Method and apparatus for selecting a flocculant for dewatering treatment

Abstract

To provide a method that can properly select a coagulant to be added to a suspension such as sludge or slurry.SOLUTION: A method includes the steps of: inputting suspension property into a learned model which is constituted with machine learning; outputting a recommended value of physical properties of coagulants from the learned model; and selecting one coagulant to be added to the suspension of the coagulants from a database on the basis of the recommended value of physical properties.SELECTED DRAWING: Figure 1

Description

[Technical field]

[0001] The present invention relates to a technique for selecting a flocculant to be used in a dewatering process, and in particular to a method and apparatus for predicting the physical property values ​​of a flocculant suitable for a suspension, such as sludge or slurry, using an AI model in an operation for separating the suspension into turbidity and liquid. [Background technology]

[0002] Conventionally, dehydrators have been used to squeeze suspensions (e.g., sludge) discharged from liquid treatment facilities such as sewage treatment plants, sewage treatment plants, and industrial wastewater treatment plants to separate water from the suspension. As a pretreatment for dehydration, a flocculant is added to the suspension in a flocculation tank, and the mixture of the flocculant and the suspension is stirred to form flocs and make the suspension in a state that is easy to separate solids and liquids. Thereafter, the suspension with the flocs formed is sent to a dehydrator, which dehydrates the suspension. A solid material with a low moisture content, called a cake or dehydrated cake, is discharged from the dehydrator.

[0003] In the dehydration equipment at water treatment sites, the optimal flocculant is uniquely determined for the suspension to be treated. However, since the properties of the suspension to be treated are constantly changing, the properties of the optimal flocculant for the suspension also change. However, it is difficult to frequently measure the suspension properties on-site and select and apply the optimal flocculant, so the current situation is that it is not always possible to meet the demand for dehydration using the optimal flocculant for the suspension.

[0004] Recently, technology has been developed to select the optimal flocculant using artificial intelligence (AI). For example, the optimal flocculant is selected using a model constructed by machine learning. Since the selection of flocculants using such AI does not depend on the skills or experience of the workers, relatively stable selection results are expected. [Prior art documents] [Patent documents]

[0005] [Patent Document 1] Patent Publication No. 2021-35669 [Patent Document 2] JP 2020-065964 A [Patent Document 3] JP 2004-177122 A [Patent Document 4] JP 2011-156529 A [Patent Document 5] Japanese Patent Application Publication No. 11-169899 Summary of the Invention [Problem to be solved by the invention]

[0006] In the technology of Patent Document 1, an image of a floc for inspection is input into a trained model, and an appropriate flocculant is selected from among multiple flocculants used in the training floc image. However, this technology can only select the flocculant used in training. In order to be able to select a flocculant that was not used in training, it is necessary to add that flocculant to the training data and run the machine learning again. Furthermore, in the technology described in Patent Document 1, image data of the floc is required as training data in constructing the machine learning model.

[0007] The technology in Patent Document 2 predicts the amount of flocculant to be added using a machine learning model, but because the physical properties of the flocculant are fixed with this technology, it is not possible to apply a flocculant with appropriate properties determined by the physical properties of the sludge at the site.

[0008] In the technology of Patent Document 3, a flocculant is selected using the absorption spectrum or emission spectrum of the measured water as control information, but in order to select a flocculant, analysis must be performed based on various information other than the spectrum.

[0009] The technology in Patent Document 4 shows that pH, conductivity, and colloidal charge are correlated with the amount of flocculant injected. However, it does not mention technology for predicting the physical properties of flocculants (viscosity, cationicity, etc.) to select the flocculant, or technology for using other physical properties such as TS in the prediction to improve prediction accuracy.

[0010] In view of the above-mentioned problems, an object of the present invention is to provide a method and an apparatus capable of appropriately selecting a flocculant to be added to a suspension such as sludge or slurry. [Means for solving the problem]

[0011] In one aspect, a method for selecting a flocculant to be added to a suspension is provided, the method comprising: inputting properties of the suspension into a trained model constructed by machine learning; outputting recommended values ​​of physical properties of the flocculant from the trained model; and selecting a flocculant to be added to the suspension from a plurality of flocculants in a database based on the recommended value of the physical properties.

