A lubricating oil blending method, system and apparatus
By using virtual modulation verification and machine learning models, the problems of long formulation development cycle and low efficiency in the lubricating oil blending process have been solved, realizing automated and reproducible lubricating oil formulation optimization, and improving the scientificity and efficiency of formulation screening.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- NEW SIXTH RING (SHANDONG) ENERGY TECH CO LTD
- Filing Date
- 2026-05-06
- Publication Date
- 2026-07-21
AI Technical Summary
In the traditional lubricant blending process, the formulation development cycle is long. The key problems that existing technologies have failed to effectively solve are the long formulation development cycle, high cost, low efficiency, difficulty in achieving the globally optimal combination, and lack of systematic quantitative evaluation methods.
Virtual modulation verification technology is used to establish a mapping relationship between additives and lubricant performance through a pre-trained machine learning model. Combined with dynamic time warping algorithm, performance deviation values are calculated to realize an automated and reproducible optimization process for the formulation.
It enables the prediction and adjustment of formulation performance in a computer environment without the need for physical sample preparation, significantly reducing reliance on personal experience, improving the scientific rigor and efficiency of formulation screening, and ensuring the reliability of optimal combinations.
Smart Images
Figure QLYQS_1
Abstract
Description
Technical Field
[0001] This invention belongs to the field of lubricating oil blending technology, specifically relating to a lubricating oil blending method, system, and apparatus. Background Technology
[0002] Lubricating oil blending is a process in which base oil is mixed with various functional additives in a certain proportion, and then processed through stirring, heating and other processes to make the additives evenly dispersed in the base oil, thereby obtaining a lubricating oil product with specific performance indicators.
[0003] Traditional methods require the actual preparation of samples and the conduct of multiple performance tests for each round of formulation adjustments. A complete formulation development cycle typically takes several weeks to months, consuming a large amount of base oil, additives, and experimental resources. Formulation screening and adjustment often rely on the personal experience of senior engineers, and the adjustment approaches of different personnel vary greatly, making standardization and reproducibility difficult. For complex multi-additive systems, the interactions between additives are difficult to accurately judge based on experience.
[0004] After extracting candidate solutions from the formula database, there is a lack of systematic quantitative evaluation methods. This makes it impossible to objectively compare and rank the performance of each solution without conducting physical experiments. Formula adjustments lack quantitative calculation tools, and the adjustment range relies on empirical judgment, resulting in low efficiency and difficulty in guaranteeing convergence. Furthermore, when multiple performance indicators deviate simultaneously, there is a lack of effective means to handle the coupling conflict of multiple objectives.
[0005] It is difficult to accurately determine which additive is the key factor affecting a certain performance index. Trial and error methods are often used, which are costly, inefficient, and prone to getting stuck in local optima, making it difficult to find the globally optimal formula combination. Summary of the Invention
[0006] The purpose of this invention is to provide a method, system, and apparatus for blending lubricating oil.
[0007] A method for blending lubricating oil includes the following steps: S1. Obtain the performance constraints of the target lubricating oil and extract at least three candidate formulations from the lubricating oil formulation database; S2. Perform virtual modulation verification on each candidate formulation scheme to obtain the expected performance index of each candidate formulation scheme; compare the expected performance index with the performance constraints of the target lubricating oil to obtain the performance deviation value of each candidate formulation scheme. S3. Select the candidate formulation with the smallest performance deviation value as the benchmark formulation, and adjust the component ratio of the benchmark formulation according to the performance deviation value to obtain the adjusted formulation. S4. The adjusted formula is virtually modulated and verified again. After confirming that its expected performance indicators meet the performance constraints of the target lubricating oil, the final formula is obtained and output as the original experimental formula.
[0008] In S2, virtual modulation verification is performed on each candidate formulation scheme, specifically as follows: A pre-trained machine learning model is used to establish the mapping relationship between each additive in the candidate formulation and various properties of the lubricating oil under standard stirring process conditions, as well as the mapping relationship between each additive in the candidate formulation and the expected performance index. Based on the candidate formulation, the contribution coefficient of each additive to various properties of the lubricating oil under standard stirring process conditions and the expected performance index of the candidate formulation are obtained.
