Calculation method of oil performance parameters and device for online monitoring of lubricating oil

Abstract

The present invention belongs to the technical field of lubricating oil monitoring, and specifically discloses a method for calculating oil performance parameters and a device for online monitoring of lubricating oil. The method comprises: collecting near-infrared spectral data of the oil, and calculating singular values ​​based on the spectral data; selecting representative characteristic data based on the singular values; and establishing a model for calculating the oil performance parameters based on the singular values. The advantages of the present invention are that it does not destroy the sample, does not consume chemical reagents, is simple and convenient, and is also efficient and low-cost. It is very suitable for rapid on-site analysis and also provides an experimental basis for non-destructive analysis.

Description

Technical Field

[0001] The invention relates to the technical field of lubricating oil monitoring, and in particular to a method for calculating oil performance parameters and a device for online monitoring of lubricating oil. Background Art

[0002] Lubricating oil can be called the "blood" of industrial equipment, and lubricating oil monitoring is the most effective way to monitor the lubrication and wear status of large-scale mechanical equipment and diagnose faults. It is of great significance to ensure the safe and reliable operation of mechanical equipment, avoid unplanned downtime, and achieve energy saving and efficiency improvement.

[0003] Existing oil monitoring has been mainly offline for many years, which is far from meeting the needs of long-term continuous monitoring of modern equipment. Therefore, online oil monitoring technology has become one of the important development trends of equipment lubrication wear failure diagnosis technology. Online lubricating oil monitoring eliminates human uncertainty factors, and sampling and testing can be carried out at the same time. As long as the installation location is selected appropriately, the problem of sampling representativeness can be basically solved.

[0004] With the development of the times, the demand for online monitoring of oil using infrared spectroscopy technology is becoming more and more urgent. In the field of online monitoring of lubricating oil, attempts have been made to use mid-infrared light for spectral analysis to achieve online monitoring. However, mid-infrared spectroscopy equipment is difficult to miniaturize, and samples still need to be prepared, making it difficult to fully achieve the goal of "online" monitoring. Summary of the invention

[0005] In order to solve the above defects and fully realize online monitoring, the present invention first proposes a method for calculating oil performance parameters, including:

[0006] collecting near infrared spectral data of the oil, and calculating singular values ​​based on the spectral data;

[0007] Selecting representative feature data based on the singular values;

[0008] A model for calculating the oil performance parameters is established based on the singular values.

[0009] In the above method, the collecting of near infrared spectral data of the oil and calculating singular values ​​based on the spectral data include:

[0010] S1. Collect m groups of spectral data, where m is a positive integer, wherein each group of spectral data is a group of n-dimensional vectors, and n is used to indicate the number of frequency points of the near-infrared light used for spectral analysis;

[0011] S2. Establish an m×n X matrix, and calculate a singular value matrix of the X matrix.

[0012] In the above method, S2 further includes:

[0013] S21, calculating the central vector of each row of data in the X matrix;

[0014] S22, subtract the center vector from each element in the X matrix row by row, to obtain matrix;

[0015] S23, in the Calculate the singular values ​​of the matrix based on the matrix.

[0016] In the above method, the central vector is the arithmetic mean of each row of data.

[0017] In the above method, the step of selecting representative feature data based on the singular values ​​includes:

[0018] S3, calculating the ratio of each singular value in the singular value matrix to the sum of all singular values;

[0019] S4. Determine the spectral data corresponding to the k singular values ​​with the largest ratios as representative characteristic data.

[0020] In the above method, in S4, the k singular values ​​are determined by the following method:

[0021] pre-specify the value of k; or

[0022] The ratios are arranged in descending order, and the ratios are accumulated one by one. When the accumulated value exceeds a predetermined value, the number of accumulated ratios is determined to be k.

[0023] In the above method, the predetermined value is 0.8-0.95.

[0024] In the above method, the model for calculating the oil performance parameter based on the singular value is established, including:

[0025] S5. Establishing a corrected model for calculating the performance parameters based on the selected k feature vectors.

[0026] In the above method, the least square method with penalty term is used to establish a model for calculating the performance parameters.

[0027] In the above method, the oil performance parameters include: viscosity, usage time, acidity, and element content of the oil.

[0028] In the above method, the spectral data includes one or more of transmission data, reflection data, and diffuse reflection data of near-infrared light.

[0029] Based on the same inventive concept, the present invention also proposes a device for online monitoring of oil performance parameters, comprising:

[0030] A near-infrared spectral analysis unit, used for acquiring spectral data of the oil in the near-infrared band;

[0031] A performance parameter calculation unit is used to implement the calculation method according to any one of claims 1 to 11 based on the spectral data.

