Base oil, lubricating oil composition, and method for lubricating sliding surfaces

Abstract

To provide base oil capable of improving the lubrication capacity of a lubricant composition.SOLUTION: The base oil according to the present invention includes one or more components selected from the group consisting of 6 to 20C paraffinic compounds, naphthenic compounds, and alkylbenzene compounds, each comprising a side chain, with the side chain of the alkyl group being branched in the paraffinic compound and in the naphthenic compound.SELECTED DRAWING: None

Description

[Technical Field]

[0001] The present invention relates to a base oil and a lubricating oil composition containing the same. [Background technology]

[0002] Industrial products that include industrial machinery such as internal combustion engines, hydraulic machines, compressors, turbines, gear elements, bearings, metalworking equipment, and refrigerators utilize various lubricating oil compositions (also called "lubricating oils"), such as engine oil, hydraulic oil, compressor oil, turbine oil, gear oil, metalworking oil, and refrigeration oil, to ensure smooth operation.

[0003] Lubricating oil compositions mainly consist of one or more base oils and additives. Examples of such lubricating oil compositions include those containing ester-based base oils and lubricating oil additives (see, for example, Patent Document 1). [Prior art documents] [Patent Documents]

[0004] [Patent Document 1] Japanese Patent Publication No. 2022-158124 [Overview of the project] [Problems that the invention aims to solve]

[0005] Here, lubricating additives are blended with the base oil in a ratio of, for example, a few percent to several tens of percent, and impart effects such as friction reduction or wear suppression to the base oil. In this sense, the performance of the lubricating oil composition can be said to be guaranteed by the additives. However, the effect of additives may differ depending on the type of base oil. When the base oil forms an oil film on the sliding parts, the density differs depending on the type of base oil. As a result, the additive may not be able to pass through the base oil and reach the lubricating parts, and the lubricating oil composition may not be able to exhibit functions such as friction reduction.

[0006] Therefore, in order for the lubricating oil composition to exhibit desired lubricating performance and the like, it is required to have a base oil that allows additives to reach the lubricated parts and can bring out the effects of the additives.

[0007] One aspect of the present invention is Forms an oil film with a sparse structure. to provide a base oil that can be obtained.

Means for Solving the Problems

[0008] One aspect of the present invention is comprising one or more components selected from the group consisting of paraffinic compounds, naphthenic compounds, and alkylbenzene compounds, having a side chain and having 6 to 20 carbon atoms, wherein the paraffinic compound and the naphthenic compound are base oils in which the alkyl group of the side chain is branched.

Effects of the Invention

[0009] One aspect of the base oil according to the present invention can bring out the effects of additives and improve the lubricating performance of the lubricating oil composition.

Brief Description of the Drawings

[0010] [Figure 1] It is a block diagram showing a schematic configuration of a learning device. [Figure 2] It is a diagram showing an example of a table in which structural formulas (SMILES) are described. [Figure 3] It is an explanatory diagram showing the difference in the adsorption state of additives due to the difference in the density of the oil film. [Figure 4] It is a diagram showing an example of a functional block diagram showing the configuration of a device for calculating feature amounts. [Figure 5] It is a diagram showing an example of creating a three-dimensional structure from the single-molecule structure of a target molecule. [Figure 6] It is a diagram showing an example of creating a liquid structure from the three-dimensional structure of a target molecule. [Figure 7] It is a diagram showing an example of the distance between two molecules. [Figure 8]It is a diagram showing an example of a radial distribution function. [Figure 9] It is a diagram showing an example of the radial distribution function of a liquid of a general normal alkane. [Figure 10] Among Fig. 9, it is a diagram showing an example of the radial distribution function focusing only on intermolecular interactions. [Figure 11] It is a functional block diagram showing the configuration of a prediction device. [Figure 12] It is a diagram showing an example of the accuracy of a predicted value predicted by a learned model. [Figure 13] It is a diagram showing an example of the accuracy of other predicted values predicted by a learned model. [Figure 14] It is a diagram showing an example of the relationship between predicted values of two feature quantities. [Figure 15] It is a block diagram showing the hardware configuration of a prediction device. <00001​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​Embodiments of the present invention will be described in detail below. In this specification, unless otherwise specified, the "~" indicating a numerical range means that the numbers before and after it are included as the lower and upper limits. Furthermore, if only the upper limit of a numerical range represented by "~" has a unit specified, it means that the lower limit also has the same unit.

[0012] <Base oil> The base oil according to this embodiment will now be described. The base oil according to this embodiment contains one or more components selected from the group consisting of paraffinic compounds, naphthenic compounds, and alkylbenzene compounds, which have side chains and have 6 to 20 carbon atoms.

[0013] Paraffin compounds have a branched structure in which the alkyl group of the side chain is located. Paraffin compounds are represented by formula (I): (CH3)n1-R1(n1 is 4~ 7 The integer R1 is a straight-chain saturated hydrocarbon with 4 to 9 carbon atoms. or 3,4-diethylhexane The following can be used: That is, the paraffinic compound can be a compound having 4 to 6 methyl groups and 8 to 15 carbon atoms.

[0014] Specifically, paraffinic compounds include 2,2,4,6,6-pentamethylheptane, 2,2,3,3-tetramethylbutane, 2,2,4,4,6,8,8-heptamethylnonane, and 3,4 - Diethylhexane and the like can be used. These may be used individually or in combination of two or more.

[0015] Naphthenic compounds have a branched structure in which the alkyl group of the side chain is located. Naphthenic compounds can be compounds that have a naphthenic ring and have 8 to 16 carbon atoms.

[0016] Specifically, naphthenic compounds include tert-butylcyclohexane, tricyclo[5.2.1.02,6]decane-3-ene, tricyclo[6.2.1.02,7]undeca-4-ene, tricyclo[5.2.1.01,5]decane-8-ene, (1R,4E,9S)-4,11,11-trimethyl-8-methylidenebicyclo[7.2.0]undeca-4-ene, 1,2,3,3a,4,5,5a,6,7,8,8a,9,10,10a,10b,10c-hexadecahydropyrene, 1,4-di(propan-2-yl)cyclohexane, 1,3-di( (Propan-2-yl)cyclohexane, 1,3,5-tri(propan-2-yl)cyclohexane, tetracyclo[5.3.0.02,6.03,10]decane, tetracyclo[5.4.0.01,3.02,7]undecane, (1R,2S,7R,8R,9S,10R)-2,6,6,9-tetramethyltetracyclo[5.4.0.02,9.08,10]undecane, 1-cyclohexyladamantane, and (2S,6S,10R,14R)-2,9-dimethyltetracyclo[6.6.0.02,6.010,14]tetradecane can be used. These may be used individually or in combination of two or more.

[0017] Alkylbenzene compounds can be compounds represented by formula (II): (CH3)3-C-Ar-C-(CH3)3 (where Ar is an aryl group). In other words, alkylbenzene compounds can be compounds having a tert-butyl group in the alkyl group.

[0018] Specifically, alkylbenzene compounds such as 1,4-di-tert-butylbenzene and 1,3-di-tert-butylbenzene can be used. These may be used individually or in combination of two or more.

[0019] The base oil according to this embodiment can be selected using a prediction device, which will be described later. The prediction device uses feature quantities that indicate the gaps between molecules constituting the base oil to predict the degree of density of the oil film containing the base oil, and can therefore predict base oils that form an oil film with a sparse structure. For this reason, the prediction device can be used to select a base oil that can bring out the effects of the additives contained in the lubricating oil composition and improve the lubrication performance of the lubricating oil composition. Even if the lubricating oil composition containing this selected base oil comes into contact with a solid surface and an oil film made of the base oil is formed on the surface, the additives contained in the lubricating oil composition can pass through the oil film and reach the solid surface. Therefore, the lubricating oil composition containing the selected base oil can bring out the effects of the additives and can exhibit functions such as friction reduction. Accordingly, the base oil according to this embodiment can be used to manufacture a lubricating oil composition with high lubrication performance that can bring out the effects of the additives.

[0020] In this embodiment, the solid is composed of metals such as aluminum and alloys, and the surface of the solid refers to the contact surface that comes into contact with the base oil.

[0021] In this embodiment, contact includes the base oil reaching the surface of a solid and the base oil penetrating the solid.

[0022] <Lubricating oil composition> The lubricating oil composition according to this embodiment comprises a base oil according to this embodiment, preferably comprises the base oil according to this embodiment and additives, more preferably substantially composed of the base oil according to this embodiment and additives, and even more preferably composed of the base oil according to this embodiment and additives. "Substantially" means that, in addition to the base oil and additives, it may also contain unavoidable impurities that may be inevitably present during the manufacturing process.

[0023] Since the lubricating oil composition according to this embodiment contains a base oil according to this embodiment, when the lubricating oil composition according to this embodiment comes into contact with a solid surface, the base oil in the oil film formed on that surface becomes sparse. Therefore, when the lubricating oil composition according to this embodiment comes into contact with a solid surface and an oil film made of the lubricating oil composition according to this embodiment is formed on that surface, the additives contained in the oil film of the lubricating oil composition can easily pass through the oil film, which has a sparse base oil, and reach the solid surface, thereby bringing out the effect of the additives. Accordingly, the lubricating oil composition according to this embodiment can be suitably used as a supplied lubricating oil to lubricate contact surfaces such as sliding surfaces of lubrication parts in industrial machinery and metalworking.

