Oil product near infrared spectrum detection method and detection device

Abstract

The invention provides an oil product near infrared spectrum detection method and a detection device.The device comprises a sample pool, a near infrared light source, a linear array type infrared light filter, a linear array camera and a circuit part, and the sample pool is used for containing an oil sample to be detected; the near-infrared light source is used for irradiating an oil sample to be detected; the linear array type infrared optical filter is a step type thin film FP cavity filter, and the step type thin film FP cavity filter is used for performing multispectral light splitting treatment on an optical signal transmitted when the near-infrared light source irradiates the oil sample to be detected to obtain a multispectral transmission signal; the linear array camera comprises a linear array type infrared image sensor and a camera circuit board, and the linear array type infrared image sensor is used for sensing the multispectral transmission signal and generating image data through the camera circuit board; and the circuit part is used for obtaining the component information of the oil sample to be detected based on the image data and the oil component inversion model, so that the function of quickly and accurately detecting the quality of the light oil on site is realized.

Description

Technical Field

[0001] The present application relates to the technical field of oil product detection devices, and in particular to a near-infrared spectrum detection method and a detection device for oil products. Background Art

[0002] Light fuel oil generally refers to a hydrocarbon mixture with a boiling point range of about 50 to 350°C. Light fuel oil is relatively clean, has a low density, a low calorific value, and is easy to ignite. Due to its good combustibility, volatility, and good stability, it is the most consumed, widely used, and most in-demand finished oil. It is widely used in people's lives, social production, and military fields such as aerospace, aviation, weapons, and ships, which brings about a very wide range of oil quality testing needs.

[0003] Currently, the oil industry mostly uses traditional laboratory analysis methods, which are limited to laboratory environments. Different indicators need to be tested separately, and the analysis time is long, the operation is cumbersome, the cost is high, and it is impossible to control the analysis in real time. Summary of the invention

[0004] The present application aims to solve one of the technical problems in the related art at least to some extent.

[0005] To this end, the first purpose of the present application is to propose a near-infrared spectroscopy detection method for oil products, so as to realize the on-site rapid and accurate detection function of the quality of light oil.

[0006] The second purpose of the present application is to provide a near-infrared spectroscopy detection device for oil products.

[0007] To achieve the above-mentioned purpose, the first embodiment of the present application proposes a near-infrared spectroscopy detection method for oil products, comprising:

[0008] Irradiate the oil sample to be tested by near infrared light;

[0009] Performing multi-spectral spectroscopic processing on the light signal transmitted by the oil sample to be tested to obtain a multi-spectral transmission signal;

[0010] sensing the multi-spectral transmission signal and generating image data;

[0011] Based on the image data, obtaining spectral features;

[0012] The spectral features are input into an oil composition inversion model to obtain oil composition information of the oil sample to be tested; the oil composition inversion model is trained based on training samples obtained from near-infrared absorption spectrum image data of different light oils and corresponding oil composition information.

[0013] To achieve the above-mentioned purpose, the second embodiment of the present application proposes a near-infrared spectrum detection device for oil products, comprising:

[0014] A sample pool, used to contain the oil sample to be tested;

[0015] A near-infrared light source, the near-infrared light source is used to emit a near-infrared signal to irradiate the oil sample to be tested in the sample pool;

[0016] A linear infrared filter, wherein the linear infrared filter is a step-type thin film FP cavity filter, and the step-type thin film FP cavity filter is used to perform multi-spectral spectroscopic processing on the light signal transmitted when the near-infrared light source irradiates the oil sample to be tested, so as to obtain a multi-spectral transmission signal;

[0017] A linear camera, the linear camera comprising a linear infrared image sensor and a camera circuit board, the incident window of the linear infrared image sensor being coupled and aligned with the stepped thin film FP cavity filter through a structural member, the linear infrared image sensor being used to sense the multi-spectral transmission signal and generate image data through the camera circuit board;

[0018] The circuit part is used to drive the near-infrared light source and obtain the component information of the oil sample to be tested based on the image data and the oil component inversion model.

