Resin manufacturing method and resin manufacturing device

Abstract

To provide a method for producing a resin and an apparatus for producing a resin which can accurately control a completion timing of a polymerization step of subjecting a resin reaction raw material to polymerization reaction.SOLUTION: A method for producing a resin includes a step of subjecting a reaction raw material to polymerization reaction, wherein the production method includes the steps of: imaging at least a part of a reaction mixture in the polymerization reaction of the reaction raw material in time series, and acquiring an image; and analyzing the acquired image by a machine learning method, and determining a degree of progress of the polymerization reaction; and finishing the polymerization reaction on the basis of the result of the degree of progress of the determined polymerization reaction. An apparatus for manufacturing a resin includes: a reaction container for subjecting the reaction raw material to the polymerization reaction; an imaging part for imaging at least a part of a reaction mixture in the polymerization reaction of the reaction raw material in time series, and acquiring an image; and a machine learning part which has a machine learning function, analyzes the acquired image by the machine learning method, and determines the degree of progress of the polymerization reaction.SELECTED DRAWING: Figure 1

Description

[Technical Field]

[0001] The present invention relates to a method and an apparatus for producing a resin. [Background technology]

[0002] Conventionally, to determine the end of a polymerization process in which raw materials for producing a resin are polymerized, workers would sample the resin during the polymerization reaction, adjust the temperature, and then measure properties such as bubble viscosity, and use the results to determine the progress of the polymerization process. [Prior art documents] [Patent documents]

[0003] [Patent Document 1] Japanese Patent Application Publication No. 2020-064589 Summary of the Invention [Problem to be solved by the invention]

[0004] However, with the above-described method, there is a time lag of several tens of minutes between the time of sampling and the time of obtaining the measurement results of properties such as bubble viscosity, and even if the progress of the polymerization process can be accurately determined, there is a problem in that it is not possible to accurately control the end time of the polymerization process.

[0005] For example, Patent Document 1 above describes a general technique for detecting process abnormalities in the production of chemical products in real time without any time lag, but does not mention a method for detecting the completion time of the polymerization reaction of resin reaction raw materials.

[0006] Therefore, an object of the present invention is to provide a resin manufacturing method and a resin manufacturing apparatus that can accurately control the end time of a polymerization step in which resin reaction raw materials are polymerized. [Means for solving the problem]

[0007] The gist and configuration of the present invention are as follows. (1) A method for producing a resin, comprising a step of polymerizing reactant materials, The manufacturing method includes: a step of acquiring images by capturing time-series images of at least a part of a reaction mixture during a polymerization reaction of the reaction raw materials using an imaging unit; a step of analyzing the acquired image by a machine learning technique using a machine learning unit and determining the progress of the polymerization reaction; a step of terminating the polymerization reaction based on the determined progress of the polymerization reaction; A method for producing a resin, comprising: Here, the "image" is a still image or a moving image.

[0008] (2) The method for producing a resin according to (1) above, wherein the number average molecular weight of the resin after the polymerization reaction is within the range of 900 to 20,000.

[0009] (3) The method for producing a resin according to (1) or (2) above, wherein the molecular weight of the reaction raw material contained in the reaction mixture is within a range of 30 to 400.

[0010] (4) The method for producing a resin according to any one of (1) to (3) above, wherein the resin is a resin for paint.

[0011] (5) The method for producing a resin according to (4) above, wherein the resin is a polyester resin for paint.

[0012] (6) A method for producing a resin described in any one of (1) to (5) above, wherein the portion of the reaction mixture that is imaged chronologically in the image acquisition step is the liquid surface portion of the reaction mixture.

[0013] (7) The method for producing a resin according to any one of (1) to (6) above, wherein the analysis and determination in the step of determining the progress of the polymerization reaction are carried out based on at least one of stirring marks on the liquid surface of the reaction mixture, the brightness of the liquid surface, and the state of bubbles on the liquid surface.

[0014] (8) The analysis and determination in the step of determining the progress of the polymerization reaction are The method for producing a resin according to any one of (1) to (7), wherein pixels containing stirring marks present on the liquid surface of the reaction mixture are extracted and analyzed based on the image to determine the progress of the polymerization reaction.

[0015] (9) The analysis and determination in the step of determining the progress of the polymerization reaction are The method for producing a resin according to any one of (1) to (8) above, wherein the analysis and judgment are performed by the machine learning unit that has been trained in advance using learning relationship data that shows the relationship between images of at least a portion of the reaction mixture taken in time series during the polymerization reaction of the reaction raw materials and the progress of the polymerization reaction.

[0016] (10) The method for producing a resin described in any one of (1) to (9) above, wherein the machine learning unit has a temperature correction function that performs correction in the analysis and / or judgment according to the temperature of the reaction mixture.

[0017] (11) a reaction vessel for polymerizing the reaction raw materials; an imaging unit that acquires images by capturing time-series images of at least a part of a reaction mixture during a polymerization reaction of the reaction raw materials; a machine learning unit having a machine learning function, which analyzes the acquired image by a machine learning method and determines the progress of the polymerization reaction; A resin manufacturing apparatus comprising: [Effects of the Invention]

[0018] According to the present invention, it is possible to provide a resin manufacturing method and a resin manufacturing apparatus that can accurately control the end time of a polymerization step in which reactive raw materials for a resin are polymerized. [Brief explanation of the drawings]

[0019] [Figure 1]1 is a flowchart of a method for producing a resin according to one embodiment of the present invention. [Figure 2] FIG. 1 is a diagram showing the liquid surface of a reaction mixture for each acid value. [Figure 3] 1 is a schematic diagram of a resin production apparatus according to one embodiment of the present invention. DETAILED DESCRIPTION OF THE INVENTION

[0020] DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS Hereinafter, embodiments of the present invention will be described in detail with reference to the accompanying drawings.

