obtain 2D NMR information
and adopts multi-echo train joint inversion processing tech-
niques to achieve fluid relaxation-diffusion (T2, D) 2D NMR
distribution in porous media. Taking full advantage of the
diffusion coefficient (D) and the relaxation time differences
between different fluids, this approach has greatly improved
the ability of fluid typing in 2D space compared to 1D T2
spectrum, and it is of great significance to determining fluids
by NMR logging in complex reservoir[10].

Received date: 2 Sep. 2011; Revised date: 18 July 2012.

* Corresponding author. E-mail: long@petrochina.com.cn
Foundation item: Supported by the PetroChina Research Institute of Petroleum Exploration and Development Program “New Techniques for 2D Nuclear Magnetic
Resonance” (115201C90001).
Copyright © 2012, Research Institute of Petroleum Exploration and Development, PetroChina. Published by Elsevier BV. All rights reserved.
 HU Falong et al. / Petroleum Exploration and Development, 2012, 39(5): 591–596

1 Method and principle because T2 of three kinds of fluids for the oil, gas, water have
some overlaps (Fig. 1a), it is difficult to identify fluid proper-
The NMR transverse relaxation time (T2) of pore fluid is
ties using 1D NMR information, but there are some differ-
mainly subject to the affection of the bulk relaxation, surface
ences between the three kinds of fluids (Fig. 1b). Therefore,
relaxation and diffusion relaxation, which can be expressed as:
we can clearly distinguish different fluids using relaxa-
1 1 1 1
  (1) tion-diffusion (T2, D) 2D NMR information (Fig. 1c, the wa-
T2 T2B T2S T2D
ter-line location is related to temperature and salinity, water
1 D(J G TE )2 diffusion coefficient increases with the increase in tempera-
where, 109 .
T2D 12 ture and decreases with the increase in salinity; oil-line loca-
Because the magnetic field gradient (G) is zero in the uni- tion is related to viscosity, diffusion coefficient of oil de-
form field, the diffusion relaxation makes no contribution to creases with the increase in viscosity).
transverse relaxation time, namely 1/T2DĬ0. Pore fluid re- Relaxation-diffusion (T2, D) 2D NMR need to collect more
laxation time of the NMR signals is mainly subject to bulk signal data than 1D NMR, and need to measure under more
relaxation and surface relaxation, known as the intrinsic re- modes without changing the pulse sequence of the tool[11,12].
laxation time, which can be expressed as: MRIL-Prime logging tool has hundreds of acquisition modes
1 1 1 with abundant acquisition parameters, and shows flexibility to
 (2)
T2int T2B T2S choose different wait time (TW) and echo spacing (TE), but
Under the gradient field, relaxation time can be written as: the existing acquisition modes and the processing approaches
1 1 1 1 D(J G TE ) 2 can only get 1D T2 distribution. If the logging information
+ = +109 (3) under different acquisition modes can be rationally combined,
T2 T2int T2D T2int 12
we can separate the diffusion coefficient from the transverse
For fluids in porous media, NMR information reflects the
relaxation time through a special inversion processing tech-
contribution of transverse relaxation time, when there is
nology to identify the fluid properties in 2D space.
two-phase or multi-phase fluid in porous media, the fluid re-
Under the gradient field, if the same TW and different TE
laxation information may show overlap phenomenon, which are used for data acquisition (measurement mode shown in
makes it difficult to discriminate the fluid composition. If we Fig. 2), then the measured traverse relaxation time includes
can "separate" the intrinsic relaxation time and diffusion coef- fluid diffusion coefficient information. And because all data
ficient from NMR transverse relaxation time and form 2D acquisitions are base on CPMG pulse sequence, the NMR
space information, we can determine the properties of fluid in echo signal of pore fluid conforms to multi-exponential decay
pore media considering that different fluids have different law, namely:
diffusion coefficient and transverse relaxation time. As shown m p

¦¦ (1  e
TW / T1, j  ti (1/ T2 int, j 1/ T2 D ,l )
in Fig. 1, gi ) f jl e  Hi
j 1 l 1

( i 1, 2," , n ) (4)
2 Multi-echo train joint inversion algorithm
The pore fluid signal using spin-echo measurement has the
form of echo train satisfying multi-exponential decay law of
Eq.3. The value of TW will affect the polarization degree of
the hydrogen nucleus and its value can be obtained from log-