[0009] In S2, the expected performance indicators are compared with the performance constraints of the target lubricating oil to obtain the performance deviation value of each formulation candidate scheme, specifically: The performance constraints of the target lubricating oil are represented as a constraint sequence, and the expected performance indicators of the candidate formulations are represented as an indicator sequence. The order of elements in the constraint sequence and indicator sequence is fixed according to the type of performance indicator; The minimum warping path distance between the standardized constraint sequence and the index sequence is calculated using the dynamic time warping algorithm, and this minimum warping path distance is used as the performance deviation value.
[0010] In S3, the component ratios of the baseline formulation are adjusted based on the performance deviation value to obtain the adjusted formulation, specifically: Obtain the performance deviation value and the minimum regularization path distance, determine the candidate formulation with the smallest performance deviation value as the baseline formulation, and read the mass fraction of each additive in the baseline formulation. Based on the warping path output by the dynamic time warping algorithm, the performance index dimension with the largest deviation between the index sequence and the constraint sequence is identified, and this performance index dimension is determined as the main adjustment target. If the main adjustment target is less than the target performance constraint, then the additive with a positive contribution coefficient and the largest absolute value is selected as the target additive. If the main adjustment target is greater than the target performance constraint, then the additive with the negative contribution coefficient and the largest absolute value is selected as the target additive. Based on the contribution coefficient of the target additive, the adjustment ratio of the target additive is calculated to obtain the adjusted formulation.
[0011] Based on the contribution coefficient of the target additive, the adjustment ratio of the target additive is calculated as follows: , in, The adjustment ratio for the target additive. Adjust the target deviation value as the main adjustment. The contribution coefficient, To adjust the step size coefficient, This is a comprehensive correction factor.
[0012] If multiple candidate formulations in S3 have the same performance deviation value, a stirring simulation is performed to simulate the expected performance index of the candidate formulations with the same performance deviation value under different stirring parameters, and the maximum value is taken to recalculate the performance deviation value.
[0013] In S4, the adjusted formula is again virtually modulated and verified, specifically as follows: Using the same virtual modulation verification method as S2, the expected performance indicators of the adjusted formulation are calculated to determine whether they meet the performance constraints of the target lubricating oil. If the conditions are met, the adjusted recipe will be output as the final recipe. If the requirements are not met, repeat steps S3 to S4 for iterative optimization until the performance constraints are met or the maximum number of iterations is reached.
[0014] S1 extracts at least three candidate formulations from the lubricating oil formulation database, specifically: The performance constraints of the target lubricating oil are transformed into a multi-dimensional performance vector. The standard performance vectors of each formula in the lubricating oil formula database are traversed, the weighted Euclidean distance is calculated, and the formulas are sorted from smallest to largest distance. The top three formulas are selected as candidate formula schemes.
[0015] A lubricating oil blending system for implementing the above-mentioned lubricating oil blending method includes: The candidate solution generation module obtains the performance constraints of the target lubricating oil and extracts at least three formulation candidate solutions from the lubricating oil formulation database. The performance verification module performs virtual modulation verification on each candidate formulation scheme to obtain the expected performance index of each candidate formulation scheme; the expected performance index is compared with the performance constraints of the target lubricating oil to obtain the performance deviation value of each candidate formulation scheme. The formulation adjustment module selects the candidate formulation with the smallest performance deviation value as the baseline formulation, and adjusts the component ratio of the baseline formulation according to the performance deviation value to obtain the adjusted formulation. The final formula output module performs virtual modulation verification on the adjusted formula again. After confirming that its expected performance indicators meet the performance constraints of the target lubricating oil, the final formula is obtained and output as the original experimental formula.
[0016] A lubricating oil blending apparatus includes a processor and a memory, wherein the processor implements a lubricating oil blending method when executing a computer program stored in the memory.
[0017] Compared with the prior art, the present invention has the following beneficial effects: This invention uses virtual modulation verification technology to predict and adjust the performance of a formulation in a computer environment, eliminating the need for physical sample preparation and experimental testing.
[0018] This invention represents the performance constraints of the target lubricating oil as a constraint sequence, providing a quantitative basis for formula screening and improving the scientific nature of decision-making. It represents the expected performance indicators of the candidate formula schemes as an indicator sequence, and uses a dynamic time warping algorithm to calculate the minimum warping path distance between the two as the performance deviation value, providing an objective and comparable quantitative basis for the screening of candidate formula schemes.
