Formula process extrapolation method, device and equipment and storage medium

Abstract

The invention discloses a formula process extrapolation method, device and equipment and a storage medium, and relates to the technical field of industrial big data, and the method comprises the steps: obtaining multi-source formula process data, and constructing a knowledge graph; a hypergraph extrapolation model is constructed according to the knowledge graph, extrapolation is carried out according to the hypergraph extrapolation model, the extrapolation score of each hyperedge is obtained, and the hypergraph extrapolation model comprises hypergraph nodes and hyperedges; generating a formula process extrapolation parameter according to the extrapolation score, and performing confrontation verification on the formula process extrapolation parameter; and when the confrontation verification is passed, obtaining new formula data and new process data. According to the method, the knowledge graph is constructed according to the multi-source formula process data, data islands are broken, and the relevance between the data is enhanced; according to the method, the hypergraph extrapolation model is constructed according to the knowledge graph, the formula process extrapolation parameters are generated through extrapolation of the hypergraph extrapolation model, and knowledge drives hypergraph extrapolation, so that cross-domain data extrapolation generation can be realized, and the efficiency and accuracy of formula process extrapolation are improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of industrial big data, and particularly to a formula process extrapolation method, device, equipment and storage medium. BACKGROUND

[0002] In intelligent industrial manufacturing, the existing formula optimization relies on expert experience, and the research and development cycle is as long as 6-12 months. There is a lack of systematic correlation among process parameters, material properties and performance indicators, and there is a data island problem. When data is generated through intelligent algorithms (such as genetic algorithms), it is easy to fall into a local optimal solution in a complex formula combination, and it is difficult to migrate across fields, and different industry formula knowledge is difficult to intercommunicate, such as medical formula cannot guide battery material development, and the accuracy of new data generation is low. SUMMARY

[0003] The main purpose of the present application is to provide a formula process extrapolation method, device, equipment and storage medium, which aims to solve the technical problem of low accuracy of new data generation in the existing formula process extrapolation method.

[0004] To achieve the above-mentioned purpose, the present application provides a formula process extrapolation method, which comprises:

[0005] Obtaining multi-source formula process data, and constructing a knowledge graph according to the multi-source formula process data;

[0006] According to the knowledge graph, a hypergraph extrapolation model is constructed, and extrapolation is performed according to the hypergraph extrapolation model to obtain an extrapolation score of each hyperedge, wherein the hypergraph extrapolation model comprises hypergraph nodes and hyperedges;

[0007] According to the extrapolation score, formula process extrapolation parameters are generated, and the formula process extrapolation parameters are subjected to adversarial verification;

[0008] When the adversarial verification is passed, new formula data and new process data are obtained.

[0009] In an embodiment, the step of constructing a hypergraph extrapolation model according to the knowledge graph, performing extrapolation according to the hypergraph extrapolation model, and obtaining an extrapolation score of each hyperedge comprises:

[0010] Mapping the knowledge graph to hypergraph nodes, wherein the hypergraph nodes comprise at least one of raw material entity nodes, process parameter nodes and performance indicator nodes;

[0011] Connecting the hypergraph nodes to obtain hyperedges;

[0012] According to a preset target and the hyperedges, multi-hop extrapolation is performed, the hypergraph nodes are reversely searched, and a confidence value of the hyperedges is obtained;

[0013] determining an edge weight of the hyperedge according to the confidence value;

[0014] determining an extrapolation score of each of the hyperedges according to the hypergraph nodes and the edge weight.

[0015] In an embodiment, the step of generating a recipe process extrapolation parameter according to the extrapolation score and performing adversarial verification on the recipe process extrapolation parameter comprises:

[0016] generating a recipe process extrapolation parameter according to the extrapolation score;

[0017] performing adversarial verification on the recipe process extrapolation parameter according to a physical constraint condition, the physical constraint condition comprising a mass conservation law condition, an energy spontaneity condition and a toxicity compliance condition;

[0018] when the recipe process extrapolation parameter satisfies the mass conservation law condition, the energy spontaneity condition and the toxicity compliance condition simultaneously, it means that the adversarial verification of the recipe process extrapolation parameter is passed.

[0019] In an embodiment, the step of performing adversarial verification on the recipe process extrapolation parameter according to a physical constraint condition comprises:

[0020] when the physical constraint condition is the mass conservation law condition, determining a total mass of a product mass and a byproduct mass according to the recipe process extrapolation parameter;

[0021] determining a total mass of reactants corresponding to the recipe process extrapolation parameter;

[0022] judging whether a difference between the total mass and the total mass of reactants is less than a preset mass threshold value;

[0023] if the difference is less than the preset mass threshold value, it means that the recipe process extrapolation parameter satisfies the mass conservation law condition.

[0024] In an embodiment, the step of performing adversarial verification on the recipe process extrapolation parameter according to a physical constraint condition further comprises:

[0025] when the physical constraint condition is the energy spontaneity condition, determining an enthalpy change and an entropy change generated by the recipe process extrapolation parameter in a preset reaction process;

[0026] determining a free energy change value according to the enthalpy change and the entropy change;

[0027] when the free energy change value is negative, it means that the recipe process extrapolation parameter satisfies the energy spontaneity condition.

[0028] In an embodiment, the step of adversarially verifying the formula process extrapolation parameter according to the physical constraint condition further comprises:

[0029] When the physical constraint condition is a toxicity compliance condition, a toxicity score of the formula process extrapolation parameter is obtained according to a preset toxicity regulation library, and a toxicity score value is obtained.

[0030] When the toxicity score value is greater than a preset toxicity threshold value, it indicates that the formula process extrapolation parameter meets the toxicity compliance condition.

