A method and system for early warning and judgment of lubricating oil oxidation state based on multimodal data feedback

Abstract

This invention discloses a method and system for early warning and judgment of lubricating oil oxidation state based on multimodal data feedback. The method includes: collecting chemical structure information of the lubricating oil; preprocessing the chemical structure information to obtain the lubricating oil molecular chain; extracting key structural parameters from the lubricating oil molecular chain to simulate the oxidation process, obtaining simulated data records; analyzing the changing trends of the oxidation process in the simulated data records; analyzing boundary intervals based on the changing trends of the oxidation process to obtain a simulated data distribution dataset; analyzing the simulated data distribution dataset to obtain the microstructural characteristics of the lubricating oil; comparing and verifying the microstructural characteristics of the lubricating oil from historically collected data with the microstructural characteristics of the lubricating oil from the oxidation process simulation, obtaining the microstructural verification results of the lubricating oil; and issuing an early warning based on the microstructural verification results of the lubricating oil. By simultaneously identifying both chemical and physical aspects, the method can more effectively provide early warnings based on the identification results of the lubricating oil.

Description

Technical Field

[0001] This invention relates to the field of lubricating oil oxidation early warning analysis, and in particular to a method and system for judging the early warning state of lubricating oil oxidation based on multimodal data feedback. Background Technology

[0002] With the continuous advancement of industrialization, lubricating oil is increasingly widely used in various mechanical equipment, especially in high-temperature, high-pressure, and extreme working environments, where its role is particularly important. However, during long-term use, lubricating oil is susceptible to oxidation reactions due to factors such as temperature, oxygen, and pressure, leading to a decline in its performance and consequently affecting the operating efficiency and lifespan of the equipment.

[0003] Lubricating oil oxidation is a significant indicator of lubricating oil aging. Oxidation causes changes in the molecular structure of the lubricating oil, forming deposits such as acidic substances, polymers, and gums, further affecting its rheological properties and lubrication performance. Traditional methods for assessing lubricating oil oxidation rely primarily on physical indicators such as viscosity, acid value, and base value. While these methods provide some information, the complexity and variability of the lubricating oil oxidation process mean that a single physical indicator often fails to comprehensively and accurately reflect the oxidation state of the lubricating oil.

[0004] In recent years, with the development of science and technology, especially the advancement of multimodal data analysis technology, it has become possible to determine the oxidation state of lubricating oil by combining chemical structural characteristics and physical properties. By collecting various data types of lubricating oil, such as infrared spectroscopy, dynamic viscosity, and molecular weight distribution, it is possible to comprehensively analyze the molecular chain structure, oxidation degree, and physical properties of lubricating oil, thereby more accurately predicting the oxidation trend and service life of lubricating oil. Summary of the Invention

[0005] The purpose of this invention is to provide a method and system for early warning and judgment of lubricating oil oxidation state based on multimodal data feedback, which solves the above-mentioned technical problems pointed out in the prior art.

[0006] This invention provides a method for early warning and judgment of lubricating oil oxidation state based on multimodal data feedback, comprising the following steps:

[0007] Chemical structure information of the lubricating oil is collected, and the chemical structure information is preprocessed to obtain the molecular chain of the lubricating oil.

[0008] Key structural parameters of the lubricating oil molecular chain were extracted to simulate the oxidation process, resulting in simulated data records. The variation trend of the oxidation process was analyzed from the simulated data records. Boundary intervals were analyzed based on the variation trend of the oxidation process to obtain a simulated data distribution dataset. The microstructural characteristics of the lubricating oil were obtained by analyzing the simulated data distribution dataset.

[0009] The simulated data record includes the carbonyl absorption peak intensity, hydroxyl absorption peak width, dynamic viscosity value, and molecular weight distribution width.

[0010] The microstructural characteristics of the lubricating oil are collected from historical data. The microstructural characteristics of the lubricating oil obtained from the historical data are compared and verified with the microstructural characteristics of the lubricating oil simulated by the oxidation process to obtain the microstructural verification results of the lubricating oil. Early warning is issued based on the microstructural verification results of the lubricating oil.

[0011] Compared with the prior art, the embodiments of the present invention have at least the following technical advantages:

[0012] Analysis of the multimodal data feedback method and system for early warning judgment of lubricating oil oxidation state provided by the present invention reveals that, in specific applications, structural parameters (such as bond length, bond angle, and intermolecular forces) in the lubricating oil molecular chain are extracted to generate a basic dataset, which provides initial conditions for subsequent oxidation reaction simulation. Molecular dynamics algorithms are used to simulate the oxidation process of lubricating oil under different oxidation conditions (such as oxygen concentration and temperature), collecting oxidation process data at different time points and extracting evolution parameters, such as absorption peak shift, chain breakage, and reaction rate. A time-series change graph is then constructed using simulated data records to analyze the changing trends of carbonyl absorption peak intensity and hydroxyl absorption peak width, thereby locating the inflection point of the hydroxyl absorption peak and identifying the key inflection stages in the oxidation process. The average molecular weight, variance, and standard deviation of the molecular weight distribution are calculated, and based on the dynamic change trend and combined with changes in dynamic viscosity values, the physical change range is further determined. Combining the chemical change range (such as the inflection point of the hydroxyl absorption peak) and the physical change range (such as the rate of change of the molecular weight distribution width), the distribution characteristics of the simulated data are obtained, thereby evaluating the oxidation state and performance of the lubricating oil.

[0013] Furthermore, by precisely analyzing the dynamic changes of lubricating oil molecular chains, early detection and performance prediction of lubricating oil oxidation and degradation processes can be achieved. Based on a comprehensive analysis of simulated and time-series data, starting from multiple parameters such as bond length, carbonyl absorption peak intensity, and dynamic viscosity, key structural parameters are extracted, inflection points are identified, and change mapping relationships are established to identify whether bond breakage, recombination, and changes in molecular weight distribution occur in the lubricating oil molecular chains. Then, using the bond length sequence in the simulated data, combined with the analysis of the mean and standard deviation, it is determined whether bond breakage has occurred in the lubricating oil molecular chains. The extraction and analysis of key structural parameters can intuitively reflect the molecular structure of lubricating oil. By analyzing the chain change trend, we can identify molecular chain breakage caused by oxidation. Furthermore, by combining the inflection points of carbonyl absorption peak intensity and dynamic viscosity values, we can establish a mapping relationship between the two, further enhancing the accuracy of identifying subtle characteristics of lubricating oil. By calculating the bond breakage and recombination time points, we can reconstruct the dynamic trajectory of lubricating oil molecular chains. This trajectory reveals the entire process of molecular chains from integrity to breakage and then recombination, providing important evidence for a deeper understanding of the degradation process of lubricating oil. Simultaneously, by analyzing the changes in the ratio of carbonyl absorption peaks to hydroxyl absorption peaks, we can further analyze the changes in molecular weight distribution and identify the characteristics of the accumulation of oxidation products and molecular aggregation phenomena in lubricating oil.

[0014] Furthermore, through a series of data preprocessing, dimensionality reduction analysis, and modeling methods, this study delves into the microstructural changes of lubricating oil during oxidation, ultimately providing effective data support for lubricating oil condition monitoring and early warning. This method not only effectively identifies key factors affecting the lubricating oil oxidation process but also comprehensively evaluates the reaction process through multi-dimensional data analysis, improving the accuracy and reliability of early warnings. Simultaneously, by organizing molecular structure evolution trajectory data into an initial data matrix, the data structure becomes clearer and more standardized. Matrix-formatted data can transform time series data into a format that machine learning algorithms can process, providing a foundation for dimensionality reduction analysis and model construction. Principal component analysis (PCA) dimensionality reduction effectively removes redundant data, reduces computational load, and extracts the most representative features, helping to identify… Key structural change factors during oxidation are identified. Next, by calculating the correlation between oxidation reaction rate and molecular structure changes, the accelerated phase of the oxidation process can be identified, providing early warnings for equipment maintenance. This helps predict the trend of lubricating oil oxidation reactions, promptly detect anomalies, and prevent equipment failure or damage. Furthermore, analysis of bond breaking and recombination event frequencies reveals important reaction events during oxidation. Additionally, the construction of a multivariate data matrix, integrating carbonyl absorption peak intensity and dynamic viscosity trends, not only comprehensively reflects the evolution of the oxidation process but also provides rich input data for dimensionality reduction analysis. Finally, by screening principal component eigenvectors, the most representative microstructural features can be extracted, providing quantitative indicators of the lubricating oil's oxidation state and improving the accuracy of data analysis. Attached Figure Description

[0015] Figure 1 This is a flowchart of a method for early warning and judgment of lubricating oil oxidation state based on multimodal data feedback, as described in Example 1.

