JP2017087716A - Prediction method for concrete quality or concrete blending condition - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a method capable of predicting concrete quality or a concrete blending condition in a short time at high accuracy.SOLUTION: A prediction method for cement quality or production condition comprises: performing learning of a neural network with sufficiently large number of times of learning such as σ<σusing the neural network having the input and output layers, learning data and monitor data; repeating learning of the neural network until σ≥σwhile reducing the number of learning times; inputting the measured value of a monitoring data to the input layer of the neural network and outputting the speculation value of an estimate data from the output layer if that measured value belongs within a numerical range formed by a specific limited condition when an analytical degree determination value does not satisfy a first set value and when the above determination value satisfy a previously-set second set value.SELECTED DRAWING: Figure 1

Description

  The present invention relates to a method for predicting concrete quality or concrete mixing conditions using a neural network.

It is common knowledge that the quality of cement, which is the main material of concrete, has a great influence on the quality of concrete, such as its fresh properties such as fluidity, compressive strength, and various durability. However, there are many unclear points about the detailed relationship between the mutual qualities, and a technique for predicting the quality of concrete from the quality of cement has not yet been established.
Furthermore, it is well known that many factors affect the quality of concrete, such as the quality of concrete materials other than cement, the blending of concrete, the concrete construction method, the service environment and the service period of the concrete.
Thus, it is difficult to predict the quality of concrete because many factors influence it. For this reason, post-management such as on-site testing and specimen testing is common for quality control of concrete. Moreover, the quality estimation methods proposed so far are also post-mortem methods for estimating from the quality of concrete after production.

For example, Patent Literature 1 discloses a concrete property estimation method for estimating the properties of ordinary concrete, in which the fresh concrete concrete is put into a hopper of a concrete pump and the fresh concrete in the hopper is stirred. The concrete property estimation method is characterized in that the fresh property of the ordinary concrete is estimated on the basis of the load value related to the rotation of the blade and the rotation number of the blade.
Patent Document 2 includes a step of obtaining a relationship between the strength of concrete and an index related to the microscopic structure of the concrete, a step of estimating the strength at the original position of the frame concrete forming the frame, and the relationship and the above Based on the strength of the concrete in-situ, the step of calculating an index related to the microscopic structure of the frame concrete, and the durability of the frame concrete based on the index related to the microscopic structure of the frame concrete. A method for estimating the durability of concrete, characterized in that it includes an estimating step.

Furthermore, as a method for predicting the quality of concrete using factors related to neural network and concrete construction, Patent Document 3 describes the covering of concrete structures, the depth of neutralization, the total chloride content at the position of the reinforcing bars, It is obtained by observing the steel bars exposed by the survey, such as cracks on the concrete surface, presence or absence of floating, annual average temperature, annual average humidity, annual precipitation, etc. of the area where structures are installed among publicly available weather information. A method for predicting the progress of corrosion of reinforcing bars in concrete by a neural network constructed using the degree of corrosion of the reinforcing bars is described.
A neural network is an information processing system that performs optimization through learning. By deepening learning, it is possible to perform more advanced, complex, and adaptive information processing.

JP 2010-249742 A Japanese Patent Laying-Open No. 2015-10918 Japanese Patent Laid-Open No. 10-21211

The methods described in Patent Documents 1 to 3 are methods for predicting the quality of concrete afterwards using the manufactured concrete, and the preventive effect expected in the prediction work, that is, the concrete before the concrete is manufactured. By predicting the quality of the product and clarifying the necessary corrections, it was not possible to prevent the production of improper concrete and to produce the concrete of suitable quality. .
The object of the present invention is to obtain a concrete quality or a concrete blending condition (concrete having a predetermined planned quality) in a short time and with high accuracy before producing concrete even if the number of data is small. It is to provide a method that can predict the selection and usage of concrete materials.

As a result of intensive studies to solve the above problems, the inventors of the present invention are methods for predicting concrete quality or concrete mixing conditions using a neural network, and using learning data and monitor data, σ _L < sigma _{M (the} meaning of this expression will be described later.) become as, after the learning of the neural network in a sufficiently large number of times of learning, in turn, sigma _L ≧ learning neural network while reducing the number of times of learning When it is repeated until σ _M and the analytical determination value is less than the first set value, the learning of the neural network is finished, the actual value of the monitoring data is input to the input layer of the neural network, and the output layer When the estimated value of the evaluation data is output and the analytical determination value is greater than or equal to the first set value and less than the second set value, the neural network Enter the measured value of the specific monitoring data to the input layer of the workpiece, according to the method of outputting the estimated value of the evaluation data from the output layer, it found that can achieve the above object, the present invention has been completed.

（上記式（１）中、学習データの平均２乗誤差（σ_Ｌ）とは、学習データの監視データの実測値を学習後のニューラルネットワークの入力層に入力して得られた評価データの推測値と学習データの評価データの実測値との平均２乗誤差（σ_Ｌ）である。評価データの推測値の平均値とは、学習データの監視データの実測値を学習後のニューラルネットワークの入力層に入力して得られた評価データの推測値の平均値である。） That is, the present invention provides the following [1] to [7].
[1] A method for predicting the quality of concrete or concrete mixing conditions using a neural network having an input layer and an output layer, wherein the input layer is used for inputting measured values of monitoring data in concrete production. Yes, the output layer is for outputting an estimated value of the evaluation data related to the evaluation of the quality of the concrete or the mixing condition of the concrete, and the combination of the monitoring data and the evaluation data is
(I) The monitoring data is one or more types of data selected from data on cement, data on physical properties of cement, data on materials of concrete other than cement, and data on blending of concrete, and the evaluation A combination where the data is data about the quality of concrete, or
(Ii) The monitoring data is one or more kinds of data selected from data on cement, data on physical properties of cement, data on materials of concrete other than cement, data on blending of concrete, and data on quality of concrete. And the evaluation data is a combination of data relating to the mixing conditions of concrete,
(A) a step of initializing the number of learning times;
(B) using a plurality of learning data that is a combination of the actual measurement value of the monitoring data and the actual measurement value of the evaluation data, and performing learning of the neural network by the number of learning times set in the previous step;
(C) The estimated value of the evaluation data obtained by inputting the actual measurement value of the monitoring data of the learning data to the input layer of the neural network obtained by the learning of the latest step (B) and the actual measurement of the evaluation data of the learning data Of the monitoring data in the monitor data, which is a combination of the measured value of the monitoring data and the evaluation value of the evaluation data for confirming the reliability of the mean square error (σ _L ) with the value and the learning result of the neural network Mean square error between the estimated value of the evaluation data obtained by inputting the actual measurement value to the input layer of the neural network obtained by learning in the most recent step (B) and the actual measurement value of the evaluation data in the monitor data (Σ _M ) is calculated, and when the calculated relationship between σ _L and σ _M is σ _L ≧ σ _M , the step (D) is performed, and the calculated relationship between σ _L and σ _M is σ _L < If σ _M , perform step (E) And a process of
(D) A learning number larger than any of the learning number set in the most recent step (A) and the reset number of learning times of the latest neural network is reset as a new learning number, and the step ( Carrying out B) to (C);
(E) resetting the number of learnings obtained by reducing the number of learnings performed in learning of the latest neural network as a new number of learnings;
(F) using the learning data used in the most recent step (B), learning the neural network for the number of times set in the most recent step (E);
(G) An actual value of the monitoring data of the learning data is input to the input layer of the neural network obtained by learning in the latest step (F), and an estimated value of the evaluation data obtained and the actual measurement of the evaluation data of the learning data Obtained by inputting the mean square error (σ _L ) with the measured value and the measured value of the monitoring data in the monitoring data to the input layer of the neural network obtained by the learning of the most recent step (F). When the mean square error (σ _M ) between the estimated value of the evaluation data and the actually measured value of the evaluation data in the monitor data is calculated, and the relationship between the calculated σ _L and σ _M is σ _L ≧ σ _M , Performing step (I), and if the calculated relationship between σ _L and σ _M is σ _L <σ _M ,
(H) If the number of learning times of the neural network in the latest step (F) exceeds a predetermined numerical value, the steps (E) to (G) are performed again, and the learning of the neural network in the latest step (F) is performed. A step of performing step (J) when the number of times is equal to or less than a predetermined value;
(I) The analytical value determination value is calculated using the following equation (1), and when the analytical value determination value is less than a predetermined first set value, the learning of the neural network is terminated, and the learned neural network The actual value of monitoring data in concrete production is input to the input layer of the network, and the estimated value of the evaluation data related to the evaluation of the quality of the concrete or the mixing condition of the concrete is output from the output layer of the neural network. Predicting the quality of the material or the mixing condition of the concrete, and the analysis degree determination value is equal to or higher than a predetermined first set value, the step of performing the step (J),
(J) A determination is made as to the number of times the step (A) has been performed. If the number is less than or equal to a preset number, the learning conditions are initialized and the steps (A) to (I) are performed again. Performing the step (K) when the number of times exceeds a preset number of times, and
(K) Of all the analytical determination values calculated in step (I), when the smallest analytical determination value is less than a predetermined second set value, the smallest analytical determination value can be obtained. If the neural network in step (I) is a learned neural network, then step (L) is performed, and if the smallest analytical determination value is greater than or equal to a predetermined second set value, the quality of the concrete Or the process of determining that the blending condition of concrete cannot be predicted and terminating the prediction,
(L) The actual measurement value and evaluation data of the monitoring data used as learning data in the step (I) in which the smallest analysis degree determination value can be obtained among all the analysis degree determination values calculated in the step (I). An uncorrelated test was performed on the combination of measured values of, and when there were two or more types of monitoring data judged to be significant at the 5% significance level, it was judged to be significant at the 5% significance level. Plot actual values of monitoring data used as learning data in a coordinate space with all types of monitoring data as coordinate axes, and include all of the monitoring data formed by connecting the plotted monitoring data in the coordinate space After setting an area formed by connecting the monitoring data so that the area is maximized as a predictable monitoring data area, the process (M) is performed, % If monitoring data is determined to be significant at a significance level of 0 or 1 kind, the step of terminating the prediction is determined that it is impossible to predict the blending conditions of the concrete quality or concrete,
(M) It is determined whether or not an actual measurement value of monitoring data in concrete production is included in the predictable monitoring data area, and an actual measurement value of monitoring data in concrete production is included in the predictable monitoring data area, The actual value of the monitoring data in concrete production is input to the input layer of the learned neural network obtained in (K), and it is related to the evaluation of the concrete quality or the concrete mixing condition from the output layer of the neural network. If the estimated value of the evaluation data is output to predict the concrete quality or the concrete mixing condition, and the actual value of the monitoring data in concrete production is not included in the predictable monitoring data area, the quality of the concrete or the concrete mixing Judging that the condition cannot be predicted Prediction method of blending conditions of quality or concrete of the concrete, which comprises a step, the to end.

(In the above equation (1), the mean square error (σ _L ) of the learning data is an estimation of the evaluation data obtained by inputting the measured value of the monitoring data of the learning data to the input layer of the neural network after learning. The mean square error (σ _L ) between the value and the actual measurement value of the evaluation data of the learning data The average value of the estimated value of the evaluation data is the input of the neural network after learning the actual measurement value of the monitoring data of the learning data (This is the average estimated value of the evaluation data obtained by entering the layer.)