[0019] This invention establishes a mapping relationship between each additive and various properties of lubricating oil under standard stirring process conditions by pre-training a machine learning model, and obtains the contribution coefficient of each additive, making the formulation adjustment process programmable, reproducible, and automated, significantly reducing reliance on personal experience.
[0020] This invention identifies the primary adjustment target through a dynamic time warping algorithm and combines it with a comprehensive quantitative assessment of the side effects of the adjustment on non-target performance, thereby achieving the optimal trade-off when multiple targets conflict.
[0021] When multiple candidate formulations have the same performance deviation value, this invention simulates the expected performance index of each formulation under different stirring parameters through stirring simulation, takes the maximum value and recalculates the performance deviation value, providing further selection basis for the tie case and ensuring the reliability of the screening results. Detailed Implementation
[0022] To further understand the content of this invention, the invention will be described in detail with reference to the embodiments.
[0023] This invention relates to a method for blending lubricating oil, comprising the following steps: S1. Obtain the performance constraints of the target lubricating oil and extract at least three candidate formulations from the lubricating oil formulation database; Specifically, the performance constraints of the target lubricating oil are obtained, and the performance constraints of the target lubricating oil are converted into a multi-dimensional performance vector. The standard performance vectors of each formula in the lubricating oil formula database are traversed, the weighted Euclidean distance is calculated, and the formulas are sorted from smallest to largest distance. The top three formulas are selected as candidate formula schemes.
[0024] Preferably, the weighted Euclidean distance in this step is less than the maximum distance threshold, which makes it possible to further compare the selected formulation candidates. If the weighted Euclidean distance of all formulation candidates is greater than or equal to the maximum distance threshold, an error is reported, which may be due to a mismatch in the lubricating oil formulation database.
[0025] S2. Perform virtual modulation verification on each candidate formulation scheme to obtain the expected performance index of each candidate formulation scheme; compare the expected performance index with the performance constraints of the target lubricating oil to obtain the performance deviation value of each candidate formulation scheme.
[0026] Furthermore, virtual modulation verification was performed on each candidate formulation, specifically as follows: A pre-trained machine learning model is used to establish the mapping relationship between each additive in the candidate formulation and various properties of the lubricating oil under standard stirring process conditions, as well as the mapping relationship between each additive in the candidate formulation and the expected performance index. Based on the candidate formulation, the contribution coefficient of each additive to various properties of the lubricating oil under standard stirring process conditions and the expected performance index of the candidate formulation are obtained.
[0027] Specifically, the process of training a machine learning model can employ the following methods; any other methods within the field capable of accomplishing this task are acceptable, and this application does not impose excessive restrictions: During pre-training, base oil viscosity, base oil density, mass fraction of each additive, and molecular weight of the additives are extracted from the candidate formulations. Standard modulation process parameters are extracted from a preset process parameter library. These standard modulation process parameters include modulation temperature, stirring speed, impeller diameter of the modulation vessel, and modulation time. Calculate the theoretical viscosity of the mixture based on the base oil viscosity, additive concentration, and mass fraction of each additive; , in, This is the theoretical viscosity of the mixture. The mass fraction of each additive. The viscosity contribution coefficient of each additive.
[0028] Calculate the stirring Reynolds number based on the base oil density, stirring speed, impeller diameter of the mixing vessel, and the theoretical viscosity of the mixture. ; , in, This is the theoretical viscosity of the mixture. Based on base oil density, This refers to the stirring speed. The diameter of the impeller in the mixing vessel.
[0029] The Reynolds number is a dimensionless number in fluid mechanics that reflects the ratio of inertial forces to viscous forces. The smaller the Reynolds number, the more viscous forces dominate, resulting in smoother flow; the larger the Reynolds number, the more inertial forces dominate, resulting in turbulent flow.
[0030] Based on the stirring Reynolds number, the flow state characteristic values of the mixture are calculated. : ; Record the stirring Reynolds number, theoretical viscosity of the mixture, flow state characteristic values, modulation temperature, stirring speed, and modulation time as an intermediate parameter set for virtual modulation verification.
[0031] These intermediate parameters form the basis for all subsequent calculations. The Reynolds number and flow state characteristics determine the shear strength and dispersion effect of the additive during modulation; the theoretical viscosity of the mixture affects pumping power consumption and mixing time.
