Oil product formulation performance prediction method and system based on historical small sample data

Abstract

The application provides an oil product formula performance prediction method and system based on historical small sample data, and relates to the technical field of oil product formula performance prediction. In the application, a first component sequence and an environmental representation vector of a historical oil product formula are acquired, a historical formula topology is constructed, a candidate formula topology is constructed for a to-be-tested oil product to calculate a topology structure difference, feature extraction is performed on the historical formula topology based on the topology structure difference to generate a structure-effect mapping parameter, and nonlinear mapping of a physicochemical dimension is performed to output a kinematic viscosity value and an anti-wear scar value of the to-be-tested oil product. The technical scheme provided by the application jointly expresses the chemical connection relationship at the molecular level and the thermodynamic constraint of the macroscopic service environment in the same topology structure, effectively describes the coupling effect of the additive molecular structure and the high-temperature service condition on the formula performance, and also realizes accurate prediction of multi-dimensional performance indexes of the to-be-tested new formula in a cross-formula generalization manner without relying on a large number of bench tests.

Description

Technical Field

[0001] This application relates to the field of oil formulation performance prediction technology, and in particular to a method and system for predicting oil formulation performance based on historical small sample data. Background Technology

[0002] As industrial equipment develops towards high-speed, heavy-load, and high-temperature operating conditions, increasingly higher requirements are placed on the kinematic viscosity stability and anti-wear performance of oils under target high-temperature environments. In the process of oil research and development, researchers need to predict the kinematic viscosity and anti-wear spot values ​​of various formulations under target high-temperature environments from a large number of candidate formulations in order to screen the target formulations that meet the requirements of equipment operating conditions.

[0003] Existing oil formulation performance prediction schemes are usually based on physical testing or simple statistical regression. For example, some schemes use a four-ball friction and wear test bench and a high-temperature kinematic viscometer to directly prepare physical samples of each candidate formulation and obtain performance data. Other schemes use linear regression models based on component dosage, and extrapolate the performance values ​​of candidate formulations based on the statistical relationship between the component ratios of a small number of historical formulations and the corresponding performance test results.

[0004] However, existing methods have significant limitations in actual prediction. First, physical testing is time-consuming and costly, requiring retesting every time the component ratio is changed, making it difficult to adapt to the needs of rapid iterative formulation screening. Second, linear regression based on component dosage only establishes a statistical relationship between component mass fraction and macroscopic performance, failing to characterize the coupling effect of additive molecular structure and high-temperature service conditions on formulation performance, resulting in a significant decrease in prediction accuracy in the extrapolation range. Third, existing methods often predict a single performance in isolation, failing to simultaneously output multi-dimensional performance indicators such as kinematic viscosity and anti-wear spots, making it difficult to support comprehensive formulation evaluation. Summary of the Invention

[0005] To address the problems existing in the prior art, the purpose of this application is to provide a method and system for predicting the performance of oil formulations based on historical small sample data. This method constructs a historical formulation topology by structurally mapping the component sequences of historical oil formulations and concatenating them with the high-temperature environment characterization vector dimension. Simultaneously, a candidate formulation topology is constructed for the oil to be tested, and its topological structure difference is compared with that of the historical formulation topology. By leveraging the transferable structure-property information in the historical samples, structure-property mapping parameters for the oil to be tested are generated. Furthermore, through nonlinear mapping of the physicochemical dimension, kinematic viscosity and anti-wear plaque values ​​are output simultaneously, thereby achieving a physically interpretable and cross-formulation generalizable quantitative prediction of the performance of the oil formulation to be tested.

[0006] To achieve the above objectives, this application adopts the following technical solution:

[0007] Firstly, this application provides a method for predicting the performance of oil formulations based on historical small-sample data, including:

[0008] Obtain the first component sequence of historical oil formulations and the environmental characterization vector of historical oil formulations under target high-temperature conditions;

[0009] The first component sequence is subjected to structural mapping processing to generate a first molecular feature, and the first molecular feature is concatenated with the environmental characterization vector to construct a historical formula topology.

[0010] The second component sequence of the oil to be tested and the characterization vector of the environment to be tested are obtained, and the second component sequence is subjected to structural mapping processing to generate the second molecular feature;

[0011] The second molecular feature is concatenated with the characterization vector of the environment to be tested to construct a candidate formulation topology, and the topological structure difference between the candidate formulation topology and the historical formulation topology is calculated.

[0012] Based on the topological differences, features are extracted from the historical formulation topology to generate structure-activity mapping parameters corresponding to the oil to be tested.

[0013] The structure-property mapping parameters are nonlinearly mapped along the physicochemical dimension to output the predicted performance values ​​of the oil under test. The predicted performance values ​​include the kinematic viscosity and anti-wear plaque values ​​of the oil under test.

[0014] Optionally, the first component sequence is subjected to structure mapping processing to generate a first molecular feature, and the first molecular feature is concatenated with the environmental representation vector to construct a historical formulation topology, including:

[0015] Analyze the covalent bond information connecting adjacent atoms in each of the compound nodes;

[0016] Based on the covalent bond information, the local connectivity graph of each compound node is extracted as the first molecular feature;

[0017] Extract the thermodynamic boundary carried in the environmental characterization vector;

[0018] The thermodynamic boundary is applied as an external constraint to the locally connected graph to construct the historical recipe topology.

[0019] Optionally, the second component sequence of the oil to be tested and the characterization vector of the environment to be tested are obtained, and the second component sequence is subjected to structure mapping processing to generate a second molecular feature, including:

[0020] Candidate nodes characterizing the additives to be tested are extracted from the formulation design list of the oil to be tested in order to combine the second component sequence.

[0021] Based on the chain interaction properties of the additive to be tested, the second component sequence is physically mapped to output the second molecular feature;

[0022] The pressure load and thermal conductivity in the environmental characterization vector to be tested are concatenated with the second molecular feature to construct a candidate formulation topology.

[0023] Optionally, the second molecular feature is concatenated dimensionally with the test environment representation vector to construct a candidate formulation topology, and the topological differences between the candidate formulation topology and the historical formulation topology are calculated, including:

[0024] Obtain a list of formulation designs for the oil to be tested, and extract candidate nodes characterizing the additives to be tested from the list of formulation designs to combine them into a second component sequence;

[0025] Based on the molecular stereoconfiguration and steric hindrance of the additive to be tested, the sequence of the second component is mapped to output the second molecular feature;

[0026] The operating pressure load and thermal distribution field are extracted from the characterization vector of the environment under test, and the operating pressure load and thermal distribution field are used as physical constraints to be dimensionally concatenated with the second molecular feature to construct the candidate formulation topology;

[0027] Extract the micro-crosslinking density of the candidate formulation topology and the micro-crosslinking density of the historical formulation topology, and compare the spatial mesh deformation rate of the candidate formulation topology and the historical formulation topology under the physical constraint term;

[0028] Based on the discrete span of the micro-crosslinking density and the spatial grid deformation rate, the structural distortion deviation between the candidate formulation topology and the historical formulation topology is calculated, and the structural distortion deviation is established as the topological difference.

[0029] Optionally, based on the topological differences, feature extraction is performed on the historical formulation topology to generate structure-property mapping parameters corresponding to the oil to be tested, including:

[0030] Based on the topological differences, locate the shared connected regions in the historical recipe topology whose internal structural changes are less than a preset limit value;

[0031] Read the force response values ​​of each node in the shared connectivity region under the action of the target high-temperature environment;

[0032] Based on the force response value, the attributes of each node in the shared connectivity region are reorganized using physical information.

[0033] The node attributes, after being reorganized by physical information, are mapped to the multi-component thermodynamic system of the oil sample to be tested, so as to output the structure-property mapping parameters of the oil sample to be tested.

[0034] Optionally, the node attributes after physical information reorganization are mapped to the multi-component thermodynamic system of the oil sample to be tested, so as to output the structure-property mapping parameters of the oil sample to be tested, including:

[0035] Extracting the concentration distribution of substances and the microscopic phase boundaries in a multi-component thermodynamic system;

[0036] Based on the substance concentration distribution and the microscopic phase boundary, the microscopic interaction energy of the node attributes after physical information recombination within the multi-component thermodynamic system is analyzed.

[0037] Based on the aforementioned micro-interaction energy, the frequency of micro-collisions and the heat conduction flux of each node under the target high-temperature environment are inferred.

[0038] Based on the micro-collision frequency, the amplitude of internal energy fluctuations at each node due to kinetic energy exchange is calculated, and the local temperature gradient of the multi-component thermodynamic system is determined based on the heat conduction flux.

