Lubricating oil thermal aging raman detection method based on combination pretreatment and enhanced fisher discrimination

Abstract

The present application relates to the technical field of lubricating oil performance rapid detection, and discloses a lubricating oil thermal aging Raman detection method based on combined pretreatment and enhanced Fisher discrimination, which acquires original Raman spectrum data of the lubricating oil to be detected; performs feature interval clipping and normalization based on preset thermal aging characteristic groups to obtain a first pretreated spectrum; performs combined pretreatment on the first pretreated spectrum by using row and column pretreatment operators, corrects scattering offset by using a row operator, and performs weight distribution and feature scaling according to signal variance by using a column operator to obtain a target feature data sequence; inputs the target feature data sequence into a detection model based on Fisher discrimination which is established offline based on adjacent aging grade discrimination score centroid weighted distance screening, and outputs the thermal aging grade of the lubricating oil to be detected. The present application does not require complex pretreatment, can accurately amplify weak features of slight aging, has extremely high accuracy in identifying critical samples, has strong anti-interference capability, and meets the demand for on-site detection.

Description

Technical Field

[0001] This invention relates to the field of rapid testing technology for lubricating oil performance, and in particular to a combined pretreatment-enhanced Fisher discrimination method for the thermal aging Raman detection of lubricating oil. Background Technology

[0002] During long-term high-temperature operation, lubricating oil undergoes thermal aging reactions such as oxidation, chain breakage, and polymerization, generating aging products such as carboxylic acids, gums, and sludge. These directly lead to a decline in lubrication performance, causing equipment wear, leaks, and even shutdowns. Currently used methods for detecting the thermal aging of lubricating oil have the following core shortcomings:

[0003] First, traditional physicochemical index testing (such as acid value, kinematic viscosity, and n-pentane insoluble matter) requires complex pretreatment processes such as extraction and centrifugation, resulting in long single-sample testing cycles and failing to meet the needs of rapid on-site testing. Second, infrared spectroscopy is susceptible to interference from moisture and impurities in the oil, and its sensitivity for mildly aged samples (such as samples with low acid values) is insufficient. Finally, existing Raman spectroscopy detection schemes mostly use raw spectra combined with single preprocessing modeling, and typically only use the overall accuracy of the model as the screening criterion, without fully considering the discrimination margin for critical states such as unaged or mildly aged samples. This leads to a high false positive rate for critical samples and insufficient reliability in practical applications. Summary of the Invention

[0004] This invention provides a Raman detection method for lubricating oil thermal aging that enhances Fisher's judgment through combined preprocessing. By employing a technical approach of feature interval directional cropping, scattering correction-feature enhancement combined preprocessing, and secondary weighted centroid distance model screening, it achieves accurate determination of lubricating oil aging levels without sample pretreatment. At the same time, it significantly improves the accuracy of critical state identification and the robustness of samples under complex working conditions, making it suitable for various application scenarios such as on-site sampling, offline periodic testing, and online monitoring.

[0005] This invention provides a combined pretreatment-enhanced Fisher discrimination method for Raman detection of lubricating oil thermal aging, comprising:

[0006] The raw Raman spectral data of the lubricating oil to be tested is obtained, and the characteristic range of the raw Raman spectral data is clipped based on the preset characteristic groups of the lubricating oil thermal aging to obtain clipped spectral data. The clipped spectral data is then normalized to obtain the first preprocessed spectrum.

[0007] The first preprocessed spectrum is preprocessed by a combination of preset row preprocessing operators and column preprocessing operators to obtain a target feature data sequence; wherein, the row preprocessing operator corrects the scattering shift of the first preprocessed spectrum to obtain intermediate corrected data, and the column preprocessing operator performs weight allocation and feature scaling on the intermediate corrected data according to the preset variance of each wavenumber signal.

[0008] The target feature data sequence is input into a pre-established lubricating oil thermal aging detection model based on Fisher discrimination, and the thermal aging level of the lubricating oil to be tested is output.

[0009] Further, the steps of acquiring the raw Raman spectral data of the lubricating oil to be tested, and cropping the raw Raman spectral data into a characteristic region based on preset characteristic groups of lubricating oil thermal aging to obtain cropped spectral data, and normalizing the cropped spectral data to obtain the first preprocessed spectrum, specifically include:

[0010] The optical signal of the lubricating oil sample to be tested is acquired by an excitation source, and the detected optical signal is converted into an electrical digital signal by a spectrometer to obtain the original Raman spectral data.

