JP2002154888A - Classifying method for granular fertilizer and classifying equipment for granular fertilizer - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a classifying method for granular fertilizers capable of extremely rapidly, accurately and easily evaluating and classifying the quality and function of the granular fertilizer and an apparatus for manufacturing the granular fertilizers. SOLUTION: The granular fertilizers are irradiated with electromagnetic waves of different wavelengths, the reflection or absorption spectral values are measured and the granular fertilizers are classified by the spectral values thereof. The apparatus for classifying has an electromagnetic wave irradiation means for irradiating the granular fertilizers with the electromagnetic waves of the different wavelengths, a measuring means for measuring the reflection or absorption spectral values of the electromagnetic waves of the granular fertilizers and a classifying means for classifying the granular fertilizers by the measured spectral values.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a method and apparatus for separating granular fertilizer, and more particularly to a granular fertilizer for separating granular fertilizer based on a spectrum value obtained by irradiating the granular fertilizer with electromagnetic waves. And a separation apparatus.

[0002]

2. Description of the Related Art Granular fertilizers, such as urea and chemical fertilizers, and coated fertilizers coated with a film containing a polymer resin as a main component and controlled in dissolution, etc., are manufactured before shipment. Various tests have been carried out to confirm this. For example, a part of the granular fertilizer produced for each production lot or each change of the prescription is extracted, quality and function inspections are performed, and those that pass the inspection are separated and supplied for sale.

[0003] There are various analysis / measurement items for such quality and function tests of granular fertilizers. For example, components and elemental analysis represented by neutralization titration method, colorimetric method, atomic absorption method, etc., as a guaranteed component quantification of fertilizer, moisture measurement by heating weight loss method as moisture quantification, and temperature and humidity as moisture absorption evaluation were specified. Direct measurement of moisture absorption rate under conditions, measurement of solidification amount over time under load as consolidation evaluation, measurement of particle size distribution by sieving as particle size evaluation, measurement of floating amount over time in water as floating evaluation, As dust evaluation, dust density measurement at the time of dust generation and the like are performed.

[0004] In the case of coated fertilizers, it is necessary to evaluate elution characteristics such as elution induction time and specific ratio elution time of fertilizer components, which are important functions. This evaluation method includes:
A method of directly measuring elution in water or soil over a long period of time has been adopted.

[0005]

In recent years, in quality and function management of fertilizers, demands for evaluation methods have been increasing in terms of speed, accuracy, and simplicity. However, as described above, there are many direct evaluation methods for evaluating the quality and function of fertilizers, so that there is a problem that it is destructive, requires a long time, and is extremely troublesome. Naturally, even after the production of granular fertilizer, it is difficult to distinguish good or bad products until this evaluation result is obtained, so it took enormous time and trouble before shipping. In particular, the evaluation of elution characteristics and the like requires a time equivalent to the actual elution time, which is extremely inefficient.

Japanese Patent Application Laid-Open No. 11-72434 discloses that a polymer resin is irradiated with visible light or near-infrared light, its reflection or absorption spectrum is Fourier-transformed, and a power spectrum is obtained. It describes a method of approximately regressing on a straight line and identifying the type and state of the resin from the slope and intercept of the straight line. However, the publication discloses the identification of the substance itself, its state,
For example, it only describes that the density of the polymer resin and whether or not it is a mixture can be identified, and there is no description suggesting that the quality and function of the test object can be separated. In addition, there is no description suggesting a granular fertilizer as a subject.

[0007] It is widely known that the elution characteristics of coated fertilizers are not enough to accurately grasp the elution characteristics of coated fertilizers, even if the type of film composition and the state of the polymer resin, etc. can be identified. It has been said that the only way to evaluate is to carry out a long-term complicated elution evaluation directly in water or soil.

In order to shorten and simplify the time and labor required for the evaluation of the dissolution characteristics, improvements have been made in evaluation methods such as raising the evaluation conditions to higher temperatures than normal fertilization conditions. However, even if the elution temperature is increased, for example, even if the ambient temperature in the evaluation system is 15 ° C. higher than the actual fertilization condition, acceleration is at most about 4 to 5 times, and accuracy and reproducibility are higher at higher temperatures. descend. In the case of a granular coated fertilizer that has the highest social demand and has a dissolution period of about 100 days, at least about 20 days must be spent for evaluation, and it takes a very long time and labor.

