WO2012086754A1 - Cement quality/manufacturing condition measurement method - Google Patents

Cement quality/manufacturing condition measurement method Download PDF

Info

Publication number
WO2012086754A1
WO2012086754A1 PCT/JP2011/079804 JP2011079804W WO2012086754A1 WO 2012086754 A1 WO2012086754 A1 WO 2012086754A1 JP 2011079804 W JP2011079804 W JP 2011079804W WO 2012086754 A1 WO2012086754 A1 WO 2012086754A1
Authority
WO
WIPO (PCT)
Prior art keywords
data
cement
neural network
quality
evaluation
Prior art date
Application number
PCT/JP2011/079804
Other languages
French (fr)
Japanese (ja)
Inventor
大亮 黒川
麻衣子 大野
宙 平尾
秀幸 菅谷
昌宏 鶴田
優 仲地
Original Assignee
太平洋セメント株式会社
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by 太平洋セメント株式会社 filed Critical 太平洋セメント株式会社
Priority to JP2012524044A priority Critical patent/JPWO2012086754A1/en
Publication of WO2012086754A1 publication Critical patent/WO2012086754A1/en

Links

Images

Classifications

    • C—CHEMISTRY; METALLURGY
    • C04—CEMENTS; CONCRETE; ARTIFICIAL STONE; CERAMICS; REFRACTORIES
    • C04B—LIME, MAGNESIA; SLAG; CEMENTS; COMPOSITIONS THEREOF, e.g. MORTARS, CONCRETE OR LIKE BUILDING MATERIALS; ARTIFICIAL STONE; CERAMICS; REFRACTORIES; TREATMENT OF NATURAL STONE
    • C04B7/00—Hydraulic cements
    • C04B7/36—Manufacture of hydraulic cements in general
    • C04B7/361—Condition or time responsive control in hydraulic cement manufacturing processes
    • F—MECHANICAL ENGINEERING; LIGHTING; HEATING; WEAPONS; BLASTING
    • F27—FURNACES; KILNS; OVENS; RETORTS
    • F27B—FURNACES, KILNS, OVENS, OR RETORTS IN GENERAL; OPEN SINTERING OR LIKE APPARATUS
    • F27B7/00—Rotary-drum furnaces, i.e. horizontal or slightly inclined
    • F27B7/20—Details, accessories, or equipment peculiar to rotary-drum furnaces
    • F27B7/42—Arrangement of controlling, monitoring, alarm or like devices

Definitions

  • the present invention relates to a method for predicting cement quality or manufacturing conditions using a computer.
  • the compressive strength of mortar is obtained by kneading cement, standard sand and water according to JIS R 5201, molding a specimen, curing for 1 day, curing for 3 days, curing for 7 days, and after curing for 28 days. At each time point, the specimen is measured on a compression tester. That is, since it takes 28 days to determine the measurement result of the mortar compressive strength, it is impossible to obtain a predicted value of the mortar compressive strength when the cement is shipped.
  • the amount of industrial waste used as a cement raw material or a burning fuel is increasing, and it is considered that there are more opportunities for the quality of cement to fluctuate. For this reason, in order to prevent abnormalities in the quality of cement to be shipped, the importance of cement quality control is increasing.
  • Patent Document 1 discloses that the powder X-ray analysis result of cement or clinker is analyzed by the profile fitting method, and the quality of the cement (specifically, based on the crystal information of the clinker mineral obtained from this). In particular, a cement quality prediction method is described which predicts changes in cement setting time and mortar compressive strength.
  • Patent Document 2 discloses information on the amount of clinker constituent minerals and additives in cement, information on the crystal structure of the clinker constituent minerals, and information on the minor components of the clinker collected as quality control information during operation of the cement manufacturing plant.
  • Patent Documents 1 and 2 cannot predict quality (for example, mortar fluidity) other than the compressive strength and setting time of mortar.
  • factors affecting the quality of cement are not limited to those listed in Patent Documents 1 and 2, and various factors such as various conditions in the manufacturing process are considered to be complicatedly related.
  • the methods 1 and 2 were not highly accurate methods.
  • there are some cement manufacturing conditions that are difficult to predict due to multiple factors such as the preheater gas flow rate, such as the hydraulic rate of the clinker raw material just before being put into the kiln. . Therefore, there is a need for a method capable of predicting cement quality or production conditions in a short time and with high accuracy in consideration of various factors.
  • An object of the present invention is to provide a method capable of predicting cement quality or production conditions in a short time and with high accuracy.
  • the present invention provides the following [1] to [6].
  • [1] A method for predicting cement quality using a neural network having an input layer and an output layer, wherein actual values of monitoring data in cement production are input to the input layer, and A method for predicting cement quality or production conditions, characterized by outputting an estimated value of evaluation data related to the evaluation of quality or production conditions.
  • [2] The cement quality or manufacturing condition prediction method according to [1], wherein the neural network is a hierarchical neural network having an intermediate layer between the input layer and the output layer.
  • the combination of the monitoring data and the evaluation data is (I)
  • the monitoring data is one or more data selected from data on cement raw materials, data on firing conditions, and data on grinding conditions
  • the evaluation data is data on clinker and data on cement. Or a combination of data on cement raw materials other than the above monitoring data, data on firing conditions, or data on grinding conditions, or (Ii)
  • a combination in which the monitoring data is data relating to cement
  • the evaluation data is data relating to physical properties of the cement-containing hydraulic composition, The method for predicting cement quality or production conditions according to the above [1] or [2].
  • [4] The method for predicting cement quality or production conditions according to any one of [1] to [3], wherein a neural network is optimized by using a plurality of combinations of actual values of monitoring data and actual values of evaluation data .
  • [5] The method according to any one of [1] to [4], wherein the cement production conditions are optimized based on the estimated value of the evaluation data obtained by artificially changing the value of the monitoring data.
  • the [1] to [1] to [1] to [1] are used to periodically check the estimated value of the evaluation data and the magnitude of the difference between the actually measured values corresponding to the estimated value and update the neural network based on the check result.
  • the cement quality or manufacturing condition prediction method of the present invention it is possible to predict cement quality or manufacturing conditions in a short period of time and with high accuracy based on various data obtained in the cement manufacturing process. Further, it is possible to manage the production conditions in real time based on the obtained predicted value, and it is possible to improve the cement quality stabilization or optimize the cement production conditions. Furthermore, the accuracy of prediction can be improved by periodically updating the neural network.
  • a neural network having an input layer for inputting an actual measurement value of monitoring data in cement production and an output layer for outputting an estimated value of evaluation data related to evaluation of cement quality or production conditions. Used to predict cement quality.
  • the neural network may be a hierarchical neural network having an intermediate layer between an input layer and an output layer. Examples of the combination of the monitoring data and the evaluation data include the following (i) and (ii).
  • the monitoring data is one or more data selected from data on cement raw materials, data on firing conditions, and data on grinding conditions, and the evaluation data is data on clinker and data on cement.
  • the monitoring data is data related to cement
  • the evaluation data is water containing cement Combinations that are data on the physical properties of hard compositions
  • Data related to cement raw material which is one of the monitoring data in the combination of (i), includes the chemical composition of the cement raw material, the remaining amount, the specific surface area of the brane (fineness), the ignition loss, from the time of input to the kiln.
  • a clinker main ingredient eg, multiple time points, such as one time point before 5 hours or four time points before 3 hours, 4 hours, 5 hours, and 6 hours
  • the chemical composition of ordinary clinker raw materials such as ordinary Portland cement
  • the supply amount of clinker main raw materials such as ordinary Portland cement
  • the supply amount of clinker secondary materials consisting of special raw materials such as waste
  • the amount of blended silo storage Remaining amount
  • the storage amount of the raw material storage silo replacement amount
  • the current value of the cyclone located between the raw material mill and the blending silo (representing the rotation speed of the cyclone, What there is a correlation between the speed of the material passing through the Ron) and the like.
  • the chemical composition of the cement raw material is 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.
  • the “data relating to the firing conditions” which is one of the monitoring data in the combination of (i) includes the kiln input CFW, the rotational speed, the outlet temperature, the firing zone temperature, the clinker temperature, the kiln average torque, and the O 2 concentration. , NO X concentration, cooler temperature, such as the pre-heater gas flow rate (which is the temperature correlated to the pre-heater) and the like. These data are used singly or in combination of two or more.
  • examples of the “data regarding pulverization conditions”, which is one of the monitoring data in the combination of (i) include pulverization temperature, water spray amount, separator air volume, gypsum addition amount, and the like.
  • data relating to clinker which is one of the evaluation data in the combination of (i), includes clinker mineral composition, ratio of two or more mineral compositions, chemical composition, wet f. CaO (free lime), weight, etc. are mentioned.
  • the mineral composition of the 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.
  • ratio of two or more mineral compositions include a ratio of C 3 S / C 2 S.
  • the mineral composition of the clinker can be obtained, for example, by the Rietveld method.
  • the chemical composition of the clinker is 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 , P in the clinker. 2 O 5 , MnO, Cl, Cr, Zn, Pb, Cu, Ni, V, As, Zr, Mo, Sr, Ba, F, and the like.
  • Examples of the “data on cement” that is one of the evaluation data in the combination of (i) include the brane specific surface area, the residue, the rate of semi-hydration of gypsum and the like.
  • Data related to cement raw materials other than monitoring data which is one of the evaluation data in the combination of (i), includes the chemical composition of the clinker raw material immediately before being put into the kiln, the brain specific surface area of the clinker raw material just before being put into the kiln The residual amount of the clinker raw material immediately before being introduced into the kiln, the decarboxylation rate of the clinker raw material immediately before being introduced into the kiln, the moisture content of the clinker raw material immediately before being introduced into the kiln, and the like.
  • Data related to firing conditions other than monitoring data which is one of the evaluation data in the combination of (i), includes the power value related to the rotation of the kiln, the maximum temperature in the kiln, the kiln outlet temperature, and the kiln outlet oxygen concentration. , Weight of clinker, and the like.
  • Evaluation data other than monitoring data that is one of the evaluation data in the combination of (i) includes the temperature in the mill, the temperature of the powder discharged from the mill, and the amount of the powder discharged from the mill. The amount of powder not discharged from the mill, the fineness of cement, the residual amount of cement, the gypsum hemihydrate rate of cement, and the like.
  • Data related to cement which is monitoring data in the combination of (ii) includes chemical composition, mineral composition, mineral crystallite diameter, mineral crystal lattice constant, wet f. Examples include CaO, ignition loss, brain specific surface area, particle size distribution, residual amount, color tone L value, color tone a value, and color tone b value. These data are used singly or in combination of two or more.
  • the chemical composition of cement refers to 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, TiO 2, P 2 O 5, MnO, is Cl, Cr, Zn, Pb, Cu, Ni, V, as, Zr, Mo, Sr, Ba, the content of such F.
  • 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), f. CaO, f.
  • the “data on the chemical composition and mineral composition of cement” which is evaluation data in the combination (i) may be used.
  • the “physical properties of the cement-containing hydraulic composition”, which is evaluation data in the combination of (ii), includes mortar compressive strength, bending strength, fluidity (flow value), heat of hydration, setting time, drying shrinkage rate, Stability, swelling in water, sulfate resistance, neutralization, ASR resistance and the like can be mentioned.
  • the target cement is not particularly limited.
  • mixed cement such as blast furnace cement and fly ash cement, and cement obtained by adding an admixture such as limestone powder and silica fume to Portland cement.
  • the manufacturing process of Portland cement is roughly divided into three processes: a raw material preparation process, a firing process, and a finishing process.
  • the raw material preparation step is a step of preparing a raw material mixture by preparing cement raw materials such as limestone, clay, silica stone, and iron oxide raw materials at an appropriate ratio and finely pulverizing them with a raw material mill.
  • the firing step is a step of supplying a raw material mixture to a rotary kiln via a suspension preheater or the like, sufficiently firing, and then cooling to obtain a clinker.
  • the finishing step is a step of adding Portland cement by adding an appropriate amount of gypsum and the like to the clinker and pulverizing with a cement mill.
  • the clinker is preferably collected from a place as close as possible to the kiln outlet and where the clinker is sufficiently cooled (usually in the middle of the clinker cooler).
  • the clinker is preferable to collect 1 kg or more of clinker and obtain a representative sample by reduction.
  • For cement it is preferable to sample from the cement mill outlet.
  • the collection interval of the sample for monitoring data is as short as possible. However, if the sampling interval is shortened, labor and the like increase.
  • the collection interval is preferably set to, for example, 15 minutes to 1 hour.
  • the cement production conditions are optimized based on the estimated value (for example, setting time) of the evaluation data obtained by artificially changing the value of the monitoring data (for example, the mineral composition of the cement). be able to.
  • the relationship between the monitoring data and the quality data is previously learned by a neural network, and the quality data is predicted based on only the monitoring data using the learning result.
  • the learning is performed by using a plurality of combinations of the actual measurement values of the monitoring data and the actual measurement values of the evaluation data.
  • the number of the combinations is, for example, 10 or more. Although the upper limit of the number of this combination is not specifically limited, For example, it is 1000.
  • the neural network periodically checks the estimated data and the magnitude of the difference between the actually measured values corresponding to the estimated value, and based on the inspection result, the neural network Is preferably updated.
  • the update cycle is preferably, for example, once per hour, and more preferably, for example, once every 30 minutes.
  • the combination (ii) (the neural network relating to the prediction of the physical properties of the cement-containing hydraulic composition), for example, preferably once a month, more preferably once a week, for example once a day. Further preferred.
  • the method for predicting cement quality or production conditions of the present invention by using a neural network, compression of clinker mineral composition and cement-containing hydraulic composition (for example, mortar) can be performed only by inputting monitoring data. Predicted values such as intensity can be obtained within one hour.
  • cement quality abnormalities can be detected at an early stage during cement production, and by optimizing various conditions in the raw material preparation process, firing process and finishing process. Can be manufactured. Specifically, when an abnormality is observed in the predicted value of the mineral composition of the clinker, the mineral composition of the clinker can be achieved by adjusting the raw material preparation, the firing conditions, and the like. It is also possible to correct the manufacturing target based on the predicted value.
  • Cement quality can be targeted.
  • a computer for controlling cement production and a computer used for carrying out the cement quality or production condition prediction method of the present invention it is possible to artificially vary the monitoring data based on the evaluation data.
  • the control system can also be automated.
  • software for performing an operation using a neural network include “Neural Network Library” (trade name) manufactured by OLSOFT.
  • Example 1 28 cements with different sampling times were sampled and kneaded according to “JIS R 5201”, and the compressive strength of the mortar at each time point after curing for 3 days, after curing for 7 days, and after curing for 28 days was actually measured. It was used as learning data. Using these 28 pieces of learning data, the neural network was learned. In the input layer of the neural network, the brain specific surface area of the sample data, the residual amount of 32 ⁇ m, the wet f.CaO, and the amount of each mineral were input.
  • the neural network learning method was executed until ⁇ L> ⁇ M.
  • the monitor data is data measured separately from the 28 learning data, and is data for confirming the reliability of the learning result of the neural network. The number of monitor data was two.
  • FIGS. 1 to 3 mean actual measurement values of the mortar compressive strength.
  • Estimated values of mortar compressive strength after 3 days curing, 7 days curing and 28 days curing obtained by entering the amounts of CaO and each mineral are 31.5 N / mm 2 , 43.8 N / mm 2 and 58.8 N / mm 2 , and the measured values and the estimated values almost coincided.
  • the amount of each mineral, lattice constant (a, b, c, ⁇ , etc.) or lattice volume is used to capture changes in crystal information of clinker minerals due to small and trace components, and multiple regression analysis
  • the quality of cement was predicted using the multiple regression equation obtained from the above.
  • a preferable example of the multiple regression equation used when predicting the compressive strength of mortar as cement quality is shown below.
  • FIGS. 4 to 6 show graphs of estimated values and teaching values (actual measurement values) of the compression strength on the 3rd, 7th, and 28th.
  • the correlation coefficient (R 2 ) between the estimated value and the teaching value is low and the accuracy of the estimated value is also low.
  • Example 2 Using 20 cements with different sampling times as samples, fluidity test of cement using a high-performance water reducing agent (steel slump cone and stab, stipulated in JIS A1171-2000, 500 mm x 500 mm acrylic plate, We used a spoon and mortar standard sand specified in JIS R5201-1997.) And used it as learning data. The fluidity was measured immediately after kneading and after 30 minutes had elapsed. Using these 20 pieces of learning data, a neural network was learned. The input layer of the neural network includes a brain specific surface area, a residual amount of 32 ⁇ m, a wet f. The amount of CaO and each mineral was input. The amount of each mineral was calculated in the same manner as in Example 1.
  • a high-performance water reducing agent stipulated in JIS A1171-2000, 500 mm x 500 mm acrylic plate, We used a spoon and mortar standard sand specified in JIS R5201-1997.
  • the fluidity immediately after kneading was measured in the same manner as described above. The result was 266 mm.
  • the fluidity immediately after kneading obtained by inputting the brane specific surface area of cement A, the residual amount of 32 ⁇ m, wet f.CaO, and the amount of each mineral into the neural network obtained above is 260 mm, The measured value and the estimated value almost coincided.
  • Example 3 Using 22 cements with different sampling times as samples, the heats of hydration after 7 days and 28 days were actually measured according to “JIS R 5203” and used as learning data. Using these 22 pieces of learning data, neural network learning was performed.
  • the input layer of the neural network includes a brain specific surface area, a residual amount of 32 ⁇ m, a wet f.
  • the amount of CaO and each mineral was input.
  • the amount of each mineral was calculated in the same manner as in Example 1. Input these data, calculate the mean square error ( ⁇ L) of the predicted value and the actual value obtained from the learning data, and calculate the mean square error ( ⁇ M) of the predicted value and the actual value obtained from the monitor data.
  • the neural network learning method was executed until ⁇ L> ⁇ M.
  • the number of monitor data was two.
  • As the neural network a hierarchical neural network having an input layer, an intermediate layer, and an output layer was used. After completion of learning, heat of hydration after 7 days and 28 days was estimated based on 22 learning data. The results are shown in FIGS.
  • Example 4 Using 20 cements with different sampling times as samples, the start and end of setting time were measured according to “JIS R 5201” and used as learning data. Using these 20 pieces of learning data, a neural network was learned. The input layer of the neural network includes a brain specific surface area, a residual amount of 32 ⁇ m, a wet f. The amount of CaO and each mineral was input. The amount of each mineral was calculated in the same manner as in Example 1. Input these data, calculate the mean square error ( ⁇ L) of the predicted value and the actual value obtained from the learning data, and calculate the mean square error ( ⁇ M) of the predicted value and the actual value obtained from the monitor data. The neural network learning method was executed until ⁇ L> ⁇ M. The number of monitor data was two. As the neural network, a hierarchical neural network having an input layer, an intermediate layer, and an output layer was used. After completion of learning, the setting time (starting and closing) was estimated based on 20 learning data. The results are shown in FIGS.
  • Example 5 [B. Data prediction for clinker or cement]
  • the content (%) of free lime (f.CaO) in the clinker was calculated based on the mineral composition and used as learning data.
  • the amount of each mineral was calculated in the same manner as in Example 1.
  • the neural network was learned. In the neural network input layer, the chemical composition of the clinker raw material, the kiln outlet temperature, the kiln firing zone temperature, and the kiln average torque immediately before being introduced into the kiln were input.
  • the content (%) of free lime (f.CaO) in the clinker was measured in the same manner as described above. The result was 0.35%.
  • the clinker obtained by inputting the chemical composition of the clinker raw material just before the clinker A is put into the kiln the kiln drop temperature, the kiln firing zone temperature, and the kiln average torque into the neural network obtained above.
  • the content (%) of the free lime (f.CaO) was 0.42%, and the measured value and the estimated value almost coincided with each other.
  • Example 6 Using 47 cements with different sampling times as samples, the semi-water ratio (%) of gypsum in the cement was calculated based on the mineral composition, and this was used as learning data. The amount of each mineral was calculated in the same manner as in Example 1. Using these 47 pieces of learning data, the neural network was learned. Into the input layer of the neural network, the amount of clinker input, the weight of clinker, the amount of gypsum added, the amount of water spray, the number of rotations of the mill, and the temperature of the powder discharged from the mill were input.
  • the semi-waterification rate (%) of gypsum in the cement was measured in the same manner as described above. The result was 67%. On the other hand, it is obtained by inputting the input amount of clinker, the weight of clinker, the addition amount of gypsum, the amount of water spray, the rotation speed of the mill, and the temperature of the powder discharged from the mill into the neural network obtained above. The semi-waterification rate (%) of gypsum in the cement was 63%, and the measured value and the estimated value almost coincided.
  • Example 7 Using 200 clinker samples with different sampling times as samples, the amount of each mineral was measured. The amount of each mineral was calculated in the same manner as in Example 1. From the results, the C 3 S / C 2 S ratio and the C 4 AF / C 3 A ratio were actually measured and used as learning data. Using these 200 pieces of learning data, a neural network was learned. The chemical composition of the raw material, the kiln firing zone temperature, and the like were input to the input layer of the neural network. The chemical composition of the raw material is SiO 2 , Al 2 O 3 , Fe 2 O 3 , CaO, MgO, SO 3 , Na 2 O, K 2 O, Na 2 Oeq, TiO 2 , P 2 O 5 , MnO.
  • ⁇ L mean square error
  • ⁇ M mean square error
  • the neural network learning method was executed until ⁇ L> ⁇ M.
  • the number of monitor data was three.
  • As the neural network a hierarchical neural network having an intermediate layer was used. After the completion of learning, the hydraulic modulus of the clinker raw material immediately before being put into the kiln was estimated based on 40 pieces of learning data. The result is shown in FIG.

