Prediction method of lubricating oil additive
By constructing a training database for lubricant and using mid-infrared spectral data, an intelligent analysis model for lubricant additives was established, and the accuracy of the detection of additive content in lubricant was solved, efficient and accurate detection results were achieved, and economic losses were reduced.
Patent Information
- Application Number
- CN202411819291.7
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-12-11
- Publication Date
- 2025-04-01
AI Technical Summary
The prior art is difficult to accurately predict the content of additives in lubricating oils, especially in complex industrial environments, resulting in equipment failures and economic losses.
By constructing a lubricant training database, using mid-infrared spectral data and machine learning technology, an intelligent analysis model for lubricant additives is established to achieve rapid and accurate detection of the types and content of lubricant additives.
It improves the accuracy and efficiency of lubricant additive content detection, reduces manpower and material costs, and realizes quality monitoring and early warning of lubricant oil in mechanical equipment.
Smart Images
Figure CN119741994A_ABST
Abstract
Description
Technical Field
[0001] The invention relates to the technical field of lubricants, and in particular to a prediction method for lubricant additives. Background Art
[0002] Modern industrial equipment is larger, faster and more complex. It is developing towards high speed, integration and intelligence. Its functions are constantly enhanced and its internal structure is becoming more sophisticated. As the operation and maintenance of these complex and precise industrial equipment become more complex and critical, various problems that the equipment may encounter during operation, such as wear of parts, loss of precision, performance degradation, etc., may cause equipment failure or shutdown, leading to economic losses or even serious casualties. By detecting the lubricating oil in the equipment, it is possible to achieve advanced monitoring of the equipment, from passive response to equipment failure to active warning of failure, timely replacement of lubricating oil and adjustment of equipment status to avoid paralysis of production system and property loss. Therefore, it is particularly important to predict the attenuation of the additive content in the lubricating oil.
[0003] Infrared spectroscopy has made great progress due to its advantages such as non-destructive detection, rapid analysis, and easy operation. It has become the preferred tool for researchers in chemical analysis. The introduction of machine learning technology has helped infrared spectroscopy to develop further. The application of infrared spectroscopy and machine learning technology in multi-component mixture systems to identify the types and detect the content of specific substances mainly focuses on distinguishing medicinal materials with similar appearances or roughly estimating the effective concentration of adulterated samples. However, for lubricant additives, their concentrations are usually very low, accounting for only 0.1% to 5.0% of the final product. And the oil samples obtained in the real production environment may have subtle differences in composition compared to the oil samples prepared in the laboratory. In summary, the model needs higher accuracy and stronger generalization ability.
[0004] In view of this, this application is hereby filed. Summary of the invention
[0005] The present invention provides a method for predicting lubricant additives, which can quickly and effectively detect the content of additives in the lubricant in use, has higher detection efficiency and detection accuracy, and realizes quality monitoring of the lubricant of mechanical equipment.
[0006] The present invention is achieved through the following technical solutions:
[0007] A prediction method for lubricating oil additives, comprising:
[0008] Acquire relevant parameters of a lubricating oil sample and construct a training database for the lubricating oil sample, wherein the relevant parameters include base oil type and content, lubricating oil additive type and content, and mid-infrared spectrum data of the lubricating oil sample;
[0009] The training database is divided into a training set 1, a training set 2 and a test set, wherein the training set 1 is a mixed sample of a single lubricant additive and a base oil, and the training set 2 and the test set are mixed samples of multiple lubricant additives and base oils;
[0010] Building an analysis model material library, wherein the analysis model material library includes a variety of existing analysis models;
[0011] The linear relationship between the single lubricant additive and the mid-infrared spectrum is found through the training set 1, and the existing analysis models in the analysis model material library are trained using the training set 2 to obtain a plurality of intermediate analysis models, and the accuracy of the intermediate analysis models is tested through the test set, and the intermediate analysis model that reaches the preset accuracy is used as the trained lubricant additive intelligent analysis model;
[0012] A lubricant prediction material library to be tested is obtained, and mid-infrared spectrum data of lubricant samples in the lubricant prediction material library is input into the lubricant additive intelligent analysis model, thereby outputting predicted additive types and contents.
