阅读说明：本技术 基于图像辅助原子发射光谱的油液元素检测方法及装置 (Oil element detection method and device based on image-assisted atomic emission spectroscopy ) 是由 傅骁 段发阶 蒋佳佳 张聪 鲍瑞伽 马凌 刘文正 郑好 刘昌文 于 2019-11-18 设计创作，主要内容包括：本发明公开一种基于图像辅助原子发射光谱的油液元素检测方法及装置,方法包括以下步骤：(1)利用摄像机和转盘电极原子发射光谱仪,同步采集油液样品的电弧等离子体图像和原子发射光谱,利用元素谱线估算等离子体状态参数,建立等离子体图像与状态参数的内在联系；(2)基于电弧等离子体图像和原子发射光谱数据,运用偏最小二乘算法(PLS),建立图像-光谱训练模型和测试模型,先通过训练模型求解得到补偿比例系数,再利用测试模型求解得到补偿后的谱线强度；(3)利用补偿后的谱线强度,结合定量分析算法,实现油液元素含量检测。该方法能够在不显著增加成本的情况下,明显提高油液元素现场检测精度,改善元素检测的重复性。(The invention discloses an oil element detection method and device based on image-assisted atomic emission spectroscopy, wherein the method comprises the following steps: (1) synchronously acquiring an arc plasma image and an atomic emission spectrum of the oil liquid sample by using a camera and a turntable electrode atomic emission spectrometer, estimating a plasma state parameter by using an element spectral line, and establishing an internal relation between the plasma image and the state parameter; (2) based on the arc plasma image and the atomic emission spectrum data, an image-spectrum training model and a test model are established by using a Partial Least Squares (PLS) algorithm, a compensation proportionality coefficient is obtained by solving through the training model, and then the compensated spectral line intensity is obtained by solving through the test model; (3) and (4) detecting the content of the oil element by using the compensated spectral line intensity and combining a quantitative analysis algorithm. The method can obviously improve the on-site detection precision of the oil liquid elements and improve the repeatability of element detection without obviously increasing the cost.)

1. The oil element detection method based on the image-assisted turntable electrode atomic emission spectrum is characterized by comprising the following steps of:

(1) synchronously acquiring an arc plasma image and an atomic emission spectrum of the oil liquid sample by using a camera and a turntable electrode atomic emission spectrometer, estimating a plasma state parameter by using an element spectral line, and establishing an internal relation between the plasma image and the state parameter;

(2) based on the arc plasma image and the atomic emission spectrum data, an image-spectrum training model and a test model are established by using a Partial Least Squares (PLS) algorithm, a compensation proportionality coefficient is obtained by solving through the training model, and then the compensated spectral line intensity is obtained by solving through the test model;

(3) and (4) detecting the content of the oil element by using the compensated spectral line intensity and combining a quantitative analysis algorithm.

2. The oil element detection method based on the image-assisted turntable electrode atomic emission spectroscopy as claimed in claim 1, wherein the step (1) comprises the following steps:

(101) through experiments, different combinations of oil standard samples with known element contents are utilized to establish a training set, quantitative analysis is carried out on oil elements, and array arc images and plasma spectrum data are obtained;

(102) the plasma state parameters are estimated by utilizing the multispectral line to average, an independent variable matrix is constructed by using full-pixel data of an arc image, a dependent variable matrix is constructed by using the plasma state parameters, a regression model is established by using PLS (partial least squares), and cross validation is performed by using a leave-one method.

3. The method for detecting oil elements based on image-assisted turntable electrode atomic emission spectroscopy as claimed in claim 2, wherein the plasma state parameters in step (102) include temperature and electron density.

4. The oil element detection method based on the image-assisted turntable electrode atomic emission spectroscopy as claimed in claim 1, wherein the step (2) comprises the following steps:

(201) establishing a training model; using p standard oil liquid samples (with known element content), acquiring q groups of images and spectra for each sample, and forming k-p × q groups of observation data, wherein the image resolution is assumed to be r-m × n; converting an m multiplied by n pixel matrix of an original image into a single-row 1 multiplied by r array, wherein k original images form a k multiplied by r training matrix P; analyzing the training matrix P by using a PLS algorithm to obtain a coefficient matrix R of R multiplied by R, extracting proper first l (l is less than or equal to R) columns to form an R multiplied by l image feature extraction matrix E, and simultaneously extracting first l main components corresponding to the training matrix P to form a k multiplied by l image feature matrix C; line intensity I for a particular wavelength_iCalculating the average intensity of the wavelength spectral line by using the spectra of k groups of training sets

The relative deviation vector I of the wavelength spectral line intensity is recorded as