[0024] <Learning device> To explain the prediction device, we will first describe the learning device that generates the pre-trained model used in the prediction device. The learning device uses features such as the molecular structure of the target molecule as explanatory variables and features indicating the intermolecular gaps of the target molecule within a simulation cell having an arbitrary parallelepiped shape as the objective variable to perform machine learning and generate a pre-trained model that predicts the intermolecular gap features of the target molecule from the target molecule.

[0025] In this embodiment, the target molecule is measured at room temperature (20°C) and atmospheric pressure (1.013 × 10⁻⁶). 5 These are molecules that make up the base oil, which is a liquid in Pa.

[0026] Base oil is an oil used as a base material for lubricating oil compositions or greases, and it has the function of uniformly and stably dissolving additives and delivering them to the parts where lubrication is needed. Base oils are mainly classified into mineral oil obtained by petroleum refining, synthetic oil obtained by chemical synthesis, and mixtures thereof. Lubricating oil compositions and greases are mixtures containing one or more types of base oil and one or more types of additives. Examples of lubricating oil compositions include engine oil, hydraulic oil, compressor oil, turbine oil, gear oil, metalworking oil, and refrigeration oil, which are used in industrial machinery such as internal combustion engines, hydraulic machines, compressors, turbines, gear elements, bearings, metalworking machines, and refrigerators. Additives include plasticizers, stabilizers, oiliness agents, friction modifiers, anti-wear agents, antioxidants, UV absorbers, lubricants, mold release agents, anti-static agents, rust inhibitors, defoamers, and viscosity index improvers.

[0027] Figure 1 is a block diagram illustrating the schematic configuration of the learning device. As shown in Figure 1, the learning device 1 comprises a first acquisition unit 11, a second acquisition unit 12, a training dataset creation unit 13, a learning unit 14, and an output unit 15. The learning device 1 generates a trained model M1 that predicts feature quantities indicating the intermolecular gaps of a target molecule within a simulation cell having an arbitrary parallelepiped shape.

[0028] The first acquisition unit 11 acquires information about the target molecule, which is a molecule that makes up the base oil, as explanatory variables.

[0029] Information about the target molecule may be obtained from the first storage unit 21.

[0030] The first storage unit 21 stores a data table containing information such as the lubricating oil composition.

[0031] Information about the lubricating oil composition includes the type of lubricating oil composition, and information about the base oils and additives that make up the lubricating oil composition.

[0032] Lubricating oil compositions include engine oil, hydraulic oil, gear oil, and refrigeration oil.

[0033] Information regarding the base oil and additives includes the type of base oil and additives, the components and structural formulas of the base oil and additives, and the ratio of the base oil and additives.

[0034] Base oils include mineral oil, synthetic oil, animal and vegetable oil, and mixtures thereof.

[0035] The types of additives include plasticizers, stabilizers, lubricity agents, friction modifiers, anti-wear agents, metal-based detergents, antioxidants, friction-resistant and wear-resistant agents, extreme pressure agents, UV absorbers, lubricants, mold release agents, ashless dispersants, lubricity improvers, anti-static agents, rust inhibitors, defoamers, viscosity index improvers, metal deactivators, and solid lubricants.

[0036] Examples of components that make up a base oil include the components that make up oils commonly used as base oils, such as mineral oil, synthetic oil, animal and vegetable oil, and mixtures thereof.

[0037] Examples of mineral oils include paraffinic crude oil, naphthenic crude oil, intermediate crude oil, aromatic crude oil, distillates, and refined oils. Examples of paraffinic crude oils include isoparaffins and paraffins. Examples of naphthenic crude oils include naphthenes.

[0038] Examples of synthetic oils include poly-α-olefins, polyisobutylene (polybutene), monoesters, diesters, polyol esters, silicate esters, polyalkylene glycols, polyphenyl ethers, silicones, fluorine compounds, alkylbenzenes, and GTL base oils.

[0039] Examples of animal and vegetable oils include vegetable oils such as castor oil, olive oil, cocoa butter, sesame oil, rice bran oil, safflower oil, soybean oil, camellia oil, corn oil, rapeseed oil, palm oil, palm kernel oil, sunflower oil, cottonseed oil, and coconut oil, as well as animal fats such as beef tallow, lard, milk fat, fish oil, and whale oil.

[0040] Examples of components that make up the additive include ester compounds of monohydric or polyhydric aliphatic carboxylic acids and monohydric or polyhydric aliphatic alcohols.

[0041] Examples of monovalent aliphatic carboxylic acids used in the synthesis of ester compounds include methaneic acid, acetic acid, propionic acid, butyric acid, pentanoic acid, caproic acid, heptanoic acid, octanoic acid, nonanoic acid, decanoic acid, undecanoic acid, dodecanoic acid, tridecanoic acid, tetradecanoic acid, pentadecanoic acid, hexadecanoic acid, heptadecanoic acid, octadecanoic acid, nonadecanoic acid, eicosanic acid, henicosanoic acid, docosanic acid, tricosanic acid, tetracosanic acid, pentacosanoic acid, hexacosanoic acid, heptacosanoic acid, octacosanic acid, nonacosanoic acid, and triacontanoic acid. Examples of saturated aliphatic carboxylic acids include unsaturated aliphatic carboxylic acids such as acrylic acid, propiolic acid, methacrylic acid, pentenoic acid, hexenoic acid, heptenic acid, octenic acid, nonenic acid, decenoic acid, undecenoic acid, dodecenoic acid, tridecenoic acid, tetradecenoic acid, pentadecenoic acid, hexadecenoic acid, heptadecenoic acid, oleic acid, nonadecenoic acid, eicosenoic acid, henicosenoic acid, docosenoic acid, tricosenoic acid, tetracosenoic acid, pentacosenoic acid, hexacosenoic acid, heptacosenoic acid, octacosenoic acid, nonacosenoic acid, and triaconthenic acid.

[0042] Examples of polyvalent aliphatic carboxylic acids used in the synthesis of ester compounds include saturated aliphatic carboxylic acids such as oxalic acid, malonic acid, succinic acid, glutaric acid, adipic acid, pimelic acid, suberic acid, azelaic acid, sebacic acid, undecanediic acid, dodecanediic acid, tridecanediic acid, tetradecanediic acid, heptadecanediic acid, and hexadecanedioic acid; and unsaturated aliphatic carboxylic acids such as hexenodioic acid, heptenodioic acid, octenodioic acid, nonenodioic acid, decenodioic acid, undecenodioic acid, dodecenodioic acid, tridecenodioic acid, tetradecenodioic acid, heptadecenodioic acid, and hexadecenodioic acid.

[0043] Examples of monohydric aliphatic alcohols used in the synthesis of ester compounds include methanol, ethanol, propanol, butanol, pentanol, hexanol, heptanol, octanol, nonanol, decanol, undecanol, dodecanol, tridecanol, tetradecanol, pentadecanol, and hexadecanol.

[0044] Examples of polyhydric aliphatic alcohols used in the synthesis of ester compounds include ethylene glycol, propylene glycol, neopentyl glycol, glycerin, trimethylolethane, trimethylolpropane, pentaerythritol, and sorbitan.

[0045] The first memory unit 21 may use, for example, RDKit included in libraries such as Anaconda®, software distributed by Anaconda, Inc. in the United States. If the structural formula is SMILES, the string "SMILES" is read into the first acquisition unit 11 using MolFromSmiles included in RDKit, and the molecular structural formula is read.

[0046] The information on the target molecule that the first acquisition unit 11 acquires from the first storage unit 21 may include the name and structural formula of the target molecule. The structural formula may be SMILES, which is a string representation of the molecular structure of the target molecule. An example of a table containing structural formulas (SMILES) is shown in Figure 2. As shown in Figure 2, the SMILES of each target molecule are displayed. The table containing the structural formula of each target molecule may be obtained from data in formats such as CSV or spreadsheet software such as Excel. The first acquisition unit 11 may input a table containing the SMILES of each target molecule as shown in Figure 2.

[0047] SMILES can be obtained from chemical databases, such as PubChem (a database of chemical substances provided by the NCBI in the United States).

[0048] The second acquisition unit 12 acquires feature quantities that indicate the intermolecular gaps of the target molecule within a simulation cell having an arbitrary shape such as a parallelepiped, using these as the target variable.

[0049] The feature quantities indicating the intermolecular gaps of the target molecule may be obtained from the second memory unit 22.

[0050] The second memory unit 22 may store a database showing the relationship between the target molecule and feature quantities that indicate the intermolecular gaps of the target molecule within a simulation cell of any shape.

[0051] The feature quantities indicating the intermolecular gaps of the target molecules are values ​​calculated based on the radial distribution function, which represents the relationship between the interatomic distance between different molecules and the relative abundance of other molecules to one molecule, in a liquid-optimized structure obtained by stabilizing a liquid structure created from the three-dimensional structures of any number of target molecules within a simulation cell. As described later, the feature quantities indicating the intermolecular gaps of the target molecules may be the midpoint or slope of the approximate formula of the radial distribution function, the first or second peak of the radial distribution function, or the value on the horizontal axis when the derivative of the radial distribution function asymptotically approaches zero (see Figure 7). The midpoint and slope of the approximate formula of the radial distribution function correlate with the interatomic distance between different molecules and can be said to correlate with the degree of density of the oil film formed using a base oil composed of the target molecules.