[0013] The method of the present invention comprises establishing a training database for lubricating oil; constructing an analysis model material library, collecting a plurality of existing analysis models; constructing an intelligent analysis model for lubricating oil additives, and training the existing analysis models in the analysis model material library through training set 1 and training set 2; outputting the trained intelligent analysis model for lubricating oil additives; constructing a prediction material library for lubricating oil to be tested, and inputting it into the trained intelligent analysis model for lubricating oil additives, and the intelligent analysis model for lubricating oil additives outputs the predicted additive content; the present invention uses the intelligent analysis model for lubricating oil additives to obtain the intrinsic connection between the mid-infrared spectrum and the type and content of lubricating oil additives, constructs a prediction model, and automatically identifies the type and content of additives in unknown oil products by inputting the mid-infrared spectrum data of unknown lubricating oil samples into the intelligent analysis model for lubricating oil additives.
[0014] As an optimization, the specific process of obtaining relevant parameters of the lubricating oil sample and constructing a training database for the lubricating oil sample is as follows:
[0015] Obtaining a lubricating oil sample, and constructing a training material library based on the lubricating oil sample;
[0016] Analyzing the components of the lubricating oil samples in the training material library and the mid-infrared spectrum data of the lubricating oil samples, wherein the components include the base oil type and content, the lubricating oil additive content, and the lubricating oil additive type;
[0017] The base oil type and content, lubricant oil additive type and content are used as influencing factors, the mid-infrared spectral data are used as input, the content of lubricant oil additives, the recognition accuracy of the characterization prediction results and the degree of fitting of the influencing factors, and the determination coefficient R 2 , root mean square error RMSE, and mean absolute percentage error MAPE are output, the influencing factors, inputs, and outputs are synthesized into a vector, and a training database of lubricating oil samples consisting of vectors corresponding to multiple lubricating oil samples is constructed.
[0018] As an optimization, the sampling requirements for the mid-infrared spectral data include:
[0019] Spectral range: 4000~400cm -1 ; The sampling method is potassium bromide window; the transmission wavelength range is 7800~400cm -1 ; Transmittance greater than 92%;
[0020] Acquisition settings: 16 scans, 4 resolution, data interval 0.482 cm -1 .
[0021] As optimization, the existing analysis models include Adaboost model, random forest model, and ridge regression model.
[0022] As an optimization, the specific process of finding the linear relationship between the single lubricant additive and the mid-infrared spectrum through the training set 1, and using the training set 2 to train the existing analysis model in the analysis model library to obtain the intermediate analysis model is as follows:
[0023] Performing first-order derivative processing on the mid-infrared spectral data in the training set 1, the training set 2 and the test set;
[0024] Performing linear regression processing on the training set 1 after the first-order derivative processing, and performing reverse linear regression processing on the training set 2 and the test set after the first-order derivative processing, so as to obtain a linear relationship between the absorbance at a single point of the lubricating oil samples in the training set 1, the training set 2 and the test set and the concentration of the lubricating oil additive;
[0025] The training set 2 subjected to reverse linear regression processing is input into a plurality of the existing analysis models, thereby obtaining a plurality of the intermediate analysis models.
[0026] As an optimization, the accuracy of the intermediate analysis model is tested by the test set, and the specific process of using the intermediate analysis model that reaches the preset accuracy as the trained lubricant additive intelligent analysis model is as follows:
[0027] Step A, setting a model prediction accuracy threshold, performing an accuracy test on the intermediate analysis model after training through the test set, and judging whether the set model prediction accuracy is reached. If the accuracy test does not reach the set model prediction accuracy, the intermediate analysis model is not output; if the test accuracy reaches the set model prediction accuracy, the intermediate analysis model is output as the trained initial lubricant additive intelligent analysis model, and then entering step B;
[0028] Step B, setting a minimum output model quantity threshold. If the number of trained initial lubricant additive intelligent analysis models outputted in step A is lower than the minimum output model quantity threshold, then correct the model prediction accuracy threshold and jump to step A for retraining. If it is not lower than the minimum output model quantity threshold, then compare the accuracy of all the initial lubricant additive intelligent analysis models and output the optimal initial lubricant additive intelligent analysis model as the optimal lubricant additive intelligent analysis model in the analysis model material library.