[0012] In one embodiment, the database includes identification names of a plurality of flocculants and data on a plurality of physical properties of each of the plurality of flocculants. In one embodiment, the physical properties of the flocculant are the viscosity and ionic index of the flocculant, and the ionic index is the cationic degree, anionic degree, or colloidal equivalent value of the flocculant. In one aspect, selecting a flocculant to be added to the suspension from among a plurality of flocculants in the database based on the recommended value of the physical property is characterized in that a flocculant having physical properties that are the same as or close to the recommended value of the physical property is selected from among a plurality of flocculants in the database. In one aspect, selecting a flocculant having physical properties that are the same as or close to the recommended values ​​of the physical properties from among the multiple flocculants in the database is characterized by plotting a recommended point identified from the recommended values ​​of the physical properties on a coordinate system having viscosity and ionic index as its coordinate axes, and selecting a flocculant that is present within a target range centered on the recommended point from among the multiple flocculants on the coordinate system. In one aspect, the trained model is a model constructed by machine learning using training data including multiple properties of multiple types of suspensions and multiple values ​​of the physical properties of multiple types of flocculants as correct labels.

[0013] In one aspect, an apparatus for selecting a flocculant to be added to a suspension includes: a physical property calculation unit having a trained model constructed by machine learning, inputting the properties of the suspension into the trained model, and outputting a recommended value of the physical property of the flocculant from the trained model; a flocculant determination unit having a database and selecting a flocculant to be added to the suspension from a plurality of flocculants in the database based on the recommended value of the physical property; a flocculant dissolution tank for dissolving the selected flocculant in a solvent; a flocculation tank for adding the flocculant dissolved in the solvent to the suspension to form flocculants; and a dewatering device for dewatering the suspension in which the flocculants have been formed. An apparatus for selecting a flocculant for a dewatering process is provided, comprising:

[0014] In one aspect, a flocculant selection device is provided for selecting a flocculant to be added to a suspension, the flocculant selection device comprising: a trained model constructed by machine learning, a physical property calculation unit that inputs the properties of the suspension into the trained model and outputs a recommended value of the physical properties of the flocculant from the trained model; and a flocculant determination unit that has a database and selects a flocculant to be added to the suspension from a plurality of flocculants in the database based on the recommended value of the physical property.

[0015] In one embodiment, the database includes identification names of a plurality of flocculants and data on a plurality of physical properties of each of the plurality of flocculants. In one embodiment, the physical properties of the flocculant are the viscosity and ionic index of the flocculant, and the ionic index is the cationic degree, anionic degree, or colloidal equivalent value of the flocculant. In one aspect, the flocculant determination unit is configured to select a flocculant having physical properties that are the same as or close to the recommended values ​​of the physical properties from among a plurality of flocculants in the database. In one aspect, the flocculant determination unit is configured to plot a recommended point identified from the recommended values ​​of the physical properties on a coordinate system having viscosity and ionic index on its coordinate axes, and to select a flocculant that is present within a target range centered on the recommended point from among multiple flocculants on the coordinate system. In one aspect, the trained model is a model constructed by machine learning using training data including multiple properties of multiple types of suspensions and multiple values ​​of the physical properties of multiple types of flocculants as correct labels.

[0016] In one aspect, there is provided a dewatering treatment system comprising: the above-mentioned flocculant selection device; a flocculant dissolution tank for dissolving the flocculant selected by the flocculant selection device in a solvent; a flocculation tank for adding the flocculant dissolved in the solvent to a suspension to form flocs; and a dewatering device for dewatering the suspension in which the flocs have been formed. Effect of the Invention

[0017] According to the present invention, a flocculant suitable for the suspension to be treated is selected based on the recommended values ​​of the physical properties of the flocculant output from the trained model. In particular, according to the present invention, a flocculant having physical properties that are the same as or close to the recommended values ​​of the physical properties can be selected from among many flocculants available on the market. As a result, flocculants suitable for the suspension to be treated can be formed, and a dehydrated cake with a low moisture content can be produced in a dewatering device. [Brief description of the drawings]

[0018] [Figure 1] FIG. 1 is a schematic diagram illustrating an embodiment of a dehydration system. [Diagram 2] FIG. 2 is a schematic diagram illustrating an example of a database. [Diagram 3] FIG. 1 is a schematic diagram illustrating an embodiment of an aggregant selection device that functions as a cloud server. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0019] Hereinafter, an embodiment of the present invention will be described with reference to the drawings. Fig. 1 is a schematic diagram showing one embodiment of a dehydration treatment system for dehydrating a suspension. Examples of suspensions include sludge, waste liquid discharged during food processing, and waste liquid discharged during the processing of industrial products such as paper. In the following description of the embodiment, sludge is used as the suspension, but the present invention is not limited to the following embodiment. Examples of sludge include primary settling sludge, final settling sludge, mixed raw sludge, drinking water sludge, sewage sludge, excess sludge from water treatment, sewage septic tank sludge, and agricultural and fishery village wastewater treatment sludge.