In such a long-term evaluation, in addition to the problem of the evaluation method itself, it becomes necessary to store and manage the granular fertilizer to be evaluated as a stock for a long time after production.
In addition, there is a problem that if the product is separated from the defective product, there is a possibility that a shortage of good products may occur.

[0010] In view of such circumstances, the present invention provides a method for separating granular fertilizer which can evaluate and separate the quality and function of granular fertilizer in a very short time and with high precision and accuracy as compared with the conventional method, and a method for separating granular fertilizer. It is intended to provide a device.

[0011]

According to the present invention, there is provided a method for separating granular fertilizer, comprising irradiating the granular fertilizer with electromagnetic waves having different wavelengths, measuring a reflection or absorption spectrum value, and separating the fertilizer based on the spectrum value. It is a feature.

The granular fertilizer sorting apparatus of the present invention includes an electromagnetic wave irradiating means for irradiating the granular fertilizer with electromagnetic waves having different wavelengths, and a measuring means for measuring the reflection or absorption spectrum value of the electromagnetic wave of the granular fertilizer. Separation means for separating the granular fertilizer based on the spectrum value.

The present inventor has surprisingly found that the quality or function of the granular fertilizer, particularly the elution induction time and the elution time in the coated granular fertilizer, can be determined by the value of the reflection or absorption spectrum obtained by irradiating the granular fertilizer with electromagnetic waves having different wavelengths. It has been found that it is possible to grasp the specific ratio elution time and / or to separate the granular fertilizer easily and accurately. The present invention is based on such findings.

In the present invention, preferably, the obtained spectrum value is Fourier-transformed into a power spectrum, which is approximately regressed to a log-log line, and the slope and / or intercept value of the line is obtained. The fertilizer is separated using the values of the slope and / or intercept of the straight line thus obtained.

Further, using the values of the slope and the intercept, the elution induction time and / or the specific ratio elution time of the granular fertilizer obtained by the perceptron type neural network and / or the multidimensional ^Ck class interpolation method are obtained. It is preferable to separate the fertilizer based on the elution induction time and / or the specific ratio elution time.

[0016]

BEST MODE FOR CARRYING OUT THE INVENTION Hereinafter, embodiments of the present invention will be described in detail.

In the present invention, the granular fertilizer is irradiated with electromagnetic waves having different wavelengths to measure the reflection or absorption spectrum value, and the fertilizer is separated based on the spectrum value. Specifically, for example, comparing the optical spectrum value obtained from the granular fertilizer whose quality and function are known in advance with the optical spectrum value of the granular fertilizer manufactured from each production lot having different production conditions, Based on the degree of divergence, it is determined whether or not the manufactured fertilizer has the desired quality and performance, and is separated.

Although the wavelength range of the electromagnetic wave applied to the granular fertilizer is arbitrary, a long wavelength portion of the electromagnetic wave wavelength, specifically 4
Electromagnetic waves in the visible to infrared region of from 00 to 25000 nm are preferable because they reflect the molecular skeleton constituting the substance and its geometrical characteristics.
Furthermore, 800 to 4000 nm, especially 1100 to 2400
Electromagnetic waves having a wavelength of nm are preferred.

The method of irradiating the electromagnetic wave is arbitrary. For example, the number of scans is set to 1 to 1 at a speed of 5000 points / second for each specific wavelength of 0.5 nm using an acousto-optic tunable wavelength filter or the like.
The measurement may be performed at 00 times and the number of measurements is 1 to 100 times / sample. When measurement is performed using an electromagnetic wave having a wavelength of 1100 to 2400 nm by this method, the number of measurement points is 2600,
The required time is slightly over 0.5 seconds. The required time when the number of scans is set to 20 is slightly over 10 seconds.

Even when the electromagnetic waves radiated on the granular fertilizer penetrate the fertilizer, for example, the case for holding the fertilizer to be measured is made of ceramic, and the electromagnetic waves can be reflected on the case to obtain a spectrum and the measurement can be performed without noise.