[0029] As an optimization, the initial value of the model prediction accuracy threshold set in step A is 0.95.
[0030] As an optimization, if the number of trained initial lubricant additive intelligent analysis models outputted in step A is lower than the minimum output model number threshold in step B, the model prediction accuracy threshold is corrected to 98% of the original model prediction accuracy threshold.
[0031] As an optimization, the ratio of the training set 2 to the test set is 3:1.
[0032] As an optimization, it also includes:
[0033] According to the types and contents of lubricant additives and the corresponding mid-infrared spectral data, a variable curve of the content and spectrum of each additive is constructed, and the type and content of the lubricant are determined based on the spectrum and the variable curve.
[0034] Compared with the prior art, the present invention has the following advantages and beneficial effects:
[0035] The present invention provides a method and system for predicting the attenuation of the content of lubricant additives in use, by constructing a training database of lubricant; collecting multiple analysis models and constructing an analysis model material library; outputting a preferred analysis model through the training set for each model in the analysis model material library; constructing a prediction material library of the lubricant to be tested, and inputting it into the trained preferred lubricant additive analysis model to predict the additive content; the present invention obtains the intrinsic connection between the mid-infrared spectrum and the type and content of lubricant additives by performing linear regression processing on the training set, and obtains the key technology for cracking the overlap and drift of the mid-infrared spectrum through the training of the preferred analysis model; the mid-infrared spectrum data of the unknown lubricant sample is input into the trained lubricant additive intelligent analysis model to automatically detect the type and content of the additives in the unknown oil. Compared with the traditional research method of the type and content of lubricant additives, it has higher detection efficiency and detection accuracy, saving manpower and material costs. BRIEF DESCRIPTION OF THE DRAWINGS
[0036] The drawings described herein are used to provide a further understanding of the embodiments of the present invention, constitute a part of this application, and do not constitute a limitation of the embodiments of the present invention. In the drawings:
[0037] Figure 1 It is an overall flow chart of a prediction method for lubricating oil additives according to the present invention;
[0038] Figure 2 The present invention is a detailed flow chart of a method for predicting lubricant additives. DETAILED DESCRIPTION
[0039] In order to make the objectives, technical solutions and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with embodiments and drawings. The exemplary embodiments of the present invention and their description are only used to explain the present invention and are not intended to limit the present invention.
[0040] This embodiment 1 provides a prediction method for lubricating oil additives, such as Figure 1 As shown, including:
[0041] S1. Obtain relevant parameters of a lubricating oil sample and construct a training database for the lubricating oil sample, wherein the relevant parameters include base oil type and content, lubricating oil additive type and content, and mid-infrared spectrum data of the lubricating oil sample;
[0042] In some embodiments, the specific process of S1 is:
[0043] S1.1, obtaining a lubricating oil sample, and constructing a training material library based on the lubricating oil sample; that is, the training material library includes mixed oil samples of various additives and base oil. In this embodiment, the lubricating oil sample can be prepared in a laboratory.
[0044] S1.2, analyzing the components of the lubricating oil samples in the training material library and the mid-infrared spectrum data of the lubricating oil samples, wherein the components include the base oil type and content, the lubricating oil additive content, and the lubricating oil additive type;
[0045] Specifically, the sampling requirements for the mid-infrared spectral data include:
[0046] Spectral range: 4000~400cm -1 ; The sampling method is potassium bromide window; the transmission wavelength range is 7800~400cm -1 ; Transmittance (1-25μm) greater than 92%;
[0047] Acquisition settings: 16 scans, 4 resolution, data interval 0.482 cm -1 .
[0048] S1.3, taking the base oil type and content, lubricant additive type and content as influencing factors, taking the mid-infrared spectral data as input, taking the content of lubricant additives, the recognition accuracy of the characterization prediction results and the degree of fitting of the influencing factors, and the determination coefficient R 2 , root mean square error RMSE, and mean absolute percentage error MAPE are output, the influencing factors, inputs, and outputs are synthesized into a vector, and a training database of lubricating oil samples consisting of vectors corresponding to multiple lubricating oil samples is constructed.