Fig. 1 The maps of 2D NMR information distribution of differ- Fig. 2 The schematic diagram of (T2, D) 2D NMR logging with
ent fluid component in the porous media different echo spacing acquisition mode
 592 
 HU Falong et al. / Petroleum Exploration and Development, 2012, 39(5): 591–596

ging acquisition parameters card, so we can determine the
value of the polarization factor. According to the acquisition
mode shown in Fig. 2, we combined the signals of multiple
echo trains measured from multi-CPMG, that is,
g ( g1 , g 2 ,",g n )T . Then we can use regularization method to
conduct joint inversion to multi-echo trains to determine the
signal amplitude f jl under (T2, D) acquisition mode,
through the following steps:
°­ 1 n ª m p
min I ( f ) min ® ¦ « ¦¦ f jl e i 2 int, j 2 D ,l 
 t (1/ T
Eq. 9 can be solved to obtain the magnitude f jl of the 2D
NMR (T2, D) model, namely 2D NMR fluid signal distribu-
tion, using the method of search for the minimum of convex
function. For different signal to noise ratio (SNR) signals of
multiple measurements, n echo strain combination can be
conducted taking into account of SNR, namely, g ( g1 ,
H1 H
g2 ," g n 1 )T , then we can strike the f jl according to the
H2 Hn
above echo strain inversion method [1315].
1/ T )

°¯ 2 i 1 ¬ j 1 l 1 3 Applications
2
º D m p ½°
g i »  ¦¦ f jl2 ¾ (5) There are many kinds of sedimentary facies existing in the
¼ 2 j 1 l 1 °¿ Nanpu Sag, such as fluvial facies, fan delta facies and near-
The above equation can be converted to: shore subaqueous fan facies. Because of the sedimentary fa-
 ti (1/ T2 int, j 1/ T2 D ,l ) ª
n m p
cies overlaying each other, reservoir lithology and physical
¦ « ¦¦ f jl e  gi @  D i f jl 0
 ti (1/ T2 int, j 1/ T2 D ,l )
e
i 1 ¬ j 1 l 1 property change quickly, it is very difficult to determine the
(i 1,2,", n; j 1,2,", m; l 1,2,", p ) (6) fluid properties using conventional logging data[1620]. On the
basis of theoretical studies and numerical simulation, we
Where, f jl is the signal amplitude of (T2, D) 2D NMR dis-
conducted multi-echo spacing 2D NMR logging in Well A of
tribution, thus its value is non-negative, so it meets:
Nanpu Sag using the MRIL-Prime tool to explore the feasibil-
ª n )º
f jl D max «0, ¦ ci e i 2 int, j 2 D ,l »
 t (1/ T 1/ T
ity of 2D NMR and determined fluid properties using
¬ i1 ¼
MRIL-Prime tool.
( j 1,2,", m; l 1,2,", p ) (7)
In reference to the existing acquisition modes of MRIL-
Combining Eq. 6 and Eq. 7, we can get: Prime tool and taking into account the viscosity of the Nanpu
m p

¦¦ e
 ti (1/ T2int, j 1/ T2 D ,l )  ti (1/ T2 int, j 1/ T2 D ,l )
e ci  oilfield, acquisition mode combination is selected using nu-
j 1 l 1 merical simulation methods: echo spacings are designed to be
D ci gi (i 1,2,", n) (8) 0.9, 1.2, 3.6 and 6.0 ms and wait time is set as 12.998 s; under
The value of ci in Eq. 8 can be calculated using the fol- the context of the combination parameters the signal separa-
lowing formula: tion of the oil, gas and water in porous medium are achieved.
1 T Conventional well logs and 1D NMR T2 information from
\ c ( M  D I )c  c ˜ g (9)
2 multi-echo spacing logging are shown in Fig. 3. Echo spacing

Fig. 3 The NMR T2 distribution of Well A with different echo spacing collected by MRIL-Prime tool
 593 
 HU Falong et al. / Petroleum Exploration and Development, 2012, 39(5): 591–596

Fig. 4 The (T2, D) 2D NMR information distribution maps of different depth in Well A

