JPH10213557A - Agricultural product nondestructive inspection method utilizing hydrogen nuclear magnetic resonance, and magnet box and device therefor - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a nondestructive inspection method which utilizes hydrogen magnetic resonance and can be used practically for agricultural product inspection, and magnet box and device for the inspection. SOLUTION: A magnet box 2 of soft magnetic substance is formed on the inner upper and lower surfaces of a case 2 of high permeability made of soft magnetic substance on the surfaces and corners of the flat plate type permanent magnets 6 and 8. Also pole faces 16 and 18 of high permeability made of soft magnetic substance, shimming plates 20 and 22, and corner steel are attached fixedly onto the surfaces and corners of the permanent magnets 6 and 8, and spare poles capable of increasing the uniformity of static magnetic field B0 , increasing performance, and passing specimens (agricultural products) between the permanent magnets 6 and 8 are formed. Then an RF/sensitive coil 34 is set so as to form the dynamic magnetic field B1 and sense the resonance signals from the specimens.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a method for nondestructive testing of agricultural products, a magnet box (sensor) and an apparatus for the same, and more particularly, to a nondestructive method for determining water content and sugar content in agricultural products by utilizing the principle of hydrogen nuclear magnetic resonance. The present invention relates to a method for nondestructive testing of agricultural products, a magnet box, and a device which can be measured continuously while there is.

[0002]

BACKGROUND OF THE INVENTION The development of a non-destructive moisture content (or sugar content) measuring device for agricultural products (or processed foods) is intended to automate drying and storage facilities, processing facilities or food processing processes. Accurate moisture content measurement of agricultural products plays a major role in determining the effective use of energy and quality during drying and storage.

[0003] Non-destructive continuous measurement of the water content during the drying process can prevent poor drying such as overdrying or undried, and can achieve uniform drying. Over-drying results in loss of produce and waste of energy, and incompletely dried produce rots during storage or leads to microbial degradation, leading to loss of quality.

There is an urgent need to develop a moisture content measuring device used for such a purpose, but at present it depends on a device for taking a certain sample and destructively measuring the moisture content. Research on the development of such a nondestructive and continuous moisture content measuring device is actively proceeding in various countries in Europe and the United States, but the use of near-infrared rays or the measurement of moisture content by electrical physical property measurement has become mainstream.

[0005] In the case of rice, it is essential to measure the moisture content of the rice in order to minimize the loss during storage at homes and shipping sites, and methods for measuring the moisture content of rice include the Obun method. Non-destructive methods such as a direct method and a method of measuring heritability change according to the moisture content of rice, and a method of measuring a change in electric resistance.

Recently, non-destructive measurement of the water content of rice has been required in treatment facilities after harvesting large-scale rice, and a method for measuring changes in electrical resistance and heritability has been proposed. The measurement of water content by the method has a small number of samples and is inferior in representativeness, and the measurement of the water content by the heritability has problems such as being greatly affected by the density of the product.

On the other hand, nuclei such as ¹ H, ¹³ C, ¹⁹ F, ³¹ P and the like in which the total number of protons or neutrons is odd rotate about one axis, and as a result, the rotational moment (An
gularMoment) and magnetic moment (Magnetic Moment)
The nucleus having such a moment absorbs energy while causing resonance in a magnetic field that changes at the same frequency as the axis rotation speed. So to speak, nuclear magnetic resonance
Magnetic Resonance (NMR) phenomenon occurs.

Before describing a non-destructive inspection method of agricultural products using NMR and a preferred embodiment of the sensor and the apparatus thereof according to the present invention, the principle of hydrogen nuclear magnetic resonance will be schematically described as follows.

If a substance contains a large amount of water, 99.98% of the hydrogen nuclei forming water molecules
Can be considered as one small magnet that exists in the ¹ H state and has a magnetic moment. Since the hydrogen nuclei are arranged arbitrarily, they do not have magnetism as shown in FIG. 1 when viewed as a whole.

If the substance is placed in a strong magnetic field (B ₀ ) of several thousand gauss, as shown in FIG. 1, one rotation magnetic moment of the nuclear hydrogen nucleus is aligned in the direction of a strong external magnetic field, and The hydrogen nuclei are precessed based on the direction of the magnetic field.
on) Exercise. Each speed of the precession has a unique numerical value according to the type of nucleus, and changes in proportion to the strength of the external magnetic field.

That is, hydrogen atoms precess at a speed of 4.26 MHz in a 1,000 gauss magnetic field. At this time, the sum of the vectors of the magnetic moments of all hydrogen nuclei in the direction of the magnetic field (Z axis) is defined as the macroscopic magnetism.
ation) or total magnetization M.

In the direction perpendicular to the strong magnetic field B ₀ (X axis), another relatively weak radio frequency (radio
oscillating magnetic field
If applying a field) B ₁ to the same frequency as the nucleus velocity of precession of the protons, undergoes resonance phenomenon, absorbs energy, tilting the macroscopic magnetization M as in FIG. 2 in the Y direction. B
_{If one} magnetic field is extinguished, the tilted M returns to the original position of the Z-axis and emits energy as shown in FIG. 3 to the outside, and a resonance signal (Resonance Signal) diverges.

[0013] The resonance signal is extracted by a sensing coil, and a chemical, qualitative and quantitative analysis of a substance to be investigated can be performed. First, the frequency of the resonance signal can be calculated using the magnetic field strength of the permanent magnet. A resonance signal used to find chemical information of a substance can determine a desired signal variable through many processes of electromagnetic signal processing.

The signal variables typically used include chemical shift and relaxation time. Even for the same hydrogen atom, the resonance frequency differs slightly depending on the bonding environment. Since the degree of this difference is called a chemical displacement, tetramethylsilane (TMS, [Si (C
The difference (ppm) is displayed with reference to the hydrogen resonance frequency emitted from H ₃ ) ₄ ]), and the composition of the chemical substance can be analyzed using this numerical value.

Mx, My, Mz on the X, Y, Z axes
While returning to the Z axis macroscopic magnetization vector M which is inclined is original position, the time Mz with two types of relaxation time to relax while returning to its original size M ₀ T ₁ (Spin- L
attice Relaxation Time), refers to the time that Mx and My relaxes while disappear T ₂ and (Spin-Spin Relaxation Time). If this relationship is represented by a mathematical expression, it is as follows.

[0016]

(Equation 1) .

T ₁ is the relaxation time when the nucleus relaxes with the environment surrounding the nucleus, and T ₂ is the time between the nuclei. T ₁ and T ₂ have different values according to the state or type of the substance, and T ₂ is always equal to or smaller than T ₁ .

The T ₂ value is short for a solid state and long for a liquid state. The time constants of water and oil have different numerical values.

The advantages of the above-mentioned nuclear magnetic resonance are that the sample to be examined can be maintained nondestructively without chemical or physical damage, and re-measured at regular time intervals when necessary. It can be used for diagnostic purposes in the medical field and analytical applications in the physics and chemistry fields.

Before such a nuclear magnetic resonance (NMR) apparatus was put to practical use, infrared spectrometers and mass spectrometers were mainly used in various chemical fields. Due to the development of chemicals, biology, medicine, etc., since the latter half of 1970, as the most powerful means for structural, quantitative analysis and chemical analysis of chemical components existing in substances with the principle of nuclear magnetic resonance, Widely used for. It is mainly used for chemical qualitative and quantitative analysis by observing nuclear magnetic resonance of hydrogen and carbon.

In addition, the nuclear magnetic resonance signal of hydrogen, which is a monovalent atom, in the structural analysis and quantitative measurement of chemical components using nuclear magnetic resonance appears most strongly, and about 99.000 of all hydrogen atoms existing in nature. 98% are such monovalent atoms.

Among them, water (H ₂ O) emits the strongest hydrogen resonance signal among the chemical substances composed of monovalent hydrogen ( ¹ H). Therefore, the measurement of the water content using nuclear magnetic resonance is as follows. It is relatively easy compared to quantitative measurements of other chemicals.

