Note: Descriptions are shown in the official language in which they were submitted.
~`~? 93/l~U~21210 ~ 7 PCT/US92/102~
IMPROVED MAGNETIC RESONANCE ANALYSIS IN
REA~ TIME, $NDUSTRIAL USAGE MODE
1 0
; ~ CROSS REFERENCE TO RS~ATED APPLICAT$ONS
~; This application is closely related to the
U.S.Patent 5,015,954 issued on 14 May 1991 to
~Dechene et al~ and to U.S. Patent 5,049,81g
granted~Sept. 17,~1991 to Dechene et al., both
entitled~"Magnetic ~Resonance Analysis in Real
Timé;,~;Industrial~ Usage Mode".~ Both~ of these
patents~ are~of~common ~assignment~ with this
- - ~20 applicàtion,~ ~and the disclosures of both are
her~e~by~incorporated heroin by reference, as
though set~ out at }ength herein.
FT~D OF ~E TNVENTION
. 25 ~ -The~ ~present invention relates to an
instrument for measurement of the type and
quantity of lattice bound and free magnetically
active nuclei within successive samples of a
process material flow through pulsed nuclear
magnetic resonancé (NMR) techniques and more
; particularly the application of such measurèment
to industrial process control of moisture
content, polymer content, crystallinity
-~ fraction, oil/fat fraction and other percentages
-35 of.components analysis and other parameters.
,, ~, :, : , , -
W093/1n4~ 2I2I027 PCT/~S9211024~
BAC~GROUND OF THE INVENTION
NMR techniques have grown extensively over
the past forty years, most notably in the
medical instrumentation areas where in ~ivo
examination of various parts of the human body
can be seen, and in clinical research raboratory
uses. In addition there has been some use and
interest in the application of these techniques
to industrial instrumentation and control tasks.
The present invention enables effective
utilization (technically and economically) of
pulsed NMR techniques in industrial areas tO
replace or complement existing optical and
radiant energy-based instrumentation.
Pulsed NMR spectroscopy is described in our
above-cited patent. This technique uses a burst
or pulse which is designed to excite the nuclei
of a particular nuclear species of a sample
being measured ~the protons, or the like, of
~ 20 such sample having first been precessed in an
- essentially static magnetic field); in other
words the precession is modified by the pulse.
After the application of the pulse thère occurs
a frèe induction decay (FID) of the
25~ magnetization associated with the excited
nuclei. Traditional Fourier Transform analysis
generates a frequency domain spectrum which can
be used to advantage in studying the nuclei of
interest. The duration of the pulses, the time
between the pulses, the pulse phase angle and
the composition of the sample are parameters
which affect the sensitivity of this technique.
These frequency domain techniques are not easily
useable in industrial applications, especially
on-line applications.
An object of this invention is an improved
measurement system which leads to accurate, fast
- ` ~ g3/lOU~ 2 1
2 1 0 2 7 PCT/~'S92/1024~
determination of the types and quantity of the
nuclear species of interest.
A further object of this invention is its
application to the industrial, on-line problems
of measuring and calibrating the controlling
processes per se.
Another object of this invention is to
utilize time domain analysis in achieving such
system.
The principal variables of interest are
moisture, oil/fat and polymer crystallinity,
density, tacticity and the molecular structure
of polypropylene and polyethylene. But other
parameters can be measured based on hydrogen or
other sensitive species including e.g. sodium
and fluorine. It is an object of this invention
to accommodate a variety of such measuring
tasXs. ~ ~
Another object is to accommodate the
dynamics of industrial on-line applications
including variations of density, temperature,
packing and size factors, friction and static
electricity, vibrations and frequent,
repetitive, cyclic and non-cyclic measurements.
25` -`~ A further object of the invention is to
- integrate a'l the features of accurate, fast
determination of the types and quantity of the
nuclear species of plastics or polymer molecular
structure of interest the use of time domain
analysis in such a system, its application to
the industrial, on-line problems of monitoring
and controlling processes, measuring free and
bound water in organic and inorganic
substances(based on hydrogen nuclei modified-
~ precession analysis) and other parametermeasurement (based on hydrogen or other
sensitive species including e.g., fluorine,
" .