[0021] <Resin manufacturing method> FIG. 1 is a flowchart of a method for producing a resin according to one embodiment of the present invention. First, the resin to be produced will be described. In this embodiment, the resin is a paint resin. The paint resin may contain, for example, one or more of acrylic resin, polyester resin, epoxy resin, amino resin, and blocked isocyanate. Note that the resin of the present invention is not limited to a paint resin as long as it is produced through a polymerization process and causes a significant change in the liquid surface portion to be imaged during the polymerization reaction.

[0022] As shown in FIG. 1, the resin manufacturing method according to this embodiment includes the steps of preparing reactive raw materials (step S101), polymerizing the reactive raw materials (step S102), capturing images of at least a portion of the reaction mixture in the polymerization reaction of the reactive raw materials in a time-series manner using an imaging unit to obtain images (step S103), analyzing the obtained images using a machine learning technique using a machine learning unit to determine the progress of the polymerization reaction (step S104), and terminating the polymerization reaction based on the determined progress of the polymerization reaction (step S105). The step of preparing the reaction raw materials (step S101) and the step of polymerizing the reaction raw materials (step S102) will be described below using acrylic resin, polyester resin, and epoxy resin as examples. Note that the term "polymerization reaction" in this specification includes various polymerization reactions such as polycondensation reaction, radical polymerization reaction, and cationic polymerization reaction.

[0023] (acrylic resin) The acrylic resin can be prepared by copolymerizing a hydroxyl group-containing monomer (a) with another monomer (b). Examples of the hydroxyl group-containing monomer (a) include 2-hydroxyethyl (meth)acrylate, hydroxypropyl (meth)acrylate, 2,3-dihydroxybutyl (meth)acrylate, 4-hydroxybutyl (meth)acrylate, reaction products of these hydroxyl group-containing (meth)acrylates with ε-caprolactone, and esters of polyhydric alcohols such as polyethylene glycol mono(meth)acrylate with acrylic acid or methacrylic acid. Furthermore, reaction products obtained by ring-opening polymerization of the monoesters of the above polyhydric alcohols with acrylic acid or methacrylic acid with ε-caprolactone can also be used. These hydroxyl group-containing monomers (a) may be used alone or in combination. In this specification, (meth)acrylate refers to acrylate or methacrylate.

[0024] Examples of other monomers (b) include carboxyl group-containing monomers such as acrylic acid, methacrylic acid, crotonic acid, maleic acid, itaconic acid, and fumaric acid, and dicarboxylic acid monoester monomers such as ethyl maleate, butyl maleate, ethyl itaconate, and butyl itaconate; (meth)acrylic acid alkyl ester monomers such as methyl (meth)acrylate, ethyl (meth)acrylate, propyl (meth)acrylate, n-, i-, or t-butyl (meth)acrylate, 2-ethylhexyl (meth)acrylate, and lauryl (meth)acrylate; alicyclic group-containing monomers such as cyclopentyl (meth)acrylate, cyclohexyl (meth)acrylate, isobornyl (meth)acrylate, tricyclodecanyl (meth)acrylate, and adamantyl (meth)acrylate; and amino (meth)acrylate. Examples of the other monomers (b) include aminoalkyl (meth)acrylate ester monomers such as ethyl (meth)acrylate, dimethylaminoethyl (meth)acrylate, and butylaminoethyl (meth)acrylate; aminoalkyl (meth)acrylate amide monomers such as aminoethyl (meth)acrylamide, dimethylaminomethyl (meth)acrylamide, and methylaminopropyl (meth)acrylamide; other amide group-containing monomers such as acrylamide, methacrylamide, N-methylolacrylamide, methoxybutylacrylamide, and diacetone acrylamide; vinyl cyanide monomers such as (meth)acrylonitrile and α-chloroacrylonitrile; saturated aliphatic carboxylic acid vinyl ester monomers such as vinyl acetate and vinyl propionate; and styrene-based monomers such as styrene, α-methylstyrene, and vinyltoluene. These other monomers (b) may be used singly or in combination of two or more. Of the other monomers (b) above, acrylic acid, methacrylic acid, methyl (meth)acrylate, ethyl (meth)acrylate, 2-ethylhexyl (meth)acrylate, lauryl (meth)acrylate, cyclohexyl (meth)acrylate, and the like are preferably used.

[0025] The polymerization method for the hydroxyl group-containing monomer (a) and the other monomer (b) can be a method commonly used by those skilled in the art. Examples of the polymerization method include bulk polymerization using a radical polymerization initiator, solution polymerization, and a two-stage bulk-suspension polymerization method in which bulk polymerization is followed by suspension polymerization. Among these, solution polymerization is particularly preferred. Examples of the solution polymerization method include a method in which the monomer mixture is heated with stirring in the presence of a radical polymerization initiator at a temperature of, for example, 80 to 200°C.