[0016] Figure 2 This is an overall flowchart of a lubricating oil oxidation state early warning and judgment method based on multimodal data feedback, as described in Example 1.

[0017] Figure 3 This is a flowchart illustrating the process of obtaining microstructural features through a change range in a multimodal data feedback method for early warning and judgment of lubricating oil oxidation state, as described in Example 1.

[0018] Figure 4 This is a schematic diagram illustrating the changing trend of a multimodal data feedback method for early warning and judgment of lubricating oil oxidation state, as described in Example 1.

[0019] Figure 5 This is a flowchart illustrating the process of obtaining microstructural features through bond breakage and recombination in a multimodal data feedback method for early warning judgment of lubricating oil oxidation state, as described in Example 1.

[0020] Figure 6 This is a schematic diagram illustrating the identification inflection point of a continuous time-series curve in a multimodal data feedback method for early warning judgment of lubricating oil oxidation state, as described in Example 1.

[0021] Figure 7 This is a flowchart illustrating the process of determining the microstructure characteristics of lubricating oil oxidation state using a multimodal data feedback method, as described in Example 1.

[0022] Figure 8 This is a flowchart of a lubricating oil oxidation state early warning and judgment system based on multimodal data feedback, as described in Example 2.

[0023] Labels: Acquisition module 10; Identification module 20; Early warning module 30. Detailed Implementation

[0024] The technical solution of the present invention will now be clearly and completely described with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. 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.

[0025] The present invention will now be described in further detail with reference to specific embodiments and accompanying drawings.

[0026] Example 1

[0027] like Figure 1 As shown, Figure 2 As shown in the figure, this embodiment of the invention provides a method for early warning and judgment of lubricating oil oxidation state based on multimodal data feedback, including the following operation steps:

[0028] S1: Collect chemical structure information of the lubricating oil, preprocess the chemical structure information to obtain the lubricating oil molecular chain;

[0029] It should be noted that characterization instruments such as infrared spectroscopy (IR) (for example, hydroxyl groups are collected using infrared spectroscopy), nuclear magnetic resonance (NMR), Raman spectroscopy, and mass spectrometry (MS) are used to analyze the composition of lubricating oil and collect chemical structure information in the samples.

[0030] By collecting chemical structure information of lubricating oil, a comprehensive understanding of its molecular chain characteristics can be obtained. Chemical structure information generally includes the molecular chain composition of the lubricating oil, as well as the attributes of its chemical structure, such as chain length and composition. The molecular chain length and composition of lubricating oil significantly affect its performance, especially during oxidation, where changes in molecular structure directly determine its service life and efficiency. Preprocessing steps may involve noise removal and data normalization, laying the foundation for subsequent oxidation process simulation and performance assessment. Accurately extracting the chemical characteristics of the molecular chains (i.e., the molecular components and molecular chain structure of the lubricating oil) to construct a molecular database allows for a more precise understanding of the structural changes in the lubricating oil. The molecular database is used to construct semantically related graph structures, mainly containing chemical substance nodes and molecular chain nodes. These graph structures are used to identify the molecular chains of the lubricating oil, thereby determining whether the lubricating oil is in a state of oxidative deterioration (i.e., the molecular chains in the lubricating oil may break or cross-link due to oxidation, leading to an increase or decrease in viscosity, resulting in poor lubrication performance and increased friction and wear).

[0031] S2: Extract key structural parameters from the molecular chain of the lubricating oil and simulate the oxidation process to obtain simulated data records; analyze the changing trend of the oxidation process from the simulated data records; analyze the boundary intervals based on the changing trend of the oxidation process to obtain a simulated data distribution dataset; analyze the simulated data distribution dataset to obtain the microstructural characteristics of the lubricating oil.

[0032] The simulated data record includes the carbonyl absorption peak intensity, hydroxyl absorption peak width, dynamic viscosity value, and molecular weight distribution width.

[0033] It should be noted that the key structural parameters include information such as bond lengths and bond angles, as well as intermolecular forces (i.e., the bond length in the key structural parameters refers to the distance between each pair of covalent bonds; the bond angle refers to the angle formed by three adjacent atoms; and the intermolecular forces refer to the interaction data such as van der Waals forces and Coulomb forces between atomic groups). By simulating the oxidation process of the lubricating oil molecular chain, data on the changes of the lubricating oil at different oxidation stages can be obtained. The simulation data records include key chemical and physical characteristics, such as the carbonyl absorption peak intensity, hydroxyl absorption peak width, dynamic viscosity, and molecular weight distribution width, which can reflect the changes in the molecular structure of the lubricating oil during the oxidation process and provide a quantitative description of the degree of oxidation.

[0034] The above process, by analyzing the changing trends of the oxidation process, can obtain the characteristic changes of lubricating oil during the oxidation process, such as the rate of change of molecular structure and the intensity of oxidation; analyzing the boundary intervals and simulated data distribution datasets can provide data support for establishing standardized early warning systems, provide quantitative standards for early warning systems of lubricating oil oxidation status, and determine whether lubricating oil has entered a dangerous oxidation stage based on data distribution.

[0035] The carbonyl absorption peak intensity in the simulation data record refers to the infrared spectral characteristic signal reflecting the formation of carbonyl groups due to oxidation; the hydroxyl absorption peak width refers to the infrared spectral characteristics indicating the hydroxyl environment and its changes; the dynamic viscosity value refers to the rheological properties of the fluid in the simulation system; the molecular weight distribution width refers to the change in molecular weight distribution caused by molecular chain breakage or polymerization induced by oxidation; the simulation data record contains data on chemical properties (i.e., carbonyl absorption peak intensity and hydroxyl absorption peak width) and data on physical properties (i.e., dynamic viscosity value and molecular weight distribution width).

[0036] S3: Collect historical data on the microstructure characteristics of the lubricating oil; compare and verify the microstructure characteristics of the lubricating oil obtained from the historical data with the microstructure characteristics of the lubricating oil simulated by the oxidation process in S2, and obtain the microstructure verification results of the lubricating oil; issue an early warning based on the microstructure verification results of the lubricating oil (i.e., the microstructure characteristics of the lubricating oil obtained from the historical data are the characteristics of the historically collected comparison samples, and the microstructure results of the lubricating oil collected in S1 are verified using these sample characteristics).

[0037] It should be noted that by collecting historical data on the microstructural characteristics of lubricating oil (such as infrared spectral data, which includes both normal and abnormal microstructures of the lubricating oil in the past; normal microstructures involve stable molecular chains, providing a uniform distribution of the oil film between moving parts and ensuring good lubrication; while abnormal microstructures involve broken or cross-linked molecular chains, leading to drastic changes in oil viscosity and causing the lubricating oil to lose its original functions (such as anti-oxidation, anti-corrosion, and anti-wear)), and comparing this data with simulated microstructural characteristics of the lubricating oil, the current oxidation state of the lubricating oil can be accurately determined. Based on the comparison results, the oxidation level of the lubricating oil can be identified, and a warning can be issued according to a preset oxidation standard.

[0038] The collection and comparison of the above-mentioned actual structural features can help accurately diagnose the actual oxidation status of lubricating oil, ensuring the accuracy and real-time performance of the system. The significance of this step lies in combining simulated data with actual data, which increases the reliability and accuracy of the early warning system. Through early warning, the continued use of excessively oxidized lubricating oil can be avoided, and measures can be taken in advance.