W093/10468 2 1 2 1 0 2 7 PCI/l'592/1U24~
sodium 23, etc.) accommodating the dynamics of
industrial on-line applications including
variations of density, temperature, packing and
size factors, friction and static electricity,
- S vibration and frequent, repetitive, cyclic and
non-cyclic measurements. ~~~
A further object of the present invention is
to extend those achievements further in relation
to industrial on-line processing, and the like,
as applied to mixed species (or mixed phases) of
NMR-active materials and more particularly
foodstuffs, where moisture within a solid,
crystalline/amorphous ratio, oils/fat or solvent
ratio, and polypropylene wherein the tacticity
lS of the polypropylene and the concurrent
molecular structure is determined on-line.
SUMMARY OF ~E INVENTION
The present invention provides a materials
~; 20 measurement system using magnetic resonance
hardware, controls (and related data capture and
data reduction means and steps) and techniques,
preferably in the time domain. The system can be
` used to-capture data from a continuous
25 ` production line or like repetitive measurements
system.
The NMR system effects a reliable extraction
of free induction decay data in a way that is
practical in an industrial on-line context and
economically practical. The system is
characterized by provision of a base magnetic
field homogeneity to a reasonable degree and
offset of inhomogeneity effects, temperature
stabilization to a reasonable degree and offsets
of thermal drift effects and use of multiple
runs (10-100) for each measurement with digital
data red~ction and use of statistical methods or
`~g3/1~U~ 2 1 2 1 0 2 7 PCT/US92/1024~
other data manipulation for industrially
effective measurement. For polypropylene, these
data can be represented, for discussion/
analysis, as a free induction decay curve ~FID)
with attention to time sequence components of a
first, very fast Gaussian, followed b~ a slower
modified Gaussian (a Gaussian multiplied by a
cosine with an argument of K~t)l/2), in turn
followed by an even slower exponential, In
materials other than polypropylene these data
can be represented, for discussion/analysis, as
an FID with time components of a first, very
fast, modified Gaussian (a Gaussian multiplied
by a sinc function sometimes called an Abragam)
followed by a slower essentially Gaussian in
turn followed by an even slower exponential
region.~These different components are
representative of proton relaxation after an
initial excltation by a pulse of transmitted and
~ 20 near resonant coupled radio frequency energy
- that induces a modification of the precession of
protons in the sample being measured in a high
static magnetic field. The calibration system
~ including the M-L technique (described below)
will still be applicable.
In the case of polypropylene, the fast
Gaussian FID portion is based on measurement
data points of magnetization decay of a
relatively immobile~structure present in the
polypropylene sample and picked up at the NMR
system's receiver. This component can be related
to the isotactic polypropylene structure. The
slower modified Gaussian component is a
transition zone between the constrained and the
3S more mobile structures. The slow exponential FID
portion is usually based on mobile structure.
This component can be related to the atactic
:::
~ ~ r
W093/1~U~ 2 1 2 1 0 2 7 PCT/US92/1024~
polypropylene structure. The fast Gaussian, the
slower modified Gaussian and exponential FID
portions and the FID as a whole can be
extrapolated to a decay origin usually set close
to the time center of the excitation pulse~
Zero time intercepts of these curves~provide
ratio data using the FID intercept and/or
intercepts of one or more of the curve portions
to determine polymer tacticity or
crystallinity. Density can also be determined
through the invention because the FID varies
predictably as a function of
tacticity/crystallinity and density.
In the case where the FID is analyzed as an
Abragam, slow Gaussian and an exponential, the
Abragam FID portion is based on measurement data
points of magnetization decay of immobile
structures of highly constrained protons ~as in
crystalline structures) present in the sample
and picked up at the-NMR system's receiver; this
portion is usually based on the response of
bound proton species such as the hydrogen or
hydrate content of chemical compounds (or
similarly for many other NMR-active species,
e.g. fluorine or sodium-23, which are chemically
bonded and highly constrained~. The slower
Gaussian region is a transition zone of somewhat
more mobile structures between the constrained
and the loosely bound protons. The slow
exponential FID portion is usually based on
loosely constrained or unconstrained structuxes
of NNR-active species such as moisture
physlcally present in a sample but essentially
chemically unbonded thereto. The Abragam,
~ ~ 35 Gaussian and exponential FID portions and the
-~ FID as a whole can be extrapolated to a decay
; ~ origin usually set close to the time center of
.