[0032] In the above device, the spectral data includes data related to the viscosity, usage time, acidity, and element content of the oil.

[0033] The present invention also proposes a computer-readable storage medium, in which at least one computer instruction is stored. The at least one instruction is loaded and executed by a processor to implement the above method.

[0034] The present invention also proposes a computer program product, which includes computer instructions. When the computer instructions are executed, the above method is implemented.

[0035] Compared with the prior art, the present invention discloses a method of first establishing a mathematical correction model (for example, establishing a corresponding relationship between infrared spectrum absorbance and lubricant performance parameters), and then using a chemometric method (including spectrum preprocessing, wavelength variable selection, underfitting / overfitting judgment and removal of abnormal samples, etc.) to quantitatively analyze lubricant performance parameters (such as viscosity, acidity element content and other key physical and chemical indicators). This method does not destroy the sample, does not consume chemical reagents, is simple and convenient, and is also efficient and low-cost. It is very suitable for on-site rapid analysis and also provides an experimental basis for non-destructive analysis. BRIEF DESCRIPTION OF THE DRAWINGS

[0036] Figure 1 is a flow chart provided according to some embodiments of the present invention;

[0037] Figure 2 is a characteristic value histogram provided according to some embodiments of the present invention;

[0038] Figure 3 is a prediction fitting curve diagram of time parameters provided according to some embodiments of the present invention;

[0039] Figure 4 is a predicted fitting curve diagram of viscosity parameters provided according to some embodiments of the present invention;

[0040] Figure 5 is a predicted fitting curve diagram of phosphorus content provided according to some embodiments of the present invention. DETAILED DESCRIPTION

[0041] First of all, it should be stated that although the terms "first", "second", etc. may be used in this article to describe various features, these features should not be limited by these terms. These terms are used only to distinguish and cannot be understood as indicating or implying relative importance or ranking. For example, without departing from the scope of the exemplary embodiments, a first feature can be called a second feature, and similarly, a second feature can be called a first feature.

[0042] It should also be stated that the steps in the methods and processes of the present invention are numbered for ease of reference, rather than to limit the order of precedence. If there is an order between the steps, the text description shall prevail.

[0043] As mentioned in the background technology, mid-infrared spectroscopy has been tried to be applied to the online monitoring of lubricating oil. However, there are two main reasons why near-infrared spectroscopy has not yet been applied to the online monitoring technology of lubricating oil: 1) Near-infrared spectra are easily affected by temperature and hydrogen bonds, and it is impossible to use traditional spectral analysis methods to identify molecular structures; 2) Near-infrared spectra contain a large number of overlapping broad bands (meaning that there is almost no "fingerprinting"), that is, the interference between near-infrared bands overlaps, and it is impossible to quantitatively analyze the oil based on the single-wavelength Lambert-Beer law.

[0044] The technical concept of the present invention is to use near-infrared light to illuminate the oil to be tested, and the oil will reflect, transmit and diffusely reflect the infrared light. The spectrum of the reflected, transmitted and diffusely reflected near-infrared light is analyzed to establish a relationship function between the spectrum and the performance of the oil (for example, viscosity, acidity, etc.), so that the performance parameters of the oil can be quantitatively calculated.

[0045] To make the objects, embodiments and advantages of the present invention more clear, the embodiments of the present invention will be described in further detail below with reference to the accompanying drawings.

[0046] When near-infrared light is used to analyze oil, the dimension of the data is relatively large. The range of near-infrared spectrum currently defined is 780nm-2526nm. If the spectrum analysis is performed with a wavelength interval of 1nm, a set of near-infrared spectrum data may reach 1747 dimensions. Even with computer-assisted calculations, it is difficult to directly extract key and effective information from the feature vector. And for spectral detection, the interval period of data collection will not be very short, so there will be a situation where the number of data is much smaller than the number of features. Therefore, in order to effectively utilize the near-infrared spectrum data of oil, the "dimensionality reduction" method can be used to select several spectral data that best represent all the data from a large number of data, that is, select the spectral data that carries the most performance (parameter) information as the object of concern for the calculation model.

[0047] The present invention collects as much near-infrared spectral data of oil as possible, first calculates singular values ​​based on the spectral data, for example, the singular value decomposition (SVD) method can be used to calculate the singular values ​​of a two-dimensional matrix. Then, representative feature data is selected based on the singular values ​​and a model for calculating the oil performance parameters is established.

[0048] Specifically, Figure 1 Shown is a flow chart of calculating performance parameters according to some embodiments of the present invention.

