Intelligent lubricating oil production management and control method and system

Abstract

The invention relates to the technical field of intelligent lubricating oil production, and discloses an intelligent lubricating oil production management and control method and system. The method comprises the following steps: receiving and analyzing a user production request to obtain a target performance set and base oil batch information; retrieving the basic oil depot to obtain attribute features, inputting the attribute features and the target performance set into a formula generation network, and calculating a theoretical feeding sequence of the additives; assigning a simulation production environment according to the sequence and operating simulation to obtain a preliminary process parameter boundary; combining the state data of the current production device with the boundary to generate an adaptive production process procedure of the batch; executing regulations and periodically collecting online physical property parameters of the mixture; inputting the online parameters into the state evaluation model, and generating a process adjustment instruction in real time; and adjusting dynamic control parameters in the regulation according to the instruction to realize closed-loop control. According to the invention, dynamic and accurate regulation and control from the formula to the production process are realized, and the flexibility, stability and product consistency of lubricating oil production are improved.

Description

Technical Field

[0001] This invention relates to the field of intelligent lubricating oil production technology, specifically to an intelligent lubricating oil production control method and system. Background Technology

[0002] Traditional lubricant production methods typically rely on fixed formulations and pre-set process parameters. In the formulation stage, additive types and amounts are usually selected based on standard base oil grades and a limited set of key performance indicators, using manual experience or simple linear calculations to create a universal formulation suitable for a specific type of oil. In the process control stage, an open-loop control model is commonly used, meaning production is executed according to pre-set parameters such as temperature, time, and stirring speed. This lacks direct feedback and control over the relationship between the real-time state of the mixture and the target performance of the finished product.

[0003] Existing technical solutions have shortcomings. Fixed formulation models cannot sensitively adapt to the natural fluctuations in the composition and properties of different batches of base oils, resulting in insufficient performance stability of the final product under the same formulation and making it difficult to achieve precise and flexible blending for specific performance combinations. Open-loop production control makes it impossible to perceive real-time changes in the mixing state of materials during the production process. It cannot correct deviations in the performance of intermediate products caused by feeding errors, uneven mixing, or different reaction degrees during the mixing stage. Problems can only be discovered after the final product is tested, resulting in unstable batch quality or production rework. How to dynamically generate precise formulations based on the individual characteristics of each batch of base oil, and how to dynamically adjust the process based on the real-time state of the mixture during production, are key issues that need to be addressed to achieve high-quality, stable, and intelligent production of lubricating oils. Summary of the Invention

[0004] The purpose of this invention is to provide an intelligent lubricating oil production control method and system to solve the problems mentioned in the background art.

[0005] To achieve the above objectives, the present invention provides an intelligent lubricating oil production control method, the method comprising: Receive and parse the user's production request to obtain the target performance set of the finished lubricating oil and the base oil batch information; Based on the base oil batch information, the base oil storage is retrieved to obtain the corresponding base oil attribute characteristics; The base oil properties and the target performance set of the finished lubricating oil are input into the formulation generation network to calculate the theoretical additive sequence. Based on the theoretical feeding sequence of the additives, parameters are assigned to the preset simulated production environment, and the mixing process simulation is run to obtain the preliminary process parameter boundaries. Collect the status data of the current production unit and dynamically couple it with the preliminary process parameter boundaries to generate a production process specification adapted to the current batch. During the production process, online physical property parameters of the mixture are periodically collected. The online physical property parameters are input into the state evaluation model to generate process adjustment instructions in real time. The dynamic control parameters in the production process procedure are adjusted according to the process adjustment instructions to achieve closed-loop control of the production process.

[0006] Preferably, the base oil depot is retrieved based on the base oil batch information to obtain the corresponding base oil attribute characteristics, including: Based on the base oil batch information, locate the corresponding base oil data record in the base oil depot; Extract the chemical composition spectrum, viscosity index data, pour point data, flash point data, and impurity content data of the base oil from the base oil data record; Feature extraction is performed on the chemical composition map to obtain the hydrocarbon distribution vector; The viscosity index data, pour point data, flash point data, and impurity content data are standardized and encoded to obtain a basic physical property vector. The hydrocarbon distribution vector is fused with the basic physical property vector to form the base oil attribute features that characterize the batch information of the base oil.

[0007] Preferably, the base oil properties and the target performance set of the finished lubricating oil are input into the formulation generation network to calculate the theoretical additive sequence, including: The pre-built recipe generation network includes a feature understanding layer, a demand alignment layer, and a sequence generation layer; The base oil property features are input into the feature understanding layer, and the base oil adaptation features are output. The target performance set of the finished lubricating oil is input into the demand alignment layer, and the target performance characteristics are output. In the sequence generation layer, the base oil adaptation features and target performance features are matched and mapped step by step. Based on the mapping results, the additive knowledge base is invoked to filter out a variety of candidate additives that meet the constraints and their theoretical addition amounts. Based on the preset feeding rule model, the various candidate additives and their theoretical addition amounts are sorted and conflict-resolved in a time sequence, and the theoretical feeding sequence of the additives, including the type of additive, the amount added, the order of addition, and the identification of the addition stage, is output.

[0008] Preferably, based on the theoretical feeding sequence of the additives, parameters are assigned to a preset simulated production environment, and a mixing process simulation is run to obtain preliminary process parameter boundaries, including: Load a preset simulated production environment that includes a reactor geometry model, a stirrer model, and a heat exchange model; The additive types, amounts, order of addition, and stage identifiers are extracted from the theoretical addition sequence of the additives and used as the input event stream for the simulation. Set the initial simulation parameters of the preset simulated production environment, including initial temperature, initial pressure, and initial base oil flow rate; The input event stream drives the preset simulated production environment to run a hybrid process simulation with multiphysics coupling. During the simulation, the temperature field distribution, shear stress distribution, additive dispersion index, and sedimentation tendency index of the simulated mixture are calculated and monitored in real time. When the simulation reaches the set termination conditions, the extreme or critical values ​​of the temperature field distribution, shear stress distribution, additive dispersion index, and sedimentation tendency index of the simulated mixture are extracted and defined as the preliminary process parameter boundaries. The preliminary process parameter boundaries include the highest safe temperature, the minimum allowable shear stress, the lowest dispersion threshold, and the maximum sedimentation rate threshold.