[0052] As shown in Figure 3, in sliding parts of industrial machinery and metalworking, the denser the oil film formed by the base oil composed of the target molecules on the sliding surface (hereinafter simply referred to as the "sliding surface of the sliding part") that rubs against the surface of other sliding parts, the denser the base oil insideAdditives have difficulty passing through the interior of the oil film and reaching the sliding surface, so their effects tend to be less pronounced (see Figure 3(a)). On the other hand, the more sparse the oil film, the more gaps are likely to form in the oil film, and the easier it is for additives to pass through these gaps. As a result, additives can migrate within the oil film and reach the sliding surface (see Figure 3(b)). Therefore, the effects of additives, such as friction reduction, tend to be more pronounced. Thus, if the characteristic quantities of the target molecules constituting the base oil are correlated not with the interatomic distances within a molecule, but with the interatomic distances between different molecules (for example, the slope and midpoint of the approximate formula of the radial distribution function), the degree of density of the oil film can be predicted from the magnitude of the characteristic quantities indicating the intermolecular gaps of the target molecules.

[0053] The characteristic quantities indicating the intermolecular gaps of the target molecule can be calculated by performing molecular simulations such as molecular dynamics using the optimal liquid structure of the single-molecule structure of the target molecule constituting the base oil.

[0054] The feature quantities indicating the intermolecular gaps of the target molecule may be values ​​calculated from, for example, a molecular feature quantity calculation device.

[0055] The molecular feature calculation device is not particularly limited as long as it can calculate feature quantities indicating the intermolecular gaps of the target molecule by performing molecular simulations using the single-molecule structure of the target molecule constituting the base oil. An example of a functional block diagram showing the configuration of the molecular feature calculation device is shown in Figure 4. As shown in Figure 4, the molecular feature calculation device 30 may have an acquisition unit 31, a single-molecule information acquisition unit 32, a liquid structure creation unit 33, an optimized structure acquisition unit 34, a feature quantity calculation unit 35, a prediction unit 36, and an output unit 37.

[0056] The acquisition unit 31 acquires information about the target molecules, which are molecules that make up the base oil, from the storage unit 38, as well as information about additives.

[0057] The memory unit 38 stores a data table containing information on the target molecule and additives. The information on the target molecule and additives may be the same as the information stored in the first memory unit 21 and the second memory unit 22. Since the information on the target molecule and additives is the same as the information stored in the first memory unit 21 and the second memory unit 22, details are omitted.

[0058] For information about the target molecule, you may use its name and structural formula. For the structural formula, you may use SMILES, etc. Since SMILES is the same as described above, details will be omitted.

[0059] The single-molecule information acquisition unit 32 acquires the three-dimensional structure of the target molecule acquired by the acquisition unit 31 as single-molecule information. For example, as shown in Figure 5, the single-molecule information acquisition unit 32 adds hydrogen to the single-molecule structure of the target molecule to create a three-dimensional structure of the target molecule with hydrogen added to the single-molecule structure, and acquires it as single-molecule information.

[0060] The single-molecule information acquisition unit 32 may acquire an optimized structure, in which the three-dimensional structure has been optimized, as single-molecule information. The single-molecule information acquisition unit 32 may consider the relationship between the coordinates and energies of the atoms constituting the target molecule, arrange them in appropriate positions, and find a structure in which the three-dimensional structure is energetically most stable. The single-molecule information acquisition unit 32 may create the optimized structure using general molecular simulation methods such as first-principles calculations or molecular mechanics calculations.

[0061] The liquid structure creation unit 33 creates a liquid structure by arranging an arbitrary number (N) of single-molecule information obtained by the single-molecule information acquisition unit 32 inside the simulation cell, as shown in Figure 6. The number of single-molecule information is not particularly limited and can be any number as appropriate, but it may be several tens to several hundred. The simulation cell may be defined to have any shape, such as a parallelepiped. The size of the simulation cell may be defined to have any size as appropriate, depending on the type of lubricating oil composition or target molecule. The simulation cell may be defined so that, for example, its density is approximately the same as that of the actual base oil, so as to reproduce the actual structure of the base oil.

[0062] The liquid structure creation unit 33 places any number (N) of single-molecule information within the simulation cell at different positions so that these N single-molecule information do not overlap. The liquid structure creation unit 33 may create a liquid structure by randomly placing any number (N) of single-molecule information in the simulation cell and rearranging them.

[0063] As shown in Figure 4, the optimized structure acquisition unit 34 may acquire an optimized liquid structure by relaxing the liquid structure created by the liquid structure creation unit 33. Relaxation refers to a general structural optimization method such as the steepest descent method or the conjugate gradient method for finding the minimum energy value in multi-space. For example, relaxation means adjusting the sum or scalar of force vectors acting on the entire liquid structure, or the stress tensor or principal components of the stress tensor, to match a predetermined pressure (external pressure) described later, under a certain threshold. The liquid structure created by the liquid structure creation unit 33 has single-molecule information stored in a simulation cell of an appropriate size, so it may not have the correct density that matches the actual state. In this state, even if the feature calculation unit 35 calculates features, the accuracy of the calculated features may be low. The optimized structure acquisition unit 34 can create a liquid optimized structure that has been adjusted to have the correct density by optimizing the liquid structure.

[0064] The optimized structure acquisition unit 34 may create a liquid optimized structure under the following two conditions when optimizing the liquid structure.

[0065] (1) The volume of the simulation cell is variable. An external force is applied to the simulation cell, and the liquid structure may be subjected to a predetermined pressure (external pressure).

[0066] (2) The optimized structure acquisition unit 34 may use (1) above as the structure at absolute zero. Molecular dynamics simulations may be performed to consider the effect of temperature. Examples of molecular dynamics simulations include structural optimization (molecular mechanics method), molecular dynamics method, Monte Carlo method, etc. The volume of the simulation cell may be variable. The temperature may be set to any appropriate value. An external force may be applied to the simulation cell, and the liquid structure may be subjected to a predetermined pressure (external pressure).

[0067] The liquid optimization structure is created under the two conditions described above. This defines the simulation cell to reproduce the density of the actual base oil. In other words, the simulation cell is defined to have the density when the liquid structure is relaxed, taking into account the given pressure and temperature.

[0068] When simulating a lubricating oil composition in contact with the device, the temperature and pressure may be set to high and high temperatures, respectively. For example, the high temperature may be 50°C to 200°C, specifically around 60°C. For example, the high pressure may be 50 MPa to 2000 MPa, specifically around 500 MPa.

[0069] The optimized structure acquisition unit 34 may create a liquid optimized structure using general molecular simulations such as first-principles calculations or molecular mechanics calculations.

[0070] The feature calculation unit 35 calculates the interatomic distances between different molecules from the liquid optimization structure and calculates a radial distribution function (RDF) that represents the relative abundance of other molecules relative to one molecule at each interatomic distance, thereby calculating a feature that indicates the intermolecular gap of the target molecule. The feature calculation unit 35 may use a feature represented by a radial distribution function that is based on the interatomic distances between different molecules, rather than the interatomic distances within a single molecule (for example, the slope and midpoint of the approximate formula of the radial distribution function, as described later).

[0071] The interatomic distance between different molecules is the distance between atoms contained in two different molecules. Let one molecule be molecule A, and another molecule different from molecule A be molecule B. In this case, the interatomic distance between different molecules may be the distance between a certain atom a in molecule A and an atom in molecule B that is in the same position as atom a, or it may be the distance between a certain atom a in molecule A and an atom in molecule B that is in the same position as an atom other than atom a.

[0072] The feature calculation unit 35 calculates the interatomic distances between different molecules from the liquid optimized structure obtained by the optimized structure acquisition unit 34, and calculates a radial distribution function that represents the relative abundance of one molecule to another at each interatomic distance. The radial distribution function may be calculated, for example, from the following equation (1).

[0073]

number

[0074] Figure 7 shows an example of a radial distribution function that illustrates the relationship between an atom a in molecule A and an atom b in another molecule B. As shown in Figure 7, as the distance from an atom a in molecule A to an atom b in molecule B increases, the probability density of molecule B's existence rises from a constant state of 0 to 1, and in the interval shown on the horizontal axis in Figure 7, a radial distribution function is obtained in which the value on the vertical axis is approximately constant at around 0.95.

[0075] As shown in Figure 8, the radial distribution function represents the probability that an atom b of another molecule B exists at a certain distance r from an atom a of molecule A. In other words, the radial distribution function represents the frequency with which an atom b of another molecule B is located at a distance r from an atom a of molecule A, and correlates with the intermolecular distance gap between an atom a of molecule A and an atom b of molecule B. Note that the radial distribution function can be plotted in multiple dimensions in addition to two dimensions.

[0076] The feature calculation unit 35 may convert the calculated radial distribution function into an approximate formula. The feature calculation unit 35 may fit the radial distribution function to the logistic function shown in equation (2) below to obtain an approximate formula for the radial distribution function. y = a / (1 + e) -b(x-x 0 ) )+c ···(2) (In the formula, y is the probability density of molecule B's existence, a and c are coefficients, b corresponds to the slope of the approximation formula, x is the distance (in Å) from an atom in molecule A to an atom in molecule B, and x0 is the position of the midpoint of the slope of the approximation formula.)

[0077] The feature calculation unit 35 may calculate a feature representing the intermolecular gap of the target molecule by extracting the parameters of the approximate formula of the transformed radial distribution function. The parameters may be the midpoint or the slope of the approximate formula of the radial distribution function. That is, when the approximate formula of the radial distribution function is obtained using the logistic function shown in equation (2) above, the logistic function shown in equation (2) above uses the position of the midpoint of the approximate formula of the radial distribution function as parameter x0 and the slope of the approximate formula as parameter b. Therefore, as shown in Figure 7, the midpoint or the slope of the approximate formula of the radial distribution function can be used as a parameter of a feature representing the intermolecular gap of the target molecule, and it can be said that the midpoint or the slope of the approximate formula of the radial distribution function correlates with the interatomic distance between different molecules and correlates with the degree of density of the oil film.