[0032] It is important to note that this step involves virtual modulation and verification of each candidate formulation, calculating the impact of additives on the expected performance indicators.
[0033] The expected performance metrics for each candidate formulation are calculated as follows: Calculate the blade linear velocity based on the stirring speed and the diameter of the impeller in the mixing vessel. ; , The blade linear velocity determines the shear strength experienced by the fluid.
[0034] Based on the blade linear velocity v and the gap δ between the blade and the vessel wall, the average shear rate during the modulation process is calculated. The shear rate reflects the deformation rate inside the fluid and directly affects the dispersion effect of the additive. ,in This is the blade shape coefficient, which is determined empirically. The cumulative shear work W is calculated based on the average shear rate, modulation time t, base oil viscosity, and mass fraction of each additive. shear ; , Cumulative shear work reflects the total mechanical energy input into the oil during the entire modulation process. The greater the shear work, the more fully the additive is dispersed, but it may also lead to additive degradation.
[0035] Calculate the thermal energy E based on the modulation temperature T and the base oil viscosity. thermal , , Where R is the gas constant.
[0036] Thermal energy reflects the contribution of molecular thermal motion to the diffusion of additives. The higher the temperature and the lower the viscosity, the greater the thermal energy.
[0037] The cumulative shear work is coupled with the thermal energy to obtain the energy coupling coefficient ξ. , The energy coupling coefficient reflects the relative contributions of shear work and thermal energy to the additive dispersion process. When ξ approaches 1, shear is dominant; when ξ approaches 0, thermal motion is dominant.
[0038] The effective dispersion energy E is calculated based on the energy coupling coefficient ξ, the stirring Reynolds number Re, the theoretical viscosity of the mixture, and the average shear rate. disp ; , Effective dispersion energy is an energy index that integrates flow regime, shear, and thermal effects, and is used to calculate the degree of activation of additives.
[0039] Polar parameters of each additive were extracted from the candidate formulations. P add, i Based on the polarity parameters of each additive and the polarity parameters of the base oil P The difference in oil values is used to calculate the polarity matching degree of each additive. ei ; The polarity parameters of additives and base oils can be expressed using Hildebrand solubility parameters or Hansen solubility parameters. The smaller the difference between the two, the easier it is for the additive to dissolve and disperse.
[0040] Based on the polarity matching degree, effective dispersion energy, modulation temperature T, and mass fraction of each additive, the effective activation mass fraction of each additive is calculated: ; , The effective activation mass fraction is one of the innovative aspects of this invention. It reflects the portion of the additive that actually plays a role in the formulation process.
[0041] Based on the effective activation mass fraction of each additive and the theoretical viscosity of the mixture, the first entanglement factor is calculated: ; The second entanglement factor is calculated based on the stirring Reynolds number and the energy coupling coefficient: ; The third entanglement factor is obtained by cross-multiplying the first entanglement factor with the second entanglement factor: ; The first entanglement factor reflects the coupling relationship between additive concentration and mixture viscosity. Higher concentrations and greater viscosity result in stronger entanglement. The second entanglement factor reflects the coupling relationship between Reynolds number and energy coupling; stronger turbulence leads to a decrease in this factor. The third entanglement factor reflects the overall degree of entanglement across multiple factors.
[0042] It should be noted that the denominator in all calculations in this application is not zero. If the denominator is zero, a small perturbation is added to change the variable to obtain an optimized result.
[0043] For each additive, its contribution coefficient to each performance index is calculated based on the first entanglement factor, the second entanglement factor, the third entanglement factor, the flow state characteristic value, the energy coupling coefficient, the effective activation rate, and the baseline contribution coefficient of each additive: , , , , Using the baseline performance value of the base oil as the initial value, the contributions of each additive are added one by one to obtain the expected performance indicators: , , , .
[0044] Based on this pre-trained machine learning model, the mapping relationship between each additive in the candidate formulation and various properties of the lubricating oil under standard stirring process conditions is established, as well as the mapping relationship between each additive in the candidate formulation and the expected performance index. Based on the candidate formulation, the contribution coefficient of each additive to various properties of the lubricating oil under standard stirring process conditions and the expected performance index of the candidate formulation are obtained.