[2] The predetermined first set value of the analytic degree determination value is 6% or less, and the predetermined second set value of the analytical degree determination value is larger than the first set value and 20 The method for predicting the quality of concrete or the mixing condition of concrete according to claim 1, which is not more than%.
[3] Prediction of concrete quality or concrete mixing condition according to [1] or [2], wherein the neural network is a hierarchical neural network having an intermediate layer between the input layer and the output layer. Method.
[4] The method according to any one of [1] to [3], wherein the concrete production conditions are optimized based on the estimated value of the evaluation data obtained by artificially changing the value of the monitoring data. A method of predicting concrete quality or concrete mix conditions.
[5] The combination of the monitoring data and the evaluation data is the combination of (i), and the evaluation data in the combination is intensity, slump, or slump flow. A method for predicting the quality of concrete or the mixing condition of concrete according to any one of the above.
[6] The monitoring data in the combination of (i) described above are cement brane specific surface area, sieve test residual amount, color tone a value, color tone L value, cement mineral composition, cement chemical composition, mortar flow, admixture amount, And the method for predicting the quality of concrete or the mixing condition of the concrete according to [5], which is at least one selected from unit water amounts.
[7] Before the step (A), (A-1) two or more types of monitoring data aggregates including one or more types of arbitrarily selected monitoring data are prepared, and the two or more types of monitoring data aggregates are prepared. For each, learning is performed for an unlearned neural network that is different from the neural network used in steps (A) to (M) by using a plurality of selection data that is a combination of the actual measurement value of the monitoring data and the actual measurement value of the evaluation data. The mean square error between the estimated value of the evaluation data obtained by inputting the actual value of the monitoring data of the selected data and the actual value of the evaluation data of the selected data is calculated in the input layer of the obtained neural network. Any of the above [1] to [6], including the step of using the monitoring data in the selection data having the smallest mean square error as the monitoring data in the steps (A) to (M) A method for predicting the quality of concrete or the mixing conditions of concrete described in any of them.

Using the method for predicting concrete quality or concrete blending conditions of the present invention, even if the number of data is small, the concrete quality or concrete blending conditions can be predicted with high accuracy in a short time before producing concrete. can do.
Further, it is possible to optimize the quality of concrete or the mixing condition of concrete in real time based on the obtained estimated value, and it is possible to improve the stabilization of the quality of concrete.
Furthermore, high prediction accuracy can be maintained by continuing learning of the neural network.

It is a flowchart which shows an example of the prediction method of this invention. It is a figure which shows the predictable monitoring data area | region set in the process (L).

Hereinafter, the present invention will be described in detail.
The prediction method of the present invention includes an input layer for inputting an actual measurement value of monitoring data in the production of concrete, and an output layer for outputting an estimated value of evaluation data related to the evaluation of the quality of concrete or the mixing condition of concrete. Is used to predict the quality of concrete or the mixing condition of concrete.
The neural network of the present invention may be a hierarchical neural network having an intermediate layer between an input layer and an output layer.

The following (i) or (ii) is mentioned as a combination of the monitoring data and the evaluation data.
(I) The monitoring data is one or more types of data selected from data on cement, data on physical properties of cement, data on materials of concrete other than cement, and data on blending of concrete, and the evaluation A combination where the data is data about the quality of concrete, or
(Ii) The monitoring data is one or more kinds of data selected from data on cement, data on physical properties of cement, data on materials of concrete other than cement, data on blending of concrete, and data on quality of concrete. Yes, and the above-mentioned evaluation data is data related to the mixing conditions of concrete

“Data related to cement” which is one of the monitoring data in the combination of (i) includes chemical composition of cement, mineral composition of cement, mineralogy and crystallographic properties of each mineral, wet f. Examples include CaO, loss on ignition, specific surface area of branes, particle size distribution, sieve test residue, gypsum hemihydrate, color tone, and the like.
In addition, data on raw materials of cement clinker contained in cement, data on firing conditions of cement clinker, data on grinding conditions of cement, and data on cement clinker can also be used as data on cement.

Here, as the chemical composition of the cement, SiO ₂ , Al ₂ O ₃ , Fe ₂ O ₃ , CaO, MgO, SO ₃ , Na ₂ O, K ₂ O, Na ₂ Oeq (total alkali) in the cement raw material, Examples include TiO ₂ , P ₂ O ₅ , MnO, Cl, Cr, Zn, Pb, Cu, Ni, V, As, Zr, Mo, Sr, Ba, and F.
These can be obtained by chemical composition analysis methods such as “JIS R 5202 (chemical analysis method for cement)” and “JIS R 5204 (fluorescence X-ray analysis method for cement)”.

As the mineral composition of the cement, 3CaO · SiO ₂ (C ₃ S), 2CaO · SiO ₂ (C ₂ S), 3CaO · Al ₂ O ₃ (C ₃ A), 4CaO · Al ₂ O ₃ · Fe ₂ O ₃ (C ₄ AF), free lime, periclase, dihydrated gypsum, hemihydrate gypsum, anhydrous gypsum, limestone powder, blast furnace slag, steelmaking slag, fly ash, natural pozzolana, silica fume, quartz clinker minerals such as silica powder, gypsum, The content rate of the small amount mixing component described in "JISR5210 (Portland cement)", a cement mixing material, etc. is mentioned.
These are measurements using observation images of an optical microscope and a scanning electron microscope; analysis using various mineral composition quantification methods described in “JIS K 0131 (general rules for X-ray diffraction analysis)”; “JIS K 0129 (heat Analysis using various thermal analysis methods described in "General Rules for Analysis"; Estimation using chemical composition values by Borg's method, etc .; Estimation using characteristic values of cement other than chemical composition such as cement tone; Cement production It can be obtained by a method such as weighing in the process.

The mineralogy and crystallographic properties of each mineral are the texture (structure), size, color, birefringence and other optical properties, lattice constants, crystallite diameter, lattice strain, and other evaluation values and measurements of each mineral. Value or calculated value.
These can be obtained by optical microscopy, various electron microscopy, powder X-ray diffraction, or the like.

The ignition loss is measured by measuring the mass of volatile components such as dissociated carbon dioxide due to thermal decomposition of moisture and limestone by the method for determining ignition loss described in "JIS R 5202 (Chemical chemical analysis method)". Value.
The specific surface area of the brain, the particle size distribution, and the residual amount of the sieving test are the values measured by the specific surface area test or the net sieving test of “JIS R 5201 (cement physical test method)” or “JIS Z 8815 (general rule of sieving test method)”. Or a particle size distribution measurement value obtained by the laser diffraction / scattering method.
The color tone (color tone L value, color tone a value, color tone b value) is a value measured by the method of “JIS Z 8722 (color measurement method—reflection and transmission object color)” or the like.

“Data on raw materials of cement clinker” means chemical composition of cement clinker raw material, hydraulic modulus, residual amount of sieving test, specific surface area of brane (fineness), loss of ignition, predetermined time from the time of input to kiln Cement clinker raw material (in transit, such as one time point before 5 hours, or multiple time points such as 1 time point before 3 hours, 4 hours ago, 5 hours ago, and 6 hours ago) The chemical composition, hydraulic modulus, supply amount, waste, etc. of the cement clinker blended raw material from which fine particles etc. are extracted by the air flow that flows in the counter flow (hereinafter referred to as cement clinker kiln raw material) Auxiliary amount of cement clinker made of raw materials, blended silo storage amount (remaining amount) of compounded raw material, storage amount of compound silo storage silo (remaining amount), raw material mill and blended raw material blend The current value of the cyclone located between the wing silos (representing the number of revolutions of the cyclone and correlating with the speed of the raw material passing through the cyclone), raw material chemistry by mixing the cement clinker kiln raw material and auxiliary raw material The composition, hydraulic modulus, brain specific surface area, sieve test residual amount, decarboxylation rate, moisture content, and the like can be mentioned.
Here, the chemical composition of the raw material of cement clinker (prepared raw material or raw material in kiln) is SiO ₂ , Al ₂ O ₃ , Fe ₂ O ₃ , CaO, MgO, SO ₃ , Na ₂ O in the raw material of cement clinker. , K ₂ O, Na ₂ Oeq (total alkali), TiO ₂ , P ₂ O ₅ , MnO, Cl, Cr, Zn, Pb, Cu, Ni, V, As, Zr, Mo, Sr, Ba, F, etc. It is a content rate.

“Data on the firing conditions of cement clinker” includes the insertion amount of cement clinker raw material, kiln rotational speed, drop outlet temperature, firing zone temperature, cement clinker temperature, kiln average torque, O ₂ concentration, NO _X concentration, clinker cooler Temperature, gas flow rate of the preheater (correlation with the temperature of the preheater), and the like.

  “Cement grinding conditions data” includes grinding temperature, amount of water spray in finishing mill, separator air volume, type of gypsum, amount of gypsum added, amount of cement clinker input, number of revolutions of finishing mill, discharged from finishing mill And the amount of powder discharged from the finishing mill, the amount of powder not discharged from the finishing mill, and the like.

“Data on cement clinker” includes cement clinker mineral composition, crystallographic properties of each mineral (such as lattice constant and crystallite diameter), ratio of two or more mineral compositions, chemical composition, wet f. CaO (free lime), weight, etc. are mentioned.
Here, the mineral composition of the cement clinker is 3CaO · SiO ₂ (C ₃ S), 2CaO · SiO ₂ (C ₂ S), 3CaO · Al ₂ O ₃ (C ₃ A), 4CaO · Al ₂ O ₃ · Fe ₂ O ₃ (C ₄ AF), f. CaO, f. The content of MgO or the like. As "the ratio of the two or more mineral composition" includes, for example, the ratio of C 3 _{S /} C 2 _S.
The mineral composition of the cement clinker can be obtained by, for example, the XRD-Riet belt method.
The chemical composition of the cement clinker means SiO ₂ , Al ₂ O ₃ , Fe ₂ O ₃ , CaO, MgO, SO ₃ , Na ₂ O, K ₂ O, Na ₂ Oeq (total alkali), TiO _{2 in} the cement clinker. , P ₂ O ₅ , MnO, Cl, Cr, Zn, Pb, Cu, Ni, V, As, Zr, Mo, Sr, Ba, F, and the like.
It is preferable to use two or more kinds of the “data regarding cement” described above from the viewpoint of improving the accuracy of prediction of evaluation data.

As "data regarding physical properties of cement" which is one of the monitoring data in the combination of (i), cement density, fineness, setting time, stability, strength (mortar compressive strength, bending strength), Examples thereof include fluidity (such as mortar flow value) and heat of hydration.
Cement density, fineness, setting time, stability, strength (mortar compressive strength, flexural strength), and fluidity are measured values according to the test method described in “JIS R 5201 (Cement physical test method)”. And yield values by a double cylindrical rotational viscometer, measured values of rheological properties such as plastic viscosity, and measured values of various fluidity tests on cement paste or mortar.
Examples of the heat of hydration of cement include measured values by the test method described in “JIS R 5203 (Method of measuring heat of hydration of cement (heat of dissolution method))”.
It is preferable to use two or more kinds of the “data relating to the physical properties of cement” described above from the viewpoint of improving the accuracy of prediction of evaluation data.

In “data on concrete materials other than cement”, which is one of the monitoring data in the combination (i), the concrete material indicates cement, aggregate, water, and various admixtures.
Details of each concrete material will be described below.
Examples of the cement include “JIS R 5210 (Portland cement)”, “JIS R 5211 (blast furnace cement)”, “JIS R 5212 (silica cement)”, “JIS R 5213 (fly ash cement)”, or “JIS R 5214”. (Eco-cement) ”: Cement that conforms to the standards of foreign standards, white cement, etc.

As aggregates, “JIS A 5308 (Ready Mixed Concrete)” Annex A, “JIS A 5002 (Lightweight Concrete Aggregate for Structure)”, “JIS A 5005 (Crumbles and Sand for Concrete)”, “JIS A 5011”. (Concrete slag aggregate) ”and other such aggregates.
Examples of water include water that conforms to the provisions of Appendix C of “JIS A 5308 (Ready Mixed Concrete)”.
As various admixtures, "JIS A 6201 (Fly Ash for Concrete)", "JIS A 6202 (Expanding Concrete)", "JIS A 6204 (Chemical Admixture for Concrete)", "JIS A 6205 (For Reinforced Concrete) Anti-corrosive agent ”,“ JIS A 6206 (Blast furnace slag fine powder for concrete) ”,“ JIS A 5008 (limestone powder for paving) ”,“ JIS K 5906 (aluminum pigment for paint) ” Etc.
It is preferable to use two or more kinds of the above-mentioned “data on concrete materials other than cement” from the viewpoint of improving the accuracy of prediction of evaluation data.