[0049] S2, the training database is divided into a training set 1, a training set 2 and a test set, wherein the training set 1 is a mixed sample of a single lubricant additive and a base oil, and the training set 2 and the test set are mixed samples of multiple lubricant additives and base oils, such as lubricant samples compounded with multiple additives. In this embodiment, the ratio of the training set 2 to the test set is 3:1.
[0050] S3, constructing an analysis model material library, wherein the analysis model material library includes a variety of existing analysis models;
[0051] In this embodiment, the existing analysis models include Adaboost model, random forest model, and ridge regression model.
[0052] S4, finding the linear relationship between the single lubricant additive and the mid-infrared spectrum through the training set 1, and using the training set 2 to train the existing analysis models in the analysis model library to obtain multiple intermediate analysis models, and testing the accuracy of the intermediate analysis models through the test set, and using the intermediate analysis model that reaches the preset accuracy as the trained lubricant additive intelligent analysis model;
[0053] In this embodiment, the specific process of finding the linear relationship between the single lubricant additive and the mid-infrared spectrum through the training set 1 and using the training set 2 to train the existing analysis model in the analysis model library to obtain the intermediate analysis model is as follows:
[0054] Performing first-order derivative processing on the mid-infrared spectral data in the training set 1, the training set 2, and the test set; here, the first-order derivative processing is to derive the wavenumber of the spectral data in the training set 1, the training set 2, and the test set;
[0055] Performing linear regression processing on the training set 1 after the first-order derivative processing, and performing reverse linear regression processing on the training set 2 and the test set after the first-order derivative processing, so as to obtain a linear relationship between the absorbance at a single point of the lubricating oil samples in the training set 1, the training set 2 and the test set and the concentration of the lubricating oil additive;
[0056] The linear regression processing is to perform single-point linear regression on training set one to obtain the linear relationship between the absorbance at a single point and the concentration of the additive. The reverse linear regression processing is to perform single-point reverse linear regression on training set two and the test set. Based on the linear relationship between the absorbance at a single point and the concentration of the additive obtained above, the concentration of the additive at each single point of the spectral data of training set two and the test set is inversely calculated.
[0057] Training set 1 is an oil sample made of a single additive and base oil. Therefore, linear regression processing is performed on training set 1 to obtain a linear relationship between concentration and spectrum.
[0058] By performing reverse linear regression on training set 2, we can get the predicted concentration value. Inputting the predicted concentration value and the true concentration value into the analysis model together can help the model solve the problem of peak stacking in the mid-infrared spectrum.
[0059] The training set 2 that has been subjected to the reverse linear regression processing after the linear regression processing is input into the plurality of the existing analysis models, thereby obtaining the plurality of the intermediate analysis models.
[0060] The essence of the above process is to use training set 1 to extract the variation law of concentration and spectrum, then use the law to process the data of training set 2, output the processed training set 2 to the existing analysis model, and then screen it through the test set to obtain the model with the highest accuracy. In a possible implementation, the linear regression equation in the linear regression processing is A=kc+Δ, where A is the absorbance data of the lubricating oil spectrum, k is the absorbance-concentration coefficient of a certain additive at this point, c is the actual concentration of a certain additive, and Δ is a redundant item in the data.
[0061] After linear regression of training set 1, the parameters k and Δ in the linear relationship parameter A=kc+Δ between additive concentration and spectral absorbance were obtained. The spectral data y of training set 2 was subjected to reverse linear regression to output the concentration value x, and the output concentration and the true concentration were input into the analysis model for training. Therefore, training set 1 was not input into the analysis model, while each analysis model had to input training set 2.
[0062] In this embodiment, the accuracy of the intermediate analysis model is tested by the test set, and the specific process of using the intermediate analysis model that reaches the preset accuracy as the trained lubricant additive intelligent analysis model is as follows:
[0063] The specific process is:
[0064] Step A, setting a model prediction accuracy threshold, specifically, the initial value of the model prediction accuracy threshold is set to 0.95, and the accuracy test is performed on the intermediate analysis model after training through the test set to determine whether the set model prediction accuracy is achieved. If the accuracy test does not achieve the set model prediction accuracy, the intermediate analysis model is not output; if the test accuracy reaches the set model prediction accuracy, the intermediate analysis model is output as the trained initial lubricant additive intelligent analysis model, and then enters step B;
[0065] Step B, setting a minimum output model quantity threshold. If the number of trained initial lubricant additive intelligent analysis models outputted in step A is lower than the minimum output model quantity threshold, then correct the model prediction accuracy threshold and jump to step A for retraining. If it is not lower than the minimum output model quantity threshold, then compare the accuracy of all the initial lubricant additive intelligent analysis models and output the optimal initial lubricant additive intelligent analysis model as the optimal lubricant additive intelligent analysis model in the analysis model material library.