Establishing regression of the image feature matrix C to the relative deviation vector e by using PLS to obtain a PLS regression coefficient vector B of l multiplied by 1; according to the Schiebe-Lomakin formula, when the self absorption of the spectral line is not considered, the content c of the element corresponding to the spectral line is known, and the proportionality coefficient K of

(202) Establishing a test model; using an oil liquid sample to be detected, and acquiring a group of images and spectral data by single acquisition; converting the m × n pixel matrix of the original image into a single-row 1 × r array P_x(ii) a Extracting image characteristic matrix E and array P_xMultiplying, extracting to obtain an image feature array C with 1 × l main components_x(ii) a Using PLS regression coefficient vector B and image feature array C_xEstimating the relative deviation e of the spectral line intensity_outIs composed of

In the formula, B (j) and C_x(j) Refer to B and C, respectively_xThe jth value of (d); using relative deviation e_outThe corrected line intensity I' is

Finally, the proportionality coefficient K in the training model is utilizedAnd the corrected spectral line intensity I' is calculated to obtain the predicted element content c_xIs composed of

c_x＝KI′ (6)。

5. An oil element on-site detection device based on image-assisted turntable electrode atomic emission spectroscopy is characterized by comprising an electric arc generator, a turntable graphite electrode, a collimating mirror, a beam splitter, a camera, a spectrum collector, an optical fiber, a spectrometer and a computer; the turntable graphite electrode carries oil to be detected into a discharge gap, an arc plasma is generated under the drive of the arc generator, the spectrum emitted by the arc plasma sequentially passes through the collimating lens and the beam splitter to form two light paths with different directions, the light is sequentially split and collected by the spectrum collector and the optical fiber entering spectrometer, the other light path is collected by the camera to obtain image data, and the spectrum data and the image data are simultaneously sent into the computer for analysis and processing.

Technical Field

The invention belongs to the field of spectral analysis, and particularly relates to an oil element on-site detection method and device based on image-assisted turntable electrode atomic emission spectroscopy.

Background

Oil (such as lubricating oil, fuel oil and the like) is blood for ensuring the healthy operation of major equipment, and an oil system has the effect of lifting the weight in large-scale rotating machinery systems such as aero-engines and gas turbines, and directly influences the operation safety of the whole machine. The element detection is the most central content of oil state monitoring, and as oil carries a large number of micro particles generated by the abrasion of mechanical parts, the element composition of the particles can effectively reflect key information such as the abrasion degree, the abrasion position, the fatigue condition and the like of related parts. Through the detection of oil elements, researchers can timely master the running state of the equipment, early warning is carried out before faults occur, and powerful basis is provided for preventing major failure, reducing maintenance cost, improving equipment availability and the like.

Among a plurality of oil detection technologies, the turntable electrode atomic emission spectroscopy technology is a relatively traditional spectral analysis technology with more stable and reliable performance, and is a mainstream technology widely adopted by field detection of oil elements at present due to the characteristics of strong environmental adaptability, no need of sample preparation, reliable performance, simple operation and the like. Nevertheless, the technology still faces many problems to be solved urgently, especially the detection process is influenced by factors such as arc excitation energy fluctuation and environmental disturbance, and the measurement repeatability cannot be reliably guaranteed.

Disclosure of Invention

The invention aims to overcome the defects in the prior art and provides an oil element detection method and device based on image-assisted turntable electrode atomic emission spectroscopy; a camera monitoring method is adopted, an arc image and an emission spectrum are collected at the same time, a Partial Least Squares (PLS) algorithm is fused, a measurement result compensation model is constructed, the on-site detection precision of oil elements is improved, and the repeatability of element detection is improved. Based on the method, the high-precision oil element on-site detection device is provided, and can be popularized and used for on-site detection of oil elements of various equipment systems including aircraft engines, gas turbines and the like.

The purpose of the invention is realized by the following technical scheme:

the oil element detection method based on the image-assisted turntable electrode atomic emission spectrum comprises the following steps:

(1) synchronously acquiring an arc plasma image and an atomic emission spectrum of the oil liquid sample by using a camera and a turntable electrode atomic emission spectrometer, estimating a plasma state parameter by using an element spectral line, and establishing an internal relation between the plasma image and the state parameter;

(2) based on the arc plasma image and the atomic emission spectrum data, an image-spectrum training model and a test model are established by using a Partial Least Squares (PLS) algorithm, a compensation proportionality coefficient is obtained by solving through the training model, and then the compensated spectral line intensity is obtained by solving through the test model;

(3) and (4) detecting the content of the oil element by using the compensated spectral line intensity and combining a quantitative analysis algorithm.

Further, the step (1) comprises the following steps:

(101) through experiments, different combinations of oil standard samples with known element contents are utilized to establish a training set, quantitative analysis is carried out on oil elements, and array arc images and plasma spectrum data are obtained;

(102) the plasma state parameters are estimated by utilizing the multispectral line to average, an independent variable matrix is constructed by using full-pixel data of an arc image, a dependent variable matrix is constructed by using the plasma state parameters, a regression model is established by using PLS (partial least squares), and cross validation is performed by using a leave-one method.