“Data concerning the blending of concrete” which is one of the monitoring data in the combination of (i) includes cement, fine aggregate, coarse aggregate, water, various admixtures (AE agent, water reducing agent) blended in concrete. , AE water reducing agent, high performance water reducing agent, high performance AE water reducing agent, fluidizing agent, setting retarder, etc.) and blending ratio of various admixtures (for example, the amount of admixture (mass%) relative to 100 mass% of cement ) And maximum composition of coarse aggregate, water cement ratio, air content, fine aggregate rate, unit water content, unit cement content, unit fine aggregate content, unit coarse aggregate content, unit Examples include admixture amount and unit admixture amount.
The maximum size of the coarse aggregate, the water cement ratio, the amount of air, and the fine aggregate rate are those defined in “JIS A 0203 (concrete term)”.
Unit water amount, unit cement amount, unit fine aggregate amount, unit coarse aggregate amount, unit admixture amount, and unit admixture amount conform to the provisions of “Unit Amount” in “JIS A 0203 (concrete term)” Is.
It is preferable to use 2 or more types of the above-mentioned “data regarding the blending of concrete” from the viewpoint of increasing the accuracy of prediction of evaluation data.

  As the monitoring data in the combination (i), only one kind of data selected from data on cement, data on physical properties of cement, data on materials of concrete other than cement, and data on blending of concrete is used. However, it is preferable to use two or more (plural) of these four types of data from the viewpoint of improving the accuracy of prediction of evaluation data.

The “data on the quality of the concrete” which is the evaluation data in the above combination (i) includes strength (compressive strength and bending strength of concrete, mortar compressive strength and mortar bending strength, etc.), slump, slump flow, air volume, Examples include chloride content, crack resistance, dynamic elastic modulus, dynamic shear elastic modulus, dynamic Poisson's ratio, cured body void amount and void diameter distribution, durability, and color tone. Of these, strength (concrete compressive strength and flexural strength, mortar compressive strength, mortar flexural strength, etc.), slump, and slump flow, which are more important as the quality of concrete, are preferable.
The compressive strength and bending strength of the concrete, such as slump, slump flow, air content, and chloride content, are measured values obtained by the test method described in “JIS A 5308 (Ready Mixed Concrete)”, Or it is a measured value obtained based on the said test method in arbitrary mixing | blendings and material ages, and mortar compressive strength and mortar bending strength are the test methods as described in "JISR5201 (physical test method of cement)". It is a measured value obtained according to the above-mentioned test method in the obtained measured value or arbitrary combination and material age.
Crack resistance is a measured value obtained by the test method described in “JIS A 1151 (Test method for drying shrinkage cracking of constrained concrete)” or a measured value obtained in accordance with the above test method at an arbitrary composition and age. It is.
The dynamic elastic modulus, dynamic shear elastic modulus, and dynamic Poisson's ratio can be obtained by the test methods described in “JIS A 1127 (Test method for dynamic elastic modulus, dynamic shear elastic modulus and dynamic Poisson ratio of concrete by resonance vibration)” and the like. It is a measured value.
The amount of voids and the pore size distribution in the cured product are measured values obtained by pore distribution by the mercury intrusion method.

Durability includes measured values of neutralization tests such as accelerated neutralization tests, measured values of salt permeability tests such as diffusion coefficient tests of chloride ions, and alkali aggregate reactions such as residual expansion tests. Test values, test values related to freeze / thaw tests such as the test method described in “JIS A 1148 (concrete freeze-thaw test method)”, test values related to water permeability such as output method, and various durability performances The measured value of a test etc. are mentioned.
Examples of the color tone include measurement values obtained by the method of “JIS Z 8722 (color measurement method—reflection and transmission object color)” and the like.

Among the “data on the quality of concrete” of the evaluation data in the combination (i), the strength (compressive strength and bending strength of concrete, mortar compressive strength, mortar bending strength, etc.), slump, or slump flow is higher. The preferred combination of monitoring data that can be predicted with accuracy is cement brain specific surface area, sieve test residual amount, color tone a value, color tone L value, cement mineral composition, cement chemical composition (more preferably in cement raw material) Of MgO, Na ₂ O, K ₂ O, Na ₂ Oeq (total alkali), P ₂ O ₅ , and TiO ₂ content of one or more), mortar flow, admixture amount (more preferably , Water reducing agent, AE water reducing agent, high performance water reducing agent, one or more selected from high performance AE water reducing agent), and unit water volume Is is one or more.

The monitoring data in the combination of (ii), “data on cement”, “data on physical properties of cement”, “data on material of concrete other than cement”, and “data on blending of concrete”, respectively, It is the same as “data regarding cement”, “data regarding physical properties of cement”, “data regarding material of concrete other than cement”, and “data regarding blending of concrete” which are monitoring data in the combination of (i). .
Further, the “data related to the quality of concrete” which is the monitoring data in the combination (ii) is the same as the “data related to the quality of concrete” which is the evaluation data in the combination (i). In addition, when using "data regarding the quality of concrete" as monitoring data, it is preferable to use 2 or more types from a viewpoint of improving the precision of prediction of evaluation data.

  The “concrete mixing condition”, which is evaluation data in the combination (ii), is the same as “data relating to the mixing of concrete” which is one of the monitoring data in the combination (i).

In the present invention, a specific type of monitoring data is selected from a plurality of types of monitoring data, and the selected monitoring data is used in the steps (A) to (M) described later, thereby predicting the accuracy of evaluation data. Can be further enhanced.
Note that specific types of monitoring data (one type or a combination of two or more types of monitoring data that can further improve the accuracy of prediction of evaluation data) differ depending on the type of evaluation data to be predicted.
As the specific type of monitoring data, it is preferable to select data having high correlation with the evaluation data to be predicted. A method for selecting a specific type of monitoring data will be described later (step (A-1)).

  In the method for predicting the quality of concrete or the mixing condition of concrete of the present invention, the target concrete is not particularly limited. For example, general structural concrete, cold concrete, mass concrete, high fluidity concrete, low heat generation concrete, expanded concrete, prestressed concrete, low shrinkage concrete, fiber reinforced concrete, polymer concrete, watertight concrete, underwater concrete, permeable drainage concrete, resin Examples include impregnated concrete, shielding concrete, lightweight concrete, pre-packed concrete, shotcrete, recycled concrete, pavement concrete, super hard kneaded concrete, dam concrete, and precast concrete. Moreover, the combination is a combination of indications.

In the present invention, the relationship between the monitoring data in concrete production and the evaluation data related to the evaluation of the quality of the concrete or the mixing condition of the concrete is learned in advance by a neural network, and the learning result is used and only based on the monitoring data. Thus, the evaluation data is predicted.
Hereinafter, the prediction method of the present invention will be described in detail with reference to FIG.
[Step (A-1)]
This step is a step optionally performed before step (A) for the purpose of further improving the accuracy of evaluation data prediction.
In this step, two or more types of monitoring data aggregates including one or more types of arbitrarily selected monitoring data are prepared, and each of the two or more types of monitoring data aggregates is subjected to steps (A) to (M). Steps (A) to (M) are performed by using a plurality of combinations of measured values of monitoring data and measured values of evaluation data (hereinafter also referred to as “selected data”) for selecting monitoring data suitable for use. Learning of an unlearned neural network different from the neural network used in the above, and an estimated value of the evaluation data obtained by inputting the actual measurement value of the monitoring data of the selected data to the input layer of the obtained neural network, Root mean square error (RMSE) with the measured value of the evaluation data of the selected data is calculated, and the numerical value of the mean square error is the smallest Monitoring data in Tsu selection data, shall be used as the monitoring data in the step (A) ~ (M).

  One or more arbitrarily selected monitoring data are listed as data on the above-mentioned cement, data on cement physical properties, data on concrete materials other than cement, data on concrete mix, and data on concrete quality. One or more (preferably two or more, more preferably three or more) data arbitrarily selected from the data.

In this step, two or more (preferably three or more, more preferably four or more) monitoring data aggregates comprising one or more arbitrarily selected monitoring data are prepared, and the two or more monitoring data aggregates are prepared. For each of the bodies, learning of an unlearned neural network is performed using a plurality of selection data, which is a combination of the actual measurement values of the monitoring data and the evaluation data of the aggregate.
Specifically, a plurality of samples for selection are prepared, and measured values of monitoring data of the samples (one or more types of arbitrarily selected monitoring data) and measured values of evaluation data to be predicted are measured. These are used as selection data. Of the selection data, the actual value of the monitoring data is input to the input layer of the neural network, the estimated value of the evaluation data output from the output layer, and the evaluation data of the selection data corresponding to the estimated value of the evaluation data Learning of the neural network is performed by performing an arbitrary number of learnings by comparing and evaluating the measured values to correct the neural network.
The number of samples for selection is preferably 10 or more, more preferably 14 or more, still more preferably 16 or more, and particularly preferably 20 or more, from the viewpoint that monitoring data that can be predicted with higher accuracy can be selected. . The upper limit of the number of samples is not particularly limited, but is, for example, 10,000 from the viewpoint of workability.
The number of learning times of the neural network is not particularly limited, but is preferably 100 to 2,000 times, more preferably 200 to 1,500 times.

[Step (A)]
In step (A), initial setting of the number of learning is performed. The number of learning times to be set is not particularly limited, but is preferably a sufficiently large number such that overlearning (overlearning) of the neural network occurs. Specifically, it is usually 5,000 to 1,000,000 times, preferably 10,000 to 100,000 times.
In step (A), it is preferable to set the number of learnings that causes over-learning of the neural network, specifically, the number of learnings such that σ _L <σ _M (details will be described later). Since the number of learning times is increased / decreased, there is no problem even if the number of learning times initially set in the step (A) is the number of learning times normally performed for learning of the neural network.
A process (B) is implemented after completion | finish of a process (A).

[Step (B)]
In the step (B), the learning of the neural network is set in the previous step by using a plurality of combinations of the actual measurement values of the monitoring data for learning and the actual measurement values of the evaluation data (hereinafter also referred to as “learning data”). Do the learning times. The number of the combinations is, for example, 5 or more, preferably 7 or more. Although the upper limit of the number of the said combination is not specifically limited, For example, it is 1,000.
Here, the “number of learnings set in the previous step” is the number of learnings set in the step (A) or the new number of learnings reset in the step (D), and the most recent step ( The number of learning times set in the step (A) or the step (D)).
Specifically, a plurality of samples for learning are prepared, and measured values of monitoring data of the samples and measured values of target evaluation data are measured and used as learning data. Among the learning data, the actual value of the monitoring data is input to the input layer of the neural network, the estimated value of the evaluation data output from the output layer, and the evaluation data of the learning data corresponding to the estimated value of the evaluation data The neural network is learned by comparing and evaluating the actual measurement values and correcting the neural network by performing the set number of learning times.
The number of learning samples is preferably 10 or more, more preferably 14 or more, still more preferably 16 or more, and particularly preferably 20 or more, from the viewpoint of performing prediction with higher accuracy. The upper limit of the number of samples is not particularly limited, but is, for example, 10,000 from the viewpoint of workability.
When the learning number is changed and the neural network is re-learned, the neural network obtained as a result of the previous learning is initialized and learning is performed again.
A process (C) is implemented after completion | finish of a process (B).