2 1 2 1 0 2 7 Pcr/~Js~ ou~
the excitation pulse. Zero time intercepts of
these curves (i.e., FID and one or more of its
component curve portions) provide ratio data
using the PID intercept and/or intercepts of one
or more of the curve portions to determine e.g.
free vs. bound water in a moist mater~a~ (e.g.
for process control of industrial chemicals,
minerals and metals, agricultural commodities,
processed foods by determining moisture content
for upstream or downstream correction or for
acceptance/rejection purposes). Instead of
determining moisture in, say a food product the
object may be to determine ratio of relatively
crystalline and non-crystalline components of a
material, e.g. hard and soft components of a
polyethylene material and that is accomplished
in a fashion analogous to the moisture
- measurement. Polymer crystallinity and density
can also be determined through the invention
~ 20 because the FID varies predictably as a function
of crystallinity and density.
Also oils,fats and solvents may appear in
foods, and residual solvents may appear in
polypropylene or other industrial materials as
an error factor to be resolved and/or as a
target parameter to be measured. The NMR
response of such solvent or oil/fat portion can
be isolated by the present invention, and to
some degree sub-components of such solvents or
oils/fats may appear as additional very slowly
decaying exponential regions in the calibration
system described herein.
The measuring system of the invention
comprises economically soaled down and
3S industrially hardened portions, relative to the
widely used laboratory systems. A magnetic
~ essentially fixed field comprises closely spaced
:
~::
W093/10468 2 1 2 1 0 2~ PCT/~S92/~024 ~
pole pieces with a 4,000 - 8,000 Gauss field
(about 4,700 Gauss, nominally). Helmholtz coils
are provided which are adjustable to provide
rapid adjustments for the precise, correct field
and overlaid with coarse, slower adiustments to
thermal environment. This is to ass6re that the
product of a materials related constant ~gamma)
multiplied by the magnetic field strength, which
is resonant frequency, will match excitation
frequency by a selected offset. Still further
fine adjustment is made in signal processing as
described below.
The present invention accommodates great
streams of data in practical ways through
features, described below, which are inter-
related to the thermal controls to provide a
measuring system meeting the foregoing objects.
The materials of construction are also
inteqrated into the reliability considerations,
as described below. Measurement of a sample is
often accomplished in approximately a minute ~in
contrast to hours-long measurements of many
prior art systems).
The-measurements made throu~h the present
2S invention based`on ratios of intercept delay
time constants and/or integrated areas under
curves and/or peak analysis are independent of
weight or volume of sample in the measuring
region or gain of the system whereas precise
weight measurement is a necessary feature (and
limitation of) many prior art systems.
Prior art efforts at industrial on-line
measurement of the same parameters as are
treated herein have involved non-NMR
gravimetric, chemical, radioactive, acoustic,
optical and electrostatic/capacitive systems,
- none wholly satisfactory for present purpose and
~ g3/l0468 2 1 2 1 0 2 7 PCr/~lS92/1024~
NMR usage in support of continuous industrial
processes has been a forcing of off-line
laboratory instruments into service (at great
expense and nevertheless with inadequate data
sampling rates) or some early efforts of the
1980's at industrially hardened pulsed NMR
instruments making use of only one or two data
points for FID analysis. The present invention
breaks out of all those dead ends.
Laboratory methods of frequency domain NMR
analysis are described for crystallinity content
determination in, e.g., Spiess, "Molecular
Dynamics of Solid Polymers As Revealed By
Deuteron NMR", 261 Colloid & Polymer Science
193-209 (1983) and Kauffman et al.,
"Determination of Transition Temperatures and
Crystalline Content of Linear High Molecular-
Weight Polyethylene by Proton NMR Spectroscopy",
27 Jl. of Polymer Science 2203-2209 (1989).
Time domain analysis using pulsed and multiple
pulsed NMR free induction decay in coals for
detection of free radicals therein is shown in
the laboratory systems of Gerstein et al. (Iowa
State Uni~ersity Ames Laboratory) reported in
- 25 "Utility of Pulsed Nuclear Magnetic Resonance In
Studying Protons In Coalsn, 81 Jl. of Phys.
Chem. 566-571 (1977~ and "lH Nuclear Magnetic
~esonance Studies of Domain Structures In
Polymers", 52(9) J. Appl. Phys. 5517-5528
(1981). The instruments or former instruments
of IBM Federal Systems Division model and
Brucker GmbH model P201 and the description in
U.S. patent 4,430,719, granted February 7, 1984,
to Pearson are the earlier attempts referred to
above as industrial use of NMR methods. The
Pearson work was embodied in 1985 industrial
plant control work of Kaiser Aluminum & Chemical
W093/10468 2 1 2 1 0 2 7 PCr/~3Sg2/1024~ ~
Corp. It was not effective as a reliable
quantitative device. Auburn International, Inc.
offered the Pearson/Kaiser product for sale in
1987-1988 and it could not meet the needs of
industrial on-line monitoring. These failures
were followed by the making of the in~ëntions of
the above cited Dechene et al. patents and these
as enhanced through the present invention, all
solve the industrial on-line problem for various
materials.