[0031] In an embodiment, the step of obtaining multi-source formula process data and constructing a knowledge graph according to the multi-source formula process data comprises:

[0032] Multi-modal fusion is performed on the obtained multi-source data to obtain multi-source formula process data.

[0033] Entity extraction is performed on the multi-source formula process data to obtain an initial graph.

[0034] Nonlinear association relationship reasoning is performed on the initial graph to obtain a knowledge graph.

[0035] In addition, to achieve the above-mentioned purposes, the present application also proposes a formula process extrapolation device, which comprises:

[0036] A graph construction module is configured to obtain multi-source formula process data and construct a knowledge graph according to the multi-source formula process data.

[0037] A hypergraph extrapolation module is configured to construct a hypergraph extrapolation model according to the knowledge graph, perform extrapolation according to the hypergraph extrapolation model, and obtain an extrapolation score of each hyperedge, wherein the hypergraph extrapolation model comprises hypergraph nodes and hyperedges.

[0038] A parameter generation module is configured to generate a formula process extrapolation parameter according to the extrapolation score and adversarially verify the formula process extrapolation parameter.

[0039] An adversarial verification module is configured to obtain new formula data and new process data when the adversarial verification is passed.

[0040] In addition, to achieve the above-mentioned purposes, the present application also proposes a formula process extrapolation device, which comprises a memory, a processor, and a computer program stored on the memory and executable on the processor, wherein the computer program is configured to implement the steps of the formula process extrapolation method as described above.

[0041] In addition, to achieve the above-mentioned purpose, the present application also provides a storage medium, which is a computer readable storage medium, and a computer program is stored on the storage medium, and the computer program realizes the steps of the formula process extrapolation method when executed by a processor.

[0042] In addition, to achieve the above-mentioned purpose, the present application also provides a computer program product, which comprises a computer program, and the computer program realizes the steps of the formula process extrapolation method when executed by a processor.

[0043] The present application provides a formula process extrapolation method, which obtains multi-source formula process data, and constructs a knowledge graph according to the multi-source formula process data; constructs a hypergraph extrapolation model according to the knowledge graph, and performs extrapolation according to the hypergraph extrapolation model to obtain an extrapolation score of each hyperedge, wherein the hypergraph extrapolation model comprises hypergraph nodes and hyperedges; generates formula process extrapolation parameters according to the extrapolation score, and performs adversarial verification on the formula process extrapolation parameters; when the adversarial verification is passed, new formula data and new process data are obtained. The present application breaks the data island and enhances the correlation between data by constructing a knowledge graph according to multi-source formula process data; constructs a hypergraph extrapolation model according to the knowledge graph, and the hypergraph extrapolation model generates formula process extrapolation parameters, and the knowledge drives the hypergraph extrapolation, so as to realize cross-domain data extrapolation generation, and improve the efficiency and accuracy of formula process extrapolation. BRIEF DESCRIPTION OF DRAWINGS

[0044] The accompanying drawings, which are incorporated into and constitute a part of the specification, illustrate embodiments consistent with the present application and, together with the specification, serve to explain the principles of the present application.

[0045] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the accompanying drawings needed to be used in the embodiments or prior art description will be briefly introduced. Obviously, for those skilled in the art, other drawings can also be obtained without creative labor based on these drawings.

[0046] Figure 1 A flowchart is provided for the formula process extrapolation method embodiment one of the present application;

[0047] Figure 2 A system architecture diagram is provided for the formula process extrapolation method of the present application;

[0048] Figure 3 A flowchart is provided for the formula process extrapolation method embodiment two of the present application;

[0049] Figure 4 An example diagram of the hypergraph generated by the formula process extrapolation method of the present application is provided;

[0050] Figure 5 The flowchart provided for the third embodiment of the formula process extrapolation method of the present application;

[0051] Figure 6 The module structure diagram of the formula process extrapolation device of the embodiment of the present application;

[0052] Figure 7 The device structure diagram of the hardware running environment involved in the formula process extrapolation method in the embodiment of the present application.

[0053] The purpose implementation, functional characteristics and advantages of the present application will be further described with reference to the accompanying drawings in conjunction with the embodiments. DETAILED DESCRIPTION

[0054] It should be understood that the specific embodiments described herein are only used to explain the technical solutions of the present application, and are not used to limit the present application.

[0055] In order to better understand the technical solutions of the present application, the following will be described in detail in conjunction with the drawings and specific embodiments of the specification.

[0056] The main solution of the embodiment of the present application is: acquiring multi-source formula process data, and constructing a knowledge graph according to the multi-source formula process data; constructing a hypergraph extrapolation model according to the knowledge graph, and performing extrapolation according to the hypergraph extrapolation model to obtain an extrapolation score of each hyperedge, the hypergraph extrapolation model including hypergraph nodes and hyperedges; generating formula process extrapolation parameters according to the extrapolation score, and performing adversarial verification on the formula process extrapolation parameters; when the adversarial verification is passed, obtaining new formula data and new process data.

[0057] In the intelligent industrial manufacturing of the prior art, the existing formula optimization relies on expert experience, and the research and development cycle is as long as 6-12 months. There is a lack of systematic correlation among process parameters, material properties and performance indicators, and there is a data island problem. When data is generated through intelligent algorithms (such as genetic algorithms), it is easy to fall into a local optimal solution in complex formula combination, and it is difficult to migrate across fields, and different industry formula knowledge is difficult to intercommunicate, such as medical formula cannot guide battery material development, and the accuracy of new data generation is low.