[0019] In some implementations, the method for manufacturing the stepped thin film FP cavity filter comprises:

[0020] A photolithography-etching process is performed on the light-transmitting layer of the filter structure through a photolithography mask with 128 equally spaced intervals, so as to obtain a stepped thin-film FP cavity filter including 128 stepped structures.

[0021] In some implementations, the stepped thin film FP cavity filter includes a stepped structure formed by alternating high refractive index materials and low refractive index materials, the high refractive index material is one of Ta2O5, Ti3O5, TiO2, α-Si, Si3N4 and Nb2O5, and the low refractive index material is one of SiO2, MgF2 and Al2O3.

[0022] In some implementations, the structural member includes a frame with a central opening, a support plate and a bracket, the bracket is arranged parallel to the frame via the support plate perpendicular to the frame, the bracket installs the linear infrared filter via a slot thereon and is fixed by a slot cover; the linear infrared image sensor is arranged at a position on the support plate between the frame and the bracket, and the frame is connected to the camera circuit board at one end away from the bracket.

[0023] In some implementations, the oil product near-infrared spectrum detection device also includes a portable handheld housing, and the sample pool, near-infrared light source, linear infrared filter, linear camera and circuit part are all fixed in the portable handheld housing.

[0024] In some implementations, when the circuit part obtains the composition information of the oil sample to be tested based on the image data and the oil composition inversion model, it is used to:

[0025] Preprocessing and feature extraction are performed on the image data to obtain spectral features;

[0026] The spectral characteristics are input into an oil composition inversion model to obtain oil composition information of the oil sample to be tested.

[0027] In some implementations, the method for constructing the oil composition inversion model includes:

[0028] Obtain near-infrared absorption spectrum image data of different light oils and corresponding oil composition information;

[0029] Preprocessing the near-infrared absorption spectrum image data and extracting data vectors to obtain vector data of the absorption spectrum image data;

[0030] Extracting spectral features based on the determined spectral feature types and the vector data of the absorption spectrum image data;

[0031] The spectral characteristics of the light oils of different grades and the corresponding oil composition information are used as the model input and model output of the initial model respectively, and the initial model is trained to obtain an oil composition inversion model.

[0032] In some implementations, the spectral features include characteristic spectral segments and characteristic wavelength variables, the characteristic spectral segments are screened and determined by one or more methods selected from the group consisting of interval partial least squares, joint interval partial least squares, and moving window partial least squares, and the characteristic wavelength variables are screened and determined by one or more methods selected from the group consisting of correlation coefficient method, genetic algorithm, ant colony algorithm, random forest, and continuous projection algorithm.

[0033] In some implementations, the preprocessing of the near-infrared absorption spectrum image data includes:

[0034] The near-infrared absorption spectrum image data is corrected by a correction model; wherein the correction model is trained based on spectrum image data samples collected at different temperatures.

[0035] The near-infrared spectral detection method and detection device for oil products provided in the present application realize multi-spectral splitting by adopting a stepped thin-film FP cavity filter. A larger number of spectral bands can obtain more comprehensive and accurate analysis results, effectively identify sample components, and can test multiple parameters of oil products at the same time, thereby realizing rapid detection of multiple parameters of oil products; the stepped thin-film FP cavity filter and the linear array camera are fixed by structural parts to improve environmental adaptability and system stability; based on the collected spectral image data and the oil quality component inversion model, rapid detection of oil quality components of oil products can be realized; the device quickly obtains spectral image data and processes it quickly to realize rapid detection of the quality of light fuel oil such as gasoline, thereby realizing a handheld portable oil product detection instrument; and the device adopts an embedded system to ensure that the measurement, storage, display, and networking functions are all available at a relatively low cost, volume, and power consumption, thereby meeting user needs and meeting the portable rapid detection needs of first-site gas stations.