According to the present invention, components such as water content and sugar content can be nondestructively measured in agricultural products using the hydrogen nuclear magnetic resonance (NMR) principle having the above-described functions and effects. The aim is to provide a method.

It is another object of the present invention to provide a practical hydrogen nuclear magnetic resonance sensor suitable for nondestructive testing of agricultural products.

[0025]

According to a first aspect of the present invention, there is provided a non-destructive inspection method for agricultural products using hydrogen nuclear magnetic resonance. After receiving a resonance signal by magnetic resonance, signal processing is performed to obtain a pure water resonance signal or a sugar content signal to measure the water content and the sugar content of the agricultural product.

The magnet box for nondestructive inspection of agricultural products using hydrogen nuclear magnetic resonance (NMR) according to claim 2 is a flat permanent magnet on the inner upper and lower surfaces of a box-shaped case (4) made of a material having high magnetic permeability. (6) (8) is adhered and fixed so that the NS magnetic poles face each other, and an air pole (10) is formed between the permanent magnets (6) and (8). The pole faces (16) and (18) having high magnetic permeability are adhered and fixed, and the shimming plates (20) and (22) having high magnetic permeability are adhered and fixed to opposing edges of the pole faces (16) and (18). Corner steel with high magnetic permeability (2)
4) is adhered and fixed to improve the magnetic field strength and uniformity of the static magnetic field (30), and an RF / sense coil (34) is installed in the air pole (10).

According to a third aspect of the present invention, there is provided an apparatus for nondestructive inspection of agricultural products using hydrogen nuclear magnetic resonance, wherein a high frequency pulse is applied to the sample, and a transducer is provided at one end of a nuclear magnetic resonance magnet box (2) for obtaining a resonance signal. 42), a preamplifier (54) for amplifying the resonance signal detected by the magnet box (2) from the sample is connected to one end of the transmitter / receiver, and the preamplifier ( The receiving and detecting unit (5)
6), amplify and detect the signal amplified from the preamplifier (54) to obtain a sufficiently large pure water resonance signal, and connect an oscilloscope (58) to the output of the reception and detection unit (56). Connected to display the pure moisture resonance signal as an image, and a high-frequency power amplifier (52), a high-frequency switching unit (50), a pulse programmer (48), and a computer (46) for controlling the system at one end of the transmission / reception converter (42). ) Are connected one after another, and a computer (36), a pulse programmer (48), and a high-frequency switching unit (5) are connected.
0) to the receiving and detecting unit (56), respectively,
During the transmission period in which the high frequency pulse is applied by the sensing coil (34), the reception is interrupted.

According to a fourth aspect of the present invention, there is provided a magnet box for nondestructive testing of agricultural products utilizing hydrogen nuclear magnetic resonance, wherein the inclined surface of the corner steel (24) is concave.

[0029]

DESCRIPTION OF THE PREFERRED EMBODIMENTS In order to achieve the above objects and the like, the present invention provides a magnet box (Magnetic box) as a sensor for generating a static magnetic field.
console) and an RF / RF for transmitting a high-frequency signal (Radio Fuency) to generate a dynamic magnetic field and detect a resonance signal.
A sensing coil (RF coil) and a signal processor (RF signal pr)
ocessor) and a control computer, and the quality can be measured without any influence on the density of agricultural products.

On the other hand, a sensor of a nuclear magnetic resonance apparatus used for a general medical diagnostic instrument, a physicochemical analysis instrument, and the like uses IT (1
Since a superconducting magnet is used to generate a magnetic field of more than 0,000 gauss, the construction and operation thereof are complicated and the maintenance cost is high. A practical magnet box conforming to the present invention is constructed by using a permanent magnet capable of forming a front and rear magnetic field. A pole face and a corner steel are provided on the surface and side edges of the permanent magnet.
) And a shimming plate to equalize the strength of the magnetic field to greatly improve the sensitivity and bandwidth.

Also, since the circuit must be appropriately changed so as to match the resonance (resonance) frequency range in a low magnetic field (around 1,000 Gauss), the tuning circuit should be designed in consideration of the size of the object to be measured. In order to send a resonance signal to the sample in the permanent magnet as a static magnetic field in the present invention, a pulse having a specific period is formed, and a 4.1 MHz pulse is applied to this signal for resonance. In addition, a dynamic magnetic field for nuclear magnetic resonance is formed by obtaining a high-frequency pulse signal using a switching circuit, amplifying the signal, and supplying the amplified signal to an RF / sensing coil, and then applying the RF / sensing coil to the RF / sensing coil by a transmission / reception converter. The transmission of the high-frequency pulse signal and the reception of the resonance signal are intermittently performed. To obtain a water (H ₂ O) signal from the resonance signal detected by the RF / sensing coil, a pre-amplifier (PRE-amplifier) is low-noise amplified to separate and remove unnecessary noise, and then subjected to signal processing and phase detection. Then, an image is displayed on an oscilloscope, and this signal is subjected to multi-stage signal processing.

There are many types of permanent magnets depending on the materials used. The ceramic 8 series can form a low magnetic field with permanent magnets at low cost, but Nd-Fe-B
(Neodymium-Iron-Beron) series and rare earth series permanent magnets can form high magnetic fields with expensive permanent magnets,
When an air gap, which is the distance between permanent magnets, is formed to 100 mm using such permanent magnets, Nd-
Even if Fe-B has a small volume, it is 4,000 to 5,000.
A high magnetic field strength of 0 Gauss can be obtained, and in the case of the ceramic 8 presented as an experiment in the present invention, 800 to 1,
A magnetic field strength of about 500 Gauss can be obtained.

Therefore, it is possible to develop an apparatus for measuring the internal quality of agricultural products which is inexpensive and has a low maintenance cost by using permanent magnets. The material of such a permanent magnet is μ (magn
Since the μ value is expressed by a nonlinear relationship (BH curve) that changes with the strength of the magnetic field and slightly changes depending on the ambient temperature, the ceramic 8 series presented in the present invention Uses a BH curve based on normal temperature (20 ° C.) as shown in FIG.

FIGS. 5 and 6 are longitudinal sectional views of a sensor (magnet box: 2) of the present invention for detecting a resonance signal of the sample after applying a high-frequency pulse to the sample to cause resonance, and a case (magnet box: 4). )
Can be considered as a substantially square shape, and it is easier to design, manufacture and install than a curved surface.

The permanent magnets 6 and 8 having a strong magnetic force are attached and fixed to the upper and lower surfaces of the conductive case 4, but the upper and lower permanent magnets 6 and 8 are slightly separated from each other to allow a sample to be placed or passed. Are formed, and through holes 12 and 14 are formed in the left and right side plates of the case 4 so that the sample can pass through the hole.

Each of the permanent magnets 6 and 8 uses a rectangular plate-shaped magnet (magnetic shell) in which the N and S magnetic poles are formed vertically, and their relative surfaces have different polarities in the Z direction (vertical direction). Create a strong magnetic field.

The magnetic flux of the magnetic field (magnetic f
lux) is determined by the homogeneity of the magnetic field, and the magnetic flux and the homogeneity are determined by the permanent magnets 6, 8 and the case 4
Case 4 has a high investment rate and has a low magnetic loss by using a soft magnetic material with a small hysteresis loss because it is influenced by the current state of materials and the standards. Permanent magnets 6,8 which reduce the influence of external magnetic fields and eliminate shocks
In the case of, there is a characteristic that the magnetic flux of the magnetic field becomes stronger as it goes to the end or corner, so the magnetic flux density of the magnetic field is non-uniform, and such non-uniformity narrows the bandwidth of the resonance signal that can be sensed. Since the decay time of the resonance signal is reduced and the overall sensitivity of the sensor 2 is reduced, uniformity of the magnetic field strength is absolutely required.

Therefore, in the present invention, a soft magnetic body made of the same material as the entire case 4 facing the permanent magnets 6, 8 is used. The uniformity of the magnetic field can be improved by forming the pole face.