[0011] According to the preset clipping rules, the original Raman spectral data is subjected to feature region clipping processing, retaining the target feature bands corresponding to the thermal aging feature groups, and removing irrelevant band data other than the target feature bands to obtain the clipped spectral data.

[0012] The cropped spectral data is subjected to full-range area normalization processing, and the total spectral signal intensity of all target feature bands contained in the cropped spectral data is uniformly normalized to 1 to obtain the first preprocessed spectrum.

[0013] Furthermore, the target characteristic band includes 2800-3000 cm⁻¹. -1 CH stretching vibration zone, 1650-1750cm -1 The carbonyl stretching vibration region and 3200-3600 cm -1 The hydroxyl stretching vibration region.

[0014] Further, the step of performing combined preprocessing of the first preprocessed spectrum using preset row preprocessing operators and column preprocessing operators to obtain the target feature data sequence includes:

[0015] A preset row preprocessing operator is invoked to mathematically correct the scattering shift of the first preprocessed spectrum along the sample dimension to obtain intermediate corrected data; wherein, the row preprocessing operator is a multivariate scattering correction or a standard normal variable transformation, used to perform sample-by-sample full-spectrum scattering interference correction processing;

[0016] A preset column preprocessing operator is invoked to perform weight allocation and feature scaling on the intermediate corrected data along the wavenumber dimension based on the signal variance, thereby obtaining the target feature data sequence; wherein, the column preprocessing operator is a variance scaling operator, used to perform wavenumber-by-wavenumber feature intensity scaling processing along the wavenumber dimension.

[0017] Furthermore, the step of inputting the target feature data sequence into a pre-established lubricating oil thermal aging detection model based on Fisher discrimination, and outputting the thermal aging level of the lubricating oil to be tested, specifically includes:

[0018] A two-stage screening based on dual-objective parameters was performed to build an offline lubricating oil thermal aging detection model based on Fisher discrimination.

[0019] The target feature data sequence is input into the lubricating oil thermal aging detection model for classification and inference, and the thermal aging level of the lubricating oil to be tested is output.

[0020] The quantitative physicochemical indicators of the lubricating oil to be tested are derived using partial least squares regression correlation mapping rules, and the data is displayed and stored in the database.

[0021] Furthermore, the step of performing two-level screening based on dual-objective parameters and constructing an offline lubricating oil thermal aging detection model based on Fisher discrimination specifically includes:

[0022] Multiple initial models were obtained by using different preprocessing operators and spectral sets of lubricating oil samples with calibrated aging levels, combined with the Fisher discrimination algorithm.

[0023] Cross-validation of the multiple initial models is performed using a test set, and the first set of candidate models whose full aging level discrimination accuracy meets the preset conditions is retained as the first-level screening.

[0024] For the candidate models in the first candidate model set, the variance contribution of their Fisher discriminant function is extracted as a weight. The average weighted distance between the centroid of the unaged sample and the centroids of the other aging level samples, as well as the average weighted distance between the centroids of adjacent aging level samples, are calculated as the second level of screening.

[0025] The candidate model with the maximum average weighted distance was selected as the final lubricating oil thermal aging detection model.

[0026] Furthermore, in the second-level screening, the formula for calculating the average weighted distance D of the centroids of adjacent aging level samples is as follows:

[0027]

[0028] Where n is the number of groups of adjacent aging levels; and These are the mean scores of the first classification of the aging level samples at level i and level i+1, respectively. and These are the mean scores of the second classification of the aging level samples at level i and level i+1, respectively. and These represent the percentage of variance contribution from the first and second classification discriminant functions, respectively.

[0029] Furthermore, the step of deriving the quantitative physicochemical indicators of the lubricating oil to be tested using partial least squares regression association mapping rules, and then displaying and storing the data, specifically includes:

[0030] Based on the pre-stored partial least squares regression correlation mapping rule between the actual acid value, kinematic viscosity change rate and Fisher discrimination score, the classification inference results of the lubricating oil to be tested are transformed into continuous quantitative physicochemical index prediction values.

[0031] The thermal aging level and the predicted values ​​of the quantified physicochemical indicators are displayed locally on the screen, and the test data is remotely uploaded to the management platform through the communication unit. The data storage unit performs local or cloud storage to form a historical test record ledger.