[0045] In addition, a more general estimation method for expected performance indicators is provided: find the historical formula that is closest to the current formula in the formula database, and use its measured performance as the expected performance of the current formula.
[0046] The specific procedure is as follows: Calculate the component distance between the current formulation and each formulation in the database, select one or more formulations that are closest in distance, and take the average of the measured performance of these formulations as the expected performance of the current formulation.
[0047] The expected performance indicators are compared with the performance constraints of the target lubricant to obtain the performance deviation value for each candidate formulation. Specifically: The performance constraints of the target lubricating oil are represented as a constraint sequence, and the expected performance indicators of the candidate formulations are represented as an indicator sequence. The order of elements in the constraint sequence and indicator sequence is fixed according to the type of performance indicator; The minimum warping path distance between the standardized constraint sequence and the index sequence is calculated using the dynamic time warping algorithm, and this minimum warping path distance is used as the performance deviation value.
[0048] Furthermore, the minimum normalized path distance between the standardized constraint sequence and the index sequence is calculated using the dynamic time warping algorithm. A specific example is as follows: Assuming performance requirements: VI=140, PP=-30, FP=220, OIT=40, Formula A: VI=135, PP=-28, FP=215, OIT=38, all deviations are too low (except for PP, which is too high); Formula B: VI=145, PP=-32, FP=225, OIT=42, all deviations are too high; When directly subtracting two recipes, the absolute values of their deviations may be the same. However, the Dynamic Time Warping algorithm considers the deviation shapes of recipes A and B to be different, thus assigning different similarity scores. This helps in selecting a more suitable candidate recipe.
[0049] Dynamic time warping algorithms allow for stretching and alignment between sequences, capturing the shape similarity between two sequences rather than just point-by-point differences. In lubricant formulations, this means that dynamic time warping algorithms can identify whether the patterns of deviation are consistent, even if each metric has deviations.
[0050] The core of the dynamic time warping algorithm is to find an optimal matching path between two sequences, allowing for non-one-to-one correspondence between sequence elements, thereby measuring the shape similarity between the two sequences, rather than just the difference between corresponding points.
[0051] There may be coupling relationships between different performance indicators. Dynamic time warping algorithm can capture this non-linear sequence similarity and reflects the overall matching degree of the formula better than simple distance calculation.
[0052] S3. Select the candidate formulation with the smallest performance deviation value as the benchmark formulation, and adjust the component ratio of the benchmark formulation according to the performance deviation value to obtain the adjusted formulation.
[0053] Specifically, the performance deviation value and the minimum regularization path distance are obtained. The candidate formulation with the smallest performance deviation value is determined as the baseline formulation. If there are multiple candidate formulations with the same performance deviation value in S3, a stirring simulation is performed to simulate the expected performance index of the candidate formulations with the same performance deviation value under different stirring parameters. The maximum value is taken to recalculate the performance deviation value, and the candidate formulation with the smallest performance deviation value is determined as the baseline formulation.
[0054] Read the mass fraction of each additive in the baseline formulation; Based on the warping path output by the dynamic time warping algorithm, the performance index dimension with the largest deviation between the index sequence and the constraint sequence is identified, and this performance index dimension is determined as the main adjustment target. If the main adjustment target is less than the target performance constraint, then the additive with a positive contribution coefficient and the largest absolute value is selected as the target additive. If the main adjustment target is greater than the target performance constraint, then the additive with the negative contribution coefficient and the largest absolute value is selected as the target additive. The sign of the contribution coefficient determines whether the adjustment direction is to increase or decrease, while the absolute value determines the adjustment efficiency. Selecting the most efficient additive for adjustment can reduce the amount of adjustment and minimize the cascading effects on other performance indicators.
[0055] Based on the contribution coefficient of the target additive, the adjustment ratio of the target additive is calculated to obtain the adjusted formulation.
[0056] After adjusting the target additive, if the deviation of the main adjustment target is less than the single adjustment threshold, then the additive with the second largest contribution coefficient is adjusted as the target additive, and the same adjustment is performed. If, after adjusting the target additive, the main adjustment target is greater than or equal to the single adjustment threshold, then the main adjustment target is adjusted again.