Landscapes

  • Engineering & Computer Science (AREA)
  • Chemical & Material Sciences (AREA)
  • Ceramic Engineering (AREA)
  • Materials Engineering (AREA)
  • Structural Engineering (AREA)
  • Organic Chemistry (AREA)
  • Mechanical Engineering (AREA)
  • General Engineering & Computer Science (AREA)
  • Curing Cements, Concrete, And Artificial Stone (AREA)

Abstract

Provided is a method enabling quick, high accuracy measurement of the quality of cement. Using a neural network having an input layer and an output layer, actual measurement values from observational data during cement manufacture are input to the input layer, and estimate values for evaluation data relating to the evaluation of the quality of the cement are output from the output layer. The combination of the observational data and the evaluation data is: (i) a combination in which the observational data is at least one type of data selected from data relating to the raw materials of the cement, data relating to the firing conditions, and data related to the milling conditions, while the evaluation data is data relating to the clinker and data relating to the cement, or data relating to the raw materials of the cement excluded from the observational data and data relating to the firing conditions, or data relating to the milling conditions; or (ii) a combination in which the observational data is data relating to the cement, while the evaluation data is data relating to the properties of a hydraulic composition containing the cement.

Description

セメントの品質または製造条件の予測方法Methods for predicting cement quality or manufacturing conditions
　本発明は、コンピュータを用いたセメントの品質または製造条件の予測方法に関する。 The present invention relates to a method for predicting cement quality or manufacturing conditions using a computer.
　従来より、セメントの品質を評価するには、コストや時間がかかるという問題があった。例えば、モルタルの圧縮強度は、セメント、標準砂および水をＪＩＳ　Ｒ　５２０１に準じて混練し、供試体を成型し、１日間養生後、３日間養生後、７日間養生後および２８日間養生後の各時点において、供試体を圧縮試験機にかけて測定される。すなわち、モルタルの圧縮強度の測定結果が判明するまでに２８日間かかるため、セメントの出荷時にモルタルの圧縮強度の予測値を得ることができない。
　特に、近年のセメント製造では、セメント原料または焼成用燃料としての産業廃棄物の使用量が増加しており、セメントの品質が変動する機会が多くなっていると考えられる。このため、出荷するセメントの品質の異常を未然に防止するために、セメントの品質管理の重要性が高まっている。 Conventionally, there has been a problem that cost and time are required to evaluate cement quality. For example, the compressive strength of mortar is obtained by kneading cement, standard sand and water according to JIS R 5201, molding a specimen, curing for 1 day, curing for 3 days, curing for 7 days, and after curing for 28 days. At each time point, the specimen is measured on a compression tester. That is, since it takes 28 days to determine the measurement result of the mortar compressive strength, it is impossible to obtain a predicted value of the mortar compressive strength when the cement is shipped.
In particular, in recent cement production, the amount of industrial waste used as a cement raw material or a burning fuel is increasing, and it is considered that there are more opportunities for the quality of cement to fluctuate. For this reason, in order to prevent abnormalities in the quality of cement to be shipped, the importance of cement quality control is increasing.
　このような問題を解決するため、特許文献１には、セメントまたはクリンカーの粉末Ｘ線解析結果を、プロファイルフィッティング法により解析し、これから得られるクリンカー鉱物の結晶情報を基に、セメントの品質（具体的にはセメントの凝結時間及びモルタル圧縮強度）の変化を予測することを特徴とするセメントの品質予測方法が記載されている。
　また、特許文献２には、セメント製造プラントの運転において、品質管理情報として収集した、セメント中のクリンカー構成鉱物及び添加材の量の情報、クリンカー構成鉱物の結晶構造の情報、クリンカーの少量成分の量の情報、およびセメントの粉末度及び４５μｍ残分の情報を、過去に蓄積されているそれらの情報及びモルタル圧縮強さ実測データの間の重回帰分析を基に求めたモルタル圧縮強さの推定式に適用することにより、モルタル圧縮強さを推定することを特徴とするセメントの品質推定方法が記載されている。 In order to solve such a problem, Patent Document 1 discloses that the powder X-ray analysis result of cement or clinker is analyzed by the profile fitting method, and the quality of the cement (specifically, based on the crystal information of the clinker mineral obtained from this). In particular, a cement quality prediction method is described which predicts changes in cement setting time and mortar compressive strength.
Patent Document 2 discloses information on the amount of clinker constituent minerals and additives in cement, information on the crystal structure of the clinker constituent minerals, and information on the minor components of the clinker collected as quality control information during operation of the cement manufacturing plant. Estimate of mortar compressive strength based on multiple regression analysis between information on quantity, cement fineness and 45 μm residual information, and information accumulated in the past and measured data of mortar compressive strength A cement quality estimation method characterized by estimating the mortar compressive strength by applying to the equation is described.
　しかしながら、特許文献１、２の方法では、モルタルの圧縮強度及び凝結時間以外の品質（例えば、モルタルの流動性等）を予測することができない。また、セメントの品質に影響を及ぼす要因は、特許文献１、２に挙げられたものに限られず、製造工程における諸条件等、様々な要因が複雑に関係していると考えられるため、特許文献１、２の方法は精度の高い方法とは言えなかった。
　一方、セメントの製造条件の中には、キルンへの投入直前のクリンカー原料の水硬率のように、プレヒータのガスの流量等の複数の要因が複雑に関係し、予測が困難なものがある。
　そこで、様々な要因を考慮に入れた上で、短時間でかつ高い精度でセメントの品質または製造条件を予測しうる方法が必要となっている。 However, the methods of Patent Documents 1 and 2 cannot predict quality (for example, mortar fluidity) other than the compressive strength and setting time of mortar. In addition, the factors affecting the quality of cement are not limited to those listed in Patent Documents 1 and 2, and various factors such as various conditions in the manufacturing process are considered to be complicatedly related. The methods 1 and 2 were not highly accurate methods.
On the other hand, there are some cement manufacturing conditions that are difficult to predict due to multiple factors such as the preheater gas flow rate, such as the hydraulic rate of the clinker raw material just before being put into the kiln. .
Therefore, there is a need for a method capable of predicting cement quality or production conditions in a short time and with high accuracy in consideration of various factors.
特開２００５－２１４８９１号公報Japanese Patent Laid-Open No. 2005-214891 特開２００７－２７１４４８号公報JP 2007-271448 A
　本発明の目的は、短時間でかつ高い精度でセメントの品質または製造条件を予測することができる方法を提供することにある。 An object of the present invention is to provide a method capable of predicting cement quality or production conditions in a short time and with high accuracy.
　本発明者らは、上記課題を解決するために鋭意検討した結果、ニューラルネットワークを用いれば、上記目的を達成しうることを見出し、本発明を完成した。
　すなわち、本発明は、以下の［１］～［６］を提供するものである。
［１］　入力層及び出力層を有するニューラルネットワークを用いたセメントの品質の予測方法であって、上記入力層に、セメント製造における監視データの実測値を入力して、上記出力層から、セメントの品質または製造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を特徴とするセメントの品質または製造条件の予測方法。
［２］　上記ニューラルネットワークが、上記入力層と上記出力層の間に中間層を有する階層型のニューラルネットワークである前記［１］に記載のセメントの品質または製造条件の予測方法。
［３］　上記監視データと上記評価データの組み合わせが、
（ｉ）上記監視データが、セメント原料に関するデータ、焼成条件に関するデータ、及び、粉砕条件に関するデータの中から選ばれる一種以上のデータであり、かつ、上記評価データが、クリンカーに関するデータ、セメントに関するデータ、または、上記監視データ以外のセメント原料に関するデータ、焼成条件に関するデータ、もしくは粉砕条件に関するデータである組み合わせ、または、
（ｉｉ）上記監視データが、セメントに関するデータであり、かつ、上記評価データが、セメント含有水硬性組成物の物性に関するデータである組み合わせ、
である前記［１］又は［２］に記載のセメントの品質または製造条件の予測方法。
［４］　監視データの実測値と評価データの実測値の組み合わせを複数用いて、ニューラルネットワークを最適化させる前記［１］～［３］のいずれかに記載のセメントの品質または製造条件の予測方法。
［５］　上記監視データの値を人為的に変動させて得られた上記評価データの推測値に基づいて、セメントの製造条件を最適化する前記［１］～［４］のいずれかに記載のセメントの品質または製造条件の予測方法。
［６］　上記評価データの推測値と、該推測値に対応する実測値の乖離の大きさを定期的に点検し、その点検結果に基づいて、上記ニューラルネットワークを更新する前記［１］～［５］のいずれかに記載のセメントの品質または製造条件の予測方法。 As a result of intensive studies to solve the above problems, the present inventors have found that the above object can be achieved by using a neural network, and have completed the present invention.
That is, the present invention provides the following [1] to [6].
[1] A method for predicting cement quality using a neural network having an input layer and an output layer, wherein actual values of monitoring data in cement production are input to the input layer, and A method for predicting cement quality or production conditions, characterized by outputting an estimated value of evaluation data related to the evaluation of quality or production conditions.
[2] The cement quality or manufacturing condition prediction method according to [1], wherein the neural network is a hierarchical neural network having an intermediate layer between the input layer and the output layer.
[3] The combination of the monitoring data and the evaluation data is
(I) The monitoring data is one or more data selected from data on cement raw materials, data on firing conditions, and data on grinding conditions, and the evaluation data is data on clinker and data on cement. Or a combination of data on cement raw materials other than the above monitoring data, data on firing conditions, or data on grinding conditions, or
(Ii) A combination in which the monitoring data is data relating to cement, and the evaluation data is data relating to physical properties of the cement-containing hydraulic composition,
The method for predicting cement quality or production conditions according to the above [1] or [2].
[4] The method for predicting cement quality or production conditions according to any one of [1] to [3], wherein a neural network is optimized by using a plurality of combinations of actual values of monitoring data and actual values of evaluation data .
[5] The method according to any one of [1] to [4], wherein the cement 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 cement quality or manufacturing conditions.
[6] The [1] to [1] to [1] to [1] are used to periodically check the estimated value of the evaluation data and the magnitude of the difference between the actually measured values corresponding to the estimated value and update the neural network based on the check result. [5] The method for predicting cement quality or production conditions according to any one of [5].
　本発明のセメントの品質または製造条件の予測方法を用いれば、セメント製造過程で得られる様々なデータに基づいて、セメントの品質または製造条件を短期間でかつ高い精度で予測することができる。
　また、得られた予測値を基にリアルタイムで製造条件を管理することが可能であり、セメントの品質安定化の向上またはセメントの製造条件の最適化を図ることができる。
　さらに、ニューラルネットワークを定期的に更新することによって、予測の精度の向上を図ることができる。 By using the cement quality or manufacturing condition prediction method of the present invention, it is possible to predict cement quality or manufacturing conditions in a short period of time and with high accuracy based on various data obtained in the cement manufacturing process.
Further, it is possible to manage the production conditions in real time based on the obtained predicted value, and it is possible to improve the cement quality stabilization or optimize the cement production conditions.
Furthermore, the accuracy of prediction can be improved by periodically updating the neural network.