[0078] The midpoint and slope of the approximation formula correspond to the distance between molecules. A larger midpoint or a smaller slope indicates that the molecules are farther apart, forming a relatively sparse oil film, and thus shortening the adsorption time for the additive to the surface. Conversely, a smaller midpoint or a larger slope indicates that the molecules are closer together, forming a relatively dense oil film, and thus lengthening the adsorption time for the additive to the surface. Therefore, a larger midpoint or a smaller slope in the approximation formula results in a shorter adsorption time for the additive. If the additive can be adsorbed onto the sliding surface of a sliding part of industrial machinery in a shorter time, the effect of the additive can be exerted. If it can be adsorbed onto the sliding surface of a sliding part in a shorter time, the effect of the additive can be exerted, protecting the surface of the object from severe friction and resulting in effects such as reduced friction and reduced wear. For this reason, it is preferable to adjust the approximation formula so that the midpoint is larger or the slope is smaller, thereby increasing the gaps between molecules and adjusting the oil film to have a sparse structure.

[0079] Furthermore, the feature calculation unit 35 may use the first peak (the first local maximum that appears when viewed from the shorter distance to the longer distance), the second peak (similarly the second local maximum) of the radial distribution function, or the value on the horizontal axis when the derivative of the radial distribution function asymptotically approaches zero, as shown in Figure 7, as a feature representing the intermolecular gap of the target molecule.

[0080] The feature calculation unit 35 uses features that correlate with the interatomic distance between different molecules (for example, the slope and midpoint of the approximate equation of the radial distribution function) as features that indicate the gaps between molecules of the target molecules constituting the base oil. By obtaining a radial distribution function that focuses only on the interatomic distance between molecules, the feature calculation unit 35 can identify the gaps between molecules because completely different features appear depending on the type of molecule. For example, in the case of the radial distribution function of a typical normal alkane liquid, as shown in Figure 9, there are many CH bonds and CC bonds within a single molecule, so in the region where the distance between atoms within a single molecule is short, there are sharp peaks originating from CH bonds and CC bonds. On the other hand, in the region where the distance between molecules is long, there are peaks related to intermolecular pairs, but these are relatively weak compared to the interatomic distance within a single molecule and are almost undetectable. By removing the interatomic distances within a single molecule, such as CH bonds and CC bonds, a radial distribution function is obtained in which only the distances between atoms of different molecules are shown, as shown in Figure 10. The radial distribution function, which represents only the distances between atoms of different molecules, such as isoparaffins, paraffins, and naphthenes, exhibits completely different shapes depending on the type of molecule, making it effective for identifying the gaps between atoms of different molecules.

[0081] Furthermore, the feature calculation unit 35 may use various physical properties of the base oil, such as viscosity, flash point, diffusion coefficient, and thermal conductivity, in addition to the midpoint or slope of the approximate formula of the radial distribution function, as feature quantities that indicate the intermolecular gaps of the target molecule.

[0082] As shown in Figure 4, the prediction unit 36 ​​predicts the degree of density of the oil film formed by the target molecules based on the magnitude of feature quantities indicating the intermolecular gaps of the target molecules, such as the midpoint and slope extracted from the approximate formula of the radial distribution function indicating the intermolecular gaps calculated by the feature quantity calculation unit 35. Specifically, the prediction unit 36 ​​predicts the degree of density of the oil film formed by the target molecules by measuring whether the feature quantities indicating the intermolecular gaps of the target molecules, calculated by the feature quantity calculation unit 35, are large or small, using the feature quantities of the base oil currently in use as a reference.

[0083] For example, if the feature quantity indicating the intermolecular gap of the target molecule is the midpoint extracted from the approximation formula of the radial distribution function, and the feature quantity calculated by the feature quantity calculation unit 35 is larger than the feature quantity of the base oil currently in use (reference feature quantity), the prediction unit 36 ​​predicts that the oil film is sparse and predicts the degree of sparseness of the oil film according to the magnitude of that feature quantity. On the other hand, if the feature quantity calculated by the feature quantity calculation unit 35 is smaller than the feature quantity of the base oil currently in use (reference feature quantity), the prediction unit 36 ​​predicts that the oil film is dense and predicts the degree of density of the oil film according to the smallness of that feature quantity.

[0084] The output unit 37 outputs the predicted results of the density of the oil film formed inside the simulation cell using the base oil composed of the target molecules, as predicted by the prediction unit 36, by displaying and transmitting them.

[0085] As shown in Figure 1, the training dataset creation unit 13 extracts the target molecule as the explanatory variable and features indicating the intermolecular gaps of the target molecule within a simulation cell having an arbitrary parallelepiped shape as the objective variable, and adds them to the training dataset. The training dataset creation unit 13 then links the input target molecule with the features indicating the intermolecular gaps of the target molecule to create the training dataset.

[0086] The learning unit 14 generates a trained model M1 by training with a training dataset in which target molecules (explanatory variables) are associated with features (target variables) that indicate the intermolecular gaps of the target molecules within a simulation cell having an arbitrary parallelepiped shape.

[0087] The pre-trained model M1 is a pre-trained model that has undergone machine learning using a training dataset (training data table) stored in a memory unit (not shown). The learning results of the correspondence between the target molecule (explanatory variable) and the feature quantity (target variable) that indicates the intermolecular gap of the target molecule in a simulation cell having an arbitrary parallelepiped shape are applied. The pre-trained model M1 is a program that models the input-output relationship between the target molecule and the feature quantity indicating the intermolecular gap of the target molecule, using the target molecule as the explanatory variable as input data and the feature quantity indicating the intermolecular gap of the target molecule as the target variable as output data. The pre-trained model M1 may also be expressed as a mathematical formula such as a function.

[0088] The trained model M1 preferably uses a supervised learning algorithm within the field of machine learning. Examples of supervised learning algorithms include linear regression, regularized regression, partial least squares regression, polynomial regression, kernel regression, logistic regression, random forest, gradient boosting regression tree, support vector machine (SVM), and neural network. For neural networks, deep learning with more than three layers can be used. Examples of neural network types include convolutional neural networks (CNN), recurrent neural networks (RNN), and general regression neural networks. Among these, gradient boosting regression trees are preferred.

[0089] The output unit 15 displays information about the training dataset used in training the trained model M1, as well as information about the trained model M1.

[0090] Thus, since the learning device 1 is equipped with a learning unit 14, it can generate a trained model M1 that predicts feature quantities indicating the intermolecular gaps of the target molecule. The trained model M1 generated by the learning device 1 predicts feature quantities indicating the intermolecular gaps of the target molecule from the input information of the target molecule.

[0091] The feature quantities indicating the intermolecular gaps of the target molecule correlate with the interatomic distance between different molecules and with the degree of density of the oil film formed using a base oil composed of the target molecule. The more sparse the oil film, the easier it is for additives to pass through the gaps in the oil film and reach the sliding surfaces of sliding parts in industrial machinery, thus making it easier for the additive's effects, such as friction reduction, to be exerted. Therefore, the learning device 1 is used to predict the degree of density of the oil film formed using a base oil composed of the input target molecule by generating a trained model M1, thereby shortening the time required to select a base oil that will bring out the effect of the additive.

[0092] For example, when predicting the degree of density of an oil film formed using a base oil composed of target molecules, which is commonly used in the past, it takes a long time (e.g., several days to several months) to calculate the features from the radial distribution function even for just one type of base oil, so predicting the degree of density of the oil film takes even longer. Using the learning device 1, the trained model M1 can calculate the features for one type of base oil in a short time, for example, a few seconds to a few minutes, so the degree of density of the oil film can be predicted more quickly and easily.

[0093] Furthermore, since the learning device 1 can generate a trained model M1, using the trained model M1 can reduce the burden and time required to predict feature quantities indicating the intermolecular gaps of the target molecule from the information of the target molecule. When manufacturing lubricating oil compositions, in order to produce a lubricating oil composition that satisfies the desired performance according to the type and application of the lubricating oil composition, various combinations of base oils and additives are actually manufactured in experiments and their degrees of density are verified. This process determines the type of base oil that brings out the effect of the additive and the composition of the base oil and additive, which requires a great deal of effort and incurs a heavy cost due to the preparation of various base oils and additives. Since the learning device 1 can predict feature quantities indicating the intermolecular gaps of the target molecule from the target molecule, it can efficiently predict feature quantities indicating the intermolecular gaps of the target molecule while reducing the burden. Therefore, the learning device 1 can reduce the burden of predicting the degree of density of an oil film formed using a base oil composed of the target molecules by using feature quantities that indicate the intermolecular gaps of the target molecules to predict the degree of density of the oil film composed of the target molecules.

[0094] The learning device 1 preferably uses as its features values ​​values ​​calculated based on a radial distribution function that represents the relationship between the interatomic distances between different molecules and the relative abundance of other molecules to one molecule, obtained by stabilizing a liquid structure created from the three-dimensional structures of any number of target molecules within a simulation cell. This allows the learning device 1 to predict features indicating the intermolecular gaps of the target molecules with greater accuracy. Thus, the learning device 1 can improve the accuracy of selecting a base oil that brings out the effect of the additive.

[0095] <Prediction device> Let me explain the prediction device. The prediction device predicts the degree of density distribution of the oil film containing the target molecule.