[0039] Substitute the internal energy fluctuation amplitude and the local temperature gradient into the non-equilibrium thermodynamic model to deduce the local potential energy shift of each node;

[0040] Based on the structural attachment relationship of each node in the multi-component thermodynamic system, the local potential energy shift of each node is spatially superimposed and extended to construct a global microscopic potential energy surface, and the extreme stationary points and barrier heights on the global microscopic potential energy surface are extracted and converted into structure-property mapping parameters.

[0041] Optionally, a nonlinear mapping of the physicochemical dimensions is performed on the structure-property mapping parameters to output predicted performance values ​​of the oil under test. These predicted performance values ​​include the kinematic viscosity and anti-wear plaque values ​​of the oil under test, including:

[0042] Analyze the bond dissociation energy of the structure-property mapping parameters in the target high-temperature environment;

[0043] Based on the bond dissociation energy, the change in flow resistance of the oil under mechanical shear force is deduced;

[0044] The change in flow resistance is used as an input parameter and projected into the rheological constitutive model to output the kinematic viscosity value;

[0045] Extract the boundary lubricating film strength, which characterizes the adsorption capacity of extreme pressure additives, from the structure-property mapping parameters;

[0046] Based on the strength of the boundary lubricating film, the micro-protrusion contact state of the metal contact surface is derived to output the anti-wear spot value.

[0047] Secondly, this application provides an oil formulation performance prediction system based on historical sample data, including:

[0048] The acquisition module is used to acquire the first component sequence of historical oil formulations and the environmental characterization vector of historical oil formulations under the target high-temperature environment.

[0049] The first mapping module is used to perform structural mapping processing on the first component sequence to generate a first molecular feature, and to concatenate the first molecular feature with the environmental characterization vector to construct a historical formula topology.

[0050] The second mapping module is used to obtain the second component sequence of the oil to be tested and the characterization vector of the environment to be tested, and to perform structural mapping processing on the second component sequence to generate the second molecular feature;

[0051] A construction module is used to concatenate the second molecular feature with the characterization vector of the environment to be tested to construct a candidate formulation topology, and to calculate the topological structure difference between the candidate formulation topology and the historical formulation topology.

[0052] The feature extraction module is used to extract features from the historical formulation topology based on the topological differences, and generate structure-property mapping parameters corresponding to the oil to be tested.

[0053] The nonlinear mapping module is used to perform nonlinear mapping of the structure-property mapping parameters in the physicochemical dimension and output the predicted performance values ​​of the oil under test. The predicted performance values ​​include the kinematic viscosity value and anti-wear spot value of the oil under test.

[0054] Thirdly, this application provides a computing device, including a processing component and a storage component; the storage component stores one or more computer instructions; the one or more computer instructions are to be invoked and executed by the processing component to implement the oil formulation performance prediction method based on historical sample data as described in the first aspect above.

[0055] Fourthly, this application provides a computer storage medium storing a computer program, which, when executed by a computer, implements the oil formulation performance prediction method based on historical sample data as described in the first aspect.

[0056] This application obtains the first molecular features by performing structure mapping processing on the component sequence of historical oil formulations, and constructs the historical formulation topology by dimensionally splicing the first molecular features with the environmental characterization vector under the target high temperature environment. This allows the molecular-level chemical connection relationship and the thermodynamic constraints of the macroscopic service environment to be jointly expressed in the same topological structure, breaking through the bottleneck of the separation between molecular structure information and service environment information in the previous formulation performance prediction, and fundamentally solving the problem that the component dosage regression method is difficult to reflect the environmental coupling effect.

[0057] Furthermore, this application constructs candidate formulation topologies for the oil samples to be tested using the same process, and calculates the topological differences between the candidate formulation topologies and the historical formulation topologies using the discrete span of the micro-crosslinking density and the spatial grid deformation rate as measures. Then, it locates the shared connected regions with small internal structural variations in the historical formulation topologies as the regions from which transferable structure-property information can be extracted. Thus, without the need for physical sample preparation and bench testing of the formulations to be tested, it uses historical samples to make physically interpretable performance inferences for new formulations, which significantly improves the iterative efficiency of formulation screening and the prediction accuracy of extrapolation intervals.

[0058] Furthermore, this application reconstructs the physical information of the force response values ​​of nodes within the shared connectivity region and maps it to a multi-component thermodynamic system, constructing a global microscopic potential energy surface containing information on extreme stationary points and barrier heights. Then, in the physicochemical dimension, it outputs kinematic viscosity and anti-wear plaque values ​​through the bond dissociation energy-rheological constitutive model pathway and the boundary lubrication film strength-micro-convexity contact pathway, respectively. This allows the prediction results to simultaneously cover the core performance of oil products at both the bulk rheology and interfacial friction levels, thus improving the physical interpretability, cross-formulation generalization ability, and multi-dimensional performance output capability of oil formulation performance prediction. This facilitates researchers in finely selecting target formulations that meet high-temperature service requirements.

[0059] These or other aspects of this application will become more apparent in the following description of the embodiments. Attached Figure Description

[0060] To more clearly illustrate the technical solutions in the embodiments of this application, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0061] Figure 1 A flowchart illustrating a method for predicting the performance of oil formulations based on historical small sample data, provided for an embodiment of this application;

[0062] Figure 2This is a schematic diagram of the overall system for predicting the performance of oil formulations provided in this application embodiment;

[0063] Figure 3 A schematic diagram illustrating the process of constructing a historical formulation topology by structural mapping of the first component sequence and concatenating it with the environmental characterization vector dimension, as provided in an embodiment of this application.

[0064] Figure 4 A schematic diagram illustrating the geometric principles for calculating the topological differences between candidate recipe topologies and historical recipe topologies, and for locating shared connected regions, provided in this application embodiment.

[0065] Figure 5 A schematic diagram illustrating the process of constructing a global microscopic potential energy surface and extracting structure-property mapping parameters based on the recombined node attributes of physical information, as provided in the embodiments of this application.

[0066] Figure 6 This is a schematic diagram of the structure of an oil formulation performance prediction system based on historical sample data, provided in an embodiment of this application. Detailed Implementation

[0067] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of this application.

[0068] To facilitate understanding of the embodiments of this application, some non-publicly known terms involved in this application will be briefly explained first.

[0069] The historical component sequence (i.e., the first component sequence) refers to the sequence of compound nodes obtained by arranging a historical oil formulation in order of component categories such as base oil, antioxidant, anti-wear agent, extreme pressure agent, and viscosity index improver. Each compound node corresponds to a representative compound of a component category. For example, the base oil node can correspond to polyalphaolefin, and the anti-wear agent node can correspond to zinc dialkyl dithiophosphate. In this application, the historical component sequence is used as input for structure mapping processing.

[0070] The aforementioned environmental characterization vector refers to a multidimensional physical quantity vector obtained by quantifying the actual service conditions of historical oil formulations. The environmental characterization vector described in this application includes at least the temperature value, pressure load, and heat transfer of the target high-temperature environment, and optionally also includes thermodynamic boundary quantities such as the upper limit of shear rate and the partial pressure of oxygen in the atmosphere.

[0071] Structure mapping processing refers to the process of converting a component sequence composed of compound nodes into a numerical feature tensor that can be subsequently spliced ​​and differentially calculated according to molecular chemical structure information. The structure mapping processing described in this application is specifically implemented by a local connected graph extraction method based on covalent bond connectivity, which will be explained in detail below.

[0072] Historical molecular features refer to the numerical feature tensors obtained after structure mapping, which are used to characterize the molecular structure of each compound node in the historical component sequence. Each dimension of the tensor corresponds to the local connected graph encoding of a compound node.

[0073] Historical formulation topology refers to a composite topological structure that carries both molecular structure information and service environment constraint information, formed by splicing historical molecular characteristics with environmental characterization vectors corresponding to historical oil formulations in terms of dimensions.

[0074] The candidate component sequence (i.e., the second component sequence), the candidate molecular feature (i.e., the second molecular feature), and the candidate formulation topology correspond to the products of the oil formulation to be tested at each stage in the structure mapping and dimension splicing process, and their physical meanings are consistent with the historical end products mentioned above.

[0075] Topological structure difference refers to the comprehensive deviation between the candidate formulation topology and the historical formulation topology in terms of the micro-crosslinking density at the molecular level and the spatial grid deformation rate under physical constraints. In this application, the topological structure difference is measured in numerical form as the degree of structural distortion deviation.

[0076] Structure-activity mapping parameters refer to intermediate bridging parameters extracted from historical formulation topologies that can map molecular structure information to macroscopic performance indicators. In this application, the structure-activity mapping parameters are specifically represented as the coordinates of extreme stationary points on the global microscopic potential energy surface and the set of potential barrier height values ​​connecting the extreme stationary points.