実施例１で予測した材齢３日の圧縮強度の推定値と教示値の比較を示すグラフである。It is a graph which shows the comparison of the estimated value of the compressive strength of the age of 3 days estimated in Example 1, and a teaching value. 実施例１で予測した材齢７日の圧縮強度の推定値と教示値の比較を示すグラフである。It is a graph which shows the comparison of the estimated value of the compressive strength of the age of 7 days estimated in Example 1, and a teaching value. 実施例１で予測した材齢２８日の圧縮強度の推定値と教示値の比較を示すグラフである。It is a graph which shows the comparison of the estimated value of the compressive strength of the age of 28 days predicted in Example 1, and a teaching value. 比較例１で予測した材齢３日の圧縮強度の推定値と教示値の比較を示すグラフである。It is a graph which shows the comparison of the estimated value of the compressive strength of the age of 3 days estimated in the comparative example 1, and a teaching value. 比較例１で予測した材齢７日の圧縮強度の推定値と教示値の比較を示すグラフである。It is a graph which shows the comparison of the estimated value of the compressive strength of the age of 7 days estimated in the comparative example 1, and a teaching value. 比較例１で予測した材齢２８日の圧縮強度の推定値と教示値の比較を示すグラフである。It is a graph which shows the comparison of the estimated value of the compressive strength of 28 days of age predicted in the comparative example 1, and a teaching value. 実施例２で予測した混練直後の流動性の推定値と教示値の比較を示すグラフである。It is a graph which shows the comparison of the estimated value of the fluidity immediately after kneading | mixing estimated in Example 2, and a teaching value. 実施例２で予測した混練３０分後の流動性の推定値と教示値の比較を示すグラフである。It is a graph which shows the comparison of the estimated value of the fluidity | liquidity 30 minutes after kneading | mixing estimated in Example 2, and a teaching value. 実施例３で予測した材齢７日の水和熱の推定値と教示値の比較を示すグラフである。6 is a graph showing a comparison between an estimated value of heat of hydration on the age of 7 days predicted in Example 3 and a teaching value. 実施例３で予測した材齢２８日の水和熱の推定値と教示値の比較を示すグラフである。It is a graph which shows the comparison of the estimated value and teaching value of the heat of hydration of the age of 28 days predicted in Example 3. 実施例４で予測した凝結時間の始発の推定値と教示値の比較を示すグラフである。It is a graph which shows the comparison of the estimated value of the start of the setting time estimated in Example 4, and a teaching value. 実施例４で予測した凝結時間の終結の推定値と教示値の比較を示すグラフである。It is a graph which shows the comparison of the estimated value of completion | finish of the setting time estimated in Example 4, and a teaching value. 実施例５で予測したクリンカー中のフリーライムの含有率（％）の推定値と教示値の比較を示すグラフである。It is a graph which shows the comparison of the estimated value of the content rate (%) of the free lime in a clinker estimated in Example 5, and a teaching value. 実施例６で予測したセメント中の石膏の半水化率（％）の推定値と教示値の比較を示すグラフである。It is a graph which shows the comparison of the estimated value of the semi-waterification rate (%) of the gypsum in the cement estimated in Example 6, and a teaching value. 実施例７で予測したクリンカーの品質（Ｃ３Ｓ／Ｃ２Ｓ比）の推定値と教示値の比較を示すグラフである。It is a graph showing a comparison of the estimated value and the teachings of the quality of the clinker predicted in Example 7 (C 3 S / C 2 S ratio). 実施例７で予測したクリンカーの品質（Ｃ４ＡＦ／Ｃ３Ａ比）の推定値と教示値の比較を示すグラフである。It is a graph showing a comparison of the estimated value and the teachings of the quality of the clinker predicted in Example 7 (C 4 AF / C 3 A ratio). 実施例８で予測したキルンへの投入直前のクリンカー原料の水硬率の推定値と教示値の比較を示すグラフである。It is a graph which shows the comparison of the estimated value of the hydraulic modulus of the clinker raw material just before injection | pouring to the kiln estimated in Example 8, and a teaching value.
　以下、本発明について詳細に説明する。
　本発明においては、セメント製造における監視データの実測値を入力するための入力層と、セメントの品質または製造条件の評価に関連する評価データの推測値を出力するための出力層を有するニューラルネットワークを用いて、セメントの品質を予測する。
　本発明において、ニューラルネットワークは、入力層と出力層の間に中間層を有する階層型のニューラルネットワークであってもよい。
　上記監視データと上記評価データの組み合わせとしては、例えば、以下の（ｉ）、（ｉｉ）等が挙げられる。
（ｉ）上記監視データが、セメント原料に関するデータ、焼成条件に関するデータ、及び、粉砕条件に関するデータの中から選ばれる一種以上のデータであり、かつ、上記評価データが、クリンカーに関するデータ、セメントに関するデータ、または、上記監視データ以外のセメント原料に関するデータ、焼成条件に関するデータ、もしくは粉砕条件に関するデータである組み合わせ
（ｉｉ）上記監視データが、セメントに関するデータであり、かつ、上記評価データが、セメント含有水硬性組成物の物性に関するデータである組み合わせ Hereinafter, the present invention will be described in detail.
In the present invention, there is provided a neural network having an input layer for inputting an actual measurement value of monitoring data in cement production and an output layer for outputting an estimated value of evaluation data related to evaluation of cement quality or production conditions. Used to predict cement quality.
In the present invention, the neural network may be a hierarchical neural network having an intermediate layer between an input layer and an output layer.
Examples of the combination of the monitoring data and the evaluation data include the following (i) and (ii).
(I) The monitoring data is one or more data selected from data on cement raw materials, data on firing conditions, and data on grinding conditions, and the evaluation data is data on clinker and data on cement. Or a combination of data related to cement raw materials other than the monitoring data, data related to firing conditions, or data related to grinding conditions (ii) The monitoring data is data related to cement, and the evaluation data is water containing cement Combinations that are data on the physical properties of hard compositions
　前記（ｉ）の組み合わせにおける監視データの一つである「セメント原料に関するデータ」としては、セメント原料の化学組成、残分量、ブレーン比表面積（粉末度）、強熱減量、キルンへの投入時から所定の時間前の時点（例えば、５時間前の１つの時点や、３時間前、４時間前、５時間前、及び６時間前の４つの時点のような複数の時点）のクリンカー主原料（例えば、普通ポルトランドセメント用クリンカー原料のような通常の調合原料）の化学組成、クリンカー主原料の供給量、廃棄物のような特殊な原料からなるクリンカー副原料の供給量、ブレンディングサイロの貯留量（残量）、原料ストレージサイロの貯留量（残量）、原料ミルとブレンディングサイロの間に位置するサイクロンの電流値（サイクロンの回転数を表し、サイクロンを通過する原料の速度と相関関係があるもの）等が挙げられる。これらのデータは、１種を単独でまたは２種以上を組み合わせて用いられる。
　ここで、セメント原料の化学組成とは、セメント原料中のＳｉＯ２、Ａｌ２Ｏ３、Ｆｅ２Ｏ３、ＣａＯ、ＭｇＯ、ＳＯ３、Ｎａ２Ｏ、Ｋ２Ｏ、Ｎａ２Ｏｅｑ（全アルカリ）、ＴｉＯ２、Ｐ２Ｏ５、ＭｎＯ、Ｃｌ、Ｃｒ、Ｚｎ、Ｐｂ、Ｃｕ、Ｎｉ、Ｖ、Ａｓ、Ｚｒ、Ｍｏ、Ｓｒ、Ｂａ、Ｆ等の含有率である。 “Data related to cement raw material” which is one of the monitoring data in the combination of (i), includes the chemical composition of the cement raw material, the remaining amount, the specific surface area of the brane (fineness), the ignition loss, from the time of input to the kiln. A clinker main ingredient (eg, multiple time points, such as one time point before 5 hours or four time points before 3 hours, 4 hours, 5 hours, and 6 hours) For example, the chemical composition of ordinary clinker raw materials such as ordinary Portland cement), the supply amount of clinker main raw materials, the supply amount of clinker secondary materials consisting of special raw materials such as waste, the amount of blended silo storage ( Remaining amount), the storage amount of the raw material storage silo (remaining amount), the current value of the cyclone located between the raw material mill and the blending silo (representing the rotation speed of the cyclone, What there is a correlation between the speed of the material passing through the Ron) and the like. These data are used singly or in combination of two or more.
Here, the chemical composition of the cement raw material is 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. , TiO 2 , P 2 O 5 , MnO, Cl, Cr, Zn, Pb, Cu, Ni, V, As, Zr, Mo, Sr, Ba, F, and the like.
　前記（ｉ）の組み合わせにおける監視データの一つである「焼成条件に関するデータ」としては、キルンの窯入ＣＦＷ、回転数、落口温度、焼成帯温度、クリンカー温度、キルン平均トルク、Ｏ２濃度、ＮＯＸ濃度、クーラー温度、プレヒーターのガスの流量（プレヒーターの温度と相関関係があるもの）等が挙げられる。これらのデータは、１種を単独でまたは２種以上を組み合わせて用いられる。
　また、（ｉ）の組み合わせにおける監視データの一つである「粉砕条件に関するデータ」としては、粉砕温度、散水量、セパレーター風量、石膏添加量等が挙げられる。これらのデータは、１種を単独でまたは２種以上を組み合わせて用いられる。
　前記（ｉ）の組み合わせにおいて、監視データとして、セメント原料に関するデータ、焼成条件に関するデータ、及び、粉砕条件に関するデータの中から選ばれるいずれか一種のデータのみを用いてもよいが、これら３種のデータのうちの２種以上（複数）のデータを用いることが、評価データの予測の精度を高める観点から、好ましい。
　前記（ｉ）の組み合わせにおける評価データの一つである「クリンカーに関するデータ」としては、クリンカーの鉱物組成、２種以上の鉱物組成の比、化学組成、湿式ｆ．ＣａＯ（フリーライム）、容重等が挙げられる。 The “data relating to the firing conditions” which is one of the monitoring data in the combination of (i) includes the kiln input CFW, the rotational speed, the outlet temperature, the firing zone temperature, the clinker temperature, the kiln average torque, and the O 2 concentration. , NO X concentration, cooler temperature, such as the pre-heater gas flow rate (which is the temperature correlated to the pre-heater) and the like. These data are used singly or in combination of two or more.
Moreover, examples of the “data regarding pulverization conditions”, which is one of the monitoring data in the combination of (i), include pulverization temperature, water spray amount, separator air volume, gypsum addition amount, and the like. These data are used singly or in combination of two or more.
In the combination (i), as the monitoring data, only one kind of data selected from data on cement raw materials, data on firing conditions, and data on grinding conditions may be used. It is preferable to use two or more kinds (two or more) of data from the viewpoint of improving the accuracy of prediction of evaluation data.
“Data relating to clinker”, which is one of the evaluation data in the combination of (i), includes clinker mineral composition, ratio of two or more mineral compositions, chemical composition, wet f. CaO (free lime), weight, etc. are mentioned.
　ここで、クリンカーの鉱物組成とは、３ＣａＯ・ＳｉＯ２（Ｃ３Ｓ）、２ＣａＯ・ＳｉＯ２（Ｃ２Ｓ）、３ＣａＯ・Ａｌ２Ｏ３（Ｃ３Ａ）、４ＣａＯ・Ａｌ２Ｏ３・Ｆｅ２Ｏ３（Ｃ４ＡＦ）、ｆ．ＣａＯ、ｆ．ＭｇＯ等の含有率である。前述の「２種以上の鉱物組成の比」としては、例えば、Ｃ３Ｓ／Ｃ２Ｓの比が挙げられる。
　なお、クリンカーの鉱物組成は、例えばリートベルト法によって得ることができる。
　クリンカーの化学組成とは、クリンカー中のＳｉＯ２、Ａｌ２Ｏ３、Ｆｅ２Ｏ３、ＣａＯ、ＭｇＯ、ＳＯ３、Ｎａ２Ｏ、Ｋ２Ｏ、Ｎａ２Ｏｅｑ（全アルカリ）、ＴｉＯ２、Ｐ２Ｏ５、ＭｎＯ、Ｃｌ、Ｃｒ、Ｚｎ、Ｐｂ、Ｃｕ、Ｎｉ、Ｖ、Ａｓ、Ｚｒ、Ｍｏ、Ｓｒ、Ｂａ、Ｆ等の含有率である。
　（ｉ）の組み合わせにおける評価データの一つである「セメントに関するデータ」としては、ブレーン比表面積、残分、石膏の半水化率等が挙げられる。
　（ｉ）の組み合わせにおける評価データの一つである「監視データ以外のセメント原料に関するデータ」としては、キルンに投入する直前のクリンカー原料の化学組成、キルンに投入する直前のクリンカー原料のブレーン比表面積、キルンに投入する直前のクリンカー原料の残分量、キルンに投入する直前のクリンカー原料の脱炭酸率、キルンに投入する直前のクリンカー原料の水分量、等が挙げられる。
　（ｉ）の組み合わせにおける評価データの一つである「監視データ以外の焼成条件に関するデータ」としては、キルンの回転に関わる電力値、キルン内の最高温度、キルンの出口温度、キルンの出口酸素濃度、クリンカーの容重、等が挙げられる。
　（ｉ）の組み合わせにおける評価データの一つである「監視データ以外の粉砕条件に関するデータ」としては、ミル内の温度、ミルから排出される粉体の温度、ミルから排出される粉体の量、ミルから排出されない粉体の量、セメントの粉末度、セメントの残分量、セメントの石膏半水化率、等が挙げられる。 Here, the mineral composition of the 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. Examples of the above-mentioned “ratio of two or more mineral compositions” include a ratio of C 3 S / C 2 S.
The mineral composition of the clinker can be obtained, for example, by the Rietveld method.
The chemical composition of the clinker is 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 , P in the clinker. 2 O 5 , MnO, Cl, Cr, Zn, Pb, Cu, Ni, V, As, Zr, Mo, Sr, Ba, F, and the like.
Examples of the “data on cement” that is one of the evaluation data in the combination of (i) include the brane specific surface area, the residue, the rate of semi-hydration of gypsum and the like.
“Data related to cement raw materials other than monitoring data”, which is one of the evaluation data in the combination of (i), includes the chemical composition of the clinker raw material immediately before being put into the kiln, the brain specific surface area of the clinker raw material just before being put into the kiln The residual amount of the clinker raw material immediately before being introduced into the kiln, the decarboxylation rate of the clinker raw material immediately before being introduced into the kiln, the moisture content of the clinker raw material immediately before being introduced into the kiln, and the like.