[0009] Preferably, the current status data of the production unit is collected and dynamically coupled with the preliminary process parameter boundaries to generate a production process specification adapted to the current batch, including: The status data of the current production device is collected in real time from the production line control system. The status data includes the real-time temperature inside the reactor, the current speed of the agitator, the current opening degree of the feed valve, and the current flow rate of the circulating pump. A process parameter adjustment mapping table is constructed, which defines the dynamic relationship between the initial process parameter boundaries and the state data; Based on the process parameter adjustment mapping table, the real-time temperature inside the reactor is compared with the maximum safe temperature to generate a temperature compensation coefficient. The current speed of the agitator, the current opening degree of the feed valve, and the current flow rate of the circulating pump are correlated with the minimum allowable shear stress and the minimum dispersion threshold to generate a coordinated adjustment strategy for flow rate and speed. By integrating the temperature compensation coefficient with the coordinated adjustment strategy of flow rate and rotation speed, and combining the theoretical feeding sequence of the additives, a production process specification adapted to the current batch is generated, which includes a time axis, target temperature at each stage, target stirring speed, target feed flow rate, and additive feeding instructions.

[0010] Preferably, the production process is executed, and online physical property parameters of the mixture are periodically collected during the material mixing process, including: The production line is controlled to operate according to the time axis, target temperature at each stage, target stirring speed, target feed flow rate and additive feeding instructions set according to the production process specification adapted to the current batch. During the mixing process, online physical property parameters of the mixture are collected at fixed intervals using an online sensor array installed inside the reactor. The online physical properties collected by the online sensor array include dielectric constant, dynamic viscosity, near-infrared spectral absorption characteristics, and ultrasonic wave propagation velocity. The dielectric constant, dynamic viscosity, near-infrared spectral absorption characteristics, and ultrasonic wave propagation velocity collected in each fixed period are preprocessed. The preprocessing includes filtering and noise reduction, outlier removal, and data normalization to form a standardized set of periodic online physical property parameters.

[0011] Preferably, the online physical property parameters are input into the state assessment model to generate process adjustment instructions in real time, including: The state assessment model is a neural network model pre-trained using historical production data, and its input dimension matches the dimension of the standardized set of periodic online physical property parameters. The standardized set of periodic online physical property parameters is input into the state assessment model; The state assessment model outputs a state assessment vector of the mixture, which includes a homogeneity score, a reaction progress estimate, and a stability deviation. The state evaluation vector is compared with the preset process standard vector to calculate the deviation vector; According to the preset adjustment strategy library, the deviation vector is mapped to specific process adjustment instructions, which include temperature fine-tuning, stirring speed adjustment, and feed flow correction.

[0012] Preferably, adjusting the dynamic control parameters in the production process specification according to the process adjustment command to achieve closed-loop control of the production process includes: Receive process adjustment instructions generated in real time by the state assessment model; Analyze the process adjustment commands and extract the temperature fine-tuning amount, stirring speed adjustment amount, and feed flow correction amount; Based on the temperature fine-tuning amount, update the target temperature values ​​for the current and subsequent stages in the production process procedure adapted to the current batch; Based on the stirring speed adjustment amount, update the stirring speed setting value in the current and subsequent stages of the production process procedure adapted to the current batch; Update the feed flow rate setting value for the current and subsequent stages in the production process procedure adapted to the current batch based on the feed flow rate correction amount; The updated target temperature value, stirring speed setting value, and feed flow rate setting value are sent to the production line control system, so that the production line executes the updated parameters and completes one closed-loop control cycle. Repeat the steps of collecting online physical property parameters, generating process adjustment instructions, adjusting production process specifications and issuing them until the production process ends.

[0013] Preferably, feature extraction is performed on the chemical composition spectrum to obtain a hydrocarbon distribution vector, including: The chemical composition spectrum is processed using a preset feature extraction algorithm to identify the characteristic peaks in the spectrum and their corresponding hydrocarbon types; For each identified hydrocarbon type, the relative percentage content of that hydrocarbon type in the base oil is calculated based on its corresponding characteristic peak area; According to the preset order of hydrocarbon types, the relative percentage content of each type of hydrocarbon is arranged in sequence to form an ordered numerical sequence. The ordered numerical sequence is normalized to eliminate dimensional differences caused by different detection equipment or conditions, resulting in a standardized hydrocarbon distribution vector.

[0014] Preferably, when the processor executes the computer program, it implements the steps of the intelligent lubricating oil production control method as described in any of the above-described methods.

[0015] Compared with the prior art, the beneficial effects of the present invention are: By introducing a formulation generation network, the specific batch-specific base oil properties and multiple target performance requirements of the finished lubricating oil are simultaneously taken as input. The network directly outputs a suitable theoretical feed sequence through internal model calculations. This eliminates the reliance on fixed formulations or simple empirical calculations, and can automatically resolve the nonlinear mapping relationship between base oil characteristics and complex performance targets. It achieves dynamic and precise formulation design, enabling each batch of production to generate customized additive addition schemes based on the actual properties of the base oil used. This ensures the attainability and accuracy of finished product performance indicators, improves the adaptability of production to different raw materials, and enhances the consistency of product performance.

[0016] By periodically collecting online physical property parameters of the mixture during the production process and inputting them into a state assessment model to generate process adjustment commands in real time, a real-time closed-loop control system is formed, which dynamically corrects the production process parameters during execution. This changes the traditional open-loop production mode that follows a fixed process, transforming production control from mechanically following process parameters to tracking and adjusting the real-time quality status of the product. It enables proactive intervention and dynamic optimization of the production process, and can promptly compensate for and correct performance deviations caused by various disturbances during the mixing stage, ensuring that the state of intermediate products always converges towards the target performance. This significantly improves the stability of the single-batch production process and the uniformity of the final product quality. Attached Figure Description

[0017] Figure 1 This is a schematic diagram illustrating the working principle of the intelligent lubricating oil production control method described in this invention. Figure 2 A flowchart for obtaining the property characteristics of base oils; Figure 3 A flowchart for generating a theoretical feeding sequence for a formula using a network calculation. Figure 4 A comparison chart of multiple indicators for assessing the status of the lubricating oil production process; Figure 5 A multi-parameter tracking diagram for closed-loop control of lubricant production. Detailed Implementation

[0018] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0019] Please see Figure 1This invention provides an intelligent lubricating oil production control method, comprising: receiving and parsing a user's production request to obtain the target performance set of the finished lubricating oil and base oil batch information; retrieving the base oil database based on the base oil batch information to obtain the corresponding base oil attribute characteristics; inputting the base oil attribute characteristics and the target performance set of the finished lubricating oil into a formulation generation network to calculate the theoretical additive feeding sequence; assigning parameter values ​​to a preset simulated production environment based on the theoretical additive feeding sequence, running a mixing process simulation to obtain preliminary process parameter boundaries; collecting the current production unit's status data and dynamically coupling it with the preliminary process parameter boundaries to generate a production process procedure adapted to the current batch; executing the production process procedure and periodically collecting online physical property parameters of the mixture during material mixing; inputting the online physical property parameters into a state evaluation model to generate process adjustment instructions in real time; adjusting the dynamic control parameters in the production process procedure according to the process adjustment instructions to achieve closed-loop control of the production process.