[0026] The number average molecular weight of the acrylic resin is 3,000 to 20,000, preferably 3,500 to 15,000, for example, 3,500 to 12,000. When the number average molecular weight of the acrylic resin is within the above range, the drying properties of the coating composition can be improved. In the present invention, an acrylic resin having a molecular weight within the above range is not simply used, but is used in combination with the extender pigment and viscosity modifier of the present invention. This combination allows the use of an acrylic resin having a molecular weight within the above range to improve the drying properties of the paint without impairing the appearance of the paint film. If the number average molecular weight is 3,000 or more, better drying properties of the coating film can be obtained and the physical properties of the resulting multi-layer coating film can also be improved, and if the number average molecular weight is 20,000 or less, better coating film appearance such as gloss can be obtained.

[0027] Furthermore, since the number average molecular weight of the acrylic resin is within the above range, the coating composition has good drying properties, and for example, it is possible to prevent dust adhesion due to stickiness of the coating composition scattered in the coating booth, thereby maintaining a good coating environment.

[0028] In this specification, the number average molecular weight is a value measured by GPC (gel permeation chromatography) and converted into polystyrene. Although only one type of acrylic resin can be used, two or more types can also be used in combination to balance the coating film performance.

[0029] The acrylic resin preferably has a hydroxyl value of 50 to 250 mgKOH / g on solid content. Having a hydroxyl value of solid content within the above range allows for appropriate reaction with curing agents such as blocked isocyanates and amino resins for paints, resulting in desired coating film properties. The acrylic resin more preferably has a hydroxyl value of 50 to 200 mgKOH / g on solid content. The acrylic resin preferably has an acid value of 2 to 50 mgKOH / g on solid content. Having an acid value of solid content within the above range allows for desired coating film properties. The acrylic resin more preferably has an acid value of 5 to 20 mgKOH / g on solid content. Note that, in this specification, the acid value and hydroxyl value of the resin are both expressed as values ​​converted into solid content, and are values ​​measured by a method in accordance with JIS K 0070.

[0030] (polyester resin) As the polyester resin, a polyester resin having two or more hydroxyl groups per molecule, generally called polyester polyol, is preferably used. Such a polyester resin can be prepared by polycondensation (esterification) of a polyhydric alcohol with a polybasic acid or its anhydride. Methods commonly used by those skilled in the art can be used as a method for polycondensation (esterification) of a polyhydric alcohol with a polybasic acid or its anhydride. Examples of polycondensation include atmospheric solution polymerization and pressurized solution polymerization. Examples of atmospheric solution polymerization include a method in which the above-mentioned monomer mixture is heated and dehydrated with stirring at a temperature of, for example, 180 to 210°C under reflux with xylene in the presence of a dibutyltin oxide catalyst.

[0031] Examples of polyhydric alcohols include ethylene glycol, diethylene glycol, polyethylene glycol, propylene glycol, dipropylene glycol, polypropylene glycol, neopentyl glycol, 1,2-butanediol, 1,3-butanediol, 2,3-butanediol, 1,4-butanediol, 1,5-pentanediol, 1,6-hexanediol, hydrogenated bisphenol A, hydroxyalkylated bisphenol A, 1,4-cyclohexanedimethanol, 2,2-dimethyl-3-hydroxybenzoates, and benzophenone-2. Examples of polyhydric alcohols include hydroxypropyl-2,2-dimethyl-3-hydroxypropionate, 2,2,4-trimethyl-1,3-pentanediol, N,N-bis-(2-hydroxyethyl)dimethylhydantoin, polytetramethylene ether glycol, polycaprolactone polyol, glycerin, sorbitol, trimethylolethane, trimethylolpropane, trimethylolbutane, hexanetriol, pentaerythritol, dipentaerythritol, tris-(hydroxyethyl)isocyanate, etc. These polyhydric alcohols may be used alone or in combination of two or more.

[0032] Examples of polybasic acids or anhydrides thereof include phthalic acid, phthalic anhydride, tetrahydrophthalic acid, tetrahydrophthalic anhydride, hexahydrophthalic acid, hexahydrophthalic anhydride, methyltetrahydrophthalic acid, methyltetrahydrophthalic anhydride, himic anhydride, trimellitic acid, trimellitic anhydride, pyromellitic acid, pyromellitic anhydride, isophthalic acid, terephthalic acid, maleic acid, maleic anhydride, fumaric acid, itaconic acid, adipic acid, azelaic acid, sebacic acid, succinic acid, succinic anhydride, lactic acid, dodecenylsuccinic acid, dodecenylsuccinic anhydride, cyclohexane-1,4-dicarboxylic acid, endo acid anhydride, etc. These polybasic acids or anhydrides thereof may be used alone or in combination of two or more.

[0033] The polyester resin may be a modified polyester resin prepared as described above and modified with lactone, fats and oils or fatty acids, melamine resin, urethane resin, or the like. For example, the fats and oils or fatty acid modified polyester resin is a polyester resin modified with fats and oils such as castor oil, dehydrated castor oil, palm oil, corn oil, cottonseed oil, linseed oil, perilla oil, poppy seed oil, safflower oil, soybean oil, or tung oil, or fatty acids extracted from these fats and oils. In producing this fats and oils or fatty acid modified polyester resin, it is preferable to add up to about 30 parts by mass of the fats and oils and / or fatty acids in total per 100 parts by mass of the polyester resin.

[0034] The polyester resin preferably has a number average molecular weight of 1,000 to 20,000. Having a number average molecular weight within the above range improves coating workability and provides desired coating film performance. The polyester resin more preferably has a number average molecular weight of 1,200 to 10,000. The polyester resin preferably has a solid content hydroxyl value of 40 to 350 mgKOH / g. Having a solid content hydroxyl value within the above range allows for appropriate reaction with blocked isocyanate, providing desired coating film performance. The polyester resin more preferably has a solid content hydroxyl value of 40 to 300 mgKOH / g.