Further, the plasma state parameters in step (102) include temperature and electron density.

Further, the step (2) comprises the following steps:

(201) establishing a training model; using p standard oil liquid samples (with known element content), acquiring q groups of images and spectra for each sample, and forming k-p × q groups of observation data, wherein the image resolution is assumed to be r-m × n; converting an m multiplied by n pixel matrix of an original image into a single-row 1 multiplied by r array, wherein k original images form a k multiplied by r training matrix P; analyzing the training matrix P by using a PLS algorithm to obtain a coefficient matrix R of R multiplied by R, extracting proper first l (l is less than or equal to R) columns to form an R multiplied by l image feature extraction matrix E, and simultaneously extracting first l main components corresponding to the training matrix P to form a k multiplied by l image feature matrix C; line intensity I for a particular wavelength_iCalculating the average intensity of the wavelength spectral line by using the spectra of k groups of training sets

Is composed of

The relative deviation vector I of the wavelength spectral line intensity is recorded as

Establishing regression of the image feature matrix C to the relative deviation vector e by using PLS to obtain a PLS regression coefficient vector B of l multiplied by 1; according to the Schiebe-Lomakin formula, when the self absorption of the spectral line is not considered, the content c of the element corresponding to the spectral line is known, and the proportionality coefficient K of

(202) Establishing a test model; using an oil liquid sample to be detected, and acquiring a group of images and spectral data by single acquisition; converting the m × n pixel matrix of the original image into a single-row 1 × r array P_x(ii) a Extracting image characteristic matrix E and array P_xMultiplying, extracting to obtain an image feature array C with 1 × l main components_x(ii) a Using PLS regression coefficientsQuantity B and image feature array C_xEstimating the relative deviation e of the spectral line intensity_outIs composed of

In the formula, B (j) and C_x(j) Refer to B and C, respectively_xThe jth value of (d); using relative deviation e_outThe corrected line intensity I' is

Finally, calculating to obtain the content c of the predicted element by utilizing the proportionality coefficient K and the corrected spectral line intensity I' in the training model_xIs composed of

c_x＝KI′ (6)。

The invention also provides another technical scheme as follows: an oil element on-site detection device based on image-assisted turntable electrode atomic emission spectroscopy comprises an arc generator, a turntable graphite electrode, a collimating mirror, a beam splitter, a camera, a spectrum collector, an optical fiber, a spectrometer and a computer; the turntable graphite electrode carries oil to be detected into a discharge gap, an arc plasma is generated under the drive of the arc generator, the spectrum emitted by the arc plasma sequentially passes through the collimating lens and the beam splitter to form two light paths with different directions, the light is sequentially split and collected by the spectrum collector and the optical fiber entering spectrometer, the other light path is collected by the camera to obtain image data, and the spectrum data and the image data are simultaneously sent into the computer for analysis and processing.

Compared with the prior art, the technical scheme of the invention has the following beneficial effects:

(1) compared with the traditional oil element detection method, the oil element field detection method based on the image-assisted turntable electrode atomic emission spectrum disclosed by the invention integrates a camera monitoring method and a spectrum analysis method, simultaneously acquires an arc plasma image and an emission spectrum by using a camera and a spectrometer, and assists the atomic spectrum analysis by using the plasma image, and specifically, realizes quantitative analysis result compensation by using the image and the spectrum data and combining a partial least square algorithm, so that the method can obviously improve the oil element field detection precision and improve the element detection repeatability under the condition of not obviously increasing the cost;

(2) the invention discloses an oil element on-site detection device based on an image-assisted turntable electrode atomic emission spectrum, which is characterized in that a camera is introduced on the basis of a traditional turntable electrode atomic emission spectrometer, algorithms such as image processing and spectral analysis are applied, high-precision quantitative analysis of oil elements is realized, the environmental adaptability of the device is strong, the performance is reliable, the oil element analysis requirements under various on-site environments can be met, and the device is an important way and an effective means for realizing wear monitoring of mechanical parts and ensuring safe navigation of important carriers such as airplanes and ships.

Drawings

FIG. 1 illustrates the process of establishing an internal relationship between an arc image and a plasma state parameter in the present invention.

Fig. 2 shows a process of using image-spectrum data to realize measurement result compensation in the present invention.

FIG. 3 is a schematic structural diagram of an oil element on-site detection device based on an image-assisted turntable electrode atomic emission spectrum.

Detailed Description

The invention is described in further detail below with reference to the figures and specific examples. It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention.