[0057] After adjusting the deviation between the main adjustment target and the target performance constraint to within the threshold range, continue to adjust the performance indicator dimension with the largest deviation between other indicator sequences and the constraint sequence until all dimensions are within the threshold range.
[0058] Based on the contribution coefficient of the target additive, the adjustment ratio of the target additive is calculated as follows: , in, The adjustment ratio for the target additive. Adjust the target deviation value as the main adjustment. The contribution coefficient, To adjust the step size coefficient, This is a comprehensive correction factor.
[0059] Calculate the new mass fraction of the target additive, adjust the mass fraction of the base oil accordingly, update the mass fraction of each additive and normalize it so that the sum of the mass fractions of each component is 100%, and obtain the adjusted formula.
[0060] S4. The adjusted formula is virtually modulated and verified again. After confirming that its expected performance indicators meet the performance constraints of the target lubricating oil, the final formula is obtained and output as the original experimental formula.
[0061] Specifically, the same virtual modulation verification method as S2 is used to calculate the expected performance index of the adjusted formula and determine whether it meets the performance constraints of the target lubricating oil. If the conditions are met, the adjusted recipe will be output as the final recipe. If the requirements are not met, repeat steps S3 to S4 for iterative optimization until the performance constraints are met or the maximum number of iterations is reached.
[0062] The final formula is obtained and used as the original experimental formula to complete the research on lubricant blending formula. If this final formula is used for production, material proportioning instructions can be generated based on the final formula and sent to the production line control system to control the mixing tank to perform lubricant mixing operations. Specifically: Read the base oil type and each additive type from the final formula, and obtain the corresponding mass fraction of each component; Obtain the target lubricating oil volume from the production order, calculate the feeding volume or feeding mass of each component based on the target lubricating oil volume and the mass fraction of each component, and generate a material proportioning instruction that includes the feeding sequence and feeding amount. The material proportioning instructions are sent to the production line control system via an industrial communication protocol. After the production line control system parses the material proportioning instruction, it sequentially controls the corresponding material storage tank valves to open, and adds each component into the mixing vessel according to the feeding sequence and amount. It then starts the stirring and heating devices and executes the lubricating oil mixing operation according to the preset mixing process parameters to obtain the target lubricating oil.
[0063] A lubricating oil blending system for implementing the above-mentioned lubricating oil blending method includes: The candidate solution generation module obtains the performance constraints of the target lubricating oil and extracts at least three formulation candidate solutions from the lubricating oil formulation database. The performance verification module performs virtual modulation verification on each candidate formulation scheme to obtain the expected performance index of each candidate formulation scheme; the expected performance index is compared with the performance constraints of the target lubricating oil to obtain the performance deviation value of each candidate formulation scheme. The formulation adjustment module selects the candidate formulation with the smallest performance deviation value as the baseline formulation, and adjusts the component ratio of the baseline formulation according to the performance deviation value to obtain the adjusted formulation. The final formula output module performs virtual modulation verification on the adjusted formula again. After confirming that its expected performance indicators meet the performance constraints of the target lubricating oil, the final formula is obtained and output as the original experimental formula.
[0064] A lubricating oil blending apparatus includes a processor and a memory, wherein the processor implements a lubricating oil blending method when executing a computer program stored in the memory.
Claims
1. A method for blending lubricating oil, characterized in that, Includes the following steps: S1. Obtain the performance constraints of the target lubricating oil and extract at least three candidate formulations from the lubricating oil formulation database; S2. Perform virtual modulation verification on each candidate formulation scheme to obtain the expected performance index of each candidate formulation scheme; The expected performance indicators are compared with the performance constraints of the target lubricant to obtain the performance deviation value of each formulation candidate. S3. Select the candidate formulation with the smallest performance deviation value as the benchmark formulation, and adjust the component ratio of the benchmark formulation according to the performance deviation value to obtain the adjusted formulation. S4. The adjusted formula is virtually modulated and verified again. After confirming that its expected performance indicators meet the performance constraints of the target lubricating oil, the final formula is obtained and output as the original experimental formula.
2. The method according to claim 1, characterized in that, In S2, virtual modulation verification is performed on each candidate formulation scheme, specifically as follows: A pre-trained machine learning model is used to establish the mapping relationship between each additive in the candidate formulation and various properties of the lubricating oil under standard stirring process conditions, as well as the mapping relationship between each additive in the candidate formulation and the expected performance index. Based on the candidate formulation, the contribution coefficient of each additive to various properties of the lubricating oil under standard stirring process conditions and the expected performance index of the candidate formulation are obtained.