[0035] (epoxy resin) The epoxy resin preferably contains at least one of a bisphenol A epoxy resin and a novolac epoxy resin. Suitable epoxy resins include those having two or more glycidyl groups per molecule, and those in which a monocarboxylic acid, a monoamine, or a monophenol compound is added to a terminal glycidyl group. Such epoxy resins can be prepared as a polymer of a low-molecular-weight polyhydric alcohol and a low-molecular-weight polyhydric glycidyl group-containing compound. Examples of preparation methods include heating a diadduct of bisphenol A and epichlorohydrin of bisphenol A in a xylene solvent at a temperature of, for example, 80 to 180°C while stirring, using a tertiary amine catalyst.

[0036] In the present invention, the epoxy resin may further contain other epoxy resins in addition to the bisphenol A epoxy resin and / or novolac epoxy resin. Specific examples of other epoxy resins include bisphenol F epoxy resin, bisphenol AD ​​epoxy resin, bisphenol S epoxy resin, biphenyl epoxy resin, naphthalene epoxy resin, glycidyl ester epoxy resin, glycidyl amine epoxy resin, and rubber-modified epoxy resin.

[0037] The amount of bisphenol A epoxy resin and / or novolac epoxy resin in the total amount of epoxy resin is preferably 10 to 100 mass %, more preferably 20 to 100 mass %. When the amount of bisphenol A epoxy resin and / or novolac epoxy resin in the total amount of epoxy resin is within the above range, adhesion to the substrate is further improved.

[0038] The polycondensation (esterification) method of a polyhydric alcohol and a polybasic acid or anhydride thereof can be a method commonly used by those skilled in the art. Examples of polycondensation include atmospheric solution polymerization and pressurized solution polymerization. Examples of atmospheric solution polymerization include a method in which the above-mentioned monomer mixture is heated and dehydrated with stirring at a temperature of, for example, 180 to 210°C in the presence of a dibutyltin oxide catalyst under reflux with xylene.

[0039] The number average molecular weight of the epoxy resin is preferably 900 to 20,000. Having a number average molecular weight within the above range has the advantage of providing better coating film properties and coating workability.

[0040] (amino resin) The amino resin used in the present invention is a conventional amino resin etherified with a monohydric alcohol, generally containing 2 to 7% free formaldehyde. Examples of the amino resin include melamine resins, acetoguanamine resins, benzoguanamine resins, urea resins, thiourea resins, melamine-acetoguanamine co-condensation resins, melamine-benzoguanamine co-condensation resins, melamine-thiourea co-condensation resins, acetoguanamine-benzoguanamine co-condensation resins, acetoguanamine-urea co-condensation resins, acetoguanamine-thiourea co-condensation resins, benzoguanamine-urea co-condensation resins, benzoguanamine-thiourea co-condensation resins, and urea-thiourea co-condensation resins, all of which are etherified with a monohydric alcohol. The monohydric alcohol may be any one that is commonly used in etherification reactions, such as methanol, ethanol, n-propanol, isopropanol, n-butanol, isobutanol, sec-butanol, n-amyl alcohol, n-octanol, 2-ethylhexyl alcohol, lauryl alcohol, ethylene glycol monomethyl ether, ethylene glycol monoethyl ether, ethylene glycol monobutyl ether, and allyl alcohol, and one or more of these may be used. The amino resin etherified with the monohydric alcohol can be produced by a conventional method. For example, a predetermined amino compound, which is the raw material for the amino resin, and formaldehyde are first reacted with stirring at a temperature of 70 to 110°C in the presence of a basic catalyst to form a methylolated product, which is then etherified with a monohydric alcohol in the presence of an acidic catalyst. Alternatively, the amino compound and formaldehyde may be simultaneously methylolated and etherified in the presence of an acidic catalyst and a monohydric alcohol. Heating and stirring may then be continued until the reaction is complete.

[0041] (Blocked isocyanate) Examples of blocked isocyanates that can be used include polyfunctional isocyanate compounds such as aliphatic, alicyclic, aromatic-group-containing aliphatic, or aromatic diisocyanates, diisocyanate dimers, and diisocyanate trimers (preferably isocyanurate-type isocyanates (so-called isocyanurates)), to which a substance having active hydrogen, such as alcohol, amine, β-diketone, amide, carboxylic acid, or phenol, is added as a blocking agent. Such blocked isocyanates may be so-called asymmetric types.

[0042] Examples of diisocyanates include diisocyanates containing 5 to 24, preferably 6 to 18, carbon atoms. Examples of such diisocyanates include trimethylene diisocyanate, tetramethylene diisocyanate, hexamethylene diisocyanate (HDI), 2,2,4-trimethylhexane diisocyanate, undecamethyldisocyanate, and the like. Diisocyanate-(1,11), lysine ester diisocyanate, cyclohexane-1,3- and 1,4-diisocyanate, 1-isocyanato-3-isocyanatomethyl-3,5,5-trimethylcyclohexane (isophorone diisocyanate: IPDI), 4,4'-diisocyanatodicyclomethane, ω,ω'-dipropyl ether diisocyanate, thiodipropyl diisocyanate, cyclohexyl-1,4-diisocyanate, dicyclohexylmethane-4,4'-diisocyanate, 1,5-dimethyl-2,4-bis(isocyanatomethyl)benzene, 1,5-trimethyl-2,4-bis(ω-isocyanatoethyl)-benzene, 1,3,5-trimethyl-2,4-bis(isocyanatomethyl) )benzene, 1,3,5-triethyl-2,4-bis(isocyanatomethyl)benzene, dicyclohexyldimethylmethane-4,4'-diisocyanate, etc. Also usable are aromatic diisocyanates such as 2,4-diisocyanatotoluene and / or 2,6-diisocyanatotoluene, 4,4'-diisocyanatodiphenylmethane, and 1,4-diisocyanatoisopropylbenzene. Examples of the isocyanurate-type isocyanates include trimers of the above-mentioned diisocyanates. Note that such blocked isocyanates may be used alone or in combination of two or more.