[0049] In S1, first, m groups of the spectral data are collected, where m is a positive integer. It can also be understood as how many times the spectral data have been collected in total. In practical applications, since it is a long-term process for oil to be contaminated or deteriorate during use, the sampling frequency of the oil will not be very frequent, that is, m is usually a small number, for example, a single digit. Each group of the spectral data is a set of n-dimensional vectors, where n is used to indicate the number of frequency points of the near-infrared light used for spectral analysis, and n may be a value of the order of thousands. In step S2, the n-dimensional spectral data will be reduced to a single digit.

[0050] In S2, an m×n matrix X is established, and a singular value matrix of the matrix X is calculated. Specifically, the SVD method is used to obtain further six accounts of singular values.

[0051] S21, calculate the central vector of each row of data in the X matrix. The central vector refers to the arithmetic mean of each row of data.

[0052] S22, subtract the center vector from each element in the X matrix row by row, to obtain Matrix. Since the near-infrared spectrum data of oil fluctuates relatively little during the entire life cycle of the oil, the difference in the light absorption rate at the corresponding wavelength length is very small. If the original data is used directly for modeling, a small measurement error will have a huge impact on the prediction results. Here, the difference between each element in the X matrix and the average value is analyzed, which will be convenient for drawing data charts later.

[0053] S23, in the Calculate the singular values ​​of the matrix based on the matrix.

[0054] In S3, the ratio of each singular value in the singular value matrix to the sum of all singular values ​​is calculated; the higher the ratio of a single singular value to the total singular values, the more spectral information the corresponding spectral data carries.

[0055] In S4, the spectral data corresponding to the k singular values ​​(k<n) with the largest ratio are determined as representative feature data. In order to reduce the dimension of the data, it is necessary to make some choices on the data. By calculating the ratio of the singular value to the sum of all singular values, the spectral data that can represent most of the information in the spectral data is selected, thereby achieving the effect of data "dimension reduction".

[0056] Specifically, the k singular values ​​may be predetermined to be the first k larger singular values, or may be determined by the following method:

[0057] The ratios are arranged in descending order, and the ratios are accumulated one by one. When the accumulated value exceeds a predetermined value, the number of accumulated ratios is determined as k. The predetermined value is used to indicate that the spectral data corresponding to the singular value can represent (or contain) most of the information in all spectral data, which can generally be 0.8 to 0.95.

[0058] The following is an example to illustrate the process of data selection.

[0059] For example, suppose the singular values ​​(arranged in descending order) obtained by the SVD method are (5, 1, 0.1, 0.1, 0.01), where the first eigenvalue (5) accounts for the total number of all eigenvalues ​​(singular values) as follows: 5 / (5+1+0.1+0.1+0.01)=0.805; the first eigenvalue (5) plus the second eigenvalue (1) together account for: (5+1) / (5+1+0.1+0.1+0.01)=0.966. If the preset predetermined value is 0.95, the spectral data corresponding to the first and second eigenvalues ​​can be taken as the result of dimensionality reduction (principal component). In other words, after the previous 5-dimensional data is selected, only 2-dimensional data that can represent the main information is retained.

[0060] In the above example, the cumulative contribution rate of the first two principal components is 96.6%, that is, 96.6% of the spectral information contained in the original spectral data is concentrated in the first two principal components, and the remaining three spectral data can be discarded in exchange for a simple data dimension (for easy calculation).

[0061] In S5, a corrected model for calculating the performance parameters is established based on the selected k feature vectors. For example, a least square method with a penalty term can be used to establish a model for calculating the performance parameters. According to the model, relatively accurate parameters can be calculated when monitoring oil performance.

[0062] In the following, we collect m near-infrared spectral data X={x1,x2,…,x i …,x m} as an example to fully illustrate the above method.

[0063] Spectral data X = {x1, x2, ..., x i …,x m Any element x in i The dimension is n, that is, x i =[x i1 ,x i2 ,…,x in ]. That is to say, for the oil viscosity, the near-infrared spectrum is sampled at n frequency points in total, and the sampling is m times in total. Therefore, X is an m×n matrix.

[0064] First, calculate the center vector of each row of data in the X matrix, and subtract the center point vector from each data to get matrix:

[0065]

[0066]

[0067] For this Matrix, using the singular value decomposition (SVD, singular value decomposition) method, Singular value decomposition of the matrix:

[0068] SVD(X)=UΣV H

[0069] Among them, U is an m*m orthogonal matrix, V is an n*n orthogonal matrix, Σ is an m*n singular value matrix with only the main diagonal containing non-zero values. The relationship between eigenvalues ​​and singular values ​​is square.