[0066] In some embodiments, the model prediction accuracy threshold in step A is: the accuracy of lubricant additive type identification is ≥ 0.95; the determination coefficient R 2 ≥0.97; RMSE≤0.05; MAPE≤0.05. If the number of trained lubricant additive intelligent analysis models output in step A is lower than the minimum output model number threshold in step B, it is corrected to 98% of the model prediction accuracy threshold in the original step A.
[0067] S5. Obtain a prediction material library of lubricating oil to be tested, and input the mid-infrared spectrum data of the lubricating oil samples in the lubricating oil prediction material library into the lubricating oil additive intelligent analysis model, thereby outputting the predicted type and content of the additive.
[0068] Example 2
[0069] On the basis of Example 1, a variable curve of the content and spectrum of each additive is constructed according to the type and content of lubricating oil additives and the corresponding mid-infrared spectrum data, and the type and content of the lubricating oil are determined according to the spectrum and the variable curve.
[0070] Compared with Example 1, Example 2 focuses on being able to predict the content of additives in finished lubricating oil.
[0071] Example 3
[0072] The difference between this embodiment and embodiment 1 is that this embodiment provides a specific implementation case based on the method flow steps of embodiment 1, as follows:
[0073] S1. The laboratory uses PAO as base oil, sulfided isobutylene, alkylated diphenylamine, thiophosphate amine salt, dialkyl dithiocarbamate molybdenum, dialkyl dithiophosphate zinc and other additives with different mass fractions (0.5%, 1%, 2%, 3%) to prepare lubricating oil samples, and uses Fourier transform infrared spectrometer to obtain its infrared spectrum data to obtain the laboratory lubricating oil training material library.
[0074] After obtaining the finished lubricant oil sample, the infrared spectrum data thereof is obtained by using a Fourier infrared spectrometer to obtain a finished lubricant oil prediction material library.
[0075] S2, divide the training database into training set 1, training set 2 and test set;
[0076] S3. Build a library of intelligent analysis models for lubricant additives, including Adaboost model, random forest model, ridge regression model, etc.
[0077] S4, using the training set 1 and the training set 2 to train the existing analysis models in the analysis model material library to obtain a plurality of intermediate analysis models, and using the test set to test the accuracy of the intermediate analysis models, and using the intermediate analysis models that reach a preset accuracy as the trained lubricant additive intelligent analysis model;
[0078] The data of the laboratory lubricant training material library is processed by first-order derivative, linear regression, and reverse linear regression, and the processed data is input into the analysis model material library for modeling. In the Python-based lubricant additive intelligent analysis model, the training set training data is imported and the test set is used to test the intelligent algorithm model to observe the test accuracy value.
[0079] Correcting the initial lubricant additive intelligent analysis model to achieve the expected accuracy, and selecting the best lubricant additive intelligent analysis model;
[0080] Set the model prediction accuracy threshold: Lubricant additive type identification accuracy ≥ 0.95; determination coefficient R 2 ≥0.97; RMSE≤0.05; MAPE≤0.05. The trained intelligent analysis model of lubricating oil additives is tested for accuracy through the test set to determine whether the set model prediction accuracy is achieved. If the accuracy test does not achieve the set model prediction accuracy, the analysis model is not output; if the test accuracy reaches the set model prediction accuracy, the trained intelligent analysis model of lubricating oil additives is output;
[0081] A minimum output model quantity threshold is set. If the number of trained lubricant additive intelligent analysis models is lower than the threshold, the model prediction accuracy threshold is corrected and the process is skipped to step three for retraining. If the minimum output model quantity threshold is reached, the accuracy of the output models is compared and the optimal lubricant additive intelligent analysis model is output as the preferred lubricant additive intelligent analysis model for the analysis model library.