[0058] The application provides a solution of acquiring multi-source formula process data, constructing a knowledge graph according to the multi-source formula process data, constructing a hypergraph extrapolation model according to the knowledge graph, extrapolating according to the hypergraph extrapolation model to obtain an extrapolation score of each hyperedge, the hypergraph extrapolation model including hypergraph nodes and hyperedges, generating formula process extrapolation parameters according to the extrapolation score, and performing adversarial verification on the formula process extrapolation parameters. When the adversarial verification is passed, new formula data and new process data are obtained. The application breaks the data island and enhances the correlation between data by constructing a knowledge graph according to multi-source formula process data. The hypergraph extrapolation model is constructed according to the knowledge graph, the formula process extrapolation parameters are generated by extrapolation, the knowledge drives the hypergraph extrapolation, and thus cross-field data extrapolation generation can be realized, and the efficiency and accuracy of formula process extrapolation are improved.

[0059] It should be noted that the execution subject of the method of the embodiment can be a computing service device with formula process extrapolation, network communication and program running functions, such as a tablet computer, a personal computer, a mobile phone and the like; or a formula process extrapolation device with the same or similar functions. The embodiment and the following embodiments will be described by taking the formula process extrapolation device as an example.

[0060] Based on this, the application embodiment provides a formula process extrapolation method, which is described below with reference to Figure 1 , Figure 1 The flowchart of the first embodiment of the formula process extrapolation method of the application is shown in FIG. 1.

[0061] In the embodiment, the formula process extrapolation method includes steps S10-S40:

[0062] Step S10, acquiring multi-source formula process data, and constructing a knowledge graph according to the multi-source formula process data.

[0063] It should be noted that the application belongs to the cross field of intelligent manufacturing and material science, and specifically relates to a system and method for realizing formula process optimization through knowledge graph construction, multi-source data fusion and dynamic extrapolation algorithm, which is suitable for industrial production scenes such as chemical materials, biological pharmaceuticals, food processing and the like that need formula iteration and upgrading. The architecture of the system includes a multi-source data input layer, a knowledge graph construction engine, a cross-latitude extrapolation module, an output and feedback layer. The architecture diagram can be referred to as shown in FIG. 2. Figure 2

[0064] It can be understood that the multi-source data input layer acquires multi-source formula process data such as a material database (such as PubChem), a formula process patent library, experimental reports, equipment sensor data and the like. The multi-source formula process data is input to the knowledge graph construction engine, the multi-source formula process data is subjected to entity extraction, and the knowledge graph is constructed.

[0065] ​In a feasible implementation, step S10 can include steps S101-S103:

[0066] Step S101, multi-modal fusion is performed on the obtained multi-source data to obtain multi-source formula process data.

[0067] It can be understood that the multi-source data input layer integrates the material database, the formula process patent library, the experimental report, the equipment sensor data and other multi-source data, performs multi-modal data fusion, and obtains the multi-source formula process data.

[0068] Step S102, entity extraction is performed on the multi-source formula process data to obtain an initial graph.

[0069] It should be understood that the knowledge graph construction engine performs entity extraction on the above multi-source formula process data to construct a knowledge graph.

[0070] Step S103, non-linear correlation reasoning is performed on the initial graph to obtain a knowledge graph.

[0071] It should be noted that the implicit non-linear correlation in the knowledge graph can also be mined by a graph neural network (GNN), for example, the non-linear relationship between the concentration of a catalyst reaction temperature product porosity can be mined. A real-time updating mechanism is also provided: every 24 hours, the latest scientific research papers and patent data worldwide are automatically captured to dynamically update the knowledge graph.

[0072] In the embodiment, the multi-modal fusion of the multi-source data is performed to ensure the richness of the data; the initial graph is obtained through entity extraction, and then the implicit non-linear correlation in the knowledge graph is mined by the graph neural network, so that a more rich and accurate knowledge graph can be constructed.

[0073] Step S20, constructing a hypergraph extrapolation model according to the knowledge graph, and performing extrapolation according to the hypergraph extrapolation model to obtain an extrapolation score of each hyperedge, wherein the hypergraph extrapolation model includes hypergraph nodes and hyperedges.

[0074] It can be understood that the hypergraph nodes and hyperedges can be generated from the knowledge graph by a cross-latitude extrapolation module to construct a hypergraph extrapolation model. The hypergraph extrapolation model realizes multi-hop reasoning through hyperedges. The multi-hop reasoning process is to search relevant processes and raw materials in the hypergraph nodes reversely to obtain the extrapolation score of each hyperedge for a given performance target.

[0075] Step S30, generating formula process extrapolation parameters according to the extrapolation score, and performing adversarial verification on the formula process extrapolation parameters.

[0076] It should be understood that the super-edge with a higher extrapolation score can also be selected by the cross-latitude extrapolation module, and a new recipe and a new process are generated corresponding to the super-graph node, that is, the recipe process extrapolation parameter is obtained. In order to improve the data effectiveness of the generated recipe process extrapolation parameter, the recipe process extrapolation parameter can also be subjected to adversarial verification to verify the feasibility of the data.

[0077] It should be noted that the supergraph generated according to the knowledge graph is a double-channel, that is, the influence between each supergraph node in the knowledge graph association graph is bidirectional, and the weight corresponding to each superedge is different. Based on different directions, both positive effects and negative effects can be represented, and the weight is optimized under different effect conditions. For example, process-performance decoupling optimization can be performed by a double-channel optimizer. Channel 1: formula element replacement based on knowledge graph, for example, replacing toxic ingredients with environmentally friendly materials; Channel 2: process parameter quantization adjustment, for example, discretizing continuous variables such as temperature / time into executable instructions.

[0078] Step S40, when the adversarial verification passes, obtaining new recipe data and new process data.