[0036] Additional aspects and advantages of the present application will be given in part in the description below, and in part will become apparent from the description below, or will be learned through the practice of the present application. BRIEF DESCRIPTION OF THE DRAWINGS

[0037] The above and / or additional aspects and advantages of the present application will become apparent and easily understood from the following description of the embodiments in conjunction with the accompanying drawings, in which:

[0038] Figure 1 A flowchart of a near-infrared spectroscopy detection method for oil products provided in an embodiment of the present application;

[0039] Figure 2 A schematic diagram of the structure of a near-infrared spectroscopy detection device for oil products provided in an embodiment of the present application;

[0040] Figure 3 A schematic diagram of a photolithography mask layout structure provided in an embodiment of the present application;

[0041] Figure 4 A schematic diagram of a three-dimensional structure of a structural member provided in an embodiment of the present application;

[0042] Figure 5 for Figure 4 Cross-sectional view of .

[0043] Figure 6 A schematic diagram of the structure of a sample pool provided in an embodiment of the present application;

[0044] Figure 7 A block diagram of a device control program provided in an embodiment of the present application;

[0045] Figure 8A flowchart of a method for constructing an oil composition inversion model provided in an embodiment of the present application. DETAILED DESCRIPTION

[0046] Embodiments of the present application are described in detail below, and examples of the embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals throughout represent the same or similar elements or elements having the same or similar functions. The embodiments described below with reference to the accompanying drawings are exemplary and are intended to be used to explain the present application, and should not be construed as limiting the present application.

[0047] The following describes the oil product near-infrared spectroscopy detection method and detection device of the embodiment of the present application with reference to the accompanying drawings.

[0048] Figure 1 This is a flow chart of a near infrared spectroscopy detection method for oil products provided in an embodiment of the present application. Figure 1 As shown, the oil product near infrared spectrum detection method may include the following steps:

[0049] Step S101, irradiating the oil sample to be tested with near infrared light.

[0050] As an implementation method, the oil sample to be tested can be irradiated by an infrared light source to obtain infrared absorption spectrum information.

[0051] Step S102, performing multi-spectral spectroscopic processing on the light signal transmitted by the oil sample to be tested to obtain a multi-spectral transmission signal.

[0052] As an implementation method, a multi-spectral spectroscopic processing can be realized through a filtering structure.

[0053] Step S103: sensing the multi-spectral transmission signal and generating image data.

[0054] As an implementation method, multi-spectral transmission signals can be sensed by an infrared image sensor to generate image data.

[0055] Step S104, obtaining spectral features based on the image data.

[0056] Step S105, inputting the spectral features into an oil composition inversion model to obtain oil composition information of the oil sample to be tested; the oil composition inversion model is trained based on training samples obtained from near-infrared absorption spectrum image data of different light oils and corresponding oil composition information.

[0057] The near-infrared spectroscopy detection method for oil products of the embodiment of the present application can realize the rapid detection of the oil components of oil products based on the collected spectral image data and the oil component inversion model; by performing multi-spectral spectroscopic processing on the light signal transmitted by the oil sample to be tested, a more comprehensive and accurate analysis result can be obtained, and the sample components can be effectively identified.

[0058] In order to implement the above embodiment, Figure 2 This is a schematic diagram of the structure of a near-infrared spectrum detection device for oil products provided in an embodiment of the present application. Figure 2 As shown, the oil product near-infrared spectrum detection device may include: a sample pool, a near-infrared light source, a linear infrared filter, a linear camera and a circuit part.

[0059] The sample pool is used to contain the oil sample to be tested; optionally, the sample pool is made of quartz glass. The oil sample to be tested is light oil, which may include but is not limited to gasoline and diesel.

[0060] According to experimental verification, the absorption path of the sample pool is about 10mm, which is better, and can ensure the absorption effect while occupying a small volume. The sample pool is designed to be low-cost to ensure that it can be easily cleaned / disposable after use.