[0070] Then, the feature dimension k after dimensionality reduction is selected according to the ratio of the eigenvalue to the sum of all eigenvalues, where k<n. For specific examples, please refer to the description of S4 above, which will not be repeated here.

[0071] Finally, for each set of acquired or new spectral data X, subtract the center vector and then multiply it by the matrix [v1,v2,…,v k ] is the feature information after dimensionality reduction. For the features after dimensionality reduction, the least squares with penalty term can be used to calculate.

[0072] For example, suppose the original spectral data are:

[0073] x1=[1.1,1.2,2.3,2.6,1.9]

[0074] x2=[1.0,1.1,2.3,2.7,2.1]

[0075] x3=[1.2,1.2,2.4,2.5,2.0]

[0076] Combine the above original data x1 to x3 into the original data matrix X:

[0077]

[0078] Then, the central vector of each row of data in X is:

[0079]

[0080] X and x mean Subtract matrix:

[0081]

[0082] According to the formula of singular value decomposition method, SVD(X)=UΣV H After that, we get the Σ matrix and V H matrix:

[0083]

[0084]

[0085] From the above V H Select the singular values ​​with a ratio of more than 95% in the matrix, that is, the eigenvalues ​​with a ratio of more than 0.95*0.95=90.25%, that is, the first two rows, and then transpose them to obtain the reduced dimension V trancated matrix:

[0086]

[0087] Furthermore, after dimensionality reduction The matrix is:

[0088]

[0089] In the case where multiple performance parameters need to be calculated, such as viscosity, service life, and metal content, these properties can also be established as a matrix Y = [y1, y2, ..., y p ] T , where p indicates the serial number of the performance parameter to be predicted, for example, in this embodiment, p = 3. We only need to use the above-mentioned principal component analysis method (PCA) to find a low-dimensional singular value matrix that can contain the maximum information in the X matrix, and then use a fitting algorithm such as Ridge Regression to build a model.

[0090] Want to pass To find the association with the performance parameter matrix Y, the weights can be calculated using the least squares method make The constant term is incorporated into the matrix to obtain:

[0091]

[0092]

[0093] According to the least squares formula,

[0094] w′=(X′ T X′) -1 X′ T Y

[0095] In order to stabilize the data, use the least squares with penalty term, for example, set λ = 0.001, or try different values ​​of λ in a wide range to make y est and Y error is minimal.

[0096] According to the least squares formula w′=(X′ T X′+λI) -1 X′ T Y, where the parameter matrix takes the value of Y H =[1,2,3], we can get:

[0097]

[0098] The predicted value of oil viscosity is calculated:

[0099]

[0100] Figure 2 It is a histogram of eigenvalues ​​provided according to some embodiments of the present invention. It is the size of the first 20 eigenvalues ​​after the eigenvalues ​​are sorted from large to small after PCA decomposition. Since the eigenvalues ​​vary greatly, logarithmic coordinates are displayed here. It can be seen from the figure that in this embodiment, the first two eigenvalues ​​can contain more than 90% of the information of all the spectrum data. In other words, Figure 2 In the illustrated embodiment, the data dimension can be reduced from 20 dimensions to 2 dimensions.

[0101] Figure 3-Figure 5 middle, Figure 3 is a predicted fitting curve diagram of time parameters provided according to some embodiments of the present invention. Figure 4 is a predicted fitting curve diagram of viscosity parameters provided according to some embodiments of the present invention. Figure 5is a prediction fitting curve diagram of phosphorus content provided according to some embodiments of the present invention. In the figure, the solid line is a linear function curve of y=x, and the dots represent the training data used to train the calculation model. The × symbol is the test data used to test whether the model is accurate. It can be seen from the figure that the fitting effect of the data is quite good.

[0102] The present invention further proposes a computer-readable storage medium, in which at least one computer instruction is stored. The at least one instruction is loaded and executed by a processor to implement the above-mentioned monitoring method.

[0103] The present invention also proposes a computer program product, which includes computer instructions. When the computer instructions are executed, the above-mentioned monitoring method is implemented.

[0104] In general, the present invention performs PCA (Principal Component Analysis), i.e., principal component analysis, on the samples to reduce the dimension of the data and transform the original variables so that a few new variables can express the data characteristics of the original variables as much as possible without losing information. Then, a model of the relevant performance parameters of the lubricating oil and the absorbance of the near-infrared spectrum is established by the partial least squares method. With the miniaturization of spectral instruments, the improvement of measurement accessories, the continuous innovation of experimental technology, and the development and popularization of chemometric methods, near-infrared spectroscopy analysis technology will play an increasingly important role in the analysis and monitoring of lubricating oils.