[0020] In one embodiment of the present invention, see [reference] Figure 2 Based on the base oil batch information, the corresponding base oil data record in the base oil warehouse is located. Chemical composition spectrum, viscosity index data, pour point data, flash point data, and impurity content data of the base oil are extracted from the base oil data record. Feature extraction is performed on the chemical composition spectrum to obtain a hydrocarbon distribution vector. This process uses a preset feature extraction algorithm to process the chemical composition spectrum, identifying characteristic peaks and their corresponding hydrocarbon types. For each identified hydrocarbon type, the relative percentage content of that hydrocarbon type in the base oil is calculated based on its corresponding characteristic peak area. The relative percentage contents of each hydrocarbon type are arranged sequentially according to a preset hydrocarbon type order, forming an ordered numerical sequence. This ordered numerical sequence is normalized to eliminate dimensional differences caused by different testing equipment or conditions, resulting in a standardized hydrocarbon distribution vector. The viscosity index data, pour point data, flash point data, and impurity content data are standardized and encoded to obtain a basic property vector. The hydrocarbon distribution vector and the basic property vector are fused to form the base oil attribute features characterizing the base oil batch information.

[0021] In practice, the received production request includes the base oil batch information "BO-2023-001". Based on the base oil batch information "BO-2023-001", the base oil database is searched to locate the corresponding base oil data record. The chemical composition spectrum, viscosity index data, pour point data, flash point data, and impurity content data of the base oil are extracted from the base oil data record. The chemical composition spectrum is stored in the form of a chromatogram, the viscosity index data is a value of 95, the pour point data is a value of -15 degrees Celsius, the flash point data is a value of 220 degrees Celsius, and the impurity content data is a percentage of 0.02%.

[0022] In some embodiments, feature extraction is performed on the chemical composition spectrum to obtain a hydrocarbon distribution vector. A preset feature extraction algorithm is used to process the chemical composition spectrum to identify the characteristic peaks in the spectrum and their corresponding hydrocarbon types. The preset feature extraction algorithm is based on peak detection and pattern matching, and the identified hydrocarbon types include n-alkanes, isoalkanes, cycloalkanes, and aromatics. For each identified hydrocarbon type, the relative percentage content of the hydrocarbon type in the base oil is calculated based on its corresponding characteristic peak area; for example, the characteristic peak area of ​​n-alkanes accounts for 30% of the total characteristic peak area, isoalkanes account for 40%, cycloalkanes account for 20%, and aromatics account for 10%. The relative percentage contents of each type of hydrocarbon are arranged sequentially according to a preset hydrocarbon type order to form an ordered numerical sequence; the preset hydrocarbon type order is n-alkanes, isoalkanes, cycloalkanes, and aromatics, so the ordered numerical sequence is [30, 40, 20, 10].

[0023] Optionally, the ordered numerical sequence can be normalized to eliminate dimensional differences caused by different detection equipment or conditions, resulting in a standardized hydrocarbon distribution vector. The normalization process uses the following formula: in: This represents the percentage content of the i-th hydrocarbon substance after normalization. This represents the percentage content of the i-th hydrocarbon substance before normalization. This represents the total number of hydrocarbon types. and It is an index variable. For the sequence [30,40,20,10], after normalization, we get the standardized hydrocarbon distribution vector [0.3,0.4,0.2,0.1].

[0024] In practice, viscosity index data, pour point data, flash point data, and impurity content data are standardized and encoded to obtain a basic property vector. The standardized encoding uses a minimum-maximum scaling method to map each data point to the interval [0,1]. For example, viscosity index data of 95 is encoded as 0.75, pour point data of -15 degrees Celsius is encoded as 0.6, flash point data of 220 degrees Celsius is encoded as 0.8, and impurity content data of 0.02% is encoded as 0.1. Therefore, the basic property vector is [0.75, 0.6, 0.8, 0.1].

[0025] In some embodiments, the hydrocarbon distribution vector and the basic property vector are fused to form the base oil attribute features characterizing the batch information of the base oil. The fusion operation is achieved by vector concatenation. The standardized hydrocarbon distribution vector is [0.3,0.4,0.2,0.1], the basic property vector is [0.75,0.6,0.8,0.1], and the concatenation yields the base oil attribute features [0.3,0.4,0.2,0.1,0.75,0.6,0.8,0.1]. It is understandable that the data comparison reflects the differences in the base oil property characteristics of different base oil batches. For example, the hydrocarbon distribution vector extracted from the chemical composition spectrum of another base oil batch, "BO-2023-002," is [0.2, 0.5, 0.2, 0.1], and the basic property vector is [0.7, 0.5, 0.9, 0.2]. After fusion, the base oil property characteristics are [0.2, 0.5, 0.2, 0.1, 0.7, 0.5, 0.9, 0.2], which are numerically different from the base oil property characteristics of "BO-2023-001." Optionally, the base oil data records in the base oil library also include other historical test data, but in this embodiment, only the specified chemical composition spectrum, viscosity index data, pour point data, flash point data, and impurity content data are extracted.

[0026] In one embodiment of the present invention, see [reference] Figure 3 The pre-constructed formula generation network includes a feature understanding layer, a demand alignment layer, and a sequence generation layer. The base oil attribute features are input into the feature understanding layer, which outputs base oil matching features. The target performance set of the finished lubricating oil is input into the demand alignment layer, which outputs target performance features. In the sequence generation layer, the base oil matching features and target performance features are matched and mapped step-by-step. Based on the mapping results, an additive knowledge base is invoked to select multiple candidate additives that meet the constraints and their theoretical addition amounts. According to a pre-defined feeding rule model, the multiple candidate additives and their theoretical addition amounts are sorted temporally and conflict resolved, outputting a theoretical feeding sequence of additives containing additive type, addition amount, addition order, and addition stage identifier.