The granular fertilizer can be any granular fertilizer, for example, a single-component granular fertilizer such as urea fertilizer,
Chemical fertilizers containing nitrogen, phosphoric acid, potassium, etc., furthermore, those whose surfaces have been subjected to anti-caking treatment or dust generation prevention treatment, and furthermore, coated fertilizers coated with a film mainly composed of a polymer resin or the like. Can be

In the present invention, a plurality of types of granular fertilizers such as a coated fertilizer having the same film composition but different core granular fertilizer components may be used as the granular fertilizer to be separated.
However, in the granular fertilizer to be separated, for example, the core granular fertilizer components in the coated fertilizer etc. are unified,
It is preferable that the number of different factors is small, because the separation can be performed with high accuracy.

For example, when urea is used as a granular fertilizer component as a core of a coated fertilizer, the same surface treatment method on the urea surface has less noise in the spectrum reflecting the elution characteristics and can be separated with high accuracy. It is preferred.

In the present invention, the granular fertilizer is easily separated by quality and function based on the optical spectrum information obtained from the granular fertilizer. The quality and function include, for example, the guaranteed component amount of the fertilizer, the water content, and the hygroscopicity. , Caking, particle size distribution,
Examples include levitation, dust generation, and elution characteristics.

Particularly, in the case of a coated fertilizer, the elution property is the most important function, and the present invention is particularly effective because it can be easily separated by the elution property in a short time in a short time. The core fertilizer component is optional, and the above-mentioned chemical fertilizer, urea, and the like can be used. The coating material is not particularly limited, but preferably contains a polymer component such as a polyolefin-based resin, a urethane-based resin, or an alkyd-based resin. Fertilizer can be easily separated. The thickness of the film is preferably 30 to 200 μm, particularly preferably about 50 to 100 μm.

The shape, size and the like of the granular fertilizer are arbitrary, but among them, spherical one is preferable. The diameter of the sphere is generally 1 to 10 mm, especially 3 to 7 mm, particularly 2
It is preferably about 4 mm.

In the present invention, whether to use the reflection or absorption spectrum is optional, and may be selected according to the state of the granular fertilizer to be separated or the situation requiring separation.
Generally, it is preferable to use a reflection spectrum, and these may be used in combination.

In the present invention, based on the optical spectrum information obtained from the granular fertilizer whose quality and function are known in advance, the optical spectrum information of the granular fertilizer produced from each production lot having different production conditions is used. Compare and judge the quality and function and separate them. In the present invention, this optical spectrum information is analyzed by 1 / f fluctuation, and the optical spectrum information is grasped as two variables (slope and intercept), so that the processing becomes easier and the accuracy of classification is improved. It is preferred.

Although the method of 1 / f fluctuation analysis is arbitrary,
The optical spectrum information is Fourier-transformed to calculate a power spectrum, and the horizontal axis (or vertical axis) is plotted as a frequency component, and the vertical axis (or horizontal axis) is plotted as a logarithm of the component intensity. Is preferable, and two variables, ie, the slope of the straight line and the intercept, are obtained. In the analysis of the function, the power spectrum is the amplitude of the Fourier transform from -T to T of the function, which is 2 which is the magnitude of the amplitude.
This is the limit when T approaches infinity, obtained by dividing the group average of the power by 2πT. This 1 / f fluctuation analysis can be easily performed by using a computer or the like.

Further, a quantitative correlation between the optical spectrum information of the granular fertilizer obtained as described above and the quality / function information of the fertilizer actually measured by a conventional method is obtained and statistically processed. Then, an extremely high correlation was found. Therefore, in the present invention, the granular fertilizer is classified by separating the data obtained by statistically processing the optical spectrum information of the granular fertilizer and the quantitative correlation between the actually measured quality and function information of the fertilizer. This is preferable because the separation accuracy is further improved.

Specifically, for example, the distribution of the quality and function of the granular fertilizer is grasped as a smooth variable distribution by complementing the irregular distribution bivariate function data, and the quantitative correlation with the optical spectrum information of the granular fertilizer is determined. Ask. As this method, for example, it is preferable to use a multidimensional function fitting technique. The multidimensional function fitting technique is to express a correlation by a linear or non-linear combination of arbitrary basis functions. In the present invention, any of the technical methods can be used. Among them, a fitting technique based on a sigmoid function or a k-th power polynomial, specifically, a perceptron type neural network and / or a multidimensional C ^k class interpolation method is actually measured. The correlation coefficient between the quality / function of the granular fertilizer and the optical spectrum information is increased, which is preferable.