“Data related to firing conditions other than monitoring data”, which is one of the evaluation data in the combination of (i), includes the power value related to the rotation of the kiln, the maximum temperature in the kiln, the kiln outlet temperature, and the kiln outlet oxygen concentration. , Weight of clinker, and the like.
“Evaluation data other than monitoring data” that is one of the evaluation data in the combination of (i) includes the temperature in the mill, the temperature of the powder discharged from the mill, and the amount of the powder discharged from the mill. The amount of powder not discharged from the mill, the fineness of cement, the residual amount of cement, the gypsum hemihydrate rate of cement, and the like.
　前記（ｉｉ）の組み合わせにおける監視データである「セメントに関するデータ」としては、セメントの化学組成、鉱物組成、各鉱物結晶子径、各鉱物結晶格子定数、湿式ｆ．ＣａＯ、強熱減量、ブレーン比表面積、粒度分布、残分量、色調Ｌ値、色調ａ値、色調ｂ値等が挙げられる。これらのデータは、１種を単独でまたは２種以上を組み合わせて用いられる。
　ここで、セメントの化学組成とは、セメント原料中のＳｉＯ２、Ａｌ２Ｏ３、Ｆｅ２Ｏ３、ＣａＯ、ＭｇＯ、ＳＯ３、Ｎａ２Ｏ、Ｋ２Ｏ、Ｎａ２Ｏｅｑ（全アルカリ）、ＴｉＯ２、Ｐ２Ｏ５、ＭｎＯ、Ｃｌ、Ｃｒ、Ｚｎ、Ｐｂ、Ｃｕ、Ｎｉ、Ｖ、Ａｓ、Ｚｒ、Ｍｏ、Ｓｒ、Ｂａ、Ｆ等の含有率である。
　セメントの鉱物組成とは、３ＣａＯ・ＳｉＯ２（Ｃ３Ｓ）、２ＣａＯ・ＳｉＯ２（Ｃ２Ｓ）、３ＣａＯ・Ａｌ２Ｏ３（Ｃ３Ａ）、４ＣａＯ・Ａｌ２Ｏ３・Ｆｅ２Ｏ３（Ｃ４ＡＦ）、ｆ．ＣａＯ、ｆ．ＭｇＯ、石膏、カルサイト等の含有率である。
　なお、セメントの化学組成及び鉱物組成のデータは、前記（ｉ）の組み合わせにおける評価データである「クリンカーに関するデータ」を利用してもよい。
　前記（ｉｉ）の組み合わせにおける評価データである「セメント含有水硬性組成物の物性」としては、モルタルの圧縮強度、曲げ強度、流動性（フロー値）、水和熱、凝結時間、乾燥収縮率、安定性、水中膨張、耐硫酸塩性、中性化、ＡＳＲ抵抗等が挙げられる。 “Data related to cement” which is monitoring data in the combination of (ii) includes chemical composition, mineral composition, mineral crystallite diameter, mineral crystal lattice constant, wet f. Examples include CaO, ignition loss, brain specific surface area, particle size distribution, residual amount, color tone L value, color tone a value, and color tone b value. These data are used singly or in combination of two or more.
Here, the chemical composition of cement refers to 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, TiO 2, P 2 O 5, MnO, is Cl, Cr, Zn, Pb, Cu, Ni, V, as, Zr, Mo, Sr, Ba, the content of such F.
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), f. CaO, f. It is the content of MgO, gypsum, calcite and the like.
In addition, as the data on the chemical composition and mineral composition of cement, “data on clinker” which is evaluation data in the combination (i) may be used.
The “physical properties of the cement-containing hydraulic composition”, which is evaluation data in the combination of (ii), includes mortar compressive strength, bending strength, fluidity (flow value), heat of hydration, setting time, drying shrinkage rate, Stability, swelling in water, sulfate resistance, neutralization, ASR resistance and the like can be mentioned.
　本発明のセメントの品質または製造条件の予測方法において、対象となるセメントとしては、特に限定されず、例えば、普通ポルトランドセメント、早強ポルトランドセメント、中庸熱ポルトランドセメント、低熱ポルトランドセメント等の各種ポルトランドセメントや、高炉セメント、フライアッシュセメント等の混合セメントや、ポルトランドセメントに石灰石粉末やシリカフューム等の混和材を添加したセメント等が挙げられる。
　ポルトランドセメントの製造工程は、原料調製工程、焼成工程、仕上げ工程の３工程に大別される。原料調製工程は、石灰石、粘土、珪石、酸化鉄原料などのセメント原料を適当な割合で調合して、原料ミルで微粉砕し、原料混合物を得る工程である。焼成工程は、原料混合物をサスペンションプレヒーター等を経由してロータリーキルンに供給し、充分に焼成した後、冷却して、クリンカーを得る工程である。仕上げ工程は、クリンカーに適当な量の石膏などを加え、セメントミルで微粉砕して、ポルトランドセメントを得る工程である。 In the method for predicting the quality or production conditions of the cement of the present invention, the target cement is not particularly limited. And mixed cement such as blast furnace cement and fly ash cement, and cement obtained by adding an admixture such as limestone powder and silica fume to Portland cement.
The manufacturing process of Portland cement is roughly divided into three processes: a raw material preparation process, a firing process, and a finishing process. The raw material preparation step is a step of preparing a raw material mixture by preparing cement raw materials such as limestone, clay, silica stone, and iron oxide raw materials at an appropriate ratio and finely pulverizing them with a raw material mill. The firing step is a step of supplying a raw material mixture to a rotary kiln via a suspension preheater or the like, sufficiently firing, and then cooling to obtain a clinker. The finishing step is a step of adding Portland cement by adding an appropriate amount of gypsum and the like to the clinker and pulverizing with a cement mill.
　本発明においては、以下の場所から、監視データ用の試料を採取することが好ましい。
　クリンカーについては、キルン落ち口にできる限り近く、かつ、クリンカーが十分に冷却されている場所（通常はクリンカクーラーの中途）から採取することが好ましい。なお、クリンカーの平均的な品質データを把握するために、１ｋｇ以上のクリンカーを採取し縮分により代表試料を得ることが好ましい。セメントについては、セメントミル出口からサンプリングすることが好ましい。なお、セメントの風化を避けるために、サンプリングからできる限り時間を空けずに分析することが好ましい。
　なお、評価データの変動を把握するためには、監視データ用の試料の採集間隔はできるだけ短い方が好ましい。しかしながら、採取間隔を短くすると、労力等が増大する。したがって、実用的には、採取間隔は、例えば１５分間～１時間とすることが好ましい。
　本発明においては、監視データ（例えば、セメントの鉱物組成）の値を人為的に変動させて得られた評価データの推測値（例えば、凝結時間）に基づいて、セメントの製造条件を最適化することができる。 In the present invention, it is preferable to collect samples for monitoring data from the following places.
The clinker is preferably collected from a place as close as possible to the kiln outlet and where the clinker is sufficiently cooled (usually in the middle of the clinker cooler). In order to grasp average quality data of clinker, it is preferable to collect 1 kg or more of clinker and obtain a representative sample by reduction. For cement, it is preferable to sample from the cement mill outlet. In order to avoid weathering of cement, it is preferable to analyze the sample with as little time as possible from sampling.
In order to grasp the fluctuation of the evaluation data, it is preferable that the collection interval of the sample for monitoring data is as short as possible. However, if the sampling interval is shortened, labor and the like increase. Therefore, practically, the collection interval is preferably set to, for example, 15 minutes to 1 hour.
In the present invention, the cement production conditions are optimized based on the estimated value (for example, setting time) of the evaluation data obtained by artificially changing the value of the monitoring data (for example, the mineral composition of the cement). be able to.
　本発明では、監視データと品質データの関係を、ニューラルネットワークによって予め学習し、その学習結果を用いて、監視データのみに基づいて、品質データを予測する。
　ここでの学習は、監視データの実測値と評価データの実測値の組み合わせを複数用いることによって行われる。該組み合わせの数は、例えば、１０以上である。該組み合わせの数の上限は、特に限定されないが、例えば、１０００である。
　ニューラルネットワークは、より精度の高い予測を行うために、評価データの推測値と、該推測値に対応する実測値の乖離の大きさを定期的に点検し、その点検結果に基づいて、ニューラルネットワークを更新することが好ましい。更新の周期は、前記（ｉ）の組み合わせ（クリンカーの鉱物組成等の予測に関するニューラルネットワーク）については、例えば１時間に一回が好ましく、例えば３０分間に一回がより好ましい。前記（ｉｉ）の組み合わせ（セメント含有水硬性組成物の物性の予測に関するニューラルネットワーク）については、例えば１月に一回が好ましく、例えば１週間に一回がより好ましく、例えば１日に一回がさらに好ましい。 In the present invention, the relationship between the monitoring data and the quality data is previously learned by a neural network, and the quality data is predicted based on only the monitoring data using the learning result.
The learning here is performed by using a plurality of combinations of the actual measurement values of the monitoring data and the actual measurement values of the evaluation data. The number of the combinations is, for example, 10 or more. Although the upper limit of the number of this combination is not specifically limited, For example, it is 1000.
In order to perform prediction with higher accuracy, the neural network periodically checks the estimated data and the magnitude of the difference between the actually measured values corresponding to the estimated value, and based on the inspection result, the neural network Is preferably updated. For the combination of (i) (a neural network relating to prediction of the clinker mineral composition and the like), the update cycle is preferably, for example, once per hour, and more preferably, for example, once every 30 minutes. With regard to the combination (ii) (the neural network relating to the prediction of the physical properties of the cement-containing hydraulic composition), for example, preferably once a month, more preferably once a week, for example once a day. Further preferred.
　本発明のセメントの品質または製造条件の予測方法によれば、ニューラルネットワークを用いることによって、監視データを入力するだけで、クリンカーの鉱物組成や、セメント含有水硬性組成物（例えば、モルタル）の圧縮強度等の予測値を、１時間以内に得ることができる。
　また、得られた予測値に基づいて、セメント製造途中においてセメントの品質異常を早期に察知し、原料調製工程、焼成工程及び仕上げ工程における諸条件の最適化を行うことにより、適正な品質のセメントを製造することができる。
　具体的には、クリンカーの鉱物組成の予測値に異常が認められた場合、原料の調合、焼成条件の調整等を行うことで、クリンカーの鉱物組成を目的のものにすることができる。
　また、予測値に基いて、製造上の目標を修正することも可能である。
　例えば、モルタルの圧縮強度が目標値に達しないと予測される場合、学習に用いた監視データ（因子）とモルタルの圧縮強度の関係を解析して、最適なセメントの処方を確認することで、セメントの品質を目的のものにすることができる。
　さらに、セメント製造を制御するコンピュータと、本発明のセメントの品質または製造条件の予測方法を実施するために用いるコンピュータを接続することによって、評価データに基づいて監視データを人為的に変動させるための制御システムを自動化することもできる。
　本発明において、ニューラルネットワークによる演算を行うためのソフトウェアとしては、例えば、ＯＬＳＯＦＴ社製の「Ｎｅｕｒａｌ　Ｎｅｔｗｏｒｋ　Ｌｉｂｒａｒｙ」（商品名）等が挙げられる。 According to the method for predicting cement quality or production conditions of the present invention, by using a neural network, compression of clinker mineral composition and cement-containing hydraulic composition (for example, mortar) can be performed only by inputting monitoring data. Predicted values such as intensity can be obtained within one hour.
In addition, based on the predicted values obtained, cement quality abnormalities can be detected at an early stage during cement production, and by optimizing various conditions in the raw material preparation process, firing process and finishing process. Can be manufactured.
Specifically, when an abnormality is observed in the predicted value of the mineral composition of the clinker, the mineral composition of the clinker can be achieved by adjusting the raw material preparation, the firing conditions, and the like.
It is also possible to correct the manufacturing target based on the predicted value.
For example, when it is predicted that the compressive strength of the mortar will not reach the target value, by analyzing the relationship between the monitoring data (factor) used for learning and the compressive strength of the mortar, and confirming the optimal cement prescription, Cement quality can be targeted.
Further, by connecting a computer for controlling cement production and a computer used for carrying out the cement quality or production condition prediction method of the present invention, it is possible to artificially vary the monitoring data based on the evaluation data. The control system 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.
　以下、実施例により本発明を説明する。
［Ａ．セメント含有水硬性組成物の物性に関するデータ予測］
［実施例１］
　サンプリング時間の異なる２８個のセメントをサンプルとして、「ＪＩＳ　Ｒ　５２０１」に準じて混練し、３日間養生後、７日間養生後および２８日間養生後の各時点におけるモルタルの圧縮強度を実際に測定し、それを学習データとした。
　これら２８個の学習データを用いて、ニューラルネットワークの学習を行った。ニューラルネットワークの入力層には、サンプルデータのブレーン比表面積、３２μｍ残分量、湿式ｆ.ＣａＯ、及び各鉱物の量を入力した。なお、各鉱物の量は、粉末Ｘ線回折装置にて、測定範囲：２θ＝１０～６５°の範囲で測定を行い、リ－トベルト解析ソフトによって計算されたＣ３Ｓ、Ｃ２Ｓ、Ｃ４ＡＦ、Ｃ３Ａ、石膏類、カルサイトの量である。
　これらのデータを入力し、学習データから得られた予測値と実測値の平均２乗誤差（σＬ）を算出し、モニターデータから得られた予測値と実測値の平均２乗誤差（σＭ）を算出し、σＬ＞σＭとなるまで、ニューラルネットワークの学習方法を実行した。
　ここで、モニターデータとは、前記の２８個の学習データとは別に測定されたデータであり、ニューラルネットワークの学習結果の信頼性を確認するためのデータである。