[0020] As shown in Fig. 1, the dewatering treatment system includes a storage tank 1 in which sludge is stored as a suspension, a coagulation tank 2 in which a coagulant is added to the sludge to form coagulated flocs, a thickener 5 that thickens the sludge from which the coagulated flocs have formed, and a dewatering device 7 that dewaters the concentrated sludge. The sludge is transferred from the storage tank 1, to the coagulation tank 2, to the thickener 5, and to the dewatering device 7, in that order.

[0021] Sludge is first supplied from a storage tank 1 to a coagulation tank 2. A coagulant is added to the sludge in the coagulation tank 2, and the coagulant and sludge are agitated by the rotation of an agitator blade 2a. By agitating the coagulant and sludge in the coagulation tank 2, suspended solids in the sludge gather to form coagulated flocs. The sludge in which the coagulated flocs have been formed is transferred to a thickener 5. The thickener 5 is a device that thickens the sludge by reducing the moisture content in the sludge. A known thickener such as a belt thickener can be used as the thickener 5. The thickener 5 may be omitted.

[0022] The concentrated sludge is transferred to the dehydrator 7. The dehydrator 7 removes water from the sludge by squeezing the sludge, and produces a dehydrated cake, which is the dehydrated sludge. Examples of the dehydrator 7 that can be used include known dehydrators such as a screw press type dehydrator, a belt press type dehydrator, and a filter press type dehydrator.

[0023] The dewatering treatment system further includes a flocculant selection device 10 for selecting a flocculant to be added to the sludge. The flocculant selection device 10 includes a storage device 10a and a computing device 10b. The storage device 10a stores a program 11, a trained model 12 constructed by machine learning, and a database 15 including information on a plurality of flocculants. The computing device 10b is configured to execute a calculation according to an instruction included in the program 11, and is further configured to execute a calculation according to the trained model 12. The storage device 10a includes a main storage device such as a random access memory (RAM), and an auxiliary storage device such as a hard disk drive (HDD) and a solid state drive (SSD). Examples of the computing device 10b include a CPU (central processing unit) and a GPU (graphic processing unit). However, the specific configuration of the flocculant selection device 10 is not limited to these examples.

[0024] The dewatering treatment system further includes a flocculant dissolving tank 17 that dissolves the flocculant selected by the flocculant selecting device 10 in a solvent. The flocculant dissolving tank 17 has an agitating blade 17a therein. Generally, the flocculant is in the form of a powder or a concentrated liquid. Therefore, the flocculant is put into a solvent (e.g., water) and mixed with the solvent by the rotation of the agitating blade 17a to obtain a flocculant that can be used in the flocculation tank 2. The flocculant produced in the flocculant dissolving tank 17 is added to the sludge in the flocculation tank 2. The type of the flocculant dissolving tank 17 is not particularly limited, but an instantaneous dissolving device can be used as the flocculant dissolving tank 17. The instantaneous dissolving device may use a filtering method or a grinding method instead of stirring. Instantaneous dissolving devices are available on the market, or can be made by oneself. The flocculant dissolving tank 17 to be used is selected based on the frequency of fluctuation in the target sludge properties at the site.

[0025] The flocculant selection device 10 includes a physical property calculation unit 21 and a flocculant determination unit 22 that are virtually constructed inside the device. The physical property calculation unit 21 has the trained model 12 stored in the storage device 10a, and the flocculant determination unit 22 has the database 15 stored in the storage device 10a. The database 15 includes identification names of a plurality of flocculants and data on a plurality of physical properties of each of the plurality of flocculants. The calculation device 10b is shared by both the physical property calculation unit 21 and the flocculant determination unit 22.