[0032] The present invention also provides a combined pretreatment enhanced Fisher discrimination lubricating oil thermal aging Raman detection device, based on the combined pretreatment enhanced Fisher discrimination lubricating oil thermal aging Raman detection method described above, the device comprising:

[0033] The Raman spectroscopy acquisition module is used to acquire the raw Raman spectral data of the lubricating oil to be tested, and to cut the raw Raman spectral data into characteristic regions based on preset characteristic groups of lubricating oil thermal aging to obtain cut spectral data. The cut spectral data is then normalized to obtain the first preprocessed spectrum.

[0034] The preprocessing module is used to perform combined preprocessing on the first preprocessed spectrum using preset row preprocessing operators and column preprocessing operators to obtain a target feature data sequence; wherein, the row preprocessing operator corrects the scattering shift of the first preprocessed spectrum to obtain intermediate corrected data, and the column preprocessing operator performs weight allocation and feature scaling on the intermediate corrected data according to preset variances of each wavenumber signal.

[0035] The result output module is used to input the target feature data sequence into a pre-established lubricating oil thermal aging detection model based on Fisher discrimination, and output the thermal aging level of the lubricating oil to be tested.

[0036] The present invention also provides a computer device, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps of the above-described method.

[0037] The present invention also provides a computer-readable storage medium having a computer program stored thereon, wherein the computer program, when executed by a processor, implements the steps of the above-described method.

[0038] The beneficial effects of this invention are as follows:

[0039] 1. This invention eliminates the need for complex pretreatment operations such as extraction and centrifugation of in-use oil samples, significantly shortening the single-sample testing cycle and greatly improving testing efficiency compared to traditional physicochemical testing. Furthermore, this invention eliminates the need for chemical organic reagents, generates no testing waste liquid, and avoids secondary environmental pollution; moreover, the amount of oil sample required for a single test is extremely small, making sampling convenient and effectively reducing equipment downtime and economic losses.

[0040] 2. By combining feature interval targeted cropping with column preprocessing (variance scaling) to amplify weak features, and in conjunction with a unique two-level weighted centroid distance screening model, this invention can accurately capture weak feature signals of low-content aging products. It effectively solves the technical problem of insignificant features in the mild aging stage, accurately distinguishes between unaged and mildly aged borderline samples, and greatly improves detection sensitivity and classification accuracy.

[0041] 3. This invention employs preprocessing techniques (such as multivariate scattering correction or standard normal variable transformation) to effectively overcome Raman spectral scattering shifts caused by micron-sized colloidal particles, trace amounts of moisture, and mechanical impurities in the oil. This significantly enhances the anti-interference capability and robustness of the detection model under complex operating conditions, eliminating the need for pre-filtration of the oil to be tested and directly adapting to complex application scenarios such as industrial field sampling and online monitoring.

[0042] 4. This invention, based on a near-infrared excitation light source and a lightweight Fisher discriminant model, significantly reduces the hardware cost and subsequent maintenance costs of the detection solution compared to traditional equipment such as gas chromatography and dedicated infrared oil detectors. Furthermore, the lightweight algorithm model supports embedded deployment, making it suitable for both handheld portable devices and online real-time monitoring via connection to equipment lubrication lines. The model also boasts low migration costs and is easily adaptable to the lubrication aging detection needs of various industrial equipment, including wind turbine gearboxes and construction machinery. Attached Figure Description

[0043] Figure 1 This is a schematic diagram of a method flow according to an embodiment of the present invention.

[0044] Figure 2 This is a schematic diagram of the two-dimensional scatter distribution obtained by using principal component analysis (PCA) to extract features after combined preprocessing of lubricating oil sample data in this invention.

[0045] Figure 3 This is a schematic diagram of the two-dimensional feature scatter distribution obtained by extracting latent factors using partial least squares (PLS) after combined preprocessing of lubricating oil sample data in this invention.

[0046] Figure 4This is a schematic diagram of the visual two-dimensional scatter distribution of lubricating oil sample data after combined preprocessing, output by the t-SEN manifold learning dimensionality reduction algorithm.

[0047] Figure 5 This is a schematic diagram of the device structure according to an embodiment of the present invention.

[0048] Figure 6 This is a schematic diagram of the internal structure of a computer device according to an embodiment of the present invention.

[0049] The realization of the objective, functional features and advantages of the present invention will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0050] It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention.