[Step (C)]
In step (C), σ _L and σ _M are calculated. From the magnitude relationship between σ _L and σ _M , it can be determined whether learning has been performed a sufficiently large number of times to cause over-learning of the neural network.
Specifically, the estimated value of the evaluation data obtained by inputting the actual measurement value of the monitoring data of the learning data to the input layer of the neural network learned in the most recent step (B) and the evaluation data of the learning data The mean square error (σ _L ) with the actual measurement value is calculated. Next, the estimated value of the monitoring data and the estimated value of the evaluation data obtained by inputting the measured value of the monitoring data of the monitoring data to the input layer of the neural network learned in the most recent step (B) The mean square error (σ _M ) is calculated. Thereafter, by comparing the calculated values of σ _L and σ _M , it can be determined whether the learning of the neural network has been performed a sufficiently large number of times.
Here, the monitor data is a combination of the actual measurement value of the monitoring data and the actual measurement value of the evaluation data obtained from a sample different from the sample used for obtaining the learning data. This is data for confirmation.
From the viewpoint of workability, the number of samples of monitor data (a combination of actual values of monitoring data and actual values of evaluation data) is preferably 5 to 50%, more preferably 10 to 30% of the number of samples of learning data. is there.

When the relationship between σ _L and σ _M calculated in the step (C) is σ _L ≧ σ _M (“No” in the overlearning determination in FIG. 1), the number of learnings in the most recently performed step (B) is It can be determined that the number of times is not sufficiently large. In this case, step (D) is performed. When the relationship between σ _L and σ _M calculated in the step (C) is σ _L <σ _M (“Yes” in the over-learning determination in FIG. 1), the number of learnings of the most recently performed step (B) is It can be determined that the number of times was sufficiently large. In this case, step (E) is performed.

[Step (D)]
In the step (D), a learning number larger than any of the learning number set in the most recent step (A) and the reset number of learning times of the latest neural network is reset as a new learning number ( For example, a number obtained by multiplying the number of learning performed in the most recent step (B) by 2.0 is set as a new number of learning. After resetting the new number of learning times, steps (B) to (C) are performed again.

[Step (E)]
In step (E), the learning number obtained by reducing the number of learnings performed in the latest neural network learning is reset as a new learning number (for example, performed in the most recent step (B) or step (F)). The number obtained by multiplying the number of learning times by 0.95 is set as the new number of learning times.)
Note that learning of the latest neural network refers to learning performed in the near past. Specifically, it refers to learning performed in the nearer past in step (B) or step (F) described later.
A process (F) is implemented after completion | finish of a process (E).

[Step (F)]
In step (F), the learning data used in the latest step (B) is used to perform neural network learning for the number of learning times set in the latest step (E).
The contents to be implemented in the step (F) are the same as those in the step (B) except that the neural network learning is performed the number of times newly set in the step (E).
A process (G) is implemented after completion | finish of a process (F).

[Step (G)]
In step (G), end determination is performed using the neural network obtained in the learning of the latest step (F). Specifically, the actual value of the monitoring data of the learning data is input to the input layer of the neural network obtained by the learning in the latest step (F), and the estimated value of the evaluation data and the evaluation data of the learning data Obtained by inputting the mean square error (σ _L ) from the actual measurement value and the actual measurement value of the monitoring data in the monitor data to the input layer of the neural network obtained by the learning of the most recent step (F). The mean square error (σ _M ) between the estimated value of the evaluation data and the actually measured value of the evaluation data in the monitor data is calculated, and the relationship between the calculated σ _L and σ _M is σ _L ≧ σ _M In this case (“Yes” in the end determination in FIG. 1), it can be determined that the number of times of learning in the most recent step (F) is no longer sufficiently large. In this case, step (I) described later is performed. When the calculated relationship between σ _L and σ _M is σ _L <σ _M (“No” in the end determination in FIG. 1), the number of times of learning in the most recent step (F) is still a sufficiently large number. It can be judged that there was. In this case, the step (H) described later is performed.

[Step (H)]
In step (H), it is determined whether the neural network learning count in the most recent step (F) has exceeded a predetermined numerical value. Step (H) is performed to avoid repeating steps (E) to (G) indefinitely. If the number of times the neural network has been learned in step (F) most recently performed in step (H) exceeds a predetermined value (“Yes” in FIG. 1), steps (E) to (G) are performed again. To do. When the number of learnings of the step (F) performed most recently in the step (H) is equal to or less than a predetermined numerical value (“No” in FIG. 1), the step (J) or (K) described later is performed.
The predetermined numerical value is not particularly limited, and may be, for example, a numerical value equal to or less than 1/100 of the number of learning times set in the step (E), 1 or less, or 0 or less.

上記式（１）中、学習データの平均２乗誤差（σ_Ｌ）とは、直近の工程（Ｇ）で算出された平均２乗誤差（σ_Ｌ）と同じである。評価データの推測値の平均値とは、学習データの監視データの実測値を、直近の工程（Ｆ）にて得られたニューラルネットワークの入力層に入力して得られた評価データの推測値の平均値である。
解析度の判定を行うことで学習を行ったニューラルネットワークを用いて、コンクリートの品質等の予測を高い精度で行うことができるか否かを判断することができる。
解析度判定値が予め定めた第一の設定値未満（図１の第一の解析度判定における「Ｙｅｓ」）であれば、解析は十分であると判断され、ニューラルネットワークの学習は終了する。
解析は十分であると判断された学習済みのニューラルネットワークは、本発明の予測方法に用いられる。
具体的には、学習済みのニューラルネットワークの入力層に、コンクリート製造における監視データの実測値を入力して、学習済みのニューラルネットワークの出力層から、コンクリートの品質またはコンクリートの配合条件の評価に関連する評価データの推測値を出力することで、コンクリートの品質またはコンクリートの配合条件を予測することができる。 [Step (I)]
In step (I), the analysis level can be determined depending on whether or not the analysis level determination value is less than a predetermined first set value. The analysis degree determination value is calculated using the following formula (1).

In the above formula (1), the average of the training data square error between (sigma _L) is the same as the mean square error calculated in the last step (G) (sigma _L). The average value of the estimated value of the evaluation data is the estimated value of the evaluation data obtained by inputting the actually measured value of the monitoring data of the learning data to the input layer of the neural network obtained in the latest step (F). Average value.
It is possible to determine whether or not the quality of the concrete can be predicted with high accuracy using the neural network that has been learned by determining the degree of analysis.
If the analysis level determination value is less than a predetermined first set value (“Yes” in the first analysis level determination in FIG. 1), it is determined that the analysis is sufficient, and the learning of the neural network ends.
A learned neural network determined to be sufficient for analysis is used in the prediction method of the present invention.
Specifically, the actual value of monitoring data in concrete production is input to the input layer of the learned neural network, and the quality of the concrete or concrete mixing conditions is evaluated from the output layer of the learned neural network. By outputting the estimated value of the evaluation data to be performed, it is possible to predict the quality of the concrete or the mixing condition of the concrete.

If the analysis degree determination value is equal to or greater than a predetermined first set value (“No” in the first analysis degree determination in FIG. 1), the neural network learned using the learning data is used as it is, and the concrete is used. It is determined that the quality and the like cannot be predicted with high accuracy, and step (J) is performed.
The predetermined first set value is not particularly limited, but is preferably 6% or less, more preferably 5% or less, and particularly preferably 3% or less from the viewpoint of performing prediction with higher accuracy.
In steps (A) to (I), the analysis degree determination value in step (I) is less than a predetermined first set value or the number of times exceeds a preset number in step (J). Repeated. The analytical determination value and the learned neural network obtained each time the step (I) is performed are used in the step (K).

[Process (J)]
In step (J), it is determined whether or not the number of times step (A) has been performed is equal to or less than a preset numerical value. By performing the determination, it is possible to avoid repeating steps (A) to (I) indefinitely.
In the step (J), when the number of times the step (A) is performed is equal to or less than the preset number (“Yes” in the number determination of FIG. 1), the learning condition is initialized, and the steps (A) to (A) to again are performed. (I) is performed, and when the number of times exceeds a preset number of times (“No” in the number of times determination of FIG. 1), step (K) is performed.
The number of times set in advance is not particularly limited, but is usually 5 times or more. The upper limit of the preset number of times is preferably 100 times or less from the viewpoint of preventing steps (A) to (I) from being repeated a great deal.

  As a method for initializing learning conditions, for example, a method of re-inputting learning data after randomly changing a threshold value of units constituting a neural network or a weight combining units, and obtaining learning data Examples include a method of inputting new learning data after increasing the number of samples, changing the type of monitoring data to be used, or excluding inappropriate learning data.

[Step (K)]
In step (K), the next prediction is performed depending on whether or not the smallest analysis degree determination value calculated in step (I) is less than a predetermined second set value. It can be determined whether or not.
By adding the determination of the step (K), even in a learned neural network that has been determined in step (I) that prediction of cement quality or the like cannot be performed with high accuracy, By performing L) to (M), it is possible to determine whether or not the cement quality and the like can be predicted with high accuracy.
When the smallest analytical determination value is less than a predetermined second set value (“Yes” in the second analytical determination of FIG. 1), the process (I ) Is obtained as a learned neural network, and then step (L) is performed.

If the smallest analysis degree determination value is equal to or greater than a predetermined second set value (“No” in the second analysis degree determination in FIG. 1), using a neural network that has learned using learning data, It is determined that the next prediction such as the quality of the concrete cannot be performed with high accuracy, and the prediction is terminated.
The predetermined second set value is larger than the first set value. The upper limit is preferably 30% or less, more preferably 20% from the viewpoint of performing prediction with higher accuracy.

[Step (L)]
In step (L), a predictable monitoring data area used in step (M) is set.
First, the actual measurement value and evaluation data of the monitoring data used as learning data in the step (I) in which the smallest analysis degree determination value can be obtained among all the analysis degree determination values calculated in the step (I). Perform a non-correlation test for the combination of actual measurements. When there are two or more types of monitoring data judged to be significant at the significance level of 5% in the uncorrelated test (“Yes” in the decorrelation test in FIG. 1), the significance is significant at the significance level of 5%. A coordinate space having all the types of monitoring data determined to be coordinate axes is created.
For example, when there are two types of monitoring data determined to be significant at the significance level of 5%, ie, the cement specific surface area of cement and the amount of C ₃ S of cement, the cement specific surface area of x is used as the x axis, A coordinate space having the y-axis as the amount of C ₃ S is created (see FIG. 2).

Next, among the actual measurement values of the monitoring data used as learning data, all the actual measurement values of the types of monitoring data determined to be significant at the significance level of 5% were plotted in the coordinate space and plotted in the coordinate space. A predictable monitoring data area is set by connecting monitoring data. The predictable monitoring data area is an area including all of the plotted monitoring data, and is an area formed by connecting the monitoring data so that the area is maximized (see FIG. 2). After setting the predictable monitoring data area, the process (M) is performed.
When the monitoring data judged to be significant at the significance level of 5% is 0 or 1 type (“No” in the uncorrelated test in FIG. 1), it is judged that the cement quality or the production condition cannot be predicted. To finish the prediction.

[Process (M)]
In the process (M), using the actual measurement value of the monitoring data in the concrete production and the coordinate space created in the process (L), the actual value of the monitoring data in the concrete production and the learned neural network obtained in the process (K). Thus, it can be determined whether or not the quality of the concrete can be predicted with high accuracy.
When the actual value of monitoring data in concrete production used for prediction of concrete quality, etc. is included in the predictable monitoring data area set in step (L) (“Yes” in the coordinate judgment of FIG. 1), concrete It is determined that the quality of the product can be predicted with high accuracy, and the actual value of the monitoring data in concrete production is input to the input layer of the learned neural network obtained in the step (K). By outputting an estimated value of evaluation data related to the evaluation of the quality of the concrete or the mixing condition of the concrete from the output layer, the quality of the concrete or the mixing condition of the concrete can be predicted.
When the actual measurement value of monitoring data in concrete production is not included in the predictable monitoring data area ("No" in the coordinate determination in Fig. 1), it is determined that the quality of concrete or the concrete mixing condition cannot be predicted. To finish the prediction.
It should be noted that the monitoring data in concrete production is determined to be significant at a significance level of 5% in the uncorrelated test using actual measurement values of types of monitoring data not used as coordinate axes in the coordinate space created in step (L). If there is a type of monitoring data that does not exist), there is no limitation on the actual measurement value of the monitoring data from the types of monitoring data that are not used as the coordinate axes.