For example, a perceptron-type neural network is a NeuroSymm / Light plus
V. If the computer software of 3.2 (manufactured by Fujitsu Limited) is used, a quantitative correlation can be easily obtained on a computer, and the computer can be used for the multi-dimensional C ^k class interpolation method as well. The method of Ninomiya et al. (Transactions of the Information Processing Society of Japan, Vol. 22, No. 6, 581-
588, 1981).

With these multidimensional function fitting techniques, for example, 1 / f fluctuation analysis can be performed from quality / function values (for example, actual measurement values such as 80% elution days) measured by a conventionally known method and optical spectrum information of the fertilizer. A quantitative correlation between the slope and the value of the intercept calculated in step (1) is obtained, and the granular fertilizer can be easily and accurately separated based on the degree of difference between these two relationships. In particular, when the separation is performed using the elution characteristic value, which is a function of the coated fertilizer, the coated fertilizer is separated, and in terms of the production conditions of the granular fertilizer and the time reduction and simplified management of the function management of the granular fertilizer. The present invention has a great effect.

Further, using this quantitative correlation, a regression value of the quality and function of the fertilizer corresponding to the optical spectrum information of the newly manufactured granular fertilizer can be obtained.
In addition, to improve the reliability of sorting, the correlation coefficient is corrected as needed by adding the optical spectrum information of the newly manufactured granular fertilizer and the new quality and function information of the corresponding fertilizer as needed. It is preferable because it can be increased.

As generally performed in the statistical processing, the present invention may employ a method of performing statistical processing after removing a small number of abnormal values. Specifically, it is a general method to obtain a correlation coefficient except for a small number of measurement objects having extremely different values due to measurement errors, foreign substances, and the like.

The time required for the separation of the granular fertilizer according to the present invention varies depending on the number of objects to be measured and the processing conditions of the optical spectrum information, and can be arbitrarily changed by setting the desired accuracy and speed of the separation. For example, in the granular fertilizer having the target quality / function, the quantitative correlation between the optical spectrum information and the quality / function is clear, and after measuring the optical spectrum of the newly manufactured granular fertilizer,
The time required for managing the quality / function by obtaining the regression value of the quality / function by the quantitative correlation for this is mainly the measurement time of the optical spectrum information and the 1 / f fluctuation analysis of the information and the regression value of the regression value. This corresponds to the sum of the computer processing time for performing numerical processing such as calculation, and is usually performed within several minutes.

The method for separating granular fertilizer of the present invention can be carried out, for example, as a process control in which quality and function are separated in parallel with the production, and also in a shipping management and the like in which quality and function are separated after production. It can also be used for quality / function confirmation (evaluation).

There is no particular limitation on the fertilizer separation apparatus according to the present invention, but any structure or system that can obtain optical spectrum information of a specimen during, for example, the intermediate or collection stage of fertilizer production may be used.

[0039]

EXAMPLES The present invention will be described in more detail with reference to the following Examples, but it should not be construed that the present invention is limited to the following Examples unless it exceeds the gist of the invention.

(1) Preparation of Coated Fertilizer Coated fertilizers were prepared using film materials having various compositions shown in Table 1. For each of these coating materials, 2 kg of a solution (film component concentration of 5% by weight, 90 ° C.) prepared by dissolving or dispersing the same in tetrachloroethylene was prepared, and this was mixed with urea granules (ASEANBINTULU FERTILEZE, Malaysia).
(R company) was coated using a jet-type coating apparatus on 1 kg. Dry flowing air temperature 90 ° C, air flow 100N
Spray coating was performed at m ³ / hour to obtain coated fertilizers 1 to 12.

(2) Evaluation method of dissolution characteristics of coated fertilizer 7.0 g of coated fertilizers 1 to 12 obtained in (1) were individually poured into 200 cc of water, and allowed to stand in a thermostat at 25 ° C.
The amount of urea eluted was measured every 1 to 3 weeks to obtain an elution characteristic value. Table 1 shows the measurement results of the elution induction days (substantially the number of days until the start of elution) and the 80% elution days.