[0051] This invention provides a combined pretreatment-enhanced Fisher's discrimination Raman detection method for lubricating oil thermal aging, addressing three practical problems in the existing lubricating oil thermal aging detection field: First, traditional physicochemical detection requires complex pretreatment such as extraction and centrifugation, with a single sample detection cycle exceeding 2 hours, resulting in low detection efficiency; precision detection methods such as gas chromatography have high equipment costs and high operational barriers, and neither of these methods can meet the needs of rapid and low-cost on-site detection; Second, infrared and ordinary Raman detection are insufficient in extracting weak features of low-content carbonyl and hydroxyl oxidation products in the mild aging stage, with a false judgment rate of over 20% for unaged / mildly aged critical samples, resulting in insufficient detection sensitivity; Third, existing Raman modeling schemes only use overall discrimination accuracy as a screening index, without considering the discrimination margin between different aging levels, resulting in poor resistance to interference from gum particles and trace moisture impurities in the oil, and low reliability in practical applications.

[0052] like Figure 1 As shown, the present invention provides a combined pretreatment-enhanced Fisher discrimination method for the Raman spectroscopy detection of lubricating oil thermal aging, which can be widely applied to scenarios such as rapid identification of the thermal aging level of in-use lubricating oil in various heavy-duty industrial equipment, including wind turbine gearboxes, industrial steam turbines, and hydraulic systems of construction machinery, as well as quantitative detection of degradation indicators and assessment of remaining service life. Specifically, it includes the following steps:

[0053] S1. Obtain the original Raman spectrum data of the lubricating oil to be tested, and perform feature interval cropping on the original Raman spectrum data based on the preset characteristic groups of lubricating oil thermal aging to obtain cropped spectrum data. Normalize the cropped spectrum data to obtain the first preprocessed spectrum.

[0054] In one embodiment of the present invention, step S1 includes the following sub-steps:

[0055] S101. Collect the raw Raman spectral data of the lubricating oil to be tested.

[0056] Optical signal acquisition was performed on the lubricating oil sample to be tested. Specifically, a 785nm near-infrared excitation source was used as the excitation source. Compared with the conventional 532nm visible light source, the 785nm near-infrared source can effectively reduce the fluorescence background intensity caused by aromatic components in the lubricating oil by more than 45%, thus avoiding the weak characteristics of low-content aging products being masked by strong fluorescence. The light-emitting end of the source was coupled to the incident fiber port of the coaxial reflective Raman probe via an SMA905 interface. The transparent quartz sample cell was fixed at the focal plane position 10mm directly in front of the Raman probe, and the sampling spot diameter was controlled to 2mm to obtain the optimal signal excitation and collection environment. The scattered signal collection fiber port of the Raman probe was coupled to the incident slit of a high-sensitivity CCD spectrometer. The CCD spectrometer converted the detected optical signal into an electrical digital signal output to obtain the raw Raman spectral data.

[0057] S102. Perform feature interval clipping to obtain clipped spectral data.

[0058] The raw Raman spectral data obtained above were subjected to feature region clipping. Based on the changes in molecular structure generated during the thermal aging of lubricating oil, a preset clipping rule was established, retaining only three feature bands strongly correlated with thermal aging, specifically:

[0059] ①2800-3000cm -1 CH stretching vibration region;

[0060] ②1650-1750cm -1 The carbonyl stretching vibration region;

[0061] ③3200-3600cm -1 The hydroxyl stretching vibration region.

[0062] By removing irrelevant band data other than the three characteristic bands mentioned above, approximately 72% of irrelevant signal redundancy is effectively trimmed, reducing the subsequent computational load and eliminating interference from irrelevant band noise, thus obtaining trimmed spectral data.

[0063] S103. Normalization is performed to obtain the first preprocessed spectrum.

[0064] The clipped spectral data undergoes full-range area normalization. Through mathematical calculation, the total spectral signal intensity of all three characteristic bands contained in the clipped spectral data is standardized to 1. This step aims to eliminate signal intensity inconsistency errors caused by physical factors such as laser power fluctuations and probe-sample cell installation deviations, ensuring that the obtained first preprocessed spectrum can objectively reflect the aging degree of the tested lubricating oil under unstable on-site testing conditions, thus avoiding result bias.

[0065] S2. The first preprocessed spectrum is preprocessed by a combination of preset row preprocessing operators and column preprocessing operators to obtain a target feature data sequence; wherein, the row preprocessing operator corrects the scattering shift of the first preprocessed spectrum to obtain intermediate corrected data, and the column preprocessing operator performs weight allocation and feature scaling on the intermediate corrected data according to the preset variance of each wavenumber signal.