According to the prediction method of the present invention, the quality of concrete or the mixing condition of concrete can be predicted with higher accuracy.
In the present invention, the learning of the neural network after the learning first with a sufficiently large number of learning (sigma _{L <learning} frequency to the extent that the sigma _M), while reducing the number of times of learning, the learning of the neural network sigma _L it is intended to be repeated until the ≧ σ _M. According to this method, even if the numerical values of σ _L and σ _M vary due to factors such as when the evaluation data is insufficient in the learning data, the variation can be corrected. Can learn appropriately.
In addition, in the process (I), even if the analysis degree determination value does not satisfy a predetermined reference value due to a small number of samples of learning data, etc., monitoring data in concrete production that is input to the input layer Is included in the predictable monitoring data area, the quality of concrete and the like can be predicted with high accuracy using the monitoring data and the learned neural network.

  In order to maintain the prediction accuracy at a high level, the neural network periodically checks the magnitude of the difference between the estimated value of the evaluation data and the actually measured value corresponding to the estimated value, and based on the inspection result, It is preferable to update the neural network.

According to the method for predicting concrete quality or concrete mixing conditions of the present invention, by using a neural network, an estimated value of evaluation data such as compressive strength of concrete can be obtained within one hour simply by inputting monitoring data. Can be obtained.
In addition, based on the estimated value of the obtained evaluation data, early detection of concrete quality abnormalities during concrete production, and optimization of various conditions in raw material composition correction, kneading method, curing method, etc. Proper quality concrete can be manufactured.
Specifically, when an abnormality is found in the estimated value of the compressive strength of the concrete, the compressive strength of the concrete can be achieved by correcting the raw material composition.
It is also possible to correct the manufacturing target based on the estimated value of the evaluation data.
For example, if it is predicted that the compressive strength of concrete will not reach the target value, analyze the relationship between the monitoring data (factors) used for learning and the compressive strength of concrete, and use concrete, raw material composition, curing method, etc. By optimizing the manufacturing recipe, the quality of the concrete can be achieved.

Furthermore, the monitoring data is artificially changed based on the evaluation data by connecting the computer for controlling the concrete production and the computer used for carrying out the method for predicting the concrete quality or the concrete mixing condition of the present invention. The control system for this can also be automated.
In the present invention, examples of software for performing an operation using a neural network include “Neural Network Library” (trade name) manufactured by OLSOFT.

EXAMPLES The present invention will be specifically described below with reference to examples, but the present invention is not limited to these examples.
[Example 1]
[Prediction of compressive strength of concrete at age 28 days]
[Materials used]
The materials used are as shown below.
(1) Ordinary Portland cement: Taiheiyo Cement Co., Ltd. (2) Fine aggregate: Mixed sand made by mixing crushed sand from Yamaguchi City, Yamaguchi Prefecture and sea sand from Higashimatsuura-gun, Saga Prefecture at a mass ratio of 1: 1. , Density 2.59 g / cm ³
(3) Coarse aggregate: hard sandstone crushed stone from Kitakyushu City, Fukuoka Prefecture, maximum size 20 mm, density 2.65 g / cm ³
(4) High-performance AE water reducing agent: BASF Japan, trade name “Master Grenium SP8SV”
(5) Antifoaming agent: BASF Japan, trade name “Master Air 404”

Concrete was manufactured using the above materials for 134 normal Portland cements having different sampling dates as samples for selection and learning (hereinafter also simply referred to as “learning samples”). The composition of each material was such that the unit cement amount was 330 kg / m ³ , the fine aggregate rate was 45% by mass, and the high-performance AE water reducing agent was 0.7% by mass with respect to 100% by mass of cement. The amount of unit water and the amount of antifoaming agent added is 8% or less, and the slump immediately after kneading obtained in accordance with “JIS A 1101 (concrete slump test method)” is 8 Adjustment was made to be ± 2.5 cm.
For the production of concrete, kneading, preparation of specimens and curing were performed in accordance with “JIS A 1138 (How to make concrete in a test room)” and “JIS A 1108 (How to make a specimen for concrete strength test)”. Thereafter, the compressive strength at the age of 28 days was measured in accordance with “JIS A 1108 (Concrete compressive strength test method)”.
The obtained compressive strength at the age of 28 days was used as an actual measurement value of evaluation data in selection data and learning data (hereinafter, also simply referred to as “learning data”).

For each of the 134 ordinary Portland cements used in the learning samples, the cement-related surface area, 31 μm net sieve test residual amount, color tone a value, amount of each mineral, and P ₂ O ₅ The amount was measured and used as an actual measurement value of monitoring data in learning data.
In addition, the color tone a value is a reflection measured by using a spectral color difference meter (manufactured by Nippon Denshoku Industries Co., Ltd., “SE6000”) in accordance with “JIS Z 8722 (color measurement method—reflection and transmission object color)”. It is a measured value.
The amount of each mineral was measured at a measurement range of 2θ = 10 to 65 ° using a powder X-ray diffractometer (manufactured by Bruker AXS, “D8 ADVANCE”), and Rietveld analysis software (Bruker The amount of C ₃ S, C ₂ S, C ₃ A, and C ₄ AF calculated by “DIFFRAC ^plus TOPAS (Ver. 3)” manufactured by AXS Co., Ltd.).

[Select monitoring data]
Among the above-mentioned monitoring data (cement specific surface area of cement, 31 µm net sieve test residual amount, color tone a value, C ₃ S, C ₂ S, C ₃ A and C ₄ AF amount, P ₂ O ₅ amount), For each of the conditions 1 to 4 (monitoring data aggregates 1 to 4) for selecting the monitoring data shown in Table 1 (in Table 1, the selected monitoring data is indicated by “◯”), the actual value of the selected monitoring data And the unlearned neural network was trained using the measured values of the evaluation data. The learning was performed 1200 times.
Mean square error between the estimated value of the evaluation data obtained by inputting the actual measurement value of the selected monitoring data to the input layer of the neural network after learning ("RMSE" in Table 1) Calculated).
The results are shown in Table 1.
Among the conditions 1 to 4, the combination of the monitoring data selected in the condition 1 in which the mean square error value was the smallest (Brain specific surface area, 31 μm screen sieve test residual amount, color tone a value, C ₃ S, C ₂ S, The amounts of C ₃ A and C ₄ AF and the amount of P ₂ O ₅ were used as monitoring data (learning data and monitoring data) used for learning of the neural network.

In addition, as a sample for monitoring (hereinafter, also referred to as “monitor sample”), 13 normal Portland cements having different sampling dates from the 134 samples, the compressive strength at the age of 28 days, Measurements were made in the same manner as the learning data, and were used as monitor data (measured values of evaluation data). Furthermore, the Blaine specific surface area, the 31 μm net sieve test residual amount, the color tone a value, the amount of each mineral, and the amount of P ₂ O ₅ of the above 13 ordinary Portland cements were measured in the same manner as the learning data, and monitored in the monitor data. The measured value of data was used.

The neural network was learned using the learning data. As the neural network, a hierarchical neural network having an input layer, an intermediate layer, and an output layer was used. The neural network is an unlearned one different from the neural network used for selecting the monitoring data.
The neural network was first learned 10,000 times using the learning data and the monitor data. When σ _L and σ _{M were} calculated using the obtained neural network, the relationship between σ _L and σ _M was σ _L <σ _M.
Thereafter, the neural network is initialized, and the learning network and the monitor data are used to perform learning of the neural network by performing the number of learning times (rounded down) by multiplying the number of learning times by 0.95. It repeated until the relationship of (sigma) _L calculated using (sigma) _L and (sigma) _M became (sigma) _L > = (sigma) _M.
As a result, the analysis degree determination value was 1.67%, so learning was terminated.

The compressive strength of the concrete at the age of 28 days was measured for the concrete manufactured under the same conditions as the concrete used in the learning process, using ordinary Portland cement A different from the above sample. The result was 58.4 N / mm ² .
On the other hand, to the input layer of the obtained learned neural network, the Blaine specific surface area of normal Portland cement A, the 31 μm net sieve test residual amount, the color tone a value, the amount of each mineral, and the amount of P ₂ O ₅ are input. The estimated value of the compressive strength of concrete at the age of 28 days was output.
The obtained estimated value was 58.3 ± 2.9 N / mm ² (deviation indicates 3σ), and the actually measured value and the estimated value almost coincided.

[Comparative Example 1]
Data on 147 pieces of normal Portland cement used in Example 1 (134 learning samples + 13 monitoring samples) as a population, and data on cement having a high correlation with compressive strength at 28 days of concrete The cement brane specific surface area, 31 μm net sieve screen residual amount, and the amount of each mineral (the amount of C ₃ S, C ₂ S, C ₃ A, and C ₄ AF), which were (cement character), were used as explanatory variables. The following regression equation (determination coefficient R ² = 0.20) was obtained.
(Compressive strength of concrete at age 28 (N / mm ² )) = 0.005 × (Brain specific surface area of cement (cm ² /g))+(−0.3)×(31 μm net sieve test residual amount (Mass%)) + 0.2 × (C ₃ S amount (mass%)) + (− 0.2) × (C ₂ S amount (mass%)) + 0.3 × (C ₃ A amount (mass%) ) + 0.2 × (C ₄ AF amount (mass%)) + 91
Concrete compression at a material age of 28 days obtained by substituting the Blaine specific surface area of the normal Portland cement A used in Example 1, the 31 μm net sieve test residual amount, and the amount of each mineral into the regression equation above, respectively. The estimated value of intensity was 58.9 ± 5.2 N / mm ² (the deviation indicates 3σ).

[Example 2]
[Prediction of concrete slump]
[Used material 1]
The materials used for concrete (ordinary Portland cement, fine aggregate, coarse aggregate, high-performance AE water reducing agent, antifoaming agent) are the same as those used in Example 1.
[Used material 2]
The materials used for mortar in the mortar flow test used as monitoring data in the monitor data are as shown below.
(6) Ordinary Portland cement: Taiheiyo Cement (same as (1) above)
(7) Fine aggregate: Cement Association (sales company) Standard sand for cement strength test (8) High performance water reducing agent: Product name "Mighty 150" manufactured by Kao Corporation
(9) Antifoaming agent: Onoda Co., Ltd., trade name “Nicofix 800”

Concrete was produced using the above materials for 18 ordinary Portland cements with different sampling dates as selection and learning samples.
The blending amount of each material is as follows: the unit cement amount is 330 kg / m ³ , the fine aggregate ratio is 45% by mass, the unit water amount is 165 kg / m ³ , and the high-performance water reducing agent is 0.7% by mass with respect to 100% by mass cement. It was made to become. Moreover, the concrete which adjusted the addition amount of the antifoamer so that air amount might be 2% or less was manufactured, and the slump of the concrete immediately after kneading | mixing was measured.
The concrete was produced in accordance with “JIS A 1138 (How to make concrete in a test room)”. The concrete slump was measured in accordance with “JIS A 1101 (Concrete slump test method)”.
The obtained slump was used as an actual measurement value (selection data and learning data) of evaluation data.