[0061] The near-infrared light source is used to emit near-infrared signals to illuminate the oil sample to be tested in the sample pool, provide excitation light for the entire device, and cover the near-infrared detection wavelength range. The near-infrared light source is required to have good stability and a sufficiently long life, and its luminous wavelength should cover the spectral range of the analysis as a near-infrared wavelength. Considering that the characteristic absorption band for measuring gasoline quality components is 1.1 to 1.7 μm, the near-infrared light source selects the IRL715 infrared light source.

[0062] It should be noted that the infrared spectrum is the absorption spectrum produced by the absorption of electromagnetic waves in the infrared light region by the molecules of a substance. The present invention selects the near-infrared (1100-1700nm) region. The infrared absorption peaks of the molecules of a substance in the near-infrared region are relatively large and have fingerprint properties. When there are slight changes in the molecular structure of a substance, the spectrum in this region will change significantly, and it has excellent performance in identifying the functional groups of the molecules of a substance. The spectral range is 1100-1700nm, the spectral resolution reaches 5nm, the wavelength accuracy is better than ±1nm, and the wavelength resolution is better than 5nm.

[0063] The linear infrared filter is a stepped thin film FP cavity filter, which is used to perform multi-spectral spectroscopic processing on the light signal transmitted when the near-infrared light source irradiates the oil sample to be tested, so as to obtain a multi-spectral transmission signal.

[0064] A linear camera, the linear camera comprising a linear infrared image sensor and a camera circuit board, the incident window of the linear infrared image sensor being coupled and aligned with the stepped thin film FP cavity filter through a structural member, the linear infrared image sensor being used to sense the multi-spectral transmission signal and generate image data through the camera circuit board;

[0065] The circuit part is used to drive the near-infrared light source and process the image data to obtain the component information of the oil sample to be tested.

[0066] The near-infrared spectral detection device for oil products in the embodiment of the present application realizes multi-spectral splitting by adopting a stepped thin-film FP cavity filter. A larger number of spectral bands can obtain more comprehensive and accurate analysis results, effectively identify sample components, and can test multiple parameters of oil products at the same time, thereby realizing rapid detection of multiple parameters of oil products. The stepped thin-film FP cavity filter and the linear array camera are fixed by structural parts to improve environmental adaptability and system stability. Based on the collected spectral image data and the oil quality component inversion model, rapid detection of oil quality components of oil products can be realized. The device quickly obtains spectral image data and processes it quickly to realize rapid detection of the quality of light fuel oils such as gasoline, thereby realizing a handheld and portable oil product detection instrument, which can meet the portable rapid detection needs of first-site such as gas stations.

[0067] In some embodiments, the method for manufacturing the stepped thin film FP cavity filter comprises: performing a photolithography-etching process on the light-transmitting layer of the filter structure through a photolithography mask with 128 equal intervals to obtain a stepped thin film FP cavity filter comprising 128 step structures.

[0068] Among them, the photolithography mask is as follows Figure 3 As shown, a 128-equally divided photolithography mask can be selected to realize the production of a stepped thin film FP cavity filter.

[0069] The working wavelength of the device is 1100-1700nm. To ensure the complete spectrum information in this spectrum, the measurement spectrum is subdivided. In this embodiment, it is subdivided into 128 spectrum segments, that is, the working wavelength is evenly divided into 128 spectrum segments, and the half-wave width of each spectrum segment is about 4nm. In other words, the 128 spectrum segments can be split by the step-type thin film FP cavity filter, and the information of each spectrum segment can be sensed by the linear infrared image sensor, so that the monitoring result of the device is more accurate.

[0070] In some embodiments, the stepped thin film FP cavity filter includes a stepped structure formed by alternating high refractive index materials and low refractive index materials, the high refractive index material is one of Ta2O5, Ti3O5, TiO2, α-Si, Si3N4 and Nb2O5, and the low refractive index material is one of SiO2, MgF2 and Al2O3.