Other objects, features, and advantages will
be apparent from the following detailed
description of preferred embodiments taken in
conjunction with the accompanying drawings in
which:
BRIEF DESCRIPTION OF ~HE DRAWING(S)
FIGS. 1 and 2 are lateral and cross-sections
of a preferred embodiment of the invention
including eiectrical block diagram components.
FIG. 3A shows the voltage-time waveforms of
the free induction decay (FID) of the embodiment
of FIGS. 1-2 in the course of operation
- analyzing polypropylene;
2S FIG. 3B shows the voltage-time waveforms of
the free induction decay (FID) of the embodiment
of FIGS. 1-2 in the course of operation
analyzing other materials, especially
polypropylene;
FIG. 4 is a flow chart of measuring steps
utilizing the FIGS. 1-2 apparatus including its
signal processing elements, ~the activity of
which is illustrated by the FIG. 3 waveforms);
FIG. 5 is a voltage - time trace for FID
3S curves derived from samples of polypropylene
with different isotactic indices (xylene or
hexane solubles);and
'-~g3/~U~ 2 1 2 1 0 2 7 PCT/US92/1024~
11
FIG. 6 is a calibration curve of
polypropylene xylene solubles, as measured by
the present invention (y axis) versus standard
methods (x axis).
DE~I!AII.ED DESC~IPTION OF PR~i:~ERRED
EMBOD IMENT S
FIGS. 1-2 show transverse and cross sections
with block diagram inserts of a first embodiment
of the apparatus and method of the invention.
An industrial process line IPL has material
flowinq as indicated by arrow A-. Some of the
material is captured by a probe P and fed
through an lnlet line LI to a sample region S1.
The said region is defined by a tube 98
typically about 30 cm long made of an
essentially non-magnetic, nonconducting material
which~ does not itself generate substantially
interfering FID signals (glass, certain
ceramics, certain fluorocarbons or hybrids).
~ ; The~sample region is defined between lnlet and
--~ outlet valves Vl and V2. Gas jets J are also
provided. These are pulsed on/off repeatedly to
agitate fluent sample materials during sample
~admission and expulsion. The region S2 is the
critical portion of the sample. It is
surrounded by a sample coil 100 tuned to
resonance and driven by a tuning circuit 102 and
related transmitter/receiver controller 104.
Grounded loops 101 are Lenz Law shields which
are provided above and below coil 100 to help
shape the field of coil 100 -- i.e., contain the
field established by an excitation pulse. The
controller 104 includes an on-board
microprocessor and required power supply
elements, memory, program and I/O decoding
suitable to interconnect to the hardware shown
WO93/10468 2 1 2 1 0 2 ~ Pcr/~ls92/l024:~
and to an external microcomputer 106 with
keyboard 108, monitor ~or other display) 110,
recorder 112 and/or process controller 114 (to
control the process at IPL). The operator
initiates and controls operation from the
display keyboard 108 and the resultin~ data and
signals are subsequently shown on the display
100 and utilized in 110, 112 and/or 114. The
computer 106 also controls instrument operation
conditions.
The region S2 of tube 98 and coil 100 are in
a static, but adjustable, crossing magnetic
field defined by a magnetic assembly 116 which
comprises a yoke 118, pole pieces 120,
surrounding Helmholtz coils 124, and a coil
current qenerator 117. The critical sample
region S2 of the tube 98 and magnet are
contained in a metallic ~but non-ferromagnetic)
~ox 126 with highly thermally conductive face-
plates 128 and internal partitions 130 and over-
all mass related to each other to minimize
harmonics and other interferences with a signal
emitted from coil 100 to a sample and/or
returned'from the sample for pick-up by coil 100
and its tuned circuit 102 and transmit/receive
controller 104.
The magnetic assembly 116 including yoke
118, and other parts therein as shown on FIGS.