[0027] In practical implementation, the pre-constructed formula generation network includes a feature understanding layer, a demand alignment layer, and a sequence generation layer. The feature understanding layer consists of a three-layer fully connected neural network, the demand alignment layer consists of a bidirectional long short-term memory network, and the sequence generation layer consists of an attention mechanism and a fully connected layer. The base oil attribute features are input into the feature understanding layer, and the output is the base oil adaptation features. For example, the base oil attribute features are an eight-dimensional vector [0.3, 0.4, 0.2, 0.1, 0.75, 0.6, 0.8, 0.1], which, after processing by the feature understanding layer, yields a five-dimensional base oil adaptation feature vector [0.12, 0.85, 0.33, 0.07, 0.61]. In some embodiments, the target performance set of the finished lubricating oil is input into the demand alignment layer, and the target performance features are output. The target performance set of the finished lubricating oil includes viscosity grade, anti-wear index, and anti-oxidation index, specifically "VG68", "anti-wear index greater than 90", and "anti-oxidation index greater than 80". The demand alignment layer encodes such text and numerical requirements into a unified feature representation and outputs a six-dimensional target performance feature vector [0.68, 0.92, 0.87, 0.15, 0.42, 0.55].

[0028] In the sequence generation layer, base oil adaptation features and target performance features are matched and mapped step-by-step. This matching and mapping is achieved by calculating attention weights between features and then weighted fusion. Based on the mapping results, an additive knowledge base is invoked to select multiple candidate additives and their theoretical addition amounts that meet the constraints. The additive knowledge base stores the performance contribution matrix, compatibility matrix, and base oil influence factor of the additives. The constraints include minimizing the target performance gap, the upper limit of the total cost of the additives, and the chemical compatibility between additives. For example, the selected candidate additives include "viscosity index improver A", "anti-wear agent B", and "antioxidant C", with theoretical addition amounts of 1.2%, 0.8%, and 0.5% by mass, respectively. Optionally, according to a preset feeding rule model, multiple candidate additives and their theoretical addition amounts are sorted temporally and conflict resolved. The output is a theoretical feeding sequence of additives containing the additive type, addition amount, addition order, and addition stage identifier. The preset feeding rule model defines the dissolution temperature range of the additives, the activation energy of the reaction with other additives, and the shear sensitivity. Conflict resolution is based on an optimization function, the function expression of which is: in: This represents the optimal feeding order that minimizes the objective function value. Represents all sorted sequences Find the sequence that minimizes the objective function value. It is an index of the current additives. It is the total number of candidate additives. Indicates in the sorted sequence The Middle Additives Normalized temperature adjustment requirements It is a dimensionless weighting coefficient used to balance the two effects. Indicates additives With the previous additive in the sequence The dimensionless process conflict score between them. After calculation, the theoretical addition sequence of the additives is: "Stage 1: Add anti-wear agent B, 0.8%; Stage 2: Add antioxidant C, 0.5%; Stage 3: Add viscosity index improver A, 1.2%".

[0029] It is understandable that data comparison reflects the difference between the mapping result of the sequence generation layer and the final theoretical feeding sequence when different base oil property characteristics or target performance sets are input. For example, when the base oil property characteristics become [0.2, 0.5, 0.2, 0.1, 0.7, 0.5, 0.9, 0.2] and the target performance set requires "VG46", the selected candidate additives become "viscosity index improver D" and "anti-wear agent B", and their theoretical addition amounts and feeding order also change accordingly. In some embodiments, the content of the additive knowledge base can be expanded through offline updates. Optionally, the network structure parameters of the feature understanding layer and the demand alignment layer can be adjusted through supervised training based on historical production data.

[0030] In one embodiment of the present invention, a preset simulated production environment, including a reactor geometry model, a stirrer model, and a heat exchange model, is loaded. The additive type, dosage, order of addition, and stage identifier are extracted from the theoretical additive feeding sequence and used as the input event stream for the simulation. Initial simulation parameters for the preset simulated production environment are set, including initial temperature, initial pressure, and initial base oil flow rate. The preset simulated production environment is driven by the input event stream to run a multiphysics-coupled mixing process simulation. During the simulation, the temperature field distribution, shear stress distribution, additive dispersion index, and sedimentation tendency index of the simulated mixture are calculated and monitored in real time. When the simulation reaches the set termination conditions, the extreme or critical values ​​of the temperature field distribution, shear stress distribution, additive dispersion index, and sedimentation tendency index of the simulated mixture are extracted and defined as the preliminary process parameter boundaries, including the maximum safe temperature, minimum allowable shear stress, minimum dispersion threshold, and maximum sedimentation rate threshold. The status data of the current production unit is collected in real time from the production line control system. This status data includes the real-time temperature inside the reactor, the current speed of the agitator, the current opening degree of the feed valve, and the current flow rate of the circulating pump. A process parameter adjustment mapping table is constructed, defining the dynamic correlation between the initial process parameter boundaries and the status data. Based on this mapping table, the real-time temperature inside the reactor is compared with the maximum safe temperature to generate a temperature compensation coefficient. The current speed of the agitator, the current opening degree of the feed valve, and the current flow rate of the circulating pump are correlated with the minimum allowable shear stress and the minimum dispersion threshold to generate a coordinated adjustment strategy for flow rate and speed. The temperature compensation coefficient and the coordinated adjustment strategy for flow rate and speed are integrated, along with the theoretical additive feeding sequence, to generate a production process specification adapted to the current batch, including a time axis, target temperatures at each stage, target agitator speeds, target feed flow rates, and additive feeding instructions.

[0031] In practical implementation, a pre-set simulated production environment is loaded, including a reaction vessel geometry model, a stirrer model, and a heat exchange model. This pre-set simulated production environment is a numerical simulation platform based on the coupling of computational fluid dynamics and the discrete element method. The additive types, amounts, order of addition, and stage identifiers are extracted from the theoretical additive addition sequence as input event streams for the simulation. For example, if the theoretical additive addition sequence is: "Stage 1: Add anti-wear agent B, 0.8%; Stage 2: Add antioxidant C, 0.5%; Stage 3: Add viscosity index improver A, 1.2%", then the input event stream is defined as follows: the "Add anti-wear agent B, 0.8%" event is triggered at simulation time 0 seconds, the "Add antioxidant C, 0.5%" event is triggered at simulation time 300 seconds, and the "Add viscosity index improver A, 1.2%" event is triggered at simulation time 600 seconds.

[0032] In some embodiments, preset initial simulation parameters for the simulated production environment are set, including an initial temperature of 60 degrees Celsius, an initial pressure of 1 standard atmosphere, and an initial base oil flow rate of 5 liters per second. The preset simulated production environment is driven by an input event stream to run a multiphysics-coupled mixing process simulation; the multiphysics coupling includes a fluid flow field, a heat transfer field, and a mass diffusion field. During the simulation, the temperature field distribution, shear stress distribution, additive dispersion index, and sedimentation tendency index of the simulated mixture are calculated and monitored in real time. The temperature field distribution of the simulated mixture is stored in a three-dimensional matrix, the shear stress distribution is recorded in a vector field, the additive dispersion index is obtained by calculating the coefficient of variation of the additive concentration within the grid cells, and the sedimentation tendency index is evaluated by simulating the Stokes settling velocity of particles in the flow field.