[0043] The blocking agent may be any of those well known in the art, for example, phenol-based blocking agents such as phenol, cresol, xylenol, chlorophenol, and ethylphenol; lactam-based blocking agents such as ε-caprolactam, δ-valerolactam, γ-butyrolactam, and β-propiolactam; active methylene-based blocking agents such as ethyl acetoacetate and acetylacetone; methanol, ethanol, propanol, butanol, amyl alcohol, ethylene glycol monomethyl ether, ethylene glycol monoethyl ether, ethylene glycol monobutyl ether, diethylene glycol monomethyl ether, propylene glycol monomethyl ether, benzyl alcohol, methyl glycolate, butyl glycolate, and diacetone alcohol. Examples of blocking agents include alcohol-based blocking agents such as chol, methyl lactate, and ethyl lactate; oxime-based blocking agents such as formaldoxime, acetaldoxime, acetoxime, methyl ethyl ketoxime, diacetyl monooxime, and cyclohexane oxime; mercaptan-based blocking agents such as butyl mercaptan, hexyl mercaptan, t-butyl mercaptan, thiophenol, methylthiophenol, and ethylthiophenol; acid amide-based blocking agents such as acetic acid amide and benzamide; imide-based blocking agents such as succinimide and maleic acid imide; imidazole-based blocking agents such as imidazole and 2-ethylimidazole; pyrazole-based blocking agents; and triazole-based blocking agents.

[0044] It is preferable that at least a portion of the blocked isocyanates contained in the blocked isocyanate is an isocyanurate-type blocked isocyanate. Furthermore, in the present invention, the above-mentioned isocyanurate-type blocked isocyanate can also be used as a mixture in combination with other aliphatic, alicyclic, aromatic-group-containing aliphatic, or aromatic polyfunctional blocked isocyanates (preferably diisocyanates). In this case, the content of the above-mentioned isocyanurate-type blocked isocyanate in the total amount of blocked isocyanates is preferably 60% by mass or more.

[0045] The content of the blocked isocyanate in the film-forming resin is preferably such that the number of isocyanate groups in the blocked isocyanate is 0.5 to 1.5 per the total number of hydroxyl groups in the film-forming resin, 1. When the content of the blocked isocyanate is within the above range, sufficient curing is achieved, and the desired coating properties are obtained.

[0046] As a polymerization method for adding an active hydrogen-containing substance to an isocyanate compound, a method commonly used by those skilled in the art can be used. For example, the polymerization method includes a method in which an active hydrogen-containing substance is added dropwise to an isocyanate compound in a dibutyltin laurate catalyst and a methyl isobutyl ketone solvent while stirring and heating at a temperature of, for example, 60 to 110°C, and then heating and stirring are continued until the reaction is completed.

[0047] For other types of resins, the preparation of reaction raw materials and the polymerization process can be carried out using known methods.

[0048] Referring back to FIG. 1, in this embodiment, the imaging unit then captures images (still images or video images) of at least a part of the reaction mixture in the polymerization reaction of the reaction raw materials in time series (step S103).

[0049] In this example, the imaging unit is a camera. The camera can be any known camera. In this example, one camera is installed directly above the reaction vessel in which the polymerization reaction is performed so that the liquid surface of the reaction mixture can be imaged from directly above. The camera can also be installed at another position as long as it is capable of imaging at least a portion of the reaction mixture. Alternatively, multiple cameras may be installed to acquire images from multiple directions.

[0050] As described above, the portion of the reaction mixture to be imaged is preferably the liquid surface (a portion including the liquid surface). In this example, the liquid surface is the top surface of the reaction mixture, but it may also be the side or bottom surface, and imaging can be performed, for example, by providing a transparent window. Furthermore, imaging is performed in a time series manner, but to eliminate time lag, it is preferable to acquire images at least every 10 minutes. The image acquired every 10 minutes may be a single image, or an image set may be formed, for example, with 150 time-series images captured every 0.033 seconds (30 FPS) for 5 seconds. Furthermore, as a method of acquiring images in a time series, a video may be first recorded and time-series images may be acquired immediately.

[0051] In this embodiment, the machine learning unit then analyzes the acquired image using a machine learning technique to determine the progress of the polymerization reaction (step S104). To eliminate the time lag, as soon as an image is acquired at a certain time in step S103, step S104 is immediately performed on the acquired image without waiting for the next image to be captured. The machine learning unit can be any known processor. The imaging unit and the machine learning unit can have a communication function for transmitting and receiving image data.

[0052] Figure 2 shows the liquid surface of the reaction mixture for each acid value. As shown in Figure 2, it was found that the image of the liquid surface of the reaction mixture changes significantly with the progress of the polymerization reaction (in the example of Figure 2, the acid value is used as an indicator).