[0105] A large number of specific details are described in this specification. However, it is understood that embodiments of the present invention can be practiced without these specific details. In some instances, well-known methods, structures and techniques are not shown in detail so as not to obscure the understanding of this specification.

[0106] Similarly, it should be understood that in order to streamline the present invention and aid in understanding one or more of the various inventive aspects, in the above description of exemplary embodiments of the present invention, various features of the present invention are sometimes grouped together into a single embodiment, figure, or description thereof. However, this disclosed method should not be interpreted as reflecting the intention that the claimed invention requires more features than those expressly recited in each claim. Rather, as reflected in the claims, inventive aspects lie in less than all of the features of the individual embodiments previously disclosed. Therefore, the claims that follow the detailed description are hereby expressly incorporated into the detailed description, with each claim itself serving as a separate embodiment of the present invention.

[0107] Those skilled in the art will appreciate that the modules in the devices in the embodiments may be adaptively changed and arranged in one or more devices different from the embodiments. The modules or units or components in the embodiments may be combined into one module or unit or component, and in addition they may be divided into a plurality of submodules or subunits or subcomponents. Except that at least some of such features and / or processes or units are mutually exclusive, all features disclosed in this specification (including the accompanying claims, abstracts and drawings) and all processes or units of any method or device disclosed in this manner may be combined in any combination. Unless otherwise expressly stated, each feature disclosed in this specification (including the accompanying claims, abstracts and drawings) may be replaced by an alternative feature providing the same, equivalent or similar purpose.

[0108] In addition, those skilled in the art will appreciate that, although some embodiments described herein include certain features included in other embodiments but not other features, the combination of features of different embodiments is meant to be within the scope of the present invention and form different embodiments. For example, in the claims, any one of the claimed embodiments may be used in any combination.

Claims

1. A method for calculating oil performance parameters, characterized in that: include: Collecting near infrared spectral data of the oil, and calculating singular values ​​based on the spectral data, specifically includes: S1. Collect m groups of spectral data, where m is a positive integer, wherein each group of spectral data is a group of n-dimensional vectors, and n is used to indicate the number of frequency points of the near-infrared light used for spectral analysis; S2, establish an m×n X matrix, and calculate the singular value matrix of the X matrix; Selecting representative feature data based on the singular values ​​specifically includes: S3, calculating the ratio of each singular value in the singular value matrix to the sum of all singular values; S4, determining the spectral data corresponding to the k singular values ​​with the largest ratio as representative characteristic data, wherein the k singular values ​​are determined by the following method: pre-specify the value of k; or Arrange the ratios in descending order, accumulate the ratios one by one, and when the accumulated value exceeds a predetermined value, the number of accumulated ratios is determined to be k; A model for calculating the oil performance parameters is established based on the singular values.

2. The calculation method according to claim 1, characterized in that: S2 further includes: S21, calculating the central vector of each row of data in the X matrix; S22, subtracting the central vector from each element in the X matrix row by row to obtain a matrix; S23. Calculate a singular value matrix based on the matrix.

3. The calculation method according to claim 2, characterized in that: The central vector is the arithmetic mean of each row of data.

4. The calculation method according to claim 1, characterized in that: The predetermined value is 0.8~0.

95.

5. The calculation method according to claim 1, characterized in that: The model for calculating the oil performance parameter based on the singular value includes: S5. Establishing a corrected model for calculating the performance parameters based on the selected k feature data.

6. The calculation method according to claim 5, characterized in that: A least square method with a penalty term is used to establish a model for calculating the performance parameters.

7. The calculation method according to any one of claims 1 to 6, characterized in that: The oil performance parameters include: viscosity, service life, acidity, and element content of the oil.

8. The calculation method according to any one of claims 1 to 6, characterized in that: The spectral data includes one or more of transmission data, reflection data, and diffuse reflection data of near-infrared light.

9. A device for online monitoring of oil performance parameters, characterized in that: include: A near-infrared spectral analysis unit, used for acquiring spectral data of the oil in the near-infrared band; A performance parameter calculation unit is used to implement the calculation method according to any one of claims 1 to 8 based on the spectral data.

10. The device according to claim 9, characterized in that The spectral data includes data related to the viscosity, usage time, acidity, and element content of the oil.

11. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores at least one computer instruction, and the at least one instruction is loaded and executed by a processor to implement the method according to any one of claims 1 to 8.

12. A computer program product, characterized in that The computer program product comprises computer instructions, and when the computer instructions are executed, the method according to any one of claims 1 to 8 is implemented.

Images