[0077] A shared connected region refers to a sub-topological region in the historical recipe topology whose internal structural changes are less than a preset limit. The structural information within this region can be regarded as the underlying physical laws that can be transferred and shared between the historical recipe and the recipe to be tested.

[0078] A multi-component thermodynamic system refers to a heterogeneous microstructure consisting of base oil phase, additive micelle phase, and interface phase, formed by mixing the components of the oil formulation to be tested in their designed proportions.

[0079] The global microscopic potential energy surface refers to the three-dimensional surface of potential energy function that covers the whole phase space of the oil formulation under test, obtained by superimposing and extending the local potential energy shifts of each node in a multi-component thermodynamic system in the spatial dimension.

[0080] Boundary lubricating film strength refers to the maximum load-bearing capacity that the boundary lubricating film formed by extreme pressure additives on the surface of metal friction pairs through chemical adsorption can maintain its integrity without being sheared and broken when subjected to external mechanical loads.

[0081] Example 1

[0082] Figure 1 This is a flowchart illustrating a method for predicting oil formulation performance based on historical small sample data, provided in Embodiment 1 of this application. Figure 1 As shown in the figure, the oil formulation performance prediction method based on historical small sample data provided in this application includes steps S100 to S600. Each step will be described in detail below with reference to the accompanying drawings. To make the specific application process of each step more intuitive, a specific application scenario will be used as an example below, namely, a scenario where the performance migration prediction is performed on a candidate formulation of a low-sulfur environmentally friendly gear oil for a new energy vehicle reducer based on the historical formulation of a 75W-90 gear oil for a heavy-duty truck transmission. In this scenario, the target high-temperature environment temperature is set to 140℃.

[0083] Step S100: Obtain the historical component sequence of historical oil formulations and the environmental characterization vector of historical oil formulations under the target high-temperature environment.

[0084] In this step, the target high-temperature environment refers to the actual operating temperature range that the oil being tested is designed to operate in. For example, the target high-temperature environment for vehicle transmission fluid applications can be set to 120°C to 150°C, depending on the equipment's operating conditions. The historical oil formulation refers to a finished oil formulation that has been actually prepared and whose performance data, such as kinematic viscosity and anti-wear plaque, has been fully recorded in historical tests or bench operations.

[0085] Specifically, the process of obtaining the first component sequence and the environmental characterization vector in this step includes:

[0086] First, the formulation records of the target historical oil product are retrieved from the historical sample database. The components in the formulation records are arranged sequentially according to categories such as base oil, antioxidants, anti-wear agents, extreme pressure agents, and viscosity index improvers. Representative compounds from each category are used as compound nodes to form the first component sequence. The historical sample database can consist of the company's own long-term accumulated experimental test records or a publicly available oil performance database, and is maintained by authorized R&D personnel.

[0087] Secondly, physical quantities such as temperature, pressure load, heat transfer, and upper limit of shear rate under the target high-temperature environment are extracted from the corresponding test condition records of the historical oil formulation. These physical quantities are then combined into a multi-dimensional numerical vector in a predetermined order, serving as the environmental characterization vector for the historical oil formulation. The dimensions of the environmental characterization vector can be set to, for example, 6 to 8 dimensions, depending on the number of thermodynamic constraint variables of interest.

[0088] Finally, the first component sequence and environmental characterization vector are bound and stored as a historical sample record, and multiple historical sample records are aggregated into a historical sample set used in subsequent steps. The reason for obtaining the first component sequence and environmental characterization vector simultaneously in this step is that the performance of oil products is essentially determined by both the molecular hierarchical structure and the macroscopic service environment. Both must be used as a whole input in order to construct a formulation topology that reflects the complete coupling relationship between structure, ring, and effect in subsequent steps.

[0089] The following example illustrates this step using a specific application instance.

[0090] In this specific application embodiment, the oil being tested is targeted at a scenario predicting the migration of gear oil from a heavy-duty truck transmission to a gear oil in a new energy vehicle reducer. The target high-temperature environment is set to a temperature of 140°C, a pressure load of 2.5 MPa, and a thermal conductivity of [missing information]. The upper limit of the shear rate is set to The atmospheric oxygen partial pressure is set to 0.21 atm, and the corresponding environmental characterization vector is: =[140, 2.5, 0.15, 1500, 0.21, 1.0]; Simultaneously, a known performance 75W-90 heavy-duty gear oil formulation was retrieved from the historical sample database as a historical oil formulation. Its first component sequence includes: compound node 1 (base oil: polyalphaolefin PAO-6, accounting for 85%), compound node 2 (antioxidant: 2,6-di-tert-butyl-p-cresol BHT, accounting for 0.3%), compound node 3 (antiwear agent: zinc dialkyl dithiophosphate ZDDP, accounting for 1.5%), compound node 4 (extreme pressure agent: sulfurized isobutylene SIB, accounting for 4%), and compound node 5 (viscosity index improver: olefin copolymer OCP, accounting for 9.2%), i.e., the first component sequence. ={PAO-6, BHT, ZDDP, SIB, OCP}. The historical sample had a measured kinematic viscosity of 15.3 mm² / s at 100℃ and a four-ball friction wear scar diameter of 0.40 mm. These measured performance data were compared with... , This data is then bound and stored as a historical sample record, serving as the basis for migration in subsequent steps and a benchmark for verifying the final prediction results.

[0091] Figure 2This is a schematic diagram of the overall system for predicting the performance of oil formulations provided in this application embodiment.

[0092] like Figure 2 As shown, the overall system for predicting the performance of oil formulations includes a historical sample database 201, an environmental condition acquisition unit 202, a structure mapping processing unit 203, a dimension splicing unit 204, a topological difference calculation unit 205, a feature extraction unit 206, a nonlinear mapping unit 207, and an interactive terminal for R&D personnel 208.

[0093] Among them, the historical sample database 201 is used to store the prescription records and corresponding performance data of historical oil formulations; the environmental condition acquisition unit 202 is used to collect physical quantities such as temperature, pressure and heat conduction of the target high-temperature environment and output environmental characterization vectors.

[0094] The structure mapping processing unit 203 is connected to the historical sample database 201 and the environmental condition acquisition unit 202 respectively, and is used to perform structure mapping processing on the historical component sequence (i.e. the first component sequence) and the candidate component sequence (i.e. the second component sequence).

[0095] The input of the dimension splicing unit 204 is connected to the output of the structure mapping processing unit 203, and is used to splice molecular features and environmental characterization vectors in a dimension to construct historical formulation topology and candidate formulation topology.

[0096] The input of the topology difference calculation unit 205 is connected to the output of the dimension splicing unit 204, and is used to calculate the topological differences between the candidate formulation topology and the historical formulation topology.

[0097] The input of the feature extraction unit 206 is connected to the output of the topology difference calculation unit 205, and is used to extract structure-property mapping parameters from the historical formulation topology based on the topological differences; the input of the nonlinear mapping unit 207 is connected to the output of the feature extraction unit 206, and is used to perform nonlinear mapping of the structure-property mapping parameters in the physicochemical dimension and output predicted performance values; the R&D personnel interactive terminal 208 is connected to the nonlinear mapping unit 207, and is used to present the kinematic viscosity value and anti-wear spot value of the oil to be tested in a visual manner, and support R&D personnel to adjust the next round of candidate formulations accordingly.

[0098] Step S200: Perform structure mapping processing on the first component sequence to generate a first molecular feature, and concatenate the first molecular feature with the environmental characterization vector to construct a historical formulation topology.

[0099] Since the first component sequence obtained in step S100 is only a discrete symbol sequence composed of compound nodes and cannot directly participate in the coupling analysis with the service environment, this step needs to first convert it into quantifiable molecular features, and then concatenate it with the environmental characterization vector to form a composite topology that simultaneously carries molecular structure information and service environment constraints. The following section will combine... Figure 3 Please provide a detailed explanation.

[0100] Figure 3 This is a schematic diagram illustrating the process of constructing a historical formulation topology by structural mapping of the first component sequence and concatenating it with the environmental characterization vector dimension, as provided in an embodiment of this application.

[0101] like Figure 3 As shown, the first component sequence 301 is composed of compound nodes 301-1, 301-2, 301-3, and 301-4 in sequence. For each compound node, the covalent bond connection relationship between adjacent atoms is analyzed and local connected graphs 302-1 to 302-4 are extracted. All local connected graphs together constitute the first molecular feature 302. The environmental characterization vector 303 carries thermodynamic boundaries such as temperature value 303-1, pressure load 303-2, and thermal conductivity 303-3. The thermodynamic boundaries are applied as external constraints 304 to all local connected graphs, so that each local connected graph is surrounded by constraint nodes, and finally the historical formula topology 305 is formed.