[0096] Figure 11 is a system configuration diagram showing the configuration of the prediction device. As shown in Figure 11, the prediction device 4 comprises an acquisition unit 41, a trained model M2, a prediction unit 42, and an output unit 43, and predicts feature quantities that indicate the intermolecular gaps of target molecules as an index representing the degree of density of the oil film containing the target molecules.

[0097] The acquisition unit 41 acquires target molecules, which are molecules that make up the base oil, as explanatory variables. Information on target molecules may be acquired from the storage unit 44. The storage unit 44 stores a data table that includes information on lubricating oil composition, similar to the first storage unit 21 described above. Since the information on target molecules is the same as the information on target molecules acquired from the first storage unit 21 by the learning device 1 described above, a detailed explanation is omitted.

[0098] The trained model M2 is trained using a pre-prepared training dataset in which target molecules (explanatory variables) are associated with features (target variables) that represent the intermolecular gaps of the target molecules within a simulation cell having an arbitrary parallelepiped shape. The trained model M2 can use the trained model M1 generated by the training device 1 described above.

[0099] The prediction unit 42 uses the learned model M2 to predict the degree of density of the oil film formed by the target molecule by inputting the target molecule to be predicted acquired by the acquisition unit 41. That is, the prediction unit 42 inputs the target molecule to be predicted acquired by the acquisition unit 41 into the learned model M2, and outputs, as the objective variable, the feature amount indicating the intermolecular gap of the target molecule to be predicted predicted by the learned model M2. The prediction unit 42 outputs the objective variable as an index representing the degree of density of the oil film formed by the base oil composed of the target molecule to be predicted. As described above, the feature amount indicating the intermolecular gap of the target molecule correlates with the interatomic distance between different molecules, and correlates with the degree of density of the oil film formed using the base oil composed of the target molecule. Therefore, the prediction unit 42 can predict the degree of density of the oil film formed by the base oil composed of the target molecule to be predicted by predicting the feature amount indicating the intermolecular gap of the target molecule corresponding to the target molecule to be predicted from the learned model M2.

[0100] The output unit 43 uses, as the objective variable, the feature amount indicating the intermolecular gap of the target molecule predicted by the learned model M2, and outputs it by display or the like as representing the degree of density of the oil film formed by the base oil composed of the target molecule.

[0101] An example of the accuracy of the predicted value predicted by the learned model M2 in the prediction unit 42 is shown in FIGS. 12 and 13. Note that FIG. 12 shows the case where the feature amount is the midpoint of the approximate formula of the radial distribution function (see FIG. 7), and FIG. 13 shows the case where the feature amount is the slope of the approximate formula of the radial distribution function (see FIG. 7). Also, in FIGS. 12 and 13, as linear saturated hydrocarbons, n-octane (C8H 18 (hereinafter, also simply referred to as C8)), n-decane (C 10 H 22 (hereinafter, also simply referred to as C10)), and n-dodecane (C 12 H 26 (hereinafter, also simply referred to as C12)) are used. The learned model M2 shows the case generated using teacher data consisting only of C12.

[0102] Figure 12 plots the relationship between the calculated values ​​obtained by actually calculating the features of C8, C10, and C12 (midpoints of the approximate radial distribution function) and the predicted values ​​obtained by predicting the features using the trained model M2. The calculated values ​​used were obtained using the molecular feature calculation device 30 shown in Figure 4 above.

[0103] As shown in Figure 12, when the feature is the midpoint of the approximation formula, the root mean square error (RMSE) of the line showing the relationship between the calculated and predicted values ​​of the multiple plotted C12 points is 0.020, and the coefficient of determination (R 2 The R₀ is 0.978. The RMSE of the line showing the relationship between the calculated and predicted values ​​of the multiple plotted C8 and C10 points is 0.037, and R 2 It is 0.848.

[0104] Figure 13 plots the relationship between the calculated values ​​obtained by actually calculating the features (slope of the approximate equation of the radial distribution function) of C8, C10, and C12, and the predicted values ​​obtained by predicting the features using the trained model M2. As with Figure 12, the calculated values ​​were obtained using the molecular feature calculation device 30 shown in Figure 4 above.

[0105] As shown in Figure 13, when the feature is the slope of the approximation formula, the RMSE of the line showing the relationship between the calculated and predicted values ​​of the multiple plotted C12 points is 0.031, and R 2 The value is 0.965. The RMSE of the line showing the relationship between the calculated and predicted values ​​of the multiple plotted C8 and C10 points is 0.041, and R 2 The value is 0.891.

[0106] Therefore, when the features of the radial distribution function are the midpoint and slope of the approximation formula, the predicted values ​​predicted by the trained model M2 have an RMSE of the order of 1 / 100 of the midpoint and slope values, which is sufficiently small, and R 2 The value is close to 1, indicating that the prediction is highly accurate. Furthermore, it can be said that the features of the radial distribution function can be predicted well even with untrained data, as seen with C8 and C10.

[0107] Figure 14 shows an example of the relationship between the predicted values ​​of two features. Note that Figure 14 shows the case where the features are the predicted value of the midpoint of the approximate radial distribution function shown in Figure 12 and the reciprocal of the predicted value of the slope of the approximate radial distribution function shown in Figure 13. Also, in Figure 14, n-undecane (C) is used as the straight-chain saturated hydrocarbon. 11 H 24 (Hereafter also simply referred to as C11.)), n-tridecane (C 13 H 28 (Hereafter, it will also simply be referred to as C13.) ~n-octadecane(C 18 H 38 The following shows the case using (hereinafter simply referred to as C18).

[0108] As shown in Figure 14, by plotting the relationship between the predicted values ​​of the two features, for example, molecules located within region A, indicated by the dashed line, can be considered to have a sparse structure in the oil film formed when used as a base oil, thereby bringing out the effect of the additive. Therefore, molecules plotted within region A can be judged as preferable candidate molecules to be used as a base oil.

[0109] Thus, the prediction device 4 is equipped with a prediction unit 42, in which the prediction unit 42 predicts feature quantities indicating the intermolecular gaps of the target molecules from the target molecules using a trained model M2, and can predict the degree of density of the oil film. By predicting the degree of density of the oil film in the prediction unit 42, the prediction device 4 can predict how well the additive will pass through the oil film. Structures with large interatomic distances between molecules in the oil film are prone to gaps in the oil film formed using a base oil composed of the target molecules. The more sparse the oil film, the more the additive can move through the gaps in the oil film, allowing it to quickly adsorb to the sliding surfaces of sliding parts of industrial machinery, etc., and making it easier for the additive to exert its effect.

[0110] Furthermore, when predicting the degree of density of an oil film formed using a base oil composed of target molecules using conventional diffusion-adsorption simulations, even for a single type of base oil, it takes a long time (for example, several days to several months) to calculate the features from the radial distribution function, and therefore, predicting the density of the oil film takes even longer. The prediction device 4 can calculate the features of a single type of base oil in a short time, for example, a few seconds to a few minutes, thus enabling a quicker and simpler prediction of the degree of density of the oil film.

[0111] Therefore, the prediction device 4 can quickly predict, from the input target molecules, the characteristic quantities indicating the intermolecular gaps of the obtained target molecules, which can be used as an indicator of the degree of density of the oil film. This reduces the time required to select a base oil that brings out the effect of the additive.

[0112] Furthermore, by including a prediction unit 42, the prediction device 4 can reduce the burden and time required to predict the feature quantities that indicate the intermolecular gaps of the target molecules. Therefore, the prediction device 4 can efficiently and easily predict the degree of density of the oil film formed by the base oil consisting of the target molecules, thereby reducing the burden when predicting the degree of density of the oil film formed by the base oil consisting of the target molecules.

[0113] The prediction device 4 can be used when selecting a base oil that brings out the effect of additives, and is therefore suitable for use in the manufacture of lubricating oil compositions. Lubricating oils used in industrial machinery and the like are used under harsh conditions such as high pressure, high speed, high load, and high temperature, so it is important to select a lubricating oil composition that can exhibit the desired lubrication performance even under such conditions. By using the prediction device 4 when selecting a base oil that brings out the effect of additives, it is possible to appropriately select a lubricating oil according to the various applications of lubricating oils used in industrial machinery and the like.

[0114] Furthermore, the learning device 1 and prediction device 4 described above may be configured as a learning system and a prediction system, respectively. That is, the learning device 1 is a standalone device such as a PC (Personal Computer) that houses each component within the device, but one or more of each component may be placed outside the device and connected via a network.

[0115] For example, the training dataset may be located on the cloud. In this case, the learning device 1 is configured as a learning system using the training dataset connected via the network.

[0116] Similarly, one or more of the components of the prediction device 4 may be located outside the device and connected via a network.

[0117] <Hardware configuration of the learning device and prediction device> Next, an example of the hardware configuration of the learning device 1 and the prediction device 4 will be described. Figure 15 is a block diagram showing the hardware configuration of the learning device 1 and the prediction device 4. As shown in Figure 15, the learning device 1 and the prediction device 4 are composed of an information processing device (computer), and physically they can be configured as a computer system including a CPU (Central Processing Unit: processor) 101 which is the arithmetic processing unit, RAM (Random Access Memory) 102 and ROM (Read Only Memory) 103 which are the main memory, an input device 104 which is the input device, an output device 105 which is the input device, a communication module 106, and an auxiliary storage device 107 such as a hard disk. These are interconnected by a bus 108. Note that the output device 105 and the auxiliary storage device 107 may be provided externally.