[0039] Specifically, such as Figure 3 As shown, in step S2, key structural parameters of the lubricating oil molecular chain are extracted to simulate the oxidation process, resulting in simulated data records; the changing trend of the oxidation process is analyzed from the simulated data records; based on the changing trend of the oxidation process, boundary intervals are analyzed to obtain a simulated data distribution dataset; the microstructural characteristics of the lubricating oil are obtained by analyzing the simulated data distribution dataset. The specific operation steps are as follows:

[0040] Step S21 mainly describes the extraction of key structural parameters from the lubricating oil molecular chain to generate a basic dataset; based on the basic dataset, pre-setting oxidation condition parameters are performed, and oxidation reaction simulations are conducted at different time points using these parameters; evolution parameters are extracted based on the oxidation process of the oxidation reaction simulation to obtain simulation data records. The specific steps are as follows:

[0041] S21: Extract key structural parameters from the lubricating oil molecular chain;

[0042] The key structural parameters are stored in the form of structured data to generate a basic dataset;

[0043] The key structural parameters include: bond lengths and bond angles, as well as intermolecular forces;

[0044] The basic dataset is input into a pre-built oxidation environment simulation framework, and oxidation operating parameters are set for the oxidation environment simulation framework based on the input basic dataset.

[0045] Oxidation operating parameters include oxygen concentration and temperature conditions (i.e., oxidation operating parameters provide the conditions for the oxidation of lubricating oil).

[0046] Based on oxidation condition parameters, molecular dynamics algorithms are used to simulate oxidation reactions of key structural parameters within the oxidation environment simulation framework.

[0047] The oxidation process of the key structural parameters was simulated at different time points.

[0048] Evolution parameters (i.e., evolution parameters refer to the oxidation process of lubricating oil as a function of time in the chemical structure information (i.e., the chemical structure information has been explained in step S1 and will not be repeated here), such as the shift of the position of the absorption peak, chain breakage, acceleration or deceleration of the reaction rate, etc.) are extracted from the oxidation process of the oxidation reaction simulation at different time points, and the simulation data is recorded.

[0049] It should be noted that the oxygen concentration in the oxidation condition parameters refers to the concentration of oxygen molecules in contact with the lubricating oil during the reaction process; the temperature condition refers to setting the simulation temperature to simulate the effect of heat accelerating the oxidation reaction in actual working conditions.

[0050] By extracting key structural parameters of the lubricating oil molecular chain (such as bond length, bond angle, intermolecular forces, etc.), the molecular-level structure of the lubricating oil can be accurately described. These structured data are stored as a basic dataset. The constructed basic dataset is the basis for subsequent simulation and analysis, providing the necessary initial conditions for simulating the entire oxidation process, helping to accurately establish the initial state of the lubricating oil molecules, and providing the necessary structured data support for simulating the changes of lubricating oil under oxidation conditions.

[0051] Steps S22-S24 mainly describe how to construct a time-series change graph using the simulated data records; how to analyze the chemical change trends of the carbonyl absorption peak intensity and the hydroxyl absorption peak width during the oxidation process using the time-series change graph; how to locate the inflection point of the hydroxyl absorption peak using the chemical change trend; and how to determine the peak boundary range of the hydroxyl absorption peak using the peak value of the inflection point of the hydroxyl absorption peak.

[0052] Calculate the average molecular weight, variance, and standard deviation of the molecular weight distribution width; calculate the rate of change of the molecular weight distribution width using the average molecular weight, variance, and standard deviation of the molecular weight distribution width; calculate the rate of change of the dynamic viscosity value using different time points; determine the physical change range using the rate of change of the molecular weight distribution width and the rate of change of the dynamic viscosity value.

[0053] By integrating the peak boundary interval with the physical change interval, the simulated data distribution dataset is obtained. The specific steps are as follows:

[0054] S22: Construct a time-series variation graph for each data point in the simulation data record (i.e., carbonyl absorption peak intensity, hydroxyl absorption peak width, dynamic viscosity value, and molecular weight distribution width);

[0055] The chemical change trend during the oxidation process is analyzed by examining the carbonyl absorption peak intensity and hydroxyl absorption peak width in the time series diagram. (This is because the time series diagram is constructed from simulated data of the oxidation process at different time points. The time series diagram can directly reflect the curves of carbonyl at different time points and can intuitively reflect the change trend of hydroxyl. Therefore, the chemical change trend reflects both carbonyl and hydroxyl, but more importantly, it reflects the change trend of hydroxyl.) The rising and falling trends and periodic fluctuations of the chemical change trend are analyzed to locate the turning point of the hydroxyl absorption peak.

[0056] Find the maximum and minimum inflection point peak values ​​of the hydroxyl absorption peak in the chemical change trend.

[0057] The peak value is calculated by finding the maximum and minimum inflection point peak values ​​of the hydroxyl absorption peak using a peak detection algorithm.

[0058] Record the corresponding time point when the specific peak occurs;

[0059] By using the corresponding time points of the specific peak values, we can find the turning points (i.e., the peak values ​​of the turning points) of the hydroxyl absorption peaks that tend to zero and begin to rise and fall, and the starting and ending points of the transformation into positive and negative values.

[0060] The peak boundary range (which can also be used as the chemical change range) is formed based on the start and end points of the hydroxyl absorption peak.

[0061] It should be noted that analyzing the changes in the intensity of the carbonyl absorption peak and the width of the hydroxyl absorption peak through time-series graphs can intuitively reflect the chemical reactions of lubricating oil during the oxidation process. Locating the inflection point and peak value changes of the hydroxyl absorption peak provides a basis for judging the key transition stages in the oxidation process. The reason for locating the inflection point of the hydroxyl absorption peak is that, during the oxidation process, the hydroxyl group (-OH) is usually a key intermediate product of the oxidation reaction, and its formation and changes are directly related to the progress of the oxidation reaction. Although the carbonyl group (C=O) is also part of the oxidation product, its intensity changes are more likely to be affected by other factors, and changes in the carbonyl group usually occur in the later stages of oxidation or in a more stable phase. Therefore, changes in the hydroxyl group can reflect important transitions in the oxidation process earlier. It is necessary to first identify the inflection point of the hydroxyl absorption peak to identify the trend of change, such as... Figure 4 As shown;

[0062] By scanning data forward from a specific peak, we can find the turning point where the signal gradually strengthens from a stable or low level. We can also use the change in the first derivative to find the starting point where the signal changes from near zero (or negative) to positive and begins to rise rapidly. We can set a relative percentage of the peak (e.g., an increase of 10%-20%) as the basis for judging the start of the rising interval. Simultaneously, using the specific peak as the center, we can scan data backward to find the turning point where the signal begins to decline from the peak and changes from a significant decline to a stable or low level. Similarly, we can use the point where the change in the first derivative, after reaching a certain negative slope, approaches zero again as the end point of the falling interval. A similar percentage standard as the rising interval can be used to ensure that the determined falling boundary truly reflects the signal attenuation process. The segment between the rising and falling boundaries is then defined as a complete "peak interval." The data changes within this interval should have a significant absorption enhancement (rising part) followed by absorption decay (falling part), thus mapping this process to specific stages of a chemical reaction in the oxidation process.

[0063] For each peak and its associated interval, a structured record table is established, including the peak occurrence time and value, the start time and signal change trend of the rising interval, the end time and decay characteristics of the falling interval, the calculated rise time, fall time, duration and rate of change of the entire interval, and other statistical data. These data are then mapped to the oxidation process stages, and different intervals are mapped to the corresponding stages in the oxidation process (e.g., initial oxidation (which may be characterized by slow absorption enhancement and a long period of stability), rapid oxidation (which may be characterized by sharp absorption enhancement and rapid decline), and advanced oxidation (which may have complex characteristics such as multiple consecutive peaks superimposed)).