1-2, is in turn contained in an environmental
control chamber 132 with optional inert gas fill
- and purge controls (not shown), an internal gas -
heater 134, a motor M driving fan 136, and a
temperature sensor 138 which car. be applied to
the yoke or other detection region whose
temperature is reflective of the temperature at
pole pieces 120 and in the sample region
therebetween. A thermal controller 140
~ g3/10468 2 1 2 1 0 2 7 PCT/US92/1024~
processes temperature signals from 138 to adjust
heating/circulation at 134/136 as a coarse
control and to adjust current through the
Helmholtz coils 124 at magnet pole pieces 120 as
a sensitive and fast fine control, as ~ell as
implementing general control instructl~ons of
computer 106. Further thermal stabilization may
be provided by a closed loop heat exchanger 142
having pump 144 and coils 146 attached to yoke
118 and coils 148 attached to the plates 128 of
box 126. Still further thermal stabilization may
be provided by directing temperature controlled
air (controller not shown) into an annular
region defined by collar 134A around the sample.
Temperature controls 134-138 and 142-146,
described above, establish very coarse and less
coarse thermal controls countering sample
temperature variations. In addition there may be
a heat guiding collar 134A that directs
controlled temperature air to the sample area.
~ Alco as stated above in another preferred
; embodiment the heater 134 and/or the fan 136 may
be disabled, thereby reducing interference, when
performing high sensitivity measurements.
- 25 The strength, consistency and constancy of
the magnetic field between poles 120 in the
region S2 of the sample is thus controlled by a
uniform base magnetic field in the entire region
S2. The Helmholtz coils 124 are energized by
the coil current controller 117 to accurately
trim the final magnitude of the field in which
the sample is placed. This field is the vector
àddition of the fields due to the magnet poles
120 and the Helmholtz coils 124. The controller
117 sets the current through the Helmholtz coils
124 using current generators. The coils 124 are
wound around the magnet pole pieces such that
WOg3/l04~ 21 21 0 2 7 PCT/~S92/1024~
14
the magnetic field created by the current in the
coils 124 can add to or subtract from the field
created by the magnet pole pieces. The
magnitude of the current through the coils 124
S determines the strength of the field added to or
subtracted from the field due to the magnet pole
pieces (and related yoke structure) alone.
The actual determination of the current
through the Helmholtz coils is accomplished by
carrying out the magnetic energy and resonance
techniques hereinafter described in preliminary
runs and ad~usting Helmholtz current until the
maximum sensitive resonance is achieved. The
~elmholtz current is then adjusted to offset the
lS system from resonance by about 1-3 KHz.
The major elements of electrical controls
are tuner 102, including coils 100 and 101 and
variable capacitors 102-1 and 102-2, resistor
102-3 and diodes 102-4 and constructed for
tuning to Q of twenty to fifty to achieve coil
100 resonance, and control 104 including a
transmit/receive switch 104-1, a transmitter
104-2 and receiver 104-3, a crystal oscillator
104-4, gated pulse generator ~GPG) 104-5, and
25- phase shifter 109-6. The crystal provides a
nominal twenty megahertz carrier which is phase
modulated or demodulated by the MOD, DEMOD
elements of transmitter 104-2 and receiver 104-
3. The receiver includes variable gain
amplifier elements 104-31 and 104-32 for
operatton. ~he analog signals received are fed
to a high speed at least 12 bit flash A/D
converter 105-1 and internal (to the instrument)
CPU element 105-2, which provides data to an
external computer 106 which has a keyboard 108,
monitor 110, modem 109, recording elements 112
- and process controller elements 114, e.g., for
2121027
y~g3/1nu* PCT/~'S92/1024
control of valves Vl, V2 via valve controls ll5
and/or to coil current controls 122, all via
digital-analog converters (not shown).
The analog signal FID curve is conditioned
by a Bessel filter which acts as a pre-filter
and an anti-aliasing filter. After dig~tlzation
the signal may be time smoothed by a fast
Fourier transform filter program. The
- combination of these filters results in a
relative improvement in signal to noise ratios
which enhance the accuracy of the system.
The excitation of coil lO0 and excitation-
precession of the sample's proton content and
subsequent relaxation/decay produces a received
signal that, after demodulation, controlled gain
amplification, A/D conversion and plotting of
points has the free induction decay (FID) curve
shape C shown in FIGS. 3A and 3B.
FIGS. 3A and 3B, voltage-time traces, show
the elements of a "cycle" of excitation of a
sample and free induction decay. The excitation
- ~ pulse center is taken as tO. The transceiver 104
electranic components do not receive effectively
until saturation effects are overcome at tl.
Then a useable curve is developed. The signal
processing equipment can add or subtract
consecutive waveforms for useful adjustment as
described below.