[0102] Specifically, the implementation process of constructing the historical recipe topology in this step includes:

[0103] First, the covalent bond information connecting adjacent atoms in each compound node is analyzed. This covalent bond information includes at least the atomic number, bond order (single, double, triple), bond length, and bond angle, among other geometric and electronic parameters. Specifically, this analysis step can be performed by encoding the molecular structure of each compound node using SMILES or InChI codes. The encoded string is then fed into an open-source cheminformatics library (e.g., RDKit) to perform a conversion from SMILES to a molecular diagram to read the bonding relationships of each compound node. The SMILES or InChI codes can be obtained by researchers from publicly available chemical databases (e.g., PubChem), and this application does not restrict the specific source.

[0104] Secondly, based on the covalent bond information, the local connected graphs of each compound node are extracted as historical molecular features. The local connected graph refers to a subgraph formed by extending outwards from the core functional group atom of the compound node (e.g., the P atom in zinc dialkyl dithiophosphate) to a predetermined depth k (e.g., 3 hops) to all atoms within k hops of the root node. The extraction method of the local connected graph in this application can be implemented, for example, using an extended connectivity fingerprint algorithm, hashing each k-hop neighborhood subgraph into a fixed-length bit vector as the local connected graph encoding of that compound node; the encoding vectors of all compound nodes are concatenated in the order of the component sequence to obtain the historical molecular features.

[0105] Next, the thermodynamic boundary carried in the environmental characterization vector is extracted. The thermodynamic boundary refers to the set of boundary conditions determined by environmental physical quantities such as temperature, pressure load, and thermal conductivity, which limit the energy and conformation range that oil molecules can occupy during service. Specifically, the environmental characterization vector is split according to the category of physical quantity, and temperature, pressure load, and thermal conductivity are taken as the three components of the thermodynamic boundary, and arranged in the order of [T, P, q].

[0106] Finally, the thermodynamic boundary is applied as an external constraint to the locally connected graph to construct the historical formulation topology. Specifically, each node in the locally connected graph is used as an anchor point, and a new set of virtual constraint nodes is added to carry the boundary values ​​of temperature, pressure, and heat conduction, respectively. Weighted constraint edges are established between the virtual constraint nodes and their corresponding locally connected graph nodes, with the weight values ​​taken as the normalized values ​​of the corresponding thermodynamic boundaries. The composite graph structure with constraint edges is the historical formulation topology. The reason for using constraint nodes instead of directly using environmental quantities as global variables is that different compound nodes have significantly different sensitivities to temperature and pressure (e.g., base oil viscosity is sensitive to temperature while anti-wear agent activity is sensitive to pressure). Applying node-level constraints is necessary to preserve this node-level sensitivity in subsequent difference comparisons.

[0107] Continuing from the historical sample of 75W-90 heavy-duty gear oil obtained in step S100, the first component sequence is... The SMILES codes of each compound node in {PAO-6, BHT, ZDDP, SIB, OCP} are fed into the open-source cheminformatics library RDKit for conversion from SMILES to molecular graphs: the SMILES corresponding to PAO-6 is CCCCCCCC(CCCCCCCC)CCCCCC, which resolves to approximately 24 carbon skeleton atoms and their single bond connections; the core structure of ZDDP is rooted at the P atom, which resolves to the bond order information of P=S double bonds, PO single bonds, and PS single bonds; ECFP encoding is performed at a depth of k=3 hops, and each compound node outputs a 1024-bit locally connected graph fingerprint vector. The fingerprint vectors of the five nodes are concatenated end-to-end in M_h order to obtain a 5120-dimensional first molecular feature. Then, thermodynamic boundaries were extracted in the order of [T=140℃, P=2.5MPa, q=0.15W / (m·K)]. Fifteen virtual constraint nodes were added, each anchored by one of the five compound nodes (each compound node corresponds to three constraint nodes: temperature constraint node, pressure constraint node, and heat conduction constraint node). Weighted constraint edges were established between each constraint node and its corresponding compound node, with weights taken from the normalized values ​​of the corresponding thermodynamic boundaries (e.g., the normalized temperature value was 140 / 200=0.70). This resulted in a historical formulation topology containing five principal molecular nodes, 15 virtual constraint nodes, and approximately 85 edges. .

[0108] Step S300: Obtain the second component sequence of the oil to be tested and the characterization vector of the environment to be tested, and perform structural mapping processing on the second component sequence to generate candidate molecular features (i.e., second molecular features).

[0109] The historical formulation topology established in step S200 records the structure-ring coupling relationship of existing oil formulations. In order to transfer this historical experience to the oil to be tested, this step needs to extract the second molecular feature of the oil to be tested according to the same paradigm, so as to prepare for structural alignment for subsequent topological comparison with the historical end.

[0110] Specifically, the process of generating the candidate molecular features (i.e., the second molecular features) in this step includes:

[0111] First, candidate nodes representing the additives to be tested are extracted from the formulation design list of the oil to be tested, in order to form the second component sequence. The formulation design list refers to the preliminary composition ratio scheme proposed by the researchers for the current oil to be tested, such as 80% base oil, 0.8% antioxidant, 1.2% anti-wear agent, and 1.5% extreme pressure agent. Representative compounds are extracted from this list in a fixed order of base oil, antioxidant, anti-wear agent, extreme pressure agent, and viscosity index improver as candidate nodes, and then combined sequentially to form the second component sequence.

[0112] Secondly, based on the chain interaction properties of the additive to be tested, the second component sequence is physically mapped to output the second molecular feature. The chain interaction property refers to the interchain interaction behavior between the long-chain alkyl or polymer side chains in the additive molecule and the base oil molecule through non-bonding interactions such as van der Waals forces, hydrogen bonds, and electrostatic attraction. The specific physical mapping method is as follows: an additional chain interaction property dimension is added to each candidate node on the local connectivity graph obtained by ECFP encoding. The value of this dimension is estimated by molecular dynamics simulation. Specifically, the average interaction potential energy between the base oil molecule and the additive molecule can be estimated by using open-source molecular simulation software (such as LAMMPS or GROMACS) with a small-batch configuration sampling method as the chain interaction property value. This value is then concatenated with the local connectivity graph encoding dimension to obtain the second molecular feature.

[0113] Next, the pressure load and thermal conductivity in the environmental characterization vector to be tested are dimensionally concatenated with the candidate molecular features to construct a candidate formulation topology. The extraction method of the environmental characterization vector to be tested is the same as that for extracting environmental characterization vectors for historical samples in step S100, but its physical quantity values ​​reflect the intended application conditions of the oil to be tested; the dimensional concatenation method is the same as that in step S200, that is, virtual constraint nodes are added with each node in the candidate molecular features as anchor points.

[0114] The reason for introducing chain interaction properties into the candidate molecular features (i.e., the second molecular features) is that the formulation to be tested often involves novel additives that do not appear directly compared with historical formulations. Relying solely on local connected graph encoding is insufficient to reflect their actual mixing behavior with base oil. The introduction of chain interaction properties provides a physical basis for the comparability of candidate molecular features between the historical end and the end to be tested.

[0115] The following example illustrates the execution results of this step using the specific application examples described above. Continuing with the aforementioned application scenario, suppose the R&D personnel need to design a low-sulfur, environmentally friendly gear oil as the test oil for a new energy vehicle reducer. The formula design list includes: base oil PAO-6 (84%), amine antioxidants (0.5%), a novel trioctyl borate anti-wear agent (1.3%), sulfurized isobutylene extreme pressure agent SIB (3%) (1% less than the historical formula), and viscosity index improver polyisobutylene PIB (11.2%). This combination forms the second component sequence. ={PAO-6, amine antioxidant, trioctyl borate, SIB, PIB}. Considering that trioctyl borate is a novel additive not directly appearing in the historical sample database, its configuration in a dilute solution (1% molar fraction) of PAO-6 base oil was sampled in a small batch of 50,000 steps using the open-source molecular simulation software LAMMPS. The average chain interaction potential energy between trioctyl borate and PAO-6 was estimated to be -2.3 kJ / mol. This value was added to the 1024-bit ECFP code of the trioctyl borate candidate node as the chain interaction attribute dimension, resulting in a 1025-dimensional trioctyl borate candidate molecular node code. The other four candidate nodes were generated with 1024-bit ECFP codes in the same way (PAO-6 and SIB are consistent with the historical end codes). All candidate node codes were concatenated end to end in M_c order to obtain the second molecular feature. The environmental characterization vector under test is based on... =[140, 2.5, 0.15, 1500, 0.21, 1.0] Extract, and The candidate recipe topology is constructed by splicing together nodes using a node-level constraint method. The initial form.