[0026] A polymer flocculant is usually used as the flocculant. The identification name of each flocculant is, for example, a trade name, a brand name, a product number, etc., which is uniquely given to each flocculant. The physical properties of each flocculant include at least the viscosity and ionic index of the flocculant, and the ionic index is the cationic degree, the anionic degree, or the colloid equivalent value.

[0027] The viscosity of a flocculant is a value measured by dissolving the flocculant in an aqueous medium at a specific concentration, and is expressed in units of mPa·s, for example. Tap water, deionized water, etc. are used as aqueous media, and the dissolution concentration is generally 0.1 to 1.0%. Rotational viscometers (B-type viscometers) are often used to measure viscosity. Since the viscosity value varies depending on the dissolution conditions, a unified dissolution method is required to compare the results. In addition, when dissolving in deionized water, etc., the obtained numerical range may become wide and the reproducibility may decrease, so the salt viscosity measured in an aqueous solution to which a certain concentration of inorganic neutral salt is added is also used. Sodium nitrate, sodium chloride, etc. are used as inorganic neutral salts. The viscosity of polymer flocculants is somewhat correlated with the molecular weight, and the higher the molecular weight, the higher the viscosity. In general, the higher the molecular weight (viscosity), the higher the cross-linking ability during flocculation, so it is effective. However, depending on the sludge dewatering machine, a higher molecular weight (viscosity) is not necessarily preferable, so it is an important indicator in the selection.

[0028] Next, regarding the anionic degree and cationic degree as ionic indicators, the molar percentage of monomers with anionic groups among the monomer components constituting the flocculant is called the anionic degree, and the molar percentage of monomers with cationic groups is called the cationic degree. Regarding cationic polymer flocculants used in sludge dewatering, those with a high ratio of monomers with cationic groups are generally called high cationic (strong cationic), those with a low ratio of monomers with cationic groups are generally called low cationic (weak cationic), and those with a medium ratio are generally called medium cationic. Nonionic monomers and some anionic monomers are copolymerized in the non-cationic part of the cationic polymer flocculant. In addition, those with both cationic and anionic groups in the molecule are generally called amphoteric polymer flocculants.

[0029] Examples of cationic monomers include neutralized salts or quaternized products of various acrylate or methacrylate monomers such as the following: dimethylaminoethyl acrylate or methacrylate, diethylaminoethyl acrylate or methacrylate, diethylaminopropyl acrylamide or methacrylamide, dimethylaminopropyl acrylamide or methacrylamide, and neutralized salts thereof with hydrogen halide, sulfuric acid, nitric acid, acetic acid, etc., and quaternized products with alkyl halides, benzyl halides, dimethyl sulfate, diethyl sulfate, etc.

[0030] Examples of the nonionic monomer include acrylamide, methacrylamide, acrylonitrile, methacrylonitrile, (meth)acrylic acid esters such as dimethylaminoethyl acrylate and dimethylaminoethyl methacrylate, and vinyl acetate.

[0031] The anionic monomer is preferably an anionic vinyl monomer, and specific examples thereof include acrylic acid, methacrylic acid, itaconic acid, maleic acid, fumaric acid, vinyl sulfonic acid, allyl sulfonic acid, methallyl sulfonic acid, styrene sulfonic acid, 2-acrylamidoethanesulfonic acid, 2-acrylamido-2-methylpropanesulfonic acid, 2-methacrylamidoethanesulfonic acid, 2-methacrylamido-2-methylpropanesulfonic acid, 2-acryloyloxyethanesulfonic acid, 3-acryloyloxypropanesulfonic acid, 4-acryloyloxybutanesulfonic acid, 2-methacryloyloxyethanesulfonic acid, 3-methacryloyloxypropanesulfonic acid, 4-methacryloyloxybutanesulfonic acid, and metal salts such as alkali metals or ammonium salts thereof.

[0032] In recent years, polyamidines, which have been widely used for dewatering difficult-to-dewater sludge, are composed entirely of cationic groups and have a cationic degree of 100 mol %. The colloid equivalent value is a standard method for measuring the actual state of dissociation by colloid titration, taking into consideration the cases where not all ionic groups are dissociated or are offset by other functional groups when the polymer flocculant is dissolved in an aqueous solution. In the case of cationic polymer flocculants, the measurement procedure is described, for example, in the Sewerage Test Method. Polymer flocculants may also be used in blends. In such cases, the cationic degree (colloid equivalent value) of each polymer flocculant may be calculated based on the blending ratio and used.