The permanent magnets 6, 8 having relatively high magnetic field strength are provided with softening shimming plates 20, 22 and corner steel 24 at the ends and corners of the magnets. Form a further improved static magnetic field B ₀ .

In the present invention, the shimming plates 20 and 2 are used.
2 and the corner steel 24 are structures presented to improve the strength and uniformity of the magnetic field. The results obtained by analyzing the magnetic field distribution of the permanent magnets 6 and 8 are pole faces 16 and 18.
The direction of the magnetic field facing the outside at the end of the pole is directed toward the center of the air pole 10, thereby maximally suppressing the external leakage of the magnetic field and improving the uniformity.

The performance of the magnet box used in the NMR sensor of the present invention is determined by the strength of the magnetic field to be formed (magnetic flux) and the homogeneity of the magnetic field. They are,
Since it is complexly affected by the material of the permanent magnet, the shape of the magnet box, and the standard, a lot of trial and error is required to design a magnet box having desired performance.

Even if the material of the permanent magnet is inexpensive, such design and manufacture through trial and error entails an increase in initial costs and a considerable increase in the development period. Therefore, if trial and error is simulated on a computer by applying the finite element analysis technique, not only can the development cost be reduced and the development period be shortened, but also an appropriate design standard can be selected. A great deal of help can be gained when designing a new shape magnet box, and the present invention applies such a finite element technique to optimize the magnetic field strength and uniformity of the permanent magnet.

The above-mentioned finite element interpretation technique has many applications for solving problems such as statics of structures, electromagnetic field heat, and fluids, and many programs and techniques are known. ABAQUS, NASTRA are the most frequently used finite element interpretation tools (computer programs).
N, FLUENT, ANSYS, etc., and the general process of finite element analysis is as follows.

(1) Geometric shape design of magnet box using CAD (design program using computer)
afting) (2) Input of characteristic samples such as magnet and iron materials (3) Finite element generation (mesh generation) (4) Finite element analysis (5) Result analysis (6) If design criteria are not satisfied Return to (1) (7) Precision design of magnet box using computer design program (eg, CAD).

According to the present invention, ANSYS which is a finite element analysis tool for manufacturing electromagnetic field thermal fluid for two-dimensional and three-dimensional shapes is used.
Finite element analysis was performed using (SASI-SWANSON Analysis System, Inc.) 50. In the present invention, the basic shape of the magnet box was selected from a box shape in consideration of manufacturing convenience, and a total of 11 design factors were selected as shown in Table 1.

In the present invention, the characteristics of the design factors and the influence of the design factors and the like on the strength and uniformity of the magnetic field are analyzed by using a finite element interpretation technique, and the optimum magnetic field is selected by appropriately selecting these. And a magnet box with uniformity were designed. In Table 1, the air gap is the limiting factor in the design of the magnet box in the space where agricultural products (samples) are placed.
acts on t).

[0047]

[Table 1] .

The governing equations used in the finite element to interpret the strength and uniformity of the magnetic field of the magnet box are Maxwell equations (1) as to the electromagnetic field.
) Was used.

[0049]

(Equation 2) .

The cross-sectional structure of the hydrogen nuclear magnetic resonance sensor 2 manufactured to test and verify the uniformity of the permanent magnets 6 and 8 of the present invention has a structure as shown in FIGS. The room temperature BH of the case 4 is the same as that of FIG. 7 and forms the framework of the corner steel 24, the shimming plates 20, 22, the pole faces 16, 18, and the sensor 2, and also serves as a magnetic path for magnetic shielding. It is as follows.

[0051]

[Table 2] .

In the front and rear plates 26 and 28 which are fixed to the case 4 by adhesive bonding or bolts, through holes 12 and 14 are formed for taking in and out of the sample. Since the strength of the magnetic field is weaker than that of the side plates 30, 32, the lateral width K of the shimming plates 20, 22 located on the side of the through holes 12, 14 is greatly extended so that the magnetic leakage can be compensated.

The front and rear side plates 26 and 28 are made of a material that conducts electricity while being a non-magnetic material such as an aluminum plate. 28 and the case 4 are electrically connected to each other so as to be electrically shielded.

On the other hand, a hard tube 36 made of a non-magnetic material and having a low electric conductivity is fitted into the RF / sensing coil 34, and both ends of the tube 36 project into the through holes 12 and 14 of both side plates. Tunnel 38 through which the sample can pass through
The thickness of the tube 36 is formed to be as thin as approximately the cross-sectional diameter of the R / sensing coil 34 so that the internal cross-sectional area of the sample passage tunnel 38 is not greatly reduced.

In the above case, the RF / sensing coil 34 can be wound around the outer surface of the tube 36 to prevent the coil from hanging down.
4 supported.

One end of the RF / sense coil 34 is grounded to the case 4 having conductivity, and the other end of the RF / sense coil 34 is connected to the series resonance circuit 40 and the transmitter / receiver 42 one after another. Insulated with insulating material.

At the bottom of the tunnel 38, a continuous or discontinuous conveying means such as a belt conveyor conveyed by a variable speed driving device is installed to continuously or discontinuously control the water content of the produce and the sugar content of the fruit. To measure,
The material of the transporting means is preferably a non-magnetic material and is preferably an insulating material.

On the other hand, in order to measure the magnetic field strength of the sensor 2 of the present invention, a magnetic flux density meter (gauss meter, model name: MG-
4D, Production company: Walkers Co., Equipment resolution: ± 0.5
Gauss or ± 0.05%) using 100 mm diameter
After examining the magnetic field strength in the circle of the above, the uniformity of the magnetic field was calculated by the equation (2), and as a result, a slight error was observed between the measured value and the predicted value. Table 3 below shows a comparison value between the actually measured difference and the result difference between the predicted value and the magnetic field strength and uniformity of the sensor 2 in the XZ plane and the YZ plane.

[0059]

(Equation 3) .

[0060]

[Table 3] .

Although there are many causes of such an error, first of all, the finite element analysis is a method in which an object is originally decomposed into finite elements and an approximate value is obtained based on this.
This is due to an error in the size, shape, and the like of the decomposed finite element.

Since the BH curve used in the finite element analysis has characteristics at room temperature (20 ° C.), there is an error due to the ambient temperature of the magnet at the time of the actual measurement. However, in the case of the permanent magnet (ceramic 8) used in the present invention, the change in the strength of the magnetic field due to the temperature is about 0.19% loss and about 1 ° C., and the temperature at the time of the actual measurement is close to room temperature. It can be said that the error due to the temperature is not so large.

FIG. 8 is a graph of analyzing the change in the magnetic field and the uniformity by performing the finite element interpretation while changing the specifications of the shimming plates 20 and 22 in order to further improve the uniformity of the permanent magnets 6 and 8 according to the present invention. The shimming plates 20, 22 newly created through finite element analysis
The XZ width (d) was 20.3 mm and the YZ width (K) was changed to 17.8 mm.

In the case of the above-mentioned newly designed sensor, the uniformity is from 1199.54 ppm to 34% in the XZ plane.
5.18 ppm, 396.84 for the YZ plane
ppm to 115.74 ppm, and the field strength decreased slightly from 958 gauss to 956 gauss.

In the above state, corner steel was attached to four corners on the XZ plane and finite element analysis was performed. As a result, the magnetic field intensity increased from 956 gauss to 962 gauss, and X
The -Z plane uniformity shows a result further improved to 103.23 ppm, and the YZ plane has the same result as before because there is no design change.

On the other hand, according to the present invention, the pole faces 16, 18
In order to verify the improved uniformity by realizing the above-described improvement of the uniformity of the magnetic field by installing the shimming plates 20 and 22 on each edge of the magnetic field of the magnet box without the shimming plates 20 and 22, The homogeneity was analyzed by finite element interpretation, and Table 2 was used as it was for the specification of the magnet box used in the analysis.

Table 4 shows a comparison between the case where the shimming plates 20 and 22 are installed and the case where the shimming plates 20 and 22 are not installed. Is periodically improved, but the strength of the magnetic field is seen to decrease considerably.