of 1.2 ms and 6.0 ms are corresponding to the MRIL-Prime
tool DTE412 mode; echo spacing of 0.9 ms and 3.6 ms are
corresponding to the MRIL-Prime tool D9TWE3 mode, which
can provide the differential spectrum and shifted spectrum
results simultaneously, so as to enhance the fluid identifica-
tion capabilities of the NMR logging. At the 2955–2970 m
hole section, T2 spectrum shapes are mainly in signal peak
distribution, fluid composition are mainly free fluid; with the
increase of echo spacing time (TE), T2 spectrum moves to the
direction of fast relaxation, and T2 spectrum after the move-
ment is still mainly in single peak distribution. Due to the
lower viscosity of crude oil of the region (average value is 4
mPas), the shifted spectrum method is difficult to separate the
oil-water signal in the T2 domain, so we consider to adopt 2D
NMR for oil-water layer recognition. After data accumulation,
orthogonal decomposition, phase angle smoothing and filter-
ing to the original NMR signals collected in four different
echo spacing conditions, four echo trains are provided with
different echo spacing at the same depth. In view of the higher
SNR of signal after data stack, we conducted joint inversion
to the above four groups of echo strains, obtaining relaxa-
tion-diffusion 2D NMR logging information (Fig. 4) and the
diffusion coefficient of water-line approximate to 2.5×105
cm2/s. After adding coefficient information, the oil-water sig-
nal was accurately separated, which indicates that 2D NMR
method has certain advantages compared with 1D shifted
 594 
spectrum method in terms of light oil recognition. Oil satura-
tion in the corresponding depth is about 62%, and the conclu-
sion of 2D NMR logging is oil. Oil testing validation the layer
is oil, and the reservoir oil production in the early day is 99 t,
which verifies the 2D NMR logging conclusion.
Large-pore reservoirs are well developed in the Dongying
Formation in the Nanpu Sag. When using 1D NMR logging
for oil-water layer identification by the differential spectrum
method, differential spectrum signals often emerge in water
layer resulting in lower coincidence rate of logging interpreta-
tion. After analysis, we considered that it is because the
commonly acquisition mode cannot completely magnetize the
pore water within short wait time. To solve this problem, we
used MRIL-Prime tool with the same acquisition parameters
and mode as Well A for 2D NMR logging fluid identification.
According to T2 distributions of Well B in different acquisi-
tion modes in Fig. 5, because acquisition mode D9TWE3 was
selected, we can directly access to the differential spectrum
signal (Fig. 5, track 10). If 1D NMR fluid analysis method is
adopted, the layer may be mistaken for oil layer.
After processing the data in Fig. 5, we can obtain 2D NMR
information for each depth point (Fig. 6, track 4 to 8). As
shown in Fig. 6, track 9 is the diffusion coefficient spectrums
information from projection, track 10 is the intrinsic relaxa-
tion time spectrums from projection which can reflect the pore
structure information. Oil signals haven’t been found from
 HU Falong et al. / Petroleum Exploration and Development, 2012, 39(5): 591–596

Fig. 5 The NMR T2 distribution of Well B with different echo spacing collected by MRIL-Prime tool

Fig. 6 The (T2, D) 2D NMR information distribution maps of Well B

(T2, D) 2D maps or from the diffusion coefficient spectrums,
and the diffusion coefficient spectrums are mainly distributed
near the water-line, showing that 1D NMR differential spec-
trum signal is water signal. Therefore, we can tell the layer
(3164.0–3175.6 m) is water layer. In addition, there are some
signals greater than the water diffusion coefficient near 170 m
of track 3 in Fig. 6. Analysis shows that the inconsistency is
caused by the different measurement targets of two acquisi-
tion processes, which can be verified from the 1D NMR T2
spectrum or quality curve. The oil test results confirm that the
 595 
daily water production of this layer is 74 m3. It can be seen
that 2D NMR has obvious advantages in identification of res-
ervoir fluid properties compared to the 1D NMR differential
spectrum method.
4 Conclusions
In the case without changing the hardware structure of
MRIL-Prime tool, (T2, D) 2D NMR information can be ob-
tained by utilizing its existing acquisition modes to multi-echo
strains measurements and 2D NMR echo strain joint inversion
 HU Falong et al. / Petroleum Exploration and Development, 2012, 39(5): 591–596