[0064] S23: Calculate the average molecular weight, variance, and standard deviation of the molecular weight distribution width;

[0065] The changes in the molecular weight distribution width over time are analyzed based on the oxidation process simulation at different time points. Anomalies in the molecular weight distribution width are recorded by combining the average molecular weight, variance, and standard deviation of the molecular weight distribution width (i.e., analyzing the changes in the molecular weight distribution width over time, that is, judging the anomalies in the molecular weight distribution width by the increase or decrease of the average molecular weight, variance, and standard deviation over time, thereby obtaining the dynamic trend of the molecular weight distribution width).

[0066] The rate of change of the molecular weight distribution width (i.e., the percentage change of the dynamic trend of the molecular weight distribution width within adjacent time periods) is calculated using the dynamic trend.

[0067] The rate of change of the dynamic viscosity value is calculated using different time points (i.e., since the dynamic viscosity value is a digital value, the rate of change can be calculated more directly by the percentage change between adjacent time periods).

[0068] Preset a significant change threshold w; determine whether the rate of change of the molecular weight distribution width and the rate of change of the dynamic viscosity value are greater than the significant change threshold w.

[0069] If so, the time point at which the rate of change of the molecular weight distribution width and the rate of change of the dynamic viscosity value are greater than the significant change threshold w is taken as the inflection point of the molecular weight distribution width and the dynamic viscosity value (i.e., dynamic viscosity).

[0070] The inflection point of the molecular weight distribution width and dynamic viscosity value (i.e., dynamic viscosity) is used to find the physical change range (i.e., the search for the physical change range is the same as the method of gradually finding the starting point and ending point to form the peak boundary range by the inflection point of the hydroxyl absorption peak in step S22 above, and will not be repeated here).

[0071] It should be noted that by analyzing the changes in molecular weight distribution width, we can understand how the molecular weight of lubricating oil molecules changes under different oxidation conditions. This helps to identify anomalies in the process of molecular chain breakage and recombination, accurately assess the dynamic changes in molecular weight distribution, and find possible anomalies, such as excessive oxidation or severe molecular chain breakage. Timely identification of anomalies in molecular weight distribution is of great significance for extending the service life of lubricating oil.

[0072] S24: Determine whether the peak boundary interval (which can also be considered as the chemical change interval) and the physical change interval deviate from the preset reference range threshold, and calculate the actual simulation data distribution characteristics (that is, the actual possible distribution of the simulation data is obtained through the peak boundary interval and the physical change interval. If the interval deviates significantly from the preset reference range threshold, the distribution may be abnormal; if the deviation is small, the lubricating oil may be relatively normal, thus better determining whether there is a problem with the lubricating oil).

[0073] By integrating the simulated data distribution characteristics of each lubricating oil molecular chain, a simulated data distribution dataset is obtained.

[0074] It should be noted that by calculating the deviation of the actual simulated data distribution and comparing it with the preset reference range, it can be determined whether the lubricating oil is within the normal oxidation range; by comparing it with the reference range, the rationality of the simulation results can be verified, and it can be determined whether there are any abnormalities in the oxidation process of the lubricating oil, and then whether the lubricating oil meets the performance requirements; it can be used to identify abnormal conditions of lubricating oil performance, such as excessive oxidation or severe deterioration of molecular structure, so as to make corresponding adjustments or replacements in actual use.

[0075] S25: Analyze the bond breakage of the lubricating oil molecular chain based on key structural parameters in the simulated data distribution dataset; further analyze the mapping relationship between the carbonyl absorption peak intensity and dynamic viscosity value in the simulated data distribution dataset, and find the time point of bond breakage through the mapping relationship; reconstitute the bonds at the time point of bond breakage; obtain the dynamic change trajectory of the lubricating oil molecular chain based on the bond breakage and reconstituted bonds; perform peak value analysis of the carbonyl absorption peak intensity and hydroxyl absorption peak width on the dynamic change trajectory of the lubricating oil molecular chain to obtain a molecular structure evolution trajectory dataset; analyze the molecular structure evolution trajectory dataset to obtain the microstructural characteristics of the lubricating oil;

[0076] It should be noted that, based on the distribution of simulated data during the oxidation process, this study analyzes the microscopic processes such as bond breaking and recombination in lubricating oil molecular chains, further analyzes the evolution trajectory of its molecular structure, identifies the microscopic structural characteristics of lubricating oil, clarifies the microscopic structural changes that occur in lubricating oil molecules during oxidation, such as the breaking and recombination of molecular chains, predicts the long-term stability and performance changes of lubricating oil, and analyzes the dynamic changes of lubricating oil molecular chains. This allows for a more accurate understanding of the oxidation process of lubricating oil and provides theoretical support for the development of more durable lubricating oil products.

[0077] Specifically, such as Figure 5 As shown, in step S25, the simulated data distribution dataset is analyzed for bond breakage in the lubricating oil molecular chain based on key structural parameters; then, the simulated data distribution dataset is analyzed to determine the mapping relationship between the carbonyl absorption peak intensity and the dynamic viscosity value, and the time point of bond breakage is found through this mapping relationship; the bonds are recombined at the time point of bond breakage; the dynamic change trajectory of the lubricating oil molecular chain is obtained based on the bond breakage and recombination; peak value analysis of the carbonyl absorption peak intensity and hydroxyl absorption peak width is performed on the dynamic change trajectory of the lubricating oil molecular chain to obtain a molecular structure evolution trajectory dataset; the microstructural characteristics of the lubricating oil are obtained by analyzing the molecular structure evolution trajectory dataset. The specific operation steps are as follows:

[0078] In steps S251 and S252, the process mainly describes how to find the bond length of the corresponding lubricating oil molecular chain by traversing each time point of the simulated data distribution dataset based on key structural parameters, and calculate the bond breakage of the lubricating oil molecular chain based on the bond length of the lubricating oil molecular chain.

[0079] The inflection point between the carbonyl absorption peak intensity and the dynamic viscosity value is found at corresponding time points; a change mapping relationship is established between the carbonyl absorption peak intensity and the dynamic viscosity value based on the corresponding time points; the change mapping relationship is judged by a preset change time threshold to identify the bond breakage of the lubricating oil molecular chain;

[0080] Based on the mapping relationship between bond breakage and change in bond length values, bond breaks with overlapping coordinates are filtered to determine the time point of bond breakage at overlapping coordinates. The specific steps are as follows:

[0081] S251: Extract key structural parameters from the simulated data distribution dataset;

[0082] The key structural parameters in the simulated data distribution dataset are traversed at each time point to find the corresponding bond length values ​​of the lubricating oil molecular chains, and sorted to form a bond length sequence.

[0083] Calculate the average value and standard deviation of the bond length in the central interval of the statistical simulation data distribution dataset of the bond length sequence (i.e., the interval is the peak boundary interval and the physical change interval. Due to the calculation of the reference range threshold in step S24, there may be anomalies in the interval. Therefore, the corresponding interval is searched for the bond length sequence formed at different time points to verify the anomalies. That is, it is to determine whether there is chain breakage in the bond length sequence at the time point corresponding to the interval. The average value of the bond length at the time corresponding to the anomaly in the interval is used as the benchmark (i.e., the benchmark is the following benchmark bond length threshold)).

[0084] The average bond length and standard deviation are used to set a baseline bond length threshold (i.e., the average bond length ± a certain percentage (e.g., ±10%) can generally be used as a reference range for the length threshold).

[0085] Determine whether the bond length value at each different time point in the bond length sequence is greater than the baseline bond length threshold.

[0086] If so, it is determined that the bond length of the bond length sequence has broken, and the corresponding lubricating oil molecule chain has broken.

[0087] It should be noted that by extracting key structural parameters from the simulation data, locating the bond lengths of the lubricating oil molecular chains, and forming a bond length sequence, the changing trend of the molecular chains can be intuitively reflected. Calculating the average value and standard deviation of the bond lengths helps to confirm whether the molecular chains are within the normal range and can identify chain breaks caused by oxidation reactions.