[0118] The CPU 101 controls the overall operation of the learning device 1 and the prediction device 4 and performs various information processing. The CPU 101 executes, for example, the learning method and prediction method or the learning program and prediction program, stored in the ROM 103 or auxiliary storage device 107, to learn the density of the oil film formed using a base oil consisting of target molecules and to predict the density of the oil film.

[0119] RAM102 is used as the work area of ​​CPU101 and may include non-volatile RAM for storing major control parameters and information.

[0120] ROM103 stores basic input / output programs and the like. Learning programs and prediction programs may also be stored in ROM103.

[0121] The input device 104 is an input device such as a keyboard, mouse, operation buttons, touch panel, or display screen, which receives information input by the user as an instruction signal and outputs that instruction signal to the CPU 101.

[0122] The output device 105 includes display devices such as monitor displays, speakers, and printing devices such as printers. In the output device 105, information such as learning results and prediction results of oil film density are displayed on a display device such as a monitor display, and the displayed screen is updated in response to input operations via the input device 104 or the communication module 106.

[0123] The communication module 106 is a data transmission and reception device such as a network card, and functions as a communication interface that receives information from an external data acquisition server and outputs the analyzed information to other electronic devices.

[0124] The auxiliary storage device 107 is a storage device such as an SSD (Solid State Drive) or HDD (Hard Disk Drive), and stores various data, files, etc., necessary for the operation of the learning device 1 and the prediction device 4.

[0125] The functions of the learning device 1 and the prediction device 4 are realized by reading predetermined computer software (including a learning program and a prediction program) from the main memory such as RAM 102 or the auxiliary storage device 107, executing it with the CPU 101, thereby reading and writing data to the main memory such as RAM 102 and the auxiliary storage device 107, and operating the input device 104, output device 105, and communication module 106.

[0126] Therefore, the learning device 1 and prediction device 4 shown in Figures 1 and 11 are realized through the collaborative action of software and hardware in a computer equipped with the learning device 1 and prediction device 4, where the processor executes predetermined computer software (including a learning program and a prediction program) that is pre-stored.

[0127] A computer program implementing at least some of the functions of the learning device 1 and prediction device 4 shown in Figures 1 and 11 may be installed on the storage of one or more computers. The CPU 101 of one or more computers may perform the functions of the learning device 1 and prediction device 4 shown in Figures 1 and 11 by reading the computer program installed on its own machine into main memory and executing it.

[0128] The learning device 1 and prediction device 4 shown in Figures 1 and 11 may be implemented by one or more CPUs 101. Here, CPU 101 may refer to one or more electronic circuits located on a single chip, or one or more electronic circuits located on two or more chips or two or more devices. When multiple electronic circuits are used, each electronic circuit may communicate by wired or wireless means.

[0129] Furthermore, the functions of each part of the learning device 1 and prediction device 4 shown in Figures 1 and 11 may be executed by a single computer, or they may be executed in a distributed manner by multiple computers. When the functions of each part of the prediction device 4 shown in Figure 4 are executed in a distributed manner by multiple computers, these multiple computers may send and receive data via a communication network including a LAN (Local Area Network), WAN (Wide Area Network), PAN (Personal Area Network), or the Internet.

[0130] The learning program and prediction program can be stored, for example, in the main memory or auxiliary storage device 107 of a computer. Alternatively, the learning program and prediction program may be stored on a computer connected to a communication line such as the Internet, and provided by allowing users to download part or all of the learning program and prediction program via the communication line. Furthermore, the learning program and prediction program may be configured to be provided or distributed via a communication line.

[0131] The learning program and prediction program may be recorded (including installed) into a computer from a state where they are stored in whole or in part on a portable storage medium such as an optical disc like a CD-ROM or DVD-ROM, or a semiconductor memory like flash memory.

[0132] <Learning Method> Next, the learning method will be explained. The learning method involves generating a trained model that predicts the features representing the intermolecular gaps of a target molecule from the target molecule itself, using a training dataset in which target molecules (explanatory variables) and features representing the intermolecular gaps of the target molecule within a simulation cell having an arbitrary parallelepiped shape (target variable) are associated, in a learning device 1 having the configuration shown in Figure 1.

[0133] Figure 16 is a flowchart illustrating the learning method. As shown in Figure 16, the first acquisition unit 11 acquires information on target molecules, which are molecules constituting the base oil, from the first storage unit 21 as explanatory variables (first acquisition step: step S11).

[0134] Next, the second acquisition unit 12 acquires feature quantities indicating the intermolecular gaps of the target molecule within a simulation cell having an arbitrary parallelepiped shape from the second storage unit 22 as the target variable (second acquisition step: step S12).

[0135] Next, the training dataset creation unit 13 extracts the target molecule as the explanatory variable and features indicating the intermolecular gaps of the target molecule within a simulation cell having an arbitrary parallelepiped shape as the objective variable, and adds them to the training dataset (training dataset creation process: step S13).

[0136] The training dataset creation unit 13 associates the input target molecules with features indicating the gaps between them to create a training dataset.

[0137] Next, the learning unit 14 generates a trained model M1 by training it using a training dataset in which the target molecules (explanatory variables) are associated with features (target variables) that indicate the intermolecular gaps of the target molecules within a simulation cell having an arbitrary parallelepiped shape (training process: step S14).

[0138] The learning unit 14 generates a trained model M1 in response to input information about the target molecule included in the training dataset, such that the output is a feature representing the intermolecular gap of the target molecule within a simulation cell having an arbitrary parallelepiped shape.

[0139] Next, the output unit 15 outputs information about the training dataset used in training the trained model M1, and information about the trained model M1, by displaying it or other means (output process: step S15).

[0140] The learning method includes a learning step (step S14), in which a trained model M1 can be generated that predicts feature quantities indicating the intermolecular gaps of target molecules within a simulation cell having an arbitrary parallelepiped shape. The learning method uses the trained model M1 generated in the learning step (step S14) to predict feature quantities indicating the intermolecular gaps of target molecules from the input information of the target molecules. The feature quantities indicating the intermolecular gaps of target molecules can be used as an index representing the degree of density of the oil film formed using a base oil composed of the target molecules. Therefore, since the learning method is used to predict the degree of density of the oil film formed using a base oil composed of the input target molecules, the time required to select a base oil that brings out the effect of additives can be shortened.

[0141] Furthermore, the learning method uses the trained model M1 generated in the learning process (step S14) to predict feature quantities indicating the intermolecular gaps of the target molecules from the information of the target molecules. These feature quantities are used to predict the degree of density of the oil film formed using a base oil composed of the target molecules, thereby reducing the burden and time required to predict the degree of density of the oil film. Therefore, by using the learning method, the burden of predicting the degree of density of the oil film composed of the target molecules can be reduced.

[0142] The learning method preferably uses values ​​calculated based on a radial distribution function that represents the relationship between the interatomic distances between different molecules and the relative abundance of other molecules, obtained by stabilizing a liquid structure created from the three-dimensional structures of any number of target molecules within a simulation cell. This allows the learning method to be used to more accurately predict features indicating the intermolecular gaps of the target molecules, thereby improving the accuracy of selecting a base oil that brings out the effect of the additive.

[0143] <Prediction Method> Next, the prediction method will be explained. The prediction method involves using a prediction device 4, which has the configuration shown in Figure 11, to predict a feature quantity that indicates the intermolecular gaps of the target molecules from the target molecules, using a trained model M2 as an indicator representing the degree of density of the oil film containing the target molecules.

[0144] Figure 17 is a flowchart illustrating the prediction method. As shown in Figure 17, the acquisition unit 41 acquires the target molecules, which are molecules that make up the base oil, as explanatory variables (acquisition step: step S21).

[0145] Next, the prediction unit 42 inputs the target molecules to be predicted, acquired by the acquisition unit 41, into the trained model M2, thereby predicting the degree of density of the oil film formed by the base oil consisting of the target molecules to be predicted, as predicted by the trained model M2 (prediction step: step S22).

[0146] Specifically, the prediction unit 42 inputs the target molecules to be predicted, acquired by the acquisition unit 41, into the trained model M2, and outputs a feature quantity indicating the intermolecular gaps of the target molecules to be predicted, as predicted by the trained model M2, as the target variable. The prediction unit 42 outputs the target variable as an index representing the degree of density of the oil film formed by the base oil consisting of the target molecules to be predicted.

[0147] Next, the output unit 43 outputs the feature quantities indicating the intermolecular gaps of the target molecules, which were predicted by the trained model M2 in the prediction step (step S22), as the target variable, and displays it as the degree of density of the oil film formed by the base oil consisting of the target molecules (output step: step S23).

[0148] The prediction method includes a prediction step (step S22), in which a trained model M2 can predict feature quantities indicating the intermolecular gaps of the target molecules from the target molecules. By predicting the feature quantities indicating the intermolecular gaps of the target molecules in the prediction step (step S22), the prediction method can predict the degree of density of the oil film, and thus predict how well the additive will pass through the oil film. Furthermore, by using the trained model M2 in the prediction step (step S22), the prediction method can predict the degree of density of the oil film more quickly and easily. Therefore, the prediction method can predict the feature quantities indicating the intermolecular gaps of the target molecules obtained from the input target molecules as an indicator of the degree of density of the oil film in a short time, thus shortening the time required to select a base oil that brings out the effect of the additive.