[0068]

[Table 4] .

That is, in order to increase the strength of the magnetic field reduced due to the installation of the shimming plates 20 and 22, the thickness (e), length (l) or width (a) of the permanent magnets 6 and 8 must be increased. It is understood that the user needs to change the information.

In the present invention, since the thickness (e) of the permanent magnet is sequentially increased so as to compensate for the strength of the magnetic field while minimizing the influence on the uniformity of the magnetic field, the strength of the magnetic field is changed. Table 5 shows the results of the changes. The thickness of the permanent magnets 6, 8 was increased by 10% (7.6 mm) to design a magnet box close to the strength of the magnetic field without the shimming plates 20, 22. can do.

[0071]

[Table 5] .

The shimming plate 20 used in the present invention,
The effect of 22 and corner steel 24 is even more pronounced when compared to a magnet box of the same uniformity designed without such structures 20,22,24.

That is, in order to improve the uniformity of the permanent magnets 6, 8, the width (a) and the length (i) of the permanent magnets are reduced to 50% and 100%, respectively, without using the shimming plates 20, 22 and the corner steel 24. Table 6 shows the results of performing the finite element analysis while increasing the number.

[0074]

[Table 6] .

From the above Table 6, it is confirmed that the uniformity is considerably improved if the width and the length of the permanent magnets 6 and 8 are increased by 50% and 100%, respectively, but the intensity of the magnetic field is sharply reduced. can do.

When the width and length of the permanent magnet are increased by 1.5 times, the same magnetic field strength as that proposed in the present invention can be obtained, and the uniformity is not better. The performance is shown, and when it is increased by a factor of 2, the uniformity of the magnetic field shows almost the same performance as that of the present invention in the case of the XZ plane.
It can be seen that the magnetic field strength is reduced by 112 gauss and the uniformity of the YZ plane cannot reach the present invention. To compensate for the reduced magnetic field strength, the permanent magnet without the shimming plates 20 and 22 and the corner steel 24 is used. Table 7 shows the results of increasing the thicknesses of the magnets 6 and 8.

[0077]

[Table 7] .

In other words, it is impossible to design a sensor exhibiting a desired magnetic field strength and uniformity by further using a permanent magnet material four times or more without the shimming plates 20 and 22 and the corner steel 24. In the present invention, about 20% of the permanent magnet material, the shimming plates 20 and 22, and the corner steel 24, the addition of a soft magnetic material makes it possible to design the sensor 2 exhibiting a desired magnetic field strength and uniformity.

Further, when the size of the permanent magnet is increased, the volume and weight of the sensor are increased and the space and installation are restricted, but such a problem can also be solved.

For synthesizing the experimental results of the sensor 2 of the present invention, shimming plates 20 and 22 and corner steel 24 were used.
Table 8 shows a comparison of sensors having substantially the same magnetic field strength and uniformity between the case where the sensor is used and the case where the sensor is not used.
It became like.

[0081]

[Table 8] .

Looking at Table 8 above, it can be seen that the mink plate 2
When not using 0,22 and corner steel 24,
The thickness of the permanent magnet is increased by 40.9% and the width and the length of the permanent magnet are increased by 100% each compared with the case where these structures 20, 22, 24 are used. It turns out that you have to use it. In the present invention, by using the structures 20, 22, 24 in the sensor 2, 1/6
The desired magnetic field strength and uniformity can be obtained with the material of
The weight and volume can be greatly reduced accordingly.

On the other hand, the air pole 10 of the sensor 2 of the present invention is 15
Since it is 2.4 mm, RF /
The maximum outer diameter of the sensing coil 34 is within about 150 mm.
Since the uniformity of the magnetic field of the RF / sensing coil 34 determines the transmission characteristics of the entire transmission / reception signal, the RF / sensing coil 34 must be manufactured so that the input sensitivity of the resonance signal and the input signal are not attenuated and the magnetic field is uniform.

Further, the air coil must be designed and installed so that the direction of generation of the magnetic field is perpendicular to the Z axis which is the direction of the magnetic flux of the permanent magnets 6 and 8. Because of the demands to prevent sagging, use is made of a relatively thick but hard conductor having good electrical conductivity, for example a conductive wire such as copper or aluminum.

FIG. 9 shows a cross section of the RF / sense coil 34. In order to obtain a uniform magnetic field strength, and to improve the sensitivity of the resonance signal and prevent signal attenuation, the intervals between the coils are made equal and uniform, and the intervals are separated by about 1 to 2 times the diameter of the coil 34. Thus, it is desirable that the self capacitance is not induced.

The RF / sensing coil 34 receives a high-frequency pulse signal (high-frequency AC signal), forms a magnetic field in the sample, generates nuclear magnetic resonance, and senses a resonance signal of the sample relaxed due to nuclear magnetic resonance ( Receive).

The RF / sensing coil 34 can be divided into a coil for applying a high-frequency pulse signal to the sample and a coil for sensing the relaxation resonance signal of the sample, and can have transmission and reception functions, respectively. Sometimes it is impossible to receive a resonance signal, and when a resonance signal is received, it is practically impossible to transmit a high-frequency pulse signal, and there are difficulties and limitations in installation. Can also serve as a transmission / reception function.

The RF / Sensing Coil 34 Used in the Present Invention
Is wound using a copper wire having a cross-section radius of 2.54 mm so that the space has a radius of 60 mm. The interval between the copper wires is set to 3.81 mm, which is 1.5 times the cross-section radius, and the entire winding is performed. The number of times was set to 14, and in this case, the inductance (L) of the coil 34 was calculated to be 7.6 μH according to the equation (3).

[0089]

(Equation 4) Here, the coil length Lg is set to 910 mm, the cross-sectional area A of the coil internal space is set to about 11.204 mm ² , and the permeability in free air (permeability of freespace) is set.
As a result of verifying a value obtained by setting μ as 4π10−7 wb / mA with an impedance measuring device (LCR meter), the value was 7.9 μH at 120 Hz and 7.7 μH at 1 kHz. Theoretically, the L value is independent of the frequency, but as the frequency increases, the L value increases.
The value tends to decrease exponentially.

In the present invention, the air pole 10 is set to 152.4 mm,
When the magnetic field strength of the permanent magnets 6 and 8 was set to 965 Gauss, the high-frequency signal for obtaining the resonance signal was calculated to 4.1 MHz by the equation (4).

[0091]

(Equation 5) .

In order to construct a circuit tuned to 4.1 MHz, the strength of the magnetic field generated in the RH / sensing coil 34 must be calculated, and the calculation formula is as shown in equation (5).

[0093]

(Equation 6) .

To obtain a resonance signal from a sample (sample) magnetized by the permanent magnets 6, 8, 4.1 MHz
A weak magnetic field B ₁ is formed in a direction perpendicular to the magnetic field B _{0 of the} permanent magnets 6 and 8 through the RF / sensing coil 34 so that B
By applying _one magnetic field, the vector M ₀ of the sample becomes X from the Z axis.
Since the inclination is to the −Y plane, the inclination degree θ can be obtained from Expression (6).

[0095]

(Equation 7) .

On the other hand, in order to obtain a magnetic field intensity B ₁ required for nuclear magnetic resonance from the RF / sensing coil 34, a sufficient current can be applied to the RF / sensing coil 34. The high frequency pulse signal applied to the sensing coil 34 is amplified to a strength sufficient to form a high frequency field (RF field).

In the above, the power P applied to the RF / sensing coil 34 is P = I ² R, I is the current, and R is the total impedance including the coil 34 and the conducting wire. Must be amplified.

The air pole 10 of the sensor 2 of the present invention is 15
When the magnetic field strength is 2.4 mm and the strength of the magnetic field is 965 gauss, the frequency of the high frequency pulse signal is calculated to be 4.1 MHz in order to obtain a resonance signal, as described above. In order to resonate at 1 MHz, the resonance circuit has the same resistance (R) -variable capacity (AV
C) The series resonance circuit is formed by connecting the RF / sensing coils 34 in series. Table 9 shows theoretical formulas ideally induced by the series resonance circuit.