[0053] The machine learning unit has a machine learning function. It is preferable to train the machine learning unit in advance using learning relational data showing the relationship between time-series images of at least a portion of a reaction mixture in a polymerization reaction of reaction raw materials and the progress of the polymerization reaction. That is, an artificial intelligence model is created in advance using the relationship data through supervised learning, with the input being the acquired image data and the output being the progress of the polymerization reaction. Then, the image data acquired in step S103 is input to the artificial intelligence model, whereby the above analysis is performed, and the progress of the polymerization reaction is output and determined. It is preferable that steps S101 to S104 are performed using the same equipment and conditions as those used to acquire the relational data. The machine learning algorithm is not particularly limited, but it is preferable to use a genetic algorithm, a neural network (including deep learning), a support vector machine, a k-nearest neighbor method, or boosting. It is preferable that the machine learning unit includes a storage unit (memory) for storing various information such as the learning relational data.

[0054] 2, it has been found that the progress of the polymerization reaction is reflected in the stirring marks on the liquid surface of the reaction mixture, the brightness of the liquid surface, the state of bubbles on the liquid surface, etc. Therefore, the analysis and determination in the step of determining the progress of the polymerization reaction are preferably performed based on at least one or more of the stirring marks on the liquid surface of the reaction mixture, the brightness of the liquid surface, and the state of bubbles on the liquid surface.

[0055] Specifically, regarding the determination of the feature amounts, it is preferable that the machine learning unit includes a feature amount extraction unit, and the feature amount extraction unit determines at least one of the stirring marks, the brightness of the liquid surface, and the state of bubbles on the liquid surface as the feature amounts. The feature amount extraction unit may use an algorithm such as Maximally Stable Extermal Regions (MSER), Features from Accelerated Segment Test (FAST), Oriented-BRIEF (ORB), Scale-Invariant Feature Transform (SIFT), Speed-Up Robust Features (SURF), or Histograms of Oriented Gradients (HOG). As a modification of this embodiment, the machine learning unit may be set in advance to use at least one of the stirring marks, the brightness of the liquid surface, and the state of bubbles on the liquid surface as feature quantities.

[0056] The extracted feature quantities can consist of, for example, several to tens of thousands of parameters. Examples of the parameters related to the stirring marks include the length, number, and curvature of the ripples on the liquid surface of the reaction mixture, the speed of movement (axial and centrifugal directions), and the height of the protrusions. These parameters change depending on the progress of polymerization and can therefore serve as good feature quantities. The speed of movement can be calculated by comparing multiple images taken over a time series. The height of the protrusions can also be calculated by placing a camera at multiple angles or by the difference in brightness of the protrusions in the images. Examples of the parameters related to the brightness of the liquid surface include overall brightness, local brightness, and brightness distribution. During polycondensation, a considerable amount of nitrogen is injected while the reaction is carried out, and the water produced by the condensation evaporates and emerges as water vapor. As the viscosity of nitrogen and water vapor increases, bubbles become more difficult to remove, and they tend to turn white and become brighter, making them good feature quantities. In addition, parameters relating to the state of bubbles include the position where bubbles appear on the liquid surface, the size and shape of the bubbles, and how the bubbles burst, etc. These also change depending on the progress of polymerization, and therefore can be good feature quantities.

[0057] In the step of determining the degree of progress of the polymerization reaction, the analysis and determination preferably involve extracting and analyzing pixels containing stirring traces present on the liquid surface of the reaction mixture based on the acquired image, and determining the degree of progress of the polymerization reaction. That is, the machine learning unit is provided with an object extraction unit, and after the object extraction unit extracts the stirring traces as objects, the feature extraction unit can extract the aforementioned parameters related to the stirring traces. In this case, the target pixels can be narrowed down, thereby improving efficiency.

[0058] The machine learning unit preferably has a temperature correction function that performs corrections in analysis and / or judgment according to the temperature of the reaction mixture. Resins can easily change viscosity with temperature (this tendency is particularly pronounced for paint resins). Therefore, if the temperature of the reaction mixture is different at the time of imaging, the accuracy of analysis and judgment may decrease. Therefore, by having the machine learning unit have a temperature correction function, it is possible to perform more accurate analysis and judgment. Specifically, examples include multiplying a predetermined coefficient according to the temperature of the reaction mixture or adding a predetermined correction term.

[0059] In this embodiment, the polymerization reaction is then terminated based on the determined progress of the polymerization reaction (step S105). As an example, the acid value is used as an indicator of the progress of the polymerization reaction, and the polymerization reaction can be terminated when the acid value reaches 3. In addition to the acid value, other indicators of the progress of the polymerization reaction can include, for example, bubble viscosity, viscosity, solid content, hydroxyl value, epoxy equivalent, isocyanate equivalent, number average molecular weight, particle size, pH, amine value, color number, SP value, solvent tolerance, and turbidity.

[0060] According to the resin production method of the present invention, the progress of the polymerization reaction can be determined without time lag, without the need for workers to sample the paint resin during the polymerization reaction, adjust the temperature, measure properties such as bubble viscosity, etc., and therefore the end time of the polymerization process in which the resin reaction raw materials are polymerized can be controlled with greater precision than in cases where there is a time lag.In addition, the above-mentioned work by workers can be omitted, thereby reducing the amount of labor required.

[0061] (Resin manufacturing equipment) FIG. 3 is a schematic diagram of a resin production apparatus according to one embodiment of the present invention. As shown in FIG. 3, the resin manufacturing apparatus 1 of this embodiment includes a reaction vessel 2 in which the reaction raw materials are polymerized, an imaging unit 3 that captures images of at least a portion of the reaction mixture in the polymerization reaction of the reaction raw materials in a time-series manner to acquire images, and a machine learning unit 4 that has a machine learning function and analyzes the acquired images by a machine learning technique to determine the progress of the polymerization reaction.