[0082] S5. Predict the content of lubricant additives in mechanical equipment.
[0083] The mid-infrared spectrum data of the finished lubricating oil to be tested is directly input into the trained intelligent analysis model of lubricating oil additives to output the predicted type and content of the additives.
[0084] In the finished lubricating oil, the mass ratio of the base oil is 0.9500-0.9700, and the mass ratio of the additive is 0.0500-0.0300.
[0085] Embodiment 4 further discloses a prediction system for lubricating oil additives, which is used to execute the method of embodiment 1, comprising:
[0086] A training database construction unit, used to obtain relevant parameters of a lubricating oil sample and construct a training database for the lubricating oil sample, wherein the relevant parameters include base oil type and content, lubricating oil additive type and content, and mid-infrared spectrum data of the lubricating oil sample;
[0087] a training database division unit, which divides the training database into a training set 1, a training set 2 and a test set, wherein the training set 1 is a mixed sample of a single lubricant additive and a base oil, and the training set 2 and the test set are mixed samples of multiple lubricant additives and base oils;
[0088] A model building unit, used to build an analysis model material library, wherein the analysis model material library includes a variety of existing analysis models;
[0089] A model training unit, used to train the existing analysis models in the analysis model material library using the training set 1 and the training set 2 to obtain a plurality of intermediate analysis models, and to test the accuracy of the intermediate analysis models using the test set, and to use the intermediate analysis models that reach a preset accuracy as the trained lubricant additive intelligent analysis model;
[0090] The prediction unit is used to obtain a prediction material library of lubricating oil to be tested, input the mid-infrared spectrum data of the lubricating oil sample in the lubricating oil prediction material library into the lubricating oil additive intelligent analysis model, and output the predicted additive type and content.
[0091] The execution process of each unit can be performed according to the process steps of a method for quantitative analysis of lubricating oil additives in Example 1, and will not be described in detail in this embodiment.
[0092] At the same time, the present invention also provides a computer-readable storage medium, which stores a computer program. When the computer program is executed by a processor, the above-mentioned method for quantitative analysis of lubricating oil additives is implemented.
[0093] Those skilled in the art will appreciate that the embodiments of the present application may be provided as methods, systems, or computer program products. Therefore, the present application may adopt the form of a complete hardware embodiment, a complete software embodiment, or an embodiment in combination with software and hardware. Moreover, the present application may adopt the form of a computer program product implemented in one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) that include computer-usable program code.
[0094] The present application is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of the processes and / or boxes in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.
[0095] These computer program instructions may also be stored in a computer readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer readable memory produce an article of manufacture including an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.
[0096] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.
[0097] The specific implementation methods described above further illustrate the objectives, technical solutions and beneficial effects of the present invention in detail. It should be understood that the above description is only a specific implementation method of the present invention and is not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.
Claims
1. A method for predicting lubricating oil additives, characterized in that: include: Acquire relevant parameters of a lubricating oil sample and construct a training database for the lubricating oil sample, wherein the relevant parameters include base oil type and content, lubricating oil additive type and content, and mid-infrared spectrum data of the lubricating oil sample; The training database is divided into a training set 1, a training set 2 and a test set, wherein the training set 1 is a mixed sample of a single lubricant additive and a base oil, and the training set 2 and the test set are mixed samples of multiple lubricant additives and base oils; Building an analysis model material library, wherein the analysis model material library includes a variety of existing analysis models; The linear relationship between the single lubricant additive and the mid-infrared spectrum is found through the training set 1, and the existing analysis models in the analysis model material library are trained using the training set 2 to obtain a plurality of intermediate analysis models, and the accuracy of the intermediate analysis models is tested through the test set, and the intermediate analysis model that reaches the preset accuracy is used as the trained lubricant additive intelligent analysis model; A lubricant prediction material library to be tested is obtained, and mid-infrared spectrum data of lubricant samples in the lubricant prediction material library is input into the lubricant additive intelligent analysis model, thereby outputting predicted additive types and contents.