data processing technology. (T2, D) 2D NMR can more clearly and applications. Texas: Gulf Publishing Company, 1999.
and accurately separate oil-water signal compared to 1D NMR [2] Akkurt R, Prammer M G, Moore M A. Selection of optimal
T2 information, but it needs more echo information. For (T2, D) acquisition parameters for MRIL logs .In: SPWLA 37th An-
nual Logging Symposium. New Orleans: SPWLA, 1996.
2D NMR fluid identification method, the diffusion coefficient
[3] Akkurt R. Determination of residual oil saturation using en-
of water-line in the vicinity of 2.5×105 cm2/s, its position is hanced diffusion. SPE 49014, 1998.
relatively fixed, temperature and salinity have little influence [4] Wang Zhongdong, Wang Hao, Xiang Tiande, et al. Integration
on its location. Due to downhole crude oil composition and of NMR DTW logs and DTE logs to improve the oilbed interpre-
measurement conditions are different from laboratory, the oil tation accuracy. Well Logging Techology, 2001, 25(5): 365–368.
peak may deviate from the oil line calculated based on labo- [5] Liu Zhonghua, Li Xia, Zhao Wenzhi, et al. Enhanced diffusion
ratory data, so large amount of 2D NMR logging tool tests are theory of nuclear magnetic resonance (NMR) and its applica-
tion to fluid identification of complex reservoirs. Petroleum
needed to accurately determine the location of the oil line.
Exploration and Development, 2010, 37(6): 703–708.
[6] Sun B, Dunn K J. Core analysis with two dimensional
Nomenclature NMR .In: International Symposium of the Society of Core
Analysts. Monterey: SCA, 2002.
T2—transverse relaxation time, ms; [7] Hürlimann M D, Venkataramanan L. Quantitative measure-
T2B—bulk relaxation time, ms; ment of two-dimensional distribution functions of diffusion
T2S—surface relaxation time, ms; and relaxation in grossly inhomogeneous fields. Journal of
T2D—diffusion relaxation time, ms; Magnetic Resonance, 2002, 157: 31–42.
D—fluid diffusion coefficient, cm2/s; [8] Xie Ranhong, Xiao Lizhi, Deng Kejun, et al. Two-dimensional
J —gyromagnetic ratio, Hz/G; NMR logging. Well Logging Techology, 2005, 29(5): 430–434.
[9] Xie Ranhong, Xiao Lizhi. The (T2, D) NMR logging method
Gümagnetic field gradient, G/cm;
for fluids characterization. Chinese Journal of Geophysics,
TE—echo spacing, ms;
2009, 52(9): 2410–2418.
T2int—intrinsic relaxation time, ms;
[10] Xiao Lizhi. Some important issues for NMR logging applica-
I —porosity, %;
tions in China. Well Logging Techology, 2007, 31(5): 401–407.
TW—wait time, ms; [11] Hu Falong, Xiao Lizhi. A method for determining NMR log
giüsignal amplitude of the ith echo train, %; signal intensity. Nuclear Electronics & Detection Technology,
nünumber of echos; 2006, 26(5): 557–560.
münumber of T2 components; [12] Hu Falong, Xiao Lizhi. Numerical simulation of static mag-
pünumber of D components; netic filed distribution of NMR logging tools. Progress in
T1,jütransverse relaxation time of the jth component, ms; Geophysics, 2008, 23(1): 173–177.
fjl—2D NMR (T2, D) signal amplitude of the jth T2, lth D compo- [13] Dunn K J, Latorraca G A. The inversion of NMR log data sets
nent, %; with different measurement errors . Journal of Magnetic Reso-
ti—measurement time of the ith echo, ms; nance, 1999, 140: 153–161.
T2int,j—intrinsic relaxation time of the jth T2 component, ms; [14] Xie Ranhong, Xiao Lizhi, Liu Jiajun, et al. A method for mul-
T2D,l—diffusion relaxation time of the lth component, ms; tiple echo trains jointing inversion of NMR relaxation measure-
i—noise signal of the ith echo, %; ments. Chinese Journal of Geophysics, 2009, 52(11): 2913–2919.
güvector of singnal intensity; [15] Tan Maojin, Zou Youlong, Liu Bingkai, et al. Inversion simu-
lation of (T2, D) 2D NMR logging and analysis of observation
—smoothing factor, dimensionless;
parameters effects in gas-water model. Well Logging Techol-
i—the ith smoothing factor, dimensionless;
ogy, 2011, 35(2): 130–136.
ciüthe ith undetermined coefficients, dimensionless;
[16] Zhou Fengming, Si Zhaowei, Ma Yuejiao, et al. Integrated
cüvector of undetermined coefficients, dimensionless; identification method for low-resistivity hydrocarbon layers in
M—matrix of diffusion relaxation time and intrinsic relaxation Nanpu Sag. Petroleum Exploration and Development, 2008,
time joint distribution; 35(6): 680–684.
I—unit matrix, dimensionless; [17] Li Changxi, Shi Yujiang, Zhou Cancan, et al. Evaluation of
SP—spontaneous potential, mV; low amplitude and low resistivity pay zones under the fresh
GR—gamma rays, API; drilling mud invasion condition. Petroleum Exploration and
dh—hole diameter, cm; Development, 2010, 37(6): 696–702.
RLLD—deep lateral resistivity, ·m; [18] Luo Hongmei, Zhu Yixiu, Mu Xing, et al. Seismic facies pre-
RLLS—shallow lateral resistivity, ·m; diction of lacustrine beach and bar reservoirs in the deep zone
RMSFL—micro spherical focused resistivity, ·m; of the Bonan Subsag, Bohai Bay Basin. Petroleum Exploration
I CNL—neutron porosity, %; and Development, 2011, 38(2): 182–190.
—density, g/cm3; [19] Sun Hefeng, Zhou Xinhuai, Peng Wenxu, et al. Late-stage
hydrocarbon accumulation and enrichment in the Huanghekou
't—acoustic travel time, us/m;
Sag, southern Bohai Sea. Petroleum Exploration and Devel-
EDSPEC—differential spectrum, ms.
opment, 2011, 38(3): 307–313.
[20] Wang Yuman, Huang Xu’nan, Cheng Kunfang, et al. Evalua-
References tion methodology of lithologic exploration areas in rift basins:
A case from the Bohai Bay Basin. Petroleum Exploration and
[1] Coates G R, Xiao L Z, Prammer M G. NMR logging principles Development, 2011, 38(1): 122–128.
 596