　モニターデータの数は、２個であった。
　ニューラルネットワークとしては、入力層、中間層及び出力層を有する階層型のニューラルネットワークを用いた。
　学習終了後、２８個の学習データに基づいて、３日間養生後、７日間養生後および２８日間養生後のモルタルの圧縮強度を推定した。その結果を図１～図３に示す。なお、図１～図３中の教示値とは、モルタルの圧縮強度の実測値を意味する。 Hereinafter, the present invention will be described by way of examples.
[A. Data prediction on physical properties of cement-containing hydraulic composition]
[Example 1]
28 cements with different sampling times were sampled and kneaded according to “JIS R 5201”, and the compressive strength of the mortar at each time point after curing for 3 days, after curing for 7 days, and after curing for 28 days was actually measured. It was used as learning data.
Using these 28 pieces of learning data, the neural network was learned. In the input layer of the neural network, the brain specific surface area of the sample data, the residual amount of 32 μm, the wet f.CaO, and the amount of each mineral were input. The amount of each mineral was measured with a powder X-ray diffractometer in the measurement range: 2θ = 10 to 65 °, and C 3 S, C 2 S, C calculated by Rietveld analysis software. 4 Amounts of AF, C 3 A, gypsum and calcite.
Input these data, calculate the mean square error (σL) of the predicted value and the actual value obtained from the learning data, and calculate the mean square error (σM) of the predicted value and the actual value obtained from the monitor data. The neural network learning method was executed until σL> σM.
Here, the monitor data is data measured separately from the 28 learning data, and is data for confirming the reliability of the learning result of the neural network.
The number of monitor data was two.
As the neural network, a hierarchical neural network having an input layer, an intermediate layer, and an output layer was used.
After the completion of learning, the compressive strength of the mortar after 3 days curing, 7 days curing and 28 days curing was estimated based on 28 learning data. The results are shown in FIGS. The teaching values in FIGS. 1 to 3 mean actual measurement values of the mortar compressive strength.
　上記サンプルとは異なるセメントＡを用いて、「ＪＩＳ　Ｒ　５２０１」に準じて混練し、３日間養生後、７日間養生後および２８日間養生後の各時点におけるモルタルの圧縮強度を測定した。その結果、３日間養生後は３２．０N/mm２、７日間養生後は４３．５N/mm２、２８日間養生後は５８．１N/mm２であった。
　一方、上記で得られたニューラルネットワークに、セメントＡのブレーン比表面積、３２μｍ残分量、湿式ｆ．ＣａＯ、及び各鉱物の量を入力して得られた３日間養生後、７日間養生後および２８日間養生後のモルタル圧縮強度の推定値は、それぞれ、３１．５N/mm２、４３．８N/mm２、５８．８N/mm２であり、実測値と推定値はほぼ一致した。 Cement A different from the above sample was used and kneaded according to “JIS R 5201”, and the compressive strength of the mortar at each time point after curing for 3 days, after curing for 7 days, and after curing for 28 days was measured. As a result, after 3 days of curing is 32.0N / mm 2, 7 days after curing is 43.5N / mm 2, 28 days after curing was 58.1N / mm 2.
On the other hand, in the neural network obtained above, the Blaine specific surface area of cement A, the residual amount of 32 μm, wet f. Estimated values of mortar compressive strength after 3 days curing, 7 days curing and 28 days curing obtained by entering the amounts of CaO and each mineral are 31.5 N / mm 2 , 43.8 N / mm 2 and 58.8 N / mm 2 , and the measured values and the estimated values almost coincided.
［比較例１］
　特開２００５－２１４８９１号公報の実施例の方法に従って、モルタルの圧縮強度の試験を行った。
　具体的には、実施例１で用いられたサンプルデータのクリンカーを採取・縮分し、振動ミルによって細かく粉砕して、粉末Ｘ線回折用のサンプルを作製した。
　このサンプルを、粉末Ｘ線回折装置にて、測定範囲：２θ＝１０～６５°の範囲で測定した。
　得られたＸ線回折プロファイルを、リートベルト解析ソフトによって計算し、各クリンカー鉱物の結晶情報のパラメータを得た。
　上記解析によって得られたパラメータのうち、各鉱物の量、格子定数（ａ，ｂ，ｃ，βなど）または格子体積により、少量・微量成分によるクリンカー鉱物の結晶情報の変化を捉え、重回帰分析により求めた重回帰式を用いてセメントの品質の予測を行った。
　セメントの品質として、モルタルの圧縮強度を予測する場合に用いられる上記重回帰式の好ましい一例を以下に示す。 [Comparative Example 1]
A test of the compressive strength of the mortar was conducted according to the method of the example of JP-A-2005-214891.
Specifically, the sample data clinker used in Example 1 was collected and contracted, and finely pulverized by a vibration mill to prepare a sample for powder X-ray diffraction.
This sample was measured with a powder X-ray diffractometer in the measurement range: 2θ = 10 to 65 °.
The obtained X-ray diffraction profile was calculated by Rietveld analysis software, and parameters of crystal information of each clinker mineral were obtained.
Of the parameters obtained by the above analysis, the amount of each mineral, lattice constant (a, b, c, β, etc.) or lattice volume is used to capture changes in crystal information of clinker minerals due to small and trace components, and multiple regression analysis The quality of cement was predicted using the multiple regression equation obtained from the above.
A preferable example of the multiple regression equation used when predicting the compressive strength of mortar as cement quality is shown below.
　モルタル圧縮強度（材齢３日）（Ｎ／ｍｍ２）＝Ａ×（エーライト量；質量％）＋Ｂ×（アルミネート相量；質量％）＋Ｃ×（エーライトの格子体積；Å３）＋Ｄ×（硫酸アルカリ量；質量％）＋Ｅ
　ここで、係数Ａ～Ｅは、Ａ＝０．６、Ｂ＝０．３、Ｃ＝０．６、Ｄ＝２、Ｅ＝６０である。 Mortar compressive strength (age 3 days) (N / mm 2 ) = A × (Alite amount; mass%) + B × (Aluminate phase amount; mass%) + C × (Lite volume of alite; 3 ) + D × (Alkali sulfate amount; mass%) + E
Here, the coefficients A to E are A = 0.6, B = 0.3, C = 0.6, D = 2, and E = 60.
　モルタル圧縮強度（材齢７日）（Ｎ／ｍｍ２）＝材齢３日の予測値＋Ａ×（エーライトの格子体積；Å３）＋Ｂ×（硫酸アルカリ量；質量％）＋Ｃ
　ここで、係数Ａ～Ｃは、Ａ＝－１、Ｂ＝－８０、Ｃ＝３２である。 Mortar compressive strength (age 7 days) (N / mm 2 ) = predicted value at 3 days of age + A × (latite volume of alite; ３ 3 ) + B × (alkaline sulfate amount: mass%) + C
Here, the coefficients A to C are A = −1, B = −80, and C = 32.
　モルタル圧縮強度（材齢２８日）（Ｎ／ｍｍ２）＝材齢７日の予測値＋Ａ×（エーライトの格子体積；Å３）＋Ｂ×（ビーライト量；質量％）＋Ｃ×（ビーライトの格子体積；Å３）＋Ｄ×（硫酸アルカリ量；質量％）＋Ｅ×（フェライト相量；質量％）＋Ｆ×（フェライト相の格子体積；Å３）＋Ｇ
　ここで、係数Ａ～Ｇは、Ａ＝－２、Ｂ＝４、Ｃ＝０．６、Ｄ＝－８０、Ｅ＝－０．２、Ｆ＝－２、Ｇ＝４７である。
　３日、７日、２８日の圧縮強度の推定値と教示値（実測値）のグラフを図４～図６に示す。
　図４～図６に示すように、特開２００５－２１４８９１号の方法では、推定値と教示値の相関係数（Ｒ２）が低く、推定値の精度も低くなると考えられる。 Mortar compressive strength (age 28 days) (N / mm 2 ) = predicted value at age 7 + A × (Lite volume of alite; Å 3 ) + B × (belite amount; mass%) + C × (belite Lattice volume of Å 3 ) + D × (alkaline sulfate amount: mass%) + E × (ferrite phase amount: mass%) + F × (lattice volume of ferrite phase; ３ 3 ) + G
Here, the coefficients A to G are A = −2, B = 4, C = 0.6, D = −80, E = −0.2, F = −2, and G = 47.
FIGS. 4 to 6 show graphs of estimated values and teaching values (actual measurement values) of the compression strength on the 3rd, 7th, and 28th.
As shown in FIGS. 4 to 6, in the method disclosed in Japanese Patent Laid-Open No. 2005-214891, it is considered that the correlation coefficient (R 2 ) between the estimated value and the teaching value is low and the accuracy of the estimated value is also low.
［実施例２］
　サンプリング時間の異なる２０個のセメントをサンプルとして、高性能減水剤を用いたセメントの流動性試験（JIS　A1171-2000で規定されている鋼製のスランプコーンおよび突き棒、500mm×500mmのアクリル平板、JIS　R5201-1997で規定されているさじおよびモルタル標準砂を使用した。）を実施し、それを学習データとした。なお、流動性の測定は、混練直後及び３０分経過後に行った。
　これら２０個の学習データを用いて、ニューラルネットワークの学習を行った。ニューラルネットワークの入力層には、ブレーン比表面積、３２μｍ残分量、湿式ｆ．ＣａＯ、各鉱物の量を入力した。なお、各鉱物の量は、実施例１と同様にして算出した。
　これらのデータを入力し、学習データから得られた予測値と実測値の平均２乗誤差（σＬ）を算出し、モニターデータから得られた予測値と実測値の平均２乗誤差（σＭ）を算出し、σＬ＞σＭとなるまで、ニューラルネットワークの学習方法を実行した。
　モニターデータの数は、２個であった。
　ニューラルネットワークとしては、入力層、中間層及び出力層を有する階層型のニューラルネットワークを用いた。
　学習終了後、２０個の学習データに基づいて、混練直後及び３０分経過後の各時点の流動性を推定した。その結果を図７及び図８に示す。 [Example 2]
Using 20 cements with different sampling times as samples, fluidity test of cement using a high-performance water reducing agent (steel slump cone and stab, stipulated in JIS A1171-2000, 500 mm x 500 mm acrylic plate, We used a spoon and mortar standard sand specified in JIS R5201-1997.) And used it as learning data. The fluidity was measured immediately after kneading and after 30 minutes had elapsed.
Using these 20 pieces of learning data, a neural network was learned. The input layer of the neural network includes a brain specific surface area, a residual amount of 32 μm, a wet f. The amount of CaO and each mineral was input. The amount of each mineral was calculated in the same manner as in Example 1.
Input these data, calculate the mean square error (σL) of the predicted value and the actual value obtained from the learning data, and calculate the mean square error (σM) of the predicted value and the actual value obtained from the monitor data. The neural network learning method was executed until σL> σM.
The number of monitor data was two.
As the neural network, a hierarchical neural network having an input layer, an intermediate layer, and an output layer was used.
After completion of learning, fluidity at each time point was estimated immediately after kneading and after 30 minutes, based on 20 learning data. The results are shown in FIGS.
　上記サンプルとは異なるセメントＡを用いて、上記と同様にして混練直後の流動性を測定した。その結果は２６６ｍｍであった。
　一方、上記で得られたニューラルネットワークに、セメントＡのブレーン比表面積、３２μｍ残分量、湿式ｆ.ＣａＯ、及び各鉱物の量を入力して得られた混練直後の流動性は、２６０ｍｍであり、実測値と推定値はほぼ一致した。 Using cement A different from the above sample, the fluidity immediately after kneading was measured in the same manner as described above. The result was 266 mm.
On the other hand, the fluidity immediately after kneading obtained by inputting the brane specific surface area of cement A, the residual amount of 32 μm, wet f.CaO, and the amount of each mineral into the neural network obtained above is 260 mm, The measured value and the estimated value almost coincided.
［実施例３］
　サンプリング時間の異なる２２個のセメントをサンプルとして、「ＪＩＳ　Ｒ　５２０３」に従って、７日後、および２８日後の水和熱を実際に測定し、それを学習データとした。
　これら２２個の学習データを用いて、ニューラルネットワークの学習を行った。ニューラルネットワークの入力層には、ブレーン比表面積、３２μｍ残分量、湿式ｆ．ＣａＯ、各鉱物の量を入力した。なお、各鉱物の量は、実施例１と同様にして算出した。
　これらのデータを入力し、学習データから得られた予測値と実測値の平均２乗誤差（σＬ）を算出し、モニターデータから得られた予測値と実測値の平均２乗誤差（σＭ）を算出し、σＬ＞σＭとなるまで、ニューラルネットワークの学習方法を実行した。
　モニターデータの数は、２個であった。
　ニューラルネットワークとしては、入力層、中間層及び出力層を有する階層型のニューラルネットワークを用いた。
　学習終了後、２２個の学習データに基づいて、７日後、および２８日後の水和熱を推定した。その結果を図９及び図１０に示す。 [Example 3]
Using 22 cements with different sampling times as samples, the heats of hydration after 7 days and 28 days were actually measured according to “JIS R 5203” and used as learning data.
Using these 22 pieces of learning data, neural network learning was performed. The input layer of the neural network includes a brain specific surface area, a residual amount of 32 μm, a wet f. The amount of CaO and each mineral was input. The amount of each mineral was calculated in the same manner as in Example 1.
Input these data, calculate the mean square error (σL) of the predicted value and the actual value obtained from the learning data, and calculate the mean square error (σM) of the predicted value and the actual value obtained from the monitor data. The neural network learning method was executed until σL> σM.
The number of monitor data was two.
As the neural network, a hierarchical neural network having an input layer, an intermediate layer, and an output layer was used.
After completion of learning, heat of hydration after 7 days and 28 days was estimated based on 22 learning data. The results are shown in FIGS.
［実施例４］
　サンプリング時間の異なる２０個のセメントをサンプルとして、凝結時間の始発・終結を「ＪＩＳ　Ｒ　５２０１」に準じて測定し、それを学習データとした。