[0088] The above steps analyze bond lengths at different time points to identify potential chain breakage locations, which helps in the early detection of lubricating oil oxidation and degradation. Furthermore, bond length is an important parameter of lubricating oil molecular structure, and detecting its changes can provide key clues in the oxidation process, enabling early detection of lubricating oil degradation.

[0089] S252: Extract the carbonyl absorption peak intensity from the simulated data distribution dataset, and form a continuous time-series curve of the carbonyl absorption peak intensity at different time points;

[0090] The continuous time-series curves are smoothed to find the inflection point of the carbonyl absorption peak intensity;

[0091] Simultaneously, the turning points of dynamic viscosity values ​​in the simulated data distribution dataset corresponding to the time points of continuous time-series curves are extracted;

[0092] Establish a mapping relationship between the inflection point of the carbonyl absorption peak intensity and the inflection point of the dynamic viscosity value (i.e., since the extraction time points are the same, observe the influence of the change in carbonyl absorption peak intensity on the change in dynamic viscosity at the same time point, so the mapping relationship reflects the magnitude of the change).

[0093] A preset change time threshold is established; it is then determined whether the change mapping relationship is greater than the change time threshold.

[0094] If so, it is determined that the molecular chain of the lubricating oil has broken.

[0095] The coordinates of the lubricating oil molecular chain are determined by comparing the bond breakage of the change mapping relationship with the bond length breakage in step S252, and bond breakage with overlapping coordinates is screened; and the number of bond breakages is recorded.

[0096] The time series of the continuous time series curves are traversed to calculate the variation range of carbonyl absorption peak intensity and dynamic viscosity value at each different time point (i.e., since the dynamic viscosity value extraction corresponds to the time point of carbonyl absorption peak intensity, the same time series can be used to calculate the variation range).

[0097] Determine whether the changes in the carbonyl absorption peak intensity and dynamic viscosity values ​​calculated at each different time point are all greater than the time threshold.

[0098] If so, then the time point is determined to be the time point when the bond of the lubricating oil molecular chain breaks;

[0099] It should be noted that extracting time-series data of carbonyl absorption peak intensity and dynamic viscosity values, and identifying inflection points, can reveal changes in lubricating oil at different stages. For example, carbonyl absorption peaks are usually related to the formation of oxidation products, and dynamic viscosity reflects the viscosity characteristics of lubricating oil. Furthermore, the mapping relationship of inflection points can help determine whether chain breakage has occurred in the molecular chains of the lubricating oil. Figure 6 As shown;

[0100] The above steps establish a mapping relationship between the carbonyl absorption peak intensity and the dynamic viscosity inflection point, identify key inflection points in the oxidation process, provide multi-angle analysis methods for monitoring the oxidation process of lubricating oil, make the determination of chain breakage more accurate, and further enhance the reliability of lubricating oil performance prediction.

[0101] In steps S253 and S254, the main steps are to reorganize the bonds according to the determined time point of bond breakage and the bond length of the broken bonds to obtain reorganized bonds; to determine the time point of the reorganized bonds; and to determine the dynamic change trajectory of the molecular chain of lubricating oil molecular chains based on the time point of bond breakage and the time point of reorganization.

[0102] Based on the time-series variation diagram, determine the peak values ​​of the carbonyl absorption peak intensity and the hydroxyl absorption peak width at each time point, and further calculate the ratio of the peak values. Construct a molecular structure evolution trajectory dataset based on the ratio of the peak values ​​of the carbonyl absorption peak intensity and the hydroxyl absorption peak width at each time point. The specific steps are as follows:

[0103] S253: Based on the time point of bond breakage in the overlapping coordinates, find the corresponding bond breakage in the bond length sequence, calculate the distance between adjacent atoms of the corresponding bond breakage, and obtain the atomic distance;

[0104] Determine whether the atomic distance is greater than the reference bond length threshold;

[0105] If so, the broken bond length is recombined to obtain a recombined bond;

[0106] The recombination time and the number of recombination bonds are recorded for each recombination bond.

[0107] A preset frequent time point threshold is set; it is determined whether the time point of bond breakage and the recombination time of the recombining bond are greater than the frequent time point threshold.

[0108] If so, it is determined that the lubricating oil molecular chain has a frequent pattern of breakage and recombination, and the dynamic change trajectory of the lubricating oil molecular chain is obtained (that is, a frequent time point threshold is set to reflect the duration of bond breakage and the time of bond recombination. When both are greater than the frequent time point threshold, it indicates that the lubricating oil molecular chain breaks and recombines too frequently, thus constructing a dynamic change trajectory of the bond length of breakage and recombination, which reflects the unstable chain trajectory of the lubricating oil molecular chain).

[0109] It should be noted that by calculating the distance between adjacent atoms based on the time point of bond breakage, it is determined whether the reference bond length threshold is exceeded and whether bond recombination is necessary. Furthermore, determining the distance between atoms further confirms the stability of the molecular chain and whether bond recombination occurs. Monitoring the breakage and recombination of molecular chains helps to deeply understand the dynamic changes of the molecular chains of lubricating oil during operation and reveals the process of molecular structure adjustment of lubricating oil.

[0110] S254: Based on the time series variation diagram and the carbonyl absorption peak intensity and hydroxyl absorption peak width corresponding to different time points, find the peak values ​​of the carbonyl absorption peak intensity and hydroxyl absorption peak width (that is, as explained in step S22, the time series variation diagram can reflect the peak values ​​of the carbonyl absorption peak intensity and hydroxyl absorption peak width, which will not be repeated here).

[0111] Calculate the ratio of the peak intensity of the carbonyl absorption peak to the peak width of the hydroxyl absorption peak at each time point;

[0112] The time series ratios are sorted according to the ratio of the sequential peak values ​​at different time points to obtain a list of time series ratios.

[0113] The time series ratio list is normalized, and a molecular weight width threshold is preset.

[0114] Determine whether the ratio of the peak values ​​in the time-series ratio list is greater than the molecular weight width threshold;

[0115] If so, the ratio of the peaks is determined to reflect the change in the molecular weight distribution width. The ratios of the selected peaks are used to construct a molecular structure evolution trajectory dataset (that is, when the ratio of the peaks in the time sequence ratio list is greater than the molecular weight width threshold, it indicates that the carbonyl absorption peak has been enhanced and the hydroxyl width has changed significantly, which is often accompanied by a change in the molecular weight distribution width).

[0116] It should be noted that by analyzing the peak ratio of carbonyl absorption peaks to hydroxyl absorption peaks in the time-series variation graph, we can understand the changes in molecular weight distribution in lubricating oil. Furthermore, by comparing the normalized ratio with the molecular weight width threshold, we can screen out a dataset of molecular structure evolution trajectories (evolutionary trajectory datasets mainly refer to the dynamic recording and description of key parameters such as the chemical structure, physical properties, and reaction rate of lubricating oil over time; their core lies in capturing the entire process of the system's internal state "from initial to evolution"; they not only reflect the trajectory of the entire evolution process presented by the system at different time points or conditions (i.e., usually with time or other evolution variables (such as temperature, load conditions, reaction progress) as the horizontal axis, recording the changes in chemical structural characteristics (such as absorption peak position, peak intensity, molecular chain length, etc.), principal component scores, and macroscopic indicators (such as viscosity, reaction rate) at various times or conditions)); the changes in the carbonyl to hydroxyl absorption peak ratio detected in the above steps reveal changes in the lubricating oil molecular chains, especially changes related to oxidation and molecular aggregation states; analyzing changes in molecular weight distribution can determine whether the lubricating oil has experienced oxidation product accumulation and molecular aggregation phenomena.