[0033] In practice, when the simulation reaches the set termination conditions, the extreme or critical values ​​of the temperature field distribution, shear stress distribution, additive dispersion index, and sedimentation tendency index of the simulated mixture are extracted and defined as the preliminary process parameter boundaries. The simulation termination condition is set as the completion of all additive events and the system state reaching a quasi-steady state. For example, the highest temperature value of 85 degrees Celsius during the entire simulation process is extracted from the temperature field distribution and defined as the maximum safe temperature; the minimum shear stress value of 0.5 Pascals at the tip of the agitator blade during the entire simulation process is extracted from the shear stress distribution and defined as the minimum allowable shear stress; the dispersion value of antioxidant C at the end of the simulation is 0.92, extracted from the additive dispersion index and defined as the minimum dispersion threshold; the maximum sedimentation velocity of anti-wear agent B particles during the simulation process is 0.001 mm / s, extracted from the sedimentation tendency index and converted into a maximum sedimentation rate threshold of 0.001 mm / s. Therefore, the preliminary process parameter boundaries include the maximum safe temperature of 85 degrees Celsius, the minimum allowable shear stress of 0.5 Pascals, the minimum dispersion threshold of 0.92, and the maximum sedimentation rate threshold of 0.001 mm / s.

[0034] Optionally, real-time status data of the current production unit can be collected from the production line control system. This status data includes the real-time temperature inside the reactor (78 degrees Celsius), the current agitator speed (200 rpm), the current opening degree of the feed valve (45%), and the current flow rate of the circulating pump (4.8 liters per second). A process parameter adjustment mapping table is constructed, defining the initial dynamic correlation between process parameter boundaries and status data. For example, one record in the mapping table is associated with the real-time temperature inside the reactor and the maximum safe temperature, defining a formula for calculating the temperature deviation. in: Indicates the temperature compensation coefficient. This indicates a maximum safe temperature of 85 degrees Celsius. This indicates that the real-time temperature inside the reactor is 78 degrees Celsius.

[0035] In some embodiments, a temperature compensation coefficient is generated by comparing the real-time temperature inside the reactor with the maximum safe temperature based on a process parameter adjustment mapping table; the temperature compensation coefficient is then calculated according to a formula. The value is (85-78) / 85, approximately equal to 0.082. A correlation analysis is performed between the current agitator speed, the current opening of the feed valve, the current flow rate of the circulating pump, and the minimum allowable shear stress and minimum dispersion threshold to generate a coordinated adjustment strategy for flow rate and speed. This correlation analysis is achieved by searching the "shear stress-speed" and "dispersion-flow rate" relationship curves in the preset process parameter adjustment mapping table. The generated coordinated adjustment strategy is: increase the agitator speed to 220 revolutions per minute, and simultaneously adjust the base oil feed flow rate to 5.1 liters per second. It can be understood that the data comparison reflects the different adjustment strategies generated when the collected state data is different; for example, if the real-time temperature inside the reactor is 82 degrees Celsius, the calculated temperature compensation coefficient will be different. The value becomes (85-82) / 85, approximately equal to 0.035, which reduces the adjustment range of the corresponding temperature control strategy. By integrating the temperature compensation coefficient with the coordinated adjustment strategy of flow rate and rotation speed, and combining it with the theoretical additive feeding sequence, a production process specification adapted to the current batch is generated, including a time axis, target temperature for each stage, target stirring speed, target feed flow rate, and additive feeding instructions. The generated production process specification text for the current batch is as follows: "0-300 seconds: target temperature 80 degrees Celsius, target stirring speed 220 rpm, target feed flow rate 5.1 L / s, execute the instruction to add anti-wear agent B; 300-600 seconds: target temperature 82 degrees Celsius, target stirring speed 220 rpm, target feed flow rate 5.1 L / s, execute the instruction to add antioxidant C; 600-900 seconds: target temperature 84 degrees Celsius, target stirring speed 210 rpm, target feed flow rate 5.0 L / s, execute the instruction to add viscosity index improver A."

[0036] In one embodiment of the present invention, the production line is controlled to operate according to the time axis, target temperature at each stage, target stirring speed, target feed flow rate, and additive feeding instructions set according to the production process specification adapted to the current batch. During the mixing process, online physical property parameters of the mixture are collected at fixed intervals by an online sensor array installed in the reactor. The online physical property parameters collected by the online sensor array include dielectric constant, dynamic viscosity, near-infrared spectral absorption characteristics, and ultrasonic wave propagation velocity. The dielectric constant, dynamic viscosity, near-infrared spectral absorption characteristics, and ultrasonic wave propagation velocity collected in each fixed period are preprocessed. The preprocessing includes filtering and noise reduction, outlier removal, and data normalization to form a standardized set of periodic online physical property parameters. The state evaluation model is a neural network model pre-trained using historical production data, and its input dimension matches the dimension of the standardized set of periodic online physical property parameters. The standardized set of periodic online physical property parameters is input into the state evaluation model, and the state evaluation model outputs a state evaluation vector of the mixture. The state evaluation vector includes a uniformity score, a reaction progress estimate, and a stability deviation. The state evaluation vector is compared with a preset process standard vector to calculate the deviation vector. Based on a preset adjustment strategy library, the deviation vector is mapped to specific process adjustment commands, including temperature fine-tuning, stirring speed adjustment, and feed flow correction.

[0037] In practice, the production line operates according to the timeline, target temperature, target stirring speed, target feed flow rate, and additive addition instructions set according to the production process specifications adapted to the current batch. For example, the production process specifications stipulate that in the 0-300 second stage, the target temperature is 80 degrees Celsius, the target stirring speed is 220 rpm, and the target feed flow rate is 5.1 liters / second. The production line control system drives the reactor heater, stirring motor, and feed pump valve to perform corresponding operations. During the mixing process, online physical property parameters of the mixture are collected at fixed intervals, with the fixed interval set to 10 seconds, through an online sensor array installed inside the reactor. The online sensor array includes a dielectric constant sensor, an online viscometer, a near-infrared spectral probe, and an ultrasonic sensor, which are integrated and installed at specific sampling points on the inner wall of the reactor.