　これら２０個の学習データを用いて、ニューラルネットワークの学習を行った。ニューラルネットワークの入力層には、ブレーン比表面積、３２μｍ残分量、湿式ｆ．ＣａＯ、各鉱物の量を入力した。なお、各鉱物の量は、実施例１と同様にして算出した。
　これらのデータを入力し、学習データから得られた予測値と実測値の平均２乗誤差（σＬ）を算出し、モニターデータから得られた予測値と実測値の平均２乗誤差（σＭ）を算出し、σＬ＞σＭとなるまで、ニューラルネットワークの学習方法を実行した。
　モニターデータの数は、２個であった。
　ニューラルネットワークとしては、入力層、中間層及び出力層を有する階層型のニューラルネットワークを用いた。
　学習終了後、２０個の学習データに基づいて、凝結時間（始発及び終結）を推定した。その結果を図１１及び図１２に示す。 [Example 4]
Using 20 cements with different sampling times as samples, the start and end of setting time were measured according to “JIS R 5201” and used as learning data.
Using these 20 pieces of learning data, a neural network was learned. The input layer of the neural network includes a brain specific surface area, a residual amount of 32 μm, a wet f. The amount of CaO and each mineral was input. The amount of each mineral was calculated in the same manner as in Example 1.
Input these data, calculate the mean square error (σL) of the predicted value and the actual value obtained from the learning data, and calculate the mean square error (σM) of the predicted value and the actual value obtained from the monitor data. The neural network learning method was executed until σL> σM.
The number of monitor data was two.
As the neural network, a hierarchical neural network having an input layer, an intermediate layer, and an output layer was used.
After completion of learning, the setting time (starting and closing) was estimated based on 20 learning data. The results are shown in FIGS.
［Ｂ．クリンカーまたはセメントに関するデータ予測］
［実施例５］
　サンプリング時間の異なる１１６個のクリンカーをサンプルとして、その鉱物組成に基いて、クリンカー中のフリーライム（ｆ．ＣａＯ）の含有率（％）を算出し、それを学習データとした。なお、各鉱物の量は、実施例１と同様にして算出した。
　これら１１６個の学習データを用いて、ニューラルネットワークの学習を行った。ニューラルネットワークの入力層には、キルンへの投入直前のクリンカー原料の化学組成、キルン落口温度、キルン焼成帯温度、及び、キルン平均トルクを入力した。
　これらのデータを入力し、学習データから得られた予測値と実測値の平均２乗誤差（σＬ）を算出し、モニターデータから得られた予測値と実測値の平均２乗誤差（σＭ）を算出し、σＬ＞σＭとなるまで、ニューラルネットワークの学習方法を実行した。
　モニターデータの数は、５個であった。
　ニューラルネットワークとしては、入力層、中間層及び出力層を有する階層型のニューラルネットワークを用いた。
　学習終了後、１１６個の学習データに基づいて、クリンカー中のフリーライム（ｆ．ＣａＯ）の含有率（％）を推定した。その結果を図１３に示す。 [B. Data prediction for clinker or cement]
[Example 5]
Using 116 clinker samples with different sampling times as samples, the content (%) of free lime (f.CaO) in the clinker was calculated based on the mineral composition and used as learning data. The amount of each mineral was calculated in the same manner as in Example 1.
Using these 116 pieces of learning data, the neural network was learned. In the neural network input layer, the chemical composition of the clinker raw material, the kiln outlet temperature, the kiln firing zone temperature, and the kiln average torque immediately before being introduced into the kiln were input.
Input these data, calculate the mean square error (σL) of the predicted value and the actual value obtained from the learning data, and calculate the mean square error (σM) of the predicted value and the actual value obtained from the monitor data. The neural network learning method was executed until σL> σM.
The number of monitor data was 5.
As the neural network, a hierarchical neural network having an input layer, an intermediate layer, and an output layer was used.
After completion of learning, the content (%) of free lime (f.CaO) in the clinker was estimated based on 116 learning data. The result is shown in FIG.
　上記サンプルとは異なるクリンカーＡを用いて、上記と同様にクリンカー中のフリーライム（ｆ．ＣａＯ）の含有率（％）を測定した。その結果は０．３５％であった。
　一方、上記で得られたニューラルネットワークに、クリンカーＡのキルンへの投入直前のクリンカー原料の化学組成、キルン落口温度、キルン焼成帯温度、及び、キルン平均トルクを入力して得られたクリンカー中のフリーライム（ｆ．ＣａＯ）の含有率（％）は、０．４２％であり、実測値と推定値はほぼ一致した。 Using a clinker A different from the above sample, the content (%) of free lime (f.CaO) in the clinker was measured in the same manner as described above. The result was 0.35%.
On the other hand, in the clinker obtained by inputting the chemical composition of the clinker raw material just before the clinker A is put into the kiln, the kiln drop temperature, the kiln firing zone temperature, and the kiln average torque into the neural network obtained above. The content (%) of the free lime (f.CaO) was 0.42%, and the measured value and the estimated value almost coincided with each other.
［実施例６］
　サンプリング時間の異なる４７個のセメントをサンプルとして、その鉱物組成に基いて、セメント中の石膏の半水化率（％）を算出し、それを学習データとした。なお、各鉱物の量は、実施例１と同様にして算出した。
　これら４７個の学習データを用いて、ニューラルネットワークの学習を行った。ニューラルネットワークの入力層には、クリンカーの投入量、クリンカーの容重、石膏の添加量、散水量、ミルの回転数、及び、ミルから排出される粉体の温度を入力した。
　これらのデータを入力し、学習データから得られた予測値と実測値の平均２乗誤差（σＬ）を算出し、モニターデータから得られた予測値と実測値の平均２乗誤差（σＭ）を算出し、σＬ＞σＭとなるまで、ニューラルネットワークの学習方法を実行した。
　モニターデータの数は、３個であった。
　ニューラルネットワークとしては、入力層、中間層及び出力層を有する階層型のニューラルネットワークを用いた。
　学習終了後、４７個の学習データに基づいて、セメント中の石膏の半水化率（％）を推定した。その結果を図１４に示す。 [Example 6]
Using 47 cements with different sampling times as samples, the semi-water ratio (%) of gypsum in the cement was calculated based on the mineral composition, and this was used as learning data. The amount of each mineral was calculated in the same manner as in Example 1.
Using these 47 pieces of learning data, the neural network was learned. Into the input layer of the neural network, the amount of clinker input, the weight of clinker, the amount of gypsum added, the amount of water spray, the number of rotations of the mill, and the temperature of the powder discharged from the mill were input.
Input these data, calculate the mean square error (σL) of the predicted value and the actual value obtained from the learning data, and calculate the mean square error (σM) of the predicted value and the actual value obtained from the monitor data. The neural network learning method was executed until σL> σM.
The number of monitor data was three.
As the neural network, a hierarchical neural network having an input layer, an intermediate layer, and an output layer was used.
After the completion of learning, the semi-waterification rate (%) of gypsum in cement was estimated based on 47 pieces of learning data. The result is shown in FIG.
　上記サンプルとは異なるセメントＡを用いて、上記と同様にセメント中の石膏の半水化率（％）を測定した。その結果は６７％であった。
　一方、上記で得られたニューラルネットワークに、クリンカーの投入量、クリンカーの容重、石膏の添加量、散水量、ミルの回転数、及び、ミルから排出される粉体の温度を入力して得られたセメント中の石膏の半水化率（％）は、６３％であり、実測値と推定値はほぼ一致した。 Using cement A different from the above sample, the semi-waterification rate (%) of gypsum in the cement was measured in the same manner as described above. The result was 67%.
On the other hand, it is obtained by inputting the input amount of clinker, the weight of clinker, the addition amount of gypsum, the amount of water spray, the rotation speed of the mill, and the temperature of the powder discharged from the mill into the neural network obtained above. The semi-waterification rate (%) of gypsum in the cement was 63%, and the measured value and the estimated value almost coincided.
［実施例７］
　サンプリング時間の異なる２００個のクリンカーをサンプルとして、各鉱物の量を測定した。なお、各鉱物の量は、実施例１と同様にして算出した。その結果より、Ｃ３Ｓ／Ｃ２Ｓ比、及び、Ｃ４ＡＦ／Ｃ３Ａ比を実際に測定し、それを学習データとした。
　これら２００個の学習データを用いて、ニューラルネットワークの学習を行った。ニューラルネットワークの入力層には、原料の化学組成、およびキルンの焼成帯温度等を入力した。なお、原料の化学組成とは、ＳｉＯ２、Ａｌ２Ｏ３、Ｆｅ２Ｏ３、ＣａＯ、ＭｇＯ、ＳＯ３、Ｎａ２Ｏ、Ｋ２Ｏ、Ｎａ２Ｏｅｑ、ＴｉＯ２、Ｐ２Ｏ５、ＭｎＯ、Ｃｌ、Ｔ－Ｃｒ、Ｚｎ、Ｐｂ、Ｃｕ、Ｎｉ、Ｖ、Ａｓ、Ｚｒ、Ｍｏ、Ｓｒ、Ｂａ、Ｆ等の含有量である。
　これらのデータを入力し、学習データから得られた予測値と実測値の平均２乗誤差（σＬ）を算出し、モニターデータから得られた予測値と実測値の平均２乗誤差（σＭ）を算出し、σＬ＞σＭとなるまで、ニューラルネットワークの学習方法を実行した。
　モニターデータの数は、５個であった。
　学習終了後、２００個の学習データに基づいて、Ｃ３Ｓ／Ｃ２Ｓ比、及び、Ｃ４ＡＦ／Ｃ３Ａ比を推定した。その結果を図１５（Ｃ３Ｓ／Ｃ２Ｓ比）及び図１６（Ｃ４ＡＦ／Ｃ３Ａ比）に示す。 [Example 7]
Using 200 clinker samples with different sampling times as samples, the amount of each mineral was measured. The amount of each mineral was calculated in the same manner as in Example 1. From the results, the C 3 S / C 2 S ratio and the C 4 AF / C 3 A ratio were actually measured and used as learning data.
Using these 200 pieces of learning data, a neural network was learned. The chemical composition of the raw material, the kiln firing zone temperature, and the like were input to the input layer of the neural network. The chemical composition of the raw material is SiO 2 , Al 2 O 3 , Fe 2 O 3 , CaO, MgO, SO 3 , Na 2 O, K 2 O, Na 2 Oeq, TiO 2 , P 2 O 5 , MnO. , Cl, T—Cr, Zn, Pb, Cu, Ni, V, As, Zr, Mo, Sr, Ba, F, and the like.
Input these data, calculate the mean square error (σL) of the predicted value and the actual value obtained from the learning data, and calculate the mean square error (σM) of the predicted value and the actual value obtained from the monitor data. The neural network learning method was executed until σL> σM.
The number of monitor data was 5.
After the learning, the C 3 S / C 2 S ratio and the C 4 AF / C 3 A ratio were estimated based on 200 pieces of learning data. The results are shown in FIG. 15 (C 3 S / C 2 S ratio) and FIG. 16 (C 4 AF / C 3 A ratio).
［Ｃ．監視データ以外のセメント原料に関するデータ予測］
［実施例８］
　キルンへの投入直前の、サンプリング時間の異なる４０個のクリンカー原料をサンプルとして、その鉱物組成に基いて水硬率を算出し、それを学習データとした。なお、各鉱物の量は、実施例１と同様にして算出した。
　これら４０個の学習データを用いて、ニューラルネットワークの学習を行った。ニューラルネットワークの入力層には、クリンカー原料のサンプリング時から３時間前、４時間前、５時間前、及び６時間前の各時点における原料ミル内のクリンカー主原料（調合原料）の水硬率、クリンカー主原料の供給量、３種のクリンカー副原料（廃棄物）の各供給量、ブレンディングサイロの残量、原料ストレージサイロの残量、原料ミルとブレンディングサイロの間に位置するサイクロンの電流値、及び、プレヒーターのガスの流量を入力した。
　これらのデータを入力し、学習データから得られた予測値と実測値の平均２乗誤差（σＬ）を算出し、モニターデータから得られた予測値と実測値の平均２乗誤差（σＭ）を算出し、σＬ＞σＭとなるまで、ニューラルネットワークの学習方法を実行した。
　モニターデータの数は、３個であった。
　ニューラルネットワークとしては、中間層を有する階層型のニューラルネットワークを用いた。
　学習終了後、４０個の学習データに基づいて、キルンへの投入直前のクリンカー原料の水硬率を推定した。その結果を図１７に示す。 [C. Data prediction for cement raw materials other than monitoring data]
[Example 8]
Using 40 clinker raw materials with different sampling times just before being put into the kiln as samples, hydraulic modulus was calculated based on the mineral composition and used as learning data. The amount of each mineral was calculated in the same manner as in Example 1.
Using these 40 pieces of learning data, a neural network was learned. In the input layer of the neural network, the hydraulic rate of the clinker main raw material (prepared raw material) in the raw material mill at each time point 3 hours, 4 hours, 5 hours, and 6 hours before sampling of the clinker raw material, Supply amount of clinker main raw material, supply amount of each of three kinds of clinker secondary raw materials (waste), remaining amount of blending silo, remaining amount of raw material storage silo, current value of cyclone located between raw material mill and blending silo, And the flow rate of the preheater gas was input.
Input these data, calculate the mean square error (σL) of the predicted value and the actual value obtained from the learning data, and calculate the mean square error (σM) of the predicted value and the actual value obtained from the monitor data. The neural network learning method was executed until σL> σM.
The number of monitor data was three.
As the neural network, a hierarchical neural network having an intermediate layer was used.
After the completion of learning, the hydraulic modulus of the clinker raw material immediately before being put into the kiln was estimated based on 40 pieces of learning data. The result is shown in FIG.