[0116] Step S400: The second molecular feature and the test environment representation vector are concatenated dimensionally to construct a candidate formulation topology, so as to calculate the topological structure difference between the candidate formulation topology and the historical formulation topology.

[0117] The second molecular feature generated in step S300 only completes the description of the molecular level of the oil product to be tested. This step needs to further couple the characterization vector of the test environment to construct a candidate formulation topology and measure its difference with the historical formulation topology, so as to locate the corresponding transferable structural regions for subsequent historical experience transfer. The following is combined with Figure 4 Please provide a detailed explanation.

[0118] Figure 4 This is a schematic diagram illustrating the geometric principles for calculating the topological differences between candidate formulation topologies and historical formulation topologies, and for locating shared connected regions, as provided in this application embodiment. Figure 4 As shown, candidate formulation topology 401 and historical formulation topology 402 are arranged side-by-side in a coplanar manner in three-dimensional conformation space; the micro-crosslinking density of candidate formulation topology 401 is calculated respectively. and the micro-crosslinking density of historical formula topology 402 The discrete span of the micro-crosslinking density was obtained. Furthermore, under the physical constraint term 403 composed of pressure load and heat distribution location, the spatial mesh deformation rate of candidate formula topology 401 is calculated respectively. And the spatial grid deformation rate of historical formula topology 402 ;Will and The structural distortion deviation 404 is synthesized according to a preset weighting method, and this structural distortion deviation is used as the topological structural difference; at the same time, based on the node-level difference distribution, sub-regions with differences less than the preset limit value are delineated on the historical recipe topology 402 as shared connected regions 405.

[0119] Specifically, the process of calculating the topological differences in this step includes:

[0120] First, candidate nodes characterizing the additives to be tested are extracted from the formulation design list of the oil to be tested to combine the second component sequence. This step is consistent with the corresponding process in step S300. Here, molecular stereoconfiguration information is introduced to supplement the structure mapping.

[0121] Secondly, based on the molecular stereoconfiguration and steric hindrance of the additive to be tested, the sequence of the second component is mapped to output a second molecular feature. The molecular stereoconfiguration includes, for example, chiral configurations R / S and cis / trans configurations E / Z; the steric hindrance is given by estimating the atomic occupancy within a preset radius (e.g., 0.5 nm) around the core functional groups of each candidate node using molecular dynamics software; the stereoconfiguration identifier and steric hindrance value are then appended to the local connectivity graph encoding of the candidate node to obtain the second molecular feature with stereoconfiguration. Introducing stereoconfiguration information can significantly improve the physical accuracy of subsequent micro-crosslink density comparisons.

[0122] Next, the operating pressure load and thermal distribution field are extracted from the environmental characterization vector to be tested, and the operating pressure load and thermal distribution field are used as physical constraints to perform dimensional concatenation with the second molecular feature to construct the candidate formulation topology. The thermal distribution field refers to the temperature gradient distribution within the oil space during the actual service of the oil product under test, which can be obtained by spatial interpolation of multi-point temperature sampling from the environmental condition acquisition unit; the dimensional concatenation method is the same as in step S200.

[0123] Next, the micro-crosslinking density of the candidate formulation topology and the micro-crosslinking density of the historical formulation topology are extracted, and the spatial mesh deformation rate of the candidate formulation topology and the historical formulation topology under the physical constraint term is compared. The micro-crosslinking density refers to the number density of chemically connected edges within a unit volume of molecular topology; the spatial mesh deformation rate refers to the average rate of change of relative length of the topological structure in each spatial direction relative to the unconstrained ground state configuration under the physical constraint term. The micro-crosslinking density can be directly obtained from the ratio of the number of edges in each locally connected graph to the volume occupied by the nodes; the spatial mesh deformation rate can be estimated by performing short-duration (e.g., 100 picoseconds to 1 nanosecond, depending on the software performance) relaxation simulation of the topological structure under the target temperature and pressure of the physical constraint term using molecular dynamics software, and using the normalized value of the root mean square value of each atomic displacement before and after relaxation.

[0124] Finally, based on the discrete span of the micro-crosslinking density and the spatial grid deformation rate, the structural distortion deviation between the candidate formulation topology and the historical formulation topology is calculated, and the structural distortion deviation is established as the topological difference. The structural distortion deviation is calculated using the following formula:

[0125] ;

[0126] Where D represents the structural distortion deviation. , The micro-crosslinking densities of the candidate formulation topology and the historical formulation topology are respectively represented. , These represent the corresponding spatial grid deformation rates. and These are preset weighting coefficients used to balance the relative importance of deviations between the molecular level and the service level; for example, they can be set to... The specific value can be adjusted according to the performance focus. The larger the structural distortion deviation D value, the greater the overall difference between the candidate formulation topology and the historical formulation topology in terms of molecular structure and environmental response behavior, and vice versa.

[0127] Building upon the aforementioned application scenarios, we further introduce the molecular stereochemistry of trioctyl borate (constrained trigonal pyramidal configuration, stereochemistry identifier set to 0) and the atomic occupancy (approximately 72%) within a 0.5 nanometer radius around its B atom as a supplementary dimension of the second molecular characteristic, thus enhancing the topology of the candidate formulation. It provides a more complete structural characterization. The historical formulation topology is analyzed using LAMMPS. Topology of candidate formulations 1-nanosecond relaxation simulations were performed under constraints of T=140℃ and P=2.5MPa, and statistical results were obtained. micro-crosslinking density =1.42 / nm³, spatial grid deformation rate =0.023, micro-crosslinking density =1.36 / nm³, spatial grid deformation rate =0.028. Use the preset weighting coefficient. =0.6、 Substituting 0.4 into the structural distortion deviation formula: This value is much smaller than the preset threshold for determining homologous samples (e.g., it can be set to 0.15), indicating that the candidate formulation topology deviates little from the historical formulation topology in terms of overall structure. This means that the historical sample of the 75W-90 heavy-duty gear oil has strong transferable reference value for the low-sulfur environmentally friendly gear oil formulation to be tested.

[0128] Step S500: Based on the topological differences, feature extraction is performed on the historical formulation topology to generate structure-activity mapping parameters corresponding to the oil to be tested.

[0129] The structural distortion deviation obtained in step S400 is only a measure of the overall difference between the candidate formulation topology and the historical formulation topology. To predict the performance of the test oil without re-testing on a bench, it is necessary to further locate local regions in the historical formulation topology that are highly similar to the candidate topology, and use the physical information of these regions as transferable underlying laws to map onto the multi-component thermodynamic system of the test oil. The following section will combine... Figure 5 Please provide a detailed explanation.

[0130] Figure 5 This is a schematic diagram illustrating the process of constructing a global microscopic potential energy surface and extracting structure-property mapping parameters based on the recombined node attributes of physical information, as provided in the embodiments of this application. Figure 5 As shown, the multi-component thermodynamic system 501 includes a base oil phase 501-1, an additive micelle phase 501-2, and an interface phase 501-3. The properties of each node in the shared connected region 502 after physical information reorganization are mapped to the multi-component thermodynamic system 501 to obtain the local potential energy shift distribution 503 of the nodes. The local potential energy shifts of each node are superimposed and extended in the spatial dimension to form a global microscopic potential energy surface 504. Several extreme stationary points 505 on the global microscopic potential energy surface 504 and the potential barrier height 506 connecting the extreme stationary points together constitute the structure-property mapping parameter 507.

[0131] Specifically, the process of generating the structure-property mapping parameters in this step includes:

[0132] First, based on the topological differences, shared connected regions in the historical formulation topology whose internal structural variations are less than a preset limit are located. Specifically, the node-level differences from step S400 are distributed across the historical formulation topology. For each node, the internal structural variation at its corresponding position in the candidate formulation topology is calculated (defined by the local change in the micro-crosslinking density between the node and its neighboring nodes). The largest connected subgraph formed by connecting all nodes whose internal structural variations are less than the preset limit (e.g., 0.1) is defined as the shared connected region. The preset limit can be adjusted according to the size of the historical sample and the diversity of the formulations.

[0133] Secondly, the stress response values ​​of each node within the shared connectivity region under the target high-temperature environment are read. These stress response values ​​refer to the bonded or unbonded stress values ​​recorded in the historical sample database for the compound corresponding to the node in the shared connectivity region under the target high-temperature environment, either through actual measurement or molecular dynamics simulation. Specifically, these values ​​can be obtained directly by querying existing measured stress records in the historical sample database, or by using molecular dynamics software to perform short-duration stress sampling on the compound fragment corresponding to the node under the target high-temperature environment. For example, the stress sampling method can be estimated using the Virial stress formula.