[0079] It can be understood that after the adversarial verification passes and the double-channel decoupling optimization is performed, the new recipe data and the new process data are output and fed back to the input layer. For example, new recipe: LiFSI-EC / EMC / FEC (3:5:2); new process: gradient temperature control (45℃→55℃, +2℃ per cycle); predicted performance: conductivity 118%, cycle life 1-800 times, etc.

[0080] In addition, expert feedback reinforcement learning can also be performed through the feedback layer. When the system generates a new recipe that is manually modified, the modification path is automatically traced back and the graph weight is updated. The entity relationship weight in the knowledge graph is updated through backward propagation of experimental data:

[0081]

[0082] ΔP k = P k - P k-1 k represents the performance change in the kth experiment.

[0083] The embodiment provides a formula process extrapolation method, multi-source formula process data is acquired, and a knowledge graph is constructed according to the multi-source formula process data; a hypergraph extrapolation model is constructed according to the knowledge graph, extrapolation is performed according to the hypergraph extrapolation model, an extrapolation score of each hyperedge is obtained, and the hypergraph extrapolation model includes hypergraph nodes and hyperedges; formula process extrapolation parameters are generated according to the extrapolation score, and the formula process extrapolation parameters are subjected to adversarial verification; when the adversarial verification is passed, new formula data and new process data are obtained. According to the multi-source formula process data, the knowledge graph is constructed, the data island is broken, and the correlation between data is enhanced; the hypergraph extrapolation model is constructed according to the knowledge graph, the formula process extrapolation parameters are generated by extrapolation of the hypergraph extrapolation model, the knowledge drives the hypergraph extrapolation, so that cross-domain data extrapolation generation can be realized, and the efficiency and accuracy of formula process extrapolation are improved.

[0084] Based on the first embodiment of the present application, in the second embodiment of the present application, the same or similar contents as the above embodiment one can refer to the above introduction, and the subsequent will not be described in detail. On this basis, please refer to Figure 3 , step S20, the formula process extrapolation method further includes steps S201-S205:

[0085] Step S201, the knowledge graph is mapped into a hypergraph node, and the hypergraph node includes at least one of a raw material entity node, a process parameter node and a performance index node.

[0086] Step S202, the hypergraph node is connected to obtain a hyperedge.

[0087] It should be noted that the formula elements (raw materials, processes) in the knowledge graph and the performance index can be mapped into a hypergraph node to obtain a hypergraph node including a raw material entity node, a process parameter node and a performance index node. Then the above hypergraph node is connected to form a hyperedge.

[0088] Step S203, according to the preset target and the hyperedge, multi-hop extrapolation is performed, the hypergraph node is searched reversely, and a confidence value of the hyperedge is obtained.

[0089] Step S204, the edge weight of the hyperedge is determined according to the confidence value.

[0090] Step S205, the extrapolation score of each hyperedge is determined according to the hypergraph node and the edge weight.

[0091] It should be noted that multi-hop reasoning can be realized through the hyperedge, the multi-hop reasoning process is to give a preset target (for example, the performance target can be “improve hardness”), the related process and raw material in the hypergraph node are searched reversely, the confidence value of the hyperedge is given, the weight of the hyperedge is determined according to the confidence value, and the high weight path is preferentially selected in the reasoning process. The reasoning calculation formula is as follows:

[0092]

[0093] wherein h i represents a hypergraph node, ε hyper represents a hyperedge, a i represents the weight of each hyperedge.

[0094] In an example, the hypergraph generated by the present application can refer to Figure 4 As shown, the multi-dimensional association of the knowledge graph is visually displayed in the form of a visual decision tree. Regarding the node system (hypergraph node), Figure 4 The red circles represent raw material entities, including lithium salt additives (LiPF6 / LiFSI), carbonate solvents (DMC / EC / EMC / FEC), and other core materials. The blue circles represent process parameters, such as temperature gradient control (45°C→55°C) and mixing time (120 min), which are key process conditions. The green circles represent performance indicators, such as electrical conductivity (12-18 mS / cm) and cycle life (500→800 times), which are target parameters. Regarding the association representation (hyperedge), Figure 4 The solid line width is positively correlated with the association strength, and example relationships include LiPF6→electrical conductivity (assuming a line width of 2.5pt and a confidence level of 92%) and gradient control→thermal stability (assuming a line width of 2.2pt and a confidence level of 87%). The dashed line represents a cross-domain migration relationship (such as pharmaceuticals→battery materials) with a confidence level of 74%.

[0095] It is worth noting that dynamic statistics can be performed on the visual decision tree to obtain the following graph dimension information: total number of nodes: 1247, association dimension: 200+, real-time update: 24 hours / once, field coverage: chemical industry / pharmaceutical industry / battery materials, average confidence level: 82.4%. The following key optimization paths can be obtained: safety improvement path: FEC additive→40% reduction in toxicity; thermal stability path: gradient temperature control→40% improvement in thermal stability; cross-domain migration path: drug release process→electrolyte pore optimization. The top three confidence rankings are as follows: LiPF6 concentration electrical conductivity (92%), FEC additive safety (89%), temperature gradient dendrite inhibition (87%).

[0096] In this embodiment, by mapping the knowledge graph into hypergraph nodes and hyperedges, a hypergraph extrapolation model can be constructed. Then, given a performance target, the relevant processes and raw materials in the hypergraph nodes are searched in reverse, and confidence values are assigned to the hyperedges. According to the confidence values, the weights of the hyperedges are determined, and in the reasoning process, high-weight paths are preferentially selected, so that the extrapolation scores of each hyperedge can be more accurately calculated.

[0097] Based on the first embodiment of the present application, in the third embodiment of the present application, the same or similar contents as the above-mentioned first embodiment can be referred to the above description, and the subsequent will not be described in detail. On this basis, please refer to Figure 5 , step S30, the recipe process extrapolation method further comprises steps S301-S303:

[0098] Step S301, generating a recipe process extrapolation parameter according to the extrapolation score.