[0062] The reaction vessel 2 may be a known reaction vessel for polymerizing reaction raw materials, but is configured to have a window or be open, for example, so that at least a part of the reaction mixture in the polymerization reaction of the reaction raw materials can be imaged by the imaging unit 3 from any direction (preferably directly above the liquid surface).

[0063] The imaging unit 3 and the machine learning unit 4 are the same as those already described in the embodiment of the resin manufacturing method, and therefore will not be described again. As in the embodiment of the resin manufacturing method, the machine learning unit preferably has a temperature correction function that performs correction according to the temperature of the reaction mixture during analysis and / or determination. The machine learning unit also preferably includes the feature extraction unit described above, the object extraction unit described above, and the memory unit and communication unit described above.

[0064] The resin manufacturing apparatus of this embodiment also makes it possible to determine the progress of the polymerization reaction without time lag, without the need for an operator to sample the paint resin during the polymerization reaction, adjust the temperature, measure properties such as bubble viscosity, etc., and therefore makes it possible to control the end time of the polymerization process in which the resin reaction raw materials are polymerized with greater precision than in cases where a time lag occurs.Furthermore, the above-mentioned tasks performed by the operator can be omitted, thereby reducing the amount of labor required.

[0065] Examples of the present invention will be described below, but the present invention is not limited to the following examples in any way. [Example]

[0066] Manufacturing example 1: Creating learning dataset 1 A 2-L Kolben reactor equipped with a stirrer, temperature controller, and condenser was charged with 23.6 parts neopentyl glycol, 13.9 parts 1,6-hexanediol, 2.8 parts pentaerythritol, and 15.9 parts ε-caprolactone, followed by 33.2 parts isophthalic acid and 19.7 parts phthalic anhydride, and the mixture was heated. The condensation reaction proceeded while removing the water produced during the reaction. The temperature was raised to 180°C over approximately 1 hour from the start of the temperature increase, and 0.2 parts dibutyltin oxide was added. The temperature was then gradually raised to 220°C, and stirring and dehydration were continued until the acid value equivalent to 4 (carboxylic acid) was reached, at which point the reaction was terminated. The reaction took 9 hours from the start of the temperature increase to the end of the reaction. Samples for acid value measurement were taken every hour from the start of the initial temperature increase until the end of the reaction, and the liquid surface of the reaction mixture was photographed from a direction approximately perpendicular to the liquid surface. Then, 42 parts of a mixed solvent of xylene, Solvesso 150, and propylene glycol methyl ether acetate was added. The obtained polyester resin (1) had a number average molecular weight of 6,600, a solid content hydroxyl value of 60 mgKOH / g, a solid content acid value of 2 mgKOH / g, and a solid content concentration of 70%. Furthermore, a learning dataset 1 was obtained in which the image of the liquid surface of the reaction mixture taken in a direction substantially perpendicular to the liquid surface was used as an explanatory variable and the acid value equivalent to carboxylic acid was used as a response variable.

[0067] Manufacturing example 2: Creating learning data 2 A 2-L Kolben reactor equipped with a stirrer, temperature controller, and condenser was charged with 15.6 parts isophthalic acid, 21.8 parts hexahydrophthalic anhydride, 15.2 parts neopentyl glycol, 18.0 parts trimethylolpropane, 6.1 parts hydroxypivalic acid neopentyl glycol ester, 7.9 parts Cardurer E10P (Versatic acid glycidyl ester, manufactured by Hexion), and 15.4 parts ε-caprolactone, and the temperature was increased. The condensation reaction was allowed to proceed while removing the water produced by the reaction. The temperature was raised to 190°C over approximately 2 hours from the start of the temperature increase, and stirring and dehydration were continued until the acid value equivalent to carboxylic acid reached 8, at which point the reaction was terminated. The time from the start of the temperature increase to the end of the reaction was 8 hours. Samples for acid value measurement were taken every hour from the start of the initial temperature increase until the end of the reaction, and the liquid surface of the reaction mixture was photographed from a direction approximately perpendicular to the liquid surface. Then, 16 parts of butyl acetate were added. The obtained polyester resin (2) had a number average molecular weight of 1,500, a solid content hydroxyl value of 230 mgKOH / g, a solid content acid value of 8 mgKOH / g, and a solid content concentration of 80%. Furthermore, a learning dataset 2 was obtained in which the image of the reaction mixture taken from a direction substantially perpendicular to the liquid surface was used as an explanatory variable and the acid value equivalent to carboxylic acid was used as a response variable.

[0068] Example 1: Preparation of polyester resin (1) The training data 1 created in Production Example 1 was input into a computer and subjected to machine learning using deep learning to generate a learning unit for polymerization reaction analysis and judgment. A 2L Kolben reactor equipped with a stirrer, temperature controller, and condenser was charged with 23.6 parts neopentyl glycol, 13.9 parts 1,6-hexanediol, 2.8 parts pentaerythritol, and 15.9 parts ε-caprolactone, followed by 33.2 parts isophthalic acid and 19.7 parts phthalic anhydride, and the mixture was heated. The condensation reaction proceeded while removing the water produced during the reaction. The temperature was raised to 180°C over approximately 1 hour, and 0.2 parts dibutyltin oxide was added. The temperature was then gradually raised to 220°C, with stirring and dehydration continuing. From the start of the initial temperature increase until the end of the reaction, images of the liquid surface of the reaction mixture were taken from a direction approximately perpendicular to the liquid surface every hour, and the image data were input into the polymerization reaction analysis and judgment learning unit. The image data taken 9 hours after the start of the temperature increase to the end of the reaction was input into the polymerization reaction analysis and judgment learning unit. The end point, a solid acid value of 2 mgKOH / g, was predicted, and the reaction was therefore terminated. Then, 42 parts of a mixed solvent of xylene, Solvesso 150, and propylene glycol methyl ether acetate was added. The polyester resin obtained had a number average molecular weight of 6,600, a solid content hydroxyl value of 60 mgKOH / g, a solid content acid value of 2 mgKOH / g, and a solid content concentration of 70%.