[0149] Furthermore, the prediction method reduces the burden and time required to predict the feature quantities indicating the intermolecular gaps of the target molecules by predicting the feature quantities indicating the intermolecular gaps of the target molecules in the prediction process (step S22). Therefore, the prediction method can efficiently and easily predict the degree of density of the oil film formed by the base oil consisting of the target molecules, thereby reducing the burden when predicting the degree of density of the oil film formed by the base oil consisting of the target molecules.

[0150] As described above, embodiments have been explained, but these embodiments are presented as examples only, and the present invention is not limited by these embodiments. The above embodiments can be implemented in various other forms, and various combinations, omissions, substitutions, and modifications are possible without departing from the spirit of the invention. The above embodiments and their variations are included in the scope and spirit of the invention, as well as in the claims of the invention and its equivalents. [Examples]

[0151] The following describes the embodiment in more detail with reference to examples, but this embodiment is not limited to these examples.

[0152] <Learning device> [Creating a training dataset] 1953 compounds were used as training data (teaching data) for compounds constituting the base oil, and SMILES representing the molecular structure of the compounds was obtained. Multiple molecules constituting the obtained compounds were created and these were randomly arranged to create a liquid structure. Subsequently, structural relaxation was performed using molecular dynamics to construct the liquid structure under desired temperature and pressure conditions. A radial distribution function focusing on the distance between molecules was calculated as a feature quantity describing the structure of the oil film produced by the base oil (also called the "density feature quantity of the oil film"). By approximating the calculated radial distribution function with a logistic function, the slope and midpoint of the region in which the radial distribution function changes from 0 to 1 were determined. The slope and midpoint correspond to the distance between the molecules of the target molecule; the smaller the slope or the larger the midpoint, the further apart the molecules of the target molecule are, meaning that a sparse oil film is formed. A training dataset was created by using the molecular structure of the obtained compounds as the explanatory variable, and the slope and midpoint of the compounds as the density features of the oil film indicating the intermolecular gaps of the obtained compounds as the dependent variables, and associating the molecular structure of the compounds with the slope and midpoint of the compounds.

[0153] [Generating a pre-trained model] The generated training dataset was trained using the training unit of the training device 1 shown in Figure 1, and a trained model was generated.

[0154] [Verification of the relationship between the density features of the oil film predicted by a trained model and the measured friction coefficient] For four molecules—paraffin, naphthene, isoparaffin, and tert-butylcyclohexane represented by formula (i) below—we used the pre-trained model to verify the relationship between the density features of the oil film predicted by the pre-trained model and the coefficient of friction.

[0155] [ka]

[0156] (Predicted values ​​of oil film density features predicted by a trained model) For four molecules—paraffin, naphthene, isoparaffin, and tert-butylcyclohexane represented by formula (i) above—predicted values ​​of the density features of the oil film obtained by the generated trained model were used from the training dataset.

[0157] (Measured values ​​of the friction coefficients of the four molecules) A rectangular test specimen and a pair of clamping members made of Al were prepared. A lubricating oil was applied to the corresponding surfaces of the pair of clamping members, using one of four compounds: paraffin, naphthene, isoparaffin, or tert-butylcyclohexane represented by formula (i) above, to form an oil film.

[0158] Subsequently, as shown in Figure 18, the test specimen was placed between the corresponding surfaces of a pair of clamping members and pressed, and the test specimen was pulled upwards to measure the coefficient of friction. The test was performed for each of the four molecular compounds, and the coefficient of friction was measured for each. After measuring the coefficient of friction, the median value of the coefficients of friction of the four compounds was used as a representative value, and the measured values ​​of the coefficients of friction of the four compounds were standardized.

[0159] (mapping) Table 1 and Figure 19 show the relationship between the predicted values ​​(in Å) of the oil film density features predicted by the trained model and the measured values ​​of the standardized friction coefficient.

[0160] [Table 1]

[0161] As shown in Table 1 and Figure 19, the coefficient of determination R is the point plotted between the density features of the oil film formed by the compound predicted by the trained model and the coefficient of friction. 2The value was approximately 0.83, indicating a correlation between the density features of the oil film formed by the compound predicted by the trained model and the coefficient of friction. Furthermore, the predicted values ​​for the density features of the oil film predicted by the trained model were within the range of 3.07 to 3.31. Therefore, when setting the density features of the oil film, which indicate the intermolecular gaps of the molecules constituting the compound predicted using the trained model M2, to a value greater than a predetermined value, the predetermined value can be determined based on the relationship between the density features of the oil film, which indicate the intermolecular gaps of the molecules constituting the compound, and the coefficient of friction of the base oil composed of the compound, and can be set to 3.0 to 3.5 Å.

[0162] Furthermore, using a trained model, we predicted the density characteristics of an oil film composed of 33 compounds from the publicly available literature, "J. Sameshita et al., The Review of Physical Chemistry of Japan, 14, 2 (1940) 55-67". The 33 compounds included alcohols, fatty acids, esters, and aromatic hydrocarbons. The friction coefficient values ​​from the literature were standardized in the same manner as above to calculate the standardized friction coefficient values. Figure 20 shows the relationship between the predicted values ​​(in Å) of the oil film density characteristics predicted by the trained model and the standardized friction coefficient values ​​for the 33 compounds. As shown in Figure 20, the coefficient of determination R of the points plotted between the oil film density characteristics predicted by the trained model and the standardized friction coefficient is shown. 2 The coefficient of friction was approximately 0.88, indicating a correlation between the density features of the oil film, consisting of compounds from known literature, as predicted by the trained model, and the coefficient of friction.

[0163] Therefore, it was confirmed that the friction coefficient of a compound can be predicted from the density features of the oil film made up of the compound, which are predicted by the trained model M1 generated by the learning device 1. Thus, by simply inputting the molecular structure of the compound, such as SMILES, into the trained model M1, the density of the oil film made up of the compound can be determined, and a base oil that can maximize the effect of the additive can be selected.

[0164] <Prediction device> [Creation of a prediction device] The trained model M1 generated by the learning device 1 shown in Figure 1 was used as the trained model M2 to fabricate the prediction device 4 shown in Figure 11.

[0165] [Prediction of oil film density and selection of compounds] The fabricated prediction device 4 predicted the density features of the oil film formed by 26,159 compounds as the midpoint of the approximate radial distribution function and the reciprocal of the predicted slope of the approximate radial distribution function. Figure 21 shows the relationship between the two predicted density features of the oil film for each compound. Figure 21 also shows the two density features for the 1,953 compounds used as training data and the two predicted density features for the 26,159 compounds. The reference compound was 3,5-diethyloctane. In Figure 21, the further to the upper right of the figure, the easier it is for the additive to pass through, resulting in a sparser oil film.

[0166] As shown in Figure 21, a correlation is observed between the predicted density features of the two oil films of the compound. By plotting the relationship between the predicted values ​​of the density features of the two oil films, it is possible to select a compound that, when used as a base oil, has a more sparse structure than 3,5-diethyloctane, which is the reference compound, and that brings out the effect of the additive.

[0167] <Example 1> Using the fabricated prediction device 4, the density features of the oil film formed by 111 available compounds (manufactured by Tokyo Chemical Industry Co., Ltd.) were further predicted as the midpoint of the approximate radial distribution function and the reciprocal of the predicted slope of the approximate radial distribution function. Figure 22 shows the relationship between the predicted density features of two oil films for the 111 compounds, added to Figure 21. As shown in Figure 22, 14 compounds were extracted from the 111 compounds as compounds that have the potential to form an oil film less dense than the reference point (see the large white circle in Figure 22). Table 2 shows the predicted midpoint of the approximate radial distribution function and the reciprocal of the predicted slope of the approximate radial distribution function for the 14 extracted compounds. Table 2 also shows the value for the reference point compound (3,5-diethyloctane).

[0168] [Table 2]

[0169] Of the 14 compounds, 11 were paraffinic compounds, naphthenic compounds, or alkylbenzene compounds.

[0170] The paraffin compounds are 2,2,4,6,6-pentamethylheptane represented by formula (1-1) below, 2,2,3,3-tetramethylbutane represented by formula (1-2) below, 2,2,4,4,6,8,8-heptamethylnonane represented by formula (1-3) below, and 3,4 - These were four compounds of diethylhexane.

[0171] [ka]

[0172] [ka]

[0173] [ka]

[0174] The naphthenic compounds were five compounds: tert-butylcyclohexane, tricyclo[5.2.1.02,6]decane-3-ene, tricyclo[6.2.1.02,7]undeca-4-ene, (1R,4E,9S)-4,11,11-trimethyl-8-methylidenebicyclo[7.2.0]undeca-4-ene, and 1,2,3,3a,4,5,5a,6,7,8,8a,9,10,10a,10b,10c-hexadecahydropyrene, represented by the following formula (2-1).

[0175] [ka]

[0176] The alkylbenzene compounds were two compounds: 1,4-di-tert-butylbenzene represented by the following formula (3-1) and 1,3-di-tert-butylbenzene represented by the following formula (3-2).

[0177] [ka]

[0178] [ka]

[0179] As shown in Figure 22, by plotting the relationship between the predicted values ​​of the density features of two oil films, six of the compounds plotted in the region where at least one of the midpoint of the approximate radial distribution function and the reciprocal of the predicted slope of the approximate radial distribution function is greater than or equal to the reference point, can be said to have a sparse structure in the oil film formed when used as a base oil. Therefore, if these compounds are used as a base oil, the inhibition of additive movement is suppressed, and even if these compounds form an oil film on a solid surface such as the sliding surface of a sliding part of industrial machinery, the additive can reach the surface. Thus, if these paraffinic compounds, naphthenic compounds, or alkylbenzene compounds are used as a base oil, the performance of the lubricating oil composition can be improved.