The symbols in Table 1 indicate the following contents. SAA: surfactant, polyoxyethylene nonylphenyl ether (manufactured by Toho Chemical Co., HLB = 14.1) A: high density polyethylene (density = 0.958, MFR = 2)
B; ethylene vinyl acetate copolymer (MFR = 2.0, V
Ac = 15% by weight) C; low density polyethylene (density = 0.923, MFR =
4.0, melting point = 111 ° C.) D; linear low density polyethylene (density = 0.925, M
FR = 20, melting point = 125 ° C.) E: low molecular weight polyethylene (average molecular weight 700) F: low molecular weight polyethylene (average molecular weight 500)

[0043]

[Table 1]

(3) Measurement of Optical Spectrum Information Optical spectral information was measured as the optical spectrum information for the coated granular fertilizers 1 to 12 obtained in (1) by the following method. Equipment used: Near infrared spectroscopy plastic type discriminator (Opto Giken Co., Ltd.), trade name: plastic scan wavelength range: 1100 nm to 2400 nm, setting conditions: about 0.5 sec / scan × 20 scans, 0.5 nm Step Measurement method: Petri dish with 6cm diameter, fertilizer layer thickness about 1cm
Add fertilizer to the extent.

The reflected light is measured by bringing the irradiation part of the measuring device into contact with the upper surface of the fertilizer layer.

At this time, a ceramic plate is laid on the bottom of the petri dish so that light does not penetrate.

FIG. 1 shows the near-infrared reflection spectra of the coated fertilizers 1 to 7.

The horizontal axis represents the measured wavelength (× 1000 nm), and the vertical axis represents the reflection spectrum intensity. There is a difference in the spectrum shown for each sample, and by comparing this with that of a granular fertilizer of which quality and function are known in advance (the target granular fertilizer), it is possible to separate newly produced coated granular fertilizer. It is.

(4) 1 / f fluctuation analysis The power spectrum obtained by Fourier transforming the reflection spectrum information obtained in (3) is regressed to an approximate log-log line, and the slope and intercept of this line are obtained. The results are shown in Table 2. FIG. 2 shows the correlation between the slope and the intercept.

As shown in FIG. 2, the correlation between the slope obtained by the 1 / f fluctuation analysis and the intercept is as follows. Regression line y = -1.5728x-5.4517 Correlation coefficient R = -0.9956

When the obtained correlation coefficient R = −0.9956 was tested, R (1) was obtained from the R table (Introduction to Biostatistics, Baifukan, first edition, 11th edition, November 30, 1983, page 262, page 262).
(0, 0.001) = 0.823, and the calculated absolute value of the correlation coefficient is higher, indicating a correlation at the significance level of 0.1%. Therefore, it is clear that the fertilizer newly manufactured using this can be accurately separated. In other words, the optical spectrum information value obtained from the newly manufactured granular fertilizer is applied to this correlation table, and it is possible to easily determine whether or not it has the intended function by the degree of deviation from the correlation. .

[0052]

[Table 2]

(5) Quantitative correlation between optical spectrum information and quality / function of coated granular fertilizer (5-1) Perceptron type neural network Actually measured values of elution induction days and 80% elution days shown in Table 1 The quantitative correlation between the slope obtained by the 1 / f fluctuation analysis shown in Table 2 and the predicted value of both days obtained using the intercept was clarified by the perceptron-type neural net method.
3 and 4.

Specifically, the relation between the slope and intercept of the regression line obtained by 1 / f fluctuation analysis with respect to the number of days of elution induction and the number of days of 80% elution by making a perceptron type neural network learn the quality and function values of the granular fertilizer. Was automatically extracted. In practice, prediction was performed by using a perceptron-type neural network that learned a data set from which one of the data shown in Table 1 was removed, that is, leave-one-out was examined.

The network structure of the perceptron type neural network is such that the input layer neuro 5 corresponding to the gradient, the intercept, the number of days of elution induction, the 80% elution time, and the acceleration parameter (always 1.0) as input data, Neuron 8 and output layer neuron 1 were used. The threshold value of the learning error is 0.0008. Table 3 shows the measured values of the number of days of elution induction and the number of days of 80% elution, and the estimated values of both days by a neural network. 3 and 4
Shows the correlation between the estimated value and the actually measured value of the elution induction time and the 80% elution time. Estimated values are rounded to the first decimal place.