2. The prediction method for lubricating oil additives according to claim 1, characterized in that: The specific process of obtaining relevant parameters of lubricating oil samples and constructing a training database for the lubricating oil samples is as follows: Obtaining a lubricating oil sample, and constructing a training material library based on the lubricating oil sample; Analyzing the components of the lubricating oil samples in the training material library and the mid-infrared spectrum data of the lubricating oil samples, wherein the components include the base oil type and content, the lubricating oil additive content, and the lubricating oil additive type; The base oil type and content, lubricant oil additive type and content are used as influencing factors, the mid-infrared spectral data are used as input, the content of lubricant oil additives, the recognition accuracy of the characterization prediction results and the degree of fitting of the influencing factors, and the determination coefficient R 2 , root mean square error RMSE, and mean absolute percentage error MAPE are output, the influencing factors, inputs, and outputs are synthesized into a vector, and a training database of lubricating oil samples consisting of vectors corresponding to multiple lubricating oil samples is constructed.
3. A prediction method for lubricating oil additives according to claim 1 or 2, characterized in that: The sampling requirements for mid-infrared spectral data include: Spectral range: 4000~400cm -1 ; The sampling method is potassium bromide window; the transmission wavelength range is 7800~400cm -1 ; Transmittance greater than 92%; Acquisition settings: 16 scans, 4 resolution, data interval 0.482 cm -1 .
4. The prediction method for lubricating oil additives according to claim 1, characterized in that: The existing analysis models include Adaboost model, random forest model and ridge regression model.
5. The prediction method for lubricating oil additives according to claim 1, characterized in that: The specific process of finding the linear relationship between the single lubricant additive and the mid-infrared spectrum through the training set 1, and using the training set 2 to train the existing analysis model in the analysis model library to obtain the intermediate analysis model is as follows: Performing first-order derivative processing on the mid-infrared spectral data in the training set 1, the training set 2 and the test set; Performing linear regression processing on the training set 1 after the first-order derivative processing, and performing reverse linear regression processing on the training set 2 and the test set after the first-order derivative processing, so as to obtain a linear relationship between the absorbance at a single point of the lubricating oil samples in the training set 1, the training set 2 and the test set and the concentration of the lubricating oil additive; The training set 2 subjected to reverse linear regression processing is input into a plurality of the existing analysis models, thereby obtaining a plurality of the intermediate analysis models.
6. The prediction method for lubricating oil additives according to claim 5, characterized in that: The specific process of testing the accuracy of the intermediate analysis model through the test set and using the intermediate analysis model that reaches the preset accuracy as the trained lubricant additive intelligent analysis model is as follows: Step A, setting a model prediction accuracy threshold, performing an accuracy test on the intermediate analysis model after training through the test set, and judging whether the set model prediction accuracy is reached. If the accuracy test does not reach the set model prediction accuracy, the intermediate analysis model is not output; if the test accuracy reaches the set model prediction accuracy, the intermediate analysis model is output as the trained initial lubricant additive intelligent analysis model, and then entering step B; Step B, setting a minimum output model quantity threshold. If the number of trained initial lubricant additive intelligent analysis models outputted in step A is lower than the minimum output model quantity threshold, then correcting the model prediction accuracy threshold and jumping to step A for retraining. If it is not less than the minimum output model quantity threshold, the accuracies of all the initial lubricant additive intelligent analysis models are compared, and the optimal initial lubricant additive intelligent analysis model is output as the optimal lubricant additive intelligent analysis model in the analysis model material library.
7. A prediction method for lubricating oil additives according to claim 6, characterized in that: The initial value of the model prediction accuracy threshold set in step A is 0.
95.
8. A prediction method for lubricating oil additives according to claim 6 or 7, characterized in that: If the number of trained initial lubricant additive intelligent analysis models outputted in step A is lower than the minimum output model number threshold in step B, the model prediction accuracy threshold is corrected to 98% of the original model prediction accuracy threshold.
9. The prediction method for lubricating oil additives according to claim 1, characterized in that: The ratio of the training set 2 to the test set is 3:
1.
10. The prediction method for lubricating oil additives according to claim 1, characterized in that: Also includes: According to the types and contents of lubricant additives and the corresponding mid-infrared spectral data, a variable curve of the content and spectrum of each additive is constructed, and the type and content of the lubricant are determined based on the spectrum and the variable curve.