3. The method according to claim 1, characterized in that, In S2, the expected performance indicators are compared with the performance constraints of the target lubricating oil to obtain the performance deviation value of each formulation candidate scheme, specifically: The performance constraints of the target lubricating oil are represented as a constraint sequence, and the expected performance indicators of the candidate formulations are represented as an indicator sequence. The order of elements in the constraint sequence and indicator sequence is fixed according to the type of performance indicator; The minimum warping path distance between the standardized constraint sequence and the index sequence is calculated using the dynamic time warping algorithm, and this minimum warping path distance is used as the performance deviation value.
4. The method according to claim 1, characterized in that, In S3, the component ratios of the baseline formulation are adjusted based on the performance deviation value to obtain the adjusted formulation, specifically: Obtain the performance deviation value and the minimum regularization path distance, determine the candidate formulation with the smallest performance deviation value as the baseline formulation, and read the mass fraction of each additive in the baseline formulation. Based on the warping path output by the dynamic time warping algorithm, the performance index dimension with the largest deviation between the index sequence and the constraint sequence is identified, and this performance index dimension is determined as the main adjustment target. If the main adjustment target is less than the target performance constraint, then the additive with a positive contribution coefficient and the largest absolute value is selected as the target additive. If the main adjustment target is greater than the target performance constraint, then the additive with the negative contribution coefficient and the largest absolute value is selected as the target additive. Based on the contribution coefficient of the target additive, the adjustment ratio of the target additive is calculated to obtain the adjusted formulation.
5. The method according to claim 4, characterized in that, Based on the contribution coefficient of the target additive, the adjustment ratio of the target additive is calculated as follows: ， in, The adjustment ratio for the target additive. Adjust the target deviation value as the main adjustment. The contribution coefficient, To adjust the step size coefficient, This is a comprehensive correction factor.
6. The method according to claim 1, characterized in that, If multiple candidate formulations in S3 have the same performance deviation value, a stirring simulation is performed to simulate the expected performance index of the candidate formulations with the same performance deviation value under different stirring parameters, and the maximum value is taken to recalculate the performance deviation value.
7. The method according to claim 1, characterized in that, In S4, the adjusted formula is again virtually modulated and verified, specifically as follows: Using the same virtual modulation verification method as S2, the expected performance indicators of the adjusted formulation are calculated to determine whether they meet the performance constraints of the target lubricating oil. If the conditions are met, the adjusted recipe will be output as the final recipe. If the requirements are not met, repeat steps S3 to S4 for iterative optimization until the performance constraints are met or the maximum number of iterations is reached.
8. The method according to claim 1, characterized in that, S1 extracts at least three candidate formulations from the lubricating oil formulation database, specifically: The performance constraints of the target lubricating oil are transformed into a multi-dimensional performance vector. The standard performance vectors of each formula in the lubricating oil formula database are traversed, the weighted Euclidean distance is calculated, and the formulas are sorted from smallest to largest distance. The top three formulas are selected as candidate formula schemes.
9. A lubricating oil blending system for implementing the lubricating oil blending method according to any one of claims 1-8, characterized in that, include: The candidate solution generation module obtains the performance constraints of the target lubricating oil and extracts at least three formulation candidate solutions from the lubricating oil formulation database. The performance verification module performs virtual modulation verification on each candidate formulation scheme to obtain the expected performance index of each candidate formulation scheme. The expected performance indicators are compared with the performance constraints of the target lubricant to obtain the performance deviation value of each formulation candidate. The formulation adjustment module selects the candidate formulation with the smallest performance deviation value as the baseline formulation, and adjusts the component ratio of the baseline formulation according to the performance deviation value to obtain the adjusted formulation. The final formula output module performs virtual modulation verification on the adjusted formula again. After confirming that its expected performance indicators meet the performance constraints of the target lubricating oil, the final formula is obtained and output as the original experimental formula.
10. A lubricating oil blending device, characterized in that, It includes a processor and a memory, wherein the processor executes a computer program stored in the memory to implement a lubricating oil blending method as described in any one of claims 1-8.