[0117] The molecular structure evolution trajectory dataset also reflects the temporal changes of chain breakage and recombination in step S253, allowing observation of the continuous evolution of the molecular chain from an intact state to breakage and then partial recombination. It also reflects the changes in molecular weight distribution in step S254 (i.e., the carbonyl absorption peak is enhanced, while the hydroxyl absorption peak width changes significantly, reflecting changes in the molecular weight distribution width). Furthermore, the changes in molecular weight distribution reflect the intensity and trends of the carbonyl and hydroxyl absorption peaks (primarily reflecting the trend of the hydroxyl absorption peak). The enhancement of the carbonyl absorption peak is often related to the formation and breakage of oxidation products, while changes in the hydroxyl absorption peak width can reflect adjustments in the molecular aggregation state. Therefore, the molecular structure evolution trajectory dataset provides multifaceted data on the lubricating oil molecular chain.

[0118] S255: Analyze the driving force data of bond breaking and recombination in the molecular structure evolution trajectory dataset; extract reaction data feature sets from the molecular structure evolution trajectory dataset based on bond breaking and recombination events; extract the intensity change trend and dynamic viscosity change trend from the reaction data feature sets; extract the microstructure features of the lubricating oil based on the intensity change trend and dynamic viscosity change trend.

[0119] It should be noted that analyzing the molecular structure evolution trajectory dataset, extracting the reaction data features of driving bond breaking and recombination, and further analyzing the trends of strength change and dynamic viscosity change; analyzing the driving force data and viscosity change trends can further understand the behavior of lubricating oil during operation, especially how to predict the performance of lubricating oil through changes in microstructure; the above steps to extract the microstructural features of lubricating oil can more accurately reflect the changes in the physicochemical properties of lubricating oil.

[0120] Specifically, such as Figure 7 As shown, in step S255, the driving force data of bond breaking and recombination are analyzed in the molecular structure evolution trajectory dataset; reaction data feature sets are extracted from the molecular structure evolution trajectory dataset based on the events of bond breaking and recombination; intensity change trends and dynamic viscosity change trends are extracted from the reaction data feature sets respectively; and the microstructure features of lubricating oil are extracted based on the intensity change trends and dynamic viscosity change trends. The specific operation steps are as follows:

[0121] S2551: Preprocess the molecular structure evolution trajectory dataset and construct an initial data matrix (that is, organize the processed data in matrix form, with each row being a record of a time point and each column corresponding to a feature (such as carbonyl absorption peak intensity, dynamic viscosity, etc. The molecular structure evolution trajectory dataset is derived step by step through steps S2 and S25, and contains data on carbonyl absorption peak intensity, hydroxyl absorption peak width, dynamic viscosity values, and molecular weight distribution width)).

[0122] The initial data matrix is ​​subjected to dimensionality reduction using principal component analysis (PCA) algorithm, and the eigenvalues ​​and corresponding eigenvectors of the principal components of each matrix unit are calculated.

[0123] The eigenvalues ​​of the principal components of each matrix unit are sorted, and the principal component with the largest eigenvalue is selected for analysis loading (i.e., the eigenvector corresponding to the largest eigenvalue is used for analysis loading). This yields the key influencing factors of the lubricating oil oxidation process (i.e., the principal component is mainly affected by the carbonyl absorption peak intensity and dynamic viscosity, and is considered a key indicator reflecting the main molecular structure changes during the oxidation process; if the eigenvalue is high and the corresponding loading is significant on the carbonyl absorption peak and dynamic viscosity, the principal component can be understood as a composite indicator of the overall oxidation reaction intensity or the level of structural change).

[0124] It should be noted that the original molecular structure evolution data (such as carbonyl absorption peak intensity, hydroxyl absorption peak width, dynamic viscosity, etc.) are organized into a matrix to construct the initial data matrix, ensuring the clarity and standardization of the data. The data matrix is ​​the core of dimensionality reduction and analysis. It transforms time series data into a format that machine learning algorithms can process, helping to better understand the relationship between changes in molecular structure and oxidation processes.

[0125] S2552: The oxidation reaction rate is calculated using the rate of change of the molecular weight distribution width and the rate of change of the dynamic viscosity value (that is, the rate of oxidation process, which is obtained in step S23 by the rate of change of the molecular weight distribution width and the rate of change of the dynamic viscosity value, and will not be repeated here; furthermore, the oxidation reaction of dynamic viscosity will cause molecular chain breakage or cross-linking, which will cause the viscosity of the lubricating oil to change; the rate of change of dynamic viscosity can also be used as an indicator of the progress of the oxidation reaction).

[0126] Based on the key influencing factors, the molecular structure change data (i.e., the frequency of molecular chain breakage and recombination events (e.g., the number of breakage and recombination events occurring in different time periods) and oxidation reaction rate of the molecular structure evolution trajectory dataset are correlated to obtain the frequency distribution data of bond breakage and recombination.

[0127] The frequency distribution data of bond breaking and recombination and the oxidation reaction rate were analyzed using a pre-designed chemical reaction driving force model to obtain driving force data;

[0128] It should be noted that the oxidation reaction rate is calculated by the change rate of molecular weight distribution width and dynamic viscosity. Dynamic viscosity and molecular weight distribution width are related to the breaking and cross-linking of molecular chains, reflecting the intensity and progress of the oxidation reaction. Understanding the oxidation reaction rate helps to identify the acceleration stage of the reaction, detect abnormalities in lubricating oil oxidation in a timely manner, and provide early warning for equipment maintenance.

[0129] Driving force data represents a sensitive indicator in lubricating oil oxidation state early warning. By capturing key indicators such as molecular structure (e.g., carbonyl groups, molecular chain breakage), local reaction rate, and viscosity changes, driving force data reflects the microscopic reactions and structural changes occurring inside the lubricating oil. It quantitatively describes the dynamic evolution of the internal structure and performance of the lubricating oil over time or in the environment. When the actual measured value of the driving force parameter exceeds the system's set safety threshold, it indicates that the oxidation reaction may be accelerating, thus triggering an early warning signal. Furthermore, by continuously monitoring the driving force parameter, the impact of molecular structure changes (i.e., breakage and recombination) on the overall oxidation reaction can be grasped in real time, which helps to detect abnormalities in a timely manner.

[0130] Step S2553 mainly clarifies the process of calculating the difference between peak values ​​of adjacent bond lengths in the driving force data using the frequency distribution data of bond breakage and recombination; determining the peak position and peak height of the peaks that protrude between adjacent bond lengths; calculating the local reaction rate using the peak position and peak height of the peaks that protrude between adjacent bond lengths; and forming a reaction data feature set using the local reaction rate and the peak position and peak height of the peaks that protrude between adjacent bond lengths. The specific steps are as follows:

[0131] S2553: Calculate the difference between the peak values ​​of adjacent bond lengths in the driving force data using the frequency distribution data of the broken and recombined bonds;

[0132] A preset salience threshold is set; it is determined whether the difference between the peak values ​​of adjacent bond lengths is greater than the preset salience threshold.

[0133] If so, it is determined that the peak protrusion between the adjacent bond lengths is significant, the event of bond breakage and bond recombination is increased, and the peak position and peak height of the peak protrusion between the adjacent bond lengths are determined.

[0134] The slope of the peak values ​​between adjacent bond lengths at the sudden increase in bond breakage and recombination events is calculated as the local reaction rate.

[0135] The peak position and peak height of the peaks that are prominent between adjacent bond lengths are combined with the local reaction rate to form a reaction data feature set;

[0136] It should be noted that the execution of step S2553 above reveals the details of molecular structure changes by analyzing the frequency distribution of bond breaking and recombination, providing a deeper understanding of the lubricating oil oxidation process. Bond breaking and recombination are key reaction events in the lubricating oil oxidation process, and understanding their frequency distribution helps to discover the changing trends of the oxidation reaction. A sudden increase in the frequency of breaking and recombination may indicate an accelerated oxidation reaction or a critical state. By monitoring these changes, early warnings can be given, thereby achieving intelligent management.