[0056]

[Table 3]

The relative relationship between the actually measured value and the estimated value of the number of days of elution induction in FIG. 3 is as follows. Regression line y = 0.9223 + 3.5517 Correlation coefficient R = 0.8283

The relative relationship between the measured value and the estimated value of the 80% elution days in FIG. 4 is as follows. Regression line y = 0.9539x + 6.294 Correlation coefficient R = 0.9283

The regression line shown in FIG. 3 was compared with a known document (Introduction to Biostatistics, Baifukan, Nov. 30, 1983, First Edition, No. 11, 2nd Edition).
Test based on pages 06-210 and Appendix 17).
The variance ratio for investigating the significance of the regression linear gradient was 21.86, and the first degree of freedom, the second degree of freedom, and 10.alpha. = 0.001.
Is 21.04, the variance ratio is larger, and it is therefore clear that the regression line in FIG. 3 is significant at the 0.1% level. Similarly, when the regression line shown in FIG. 4 is also tested, the variance ratio for investigating the significance of the gradient of the regression line is 62.32, and the first degree of freedom, the second degree of freedom, 10 · α = 0.001. Since the value is 21.04, the dispersion ratio is larger. Therefore, FIG.
Is also significant at the 0.1% level. Therefore, it is clear that this can be used to accurately separate newly manufactured fertilizer, and the optical spectrum information value obtained from the newly manufactured granular fertilizer is applied to this correlation table, and the deviation from the correlation is determined. Depending on the condition, it can be easily determined whether or not it has a desired function.

(5-2) C ^k Class Interpolation Method A neural network is a non-linear multi-dimensional fitting and an interpolation based thereon. Therefore in order to compare the linear multidimensional fitting, which extends is currently used widely in computer graphics, C ^k grade interpolation for random distribution bivariate function data described above were Ninomiya et al to n-dimensional Was used in the present invention. Specifically, Sato and C ^k Class interpolation for random distribution bivariate function data of Ninomiya et al., Randomly distributed the N 2 variable data (x _i, y
_{i), f i = f (} x i, y i), (i = 1, N) to the calculation of the partial derivatives up to k floor in creation and each data point of the triangular meshes having vertices each data point After that, an interpolation function of C ^k class is set for each triangular element over the entire domain, and interpolation is performed within the domain. However, for interpolation of a point outside the defined area, an interpolation value is obtained by a least square method using a (k-1) -order polynomial. In this example, this was extended to n dimensions to estimate the elution induction time and 80% elution time. However, in the case of the present embodiment, since the number of samples is 12 due to a three-dimensional problem, k = 3.

Table 4 shows the results obtained by using the C ^k class interpolation method.
Estimated values are rounded to the first decimal place. Excluding the sample 6, the extrapolation is for points outside the learning definition area. Except that the estimation of the data of the sample 3 is extremely bad, the estimation can be made with almost the same error as the prediction by the neural network.

[0062]

[Table 4]

FIGS. 5 and 6 show the correlations except for the data of the sample 3. FIG.

[0064] The relative relationship of the measured values of the eluted induction days by C ^k class interpolation of Figure 5 and estimated value (Sample 3 exclusion) is as follows. Regression line y = 1.3694x-20.977 Correlation coefficient R = 0.8433

The relative relationship between the actually measured value and the estimated value of the 80% elution days by the ^Ck class interpolation method shown in FIG. 6 (excluding sample 3) is as follows. Regression line y = 0.161x + 27.625 Correlation coefficient R = 0.850

When the regression line in FIG. 5 is tested in the same manner as in FIG. 3, the variance ratio for investigating the significance of the regression line gradient is 24.62, and the first and second degrees of freedom are 9 and 9. -Since the value of α = 0.001 is 22.86, the dispersion ratio is larger. Therefore, it is clear that the regression line in FIG. 5 is also significant at the 0.1% level. Similarly, when the regression line in FIG. 6 is also tested, the variance ratio for investigating the significance of the regression line gradient is 1
8.41, the first degree of freedom 1 / the second degree of freedom 9 · α =
Since the value at 0.005 is 13.61, the dispersion ratio is larger. Therefore, it is clear that the regression line in FIG. 6 is significant at the 0.5% level.

From the above, it is clear that the newly manufactured fertilizer can be accurately separated by using the C ^k class interpolation method as in the case of the perceptron-type neural network method.
The optical spectrum information value obtained from the newly manufactured granular fertilizer is applied to this correlation table, and it is possible to easily determine whether or not the fertilizer has a desired function depending on the degree of deviation from the correlation.