The digitized F~D curve data of FIG. 3A are
stored in the external computer lO6 where a
program finds the best curve to fit each stored
~ID curve. The FID curve has three primary
component parts shown as A, B and E in FIG. 3A.
The A curve which dominates the first part of
the FID curve is a fast Gaussian, while the B
curve that dominates the middle part of the
curve is a slower modified Gaussian, and E that
~'
WO g3/l~U~ 2 1 2 1 0 2 7 PCT/US92/1024~
16
dominates the later part of the FID curve is an
exponential decay. The fast Gaussian and
exponential portions are associated with the
relatively immobile and mobile portions,
respectively, of the polypropylene samples. The
slower modified Gaussian is a transitlon zone
between the two. The determination of the types
of curves which make up the FID curve is
important because once the curves are known they
can be extended back to a time origin (shown as
Ao, Bo and Eo at tO, i.e., excitation of a Cycle
l), which is close to the center of the
transmitted burst signal. This is important
since the saturation effects of the instrument's
electronic components (which occur during and
after the burst signal to tli make direct
measurements of the intercepts inaccurate. The
portion of the curve of interest extends from tO
to t4 beyond which the curve i8 too small to
matter and the electronics needs recovery time
to prepare for the next cycle ~beqinning with a
pulse centered at tO#2).
The digitized FID curve data of FIG. 3B is
storèd iA the external computer 106 where a
program finds the best curve to fit each stored
F D curve. The FID curve C has three primary
component parts shown as A, B and E in FIG. 3B.
The A curve which dominates the first part of
the FID curve is an Abragam curve while the 8
curve that dominates the middle part of the
curve is a slower Gaussian (i.e. slower than the
Gaussian component of the Abragam), and E that
dominates the later part of the FID curve is an
exponential decay. The Abragam and exponential
portions are respectively controlled by bound
and unbound proton content of the samples,
(e.g., (l) water or hydration-molecules and
~ t 93/10468 2 1 2 1-0 2 7 PCI`/I 'S92/1024~
17
other water (moisture) content of the sample
mass, (2) crystalline and amorphous contents
where they both occur, including mixtures of
highly and lightly polymerized materials and (3)
components of a mixed elastomer-polymer). The
slower Gaussian is a transition zone ~etween the
two and is controlled by protons which are
partially bound or those in a state of becoming
unbound. The determination of the type of curves
which makes up the FID curve C is important
because once the curves are known they can be
extended back to a time origin (shown as Ao, Bo
and E0 at t0, i.e., excitation of a Cyc-le 1),
which is close to the center of the transmitted
burst signal. This is important since there are
saturation effects of the instrument's
; electronic gear which occur during and after the burst signal to tl. During this time
measurements cannot be accurately taken, yet the
area of interest under the curve, which is a
measure of the number of nuclei in the sample,
extends from t0 to t4 beyond which the~curve is
too small to matter and the electronics need
recovery time to prepare for the next cycle
(beginning with a pulse centered at t#02).
Since the~system is offset from resonance,
the recèived signal is the product of the in-
resonance FID and a cosine term at the offset
frequency. Operation at resonance is difficult
since many factors can bring the system out of
resonance easily whereas operation at near
resonance is relatively easily achieved and
maintained. The near resonance operation of the
- ~ system still yields good results since physical
variations cause only minor secondary effects.
Referring to both FIGS. 3A and 3B, each
(sub) cycle goes on to t5 to allow for recovery
~ .
W093/1~U~ 2 1 2 1 0 2 7 PCT/US92/1024~
18
i.e., essentially full relaxation of the protons
of the sample before beginning a new transmit
signal burst (tO#2). Typically, an excitation
pulse interval is five to ten microseconds, the
tO-tl time is five to fifteen microseconds (the
shorter the better), tl-t2, where effécts due to
the relatively immobile structure are
predominant, is five to fifteen microseconds
duration; t2-t3 is transition region of fifteen
to twenty-five microseconds duration,
characterized by phase cancellation effects; and
t3-t4 is a region of fifty to five hundred
microseconds, where more mobile (exponential)
structure predominates. There is a final t4-t5
region, to allow recovery (re-equilibration) of
the sample material, which can be on the order
of hundreds of milliseconds up to several
seconds.
The entire FID curve is composed of several
major components as shown in Equation E~. l as
applied to polypropylene and like materials,
although as described there may be fewer
components on some occasions. Furthermore,
fùnctions other than the fast Gaussian or
Abragam, slower modified Gaussian or
exponential may~be present in addition to or in
lieu of one or more of these components, e.g.
ce-lkt) , where X lies between O and l, or
between l and 2, and these additional components
can have a strong influence on the overall FID
cur~e in early, mid or late time portions of a
decay cycle.