[0099]

[Table 9] .

FIG. 11 is a circuit block diagram of the measuring apparatus of the present invention. A pulse programmer 48 for intermittently transmitting and receiving a duty cycle is connected to a computer 46 which controls the apparatus and analyzes a resonance signal.

The transmission / reception operation cycle is determined by a transmission time of an “H” pulse of about 10 μs and an “L” pulse of about 100 ms, and the transmission unit operates during the “H” period of the signal to operate at a high frequency of 4.1 Mz. After the AC signal is generated and amplified, R
Nuclear magnetic resonance is generated by applying the signal to the F sensing coil 34, and during the "L" period of the signal, a resonance signal of the sample is sensed by the RF / sensing coil 34 and detected. Obviously, a resonance signal is obtained, and the transmission / reception operation cycle appropriately varies depending on the target of the sample and the target signal.

A high-frequency switching unit (RF switching unit: 50) is connected to the output of the pulse programmer 48, and the operation cycle of the pulse programmer 48 is "H" while the operation period is 4.1.
A high-frequency power amplifier 52 is connected to the output of the high-frequency switching unit 50 by generating a high-frequency pulse signal of MHz.
The 4.1 MHz high frequency pulse signal is amplified through the RF / sensing coil 34 so as to generate a magnetic field B1 having a sufficient magnitude required for nuclear magnetic resonance.

A transmission / reception converter is connected to the output of the high-frequency power amplifier 52.
(T / R switch or duplexer: 42) to connect the RF / sensing coil 34.

At one end of the transmission / reception converter 42, a preamplifier (Pream
plifier: 54) to amplify the resonance signal sensed from the RF / sensing coil 34 with low noise.

The low noise preamplifier 54 suppresses the amplification of unnecessary noise in the signal because the resonance signal input from the RF / sense coil 34 is as weak as several μv, and the resonance signal is amplified to a large signal. Therefore, it can be said that it is a main circuit particularly as a preceding stage for signal processing.

Since the signal amplified by the preamplifier 45 is a signal obtained by mixing a resonance signal and a carrier of 4.1 MHz,
A reception and detection unit 56 is connected to the output of the preamplifier 54 to obtain a pure water resonance signal of several KHz, which is a necessary signal in the mixed signal.

A digital oscilloscope (58) is connected to the output of the receiving and detecting unit 56 to display a pure water resonance signal subjected to signal processing.

A high-frequency switching unit 50 is connected to one end of the reception and detection unit, and a 4.1 MHz reference frequency is input to remove the 4.1 MHz carrier signal from the received signal. A pulse programmer 48 is connected to one end of the terminal so that the operation of the receiving unit is interrupted and protected in the transmission state.

The oscilloscope 58 is a digitizer.
(digitizer: 60) is built in and converts a pure moisture resonance signal inputted in an analog form into a digital signal to detect a digital XY coordinate table.
-Output video as Y coordinate information.

The digitizer 60 is connected to the computer 46 so that a pure moisture resonance signal converted to a digital signal that can be recognized by the computer is input to perform signal analysis and signal processing on the computer. So that you can do it.

A receiving and detecting section 56 is connected to a signal output section of the computer 46 so that the signal analyzed and processed by the computer 46 can be output not only through the oscilloscope 58 but also through an output device such as a monitor and a print separately connected to the computer 46. Is also output.

The present invention has a structure in which the entire system can be controlled by the computer 46. The high frequency pulse signal generated by the computer program has a pulse width of about 10 μs and a delay time of 100 ms, and the entire system must have a transmission / reception cost (Duty cycle) that satisfies the pulse width and increase the signal-to-noise ratio value. Therefore, it must be possible to generate a number of repetitive resonance signals.

When a 48 pulse signal is generated by the pulse programmer and sent to the high frequency switching section 50, the high frequency circuit receives the gate signal from the pulse programmer 48 and the 4.1M signal.
The Larmor Precession frequency in Hz is mixed with the gate signal to generate a signal.

That is, a sine wave of 4.1 MHz is generated only while the gate signal is present. Pulse programmer 48
Simply produces a pulse having a constant width, and this signal is a gate signal and a 4.1 MHz high-frequency signal because two signals are generated from the high-frequency switching unit 50.

The output of the signal output from the high-frequency switching unit 50 is within several watts, and this signal is output in a high-frequency field (RF fi
Since the eld) cannot be formed, the output of the signal must be amplified. The part in charge of this function is the high-frequency power amplifier 52. The amplified signal averages 800 to 10
It becomes a signal of 00 watts. The amplified signal is a static magnetic field (sta
tic magnetic field) to form a high-frequency field that forms a magnetic field that is perpendicular to the static magnetic field,
Enter the RF / sense coil 34.

A conventional NMR test apparatus converts this signal into an RF signal.
Since there was a separate part for sending to the sensing coil 34 and a part for sensing the resonance signal, two devices for sending and receiving signals to the sample were required.
Since the F / sensing coil 34 can transmit a high-frequency field signal and receive a resonance signal, the apparatus is simplified.

Therefore, there must be a device capable of transmitting the high-frequency pulse signal generated from the high-frequency power amplifier 52 by the RF / sensing coil 34 and separating and receiving the resonance signal. is there. The transmission / reception converter 42 has a built-in damping (Damping).
The circuit must be used to separate and obtain only the signal due to the next successive resonance following the high frequency signal.

The signal generated in the sample after the transmission of the resonance signal is sensed by the RF / sensing coil 34, and this signal is transmitted to the duplexer 42. The signal output from the duplexer 42 is input to a preamplifier 54. RF / sensing coil 3
Are weak signals within several μv, and are amplified to several tens mV through the preamplifier 54.

The signal amplified from the preamplifier 54 is RF
/ Since the signal is a mixture of the 4.1 MHz high frequency pulse signal applied to the sensing coil 34 and the sample resonance signal,
After separating and removing the 4.1 MHz signal from the reception and detection unit 56, only a pure resonance signal is obtained.

Since the signal obtained above is an analog signal, it must be converted to a digitizer signal that can be recognized by the computer 46 using the digitizer 60.

The digitizer 60 has a sampling rate (sam) of twice or more the frequency finally generated for signal analysis.
pling Rate) is satisfactory, and usually a few M
A digitizer with a sampling rate within Hz is used.

The signal generation method used in the present invention is shown in FIG.
FID (Free Inductio
n Decay) signal. That is, as described above, after the width of the gate signal generated from the pulse programmer 48 is fixed, the signal is sensed after the end of the width. This is the time required to make a constant angle (usually 60 ° -90 °) with the field.

In the FID, a signal is normally generated so as to form 90 °, and a signal is obtained for the sample. However, in an actual experiment, a signal generated after having a slope of 40 ° to 90 ° was obtained. Therefore, signal analysis was easiest when the angle was within 60 °. In the present invention, the pulse width is set to 6.5 μs so as to have a slope of 60 °, and the delay time between pulses is 1
00 ms.

Experimental examples and results of measuring the water content of agricultural products and the sugar content of fruits using the present invention thus constituted are as follows.

Experimental Example 1: Measurement of moisture content of rice Sample preparation Ten samples were prepared from flowering varieties of rice with different moisture content. The manufacturing method is as follows. After adding water to a container such as a bag or bottle that is shut off from air, store it in a closed space at 40 ° C for 7 days so that the rice gradually absorbs water and the overall moisture content of the rice is uniform. So that Samples taken from the container are removed using a device such as a dessicate to remove moisture absorbed by the grains.

In this example, a grain sample was spread on newspaper for about one hour to remove water absorbed by the grain.
After the sample is taken out of the desiccator, the sample is placed in a 30 mm test tube, and the sample is naturally put without applying force so that the amount of the sample becomes constant.

To prepare the samples and simultaneously measure the water content of each sample, the sample was dried at a temperature of 105 ° C. for 5 hours. Are as shown in Table 10.