[0134] Next, based on the force response values, the attributes of each node within the shared connectivity region are physically reorganized. This physical information reorganization refers to using a preset transformation matrix to map the low-dimensional force response values ​​to a dimensional space with the same dimensions as the original attributes, generating a response feature vector of the same dimension. Then, the original attributes and the response feature vector are weighted and fused to obtain new node attributes carrying the physical service response. The specific calculation formula is as follows:

[0135] ;

[0136] in, This represents the new node attribute vector after physical information reorganization. Represents a node The original attribute vector (such as a 1025-dimensional high-dimensional feature tensor); Represents a node The force response vector (such as a one-dimensional scalar or a low-dimensional stress tensor). This represents a preset dimension transformation matrix, used to transform... Dimensional alignment to Same dimensions; and These represent the preset original attribute fusion weights and response feature fusion weights, respectively. The preset fusion weights can be calibrated based on the principle of minimizing regression error on a historical sample validation set; for example, they can be set as follows: =0.7、 =0.3. The purpose of physical information reorganization is to enable node attributes to simultaneously carry molecular structure information and service stress response, so as to avoid the loss of service-related physical quantities when migrating to the oil being tested.

[0137] Finally, the node attributes, after being reorganized from physical information, are mapped onto the multi-component thermodynamic system of the oil sample to be tested, so as to output the structure-property mapping parameters of the oil sample. The specific implementation process of the mapping includes:

[0138] Extracting the concentration distribution and microscopic phase boundaries of substances in a multi-component thermodynamic system. The concentration distribution refers to the molecular number density distribution of each component in the oil formulation to be tested, calculated according to its design proportion; the microscopic phase boundaries refer to the phase interfaces between the base oil phase, the additive micelle phase, and the interfacial phase, which can be obtained by fitting the density abrupt change surface observed in short-term molecular dynamics equilibrium simulations.

[0139] Based on the aforementioned substance concentration distribution and the aforementioned microscopic phase boundaries, the microscopic interaction energy of the nodal attributes after physical information reorganization within the multi-component thermodynamic system is analyzed. The microscopic interaction energy is obtained using the following formula:

[0140] ;

[0141] in, For nodes With nodes The microscopic interaction energy between them Preset the interaction potential function (for example, the Lanner-Jones potential superimposed with the Coulomb potential can be used). For nodes and Spatial distance between them and These are the node attributes after the physical information has been reorganized.

[0142] Based on the microscopic interaction energy, the microscopic collision frequency and heat conduction flux of each node under the target high-temperature environment are inferred. The microscopic collision frequency is estimated by substituting the microscopic interaction energy into the frequency formula of the Kinetic Collision Theory; the heat conduction flux is estimated by multiplying the partial derivative of the microscopic interaction energy along the temperature gradient direction by the molecular number density.

[0143] Based on the microscopic collision frequency, the amplitude of internal energy fluctuations caused by kinetic energy exchange at each node is calculated, and the local temperature gradient of the multi-component thermodynamic system is determined based on the heat conduction flux. The amplitude of internal energy fluctuations is obtained by multiplying the microscopic collision frequency by the average kinetic energy exchange per collision; the local temperature gradient is obtained by dividing the heat conduction flux by the system heat capacity.

[0144] The amplitude of the internal energy fluctuations and the local temperature gradient are substituted into the non-equilibrium thermodynamic model to deduce the local potential energy shifts at each node. The non-equilibrium thermodynamic model is specifically constructed based on the Onsag reciprocal relation of linear irreversible thermodynamics, and its specific derivation process is as follows:

[0145] First, the amplitude of the internal energy fluctuations is converted into a first generalized driving force characterizing the energy non-uniformity within a multi-component thermodynamic system. The local temperature gradient is transformed into a second generalized driving force characterizing the driving force of heat conduction. ;

[0146] Secondly, a set of linear phenomenological equations satisfying the Onsager reciprocal relation is constructed to describe energy flux. and material flow flux :

[0147] ;

[0148] ;

[0149] in, , , , These are phenomenological coefficients, and due to microscopic invertibility, their cross-term phenomenological coefficients satisfy Onsag reciprocal symmetry, i.e. (when When taking specific coupling terms);

[0150] Finally, under the boundary condition of a local steady state in the multi-component thermodynamic system (i.e., flux divergence is zero), the energy distribution in each space within the system is solved by simultaneously solving the aforementioned phenomenological equations, in order to calculate the local potential energy shift at each node caused by non-equilibrium thermodynamic transport. .

[0151] Based on the structural attachment relationships of each node within the multi-component thermodynamic system, the local potential energy shifts of each node are spatially superimposed and extended to construct a global microscopic potential energy surface. The extreme stationary points and barrier heights on this global microscopic potential energy surface are then extracted and converted into structure-property mapping parameters. The structural attachment relationship refers to the topological adjacency relationship established between nodes through covalent bonds, hydrogen bonds, and van der Waals interactions. Specifically, the spatial superposition and extension involves taking the space of the multi-component thermodynamic system as the domain, and for each spatial point, taking the potential energy value as the weighted sum of the local potential energy shifts of all adjacent nodes according to the structural attachment relationships to obtain the global potential energy value of that spatial point. Traversing the entire domain yields the global microscopic potential energy surface.

[0152] The extreme stationary points refer to local minima or saddle points where the first derivative of the global microscopic potential energy surface is zero. Each extreme stationary point corresponds to a stable or metastable microscopic configuration in the oil system. The potential barrier height refers to the highest potential energy value on the path connecting adjacent extreme stationary points minus the starting potential energy value, reflecting the energy threshold required for the transformation between two microscopic configurations. This application uses the spatial coordinates of all extreme stationary points and the set of potential barrier heights connecting them as the structure-property mapping parameters of the oil under test. The reason for using potential energy surface stationary points and potential barriers as structure-property mapping parameters is that the macroscopic properties of oil are physically controlled by the energy landscape of its microscopic system, and potential energy surface stationary points and potential barriers can most concisely express the causal path from microscopic configuration to macroscopic properties.

[0153] The following example illustrates the execution results of this step using the specific application examples described above. Continuing with the aforementioned application scenario, since the tested low-sulfur environmentally friendly gear oil formulation shares PAO-6 base oil and SIB extreme pressure additives with the historical formulation of 75W-90 heavy-duty gear oil, the historical formulation topology... Each node in the topology calculates its position in the candidate recipe. For the internal structural variations at corresponding locations, a preset limit value of 0.1 was used to connect nodes with variations less than this value. This ultimately delineated the shared connectivity region SR, which includes the PAO-6 base oil main chain node (node ​​1 and its surrounding temperature / pressure / heat conduction virtual constraint nodes) and the SIB sulfur bridge-SS-connection region node (node ​​4 and its surrounding virtual constraint nodes), totaling 8 molecular main nodes and their corresponding virtual constraint nodes. Using LAMMPS, short-duration Virial stress sampling was performed on each node within the SR at 140℃. The stress response values ​​for the PAO-6 base oil main chain node were approximately 5.2 MPa, and for the SIB sulfur bridge node, approximately 12.8 MPa. A preset fusion weight was then used. =0.7、 =0.3 Physical information is reorganized for each node attribute within SR to obtain new node attributes with service stress response. The recombined node attributes were mapped to the multi-component thermodynamic system of the oil sample, where the base oil phase mass fraction was 84%, the additive micelle phase mass fraction was approximately 5%, and the interface phase mass fraction was approximately 0.5%. The frequency of microscopic collisions (on the order of magnitude of approximately...) was estimated using Kingston collision theory. The local temperature gradient was estimated based on the system's heat capacity, and then the local potential energy shift at each node was derived by incorporating the Onsag reciprocal relation. Then, through spatial dimensional superposition and extension, a global microscopic potential energy surface is constructed. Finally, 17 extreme stationary points were identified on this potential energy surface (9 located in the base oil phase, 5 in the additive micelle phase, and 3 in the interface phase), along with 23 potential barriers connecting adjacent stationary points. The potential barriers are highly aggregated. The values ​​range from 40 kJ / mol to 310 kJ / mol. The set of stationary point coordinates and barrier height is used as the structure-activity mapping parameter of the oil to be tested.

[0154] Step S600: Perform nonlinear mapping of the physicochemical dimension on the structure-property mapping parameters, and output the predicted performance values ​​of the oil to be tested. The predicted performance values ​​include the kinematic viscosity value and anti-wear spot value of the oil to be tested.