[0099] Step S302, the recipe process extrapolation parameter is verified by the physical constraint condition, and the physical constraint condition includes the mass conservation law condition, the energy spontaneity condition and the toxicity compliance condition.

[0100] It should be noted that the generative adversarial network (GAN) structure can be introduced for adversarial verification, which includes a generator (Generator) and a discriminator (Discriminator). The generator generates a recipe process extrapolation parameter (new recipe and process parameters, etc.) according to the extrapolation score. The discriminator takes the recipe process extrapolation parameter output by the generator as input, and verifies the feasibility combined with the physical equation. The physical constraint conditions in the recipe adversarial verification include three kinds: mass conservation law condition, energy spontaneity condition and toxicity compliance condition.

[0101] Step S303, when the recipe process extrapolation parameter satisfies the mass conservation law condition, the energy spontaneity condition and the toxicity compliance condition at the same time, it means that the adversarial verification of the recipe process extrapolation parameter is passed.

[0102] It can be understood that the above three physical constraint conditions are verified in coordination, and when the three conditions are satisfied at the same time, it means that the adversarial verification of the recipe process extrapolation parameter is passed. For example, the recipe passing condition can be set as: Δm≤1e-6, ΔG<0 and toxicity score≥85 / 100. Through such constraint mode, the following technical advantages can be achieved: physical feasibility guarantee: avoid generating recipes that violate basic scientific laws (such as perpetual motion machine process); legal risk avoidance: ensure that the new recipe complies with international chemical management regulations; calculation efficiency optimization: reduce the calculation amount of invalid schemes by 92% through pre-screening mechanism.

[0103] In the embodiment, the formulation process extrapolation parameters are generated according to the extrapolation score; the formulation process extrapolation parameters are verified against physical constraint conditions, including mass conservation law condition, energy spontaneity condition and toxicity compliance condition; when the formulation process extrapolation parameters satisfy the mass conservation law condition, the energy spontaneity condition and the toxicity compliance condition at the same time, it means that the verification of the formulation process extrapolation parameters against the physical constraint conditions is passed. In the embodiment, the physical constraint conditions such as the mass conservation law condition, the energy spontaneity condition and the toxicity compliance condition are set, so that the feasibility of the generated formulation process extrapolation parameters can be verified, and the effectiveness of the generated new data is ensured.

[0104] In a feasible implementation, step S302 can include steps S11-S14.

[0105] Step S11, when the physical constraint condition is the mass conservation law condition, the total mass of the product and byproduct is determined according to the formulation process extrapolation parameters.

[0106] Step S12, the total mass of the reactants corresponding to the formulation process extrapolation parameters is determined.

[0107] It should be noted that when the physical constraint condition is the mass conservation law condition, the mass conservation of the formulation process extrapolation parameters is verified. Since in a closed system, the total mass of the substances participating in the chemical reaction is conserved:

[0108] ∑m 反应物 = ∑m 生成物 + ∑m 副产物 .

[0109] It can be understood that since in actual reactions, slight errors are allowed, the total mass of the product and byproduct generated in the chemical reaction by the formulation process extrapolation parameters is calculated first. Then the total mass of the reactants before the chemical reaction by the formulation process extrapolation parameters is calculated.

[0110] Step S13, it is judged whether the difference between the total mass and the total mass of the reactants is less than a preset mass threshold.

[0111] Step S14, if it is less than the preset mass threshold, it means that the formulation process extrapolation parameters satisfy the mass conservation law condition.

[0112] It can be understood that it can be judged whether the difference between the total mass and the total mass of the reactants is less than a preset mass threshold (1e-6), and if it does not exceed this mass threshold, it means that the formulation process extrapolation parameters satisfy the mass conservation law condition.

[0113] For example, the material balance calculation engine verifies the feasibility by the following formula:

[0114]

[0115] wherein, represents the input component molar concentration, represents the output component molar concentration. For example, if the lithium battery electrolyte formula is verified, when the LiPF6 molar number deviation is > 0.1%, the constraint alarm of the above formula 2 is triggered, and the system is automatically corrected to Δm≤1e-6.

[0116] In the embodiment, by judging the mass difference between the reactants and all the products corresponding to the formula process extrapolation parameters, it is judged whether the law of conservation of mass is satisfied, so as to guarantee the physical rationality of extrapolation and avoid generating formulas that violate basic scientific laws.

[0117] In another possible implementation, step S302 can further include steps S21-S23:

[0118] Step S21, when the physical constraint condition is the energy spontaneity condition, determining the enthalpy change and entropy change generated by the formula process extrapolation parameters in the preset reaction process.

[0119] It is worth noting that entropy is a measure of the degree of disorder of a system. Microscopically, entropy is related to the degree of disorder of molecular thermal motion; macroscopically, entropy reflects the degree of energy dispersion. The irreversibility of entropy change in a closed system, that is, entropy tends to increase or remain unchanged, for example: ice melting (order → disorder, entropy increase), gas diffusion (concentration → dispersion, entropy increase). If the system exchanges matter or energy with the outside world, the local entropy may decrease (such as the growth of a living body), but the total entropy (system + environment) will still increase. Among them, the heat exchanged between the system and the environment during the preset reaction process of the formula process extrapolation parameters is the enthalpy change ΔH, and the entropy change ΔS generated by the formula process extrapolation parameters in the preset reaction process satisfies the following formula:

[0120] ΔS 总 = ΔS 系统 + ΔS 环境 ≥ 0.

[0121] Step S22, determining the free energy change value according to the enthalpy change and the entropy change.