[0128]

[Table 10] .

Since the values of the measured results are different from each other, it is considered that the errors are caused by the influence of water that is not completely removed from the rice, together with the errors of these simple measuring instruments.

On the other hand, the samples produced in the above process were used to obtain a water resonance signal using the NMR apparatus of the present invention, and data was obtained every two times to verify the operability of the apparatus. After the entire system was first operated (ON), the tuning circuit was adjusted to adjust the tuning circuit to 4.1 MHz by tuning the vacuum variable capacitor (AVC), and then a moisture resonance signal was obtained.

Signal Acquisition The signal obtained through moisture resonance is shown in FIG.
As shown in the drawing, data obtained from a sample having a moisture content of 0.4% cannot be used as it is, and the signal must be processed to obtain an envelope. FIG. 14 is an envelope waveform chart obtained from the data shown in FIG. 13, and FIGS. 14, 15, and 16 are envelope waveform charts obtained from six rice samples having different moisture contents.

FIG. 17 is a drawing showing the correlation between the water content and the maximum value. The coefficient of determination was 0.904. FIG.
Represents the relationship between the arithmetic sum of the four points after the maximum value and the water content, and the coefficient of determination is 0.947.

Results Analysis FIGS. 17, 18, and 19 show FID (Free Induction Deca).
y): magnetic induction attenuation signal) The correlation between the arithmetic sum of the intensity after the maximum value and the water content is obtained, and each coefficient of determination R ²
Are 0.904, 0.942 and 0.947, and it can be seen that such a coefficient of determination shows a fairly good correlation when considering surrounding noise and measurement errors.

Based on the above results, the water content of agricultural products could be continuously measured by a relatively simple method.

Example 2: Production of a sample for measuring the sugar content of fruits Fruits and vegetables are made of starch (starch), fructose, and glucose (Glu), which are carbohydrates of saccharides as the ripeness increases.
The total sugar content increases while changing to sucrose, which is a monosaccharide and disaccharide such as cose). Most fruits and vegetables comprise 80-90% water, monosaccharides and disaccharides usually make up 10-20%, and are composed of some other fibrous material. Sugar is usually called sugar.
rose).

Sucrose is a bond between fructose and glucose, which are monosaccharides, and generally has 22 hydrogen atoms. For preliminary experiments on fruits and vegetables, 14 sugar water samples were prepared and sugar water was prepared. Glycerol samples were used that were used to set the overall system (device) to optimal resonance prior to performing experiments on. The experimental sample was prepared by distributing the same mass of distilled water in a glass tube having a diameter of 30 mm and changing the mass of sugar (only pure sucrose was used in the experiment) so that the sugar content was uniformly distributed.

The sugar content of the sugar water used in the experiment was as follows: pure distilled water, 3.1% sugar water, 5.8% sugar water, 1% sugar water.
7.6% sugar water, 18.9% sugar water, 19.7% sugar water, 22.2% sugar water, 22.5% sugar water, 2
6.8% sugar water and glycerol.

Preliminary experiments with glycerol Experiments on glycerol were performed first. After the whole system was operated (ON), a signal was obtained after adjusting a vacuum variable capacitor (AVC) to tune to a tuning circuit of 4.1 MHz.

Whether the tuning is confirmed or not is determined by the frequency counter.
(frequency counter) and Q meter (quality factor me
ter) or a measuring instrument such as a standing wave detector, and was considered tuned when the largest signal for the current circuit appeared.

Equipment for acquiring experimental data is a digital oscilloscope (58: Philips PM3394A model)
, The reference time (Time Base) is set to 10 μs, 8,192 data are obtained, and the FID of 1.6 ms period is obtained.
A resonance signal was obtained. Since a signal obtained at one time is a signal weak to noise, one signal could not be used, and a large number of signals were averaged on the digital oscilloscope 58.
The averaging method is the same as that of (7), and adopts a method of adding the influence of the newly input signal (New Data) to the previous signal (old data).

[0141]

(Equation 8) .

The signal is processed on the digital oscilloscope 58 with the average factor value set to 128, and the number of data to be averaged is 50 or more, and most of the noise can be removed by averaging. did it.

FIG. 20 shows a signal for glycerol in a period of 1.4 ms, and it can be seen that the signal appears after about 40 μs.

Since the signal before 40 μs is divided into two classes, the input signal (6.5 μs) waits until the RF / sense coil 34 detects a signal derived from the sample after the input of the signal is cut off because of the FID. This is a state signal. Since a lot of information includes this waiting time, shortening this time improves overall system performance.

The acquired signal represents an image of an oscillating signal that decreases exponentially. Also obtained with the following (8) T ₂ values of the degree of the attenuation is obtained in uniformity of a magnetic field that is bestowed.

[0146]

(Equation 9) Here, A is the output signal, A ₀ is the maximum value of the output signal, T ₂ is a value unique to the object, and the value slightly varies according to the uniformity of the magnetic field. By analyzing this value, the sugar content and water content of the fruits and vegetables can be measured, but the water content and sugar content of the fruits and vegetables are mixed. In the case of the semi-solid state, two or more T ₂ values are observed, and the higher the degree of uniformity of the magnetic field B ₀ of the permanent magnets 6, 8, the more accurate the T ₂ value can be measured.

Since the signal obtained above oscillates with time, the glycerol signal has a large initial attenuation as shown in FIG. 21 without using the signal as it is and obtaining the envelope as shown in FIG. This indicates that the attenuation is completed after 800 μs, which means that most of the signals are present before 800 μs.

It can be seen from the experiment on glycerol that a signal before 800 μs must be sensed in order to obtain and analyze a meaningful signal. However, this value should be 800 μs or more to increase the uniformity of the magnetic field between the permanent magnets 6 and 8 and to make the RF / sensing coil 34 more precise to increase the signal input sensitivity and to accurately match the impedance. Can be extended.

Experiments and Analysis on Sugar Water After setting the environment of the whole system with glycerol, experiments on 14 sugar waters were performed.
24 is a graph showing the envelope of six sugar water samples containing glycerol.

As can be seen from FIGS.
After s, the sample with high sugar content and the sample with low sugar content appear to decay to the same value, and the decay of pure water is slow, while the sample with high sugar content has the appearance of decay faster and the T _{2 of} pure water is reduced. It appears much larger than a sample with a high sugar content. This means that hydrogen atoms in samples with high sugar content are less active than pure water.

A number of simple methods have been tried to easily predict the sugar content. As an example, with respect to sugar water predicted sugar content in inclination degree for two points on the envelope without asking the T _2.

FIG. 25 shows the relationship between the inclination with respect to two points and the sugar content, and the sugar content for sugar water can be easily measured using a linear model by a simple method.

Experimental Analysis and Results for Fruits and Vegetables Fruits and vegetables used as experimental subjects are shown in Table 11.

[0154]

[Table 11] .

FIG. 26 is a comparison of the envelopes of water, apple and kiwi. The appearance of the signal is different for each object, and the decay time of the liquid water is shorter than that of the semi-solid apple or kiwi. Can be seen to appear relatively long.

(A) Experimental analysis and results on apples 30 apple samples were made to have the same mass (25.5 g ± 0.2 g) in order to eliminate the deviation caused by the mass, and the sugar content was digitally refracted. Saccharimeter (Model name: TRM
-100) twice, and the sugar content was averaged from the two values. The sugar content of the apple sample is 10.6%
It is distributed within １５15.9%, and the sugar content distribution is as follows. 10% to 11%: 1 piece, 11% to 12%: 1 piece, 1
2% to 13%: 4 pieces, 13% to 14%: 5 pieces, 14% to
15%: 8 pieces, 15% to 16%: 11 pieces, and the sugar content of the whole fruit was 14% to 16%, and many fruits were distributed.

Since apple is in a semi-solid state unlike sugar water, the envelope and data are directly used without using the inclination at two points as easily as in signal processing using a liquid sample such as sugar water.