[0066] In one embodiment of the present invention, step S2 includes the following sub-steps:

[0067] S201. Call the preset row preprocessing operator to perform mathematical correction on the scattering offset of the first preprocessed spectrum along the sample dimension to obtain intermediate correction data.

[0068] Specifically, in the preset spectral preprocessing operator library (which includes six sets of operators: no preprocessing, multivariate scattering correction (MSC), standard normal variable transformation (SNV), variance scaling, a combination of MSC and variance scaling, and a combination of SNV and variance scaling), the row preprocessing operator called in this step is either multivariate scattering correction (MSC) or standard normal variable transformation (SNV). At the underlying data processing logic, sample-by-sample full-spectrum scattering interference correction is performed along the sample dimension to eliminate Raman spectral scattering shifts caused by micron-sized colloidal particles generated by the thermal aging of lubricating oil, while also resisting scattering noise from trace amounts of moisture and mechanical impurities that may be present in the oil sample. This significantly improves the spectral signal-to-noise ratio, resulting in denoised intermediate corrected data.

[0069] S202. Call the preset column preprocessing operator to perform weight allocation and feature scaling on the intermediate correction data along the wavenumber dimension based on the signal variance to obtain the target feature data sequence.

[0070] Specifically, the column preprocessing operator invoked is the variance scaling operator. After completing the row preprocessing (i.e., full-spectrum scattering correction), wavenumber-by-wavenumber feature intensity scaling is further performed along the wavenumber dimension. Specifically, weights are assigned according to the variance of the spectral signal at each wavenumber position, specifically amplifying the weak feature signals of low-concentration carbonyl and hydroxyl aging products. Through this mathematical amplification, the technical challenge of insignificant features of low-content oxidation products in the unaged and mildly aged critical stages is effectively solved. The final output is a target feature data sequence with dual row and column enhancement, which can then be accurately identified by the subsequent classification model.

[0071] S3. Input the target feature data sequence into the pre-established lubricating oil thermal aging detection model based on Fisher discrimination, and output the thermal aging level of the lubricating oil to be tested.

[0072] In one embodiment of the present invention, step S3 includes the following sub-steps:

[0073] S301. Based on dual objective parameters, a two-level screening is performed to build an offline lubricating oil thermal aging detection model based on Fisher discrimination.

[0074] Before conducting actual tests on the lubricating oil to be tested, the offline training and screening of the testing model must be completed in advance. The specific construction process is as follows:

[0075] 1) Prepare lubricating oil samples with different degrees of thermal aging, and measure their true acid value, kinematic viscosity change rate at 40℃, and n-pentane insoluble matter to determine the true aging level (no aging, light aging, moderate aging, and heavy aging). Divide the samples into training and testing sets in a 7:3 ratio. For six different preprocessing operators, establish six corresponding initial Fisher discriminant models. The Fisher discriminant algorithm achieves multi-class classification by maximizing inter-class variance and minimizing intra-class variance. Its computational complexity is 80% lower than that of Support Vector Machines (SVM) or neural networks, meeting the embedded deployment requirements of portable devices.

[0076] To verify the enhancement effect of the combined preprocessing method described in this invention on sample features, such as... Figures 2 to 4 As shown, target feature data of samples with different aging levels (legend 0.0 represents non-aged / mildly aged class, legend 1.0 represents moderate / severely aged class) were extracted after combined preprocessing, and dimensionality reduction visualization analysis was performed using PCA, PLS, and t-SEN algorithms, respectively. Figure 2 PCA feature scatter plot, Figure 3 PLS feature scatter plot and Figure 4 As shown in the t-SEN visualization scatter plot, after the row and column combination preprocessing of this invention, samples of different categories form independent clusters with clear boundaries and no overlap in the two-dimensional projection space. This proves that the preprocessing operator of this method can effectively amplify extremely weak oxidation features, improve inter-class dispersion, and provide a data foundation for the high-precision discrimination of the subsequent Fisher model.

[0077] 2) Perform the first-level screening (rejection): call the test set with the above-mentioned aging level to perform cross-validation on the 6 initial models, set a strict rejection threshold, and only retain the model with a full aging level discrimination accuracy of 100% on both the training set and the test set as the candidate model, thereby eliminating models with overfitting or insufficient generalization performance.