[0122] Step S23, when the free energy change value is negative, indicating that the formula process extrapolation parameters satisfy the energy spontaneity condition.

[0123] It can be understood that ΔG is the Gibbs free energy change of the reaction or process, which is used to judge the spontaneity of the process. Specifically, the energy flow analysis module executes the following formula to calculate and judge the rationality of the free energy change value:

[0124] ΔG = ΔH - TΔS < 0,

[0125] wherein, AH represents enthalpy change, AS represents entropy change, T represents absolute temperature of reaction (unit: K), which influences the contribution of entropy change to free energy. The quantum chemistry calculation is realized by VASP software to predict by DFT.

[0126] For example, in the case of screening of drug sustained-release formula, the synthesis scheme of nanocarrier is rejected automatically, and AG>0 indicates that the process is not spontaneous and requires external energy input. After optimization, AG=-32 kJ / mol→ verification is passed. AG<0 indicates that the process is spontaneous.

[0127] In this embodiment, by determining the enthalpy change and entropy change generated by the formula process extrapolation parameter in the preset reaction process, and then determining the free energy change value, the energy spontaneity of the reaction process can be judged, so as to ensure that the new formula meets the spontaneity.

[0128] In another possible implementation, step S302 can further include steps S31-S32:

[0129] Step S31, when the physical constraint condition is a toxicity compliance condition, performing toxicity scoring on the formula process extrapolation parameter according to a preset toxicity regulation library to obtain a toxicity scoring score.

[0130] It should be noted that when the physical constraint condition is a toxicity compliance condition, according to the toxicity regulations of various countries and regions, a toxicity evaluation model is constructed: toxicity score = 0.3 LD50 oral + 0.2 LD50 skin +.... The toxicity score of the formula process extrapolation parameter can be obtained by the above formula. Real-time regulation synchronization can be performed, and the system synchronization period is less than or equal to 5 minutes.

[0131] Step S32, when the toxicity scoring score is greater than a preset toxicity threshold, indicating that the formula process extrapolation parameter meets the toxicity compliance condition.

[0132] It can be understood that when the toxicity scoring score is greater than the preset toxicity threshold, it indicates that the formula process extrapolation parameter meets the toxicity compliance condition. For example, if textile dye formula optimization is performed, if it is detected that AZO dye C.I. Disperse Yellow 23→ is automatically replaced by C.I. Solvent Yellow 21, the toxicity score is from 72→ 94 (OEKO- Class 1 level).

[0133] In this embodiment, by performing toxicity scoring on the formula process extrapolation parameter according to the preset toxicity regulation library, legal risks can be avoided, and it can be ensured that the new formula meets international chemical management regulations.

[0134] It should be noted that the above is only three possible embodiments of step S302 provided by the present embodiment, and the present embodiment does not specifically limit the specific embodiments of step S302.

[0135] It should be noted that the present application can achieve the following beneficial effects compared to the existing scheme, as shown in the following table (1) and table (2):

[0136] Table (1)

[0137]

[0138] Table (2)

[0139] Test items Existing methods The present system New recipe generation speed 72 hours 1.5 hours Material cost reduction rate 5-8％ 22-35％ Process parameter optimization dimension 6 dimensions 58 dimensions Cross-domain knowledge reference rate 1.2％ 79.3％

[0140] It should be noted that the above examples are only for understanding the present application and do not constitute a limitation on the extrapolation method of the formula process, and more forms of simple transformation based on this technical concept are within the protection scope of the present application.

[0141] The present application also provides a formula process extrapolation device, please refer to Figure 6 , the formula process extrapolation device comprises:

[0142] The atlas construction module 10 is used for acquiring multi-source formula process data, and constructing a knowledge graph according to the multi-source formula process data;

[0143] The hypergraph extrapolation module 20 is used for constructing a hypergraph extrapolation model according to the knowledge graph, performing extrapolation according to the hypergraph extrapolation model, and obtaining an extrapolation score of each hyperedge, wherein the hypergraph extrapolation model comprises hypergraph nodes and hyperedges;

[0144] The parameter generation module 30 is used for generating formula process extrapolation parameters according to the extrapolation score, and performing adversarial verification on the formula process extrapolation parameters;

[0145] The adversarial verification module 40 is used for obtaining new formula data and new process data when the adversarial verification is passed.

[0146] The formula process extrapolation device provided by the present application adopts the formula process extrapolation method in the above embodiments, which can solve the technical problems. Compared with the prior art, the beneficial effects of the formula process extrapolation device provided by the present application are the same as those of the formula process extrapolation method provided by the above embodiments, and other technical features in the formula process extrapolation device are the same as those disclosed in the above embodiments. The features are not repeated here.

[0147] The application provides a recipe process extrapolation device, which comprises at least one processor and a memory connected with the at least one processor; wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to perform the recipe process extrapolation method in the above embodiment one.

[0148] Reference will now be made to the drawings Figure 7 , which show structural diagrams of a recipe process extrapolation device suitable for implementing embodiments of the application. The recipe process extrapolation device in the embodiments of the application can include, but is not limited to, mobile terminals such as mobile phones, notebook computers, digital broadcast receivers, PDAs (Personal Digital Assistants), PADs (Portable Application Descriptions), PMPs (Portable Media Players), vehicle-mounted terminals (such as vehicle-mounted navigation terminals), and the like, and fixed terminals such as digital TVs, desktop computers, and the like. Figure 7 The illustrated recipe process extrapolation device is merely an example and should not impose any limitation on the functions and use range of the embodiments of the application.