A multi-linear regression analysis (Multipl) based on a statistical processing program (SAS) created by using 20 data to find a coefficient having a correlation between the envelope and the sugar content.
e linear regression) to develop a predictive model.
When using 18 coefficients, the coefficient of determination R ^{2 of the} prediction model
= 0.931. However, the results obtained by verifying the sugar content of the remaining 10 samples using the developed multiple linear regression model are shown in FIG. 28. The standard deviation value is about 2.4%, and the sugar content range is 10% to 16%. In the case of an apple having a sugar content, it is difficult to classify the apple into a second grade if the sugar content is measured using such a method and the grade is determined. Therefore, the sugar content prediction using such a statistical processing program (SAS) Did not agree.

Therefore, the resonance signals of 30 apples were predicted using the neural network. The input coefficient is calculated by taking the slope of two points from the sugar water as the slope between the initial value and the adjacent value.
The number of hidden nodes (Hidden Nodes) was four, the output value was one sugar content, and learning was performed. The number of input patterns was 20, and the remaining 10 were used for verification.

FIG. 28 is a graph showing 20 samples from 30 sample data.
The results obtained by taking individual samples and verifying them based on the results learned by a neural network are shown, and the possibility of classifying using the neural network based on the verified results is foreseen.

The experimental results for apples are as follows.

The result of predicting and verifying a coefficient having correlation using a statistical processing program (SAS) showed that it was not appropriate to classify the sugar content by a relatively simple linear equation. This is considered to be because RF / signal existing in the surrounding space causes the coil to form an induced current to generate noise.
Therefore, it is necessary to completely shield the RF / sense coil (34) and the tuning circuit from external noise.

After learning using a gradient value and an actual sugar content value using a neural network, prediction is performed, and a prediction standard error is 0.5.
65%. The measurement error was 1.13% with about 95% reliability.

If the above results are used, it is determined that sorting can be performed at about grade 3 when the sugar content of apples is generally in the range of 10% to 17%.

B. Experiment and analysis on pears 20 pear samples were made, and the mass of each was made to be 40.4g ± 0.2 to reduce errors due to the mass, and the sugar content was measured with a digital refractometer (model name: TRM-100). After measuring twice, taking the average value,
The overall sugar content is 9.8% to 13.1% lower than apple
(Brix%), and the sugar content distribution is as follows.

9.8% to 10.0%: 3 pieces, 10% to 1
1%: 6 pieces, 11% to 12%: 6 pieces, 12% to 13.1
%: Five pieces showed a more uniform sugar content distribution than apples.

Learning was performed on the sample using a neural network, and verification was performed using the result of learning. When the number of samples is 20, learning is performed on as few as 12 samples, and the result of verification with 8 samples is as shown in FIG. The coefficients used in the neural network are calculated using the statistical processing program (S
AS) were used as inputs, the number of hidden nodes was 7, and the output was 1 sugar content value.

The results of the experiment on pears are as follows.

The standard error of the result verified by the result of learning using the neural network is 0.392%. That is, the measurement error is within 0.78% with 95% reliability.

Compared with the apple sample, the mass was larger and better results could be obtained. This was determined to be because the effect on the surrounding noise (Noise) was relatively reduced by increasing the intensity of the resonance signal as the mass increased. Based on the obtained result, it was determined that the third grade according to the sugar content could be determined.

(C) Experiment and Analysis on Kiwi [0169] Kiwi samples were made into thirteen pieces, and the whole was used as a sample without cutting a part like apples and pears into a sample. Since the kiwi has a rectangular shape longer than the end direction and a diameter in the end direction of approximately 500 mm or less, the whole can be put into the RF / sensing coil 34 for the experiment. The diameter of the fabricated RF / sensing coil is 6
0 mm. The sugar content is measured by a digital refractometer (model name:
7. measured once using TRM-100);
It is distributed over 7% to 12.2%, and the sugar content distribution is as follows.

8.7% to 9%: 1 piece, 9% to 10%: 3
10% to 11%: 5 pieces, 11% to 12.2%: 4 pieces, and it was determined that the sugar content was lower than that of apple.

The mass of the kiwi is uniformly distributed from 55 g to 83 g. In order to find the similarity between the mass of the kiwi and the resonance signal, the correlation between the intensity of the entire signal (the maximum value of the resonance signal) and the mass was analyzed. The result of the analysis is as shown in FIG. 30, and the coefficient of determination R ² was 0.902 as shown in the figure.

According to the result, free induction decline (FID)
It can be seen that the mass can be predicted using the signal intensity, which means that the intensity of the resonance signal of the whole hydrogen atom is proportional to the hydrogen atom in the sample.

A method of predicting a kiwi using a learning model based on a neural network was tested. The prediction utilized 8 samples and the validation used 5 samples. FIG.
Numeral 1 is a graph for the verified result, and the result of the experiment is as follows.

The correlation between the maximum value and the mass was determined by the free induction decay (FID) signal. The correlation was 0.902, indicating that simple analysis of the resonance signal allows mass to be determined by signal intensity.

The standard error of the prediction of the sugar content of the kiwi in the neural network is 0.415, and it can be considered that the sugar content of the kiwi can be divided into about three stages as in the case of apples and pears according to the prediction result.

(D) Experiment and Analysis on Kumquat There are 35 samples of kumquat. Kumquat, like kiwi, is a fruit or vegetable smaller than the tuning coil diameter and has a diameter of 20 mm to 30 mm, and can be put into the RF / sensing coil 34 for experimentation. The mass of kumquat is uniformly distributed in the range of 10,390 g to 18,645 g, the sugar content is distributed from 11.2% to 19.7%, and the distribution of the sugar content is as follows.

11.2 to 13%: 2 pieces, 13% to 14
%: 1 piece, 14% to 15%: 11 pieces, 15% to 16%:
10 pieces, 16% to 17%: 7 pieces, 17% to 18%: 1
Pcs, 18% to 19.7%: 3 pcs.

A prediction experiment using a neural network was performed. The number of input nodes is ten, and the number is determined by the statistical processing program (SA
Using S), correlated data was input, the number of nodes in the hidden layer was 5, and the output was 1 sugar content. FIG. 32 shows the results of learning with 25 samples, prediction of sugar content using 10 samples, and comparison with actual measured values. The standard error, which is a measure of the difference between the predicted sugar content and the actual value, was slightly higher at 1.461.

The results of the experiment are as follows.

The sugar content was predicted by using a neural network model using 10 coefficients. The SEP is expressed as 1.461, and the reason why the error was somewhat large is considered that the mass of kumquat was small, the resonance signal was weak, and the influence of noise was large.

[0183] It was determined that sorting of about 2 grades was possible using the given results.

[0184] Such a fruit having a small mass can improve the uniformity of the magnetic field by increasing the strength of the magnetic field by decreasing the number of the empty poles 10 of the magnet box, so that better results can be obtained. to decide.

In the present invention, the corner steel 24 is formed in a triangular shape. However, if the inclined surface (45 °) is formed in a concave and curved shape as shown in FIG. 33, the flow of the magnetic field becomes smooth and the uniformity increases. I do.

In the present invention, a large amount of heat is generated due to a high-output high-frequency pulse signal.
In this case, the diameter of the RF / sensing coil 34 is set to 120 mm or 60 mm without affecting the moisture content and the sugar content measurement. It is preferable to increase the diameter of the sensing coil 34 to increase the strength of the permanent magnets 6, 8.

In the present invention, non-destructive inspection is described by measuring the moisture content of agricultural products and the sugar content of fruits and vegetables as examples. The content of (edible oil or non-edible oil) can also be measured.

[0188]

As described above, the present invention has an effect that the water content and the sugar content of agricultural products can be continuously and simply measured using practical hydrogen nuclear magnetic resonance without destruction.

[Brief description of the drawings]

FIG. 1 is a drawing for explaining nuclear magnetic resonance. (1) The substance is not in a magnetic field. (2) The substance is in a magnetic field.

FIG. 2 is a drawing for explaining nuclear magnetic resonance, showing a macroscopic magnetization M
FIG.