[0078] 3) Perform a second-level screening (optimization). For the candidate models selected in the first-level screening, extract the variance contribution of their Fisher discriminant function as weights, and calculate two core indicators: the first is the average weighted distance Ravg between the centroid of the unaged sample and the centroids of the other three aging levels; the second is the average weighted distance Davg between adjacent aging level centroids. The formula for calculating the weighted distance between centroids of adjacent aging level discrimination scores is as follows:

[0079]

[0080] In the formula, n is the number of groups of adjacent aging levels (in this embodiment, n=3, corresponding to non-aging-mild, mild-moderate, and moderate-severe). and These are the mean scores of the first classification of the aging level samples at level i and level i+1, respectively. and These are the mean scores of the second classification of the aging level samples at level i and level i+1, respectively. and These represent the percentage of variance contribution from the first and second classification discriminant functions, respectively.

[0081] 4) Select the candidate model with the maximum value of both Ravg and Davg as the final lubricating oil thermal aging detection model. This screening logic solves the problem of misjudging critical samples caused by using only the overall accuracy rate as an indicator, and improves the identification accuracy of critical samples with mild aging to 100%.

[0082] S302. Input the target feature data sequence into the detection model for classification and inference, and output the thermal aging level.

[0083] During the on-site testing phase, the target feature data sequence output in step S2 is input into the lubricating oil thermal aging test model that has been solidified and deployed in step S301. Based on the extracted target feature data sequence, the model calculates the discrimination score of the current sample under test in each classification dimension, classifies it according to the centroid distance, and directly outputs the heat dissipation aging level label of the lubricating oil under test (such as no aging, slight aging, etc.).

[0084] S303. Use partial least squares regression correlation mapping rules to derive quantitative physicochemical indicators, and perform data display and storage.

[0085] While outputting classification labels, the model pre-stores a partial least squares regression (PLS) correlation mapping rule between the actual acid value, the rate of change of kinematic viscosity at 40℃, and the Fisher classification score. Through this mapping rule, the Fisher classification result of the current test sample is further transformed into a continuous quantitative physicochemical index prediction value.

[0086] The output end uses an LCD screen to display localized data, intuitively presenting the thermal aging level and quantitative physicochemical indicators; at the same time, the test data is remotely uploaded to the equipment operation and maintenance management platform via Bluetooth or 4G communication unit; the test data is finally stored locally or in the cloud by the data storage unit to form a historical test record ledger, providing a data foundation for subsequent assessment of the remaining service life of the equipment.

[0087] like Figure 5 As shown, the present invention also provides a combined pretreatment enhanced Fisher discrimination lubricating oil thermal aging Raman detection device, based on the combined pretreatment enhanced Fisher discrimination lubricating oil thermal aging Raman detection method described above, the device comprising:

[0088] Raman spectroscopy acquisition module 1 is used to acquire the original Raman spectral data of the lubricating oil to be tested, and to cut the original Raman spectral data into characteristic regions based on preset characteristic groups of lubricating oil thermal aging to obtain cut spectral data. The cut spectral data is then normalized to obtain the first preprocessed spectrum.

[0089] The preprocessing operation module 2 is used to perform combined preprocessing on the first preprocessed spectrum using preset row preprocessing operators and column preprocessing operators to obtain a target feature data sequence; wherein, the row preprocessing operator corrects the scattering shift of the first preprocessed spectrum to obtain intermediate corrected data, and the column preprocessing operator performs weight allocation and feature scaling on the intermediate corrected data according to preset variances of each wavenumber signal.

[0090] The result output module 3 is used to input the target feature data sequence into a pre-established lubricating oil thermal aging detection model based on Fisher discrimination, and output the thermal aging level of the lubricating oil to be tested.

[0091] Each of the above modules is used to perform the respective steps in the above combined preprocessing enhanced Fisher discrimination lubricating oil thermal aging Raman detection method. The specific implementation method is as described in the above method embodiment, and will not be repeated here.

[0092] like Figure 6 As shown, the present invention also provides a computer device, which may be a server, and its internal structure may be as follows: Figure 6As shown, the computer device includes a processor, memory, network interface, and database connected via a system bus. The processor provides computational and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system, computer programs, and database. The internal memory provides the environment for the operation of the operating system and computer programs in the non-volatile storage media. The database stores all data required for the combined pretreatment-enhanced Fisher discrimination method for Raman detection of lubricating oil thermal aging. The network interface is used for communication with external terminals via a network connection. The computer program is executed by the processor to implement the combined pretreatment-enhanced Fisher discrimination method for Raman detection of lubricating oil thermal aging.