[0137] During the execution of step S2553 above, the local reaction rate refers to the speed of signal change in a certain local area or time period during the reaction process; it is obtained by differentiating or calculating the slope of data near the detected peak; for example, near a sudden increase event, the rate is calculated using several sampling points before and after, reflecting the intensity of the reaction in that local area; the reaction data feature set, by integrating multiple key features, can quantitatively describe the entire reaction process from multiple perspectives; step S2553 above, by real-time monitoring of various indicators in the feature set, can promptly detect abnormal changes, providing a quantitative basis for early warning of the oxidation state of lubricating oil;

[0138] S2554: Use the carbonyl absorption peak intensity to filter the reaction data feature set for regions with the same wavenumber;

[0139] The intensity difference of the carbonyl absorption peak at each moment is calculated in time sequence for this wavenumber region, and the intensity difference of the carbonyl absorption peak at each moment is aggregated to obtain the intensity change trend;

[0140] The rate of change of dynamic viscosity values ​​at each moment corresponding to the wavenumber region is collected to obtain the trend of dynamic viscosity change (that is, the rate of change of dynamic viscosity values ​​itself reflects the change of dynamic viscosity at each moment; by collecting the rate of change of dynamic viscosity values ​​at each time length corresponding to the wavenumber region, the trend of dynamic viscosity change at each time length corresponding to the wavenumber region is obtained).

[0141] It should be noted that combining the frequency distribution data of bond breaking and recombination with the oxidation reaction rate yields a new set of reaction data features. The multi-dimensional feature set can quantitatively describe the oxidation process from multiple aspects, which helps to comprehensively assess the complexity of the reaction process. By integrating multiple key features, the oxidation changes of lubricating oil can be captured more comprehensively, thereby enhancing the accuracy and reliability of early warning.

[0142] S2555: Integrate the intensity change trend with the dynamic viscosity change trend to obtain a multivariate data matrix (i.e., each row represents a time point or sample, and each column represents a characteristic index (e.g., absorption peak intensity, viscosity matching degree, etc.)).

[0143] The multivariate data matrix is ​​then subjected to dimensionality reduction again using principal component analysis algorithm to calculate eigenvalues ​​and corresponding eigenvectors.

[0144] The eigenvalues ​​are sorted, and the eigenvectors corresponding to the first k eigenvalues ​​are selected for projection to obtain the projection matrix.

[0145] The absolute value of the eigenvector corresponding to each principal component is analyzed in the projection matrix;

[0146] The principal component score is obtained by assigning weights based on the absolute values ​​of the feature vectors. The calculation formula is as follows:

[0147] ;

[0148] In the formula, The number of principal components selected (i.e., selecting k principal components and summing them up);

[0149] For the first Eigenvalues ​​of each principal component;

[0150] The constant used for normalization (i.e., usually a suitable constant is chosen (e.g., the largest eigenvalue or the sum of all eigenvalues) for normalization) Normalization is performed to ensure that the input values ​​of the exponential function are within a reasonable range, thereby obtaining appropriate weights; the dimensions of the eigenvalues ​​of different principal components are adjusted to ensure... Numerically, it does not decay too quickly or become ineffective, thus balancing the importance of each principal component.

[0151] This is represented as an exponential function used to map normalized eigenvalues ​​to a decaying weight;

[0152] m is the total number of original variables (i.e., the original variables are data such as the intensity or wavelength position of absorption peaks, the frequency of bond breaking and recombination, etc., obtained step by step in the above steps, and these data are used as inputs to calculate the principal component score).

[0153] For the first The first principal component Loadings of the original variables (i.e., representing the original variables) (coefficients or contributions in the direction of the i-th principal component).

[0154] For the first The normalized or standardized values ​​of the original variables;

[0155] It should be noted that a multivariate data matrix is ​​generated by combining the trend of carbonyl absorption peak intensity change with the trend of dynamic viscosity change; the changes in intensity and viscosity are key characteristics of the lubricating oil oxidation process, and integrating them can more comprehensively reflect the evolution of the oxidation process;

[0156] Multivariate data matrices can comprehensively reveal the relationships between different features and provide rich input data for dimensionality reduction analysis, helping to discover key factors;

[0157] S2556: Select the feature vector with the highest principal component score from all feature vectors and use it as the microstructure feature.

[0158] It should be noted that the dimensionality of the multivariate data matrix is ​​reduced by principal component analysis, and the most important eigenvectors are extracted as representative indicators of the microstructure of lubricating oil.

[0159] Dimensionality reduction can reduce redundant data and extract the most representative information, which helps to improve the efficiency and accuracy of data analysis. The above steps for principal component analysis can extract key structural features from complex data, thereby gaining a deeper understanding of the microstructural changes of lubricating oil during oxidation.

[0160] Example 2

[0161] like Figure 8 As shown, the present invention also provides a method and system for early warning and judgment of lubricating oil oxidation state based on multimodal data feedback, including: a data acquisition module 10; an identification module 20; and an early warning module 30.

[0162] The acquisition module 10 is used to acquire chemical structure information of the lubricating oil, preprocess the chemical structure information to obtain the molecular chain of the lubricating oil;

[0163] The identification module 20 is used to extract key structural parameters from the molecular chain of the lubricating oil to simulate the oxidation process and obtain simulated data records; analyze the changing trend of the oxidation process from the simulated data records; analyze the boundary intervals based on the changing trend of the oxidation process to obtain a simulated data distribution dataset; and analyze the simulated data distribution dataset to obtain the microstructural characteristics of the lubricating oil.

[0164] The early warning module 30 is used to collect historical data on the microstructure characteristics of the lubricating oil; compare and verify the microstructure characteristics of the lubricating oil based on the historical data with the microstructure characteristics of the lubricating oil simulated by the oxidation process to obtain the microstructure verification results of the lubricating oil; and issue an early warning based on the microstructure verification results of the lubricating oil.

[0165] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; those skilled in the art can modify the technical solutions described in the foregoing embodiments, or make equivalent substitutions for some or all of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for early warning and judgment of lubricating oil oxidation state based on multimodal data feedback, characterized in that, The following steps are included: Chemical structure information of the lubricating oil is collected, and the chemical structure information is preprocessed to obtain the molecular chain of the lubricating oil. Key structural parameters of the lubricating oil molecular chain were extracted and the oxidation process was simulated to obtain simulation data records; The simulation data records were analyzed to determine the changing trends of the oxidation process. Based on the changing trend of the oxidation process, the boundary interval is analyzed to obtain a simulated data distribution dataset; the simulated data distribution dataset is analyzed to obtain the microstructural characteristics of the lubricating oil. The simulated data record includes the carbonyl absorption peak intensity, hydroxyl absorption peak width, dynamic viscosity value, and molecular weight distribution width. Collect historical data on the microstructural characteristics of lubricating oil; The microstructural characteristics of the lubricating oil were compared and verified using historical data collected in the past and the microstructural characteristics of the lubricating oil simulated by the oxidation process, and the microstructural verification results of the lubricating oil were obtained. Early warnings are issued based on the microscopic verification results of the lubricating oil.

2. The method for early warning and judgment of lubricating oil oxidation state based on multimodal data feedback according to claim 1, characterized in that, Key structural parameters of the lubricating oil molecular chain were extracted to simulate the oxidation process, and simulation data was recorded. The specific operation steps are as follows: Key structural parameters are extracted from the lubricating oil molecular chain to generate a basic dataset; based on the basic dataset, pre-set oxidation condition parameters are used to simulate oxidation reactions at different time points. Evolution parameters are extracted from the oxidation process simulated by the oxidation reaction to obtain simulation data records.

3. The method for early warning and judgment of lubricating oil oxidation state based on multimodal data feedback according to claim 2, characterized in that, The simulation data records are analyzed to determine the changing trend of the oxidation process; based on the changing trend of the oxidation process, the boundary intervals are analyzed to obtain the simulation data distribution dataset. The specific operation steps are as follows: A time-series variation diagram is constructed using the simulated data; the chemical change trends of the carbonyl absorption peak intensity and the hydroxyl absorption peak width during the oxidation process are analyzed using the time-series variation diagram; the inflection point of the hydroxyl absorption peak is located using the chemical change trends; The peak boundary range of the hydroxyl absorption peak is determined by the peak value at the inflection point of the hydroxyl absorption peak. Calculate the average molecular weight, variance, and standard deviation of the molecular weight distribution width; calculate the rate of change of the molecular weight distribution width using the average molecular weight, variance, and standard deviation of the molecular weight distribution width; calculate the rate of change of the dynamic viscosity value using different time points; determine the physical change range using the rate of change of the molecular weight distribution width and the rate of change of the dynamic viscosity value. By integrating the peak boundary interval with the physical change interval, a simulated data distribution dataset is obtained.