[0155] Although the structure-property mapping parameters obtained in step S500 contain energy landscape information of the oil under test, they are still intermediate physical parameters. They need to be further converted into kinematic viscosity and anti-wear plaque values ​​that can be directly used by researchers through a nonlinear mapping pathway with clear physicochemical meaning. This application uses two parallel pathways to process bulk rheological properties and interfacial frictional properties respectively.

[0156] Specifically, the process of outputting the predicted performance value in this step includes:

[0157] First, the bond dissociation energy of the structure-property mapping parameters in the target high-temperature environment is analyzed. The bond dissociation energy refers to the Boltzmann weighted average of the potential barrier height set in the structure-property mapping parameters at the target high-temperature environment temperature, physically reflecting the effective energy required for bond breaking or configurational transitions of oil molecules at that temperature. The specific calculation method is as follows:

[0158] ;

[0159] in, For bond dissociation energy, The first parameter in the structure-effect mapping Barrier height, Boltzmann's constant, The target high-temperature environment temperature.

[0160] Secondly, based on the bond dissociation energy, the change in flow resistance of the tested oil under mechanical shear force is deduced. The change in flow resistance refers to the magnitude of the change in apparent viscosity of the oil relative to zero shear viscosity when subjected to mechanical shear. The specific deduction method is as follows: the bond dissociation energy is substituted into the shear flow model of Eyring's absolute reaction rate theory to obtain the exponential relationship between shear rate and flow resistance. Eyring's absolute reaction rate theory regards molecular flow as a chemical reaction process that overcomes a potential barrier, and its effective energy barrier is the bond dissociation energy. Shear stress can cause this energy barrier to decrease, thereby inducing non-Newtonian flow.

[0161] Next, the change in flow resistance is used as an input parameter and projected into the rheological constitutive model to output the kinematic viscosity value. The rheological constitutive model can be, for example, a Cross model, which has the following form:

[0162] ;

[0163] in, For kinematic viscosity, Shear rate, , These are zero-shear viscosity and infinite-shear viscosity, respectively. It is a time constant. For exponential parameters; as described in this application and It is calculated from the change in flow resistance using a pre-calibrated regression function. The kinematic viscosity is estimated by the ratio of bulk density to bond dissociation energy under the target high-temperature environment; finally, the kinematic viscosity is obtained by substituting the shear rate of the target under test into the Cross model.

[0164] Next, the boundary lubricating film strength, which characterizes the adsorption capacity of extreme pressure additives, is extracted from the structure-property mapping parameters. The boundary lubricating film strength is characterized by the potential well depth corresponding to the extreme stationary point located in the interfacial phase of the multi-component thermodynamic system in the structure-property mapping parameters. The deeper the potential well, the stronger the chemical adsorption of the extreme pressure additive on the metal surface, and the higher the strength of the boundary lubricating film formed.

[0165] Finally, based on the boundary lubricating film strength, the contact state of the micro-protrusions on the metal contact surface is derived to output the anti-wear scar value. The micro-protrusion contact state is described using the Greenwood-Williamson rough contact model, which determines whether the micro-protrusion is in one of three states: elastic contact, plastic contact, or adhesive wear, based on the relative magnitude of the boundary lubricating film strength and the applied load. The specific derivation process is as follows: the boundary lubricating film strength is used as the yield limit of the contact interface as input to the model. Combined with the contact load under the test condition, the proportion of plastically deformed micro-protrusions per square centimeter of the contact area is calculated. This proportion is then converted into the anti-wear scar value of the tested oil through an anti-wear scar regression function (e.g., based on statistical regression of the boundary lubricating film strength and wear scar diameter in historical samples). The anti-wear scar value numerically represents the predicted wear scar diameter in the four-ball friction test, with units such as millimeters.

[0166] The execution results of this step are illustrated below using the specific application examples described above. Continuing with the aforementioned application scenario, the 23 barrier heights in the structure-property mapping parameters are... Substituting into the Boltzmann weighted formula for bond dissociation energy (taking T = 413.15 K, i.e., the Kelvin temperature corresponding to 140 °C, and the Boltzmann constant), =8.314 J / (mol·K)), the bond dissociation energy of the oil sample to be tested was calculated. ≈240kJ / mol.

[0167] Substituting the bond dissociation energy into the shear flow model of Eyring's absolute reaction rate theory, the shear rate is derived. The change in flow resistance is approximately -12% of the zero shear viscosity (i.e., exhibiting shear thinning behavior). The change in flow resistance is further converted into Cross model parameters using a pre-calibrated regression function (taking...). , , , ), with the target shear rate to be measured Substituting into the Cross model yields the kinematic viscosity. The kinematic viscosity was approximately 18 mPa·s (equivalent to a kinematic viscosity of approximately 14.5 mm² / s at 100℃). In the anti-wear scar pathway, the average potential well depth corresponding to the three extreme stagnation points within the interface phase of the multi-component thermodynamic system was approximately 1.2 GPa. This was used as the boundary lubrication film strength input to the Greenwood-Williamson rough contact model. Combined with a 4 MPa contact load, the proportion of plastic deformation micro-protrusions was calculated to be approximately 6%. After transformation using the anti-wear scar regression function, the predicted diameter of the four-ball friction wear scar of the tested oil was approximately 0.41 mm. The relative error between this prediction result (kinematic viscosity 14.5 mm² / s, wear scar diameter 0.41 mm) and the performance results obtained from subsequent physical preparation and testing of the candidate formulation (measured kinematic viscosity 14.6 mm² / s, measured wear scar diameter 0.42 mm) was less than 2.5%, verifying the effectiveness of the method in this application for predicting oil formulation performance.

[0168] In summary, the embodiments of this application construct a historical formulation topology by mapping the component sequence structure and splicing the environmental characterization vector dimension of historical oil formulations. Then, for the oil to be tested, candidate formulation topologies are constructed according to the same paradigm, the differences in topological structure are calculated, shared connected regions are extracted, and structure-property mapping parameters are generated by physical information recombination and global microscopic potential energy surface construction. Finally, kinematic viscosity values ​​and anti-wear spot values ​​are simultaneously output through a dual-pathway system of bond dissociation energy-rheological constitutive model and boundary lubrication film strength-rough contact model. Ultimately, without relying on a large number of bench tests, a physically interpretable and cross-formulation generalizable quantitative prediction of the performance of the oil formulation to be tested is achieved.

[0169] Example 2

[0170] The following is combined Figure 6 The oil formulation performance prediction system based on historical sample data provided in the embodiments of this application will be described. Figure 6 This is a schematic diagram of the structure of an oil formulation performance prediction system based on historical sample data, provided as an embodiment of this application. Figure 6As shown, the oil formulation performance prediction system 600 based on historical sample data includes an acquisition module 601, a first mapping module 602, a second mapping module 603, a construction module 604, a feature extraction module 605, and a nonlinear mapping module 606.

[0171] The acquisition module 601 is used to acquire the first component sequence of historical oil formulations and the environmental characterization vector of historical oil formulations under a target high-temperature environment. The acquisition module 601 is communicatively connected to the historical sample database and the environmental condition acquisition unit. Its specific implementation is the same as the aforementioned method step S100, and will not be repeated here.

[0172] The first mapping module 602 is used to perform structure mapping processing on the first component sequence to generate a first molecular feature, and to concatenate the first molecular feature with the environmental characterization vector to construct a historical formulation topology. The input of the first mapping module 602 is connected to the output of the acquisition module 601, and its specific implementation is the same as the aforementioned method step S200.

[0173] The second mapping module 603 is used to acquire the second component sequence of the oil to be tested and the characterization vector of the environment to be tested, and to perform structural mapping processing on the second component sequence to generate the second molecular feature. The input end of the second mapping module 603 is connected to the interactive terminal of the R&D personnel, and its specific implementation method is the same as the aforementioned method step S300.

[0174] The construction module 604 is used to concatenate the second molecular feature with the test environment representation vector to construct a candidate formulation topology, so as to calculate the topological structure difference between the candidate formulation topology and the historical formulation topology. The input of the construction module 604 is connected to the output of the first mapping module 602 and the second mapping module 603 respectively, and its specific implementation is the same as the aforementioned method step S400.

[0175] The feature extraction module 605 is used to extract features from the historical formulation topology based on the topological differences, and generate structure-property mapping parameters corresponding to the oil to be tested. The input of the feature extraction module 605 is connected to the output of the construction module 604, and its specific implementation is the same as the aforementioned method step S500.

[0176] The nonlinear mapping module 606 is used to perform nonlinear mapping of the structure-property mapping parameters along the physicochemical dimensions, and outputs predicted performance values ​​of the oil under test. These predicted performance values ​​include the kinematic viscosity and anti-wear plaque values ​​of the oil under test. The input of the nonlinear mapping module 606 is connected to the output of the feature extraction module 605, and the output of the nonlinear mapping module 606 is connected to the interactive terminal for R&D personnel. Its specific implementation is the same as the aforementioned method step S600.