[0069] Example 2: Preparation of polyester resin (2) The training data 2 created in Production Example 2 was input into a computer and subjected to machine learning using a genetic algorithm to generate a learning unit for polymerization reaction analysis and judgment. A 2L Kolben reactor equipped with a stirrer, temperature controller, and condenser was charged with 15.6 parts of isophthalic acid, 21.8 parts of hexahydrophthalic anhydride, 15.2 parts of neopentyl glycol, 18.0 parts of trimethylolpropane, 6.1 parts of hydroxypivalic acid neopentyl glycol ester, 7.9 parts of Cardurer E10P (Versatic acid glycidyl ester, manufactured by Hexion), and 15.4 parts of ε-caprolactone, and the mixture was heated. The condensation reaction was allowed to proceed while removing the water produced by the reaction. The temperature was raised to 190°C over approximately 2 hours from the start of the temperature increase, and stirring and dehydration were continued. The time from the start of the temperature increase to the end of the reaction was 8 hours. Images of the liquid surface of the reaction mixture were taken from a direction approximately perpendicular to the liquid surface every hour from the start of the initial temperature increase until the end of the reaction. The image data taken 8 hours after the start of the temperature increase to the end of the reaction was input into the polymerization reaction analysis and judgment learning unit. The end point, a solid acid value of 8 mgKOH / g, was predicted, so the reaction was terminated. Then, 16 parts of butyl acetate were added. The obtained polyester resin (2) had a number average molecular weight of 1,500, a solid content hydroxyl value of 230 mgKOH / g, a solid content acid value of 8 mgKOH / g, and a solid content concentration of 80%.

[0070] Comparative Example 1 Hydrolysis reaction of ester compound Methyl benzoate and ion-exchanged water were charged into a 2 L Kolben reaction vessel equipped with a stirrer, a temperature controller, and a condenser, and then sodium hydroxide was added to carry out a hydrolysis reaction. This reaction was observed using the same procedures as in Production Example 1 and Example 1, with the liquid surface of the reaction mixture photographed from a direction approximately perpendicular to the liquid surface. However, no significant changes were observed in the photographed images, making it difficult to determine the end point of the process from the images. [Explanation of symbols]

[0071] 1: Resin manufacturing equipment, 2: reaction vessel, 3: imaging unit, 4: Machine Learning Department

Claims

1. A method for producing a resin, comprising a step of polymerizing reactant materials, The manufacturing method includes: a step of acquiring images by capturing time-series images of the liquid surface of the reaction mixture during the polymerization reaction of the reaction raw materials using an imaging unit; a step of analyzing the acquired image by a machine learning technique using a machine learning unit and determining the progress of the polymerization reaction; a step of terminating the polymerization reaction based on the determined progress of the polymerization reaction; Including, The method for producing a resin, wherein the resin is a paint resin.

2. The method for producing a resin according to claim 1, wherein the number average molecular weight of the resin after the polymerization reaction is in the range of 900 to 20,000.

3. 3. The method for producing a resin according to claim 1, wherein the molecular weight of the reaction raw materials contained in the reaction mixture is within a range of 30 to 400.

4. The method for producing a resin according to claim 1 , wherein the resin is a polyester resin for paint.

5. The method for producing a resin according to any one of claims 1 to 4, wherein the analysis and determination in the step of determining the progress of the polymerization reaction are performed based on at least one of stirring marks on a liquid surface of the reaction mixture, brightness of the liquid surface, and a state of bubbles on the liquid surface.

6. The analysis and determination in the step of determining the progress of the polymerization reaction include: The method for producing a resin according to any one of claims 1 to 5, wherein pixels including stirring traces present on the liquid surface of the reaction mixture are extracted and analyzed based on the image, and a degree of progress of the polymerization reaction is determined.

7. The analysis and determination in the step of determining the progress of the polymerization reaction include:

7. The method for producing a resin according to claim 1, wherein the analysis and judgment are performed by the machine learning unit that has been trained in advance using learning relationship data that shows a relationship between images of at least a portion of the reaction mixture in a polymerization reaction of the reaction raw materials in time series and the progress of the polymerization reaction.

8. The method for producing a resin according to any one of claims 1 to 7, wherein the machine learning unit has a temperature correction function that performs correction in the analysis and / or determination according to the temperature of the reaction mixture.

9. A resin manufacturing apparatus, a reaction vessel in which the reaction raw materials are polymerized; an imaging unit that captures images of the liquid surface of the reaction mixture in a time series during the polymerization reaction of the reaction raw materials; a machine learning unit having a machine learning function, which analyzes the acquired image by a machine learning method and determines the progress of the polymerization reaction; Equipped with 10. The resin manufacturing apparatus, wherein the resin is a paint resin.

Images