[0093] Those skilled in the art will understand that Figure 6 The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the computer equipment on which the present application is applied.

[0094] An embodiment of this application also provides a computer-readable storage medium storing a computer program thereon, which, when executed by a processor, implements any of the above-described combined preprocessing enhanced Fisher discrimination lubricating oil thermal aging Raman detection methods.

[0095] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by hardware related to computer program instructions. The computer program can be stored in a non-volatile computer-readable storage medium, and when executed, it can include the processes of the above method embodiments. Any references to memory, storage, databases, or other media provided in this application and used in the embodiments can include non-volatile and / or volatile memory. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory can include random access memory (RAM), such as dynamic RAM (used as main storage) or static RAM (commonly used as cache memory). By way of illustration and not limitation, RAM has various forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDR SDRAM), and Rambus DRAM (RDRAM).

[0096] It should be noted that, in this document, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, apparatus, article, or method that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, apparatus, article, or method. Unless otherwise specified, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, apparatus, article, or method that includes that element.

[0097] The above description is merely a preferred embodiment of the present invention and does not limit the patent scope of the present invention. Any equivalent structural or procedural transformations made based on the content of the present invention's specification and drawings, or direct or indirect applications in other related technical fields, are similarly included within the patent protection scope of the present invention.

Claims

1. A combined pretreatment-enhanced Fisher's discrimination method for the Raman detection of lubricating oil thermal aging, characterized in that, include: The raw Raman spectral data of the lubricating oil to be tested is obtained, and the characteristic range of the raw Raman spectral data is clipped based on the preset characteristic groups of the lubricating oil thermal aging to obtain clipped spectral data. The clipped spectral data is then normalized to obtain the first preprocessed spectrum. The first preprocessed spectrum is preprocessed by a combination of preset row preprocessing operators and column preprocessing operators to obtain a target feature data sequence; wherein, the row preprocessing operator corrects the scattering shift of the first preprocessed spectrum to obtain intermediate corrected data, and the column preprocessing operator performs weight allocation and feature scaling on the intermediate corrected data according to the preset variance of each wavenumber signal. The target feature data sequence is input into a pre-established lubricating oil thermal aging detection model based on Fisher discrimination, and the thermal aging level of the lubricating oil to be tested is output.

2. The combined pretreatment enhanced Fisher discrimination method for Raman detection of lubricating oil thermal aging according to claim 1, characterized in that, The steps of acquiring the raw Raman spectral data of the lubricating oil to be tested, cropping the raw Raman spectral data into a characteristic region based on preset characteristic groups of lubricating oil thermal aging to obtain cropped spectral data, and normalizing the cropped spectral data to obtain a first preprocessed spectrum specifically include: The optical signal of the lubricating oil sample to be tested is acquired by an excitation source, and the detected optical signal is converted into an electrical digital signal by a spectrometer to obtain the original Raman spectral data. According to the preset clipping rules, the original Raman spectral data is subjected to feature region clipping processing, retaining the target feature bands corresponding to the thermal aging feature groups, and removing irrelevant band data other than the target feature bands to obtain the clipped spectral data. The cropped spectral data is subjected to full-range area normalization processing, and the total spectral signal intensity of all target feature bands contained in the cropped spectral data is uniformly normalized to 1 to obtain the first preprocessed spectrum.

3. The combined pretreatment-enhanced Fisher discrimination method for Raman detection of lubricating oil thermal aging according to claim 2, characterized in that, The target characteristic band includes 2800-3000 cm. -1 CH stretching vibration zone, 1650-1750cm -1 The carbonyl stretching vibration region and 3200-3600 cm -1 The hydroxyl stretching vibration region.

4. The combined pretreatment-enhanced Fisher discrimination method for Raman detection of lubricating oil thermal aging according to claim 1, characterized in that, The step of performing combined preprocessing of the first preprocessed spectrum using preset row and column preprocessing operators to obtain the target feature data sequence includes: A preset row preprocessing operator is invoked to mathematically correct the scattering shift of the first preprocessed spectrum along the sample dimension to obtain intermediate corrected data; wherein, the row preprocessing operator is a multivariate scattering correction or a standard normal variable transformation, used to perform sample-by-sample full-spectrum scattering interference correction processing; A preset column preprocessing operator is invoked to perform weight allocation and feature scaling on the intermediate corrected data along the wavenumber dimension based on the signal variance, thereby obtaining the target feature data sequence; wherein, the column preprocessing operator is a variance scaling operator, used to perform wavenumber-by-wavenumber feature intensity scaling processing along the wavenumber dimension.