[0177] This application also provides a computing device, including a processing component and a storage component. The processing component may be implemented using a general-purpose microprocessor, digital signal processor, or field-programmable gate array (FPGA); the storage component may be implemented using a non-volatile memory. The storage component stores one or more computer instructions, which are invoked and executed by the processing component to implement the oil formulation performance prediction method based on historical sample data described in any of the above embodiments of this application.

[0178] This application also provides a computer storage medium storing a computer program. When the computer program is executed by a computer, it implements the oil formulation performance prediction method based on historical sample data described in any of the above embodiments of this application. The computer storage medium can be any physical medium capable of storing a computer program, such as a read-only memory, random access memory, disk, or optical disk.

[0179] The specific embodiments described above further illustrate the purpose, technical solution, and beneficial effects of this application. It should be understood that the above description is only a specific embodiment of this application and is not intended to limit the scope of protection of this application. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the scope of protection of this application.

Claims

1. A method for predicting the performance of oil formulations based on historical small sample data, characterized in that, include: Obtain the first component sequence of historical oil formulations and the environmental characterization vector of historical oil formulations under target high-temperature conditions; The first component sequence is subjected to structural mapping processing to generate a first molecular feature, and the first molecular feature is concatenated with the environmental characterization vector to construct a historical formula topology. The second component sequence of the oil to be tested and the characterization vector of the environment to be tested are obtained, and the second component sequence is subjected to structural mapping processing to generate the second molecular feature; The second molecular feature is concatenated with the characterization vector of the environment to be tested to construct a candidate formulation topology, and the topological structure difference between the candidate formulation topology and the historical formulation topology is calculated. Based on the topological differences, features are extracted from the historical formulation topology to generate structure-activity mapping parameters corresponding to the oil to be tested. The structure-property mapping parameters are nonlinearly mapped along the physicochemical dimension to output the predicted performance values ​​of the oil under test. The predicted performance values ​​include the kinematic viscosity and anti-wear plaque values ​​of the oil under test.

2. The method according to claim 1, characterized in that, The first component sequence is subjected to structure mapping processing to generate a first molecular feature, and the first molecular feature is concatenated with the environmental representation vector to construct a historical formulation topology, including: Analyze the covalent bond information between adjacent atoms in each compound node of historical oil formulations; Based on the covalent bond information, the local connectivity graph of each compound node is extracted as the first molecular feature; Extract the thermodynamic boundary carried in the environmental characterization vector; The thermodynamic boundary is applied as an external constraint to the locally connected graph to construct the historical recipe topology.

3. The method according to claim 1, characterized in that, The second component sequence of the oil sample to be tested and the characterization vector of the test environment are obtained. The second component sequence is then subjected to structure mapping processing to generate second molecular features, including: Candidate nodes characterizing the additives to be tested are extracted from the formulation design list of the oil to be tested in order to combine the second component sequence. Based on the chain interaction properties of the additive to be tested, the second component sequence is physically mapped to output the second molecular feature; The pressure load and thermal conductivity in the environmental characterization vector to be tested are concatenated with the second molecular feature to construct a candidate formulation topology.

4. The method according to claim 1, characterized in that, The second molecular feature is concatenated dimensionally with the test environment representation vector to construct a candidate formulation topology. The topological differences between the candidate formulation topology and the historical formulation topology are then calculated, including: Obtain a list of formulation designs for the oil to be tested, and extract candidate nodes characterizing the additives to be tested from the list of formulation designs to combine them into a second component sequence; Based on the molecular stereoconfiguration and steric hindrance of the additive to be tested, the sequence of the second component is mapped to output the second molecular feature; The operating pressure load and thermal distribution field are extracted from the characterization vector of the environment under test, and the operating pressure load and thermal distribution field are used as physical constraints to be dimensionally concatenated with the second molecular feature to construct the candidate formulation topology; Extract the micro-crosslinking density of the candidate formulation topology and the micro-crosslinking density of the historical formulation topology, and compare the spatial mesh deformation rate of the candidate formulation topology and the historical formulation topology under physical constraints; Based on the discrete span of the micro-crosslinking density and the spatial grid deformation rate, the structural distortion deviation between the candidate formulation topology and the historical formulation topology is calculated, and the structural distortion deviation is established as the topological difference.

5. The method according to claim 1, characterized in that, Based on the aforementioned topological differences, feature extraction is performed on the historical formulation topology to generate structure-property mapping parameters corresponding to the oil to be tested, including: Based on the topological differences, locate the shared connected regions in the historical recipe topology whose internal structural changes are less than a preset limit value; Read the force response values ​​of each node in the shared connectivity region under the action of the target high-temperature environment; Based on the force response value, the attributes of each node in the shared connected region are reorganized using physical information. The node attributes, after being reorganized by physical information, are mapped to the multi-component thermodynamic system of the oil sample to be tested, so as to output the structure-property mapping parameters of the oil sample to be tested.

6. The method according to claim 1, characterized in that, The node attributes, after being reorganized from physical information, are mapped to the multi-component thermodynamic system of the oil sample to be tested, so as to output the structure-property mapping parameters of the oil sample, including: Extracting the concentration distribution of substances and the microscopic phase boundaries in a multi-component thermodynamic system; Based on the substance concentration distribution and the microscopic phase boundary, the microscopic interaction energy of the node attributes after physical information recombination within the multi-component thermodynamic system is analyzed. Based on the aforementioned micro-interaction energy, the frequency of micro-collisions and the heat conduction flux of each node under the target high-temperature environment are inferred. Based on the micro-collision frequency, the amplitude of internal energy fluctuations at each node due to kinetic energy exchange is calculated, and the local temperature gradient of the multi-component thermodynamic system is determined based on the heat conduction flux. Substitute the internal energy fluctuation amplitude and the local temperature gradient into the non-equilibrium thermodynamic model to deduce the local potential energy shift of each node; Based on the structural attachment relationship of each node in the multi-component thermodynamic system, the local potential energy shift of each node is spatially superimposed and extended to construct a global microscopic potential energy surface, and the extreme stationary points and barrier heights on the global microscopic potential energy surface are extracted and converted into structure-property mapping parameters.

7. The method according to claim 1, characterized in that, A nonlinear mapping of the physicochemical dimensions is performed on the structure-property mapping parameters to output the predicted performance values ​​of the oil under test. These predicted performance values ​​include the kinematic viscosity and anti-wear plaque values ​​of the oil under test, including: Analyze the bond dissociation energy of the structure-property mapping parameters in the target high-temperature environment; Based on the bond dissociation energy, the change in flow resistance of the oil under mechanical shear force is deduced; The change in flow resistance is used as an input parameter and projected into the rheological constitutive model to output the kinematic viscosity value; Extract the boundary lubricating film strength, which characterizes the adsorption capacity of extreme pressure additives, from the structure-property mapping parameters; Based on the boundary lubricating film strength, the micro-protrusion contact state of the metal contact surface is derived to output the anti-wear spot value.

8. A system for predicting the performance of oil formulations based on historical sample data, characterized in that, include: The acquisition module is used to acquire the first component sequence of historical oil formulations and the environmental characterization vector of historical oil formulations under the target high-temperature environment. The first mapping module is used to perform structural mapping processing on the first component sequence to generate a first molecular feature, and to concatenate the first molecular feature with the environmental characterization vector to construct a historical formula topology. The second mapping module is used to obtain the second component sequence of the oil to be tested and the characterization vector of the environment to be tested, and to perform structural mapping processing on the second component sequence to generate the second molecular feature; A construction module is used to concatenate the second molecular feature with the characterization vector of the environment to be tested to construct a candidate formulation topology, and to calculate the topological structure difference between the candidate formulation topology and the historical formulation topology. The feature extraction module is used to extract features from the historical formulation topology based on the topological differences, and generate structure-property mapping parameters corresponding to the oil to be tested. The nonlinear mapping module is used to perform nonlinear mapping of the structure-property mapping parameters in the physicochemical dimension and output the predicted performance values ​​of the oil under test. The predicted performance values ​​include the kinematic viscosity value and anti-wear spot value of the oil under test.

9. A computing device, characterized in that, It includes a processing component and a storage component; the storage component stores one or more computer instructions; the one or more computer instructions are invoked and executed by the processing component to implement the oil formulation performance prediction method based on historical sample data as described in any one of claims 1-7.

10. A computer storage medium, characterized in that, The system contains a computer program that, when executed by a computer, implements a method for predicting the performance of oil formulations based on historical sample data as described in any one of claims 1-7.

Images