In addition, for each of the 18 ordinary Portland cements, as cement-related data, the Blaine specific surface area of the cement, the 31 μm net sieve test residual amount, the amount of each mineral were measured, and the mortar flow was measured as data related to the physical properties of the cement. The measured value was used as the actual measurement value of the monitoring data in the learning data.
The amount of each mineral is measured with a powder X-ray diffractometer (D8 ADVANCE manufactured by Bruker AXS Co., Ltd.) in the measurement range: 2θ = 10 to 65 °, and Rietveld analysis software ( The amount of C ₃ S and C ₃ A calculated by DIFFRAC ^plus TOPAS (Ver. 3) manufactured by Bruker AXS Co., Ltd., and the mortar flow is the value immediately after kneading by the following method and 30 minutes after kneading. Is the value of
[Mortar flow test]
The above material has a water cement ratio of 0.35, a fine aggregate to cement mass ratio (fine aggregate / cement) of 2.0, a high-performance water reducing agent of 1.2% by mass with respect to 100% by mass of cement, A mortar prepared by mixing so that the defoamer is 0.1% by mass with respect to 100% by mass of cement, and kneaded according to the flow test of “JIS A 1171 (Testing method for polymer cement mortar)”. Immediately after and 30 minutes after kneading, the mortar flow was measured.

[Select monitoring data]
The monitoring data shown in Table 2 was selected from the above monitoring data (cement specific surface area of cement, 31 μm net sieve test residual amount, amounts of C ₃ S and C ₃ A, mortar flow (immediately, after 30 minutes)). For each of the conditions 1 to 4 (in Table 2, the selected monitoring data is indicated by “◯”), learning of an unlearned neural network using the actual value of the selected monitoring data and the actual value of the evaluation data Went. The learning was performed 900 times.
Mean square error between the estimated value of the evaluation data obtained by inputting the actual measurement value of the selected monitoring data to the input layer of the neural network after learning ("RMSE" in Table 2) Calculated).
The results are shown in Table 2.
Among the conditions 1 to 4, the combination of the monitoring data selected in the condition 1 having the smallest mean square error value (cement specific surface area of the cement, 31 μm net sieve test residual amount, amounts of C ₃ S and C ₃ A, Mortar flow (immediately, after 30 minutes)) was used as monitoring data (learning data and monitor data) used for learning of the neural network.

In addition, using two normal Portland cements with different sampling dates as the 18 samples, the concrete slump was measured in the same way as the learning data, and the monitor data (actual value of the evaluation data) was measured. ).
Furthermore, the Blaine specific surface area, 31 μm mesh sieve test residual amount, the amount of each mineral, and the mortar flow of the above two ordinary Portland cements were measured in the same manner as the learning data to obtain monitor data (actual value of monitoring data). .

The neural network was learned using the learning data. As the neural network, a hierarchical neural network having an input layer, an intermediate layer, and an output layer was used. The neural network is an unlearned one different from the neural network used for selecting the monitoring data.
The neural network was first learned 10,000 times using the learning data and the monitor data. When σ _L and σ _{M were} calculated using the obtained neural network, the relationship between σ _L and σ _M was σ _L <σ _M.
Thereafter, the neural network is initialized, and the learning network and the monitor data are used to perform learning of the neural network by performing the number of learning times (rounded down) by multiplying the number of learning times by 0.95. It repeated until the relationship of (sigma) _L calculated using (sigma) _L and (sigma) _M became (sigma) _L > = (sigma) _M. As a result, the analysis level determination value was 3.82%, so learning was terminated.
Using normal Portland cement B different from the above sample, slump was measured for the concrete manufactured under the same conditions as the concrete used in the learning process. The result was 9.0 cm.
On the other hand, the learned neural network obtained was obtained for the Blaine specific surface area, the 31 μm net sieve test residual amount, the amount of each mineral, and the mortar fluidity of 20 ordinary Portland cements used as learning samples and monitoring samples. The estimated slump value was output from the output layer.
The estimated value of the obtained slump was 8.4 ± 0.9 cm (the deviation indicates 3σ).

[Comparative Example 2]
Data (cement character) and cement related to cement having high correlation with concrete slump, with data on 20 normal portland cements used in Example 2 (18 learning samples + 2 monitoring samples) as a population. Explains the Blaine specific surface area of cement, 31 μm net sieve test residual amount, amount of each mineral (amount of C ₃ S and C ₃ A), and mortar flow (immediately, after 30 minutes) Multiple regression analysis using variables was performed to obtain the following regression formula (determination coefficient R ² = 0.47).
(Concrete slump (cm)) = (− 0.0002) × (cement brane specific surface area (cm ² /g))+(−0.2)×(31 μm net sieve test residual amount (mass%)) + 0 0.03 × (C ₃ S amount (mass%)) + (− 0.4) × (C ₃ A amount (mass%)) + 0.002 × (mortar flow (immediately after) value (mm)) + 0.01 × (Mortar flow (after 30 minutes) value (mm)) + 15
The estimated value of the slump obtained by substituting the Blaine specific surface area of the 20 ordinary Portland cements used in Example 2, the 31 μm net sieve test residual amount, the amount of each mineral, and the mortar flow into the above regression equation, It was 8.0 ± 1.5 cm (the deviation shows 3σ).

The estimated value obtained in Example 1 (58.3 ± 2.9 N / mm ³ ) and the estimated value obtained in Comparative Example 1 (58.9 ± 5) regarding the prediction of compressive strength of concrete at 28 days of age. .2 N / mm ³ ), it can be seen that Example 1 is closer to the actually measured value (58.4 N / mm ² ) of the evaluation data and more reliable than Comparative Example 1.
Furthermore, when comparing the estimated value (8.4 ± 0.9 cm) obtained in Example 2 regarding prediction of concrete slump with the estimated value (8.0 ± 1.5 cm) obtained in Comparative Example 2, It can be seen that Example 2 is closer to the actually measured value (9.0 cm) of the evaluation data and has higher reliability than Comparative Example 2.

[Example 3]
[Prediction of concrete slump]
[Used material 1]
The materials used for concrete (ordinary Portland cement, high-performance AE water reducing agent, fine aggregate, coarse aggregate, antifoaming agent) are the same as those used in Example 1 (hereinafter the same applies to Examples 4 to 6). .
[Used material 2]
Moreover, the material used of the mortar in the mortar flow test used as monitoring data in the monitor data is the same as the material used in Example 2 (hereinafter, the same applies to Examples 4 to 6).

Concrete was produced using the above materials for 15 ordinary Portland cements with different sampling dates as selection and learning samples.
The blending amount of each material is such that the unit cement amount is 330 kg / m ³ , the fine aggregate rate is 45% by mass, the unit water amount is 165 kg / m ³ , and the amount of the high-performance AE water reducing agent is 100% by mass in concrete. It was made to be -1.0 mass%. Moreover, the addition amount of the antifoamer was adjusted so that the air amount would be 2% or less. The concrete slump immediately after kneading was measured in the same manner as in Example 2.
The obtained slump was used as an actual measurement value (selection data and learning data) of evaluation data.

For the 15 learning samples, the amount of high-performance AE water reducing agent as data related to the mix of concrete, the amount of each mineral as data related to cement, and the mortar flow as data related to the physical properties of cement, the monitoring data in the learning data The measured value was used.
The amount of the high-performance AE water reducing agent is an amount obtained by adding the target slump as 8 ± 1 cm in the concrete manufactured as the learning sample.
The amount of each mineral is the amount of C ₃ S and C ₃ A calculated in the same manner as in Example 2 for the 15 ordinary Portland cements.
The mortar flow is a value immediately after kneading and a value after 30 minutes after kneading (30 minutes after the end of kneading), measured in the same manner as in Example 2 for the 15 ordinary Portland cements.

[Select monitoring data]
Of the monitoring data described above (the amount of high-performance AE water reducing agent, the amount of C ₃ S and C ₃ A, the mortar flow (immediately, after 30 minutes)), the conditions 1 to 4 for selecting the monitoring data shown in Table 3 For each (the selected monitoring data is indicated by “◯” in Table 3), an unlearned neural network was trained using the measured value of the selected monitoring data and the measured value of the evaluation data. The learning was performed 1300 times.
Mean square error between the estimated value of the evaluation data obtained by inputting the actual measurement value of the selected monitoring data to the input layer of the neural network after learning ("RMSE" in Table 3) Calculated).
The results are shown in Table 3.
Among the conditions 1 to 4, the combination of the monitoring data selected in the condition 1 in which the value of the mean square error was the smallest (amount of high-performance AE water reducing agent, amounts of C ₃ S and C ₃ A, mortar flow (immediately, 30 minutes later)) was used as monitoring data (learning data and monitor data) used for learning of the neural network.

In addition, using two normal Portland cements with different sampling dates as the 15 samples, the concrete slump was measured in the same way as the learning data, and the monitor data (actual value of the evaluation data was measured). ).
Furthermore, the amount of the high-performance AE water reducing agent of the concrete using the two ordinary Portland cements, the amount of each mineral of the two ordinary Portland cements, and the mortar flow using the two ordinary Portland cements, It was obtained in the same manner as the learning data and used as monitor data (actually measured value of monitoring data).

The neural network was learned using the learning data. As the neural network, a hierarchical neural network having an input layer, an intermediate layer, and an output layer was used. The neural network is an unlearned one different from the neural network used for selecting the monitoring data.
The neural network was first learned 10,000 times using the learning data and the monitor data. When σ _L and σ _M were calculated using the obtained neural network, the relationship between σ _L and σ _M was σ _L <σ _M.
After that, the neural network is initialized, and the learning data and the monitor data are used to learn the neural network by performing the number of learning times (rounded down) by multiplying the number of learning times by 0.95. The relationship between σ _L and σ _M calculated using the network was repeated until σ _L ≧ σ _M. As a result, the analysis level determination value was 2.56%, so the learning was terminated.
Using normal Portland cement B different from the above sample, slump was measured for the concrete manufactured under the same conditions as the concrete used in the learning process. The result was 8.5 cm.
On the other hand, in the learning data and the monitor data, the combination of the actual measurement values of the 17 monitoring data (a combination of the amount of high-performance AE water reducing agent, the amount of each mineral, and the mortar flow) Input to the input layer and output the estimated slump value from the output layer.
The estimated value of the obtained slump was 8.2 ± 0.6 cm (deviation shows 3σ), and the actually measured value and the predicted value almost coincided.

[Comparative Example 3]
Data related to concrete blending, which has a high correlation with concrete slump, with data regarding 17 ordinary Portland cements used in Example 3 (15 learning samples + 2 monitoring samples) as a population, and cement Explains the amount of high-performance AE water reducing agent, amount of each mineral (amount of C ₃ S and C ₃ A), and mortar flow (immediately after 30 minutes) Multiple regression analysis was performed using variables, and the following regression equation (determination coefficient R ² = 0.42) was obtained.
(Concrete slump (cm)) = (8.95) × (Amount of high performance AE water reducing agent (mass%)) + (0.04) × (C ₃ S amount (mass%)) − 1.09 × (C ₃ A amount (% by mass)) + 1.9 × 10 ⁻³ × (mortar flow (immediately after) value (mm)) − 1.5 × 10 ⁻³ × (mortar flow (after 30 minutes) value (mm) ) +10.7
The estimated value of the slump obtained by substituting the amount of the high-performance AE water reducing agent, the amount of each mineral, and the mortar flow, which is data on the 17 ordinary Portland cements used in Example 3, into the regression equation is as follows. 8.1 ± 1.4 cm (deviation shows 3σ).

[Example 4]
[Prediction of concrete slump]
Concrete was produced using the above materials for 15 ordinary Portland cements with different sampling dates as selection and learning samples.
The blending amount of each material is such that the mass ratio of water to cement (water / cement) is 0.3, the fine aggregate ratio is 45% by mass, and the amount of high-performance AE water reducing agent is 0.7% with respect to 100% by mass of cement. It was made to become mass%. Moreover, the concrete which adjusted the addition amount of the antifoamer so that air amount might be 2% or less was manufactured, and the slump of the concrete immediately after kneading | mixing was measured.
The concrete production and the slump measurement method are the same as in Example 2.
The obtained slump was used as an actual measurement value (selection data and learning data) of evaluation data.