E~- 1 F(tj = {cos ~f)t~llfast Gaussian or
Abragam~ ~ lslower modified Gaussian or slow
Gaussian] + [exponential]~
The entire curve is fitted by an iterative
process based upon the Marquardt-Levenberg (M-L)
~`- g3/l~K~ 2 1 2 1 0 2 7 PCT/US92/1024~
19
approximation technique applied automatically
through a structured implementation in software.
This technique is used to determine the
magnitude of all the parameters, constants,
frequencies, etc. contained in Eq. 1 whlch best
fit the FID curve - to some given erro'r range.
This is an iterative technique where the entire
curve is determined at once. The M-L technique
is documented in the following references: Ind.
Appl. ~ath., vol. ll,`pp. 431-441 by D.W.
Marquardt 1963; Data Reduction and Error
Analysis for the Physical Sciences ~New York,
McGraw Hill), Chapter 11 by Philip R. Be~ington
1969; and The State of the Art in Numerical
Analysis ~London: Academic Press, David A.H.
Jacobs, ed 1977), chapter III.2 by J.E. Dennis.
As applied to the measurement regime of interest
he~rein in a preferred embodiment of the present
invention the selected parameters are the y-axis
~ 20 intercept ratios and time constants,
temperature coefficients, frequency terms and
- other parameters described below.
It was established that there are
approximate limits placed upon the M-L technique
which constrain the coefficients to a limited
range for the types of samples described herein.
If the technique is not converging after a given
number of iterations, 30 in the preferred
embodiment, that sample is discarded. In
30` addition the technique may fail when the system
is too close to resonance.
Once the equation of the FID curve is known,
each component can be extrapolated back to t0 to
establish the intercept of each said component.
The resulting data utilized in the computer
106 (FIGS. 1-2) is the equation for the FID
curve as composed of the three ~excluding the
WO93/10468 2 1 2 1 0 2 7 PCr/~lS92/1024~
~ 20
cosine term) components shown in Eq. 1. Each of
these curves ~and their intercepts) has been
experîmentally related to the same nuclei of
interest. In particular, when the FID curve
equation is determined, the following ratios of
the y-axis intercepts of the indicate'd curves
are formed: the exponential/fast Gaussian(or
Abragam), (R1), and the modified Gaussian/fast
Gaussian(or Abragam), (R2). The ratios (R1) and
(R2), the cross product (R12), together with the
- squares of these ratios (R11) and (R22), the
qecay times for each of the three curve
components, the product temperature and the ~-
cosine frequency term of the modified Gaussian
~ lS form a ten dimensional model. A regression
-~ analysis relates these~ten terms to the
isotactic index (xylene solubles) of the
polypropylene or to density in polyethylene, or
to the same nuclei of interest wherein some are
tightly bound, others loosely bound and some in
between for other material. The results are
independent of the amount of sample, and the
gain of the system, which obviates the need to
measure these physical quantities.
2S The M-L iteration process performs the curve
fitting by minimizing the Chi-Squared function
(the~sum of the squared differences between the
measured data points and the data points from
the derived equation), a technique well known in
the art.
Calibration of the system is accomplished by
measuring a number of known samples and using
the M-L technique to derive the model equation
constants associated with each known sample.
Regression analysis is then performed to derive
the coefficients relating the various model
equation constants to the desired measured
' ~:
~ g3/10468 2 1 2 1 0 ~ 7 Pcr/~1s92llo2~ l
quantity or quantities. A calibration equation
is then prepared using these coefficients.
In operation, a FID is obtained from the
test sample and by the M-L technique the
constants of the model equation are determined.
These constants are then input to the'
- calibration equation and the desired parameters
are calculated.
The data can be used as a QC type
measurement or as an on-line control parameter
which is fed back to control a process, back in
line IPL (FIG. 1~ or related equipment.
The form of the input operating parameters
of the system can be wide reaching to include
previously stored parameters in PROMs or ROMs or
in magnetic storage media such as disks or tapes
or inputs sent in over telephone line and modem
109. The generation of the excitation pulse can
- be accomplished`with many techniques including a
~ 20 coil or antenna arrangement. The steady magnetic
;~ field can be generated by electromagnets,
permanent magnets, electromagnets with
superconducting winding or other standard
techniques of generating magnetic fields.