[0149] As shown in Figure 7 , the recipe process extrapolation device can include a processing apparatus 1001 (such as a central processor, a graphics processor, or the like) that can perform various appropriate actions and processes according to programs stored in a read-only memory 1002 or loaded from a storage apparatus 1003 into a random access memory 1004. Various programs and data required for the operation of the recipe process extrapolation device are also stored in the random access memory 1004. The processing apparatus 1001, the read-only memory 1002, and the random access memory 1004 are connected to each other through a bus 1005. An input / output interface 1006 is also connected to the bus. Generally, the following systems can be connected to the input / output interface 1006: input apparatuses 1007 including, for example, a touch screen, a touch pad, a keyboard, a mouse, an image sensor, a microphone, an accelerometer, a gyroscope, and the like; output apparatuses 1008 including, for example, a liquid crystal display (LCD), a speaker, a vibrator, and the like; the storage apparatus 1003 including, for example, a magnetic tape, a hard disk, and the like; and a communication apparatus 1009. The communication apparatus 1009 can allow the recipe process extrapolation device to communicate with other devices wirelessly or by wire to exchange data. Although the recipe process extrapolation device having various systems is shown in the figure, it should be understood that all the illustrated systems are not required to be implemented or provided. More or fewer systems can be alternatively implemented or provided.

[0150] In particular, according to embodiments of the present disclosure, the processes described above with reference to the flowcharts can be implemented as a computer software program. For example, embodiments of the present disclosure include a computer program product comprising a computer program carried on a computer readable medium, the computer program comprising program code for performing the methods illustrated by the flowcharts. In such embodiments, the computer program can be downloaded and installed from a network via a communication device, or installed from a storage device 1003, or installed from a read-only memory 1002. When the computer program is executed by the processing device 1001, the above-mentioned functions defined in the methods of the embodiments of the present disclosure are performed.

[0151] The formula process extrapolation device provided by the present application adopts the formula process extrapolation method in the above-mentioned embodiments, and can solve the technical problem of formula process extrapolation. Compared with the prior art, the beneficial effects of the formula process extrapolation device provided by the present application are the same as those of the formula process extrapolation method provided by the above-mentioned embodiments, and other technical features in the formula process extrapolation device are the same as those disclosed in the previous embodiment method, which will not be repeated here.

[0152] It should be understood that parts of the present disclosure can be realized by hardware, software, firmware or a combination thereof. In the description of the above-mentioned embodiments, specific features, structures, materials or characteristics can be combined in any one or more embodiments or examples in a suitable manner.

[0153] The above is merely specific embodiments of the present application, but the protection scope of the present application is not limited thereto, and any person skilled in the art can easily think of changes or replacements within the technical scope disclosed by the present application, which should be covered within the protection scope of the present application. Therefore, the protection scope of the present application should be subject to the protection scope of the claims.

[0154] The present application provides a computer readable storage medium having stored thereon computer readable program instructions (i.e. computer program) for performing the formula process extrapolation method in the above-mentioned embodiments.

[0155] The computer readable storage medium provided in the application may, for example, be a U disk, but is not limited to an electric, magnetic, optical, electromagnetic, infrared, or semiconductor system, system, or device, or any combination of the above. More specific examples of the computer readable storage medium can include, but are not limited to, an electric connection with one or more conductive lines, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the above. In the embodiment, the computer readable storage medium can be any tangible medium containing or storing a program that can be used by or in combination with an instruction execution system, system, or device. The program code contained on the computer readable storage medium can be transmitted by any suitable medium, including but not limited to an electric wire, an optical cable, an RF (Radio Frequency), and the like, or any suitable combination of the above.

[0156] The computer readable storage medium described above can be included in the formula process extrapolation device, or can exist separately without being assembled into the formula process extrapolation device.

[0157] The computer readable storage medium described above carries one or more programs, which, when executed by the formula process extrapolation device, cause the formula process extrapolation device to: acquire multi-source formula process data, and construct a knowledge graph according to the multi-source formula process data; construct a hypergraph extrapolation model according to the knowledge graph, perform extrapolation according to the hypergraph extrapolation model, obtain an extrapolation score of each hyperedge, and the hypergraph extrapolation model includes hypergraph nodes and hyperedges; generate formula process extrapolation parameters according to the extrapolation score, and perform adversarial verification on the formula process extrapolation parameters; when the adversarial verification is passed, obtain new formula data and new process data.

[0158] Computer program code for carrying out operations of the present application can be written in any combination of one or more programming languages, including an object oriented programming language such as Java, Smalltalk, C++ or the like and conventional procedural programming languages, such as the "C" programming language or similar programming languages. The program code can execute entirely on the user's computer, partly on the user's computer, as a stand-alone software package, partly on the user's computer and partly on a remote computer or entirely on the remote computer or server. In the latter scenario, the remote computer can be connected to the user's computer through any type of network, including a local area network (LAN) or a wide area network (WAN), or the connection can be made to an external computer (for example, through the Internet using an Internet Service Provider).

[0159] The computer program instructions can also be loaded onto a computer, other programmable data processing apparatus, or other devices to cause a series of operational steps to be performed on the computer, other programmable apparatus or other devices to produce a computer implemented process such that the instructions which execute on the computer or other programmable apparatus provide processes for implementing the functions / acts specified in the flowchart and / or block diagram block or blocks.

[0160] The modules involved in the embodiments of the present application can be implemented in software or hardware. In some cases, the name of the module does not constitute a limitation on the module itself.

[0161] The readable storage medium provided by the present application is a computer readable storage medium, which stores computer readable program instructions (i.e. computer programs) for executing the above-mentioned formula process extrapolation method, and can solve the technical problems. Compared with the prior art, the computer readable storage medium provided by the present application has the same beneficial effects as the formula process extrapolation method provided by the above-mentioned embodiments, which will not be described here.