[0071] In some embodiments, Figure 4 , Figure 5 As shown, the structural member includes a frame 1 with a central opening, a support plate 2 and a bracket 3. The bracket 3 is arranged parallel to the frame 1 through the support plate 2 perpendicular to the frame 1. The bracket 3 installs the linear infrared filter 5 through the slot thereon and is fixed by the slot cover 4; the linear infrared image sensor 6 is arranged at a position between the frame 1 and the bracket 3 on the support plate 2, and the end of the frame 1 away from the bracket 3 is connected to the camera circuit board 7.

[0072] Thus, the front end of the structural member is equipped with a linear infrared filter, which is contained in a slot and fixed in position by adhesive and then covered with a slot cover; the rear end of the structural member is connected to a camera circuit board, and a linear infrared image sensor is arranged between the linear infrared filter and the camera circuit board to realize the sensing light path. The camera circuit board fixed to the linear camera can be connected through the four positioning holes on the frame.

[0073] In some embodiments, the line array camera is a short wave infrared line array camera.

[0074] In some embodiments, the oil product near-infrared spectrum detection device also includes a portable handheld housing, and the sample pool, near-infrared light source, linear infrared filter, linear camera and circuit part are all fixed in the portable handheld housing.

[0075] Therefore, this device provides a portable oil analyzer, which breaks through the shortcomings of traditional oil detection spectral equipment such as "cannot be carried", "too slow to measure" and "cannot move", and realizes handheld and portable measurement. The device is small in size and light in weight, so as to meet the portable and rapid detection needs of the first scene such as gas stations.

[0076] In some embodiments, in the early technical verification stage, such as Figure 6 As shown, the sample cell is a reinforced cuvette that is resistant to chemical corrosion.

[0077] When a point light source is selected as the near-infrared light source, in some embodiments, the device also includes an auxiliary optical device, which may include one or more of the optical devices such as attenuation, dispersion, and collimation. The auxiliary optical device is arranged between the sample pool and the near-infrared light source so that the incident light irradiated to the sample pool meets the requirements of the subsequent photosensitive components.

[0078] In some embodiments, when the circuit part processes the image data to obtain the component information of the oil sample to be tested, it is used to:

[0079] Preprocessing and feature extraction are performed on the image data to obtain spectral features;

[0080] The spectral characteristics are input into an oil composition inversion model to obtain oil composition information of the oil sample to be tested.

[0081] In some embodiments, the circuit part includes a main control board (i.e., a controller), a driving circuit, and a signal conditioning circuit. The main control board runs an operating system and embedded software, and the main control board is a controller; the circuit part assumes the functions of controlling each component, collecting information, processing information, and displaying information. The main control board may include a processor CPU and a UART interface, a USB interface, an HDMI output interface, a RAM memory, a data storage module, and a Gigabit Ethernet interface connected to the CPU, wherein the CPU selects a high-performance ARM, with sufficient RAM memory and data storage space, and can run an operating system of the Linux kernel to run the camera SDK, so that the functions of the whole machine can be realized; at least one Gigabit Ethernet interface is used for debugging and connecting a linear camera, a UART interface is used to control peripheral modules of the main control board (such as a driving circuit, etc.), an HDMI output interface is used to output a user graphical interface, and a USB interface is used to receive touch screen information and connect other necessary devices during debugging.

[0082] Optionally, NanoPi R6C is used as the main control board, which is a single-board computer based on the high-performance ARM processor Rockchip RK3588. It has 32GB on-board eMMC storage, 8GB RAM, and all necessary peripheral interfaces such as HDMI. Its dual Ethernet port design ensures that it can control the photoelectric sensor component (infrared array camera) while having an additional Ethernet port as a debugging terminal.

[0083] The main control software of this device runs on the main control board and has the following functions:

[0084] Calibration: Due to the absorption principle, whether it is collecting database data or actual measurement, this device needs to first collect the empty light path (including the sample cell) to obtain its spectral information, and then use the sample cell to hold the standard or the product to be tested to collect data. The ratio of the two is the effective "absorption spectrum" data to achieve device calibration.