[0038] In some embodiments, the online physical property parameters acquired by the online sensor array include dielectric constant, dynamic viscosity, near-infrared spectral absorption characteristics, and ultrasonic wave propagation velocity. Specific raw data obtained within one acquisition cycle are, for example: dielectric constant 2.3, dynamic viscosity 65 centistokes, near-infrared absorption peak at a specific wavenumber of 0.45, and ultrasonic wave propagation velocity 1450 meters per second. The dielectric constant, dynamic viscosity, near-infrared spectral absorption characteristics, and ultrasonic wave propagation velocity acquired in each fixed cycle are preprocessed. Preprocessing includes filtering and denoising, outlier removal, and data normalization to form a standardized set of periodic online physical property parameters. Filtering and denoising employs a moving average filtering algorithm, outlier removal is based on the Laida criterion, and data normalization linearly transforms each parameter to the [0,1] interval. Refer to Table 1 for an example of a processed standardized set of periodic online physical property parameters.

[0039] Table 1: Standardized Online Physical Property Parameters Table Time point (seconds) Dielectric constant (normalized) Dynamic viscosity (normalized) Near-infrared absorption characteristics (normalized) Ultrasonic wave propagation velocity (standardized) 10 0.12 0.15 0.08 0.22 20 0.18 0.23 0.15 0.30 ... ... ... ... ... In practical implementation, the state assessment model is a neural network model pre-trained using historical production data. Its input dimension matches the dimension of a standardized set of periodic online physical property parameters. Specifically, the training method for the state assessment model using historical production data involves pre-collecting a large number of historical lubricant production batch data records. Each data record contains a standardized set of periodic online physical property parameters as model input, and a pre-defined process standard vector based on historical successful production experience as the model's output target value. This data is used to construct a training dataset, and the neural network model is trained using a supervised learning algorithm. During training, the connection weights and bias parameters between network layers are adjusted to minimize the difference between the model's predicted state assessment vector and the pre-defined process standard vector, enabling the model to accurately identify the homogeneity, reaction progress, and stability of the mixture from real-time online physical property parameters. The neural network model has an input layer with 128 nodes, corresponding to 4 parameters at each sampling time. If a time window contains 5 consecutive periods, the input dimension is 20. The standardized set of periodic online physical property parameters is input into the state assessment model. When inputting, the standardized parameter values ​​of the most recent five consecutive periods are arranged in chronological order into a one-dimensional vector, which is used as one input to the state assessment model.

[0040] Optionally, the state assessment model outputs a state assessment vector for the mixture, which includes a homogeneity score, reaction progress estimate, and stability deviation. The last layer of the state assessment model is an output layer with three nodes, corresponding to the three assessment indicators mentioned above. For example, for a vector input at a certain moment, the state assessment model outputs a state assessment vector of [0.87, 0.45, 0.12]. The state assessment vector is compared with a preset process standard vector to calculate the deviation vector. The preset process standard vector defines the ideal homogeneity score, reaction progress estimate, and stability deviation at the current production stage. For example, in the initial mixing stage, the preset process standard vector is [0.95, 0.30, 0.05]. The deviation vector is calculated by element-wise subtraction. in: Represents the deviation vector. This represents the actual state evaluation vector output by the state evaluation model. This represents the preset process standard vector. Based on the above data, the deviation vector is calculated to be [0.87-0.95, 0.45-0.30, 0.12-0.05], which is [-0.08, 0.15, 0.07].

[0041] In some embodiments, the deviation vector is mapped to specific process adjustment instructions according to a preset adjustment strategy library. The process adjustment instructions include temperature fine-tuning, stirring speed adjustment, and feed flow correction. The preset adjustment strategy library is a lookup table that defines the adjustment instructions corresponding to different ranges of the deviation vector. For example, for the deviation vector [-0.08, 0.15, 0.07], the process adjustment instructions obtained by querying the adjustment strategy library are: temperature fine-tuning +1.5 degrees Celsius, stirring speed adjustment +5 rpm, and feed flow correction -0.05 liters / second. It is understandable that data comparisons show that state evaluation vectors at different times or in different batches will lead to different process adjustment instructions. For example, at another time, if the state evaluation vector is [0.97, 0.28, 0.02], and the calculated deviation vector is [0.02, -0.02, -0.03], then the process adjustment instructions obtained from the adjustment strategy library would be: temperature fine-tuning -0.5 degrees Celsius, stirring speed adjustment -2 rpm, and feed flow correction +0.01 L / s. Optionally, the content of the preset adjustment strategy library can be optimized and updated through rule-based learning algorithms.

[0042] See Figure 4This is a multi-indicator comparison chart for assessing the state of the lubricating oil production process, used to show the differences between the actual state and the process standards at different production time points. The actual uniformity is generally close to the standard value, but gradually decreases with production time, indicating a slight decrease in mixing uniformity in the later stages. The actual reaction progress continuously increases over time, significantly exceeding the standard value, indicating that the reaction process is faster than expected. The actual stability deviation gradually increases over time, approaching the standard value in the later stages, indicating a decrease in the stability of the mixture in the later stages of production. This type of chart is a core reference tool for closed-loop control of lubricating oil production, intuitively comparing the deviation between the actual state and the process standards, quickly locating production anomalies, assessing the compliance of the production process, and ensuring that the finished product performance meets standards.

[0043] In one embodiment of the present invention, a process adjustment command generated in real time by the state evaluation model is received. The process adjustment command is parsed to extract the temperature fine-tuning amount, stirring speed adjustment amount, and feed flow correction amount. Based on the temperature fine-tuning amount, the target temperature value for the current and subsequent stages of the production process adapted to the current batch is updated. Based on the stirring speed adjustment amount, the stirring speed setting value for the current and subsequent stages of the production process adapted to the current batch is updated. Based on the feed flow correction amount, the feed flow setting value for the current and subsequent stages of the production process adapted to the current batch is updated. The updated target temperature value, stirring speed setting value, and feed flow setting value are sent to the production line control system, causing the production line to execute the updated parameters, completing one closed-loop control cycle. The steps of collecting online physical property parameters, generating process adjustment commands, adjusting the production process, and sending them are repeated until the production process ends.

[0044] In practical implementation, the system receives real-time process adjustment commands generated by the state assessment model. For example, when the production line reaches the 350th second of operation, the received process adjustment command is "temperature fine-tuning +1.5 degrees Celsius, stirring speed adjustment +5 rpm, feed flow correction -0.05 liters / second". The system parses the process adjustment command, extracting the temperature fine-tuning, stirring speed adjustment, and feed flow correction. The command decoding module converts the text commands into numerical variables: the temperature fine-tuning is interpreted as +1.5, the stirring speed adjustment as +5, and the feed flow correction as -0.05.