4. The method for early warning and judgment of lubricating oil oxidation state based on multimodal data feedback according to claim 3, characterized in that, The microstructural characteristics of the lubricating oil are obtained by analyzing the simulated data distribution dataset. The specific operation steps are as follows: The simulated data distribution dataset is analyzed for bond breakage in the lubricating oil molecular chain based on key structural parameters. The dataset is then analyzed to determine the mapping relationship between carbonyl absorption peak intensity and dynamic viscosity values, identifying the time points of bond breakage. Bond recombination is then performed at these breakage time points. The dynamic evolution trajectory of the lubricating oil molecular chain is obtained based on the bond breakage and recombination. Peak value analysis of the carbonyl absorption peak intensity and hydroxyl absorption peak width is then performed on the dynamic evolution trajectory of the lubricating oil molecular chain to obtain a molecular structure evolution trajectory dataset. Finally, the microstructural characteristics of the lubricating oil are analyzed using this molecular structure evolution trajectory dataset.

5. The method for early warning and judgment of lubricating oil oxidation state based on multimodal data feedback according to claim 4, characterized in that, The simulated data distribution dataset is analyzed for bond breakage in lubricating oil molecular chains based on key structural parameters. Then, the simulated data distribution dataset is analyzed to determine the mapping relationship between the carbonyl absorption peak intensity and dynamic viscosity values. The time points of bond breakage are identified through this mapping relationship. The specific steps are as follows: Based on the key structural parameters, the simulation data distribution dataset is traversed to find the corresponding bond length value of the lubricating oil molecular chain at each time point. Based on the bond length value of the lubricating oil molecular chain, the bond breakage of the lubricating oil molecular chain is calculated. Find the turning point between the carbonyl absorption peak intensity and the dynamic viscosity value at corresponding time points; establish a change mapping relationship between the carbonyl absorption peak intensity and the dynamic viscosity value based on the corresponding time points; The change mapping relationship is judged by a preset change time threshold, and the bond breakage of the lubricating oil molecular chain is identified. Based on the mapping relationship between bond breakage and change in bond length value, bond breakage with overlapping coordinates is filtered to determine the time point of bond breakage with overlapping coordinates.

6. The method for early warning and judgment of lubricating oil oxidation state based on multimodal data feedback according to claim 5, characterized in that, The bonds are recombinated at the time points of bond breakage; the dynamic trajectory of the lubricating oil molecular chain is obtained based on the bond breakage and recombination; peak value analysis of the carbonyl absorption peak intensity and hydroxyl absorption peak width is performed on the dynamic trajectory of the lubricating oil molecular chain to obtain a molecular structure evolution trajectory dataset. The specific operation steps are as follows: Recombination is performed based on the determined time point of bond breakage and the bond length of the broken bond to obtain a recombined bond; the time point of the recombined bond is determined; the dynamic change trajectory of the molecular chain of lubricating oil molecule chain is determined based on the time point of bond breakage and the time point of recombination. Based on the time-series variation diagram, determine the peak values ​​of the carbonyl absorption peak intensity and the hydroxyl absorption peak width at each time point, and further calculate the ratio of the peak values; construct a molecular structure evolution trajectory dataset based on the ratio of the peak values ​​of the carbonyl absorption peak intensity and the hydroxyl absorption peak width at each time point.

7. The method for early warning and judgment of lubricating oil oxidation state based on multimodal data feedback according to claim 6, characterized in that, The microstructural characteristics of the lubricating oil were obtained by analyzing the molecular structure evolution trajectory dataset. The specific operation steps are as follows: The molecular structure evolution trajectory dataset is analyzed to obtain driving force data of bond breaking and recombination; based on the events of bond breaking and recombination, reaction data feature sets are extracted from the molecular structure evolution trajectory dataset. The intensity change trend and dynamic viscosity change trend are extracted from the reaction data feature set respectively; the microstructure features of the lubricating oil are extracted based on the intensity change trend and dynamic viscosity change trend.

8. The method for early warning and judgment of lubricating oil oxidation state based on multimodal data feedback according to claim 7, characterized in that, The molecular structure evolution trajectory dataset is analyzed to obtain driving force data of bond breaking and recombination; based on the events of bond breaking and recombination, reaction data feature sets are extracted from the molecular structure evolution trajectory dataset. The intensity variation trend and dynamic viscosity variation trend are extracted from the reaction data feature set, respectively; based on the intensity variation trend and dynamic viscosity variation trend, the microstructural features of the lubricating oil are extracted. The specific operation steps are as follows: The molecular structure evolution trajectory dataset is preprocessed, and an initial data matrix is ​​constructed. The initial data matrix is ​​subjected to dimensionality reduction using principal component analysis (PCA) algorithm, and the eigenvalues ​​and corresponding eigenvectors of the principal components of each matrix unit are calculated. The eigenvalues ​​of the principal components of each matrix unit are sorted, and the principal component with the largest eigenvalue is selected for analysis loading to obtain the key influencing factors of the oxidation process of lubricating oil. The oxidation reaction rate is calculated using the rate of change of the molecular weight distribution width and the rate of change of the dynamic viscosity value. Based on the aforementioned key influencing factors, the molecular structure change data and oxidation reaction rate of the molecular structure evolution trajectory dataset are correlated to obtain the frequency distribution data of bond breaking and recombination. The frequency distribution data of bond breaking and recombination and the oxidation reaction rate are analyzed using a pre-designed chemical reaction driving force model to obtain driving force data.

9. The method for early warning and judgment of lubricating oil oxidation state based on multimodal data feedback according to claim 8, characterized in that, Based on bond breaking and recombination events, reaction data feature sets are extracted from the molecular structure evolution trajectory dataset. Intensity change trends and dynamic viscosity change trends are then extracted from the reaction data feature sets. The specific steps are as follows: The difference between peak values ​​of adjacent bond lengths in the driving force data is calculated using the frequency distribution data of bond breakage and recombination; the peak position and peak height of the peak protrusion between the adjacent bond lengths are determined. The local reaction rate is calculated by the peak position and peak height of the peak protrusion between adjacent bond lengths; a reaction data feature set is formed by the local reaction rate and the peak position and peak height of the peak protrusion between adjacent bond lengths. The reaction data feature set was screened for regions with the same wavenumber by using the carbonyl absorption peak intensity; The intensity difference of the carbonyl absorption peak at each moment is calculated in time sequence for this wavenumber region, and the intensity difference of the carbonyl absorption peak at each moment is aggregated to obtain the intensity change trend; The rate of change of dynamic viscosity values ​​at each time step in the wavenumber region is aggregated to obtain the trend of dynamic viscosity change.

10. The method for early warning and judgment of lubricating oil oxidation state based on multimodal data feedback according to claim 9, characterized in that, The microstructural features of the lubricating oil are extracted based on the trends in intensity and dynamic viscosity. The specific steps are as follows: The intensity variation trend is integrated with the dynamic viscosity variation trend to obtain a multivariate data matrix; The multivariate data matrix is ​​then subjected to dimensionality reduction using principal component analysis to calculate eigenvalues ​​and corresponding eigenvectors. The eigenvalues ​​are sorted, and the eigenvectors corresponding to the first k preset eigenvalues ​​are projected to obtain a projection matrix. The absolute value of the eigenvector corresponding to each principal component is analyzed in the projection matrix; The principal component scores are obtained by assigning weights based on the absolute values ​​of the feature vectors. The feature vector with the highest principal component score among all feature vectors is selected as the microstructure feature.

Images