[0033] The physical property calculation unit 21 is configured to generate recommended values ​​of the physical properties of a flocculant suitable for the sludge in the storage tank 1 (i.e., the sludge to be treated) in accordance with the trained model 12. A flocculant suitable for the sludge to be treated is a flocculant that causes the moisture content of the dehydrated cake generated by the downstream dewatering device 7 to fall within a target range. In other words, when good flocculation flocs are formed in the flocculation tank 2, it is expected that the moisture content of the dehydrated cake discharged from the dewatering device 7 will fall within the target range. The target range for the moisture content of the dehydrated cake is determined in advance by an operator.

[0034] The physical property calculation unit 21 is configured to input the properties of the sludge to which the flocculant is to be added (i.e., the sludge to be treated) to the trained model 12, and execute calculations according to the trained model 12, thereby outputting a recommended value for the physical properties of the flocculant from the trained model 12. The properties of the sludge include at least one of the items listed below. TS(g / L):Total Solids SS(g / L):Suspended Solids ·CST(sec):Capillary Suction Time Colloid charge (meq / L) ·Electrical conductivity (mS / m) Alkalinity (mg / L) pH ·VSS(%):volatile suspended solids ·VTS(%):Volatile Total Solids Fiber content (%) ·Coarse particles (%) Addition amount of inorganic flocculant -Whether inorganic flocculants are added

[0035] The physical property calculation unit 21 may further input additional information consisting of any one of the items listed below to the trained model 12 in addition to the properties of the sludge. Area where sludge is obtained -Sludge treatment target areas (e.g. sewage, food, etc.) 7 types of dehydration equipment

[0036] The above properties of the coagulant can be measured by an operator using a measuring device in the storage tank 1 and manually entered into the physical property calculation unit 21, or a measurement sensor can be installed in the storage tank 1 and measurement data of the sludge properties can be automatically transmitted from the measurement sensor to the physical property calculation unit 21 and input therein.

[0037] The physical property calculation unit 21 inputs the above-mentioned sludge properties (and additional information) into the trained model 12 and executes calculations according to the trained model 12 to calculate recommended values ​​of the physical properties of the flocculant. The recommended values ​​of the physical properties are sent to the flocculant determination unit 22. The flocculant determination unit 22 is configured to select a flocculant to be added to the sludge to be treated from a plurality of flocculants in the database 15 based on the recommended values ​​of the physical properties.

[0038] FIG. 2 is a schematic diagram showing an example of the database 15. The database 15 includes identification names of a plurality of flocculants A to H, and data on a plurality of physical properties of each of the plurality of flocculants A to H. In this embodiment, the database 15 is expressed in the form of a coordinate system having the viscosity and ionic index on the coordinate axes, and the plurality of flocculants A to H are plotted in advance on the coordinate system. In the example shown in FIG. 2, the ionic index is expressed by the cationic degree and anionic degree [mol%] of the flocculant. In one embodiment, the ionic index may be expressed by the colloid equivalent value [meq / g] of the flocculant. When it is planned to use only either a cationic flocculant or an anionic flocculant, the ionic index of the database 15 may be expressed by only one of the cationic degree, anionic degree, or colloid equivalent value of the flocculant.

[0039] The flocculant determination unit 22 is configured to select a flocculant having the same or close physical property as the recommended value of the physical property determined by the physical property calculation unit 21 from among the multiple flocculants in the database 15, and output the identification name of the selected flocculant. In the embodiment shown in FIG. 2, the flocculant determination unit 22 plots a recommended point P1 determined from the recommended value of the physical property on a coordinate system, and selects a flocculant present within a target range R centered on the recommended point P1 from among the multiple flocculants on the coordinate system. The position of the recommended point P1 on the coordinate system is specified from the viscosity and cationic degree or anionic degree or colloid equivalent value constituting the recommended value of the physical property. The target range R is a rectangular area defined by a predetermined range of the viscosity and a predetermined range of the ionic index. In one embodiment, the target range R may be a circular area centered on the recommended point P1. The size of the target range R is predetermined, but can be changed as appropriate. In addition, the target range R may be a predetermined size, or may be determined based on the evaluation index of the trained model, for example, set to the same value as the mean absolute error / root mean square error, or set to a coefficient multiple.