[0180] <Example 2> Using the fabricated prediction device 4, 132 additional compounds (manufactured by Sigma-Aldrich Japan LLC), which were separately available, were predicted as density features of the compounds, similar to Example 1. These features were the midpoint of the approximate radial distribution function and the reciprocal of the predicted slope of the approximate radial distribution function. Figure 23 shows the relationship between the two density features of the 132 compounds, added to Figure 21. As shown in Figure 23, 33 compounds were extracted from the 132 compounds as compounds that have the potential to form an oil film less dense than the reference point (see the large white circle in Figure 23). Table 3 shows the predicted values ​​for the midpoint of the approximate radial distribution function and the reciprocal of the predicted slope of the approximate radial distribution function for the extracted 33 compounds. Table 3 also shows the values ​​for the reference point compound (3,5-diethyloctane).

[0181] [Table 3]

[0182] Of the 33 compounds, 18 were paraffinic compounds, naphthenic compounds, or alkylbenzene compounds.

[0183] The paraffinic compounds were two compounds: 2,2,3,3-tetramethylbutane, represented by formula (1-2) below, and 2,2,4,4,6,8,8-heptamethylnonane, represented by formula (1-3) below.

[0184] [ka]

[0185] [ka]

[0186] Naphthenic compounds include 1,2,3,3a,4,5,5a,6,7,8,8a,9,10,10a,10b,10c-hexadecahydropyrene represented by formula (2-1) below, 1,4-di(propan-2-yl)cyclohexane represented by formula (2-2) below, 1,3-di(propan-2-yl)cyclohexane represented by formula (2-3) below, 1,3,5-tri(propan-2-yl)cyclohexane represented by formula (2-4) below, tetracyclo[5.3.0.02,6.03,10]decane represented by formula (2-5) below, 1-cyclohexyladamantane represented by formula (2-6) below, and (2S,6S,10R,14R)-2,9 represented by formula (2-7) below. The 13 compounds were -dimethyltetracyclo[6.6.0.02,6.010,14]tetradecane, tetracyclo[5.4.0.01,3.02,7]undecane represented by the following formula (2-8), and (1R,2S,7R,8R,9S,10R)-2,6,6,9-tetramethyltetracyclo[5.4.0.02,9.08,10]undecane represented by the following formula (2-9), tert-butylcyclohexane, tricyclo[5.2.1.02,6]decane-3-ene, tricyclo[5.2.1.01,5]decane-8-ene, and (1R,4E,9S)-4,11,11-trimethyl-8-methylidenebicyclo[7.2.0]undeca-4-ene.

[0187] [ka]

[0188] [ka]

[0189] [ka]

[0190] [ka]

[0191] [ka]

[0192] [ka]

[0193] [ka]

[0194] [ka]

[0195] [ka]

[0196] The alkylbenzene compounds were two compounds: 1,4-di-tert-butylbenzene represented by the following formula (3-1) and 1,3-di-tert-butylbenzene represented by the following formula (3-2).

[0197] [ka]

[0198] [ka]

[0199] As shown in Figure 23, by plotting the relationship between the predicted values ​​of the density features of the two oil films, similar to Example 1, 14 paraffinic compounds, naphthenic compounds, or alkylbenzene compounds, among those plotted in the region where at least one of the midpoint of the approximate radial distribution function and the reciprocal of the predicted slope of the approximate radial distribution function is greater than or equal to the reference point, can be said to have a sparse structure in the oil film formed when used as a base oil. Therefore, if these paraffinic compounds, naphthenic compounds, or alkylbenzene compounds are used as a base oil, the lubrication performance of the lubricating oil composition can be improved, similar to Example 1.

[0200] The embodiments of the present invention are, for example, as follows. <1> It contains one or more components selected from the group consisting of paraffinic compounds, naphthenic compounds, and alkylbenzene compounds, which have a side chain and have 6 to 20 carbon atoms. The paraffinic compound and the naphthenic compound are base oils in which the alkyl group of the side chain is branched. <2> The aforementioned paraffinic compound is (CH3)n1-R1(n1 is 4~ 7 The integer R1 is a straight-chain saturated hydrocarbon with 4 to 9 carbon atoms. or 3,4-diethylhexane And, The naphthenic compound is a compound having a naphthenic ring with 6 carbon atoms and having 8 to 16 carbon atoms. The alkylbenzene compound is a compound represented as (CH3)3-C-Ar-C-(CH3)3 (where Ar is an aryl group). <1> The base oil described above. <3> The paraffinic compound comprises at least one of 2,2,4,6,6-pentamethylheptane, 2,2,3,3-tetramethylbutane, 2,2,4,4,6,8,8-heptamethylnonane, and 3,4-diethylhexane. <1> or <2> The base oil described above. <4> The naphthenic compounds mentioned above are tert-butylcyclohexane, tricyclo[5.2.1.02,6]decane-3-ene, tricyclo[6.2.1.02,7]undeca-4-ene, tricyclo[5.2.1.01,5]decane-8-ene, (1R,4E,9S)-4,11,11-trimethyl-8-methylidenebicyclo[7.2.0]undeca-4-ene, 1,2,3,3a,4,5,5a,6,7,8,8a,9,10,10a,10b,10c-hexadecahydropyrene, 1,4-di(propan-2-yl)cyclohexane, and 1,3-di(propan-2-yl) ) comprising at least one selected from the group consisting of cyclohexane, 1,3,5-tri(propan-2-yl)cyclohexane, tetracyclo[5.3.0.02,6.03,10]decane, tetracyclo[5.4.0.01,3.02,7]undecane, (1R,2S,7R,8R,9S,10R)-2,6,6,9-tetramethyltetracyclo[5.4.0.02,9.08,10]undecane, 1-cyclohexyladamantane and (2S,6S,10R,14R)-2,9-dimethyltetracyclo[6.6.0.02,6.010,14]tetradecane, <1> ~ <3> The base oil listed in any one of the following. <5> The alkylbenzene compound comprises at least one of 1,4-di-tert-butylbenzene and 1,3-di-tert-butylbenzene. <1> ~ <4> The base oil listed in any one of the following. <6> <1> ~ <5> A lubricating oil composition comprising a base oil as described in any one of the following. <7> In contact with a solid surface, <6> The lubricating oil composition described above. <8> A base oil comprising one or more components selected from the group consisting of paraffinic compounds, naphthenic compounds, and alkylbenzene compounds, A base oil in which the feature quantities indicating the intermolecular gaps of the molecules constituting the component, predicted using a trained model generated with a training dataset that associates target molecules, which are molecules constituting the base oil, with feature quantities indicating the intermolecular gaps of the target molecules in a simulation cell having a parallelepiped shape, are greater than or equal to a predetermined value. <9> The predetermined value is 3.0 to 3.5 Å. <8> The base oil described above. <10> The predetermined value is determined based on the relationship between a characteristic quantity indicating the intermolecular gap of the target molecule and the friction coefficient of the base oil composed of the target molecule. <9> The base oil described above. <11> The aforementioned characteristic quantity represents the degree of density of the oil film formed by the molecules constituting the aforementioned component. <8> ~ <10> The base oil listed in any one of the following. <12> The features included in the training dataset are obtained by calculating a radial distribution function that represents the relative abundance of one molecule in terms of interatomic distances between different molecules, which is included in the liquid-optimized structure obtained by optimizing the liquid structure created from the three-dimensional structure of the target molecule. <8> ~ <11> The base oil listed in any one of the following. <13> The three-dimensional structure is created by arranging any number of the molecules within the simulation cell. <12> The base oil described above. [Explanation of Symbols]

[0201] 1. Learning device 4. Prediction device 11 First acquisition part 12 Second acquisition part 13. Creation of training datasets 14. Learning Department 15, 43 Output section 41 Acquisition Department 42 Prediction Section M1, M2 pre-trained models

Claims

1. tert-butylcyclohexane, tricyclo[5.2.1.02,6]decane-3-ene, tricyclo[6.2.1.02,7]undeca-4-ene, tricyclo[5.2.1.01,5]decane-8-ene, (1R,4E,9S)-4,11,11-trimethyl-8-methylidenebicyclo[7.2.0]undeca-4-ene, 1,2,3,3a,4,5,5a,6,7,8,8a,9,10,10a,10b,10c-hexadecahydropyrene, 1,4-di(propan-2-yl)cyclohexane, 1,3-di(propan-2-yl)cyclohexane, 1,3,5-tri(propan-2-yl)cyclo A base oil used in any of the following: engine oil, hydraulic oil, compressor oil, turbine oil, gear oil, metalworking oil, and refrigeration oil, comprising one compound selected from the group consisting of hexane, tetracyclo[5.3.0.02,6.03,10]decane, tetracyclo[5.4.0.01,3.02,7]undecane, (1R,2S,7R,8R,9S,10R)-2,6,6,9-tetramethyltetracyclo[5.4.0.02,9.08,10]undecane, 1-cyclohexyladamantane, and (2S,6S,10R,14R)-2,9-dimethyltetracyclo[6.6.0.02,6.010,14]tetradecane.

2. A lubricating oil composition substantially comprising the base oil and additives described in claim 1.

3. The lubricating oil composition according to claim 2, wherein the oil film formed by the base oil comes into contact with a solid surface.

4. A method for lubricating a sliding surface, wherein the additive present in the base oil described in claim 1 is made to migrate within the oil film formed by the base oil, thereby facilitating its arrival at the sliding surface of the sliding part.

Images