Claims (6)

  1. 　入力層及び出力層を有するニューラルネットワークを用いたセメントの品質の予測方法であって、上記入力層に、セメント製造における監視データの実測値を入力して、上記出力層から、セメントの品質または製造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を特徴とするセメントの品質または製造条件の予測方法。 A method for predicting cement quality using a neural network having an input layer and an output layer, wherein an actual value of monitoring data in cement production is input to the input layer, and the quality or manufacture of cement is obtained from the output layer. A method for predicting cement quality or manufacturing conditions, characterized by outputting an estimated value of evaluation data related to condition evaluation.
  2. 　上記ニューラルネットワークが、上記入力層と上記出力層の間に中間層を有する階層型のニューラルネットワークである請求項１に記載のセメントの品質または製造条件の予測方法。 The method for predicting cement quality or manufacturing conditions according to claim 1, wherein the neural network is a hierarchical neural network having an intermediate layer between the input layer and the output layer.
  3. 　上記監視データと上記評価データの組み合わせが、
    （ｉ）上記監視データが、セメント原料に関するデータ、焼成条件に関するデータ、及び、粉砕条件に関するデータの中から選ばれる一種以上のデータであり、かつ、上記評価データが、クリンカーに関するデータ、セメントに関するデータ、または、上記監視データ以外のセメント原料に関するデータ、焼成条件に関するデータ、もしくは粉砕条件に関するデータである組み合わせ、または、
    （ｉｉ）上記監視データが、セメントに関するデータであり、かつ、上記評価データが、セメント含有水硬性組成物の物性に関するデータである組み合わせ、
    である請求項１又は２に記載のセメントの品質または製造条件の予測方法。     The combination of the monitoring data and the evaluation data is
    (I) The monitoring data is one or more data selected from data on cement raw materials, data on firing conditions, and data on grinding conditions, and the evaluation data is data on clinker and data on cement. Or a combination of data on cement raw materials other than the above monitoring data, data on firing conditions, or data on grinding conditions, or
    (Ii) A combination in which the monitoring data is data relating to cement, and the evaluation data is data relating to physical properties of the cement-containing hydraulic composition,
    The method for predicting cement quality or production conditions according to claim 1 or 2.
  4. 　監視データの実測値と評価データの実測値の組み合わせを複数用いて、ニューラルネットワークを最適化させる請求項１～３のいずれか１項に記載のセメントの品質または製造条件の予測方法。 The method for predicting cement quality or manufacturing conditions according to any one of claims 1 to 3, wherein the neural network is optimized by using a plurality of combinations of actual measurement values of monitoring data and actual measurement values of evaluation data.
  5. 　上記監視データの値を人為的に変動させて得られた上記評価データの推測値に基づいて、セメントの製造条件を最適化する請求項１～４のいずれか１項に記載のセメントの品質または製造条件の予測方法。 The cement quality according to any one of claims 1 to 4, wherein a cement production condition is optimized based on an estimated value of the evaluation data obtained by artificially changing the value of the monitoring data. Manufacturing method prediction method.
  6. 　上記評価データの推測値と、該推測値に対応する実測値の乖離の大きさを定期的に点検し、その点検結果に基づいて、上記ニューラルネットワークを更新する請求項１～５のいずれか１項に記載のセメントの品質または製造条件の予測方法。 6. The system according to claim 1, wherein the estimated value of the evaluation data and the magnitude of the difference between the actually measured values corresponding to the estimated value are periodically inspected, and the neural network is updated based on the inspection result. The method for predicting the cement quality or production conditions according to the item.
PCT/JP2011/079804 2010-12-24 2011-12-22 Cement quality/manufacturing condition measurement method WO2012086754A1 (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
JP2012524044A JPWO2012086754A1 (en) 2010-12-24 2011-12-22 Methods for predicting cement quality or manufacturing conditions