[0045] Based on the temperature fine-tuning amount, update the target temperature values ​​for the current and subsequent stages in the production process specification adapted to the current batch. Before the update, the production process specification adapted to the current batch specified that the target temperature for the 300-600 second stage was 82 degrees Celsius, and the target temperature for the 600-900 second stage was 84 degrees Celsius. The temperature fine-tuning amount of +1.5 degrees Celsius means that the target temperature for the current and subsequent unexecuted stages needs to be uniformly increased by 1.5 degrees Celsius. After the update, the target temperature value for the 300-600 second stage becomes 83.5 degrees Celsius, and the target temperature value for the 600-900 second stage becomes 85.5 degrees Celsius. Based on the stirring speed adjustment, update the stirring speed setting values ​​for the current and subsequent stages in the production process specification adapted to the current batch. Before the update, the target stirring speed for the 300-600 second stage was 220 rpm, and the target stirring speed for the 600-900 second stage was 210 rpm. After applying the stirring speed adjustment of +5 rpm, the stirring speed setting value for the 300-600 second stage becomes 225 rpm, and the stirring speed setting value for the 600-900 second stage becomes 215 rpm. Based on the feed flow rate correction, update the feed flow rate settings for the current and subsequent stages of the production process specification adapted to the current batch. Before the update, the target feed flow rate for the 300-600 second stage was 5.1 L / s, and the target feed flow rate for the 600-900 second stage was 5.0 L / s. After applying the feed flow rate correction of -0.05 L / s, the feed flow rate setting for the 300-600 second stage becomes 5.05 L / s, and the feed flow rate setting for the 600-900 second stage becomes 4.95 L / s. The general logic for parameter updates can be expressed as follows: in: Indicates the post-update phase A certain dynamic control parameter setting value, Indicates the pre-update stage The original setting value, This represents the corresponding fine-tuning amount parsed from the process adjustment command. This represents the identifier for the current and all subsequent production stages.

[0046] In some embodiments, updated target temperature values, stirring speed setpoints, and feed flow rate setpoints are sent to the production line control system, enabling the production line to execute the updated parameters and complete one closed-loop control cycle. The sent actions are written into the corresponding control loop of the production line control system via an industrial communication protocol. For example, the reactor temperature controller setpoint is modified from 82 degrees Celsius to 83.5 degrees Celsius, the stirring motor speed setpoint is modified from 220 rpm to 225 rpm, and the feed flow rate controller setpoint is modified from 5.1 L / s to 5.05 L / s. The steps of collecting online physical property parameters, generating process adjustment commands, adjusting the production process specifications, and issuing them are repeated until the production process ends. For example, after the first adjustment is completed at 350 seconds, the system continues to collect online physical property parameters at fixed intervals such as 360 seconds and 370 seconds, and triggers new process adjustment commands, thereby continuously and dynamically adjusting the production process specifications.

[0047] It is understandable that data comparisons show that different process adjustment commands received at different times will lead to different modifications to the production process specifications. For example, if the process adjustment command received at another time is "temperature fine-tuning -0.5 degrees Celsius, stirring speed adjustment -2 rpm, feed flow correction +0.01 L / s", then the updated target temperature value for the 300-600 second stage will become 81.5 degrees Celsius, the stirring speed setting will become 218 rpm, and the feed flow setting will become 5.11 L / s. Optionally, the parameter update logic can be attenuated or amplified according to preset rules when applied to subsequent stages. It is understandable that after each parameter update and issuance, the production line control system will drive the actuators to move to approach the new setpoint. In specific implementation, the updated production process specifications are kept in memory as the current valid version, serving as the basis for the next adjustment.

[0048] See Figure 5 This is a multi-parameter tracking chart for closed-loop control in lubricant production, used to display the dynamic changes of the original setpoints and the adjusted setpoints for temperature, speed, and flow rate during the production process. After a 600-second production stage switch, the updated setpoints for temperature, speed, and flow rate all deviate from the original setpoints, demonstrating the closed-loop control's parameter adaptation to different production stages. Temperature and speed fluctuate in opposite directions, while flow rate shows a step-like adjustment with stage switching, reflecting the logic of multi-parameter collaborative control. The updated setpoints are dynamically fine-tuned around the original setpoints, maintaining the process baseline while optimizing the production state through real-time adjustments. This type of chart is a core monitoring tool for closed-loop control in lubricant production, intuitively tracking the dynamic adjustment process of multiple parameters and verifying the effectiveness of closed-loop control.

[0049] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus.

[0050] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A method for intelligent production control of lubricating oil, characterized in that, The method includes: Receive and parse the user's production request to obtain the target performance set of the finished lubricating oil and the base oil batch information; Based on the base oil batch information, the base oil storage is retrieved to obtain the corresponding base oil attribute characteristics; The base oil properties and the target performance set of the finished lubricating oil are input into the formulation generation network to calculate the theoretical additive sequence. Based on the theoretical feeding sequence of the additives, parameters are assigned to the preset simulated production environment, and the mixing process simulation is run to obtain the preliminary process parameter boundaries. Collect the status data of the current production unit and dynamically couple it with the preliminary process parameter boundaries to generate a production process specification adapted to the current batch. During the production process, online physical property parameters of the mixture are periodically collected. The online physical property parameters are input into the state evaluation model to generate process adjustment instructions in real time. The dynamic control parameters in the production process procedure are adjusted according to the process adjustment instructions to achieve closed-loop control of the production process.

2. The intelligent lubricating oil production control method as described in claim 1, characterized in that, Based on the base oil batch information, the base oil storage facility is retrieved to obtain the corresponding base oil attribute characteristics, including: Based on the base oil batch information, locate the corresponding base oil data record in the base oil depot; Extract the chemical composition spectrum, viscosity index data, pour point data, flash point data, and impurity content data of the base oil from the base oil data record; Feature extraction is performed on the chemical composition map to obtain the hydrocarbon distribution vector; The viscosity index data, pour point data, flash point data, and impurity content data are standardized and encoded to obtain a basic physical property vector. The hydrocarbon distribution vector is fused with the basic physical property vector to form the base oil attribute features that characterize the batch information of the base oil.