Regarding the above 15 learning samples, the unit water amount as the data regarding the blending of the concrete, the amount of each mineral as the data regarding the cement, and the mortar flow as the data regarding the physical properties of the cement were used as the actual values of the monitoring data in the learning data. .
The unit water amount is a value adjusted by conducting a test so as to be as small as possible in a range where a required slump (8 ± 1 cm) can be obtained in the concrete manufactured as a learning sample.
The amount of each mineral is the amount of C ₃ S and C ₃ A calculated in the same manner as in Example 2 for the 15 ordinary Portland cements.
The mortar flow is a value immediately after kneading and a value after 30 minutes of kneading, measured in the same manner as in Example 2 for the 15 ordinary Portland cements.

[Select monitoring data]
Of each of the monitoring data (unit water amount, C ₃ S and C ₃ A amount, and mortar flow (immediately, after 30 minutes)) described above, each of the conditions 1 to 4 for selecting the monitoring data shown in Table 4 (Table 4, the selected monitoring data is indicated by “◯”.) Using the actually measured value of the monitoring data and the actually measured value of the evaluation data, an unlearned neural network was learned. The learning was performed 800 times.
Mean square error between the estimated value of the evaluation data obtained by inputting the actual measurement value of the selected monitoring data to the input layer of the neural network after learning ("RMSE" in Table 4) Calculated).
The results are shown in Table 4.
Among conditions 1 to 4, combination of monitoring data selected in condition 1 with the smallest mean square error value (unit water amount, amount of C ₃ S and C ₃ A, mortar flow (immediately, after 30 minutes)) Was used as monitoring data (learning data and monitor data) used for learning of the neural network.

In addition, using two normal Portland cements with different sampling dates as the 15 samples, the concrete slump was measured in the same way as the learning data, and the monitor data (actual value of the evaluation data was measured). ).
Furthermore, the unit water amount of the concrete using the above two ordinary Portland cements, the amount of each mineral of the above two ordinary Portland cements, and the mortar flow using the above two ordinary Portland cements are similar to the learning data. Measurement was made to monitor data (actual value of monitoring data).

The neural network was learned using the learning data. As the neural network, a hierarchical neural network having an input layer, an intermediate layer, and an output layer was used. The neural network is an unlearned one different from the neural network used for selecting the monitoring data.
The neural network was first learned 10,000 times using the learning data and the monitor data. When σ _L and σ _{M were} calculated using the obtained neural network, the relationship between σ _L and σ _M was σ _L <σ _M.
After that, the neural network is initialized, and the learning data and the monitor data are used to learn the neural network by performing the number of learning times (rounded down) by multiplying the number of learning times by 0.95. The relationship between σ _L and σ _M calculated using the network was repeated until σ _L ≧ σ _M. As a result, the analysis level determination value was 4.66%, so the learning was terminated.
Using normal Portland cement B different from the above sample, slump was measured for the concrete manufactured under the same conditions as the concrete used in the learning process. The result was 7.5 cm.
On the other hand, a combination of measured values of 17 monitoring data (a combination of unit water amount, amount of each mineral, and mortar flow) in learning data and monitor data is input to the input layer of the obtained learned neural network. The estimated slump value was output from the output layer.
The estimated value of the obtained slump was 7.3 ± 1.2 cm (the deviation indicates 3σ), and the actually measured value and the predicted value almost coincided.

[Comparative Example 4]
Data on concrete blending, which has a high correlation with concrete slump, with data on 17 normal portland cements used in Example 4 (15 learning samples + 2 monitoring samples) as a population, and on cement Multiple regression using data and data related to the physical properties of cement, with the unit water amount, the amount of each mineral (the amount of C ₃ S and C ₃ A), and the mortar flow (immediately, after 30 minutes) as explanatory variables Analysis was performed to obtain the following regression equation (decision coefficient R ² = 0.47).
(Concrete slump (cm)) = (0.16) × (unit water amount (kg / m ³ )) − (0.15) × (C ₃ S amount (mass%)) + 1.20 × (C ₃ A Amount (% by mass))-0.01 × (mortar flow (immediately after) value (mm)) + 0.02 × (mortar flow (after 30 minutes) value (mm)) + 2.82
The estimated value of slump obtained by substituting the unit water amount, the amount of each mineral, and the mortar flow, which are data on the 17 ordinary Portland cements used in Example 4, into the regression equation is 7.1 ± 1. 0.7 cm (deviation shows 3σ).

[Example 5]
[Prediction of slump flow]
Concrete was produced using the above materials for 15 ordinary Portland cements with different sampling dates as selection and learning samples.
The blending amount of each material was such that the unit cement amount was 330 kg / m ³ , the fine aggregate rate was 45 mass%, and the unit water amount was 165 kg / m ³ . Moreover, the concrete which adjusted the addition amount of the antifoamer so that air amount might be 2% or less was manufactured, and the slump flow of the concrete immediately after kneading | mixing was measured.
The method for producing concrete is the same as in Example 2. The slump flow was measured in accordance with “JIS A 1150 (Concrete slump flow test method)”.
The obtained slump flow was used as an actual measurement value (selection data and learning data) of evaluation data.

In addition, regarding the above 15 learning samples, the amount of high-performance AE water reducing agent as data relating to the blending of concrete, the amount of each mineral (amount of C ₃ S and C ₃ A) as data relating to cement, and the physics of cement The mortar flow (immediately, after 30 minutes) was used as the data relating to the characteristics as the measured value of the monitoring data in the learning data.
The method for obtaining these data is the same as in the third embodiment.

[Select monitoring data]
Of the monitoring data described above (the amount of high-performance AE water reducing agent, the amount of C ₃ S and C ₃ A, the mortar flow (immediately, after 30 minutes)), the monitoring data shown in Table 5 were selected under conditions 1 to 4 For each (the selected monitoring data is indicated by “◯” in Table 5), an unlearned neural network was trained using the measured value of the monitored data and the measured value of the evaluation data. The learning was performed 1500 times.
Mean square error between the estimated value of the evaluation data obtained by inputting the actual measurement value of the selected monitoring data to the input layer of the neural network after learning ("RMSE" in Table 5) Calculated).
The results are shown in Table 5.
Among the conditions 1 to 4, the combination of the monitoring data selected in the condition 1 in which the value of the mean square error was the smallest (amount of high-performance AE water reducing agent, amounts of C ₃ S and C ₃ A, mortar flow (immediately, 30 minutes later)) was used as monitoring data (learning data and monitor data) used for learning of the neural network.

In addition, using two normal Portland cements with different sampling dates as the 15 samples, the concrete slump flow was measured in the same way as the learning data, and the monitor data (actually measured evaluation data) was measured. Value).
Furthermore, learn the amount of high-performance AE water reducing agent of concrete using the above two ordinary Portland cements, the amount of each mineral of the above two ordinary Portland cements, and the mortar flow using the above two ordinary Portland cements. Measurement was performed in the same manner as the data to obtain monitor data (actual value of monitoring data).

The neural network was learned using the learning data. As the neural network, a hierarchical neural network having an input layer, an intermediate layer, and an output layer was used. The neural network is an unlearned one different from the neural network used for selecting the monitoring data.
The neural network was first learned 10,000 times using the learning data and the monitor data. When σ _L and σ _{M were} calculated using the obtained neural network, the relationship between σ _L and σ _M was σ _L <σ _M.
After that, the neural network is initialized, and the learning data and the monitor data are used to learn the neural network by performing the number of learning times (rounded down) by multiplying the number of learning times by 0.95. The relationship between σ _L and σ _M calculated using the network was repeated until σ _L ≧ σ _M. As a result, the analysis level determination value was 0.34%, so learning was terminated.
The slump flow was measured about the concrete manufactured on the same conditions as the concrete used by the said learning process using normal Portland cement B different from the said sample. The result was 58.5 cm.
On the other hand, a learned neural network obtained by combining a combination of measured values of 17 monitoring data (amount of high-performance AE water reducing agent, amount of each mineral, and mortar flow) in learning data and monitor data. The estimated slump flow was output from the output layer.
The estimated value of the obtained slump flow was 58.6 ± 0.6 cm (deviation shows 3σ), and the actually measured value and the predicted value almost coincided.

[Comparative Example 5]
Data on concrete blending, which has a high correlation with the slump flow of concrete, with the data on 17 ordinary portland cements used in Example 5 (15 learning samples + 2 monitoring samples) as the population, cement The amount of high-performance AE water reducing agent, the amount of each mineral (amount of C ₃ S and C ₃ A), and the mortar flow (immediately, after 30 minutes) Multiple regression analysis was performed using explanatory variables, and the following regression equation (decision coefficient R ² = 0.51) was obtained.
(Slump flow of concrete (cm)) = (15.9) × (Amount of high-performance AE water reducing agent (mass%)) − (0.11) × (Amount of C ₃ S (mass%)) − (0. 19) × (C ₃ A amount (mass%)) + (0.05) × (mortar flow (immediately after) value (mm)) − (0.02) × (mortar flow (after 30 minutes) value (mm) ) +33.9
The estimated value of the slump flow obtained by substituting the amount of the high-performance AE water reducing agent, the amount of each mineral, and the mortar flow, which is data on the 17 ordinary Portland cements used in Example 5, into the regression equation. Was 59.1 ± 1.6 cm (deviation shows 3σ).

[Example 6]
[Prediction of slump flow]
Concrete was produced using the above materials for 15 ordinary Portland cements with different sampling dates as selection and learning samples.
The blending amount of each material is such that the mass ratio of water to cement (water / cement) is 0.3, the fine aggregate ratio is 45% by mass, and the amount of high-performance AE water reducing agent is 0.7% with respect to 100% by mass of cement. It was made to become mass%. Moreover, the concrete which adjusted the addition amount of the antifoamer so that air amount might be 2% or less was manufactured, and the slump flow of the concrete immediately after kneading | mixing was measured.
The concrete production and slump flow measurement methods are the same as in Example 5.
The obtained slump flow was used as an actual measurement value (selection data and learning data) of evaluation data.

Further, regarding the above 15 learning samples, the amount of unit water as data relating to the blending of concrete, the amount of each mineral (amount of C ₃ S and C ₃ A) as data relating to cement, and the mortar as data relating to physical properties of cement The flow (immediately, after 30 minutes) was used as the actual measurement value (selection data and learning data) of the monitoring data.
The method for obtaining these data is the same as in the fourth embodiment.

[Select monitoring data]
Among the above-described monitoring data (unit water amount, C ₃ S and C ₃ A amounts, and mortar flow (immediately, after 30 minutes)), each of the conditions 1 to 4 for selecting the monitoring data shown in Table 6 (Table 6, the selected monitoring data is indicated by “◯”.) Using the actually measured value of the monitoring data and the actually measured value of the evaluation data, learning of an unlearned neural network was performed. The learning was performed 900 times.
Mean square error between the estimated value of the evaluation data obtained by inputting the actual measurement value of the selected monitoring data to the input layer of the neural network after learning ("RMSE" in Table 6) Calculated).
The results are shown in Table 6.
Among conditions 1 to 4, combination of monitoring data selected in condition 1 with the smallest mean square error value (unit water amount, amount of C ₃ S and C ₃ A, mortar flow (immediately, after 30 minutes)) Was used as monitoring data (learning data and monitor data) used for learning of the neural network.

In addition, using two normal Portland cements with different sampling dates as the 15 samples, the concrete slump flow was measured in the same way as the learning data, and the monitor data (actually measured evaluation data) was measured. Value).
Furthermore, the unit water amount of the concrete using the above two ordinary Portland cements, the amount of each mineral of the above two ordinary Portland cements, and the mortar flow using the above two ordinary Portland cements are similar to the learning data. Measurement was made to monitor data (actual value of monitoring data).