5. The combined pretreatment enhanced Fisher discrimination method for Raman detection of lubricating oil thermal aging according to claim 1, characterized in that, The step of inputting the target feature data sequence into a pre-established lubricating oil thermal aging detection model based on Fisher discrimination, and outputting the thermal aging level of the lubricating oil to be tested, specifically includes: A two-stage screening based on dual-objective parameters was performed to build an offline lubricating oil thermal aging detection model based on Fisher discrimination. The target feature data sequence is input into the lubricating oil thermal aging detection model for classification and inference, and the thermal aging level of the lubricating oil to be tested is output. The quantitative physicochemical indicators of the lubricating oil to be tested are derived using partial least squares regression correlation mapping rules, and the data is displayed and stored in the database.

6. The combined pretreatment enhanced Fisher discrimination method for Raman detection of lubricating oil thermal aging according to claim 5, characterized in that, The steps of performing two-level screening based on dual-objective parameters and constructing an offline lubricating oil thermal aging detection model based on Fisher discrimination specifically include: Multiple initial models were obtained by using different preprocessing operators and spectral sets of lubricating oil samples with calibrated aging levels, combined with the Fisher discrimination algorithm. Cross-validation of the multiple initial models is performed using a test set, and the first set of candidate models whose full aging level discrimination accuracy meets the preset conditions is retained as the first-level screening. For the candidate models in the first candidate model set, the variance contribution of their Fisher discriminant function is extracted as a weight. The average weighted distance between the centroid of the unaged sample and the centroids of the other aging level samples, as well as the average weighted distance between the centroids of adjacent aging level samples, are calculated as the second level of screening. The candidate model with the maximum average weighted distance was selected as the final lubricating oil thermal aging detection model.

7. The combined pretreatment enhanced Fisher discrimination method for Raman detection of lubricating oil thermal aging according to claim 6, characterized in that, In the second-level screening, the formula for calculating the average weighted distance D between the centroids of adjacent aging level samples is as follows: Where n is the number of groups of adjacent aging levels; and These are the mean scores of the first classification of the aging level samples at level i and level i+1, respectively. and These are the mean scores of the second classification of the aging level samples at level i and level i+1, respectively. and These represent the percentage of variance contribution from the first and second classification discriminant functions, respectively.

8. The combined pretreatment enhanced Fisher discrimination method for Raman detection of lubricating oil thermal aging according to claim 5, characterized in that, The steps of deriving the quantitative physicochemical indicators of the lubricating oil to be tested using partial least squares regression correlation mapping rules, and then displaying and storing the data, specifically include: Based on the pre-stored partial least squares regression correlation mapping rule between the actual acid value, kinematic viscosity change rate and Fisher discrimination score, the classification inference results of the lubricating oil to be tested are transformed into continuous quantitative physicochemical index prediction values. The thermal aging level and the predicted values ​​of the quantified physicochemical indicators are displayed locally on the screen, and the test data is remotely uploaded to the management platform through the communication unit. The data storage unit performs local or cloud storage to form a historical test record ledger.

9. A combined pretreatment enhanced Fisher discrimination lubricating oil thermal aging Raman detection device, based on the combined pretreatment enhanced Fisher discrimination lubricating oil thermal aging Raman detection method according to any one of claims 1 to 8, characterized in that, The device includes: The Raman spectroscopy acquisition module is used to acquire the raw Raman spectral data of the lubricating oil to be tested, and to cut the raw Raman spectral data into characteristic regions based on preset characteristic groups of lubricating oil thermal aging to obtain cut spectral data. The cut spectral data is then normalized to obtain the first preprocessed spectrum. The preprocessing module is used to perform combined preprocessing on the first preprocessed spectrum using preset row preprocessing operators and column preprocessing operators to obtain a target feature data sequence; wherein, the row preprocessing operator corrects the scattering shift of the first preprocessed spectrum to obtain intermediate corrected data, and the column preprocessing operator performs weight allocation and feature scaling on the intermediate corrected data according to preset variances of each wavenumber signal. The result output module is used to input the target feature data sequence into a pre-established lubricating oil thermal aging detection model based on Fisher discrimination, and output the thermal aging level of the lubricating oil to be tested.

10. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the method according to any one of claims 1 to 8.

Images