[0040] 2, flocculant F exists within the target range R. Therefore, the flocculant determination unit 22 selects flocculant F as the flocculant to be added to the sludge, and outputs the identification name of flocculant F. In other words, flocculant F is a flocculant recommended for forming flocs in the sludge to be treated.

[0041] The flocculant F is introduced into the flocculant dissolution tank 17 shown in FIG. 1, and the flocculant F is mixed with a solvent (e.g., water) to generate a flocculant F of an appropriate concentration. The flocculant F thus obtained is introduced into the flocculation tank 2 (see FIG. 1) and mixed with sludge to form flocculated flocs. According to this embodiment, the flocculant F is determined based on recommended values ​​of the physical properties, so that the sludge is appropriately dehydrated in the downstream dewatering device 7, and it is expected that the moisture content of the dehydrated cake will fall within a predetermined target range.

[0042] Data on new flocculants can be added to the database 15. For example, data on a flocculant available on the market as a new product can be easily added to the database 15 by plotting the data on the flocculant on the coordinate system of the database 15 based on its physical properties. Therefore, the flocculant determination unit 22 can select a flocculant having the same or close physical properties as the recommended physical properties from among many flocculants. As a result, flocculants suitable for the sludge to be treated can be formed, and a dehydrated cake with a low moisture content can be produced in the dewatering device 7.

[0043] It is possible that multiple flocculants exist within the target range R. In such a case, the flocculant determination unit 22 selects multiple flocculants existing within the target range R as recommended flocculants, and outputs the identification names of each of the selected multiple flocculants. In this case, the identification names of the multiple flocculants may be output by prioritizing them in the order of shortest distance from the recommended point P1. The flocculants actually added to the sludge may be all of the multiple flocculants selected by the flocculant determination unit 22, or any one of them.

[0044] When multiple flocculants are added to the sludge, the concentrated liquids and / or powders of these flocculants may be mixed and then charged into the flocculant dissolution tank 17, or the concentrated liquids and / or powders of the flocculants may be charged into the flocculant dissolution tank 17 separately without being mixed, and then mixed inside the flocculant dissolution tank 17.

[0045] In one embodiment, the flocculant determination unit 22 may select at least one of a liquid flocculant, a powder flocculant, a polyamidine-based flocculant, and a cross-linked or non-cross-linked flocculant from among the multiple flocculants present on the coordinate system, and output an identification name of the selected flocculant.

[0046] Next, a method for creating the trained model 12 will be described. The trained model 12 is a model constructed by machine learning using training data. The training data includes multiple properties of multiple types of sludge (suspension) and multiple values ​​of the physical properties of multiple types of flocculants as correct labels. The multiple types of flocculants correspond to multiple types of sludge, respectively. That is, each flocculant has been used to form aggregated flocs of the corresponding sludge, and good dewatering results have been obtained. The multiple types of sludge refer to sludge with different properties, and may include sludge with the same components but different properties, or sludge with different components and properties.

[0047] The multiple types of sludge used in the training data are sludge (suspension) that has actually been treated in the dewatering treatment system shown in FIG. 1 or another dewatering treatment system. The specific items of the properties of these sludge are the same as the above-mentioned property items input to the trained model 12. If the above-mentioned additional information (e.g., the area where the sludge was obtained) is input to the trained model 12 in addition to the sludge properties, the training data further includes the same items of additional information. The sludge properties and the additional information are explanatory variables for deriving recommended values ​​for the physical properties of the coagulant to be predicted.

[0048] The physical property value of the flocculant as the correct label is the objective variable, and more specifically, is the physical property value of the flocculant that allows the moisture content of the dehydrated cake actually produced by the dewatering device 7 to fall within the target range. The multiple types of flocculants used as the correct label are the flocculants that were actually added to the multiple types of sludge. These multiple types of flocculants used in the training data are respectively associated with the multiple types of sludge that were actually treated as explanatory variables. The training data created in this manner is stored in the storage device 10a of the flocculant selection device 10. The training data is also called learning data.