FIG. 3 is a drawing for explaining nuclear magnetic resonance, showing a macroscopic magnetization M
Signal generated by the relaxation of the body.

FIG. 4 is a BH curve diagram of the permanent magnet of the present invention.

FIG. 5 is a longitudinal sectional configuration view of a magnet box (sensor) of the present invention.

FIG. 6 is a longitudinal sectional configuration view of a magnet box (sensor) of the present invention.

FIG. 7 is a BH curve diagram of the magnet box (sensor) case of the present invention.

FIG. 8 is a shimming plate of the permanent magnet of the present invention.
The graph which analyzed the change of the uniformity of the magnetic field and the uniformity according to the width of the plate).

FIG. 9 is a longitudinal sectional view of the RF / sense coil of the present invention.

FIG. 10 is a series resonance circuit diagram of the RF / sense coil of the present invention.

FIG. 11 is a circuit block diagram of the present invention.

FIG. 12 is an FID signal diagram used in the present invention.

FIG. 13 is a waveform diagram of a moisture resonance signal obtained in the rice moisture content test of the present invention.

14 is an envelope obtained by subjecting the moisture resonance signal of FIG. 13 to signal processing.

FIG. 15 is an envelope waveform chart obtained from rice samples having different moisture contents in the rice moisture content test of the present invention.

FIG. 16 is an envelope waveform chart obtained from rice samples having different moisture contents in the rice moisture content test of the present invention.

FIG. 17 is a graph showing the correlation between the water content and the maximum value in the rice water content test of the present invention.

FIG. 18 is a graph showing the relationship between the arithmetic content of four points after the maximum value and the water content in the rice water content test of the present invention.

FIG. 19 is a graph showing the relationship between the arithmetic content of four points after the maximum value and the water content in the rice water content test of the present invention.

FIG. 20 is a signal waveform diagram relating to glycerol in the sugar content test of the fruit of the present invention.

FIG. 21 is an envelope obtained by subjecting the glycerol resonance signal shown in FIG. 20 to signal processing.

FIG. 22 shows an envelope for a sugar sample containing glycerol in the fruit sugar content test of the present invention.

FIG. 23 shows an envelope for a saccharide sample containing glycerol in the fruit sugar content test of the present invention.

FIG. 24 shows an envelope for a sugar sample containing glycerol in the fruit sugar content test of the present invention.

FIG. 25 is a graph showing the relationship between the inclination and the sugar content for two points in the sugar content test of the present invention.

FIG. 26 is a diagram showing a comparison of emplops with water, apple, and kiwi in the sugar content test of the present invention.

FIG. 27 is a graph showing the results of verifying the sugar content of the remaining 10 samples using the multiple linear regression model in the sugar content test of the present invention.

FIG. 28 is a graph showing the sugar content of an apple according to the neural network in the sugar content test of the present invention.

FIG. 29 is a graph showing a sugar content of a pear based on a neural network in the sugar content test of the present invention.

FIG. 30 shows a correlation between the FID signal intensity and the mass in the sugar content test of the present invention.

FIG. 31 is a graph showing the sugar content of kiwi based on the neural network in the sugar content test of the present invention.

FIG. 32 is a graph showing the sugar content of kumquat based on the neural network in the sugar content test of the present invention.

FIG. 33 is a sectional view of another embodiment of a corner steel according to the present invention.

[Explanation of symbols]

2 Magnet box (sensor) 4 Case 6,8 Permanent magnet 10 Air gap 12,14 Through hole 10,18 Pole face 20,22 Shimming plate 24 Corner steel (CORNER STEEL) 26,28 Around Side plates 30, 32 Left / right plate 34 RF / sensing coil 36 Tube 38 Tunnel 40 Series resonant circuit 42 Transmitter / receiver 44 Insulation material 46 Computer 48 Pulse programmer (palse plogrammer) 50 High frequency switching unit 52 High frequency power amplifier 54 Preamplifier 56 Reception and detection unit 58 Oscilloscope 60 Digitizer (A / D converter) 62 Image display B ₀ Magnetic field strength of static magnet (static magnetic field) B ₁ Magnetic field of RF / sensing coil (dynamic magnetic field) )

────────────────────────────────────────────────── ───

[Procedure amendment]

[Submission date] February 3, 1997

[Procedure amendment 1]

[Document name to be amended] Statement

[Correction target item name]

[Correction method] Change

[Correction contents]

The signal obtained through moisture resonance is shown in FIG. 13, which is data obtained from a sample having a water content of 20.4%, and the signal obtained as shown in the drawing cannot be used as it is, and the signal It must be processed to find the envelope. FIG. 14 is an envelope waveform chart obtained from the data of FIG. 13, and FIGS. 15 and 16 are envelope waveform charts obtained from four rice samples having different moisture contents.

[Procedure amendment 2]

[Document name to be amended] Drawing

[Correction target item name] All figures

[Correction method] Change

[Correction contents]

FIG.

FIG. 2

FIG. 3

FIG. 33

FIG. 4

FIG. 5

FIG. 6

FIG. 7

FIG. 8

FIG. 9

FIG. 10

FIG. 11

FIG.

FIG. 13

FIG. 14

FIG.

FIG. 16

FIG.

FIG.

FIG.

FIG.

FIG. 21

FIG.

FIG. 23

FIG. 24

FIG. 25

FIG. 26

FIG. 27

FIG. 28

FIG. 29

FIG.

FIG. 31

FIG. 32

Claims (4)

[Claims] 1. A sample is put into a hydrogen nuclear magnetic resonance sensor having a hydrogen nuclear magnetic resonance apparatus, and then subjected to nuclear magnetic resonance to receive a resonance signal. The signal is processed to obtain a pure water resonance signal or a sugar content signal. A non-destructive inspection method for agricultural products using hydrogen nuclear magnetic resonance, wherein the moisture content and the sugar content of the agricultural products are measured by using the method. 2. A flat permanent magnet (6) (8) on the upper and lower inner surfaces of a box-shaped case (4) made of a material having high magnetic permeability.
Are fixed so that the NS magnetic poles are opposed to each other, an air pole (10) is formed between the permanent magnets (6) and (8), and a pole face having high magnetic permeability is formed on the opposing surfaces of the permanent magnets (6) and (8). (16) Attach and fix (18) and fix the pole face (1).
6) Attach and fix shimming plates (20) and (22) having high magnetic permeability to the opposite edges of (18), and attach and fix corner steel (24) having high magnetic permeability to the inner space of the case (4). Improve the strength and uniformity of the static magnetic field (30),
A magnet box for nondestructive testing of agricultural products using hydrogen nuclear magnetic resonance (NMR), in which an RF / sensing coil (34) is installed in the air pole (10). 3. A transmitter-receiver (42) is connected to one end of a nuclear magnetic resonance magnet box (2) for obtaining a resonance signal after applying a high-frequency pulse to a sample so that the sample can be switched to a transmission / reception state. A preamplifier (54) for amplifying a resonance signal detected by the magnet box (2) from the sample is connected to one end of the vessel, and a reception and detection unit (56) is connected to the output of the preamplifier (54). Amplifying and detecting the amplified signal from the preamplifier (54) to obtain a pure water resonance signal of sufficient magnitude;
An oscilloscope (5) is connected to the output of the reception and detection unit (56).
8) is connected, and the pure water resonance signal is displayed on the image.
High-frequency power amplifier (52) at one end of transceiving converter (42)
, A high frequency switching unit (50), a pulse programmer (48), and a computer (46) for controlling the system are connected one after another to receive and detect the computer (36), the pulse programmer (48), and the high frequency switching unit (50). Section (56), respectively, during a transmission period during which a high frequency pulse is applied at the RF / sensing coil (34),
An apparatus for nondestructive testing of agricultural products that uses hydrogen nuclear magnetic resonance so that reception is interrupted. 4. The magnet box for nondestructive testing of agricultural products using hydrogen nuclear magnetic resonance according to claim 2, wherein the inclined surface of the corner steel (24) is concave.