[0162] The application also provides a computer program product comprising a computer program which, when executed by a processor, implements the steps of the recipe process extrapolation method as described above.

[0163] The computer program product provided by the application can solve the technical problem. Compared with the prior art, the computer program product provided by the application has the same beneficial effects as the recipe process extrapolation method provided by the above-mentioned embodiments, which will not be repeated here.

[0164] The above only describes some embodiments of the application, and does not limit the patent scope of the application. Any equivalent structural transformation, direct / indirect application in other related technical fields, or direct / indirect application in other related technical fields based on the technical concept of the application and the content of the specification and drawings are included in the patent protection scope of the application.

Claims

1. A method for extrapolating a formulation process, characterized in that, The method includes: Acquire multi-source formulation and process data, and construct a knowledge graph based on the multi-source formulation and process data; A hypergraph extrapolation model is constructed based on the knowledge graph. Extrapolation is performed based on the hypergraph extrapolation model to obtain the extrapolation score of each hyperedge. The hypergraph extrapolation model includes hypergraph nodes and hyperedges. Formula and process extrapolation parameters are generated based on the extrapolation score, and adversarial verification is performed on the formula and process extrapolation parameters. When the countermeasures are successfully validated, new formulation data and new process data are obtained.

2. The method as described in claim 1, characterized in that, The steps of constructing a hypergraph extrapolation model based on the knowledge graph, performing extrapolation based on the hypergraph extrapolation model, and obtaining the extrapolation score for each hyperedge include: The knowledge graph is mapped to hypergraph nodes, and the hypergraph nodes include at least one of raw material entity nodes, process parameter nodes, and performance index nodes. Connect the nodes of the hypergraph to obtain a hyperedge; Based on the preset target and the hyperedge, perform multi-hop extrapolation, reverse search the hypergraph nodes, and obtain the confidence value of the hyperedge; The edge weight of the hyperedge is determined based on the confidence value; The extrapolation score of each hyperedge is determined based on the hypergraph node and the edge weight.

3. The method as described in claim 1, characterized in that, The step of generating formulation and process extrapolation parameters based on the extrapolation score and performing adversarial verification on the formulation and process extrapolation parameters includes: Generate extrapolation parameters for the formulation and process based on the extrapolation score; The extrapolated parameters of the formulation process are subjected to countermeasure verification based on physical constraints, including the law of conservation of mass, the energy spontaneity condition, and the toxicity compliance condition. When the extrapolated parameters of the formulation process simultaneously satisfy the mass conservation law condition, the energy spontaneity condition, and the toxicity compliance condition, it indicates that the countermeasure verification of the extrapolated parameters of the formulation process is passed.

4. The method as described in claim 3, characterized in that, The step of performing adversarial verification on the extrapolated parameters of the formulation process based on physical constraints includes: When the physical constraint condition is the law of conservation of mass, the total mass of the product and the by-product is determined according to the extrapolated parameters of the formulation process. Determine the total mass of reactants corresponding to the extrapolated parameters of the formulation process; Determine whether the difference between the total mass and the total mass of the reactants is less than a preset mass threshold; If the mass is less than the preset mass threshold, it means that the extrapolated parameters of the formula process satisfy the mass conservation law condition.

5. The method as described in claim 3, characterized in that, The step of performing adversarial verification on the extrapolated parameters of the formulation process based on physical constraints further includes: When the physical constraint is an energy spontaneity condition, determine the enthalpy change and entropy change of the extrapolated parameters of the formulation process during the preset reaction process; The change in free energy is determined based on the enthalpy change and the entropy change; When the change in free energy is negative, it indicates that the extrapolated parameters of the formulation process satisfy the energy spontaneity condition.

6. The method as described in claim 3, characterized in that, The step of performing adversarial verification on the extrapolated parameters of the formulation process based on physical constraints further includes: When the physical constraint is a toxicity compliance condition, the toxicity score is obtained by extrapolating the formulation process parameters according to the preset toxicity regulation library. When the toxicity score is greater than the preset toxicity threshold, it indicates that the extrapolated parameters of the formulation process meet the toxicity compliance conditions.

7. The method as described in claim 1, characterized in that, The steps of acquiring multi-source formulation process data and constructing a knowledge graph based on the multi-source formulation process data include: Multimodal fusion is performed on the acquired multi-source data to obtain multi-source formulation and process data; Entity extraction is performed on the multi-source formulation process data to obtain an initial spectrum; Nonlinear relational reasoning is performed on the initial graph to obtain a knowledge graph.

8. A formula process extrapolation device, characterized in that, The formulation process extrapolation device includes: The knowledge graph construction module is used to acquire multi-source formulation and process data, and to construct a knowledge graph based on the multi-source formulation and process data. The hypergraph extrapolation module is used to construct a hypergraph extrapolation model based on the knowledge graph, perform extrapolation based on the hypergraph extrapolation model, and obtain the extrapolation score of each hyperedge. The hypergraph extrapolation model includes hypergraph nodes and hyperedges. The parameter generation module is used to generate formula and process extrapolation parameters based on the extrapolation score, and to perform adversarial verification on the formula and process extrapolation parameters; The adversarial verification module is used to obtain new formula data and new process data when the adversarial verification is successful.

9. A formula process extrapolation device, characterized in that, The device includes: a memory, a processor, and a computer program stored in the memory and executable on the processor, the computer program being configured to implement the steps of the formulation process extrapolation method as described in any one of claims 1 to 7.

10. A storage medium, characterized in that, The storage medium is a computer-readable storage medium, and a computer program is stored on the storage medium. When the computer program is executed by a processor, it implements the steps of the formulation process extrapolation method as described in any one of claims 1 to 7.

Images