3. The intelligent lubricating oil production control method as described in claim 2, characterized in that, The base oil properties and the target performance set of the finished lubricating oil are input into the formulation generation network to calculate the theoretical additive sequence, including: The pre-built recipe generation network includes a feature understanding layer, a demand alignment layer, and a sequence generation layer; The base oil property features are input into the feature understanding layer, and the base oil adaptation features are output. The target performance set of the finished lubricating oil is input into the demand alignment layer, and the target performance characteristics are output. In the sequence generation layer, the base oil adaptation features and target performance features are matched and mapped step by step. Based on the mapping results, the additive knowledge base is invoked to filter out a variety of candidate additives that meet the constraints and their theoretical addition amounts. Based on the preset feeding rule model, the various candidate additives and their theoretical addition amounts are sorted and conflict-resolved in a time sequence, and the theoretical feeding sequence of the additives, including the type of additive, the amount added, the order of addition, and the identification of the addition stage, is output.

4. The intelligent lubricating oil production control method as described in claim 3, characterized in that, Based on the theoretical feeding sequence of the additives, parameters are assigned to the preset simulated production environment, and the mixing process simulation is run to obtain preliminary process parameter boundaries, including: Load a preset simulated production environment that includes a reactor geometry model, a stirrer model, and a heat exchange model; The additive types, amounts, order of addition, and stage identifiers are extracted from the theoretical addition sequence of the additives and used as the input event stream for the simulation. Set the initial simulation parameters of the preset simulated production environment, including initial temperature, initial pressure, and initial base oil flow rate; The input event stream drives the preset simulated production environment to run a hybrid process simulation with multiphysics coupling. During the simulation, the temperature field distribution, shear stress distribution, additive dispersion index, and sedimentation tendency index of the simulated mixture are calculated and monitored in real time. When the simulation reaches the set termination conditions, the extreme or critical values ​​of the temperature field distribution, shear stress distribution, additive dispersion index, and sedimentation tendency index of the simulated mixture are extracted and defined as the preliminary process parameter boundaries. The preliminary process parameter boundaries include the highest safe temperature, the minimum allowable shear stress, the lowest dispersion threshold, and the maximum sedimentation rate threshold.

5. The intelligent lubricating oil production control method as described in claim 4, characterized in that, Collect the current status data of the production unit and dynamically couple it with the preliminary process parameter boundaries to generate a production process specification adapted to the current batch, including: The status data of the current production device is collected in real time from the production line control system. The status data includes the real-time temperature inside the reactor, the current speed of the agitator, the current opening degree of the feed valve, and the current flow rate of the circulating pump. A process parameter adjustment mapping table is constructed, which defines the dynamic relationship between the initial process parameter boundaries and the state data; Based on the process parameter adjustment mapping table, the real-time temperature inside the reactor is compared with the maximum safe temperature to generate a temperature compensation coefficient. The current speed of the agitator, the current opening degree of the feed valve, and the current flow rate of the circulating pump are correlated with the minimum allowable shear stress and the minimum dispersion threshold to generate a coordinated adjustment strategy for flow rate and speed. By integrating the temperature compensation coefficient with the coordinated adjustment strategy of flow rate and rotation speed, and combining the theoretical feeding sequence of the additives, a production process specification adapted to the current batch is generated, which includes a time axis, target temperature at each stage, target stirring speed, target feed flow rate, and additive feeding instructions.

6. The intelligent lubricating oil production control method as described in claim 5, characterized in that, During the production process, the online physical property parameters of the mixture are periodically collected during the material mixing process, including: The production line is controlled to operate according to the time axis, target temperature at each stage, target stirring speed, target feed flow rate and additive feeding instructions set according to the production process specification adapted to the current batch. During the mixing process, online physical property parameters of the mixture are collected at fixed intervals using an online sensor array installed inside the reactor. The online physical properties collected by the online sensor array include dielectric constant, dynamic viscosity, near-infrared spectral absorption characteristics, and ultrasonic wave propagation velocity. The dielectric constant, dynamic viscosity, near-infrared spectral absorption characteristics, and ultrasonic wave propagation velocity collected in each fixed period are preprocessed. The preprocessing includes filtering and noise reduction, outlier removal, and data normalization to form a standardized set of periodic online physical property parameters.

7. The intelligent lubricating oil production control method as described in claim 6, characterized in that, The online physical property parameters are input into the state assessment model to generate process adjustment instructions in real time, including: The state assessment model is a neural network model pre-trained using historical production data, and its input dimension matches the dimension of the standardized set of periodic online physical property parameters. The standardized set of periodic online physical property parameters is input into the state assessment model; The state assessment model outputs a state assessment vector of the mixture, which includes a homogeneity score, a reaction progress estimate, and a stability deviation. The state evaluation vector is compared with the preset process standard vector to calculate the deviation vector; According to the preset adjustment strategy library, the deviation vector is mapped to specific process adjustment instructions, which include temperature fine-tuning, stirring speed adjustment, and feed flow correction.

8. The intelligent lubricating oil production control method as described in claim 7, characterized in that, Adjusting the dynamic control parameters in the production process specification according to the process adjustment instructions to achieve closed-loop control of the production process includes: Receive process adjustment instructions generated in real time by the state assessment model; Analyze the process adjustment commands and extract the temperature fine-tuning amount, stirring speed adjustment amount, and feed flow correction amount; Based on the temperature fine-tuning amount, update the target temperature values ​​for the current and subsequent stages in the production process procedure adapted to the current batch; Based on the stirring speed adjustment amount, update the stirring speed setting value in the current and subsequent stages of the production process procedure adapted to the current batch; Update the feed flow rate setting value for the current and subsequent stages in the production process procedure adapted to the current batch based on the feed flow rate correction amount; The updated target temperature value, stirring speed setting value, and feed flow rate setting value are sent to the production line control system, so that the production line executes the updated parameters and completes one closed-loop control cycle. Repeat the steps of collecting online physical property parameters, generating process adjustment instructions, adjusting production process specifications and issuing them until the production process ends.

9. The intelligent lubricating oil production control method as described in claim 2, characterized in that, Feature extraction is performed on the chemical composition spectrum to obtain hydrocarbon distribution vectors, including: The chemical composition spectrum is processed using a preset feature extraction algorithm to identify the characteristic peaks in the spectrum and their corresponding hydrocarbon types; For each identified hydrocarbon type, the relative percentage content of that hydrocarbon type in the base oil is calculated based on its corresponding characteristic peak area; According to the preset order of hydrocarbon types, the relative percentage content of each type of hydrocarbon is arranged in sequence to form an ordered numerical sequence. The ordered numerical sequence is normalized to eliminate dimensional differences caused by different detection equipment or conditions, resulting in a standardized hydrocarbon distribution vector.

10. An intelligent lubricating oil production control system, comprising a memory, a processor, and a computer program stored in the memory and running on the processor, characterized in that, When the processor executes the computer program, it implements the steps of the intelligent lubricating oil production control method as described in any one of claims 1 to 9.

Images