[0049] The physical property calculation unit 21 uses the training data to create a model according to a machine learning algorithm. The model constructed by machine learning is stored in the storage device 10a as the trained model 12. Examples of machine learning algorithms include, but are not limited to, the SVR method (support vector regression method), the PLS method (partial least squares method), the deep learning method, the random forest method, the decision tree method, the gradient boosting method, and various ensemble learning methods. The structure of the trained model 12 differs depending on the machine learning used. For example, when the random forest method is used for machine learning, the trained model 12 is a decision tree. In another example, when the deep learning method is used for machine learning, the trained model 12 is a neural network having an input layer, multiple intermediate layers (hidden layers), and an output layer.

[0050] The flocculant selection device 10 is composed of at least one computer. The at least one computer may be one server or multiple servers. The flocculant selection device 10 may be an edge server arranged in a factory in which the flocculation tank 2 and the dewatering device 7 are installed, or may be a cloud server connected to a communication network such as the Internet or a local area network.

[0051] The flocculant selection device 10 may be at least one computer connected by a communication network such as the Internet or a local area network. For example, the flocculant selection device 10 may be a cloud server, or a combination of an edge server and a cloud server. The physical property calculation unit 21 and the flocculant determination unit 22 may be configured in multiple computers installed in different locations. Each of the physical property calculation unit 21 and the flocculant determination unit 22 may include a storage device 10a and a calculation device 10b.

[0052] Fig. 3 is a schematic diagram showing one embodiment of the flocculant selection device 10 functioning as a cloud server. The configuration of this embodiment that is not particularly described is the same as the embodiment described with reference to Fig. 1 and Fig. 2, so duplicated description will be omitted. As shown in Fig. 3, the flocculant selection device 10 is placed in a remote location away from the treatment facility in which the flocculation tank 2, the dewatering device 7, etc. are installed, and is connected to a communication network such as the Internet or a local area network.

[0053] A user can connect a terminal device 30 (e.g., a laptop computer, a smartphone, or a tablet device) to the flocculant selection device 10 via a communication network. The terminal device 30 transmits property data of the sludge to be treated to the flocculant selection device 10 via the communication network, and the property data of the sludge is input to a physical property calculation unit 21. The flocculant determination unit 22 selects an optimal flocculant for the sludge from a plurality of flocculants in the database 15, and transmits an identification name of the selected flocculant to the terminal device 30.

[0054] The selected flocculant is added to the sludge in the flocculation tank 2 to form flocculants. The sludge from which the flocculants have been formed is thickened by the thickener 5 and then dewatered by the dewatering device 7. The moisture content of the dewatered cake discharged from the dewatering device 7 is estimated to be within a target range. Therefore, the physical property values ​​(viscosity and ionic index) of the flocculant selected by the flocculant determination unit 22 may be added to the training data together with the properties of the sludge input to the physical property calculation unit 21.

[0055] The property data of the sludge to be treated may be input by an operator into a terminal device 30 and transmitted to the physical property calculation unit 21 of the flocculant selection device 10 via a communications network, or a measurement sensor may be installed in the storage tank 1 and the property data of the sludge obtained by the measurement sensor may be transmitted to the physical property calculation unit 21 of the flocculant selection device 10 via a communications network.

[0056] In this way, the user can use the flocculant selection device 10 even from a remote location. The method of accessing the flocculant selection device 10 from the terminal device 30 may be by using an application or a web browser. The flocculant selection device 10 may have a security function that allows access only from terminal devices registered in the flocculant selection device 10.

[0057] In the embodiment shown in Fig. 3, the flocculant selection device 10 is connected to other dewatering treatment systems via a communication network. Data on the properties of sludge used in these other dewatering treatment systems and data on the physical properties of flocculants that have provided good dewatering results (i.e., the moisture content of the dewatered cake falls within the target range) are transmitted to the flocculant selection device 10 via the communication network and added to the training data described above.

[0058] A user may operate the terminal device 30 to issue a command to the flocculant selection device 10 to execute machine learning. Furthermore, the flocculant selection device 10 may periodically or irregularly execute machine learning using the latest training data to update the trained model 12. For example, when the moisture content of the dehydrated cake generated by the dewatering device 7 is no longer within the target range, the flocculant selection device 10 may execute machine learning to update the trained model 12.

[0059] In the embodiments described so far, sludge is used as an example of the suspension, but the suspension is not limited to sludge. For example, the suspension may be a suspension discharged in the processing of food such as fruit and oil, or a suspension discharged in the processing of industrial products such as recycling waste paper.