The invention provides an oil element on-site detection method based on image-assisted turntable electrode atomic emission spectroscopy, which comprises the following steps:

the first step is as follows: synchronously acquiring an arc plasma image and an atomic emission spectrum of an oil liquid sample by using a camera and a turntable electrode atomic emission spectrometer, estimating a plasma state parameter by using an element spectral line, and establishing an internal relation between the plasma image and the state parameter, wherein the process comprises two links as shown in figure 1;

(101) through a large number of experiments, different combinations of oil standard samples with known element contents are utilized to establish a training set, quantitative analysis is carried out on oil elements, and an array of arc images 2 and plasma spectrum data 3 are obtained for arc plasmas 1;

(102) the plasma state parameters are estimated by utilizing the multispectral line to average, an independent variable matrix 4 is constructed by full-pixel data of an arc image, a dependent variable matrix 5 is constructed by the plasma state parameters (temperature and electron density), a regression model 6 is established by utilizing PLS, and the leave-one-out cross validation is adopted.

The second step is that: based on an arc plasma image and atomic emission spectrum data, an image-spectrum training and testing model is established by using a Partial Least Squares (PLS) algorithm, a training model is firstly used for solving to obtain a compensation proportionality coefficient, then a testing model is used for solving to obtain a compensated spectral line intensity, and the process is shown in figure 2 and comprises two links;

(201) and establishing a training model. Using p standard oil liquid samples (with known element content), acquiring q groups of images and spectra for each sample, and forming k-p × q groups of observation data, wherein the image resolution is assumed to be r-m × n; converting an original image 7m multiplied by n pixel matrix into a single-row 1 multiplied by r array, wherein k original images form a k multiplied by r training matrix P8; analyzing the training matrix P by using a PLS algorithm to obtain a coefficient matrix R9 of R multiplied by R, extracting appropriate first l (l is less than or equal to R) columns to form an R multiplied by l image feature extraction matrix E10, and simultaneously extracting first l principal components corresponding to the training matrix P to form a k multiplied by l image feature matrix C11; line intensity I for a particular wavelength_iThe average intensity of the wavelength spectral line is calculated by using the spectra 12 of the k groups of training sets

13 is

The relative deviation vector e 14 of the wavelength line intensity is recorded as

Establishing a regression 15 of the image feature matrix C to the relative deviation vector e by using PLS to obtain a PLS regression coefficient vector B16 of l multiplied by 1; according to the Schiebe-Lomakin formula, when the self absorption of the spectral line is not considered, the content c of the element corresponding to the spectral line is known, and the proportionality coefficient K17 can be obtained

(202) And establishing a test model. Using an oil liquid sample to be detected, and acquiring a group of images and spectral data by single acquisition; converting the 18m × n pixel matrix of the original image into a single-row 1 × r array P_x19; extracting image characteristic matrix E and array P_xMultiplying, extracting to obtain an image feature array C with 1 × l main components_x20; using PLS regression coefficient vector B and image feature array C_xEstimating the relative deviation e of the spectral line intensity_out21 is

In the formula, B (j) and C_x(j) Refer to B and C, respectively_xThe jth value of (d); using relative deviation e for test set raw spectral line intensity 22_outThe corrected line intensity I' 23 is

The third step: the compensated spectral line intensity is utilized, and a quantitative analysis algorithm is combined to realize the detection of the oil element content;

(301) calculating to obtain the predicted element content c by using the proportionality coefficient K and the corrected spectral line intensity I' in the training model_x24 is

c_x＝KI′ (6)

In addition, the invention also provides an oil element on-site detection device based on an image-assisted turntable electrode atomic emission spectrum, which mainly comprises three parts, namely an arc discharge part, an image-assisted part and a spectrum detection part, as shown in fig. 3, and the specific equipment or devices comprise an arc generator 25, a turntable graphite electrode 26, a collimating mirror 27, a beam splitter 28, a camera 29, a spectrum collector 30, an optical fiber 31, a spectrometer 32, a computer 33 and the like. The working process is as follows: the turntable graphite electrode 26 carries oil to be detected into a discharge gap, an arc plasma is generated under the drive of the arc generator 25, a plasma emission spectrum forms two light paths with different directions through the collimating lens 27 and the beam splitter 28, one light path enters the spectrometer 32 through the spectrum collector 30 and the optical fiber 31 for light splitting and collection to obtain spectrum data, the other light path enters the camera 29 for camera shooting and collection to obtain image data, and the spectrum and the image data are simultaneously sent to the computer 33 for analysis and processing.

The present invention is not limited to the above-described embodiments. The foregoing description of the specific embodiments is intended to describe and illustrate the technical solutions of the present invention, and the above specific embodiments are merely illustrative and not restrictive. Those skilled in the art can make many changes and modifications to the invention without departing from the spirit and scope of the invention as defined in the appended claims.