The neural network was learned using the learning data. As the neural network, a hierarchical neural network having an input layer, an intermediate layer, and an output layer was used. The neural network is an unlearned one different from the neural network used for selecting the monitoring data.
The neural network was first learned 10,000 times using the learning data and the monitor data. When σ _L and σ _{M were} calculated using the obtained neural network, the relationship between σ _L and σ _M was σ _L <σ _M.
After that, the neural network is initialized, and the learning data and the monitor data are used to learn the neural network by performing the number of learning times (rounded down) by multiplying the number of learning times by 0.95. The relationship between σ _L and σ _M calculated using the network was repeated until σ _L ≧ σ _M. As a result, the analysis level determination value was 1.07%, so learning was terminated.
The slump flow was measured about the concrete manufactured on the same conditions as the concrete used by the said learning process using normal Portland cement B different from the said sample. The result was 59.8 cm.
On the other hand, a combination of measured values of 17 monitoring data (a combination of unit water amount, amount of each mineral, and mortar flow) in learning data and monitor data is input to the input layer of the obtained learned neural network. The estimated slump flow was output from the output layer.
The estimated value of the obtained slump flow was 59.6 ± 1.0 cm (the deviation indicates 3σ), and the actually measured value and the predicted value almost coincided.

[Comparative Example 6]
Data on concrete blending, which has a high correlation with the slump flow of concrete, with the data on 17 normal Portland cements used in Example 6 (15 learning samples + 2 monitoring samples) as the population, cement The weight of unit water, the amount of each mineral (amount of C ₃ S and C ₃ A), and the mortar flow (immediately, after 30 minutes) were explanatory variables Regression analysis was performed to obtain the following regression formula (determination coefficient R ² = 0.39).
(Slump flow of concrete (cm)) = (0.03) × (unit water amount (kg / m ³ )) − (0.03) × (C ₃ S amount (mass%)) − 1.55 × (C ₃ A amount (% by mass))-3.1 × 10 ⁻³ × (mortar flow (immediately after) value (mm)) − 4.2 × 10 ⁻⁴ × (mortar flow (after 30 minutes) value (mm)) +18.4
The estimated value of the slump flow obtained by substituting the unit water amount, the amount of each mineral, and the mortar flow, which is data on the 17 ordinary Portland cements used in Example 6, into the regression equation is 59.0 ±. It was 2.1 cm (deviation shows 3σ).

Comparing the estimated value (8.2 ± 0.6 cm) obtained in Example 3 regarding prediction of concrete slump with the estimated value (8.1 ± 1.4 cm) obtained in Comparative Example 3, the example 3 is closer to the actual measurement value (8.5 cm) of the evaluation data than Comparative Example 3, and it is understood that the reliability is high.
Comparing the estimated value (7.3 ± 1.2 cm) obtained in Example 4 with respect to the prediction of the slump of concrete and the estimated value (7.1 ± 1.7 cm) obtained in Comparative Example 4, the Example 4 is closer to the actually measured value (7.5 cm) of the evaluation data than Comparative Example 4, and it can be seen that the reliability is high.
Furthermore, when the estimated value (58.6 ± 0.6 cm) obtained in Example 5 regarding prediction of the slump flow of concrete is compared with the estimated value (59.1 ± 1.6 cm) obtained in Comparative Example 5, It can be seen that Example 5 is closer to the actually measured value (58.5 cm) of the evaluation data and has higher reliability than Comparative Example 5.
When the estimated value (59.6 ± 1.0 cm) obtained in Example 6 regarding prediction of the slump flow of concrete was compared with the estimated value (59.0 ± 2.1 cm) obtained in Comparative Example 6, It can be seen that Example 6 is closer to the actually measured value (59.8 cm) of the evaluation data and has higher reliability than Comparative Example 6.

Claims (7)

A method for predicting concrete quality or concrete mixing conditions using a neural network having an input layer and an output layer,
The input layer is for inputting actual measurement values of monitoring data in concrete production, and the output layer is for outputting an estimated value of evaluation data related to evaluation of concrete quality or concrete mixing conditions. Is,
The combination of the monitoring data and the evaluation data is
(I) The monitoring data is one or more types of data selected from data on cement, data on physical properties of cement, data on materials of concrete other than cement, and data on blending of concrete, and the evaluation A combination where the data is data about the quality of concrete, or
(Ii) The monitoring data is one or more kinds of data selected from data on cement, data on physical properties of cement, data on materials of concrete other than cement, data on blending of concrete, and data on quality of concrete. And the evaluation data is a combination of data relating to the mixing conditions of concrete,
(A) a step of initializing the number of learning times;
(B) using a plurality of learning data that is a combination of the actual measurement value of the monitoring data and the actual measurement value of the evaluation data, and performing learning of the neural network by the number of learning times set in the previous step;
(C) The estimated value of the evaluation data obtained by inputting the actual measurement value of the monitoring data of the learning data to the input layer of the neural network obtained by the learning of the latest step (B) and the actual measurement of the evaluation data of the learning data Of the monitoring data in the monitor data, which is a combination of the measured value of the monitoring data and the evaluation value of the evaluation data for confirming the reliability of the mean square error (σ _L ) with the value and the learning result of the neural network Mean square error between the estimated value of the evaluation data obtained by inputting the actual measurement value to the input layer of the neural network obtained by learning in the most recent step (B) and the actual measurement value of the evaluation data in the monitor data (Σ _M ) is calculated, and when the calculated relationship between σ _L and σ _M is σ _L ≧ σ _M , the step (D) is performed, and the calculated relationship between σ _L and σ _M is σ _L < If σ _M , perform step (E) And a process of
(D) A learning number larger than any of the learning number set in the most recent step (A) and the reset number of learning times of the latest neural network is reset as a new learning number, and the step ( Carrying out B) to (C);
(E) resetting the number of learnings obtained by reducing the number of learnings performed in learning of the latest neural network as a new number of learnings;
(F) using the learning data used in the most recent step (B), learning the neural network for the number of times set in the most recent step (E);
(G) An actual value of the monitoring data of the learning data is input to the input layer of the neural network obtained by learning in the latest step (F), and an estimated value of the evaluation data obtained and the actual measurement of the evaluation data of the learning data Obtained by inputting the mean square error (σ _L ) with the measured value and the measured value of the monitoring data in the monitoring data to the input layer of the neural network obtained by the learning of the most recent step (F). When the mean square error (σ _M ) between the estimated value of the evaluation data and the actually measured value of the evaluation data in the monitor data is calculated, and the relationship between the calculated σ _L and σ _M is σ _L ≧ σ _M , Performing step (I), and if the calculated relationship between σ _L and σ _M is σ _L <σ _M ,
(H) If the number of learning times of the neural network in the latest step (F) exceeds a predetermined numerical value, the steps (E) to (G) are performed again, and the learning of the neural network in the latest step (F) is performed. A step of performing step (J) when the number of times is equal to or less than a predetermined value;
(I) The analytical value determination value is calculated using the following equation (1), and when the analytical value determination value is less than a predetermined first set value, the learning of the neural network is terminated, and the learned neural network The actual value of monitoring data in concrete production is input to the input layer of the network, and the estimated value of the evaluation data related to the evaluation of the quality of the concrete or the mixing condition of the concrete is output from the output layer of the neural network. Predicting the quality of the material or the mixing condition of the concrete, and the analysis degree determination value is equal to or higher than a predetermined first set value, the step of performing the step (J),
(J) A determination is made as to the number of times the step (A) has been performed. If the number is less than or equal to a preset number, the learning conditions are initialized and the steps (A) to (I) are performed again. Performing the step (K) when the number of times exceeds a preset number of times, and
(K) Of all the analytical determination values calculated in step (I), when the smallest analytical determination value is less than a predetermined second set value, the smallest analytical determination value can be obtained. If the neural network in step (I) is a learned neural network, then step (L) is performed, and if the smallest analytical determination value is greater than or equal to a predetermined second set value, the quality of the concrete Or the process of determining that the blending condition of concrete cannot be predicted and terminating the prediction,
(L) The actual measurement value and evaluation data of the monitoring data used as learning data in the step (I) in which the smallest analysis degree determination value can be obtained among all the analysis degree determination values calculated in the step (I). An uncorrelated test was performed on the combination of measured values of, and when there were two or more types of monitoring data judged to be significant at the 5% significance level, it was judged to be significant at the 5% significance level. Plot actual values of monitoring data used as learning data in a coordinate space with all types of monitoring data as coordinate axes, and include all of the monitoring data formed by connecting the plotted monitoring data in the coordinate space After setting an area formed by connecting the monitoring data so that the area is maximized as a predictable monitoring data area, the process (M) is performed, % If monitoring data is determined to be significant at a significance level of 0 or 1 kind, the step of terminating the prediction is determined that it is impossible to predict the blending conditions of the concrete quality or concrete,
(M) It is determined whether or not the actual value of monitoring data in concrete production is included in the predictable monitoring data region, and when the actual value of monitoring data in cement manufacturing is included in the predictable monitoring data region, The actual value of the monitoring data in concrete production is input to the input layer of the learned neural network obtained in (K), and it is related to the evaluation of the concrete quality or the concrete mixing condition from the output layer of the neural network. If the estimated value of the evaluation data is output to predict the concrete quality or the concrete mixing condition, and the actual value of the monitoring data in concrete production is not included in the predictable monitoring data area, the quality of the concrete or the concrete mixing Judge that the condition cannot be predicted A step of Ryosuru,
A method for predicting the quality of concrete or the mixing condition of concrete, characterized by comprising:

(In the above equation (1), the mean square error (σ _L ) of the learning data is an estimation of the evaluation data obtained by inputting the measured value of the monitoring data of the learning data to the input layer of the neural network after learning. The mean square error (σ _L ) between the value and the actual measurement value of the evaluation data of the learning data The average value of the estimated value of the evaluation data is the input of the neural network after learning the actual measurement value of the monitoring data of the learning data (This is the average estimated value of the evaluation data obtained by entering the layer.)   The predetermined first setting value of the analysis degree determination value is 6% or less, and the predetermined second setting value of the analysis degree determination value is larger than the first setting value and 20% or less. The method for predicting concrete quality or concrete mixing conditions according to claim 1.   The method according to claim 1 or 2, wherein the neural network is a hierarchical neural network having an intermediate layer between the input layer and the output layer.   The concrete quality according to any one of claims 1 to 3, wherein a concrete production condition is optimized based on an estimated value of the evaluation data obtained by artificially changing the value of the monitoring data. Prediction method of concrete mixing conditions.   The combination of the monitoring data and the evaluation data is the combination of (i), and the evaluation data in the combination is intensity, slump, or slump flow. To predict the quality of concrete or the mixing conditions of concrete.   The monitoring data in the combination of (i) above are cement brain surface area, sieve test residual amount, color tone a value, color tone L value, cement mineral composition, cement chemical composition, mortar flow, admixture amount, and unit water amount. The method for predicting the quality of concrete or the mixing condition of concrete according to claim 5, wherein the method is one or more selected from the group consisting of: Prior to step (A), (A-1) two or more types of monitoring data aggregates comprising one or more types of arbitrarily selected monitoring data are prepared, and for each of the two or more types of monitoring data aggregates,
By using a plurality of selection data that is a combination of the actual measurement value of the monitoring data and the actual measurement value of the evaluation data, an unlearned neural network different from the neural network used in the steps (A) to (M) is learned and obtained. Calculating the mean square error between the estimated value of the evaluation data obtained by inputting the actual value of the monitoring data of the selected data to the input layer of the neural network and the actual value of the evaluation data of the selected data,
The quality of concrete of any one of Claims 1-6 including the process of using the monitoring data in the selection data in which the numerical value of the mean square error was the smallest as monitoring data in the processes (A) to (M). Or a method for predicting the mixing conditions of concrete.

Images