[0060] The above-described embodiments have been described for the purpose of enabling a person having ordinary skill in the art to practice the present invention. Various modifications of the above-described embodiments are naturally possible for a person skilled in the art, and the technical idea of ​​the present invention can be applied to other embodiments. Therefore, the present invention is not limited to the described embodiments, but is to be interpreted in the broadest scope according to the technical idea defined by the claims. [Explanation of symbols]

[0061] 1. Storage tank 2 Coagulation tank 5 Concentrator 7 Dehydration equipment 10. Flocculant selection device 11 Program 12 Pre-trained models 15 Database 17 Coagulant dissolving tank 21 Physical Properties Calculation Department 22 Coagulant Determination Section 30 Terminal Equipment

Claims

1. 1. A method for selecting a flocculant to be added to a suspension, comprising the steps of: The properties of the suspension are input into a trained model constructed by machine learning; Outputting recommended values ​​of the physical properties of the flocculant from the trained model; selecting a flocculant to be added to the suspension from a plurality of flocculants in a database based on the recommended values ​​of the physical properties; The properties of the suspension are as follows: TS (g / L): Total Solids SS (g / L): Suspended Solids CST (sec): Capillary Suction Time Colloid charge (meq / L) Conductivity (mS / m) Alkalinity (mg / L) pH VSS (%): volatile suspended solids VTS (%): Volatile Total Solids Fiber content (%) Coarse particles (%) Addition amount of inorganic flocculant Addition of inorganic flocculant At least one of A method for selecting a flocculant for dewatering treatment, characterized in that the learned model is a model constructed by machine learning using training data including multiple properties of multiple types of suspensions and multiple values ​​of the physical properties of multiple types of flocculants as correct labels.

2. The method according to claim 1, characterized in that selecting a flocculant to be added to the suspension from among a plurality of flocculants in the database based on the recommended value of the physical property comprises selecting a flocculant from among a plurality of flocculants in the database having physical properties that are the same as or close to the recommended value of the physical property.

3. Selecting a flocculant having a physical property that is the same as or close to the recommended value of the physical property from a plurality of flocculants in the database Plotting recommended points identified from the recommended values ​​of the physical properties on a coordinate system having the viscosity and ionic index as coordinate axes; The method according to claim 2, characterized in that a flocculant present within a target range centered on the recommended point is selected from a plurality of flocculants on the coordinate system.

4. 1. An apparatus for selecting a flocculant to be added to a suspension, comprising: A physical property calculation unit having a trained model constructed by machine learning, inputting the properties of the suspension into the trained model, and outputting a recommended value of the physical property of the flocculant from the trained model; a flocculant determination unit having a database and selecting a flocculant to be added to the suspension from a plurality of flocculants in the database based on the recommended value of the physical property; a flocculant dissolving tank for dissolving the selected flocculant in a solvent; a flocculation tank for adding the flocculant dissolved in the solvent to the suspension to form flocs; a dewatering device for dewatering the suspension in which the flocs have been formed; It is equipped with The properties of the suspension are as follows: TS (g / L): Total Solids SS (g / L): Suspended Solids CST (sec): Capillary Suction Time Colloid charge (meq / L) Conductivity (mS / m) Alkalinity (mg / L) pH VSS (%): volatile suspended solids VTS (%): Volatile Total Solids Fiber content (%) Coarse particles (%) Addition amount of inorganic flocculant Addition of inorganic flocculant At least one of An apparatus for selecting a flocculant for dewatering treatment, characterized in that the trained model is a model constructed by machine learning using training data including multiple properties of multiple types of suspensions and multiple values ​​of the physical properties of multiple types of flocculants as correct labels.

5. An apparatus for selecting a flocculant to be added to a suspension, comprising: A physical property calculation unit having a trained model constructed by machine learning, inputting the properties of the suspension into the trained model, and outputting a recommended value of the physical property of the flocculant from the trained model; a flocculant determination unit having a database and selecting a flocculant to be added to the suspension from a plurality of flocculants in the database based on the recommended value of the physical property; An instantaneous dissolution device as a flocculant dissolution tank for dissolving the selected flocculant in a solvent; a flocculation tank for adding the flocculant dissolved in the solvent to the suspension to form flocs; a dewatering device for dewatering the suspension in which the flocs have been formed; An apparatus for selecting a flocculant for dewatering treatment, comprising:

Images