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
JP2010-287822 2010-12-24
JP2010287822 2010-12-24

Publications (1)

Publication Number Publication Date
WO2012086754A1 true WO2012086754A1 (en) 2012-06-28

Family

ID=46314021

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/JP2011/079804 WO2012086754A1 (en) 2010-12-24 2011-12-22 Cement quality/manufacturing condition measurement method

Country Status (2)

Country Link
JP (1) JPWO2012086754A1 (en)
WO (1) WO2012086754A1 (en)

Cited By (9)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JP2015124129A (en) * 2013-12-27 2015-07-06 三菱マテリアル株式会社 Clinker production method whose free lime amount controlled based on quartz crystallite diameter
JP2017066026A (en) * 2015-09-29 2017-04-06 太平洋セメント株式会社 Method for predicting quality or manufacturing condition of cement
JP6208403B1 (en) * 2016-09-30 2017-10-04 太平洋セメント株式会社 Methods for predicting cement quality or manufacturing conditions
JP2017178648A (en) * 2016-03-29 2017-10-05 太平洋セメント株式会社 Method for prospecting quality of cement or product condition
JP2017178651A (en) * 2016-03-29 2017-10-05 太平洋セメント株式会社 Method of predicting manufacturing conditions of cement clinker
WO2018061239A1 (en) * 2016-09-30 2018-04-05 太平洋セメント株式会社 Prediction method for cement quality or manufacturing conditions
JP2019156699A (en) * 2018-03-16 2019-09-19 太平洋セメント株式会社 Manufacturing method of cement
JP2020144132A (en) * 2018-09-26 2020-09-10 太平洋セメント株式会社 Method of predicting quality of ready mixed concrete
ES2908114A1 (en) * 2020-10-26 2022-04-27 Univ Sevilla Tool to manage efficient incorporation Alternative fuels in cement plants (Machine-translation by Google Translate, not legally binding)

Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JPH01282672A (en) * 1988-05-10 1989-11-14 Nippon Telegr & Teleph Corp <Ntt> Neural network device
JP2005214891A (en) * 2004-01-30 2005-08-11 Mitsubishi Materials Corp Method for predicting cement quality and manufacturing management system for cement
JP2007271448A (en) * 2006-03-31 2007-10-18 Sumitomo Osaka Cement Co Ltd Quality estimation method of cement, and quality control method using same

Family Cites Families (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JPH05250164A (en) * 1992-03-04 1993-09-28 Hitachi Ltd Learning method

Patent Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JPH01282672A (en) * 1988-05-10 1989-11-14 Nippon Telegr & Teleph Corp <Ntt> Neural network device
JP2005214891A (en) * 2004-01-30 2005-08-11 Mitsubishi Materials Corp Method for predicting cement quality and manufacturing management system for cement
JP2007271448A (en) * 2006-03-31 2007-10-18 Sumitomo Osaka Cement Co Ltd Quality estimation method of cement, and quality control method using same

Non-Patent Citations (1)

* Cited by examiner, † Cited by third party
Title
YOSHIO KASAI: "Concrete Soran, 1st edition", GIJUTSU SHOIN, 10 June 1998 (1998-06-10), pages 10 *

Cited By (11)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JP2015124129A (en) * 2013-12-27 2015-07-06 三菱マテリアル株式会社 Clinker production method whose free lime amount controlled based on quartz crystallite diameter
JP2017066026A (en) * 2015-09-29 2017-04-06 太平洋セメント株式会社 Method for predicting quality or manufacturing condition of cement
JP2017178648A (en) * 2016-03-29 2017-10-05 太平洋セメント株式会社 Method for prospecting quality of cement or product condition
JP2017178651A (en) * 2016-03-29 2017-10-05 太平洋セメント株式会社 Method of predicting manufacturing conditions of cement clinker
JP6208403B1 (en) * 2016-09-30 2017-10-04 太平洋セメント株式会社 Methods for predicting cement quality or manufacturing conditions
WO2018061239A1 (en) * 2016-09-30 2018-04-05 太平洋セメント株式会社 Prediction method for cement quality or manufacturing conditions
JP2019156699A (en) * 2018-03-16 2019-09-19 太平洋セメント株式会社 Manufacturing method of cement
JP2020144132A (en) * 2018-09-26 2020-09-10 太平洋セメント株式会社 Method of predicting quality of ready mixed concrete
JP2020144133A (en) * 2018-09-26 2020-09-10 太平洋セメント株式会社 Method of predicting quality of ready mixed concrete
JP2020144099A (en) * 2018-09-26 2020-09-10 太平洋セメント株式会社 Method of predicting quality of ready mixed concrete
ES2908114A1 (en) * 2020-10-26 2022-04-27 Univ Sevilla Tool to manage efficient incorporation Alternative fuels in cement plants (Machine-translation by Google Translate, not legally binding)

Also Published As

Publication number Publication date
JPWO2012086754A1 (en) 2014-06-05

Similar Documents

Publication Publication Date Title
WO2013179702A1 (en) Method for predicting quality or manufacturing condition of cement
WO2012086754A1 (en) Cement quality/manufacturing condition measurement method
JP6636358B2 (en) How to predict the quality or manufacturing conditions of fly ash cement
JP6794039B2 (en) Prediction method of concrete compounding conditions
JP6674356B2 (en) Prediction method of concrete quality or concrete mixing condition
Kirgiz Use of ultrafine marble and brick particles as raw materials in cement manufacturing
Torres et al. Assessing the pozzolanic activity of sugarcane bagasse ash using X-ray diffraction
JP2007271448A (en) Quality estimation method of cement, and quality control method using same
JP4336591B2 (en) Cement quality prediction method and cement production management system
JP6732380B2 (en) Method of predicting cement quality or manufacturing conditions
JP2011209022A (en) Method of manufacturing mortar or concrete and method of estimating final adiabatic temperature rise amount in mortar or concrete
Cardoso et al. Evaluation of the use of blast furnace slag as an additive in mortars
JP5323290B1 (en) Methods for predicting cement quality or manufacturing conditions
JP6639987B2 (en) How to predict cement quality or manufacturing conditions
JP6364918B2 (en) Blast furnace slag sorting method and blast furnace cement production method
JP2007076931A (en) Cement quality control method, and method of producing cement
JP2011209024A (en) Method of manufacturing mortar or concrete, and method of estimating adiabatic temperature rising rate in mortar or concrete
WO2018061239A1 (en) Prediction method for cement quality or manufacturing conditions
JP6208403B1 (en) Methods for predicting cement quality or manufacturing conditions
CN112805259A (en) Cement composition and method for producing cement composition
JP2008156231A (en) Cement composition
Paul Application of the Rietveld method in the cement industry
JP2023056609A (en) Cement quality prediction method
JP2018168021A (en) Method for predicting quality or manufacturing condition of cement
Almeida et al. Composition of renders and plasters of award-winning buildings in Lisbon (Portugal): A contribution to the knowledge of binders used in the 20th Century

Legal Events

Date Code Title Description
WWE Wipo information: entry into national phase

Ref document number: 2012524044

Country of ref document: JP

121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 11852194

Country of ref document: EP

Kind code of ref document: A1

NENP Non-entry into the national phase

Ref country code: DE

122 Ep: pct application non-entry in european phase

Ref document number: 11852194

Country of ref document: EP

Kind code of ref document: A1