FIG. 4 is an expanded flow chart showing the
steps of measurement to establish effective
industrial measurement. First a single free
induction decay curve C is established to see if
the sample area is clear (Quick FID) in an
abbreviated cycle of attempting to establish a
- curve C. If the sample region is not clear (N),
measurement is interrupted to allow valve V2
(re)opening and operation of jets J and gravity
to clear the region. A new Quick FID step
3~ establishes clearance. Then a sample is
admitted by closing valve V2, opening valve Vl
and making such adjustments of probe P and line
:
WO93/10468 2 1 2 1 0 2 7 PCr/US92/1024~
22
LI as may be necessary (if any) to assure sample
acquisition. Jets J adjust and stabilize the
new sample.
An electronic signal processing apparatus
baseline is established in 3-4 cycles ~each
having (+) and (-) sub-cycles with add~ion of
(C+) and (C-) to detect a baseline offset and
compensate for it). It would be feasible to
avoid this baseline offset determination and
simply deal with it as an additional (i.e.,
eleventh) dimension in the M-L analysis, but
this would increase iteration time.
Further adjustment is established by coils
124 to adjust H0 and this is enabled by ten to
twenty field check cycles of FID curve
generation. The (C-) FID is subtracted from the
(C+) FID, i.e., the absolute C values are added
to obtain a workable digitized FID signal -
which has a maximum value at resonance. H0 is
adjusted via coil current generator 117 and
coils 124 until such maximum is achieved, and
then H0 is changed to offset the system a known
amount from resonance. These measurements are
taken in a reliable region for such purpose,
i.e., the exponential region of t3-t4 [the above
baseline measurements are also taken there~.
Adequate field adjustment is usually made in
less than seven cycles.
Then multiple measurement cycles are usually
performed to obtain useable data. Each cycle
involves a modulated
transmission/reception/flash A-D conversion, and
storage of data. The curves are then averaged
for M-L curve fitting, and the above listed
intercepts and ratios are established. Similar
cycles, but somewhat abbreviated can be applied
for Quick FID, field check and-baseline
T`-`~93/l~u~ 2 1 ?1 0 2 7 PCT/US92/1024
23
correction purposes. Each of the sub-cycles
l(+) and ~-)] of each such cycle involves a
capture and utilization of thousands of FID
points in data reduction. Each sub-cycle occurs
S on the order of a second and the number of such
sub-cycles employed depends on the deg~red
smoothing and signal to noise ratio (S/N);
generally S/N improves in a square root
relationship to the number of cycles
accumulated.
As noted in above cited Dechene et al.
patents, in requiring greater accuracy and
reliability, sample tube composition can distort
readings. If glass is not used (and it is
lS preferred to avoid glass in industrial usage),
then the replacement should not contain
hydrogen. But fluorocarbons can be effective
in several applications since signals from
fluorine appear far from resonance. These
signals can be distinguished from hydrogen at
the levels of sensitivity required and if
desired can be filtered (or distinguished). In
other cases of higher sensitivity measurements,
e.g., for gauging relative proportions of
amorphous and crystalline species in mixtures
thereof, the sample container should be glass or
non-protonic ceramic. In some instances,
however, fluorocarbon or reinforced fluorocarbon
can be used acceptably for polymer measurements.
In all such cases the point is to avoid sample
containers with species that can couple with
transmitted energy and generate a FID decay
curve mimicking the samples.
Since the regression analysis involves ten
variables residing in an eleven dimensional
space, the results cannot be graphed. But the
measurements obtained from the model represented
WO93/10468 2 1 2 1 0 2 7 Pcr/usg2/lo24~
24
in the present invention can be favorably
compared to the results obtained from accepted
off-line measuring techniques.
FIG. 5 shows FID's taken for polypropylene
samples of varying isotactic indices. The curves
are are sufficiently distinct and the~resulting
- three component parts found (by M-L) for each
curve result, after regression, in values from
which actual polypropylene isotactic index
(xylene solubles~ is computed from the
calibration equation.
FIG. 6 shows a calibration curve of such
polypropylene xylene solubles, as measured by
the present invention (y axis) vs. standard
methods, illustrating the efficacy of the
present invention.
- It will now be apparent to those skilled in
the art that other embodiments, improvements,
details, and uses can be made consistent with
the letter and spirit of the foregoing
disclosure and within the scope of this patent,
which is limited only by the following claims,
construed in accordance with the patent law,
includi~g the doctrine of equivalents.
What is claimed is:
,