[0068]

As described above, according to the present invention, the reflection or absorption spectrum value obtained by irradiating the granular fertilizer with electromagnetic waves having different wavelengths is measured, and the fertilizer can be separated easily and in a short time based on the spectrum value. . In particular, the relationship between the slope and / or intercept value and the elution characteristic value of a line obtained by approximately regressing a power spectrum obtained by Fourier transforming a reflection or absorption spectrum value into a log-log line is expressed by a perceptron-type neural network and / or Or multidimensional C ^k
It can be grasped by the class interpolation method, and it can be easily and quickly estimated from the new spectrum information with high accuracy.

[Brief description of the drawings]

FIG. 1 shows the near-infrared reflection spectrum of a coated fertilizer.

FIG. 2 shows a correlation between a slope and an intercept of a power spectrum obtained by performing a Fourier transform on a reflection spectrum value and regressing the power spectrum onto an approximate log-log line.

FIG. 3 shows measured values of the days of elution induction and the days of 80% elution,
The quantitative correlation between the slope obtained by the 1 / f fluctuation analysis and the predicted value of both days calculated using the intercept is shown.

FIG. 4 shows measured values of the number of days of elution induction and the number of days of 80% elution,
The quantitative correlation between the slope obtained by the 1 / f fluctuation analysis and the predicted value of both days calculated using the intercept is shown.

FIG. 5 is a correlation diagram between measured values and predicted values of the number of days of elution induction and 80% of days of elution by the C ^k class interpolation method.

FIG. 6 is a correlation diagram between actually measured values and predicted values of the number of days of elution induction and 80% of days of elution by the C ^k class interpolation method.

 ──────────────────────────────────────────────────続 き Continued on the front page F term (reference) 2G059 AA05 BB09 DD01 EE01 EE02 EE12 HH01 HH02 HH06 JJ02 MM01 MM02 3F079 AB00 CA09 CB25 CB33 CB34 CB35 4H061 AA02 AA03 DD01 DD18 EE35 FF08 GG62GG

Claims (8)

[Claims] 1. A method for separating granular fertilizer, comprising irradiating the granular fertilizer with electromagnetic waves having different wavelengths, measuring a reflection or absorption spectrum value, and separating the fertilizer based on the spectrum value. 2. The power spectrum according to claim 1, wherein a reflection or absorption spectrum value is Fourier-transformed to calculate a power spectrum.
A method for separating granular fertilizer, comprising regressing this approximately into a log-log line and separating the fertilizer based on the slope and / or intercept value of the line. 3. The elution induction time and / or identification of a granular fertilizer according to claim 2, which is obtained by a perceptron-type neural network and / or a multi-dimensional ^Ck class interpolation method using values of a slope and / or an intercept. A method for separating a granular fertilizer, comprising separating a granular fertilizer using a ratio elution time. 4. The method for separating particulate fertilizer according to claim 1, wherein the electromagnetic wave having a wavelength of 400 to 4000 nm is irradiated. 5. The method for separating granular fertilizer according to claim 1, wherein the granular fertilizer is a coated fertilizer coated with a film containing a polymer resin. 6. An electromagnetic wave irradiating means for irradiating the granular fertilizer with electromagnetic waves having different wavelengths, a measuring means for measuring a reflection or absorption spectrum value of the electromagnetic wave of the granular fertilizer, and separating the granular fertilizer based on the measured spectrum value. An apparatus for separating granular fertilizer, comprising: 7. The method according to claim 6, wherein the classification means calculates a power spectrum by performing a Fourier transform on the obtained spectrum value, approximately regresses the power spectrum on a log-log line, and calculates the slope and / or the slope of the line. Alternatively, a granular fertilizer separation apparatus is characterized in that a value of an intercept is obtained, and the fertilizer is separated based on a slope of the straight line and / or an intercept value. 8. The method according to claim 7, wherein the separation means comprises a time period for inducing elution of the granular fertilizer obtained by a perceptron-type neural network and / or a multidimensional ^Ck- class interpolation method using the values of the slope and the intercept. An apparatus for separating granular fertilizer, comprising means for determining a specific ratio elution time, wherein the means for separating fertilizers separates fertilizer based on the elution induction time and / or the specific ratio elution time.