[0085] Measurement: During actual measurement, place the sample pool containing the oil sample into the light path and collect the absorption of each spectral band.

[0086] Transmission and display: internal storage, display (LCD screen, etc.), transmission (low-speed interface, high-speed interface, etc.) and other functions can be customized according to user needs.

[0087] Algorithm operation: It has secondary development function and reserved algorithm interface, which allows users to process the raw data obtained by measurement (raw data 1×128 vector, which represents the absorption spectrum of the sample in the sample pool) and call the transmission interface for external transmission.

[0088] The control software of this device is divided into functions, such as Figure 7 As shown, it includes a user instruction processing module, a light source control module, an infrared sensing system control module, an infrared information processing module, and a storage, transmission, and display module. The user instruction processing module is used to obtain and process user instructions through buttons or external data input interfaces. The light source control module is used to control the near-infrared light source to work through a driving circuit according to user instructions. The infrared sensing system control module is used to control the linear array camera to work according to user instructions. The infrared information processing module is used to process image data output by the infrared sensing system.

[0089] The control software of the device of this embodiment runs on an embedded operating system, and receives external user commands through buttons and external data input interfaces (such as serial ports, mainly for debugging), and is parsed by a user command processing module. The user command processing module mainly controls the light source and imaging system, and can also operate the storage, display and transmission modules to achieve a "trigger acquisition" operation.

[0090] Among them, the light source control module controls the near-infrared light source to emit light and turn off through PWM or voltage control. The infrared sensor system control module is formed by encapsulating the necessary functions in the camera SDK (software development kit) running on the embedded system. Its main function is to accept external instructions, control the infrared sensor imaging component (i.e., the linear array camera), and realize parameter setting and trigger shooting functions. The acquired original image is processed by the infrared information processing module, and the linear infrared spectrum image obtained by the linear array camera is calculated to obtain the processing result, which is then handed over to the storage, transmission and display module. Among them, the image preprocessing method mainly includes but is not limited to general methods such as baseline correction, normalization, derivative, smoothing, standard overall variable transformation and multivariate scattering correction. Spectral data processing is to convert the preprocessed image data into a one-dimensional spectral data vector. The storage, transmission and display module calls the operating system related interface, displays the result on the display screen, and outputs it to the specified external data interface. The related functions such as viewing the result (original image, spectral curve, processing result), operating storage and transmission are controlled by the user instruction processing module.

[0091] In some embodiments, Figure 8 As shown, the method for constructing the oil composition inversion model includes:

[0092] Step S201, obtaining near-infrared absorption spectrum image data of different light oils and corresponding oil composition information;

[0093] Step S202, preprocessing the near-infrared absorption spectrum image data and extracting data vectors to obtain vector data of the absorption spectrum image data;

[0094] Step S203, extracting spectral features based on the determined spectral feature type and the vector data of the absorption spectrum image data;

[0095] Step S204, using the spectral characteristics of the light oils of different grades and the corresponding oil composition information as the model input and model output of the initial model respectively, training the initial model, and obtaining an oil composition inversion model.

[0096] It should be noted that spectral analysis is the only way to achieve qualitative and quantitative detection of various types of oil products. The present invention proposes the concept of mathematical separation. Under the condition that complex matrix and background coexist and the original system equilibrium is not disturbed, a new chemometric method is developed to reasonably analyze the infrared spectral signal and accurately extract the spectral information of the substance to be tested. The oil quality detection algorithm is mainly divided into the following three parts: establishment of standardized data set, near-infrared spectral preprocessing, spectral feature analysis and establishment of oil quality component inversion model.

[0097] It should be noted that when obtaining the near-infrared absorption spectrum image data and corresponding oil composition information of different light oils, the infrared spectrum of the target oil sample can be first collected, and the key property test values ​​of the target oil sample can be obtained by laboratory testing methods; then, the infrared spectrum data of the target oil sample can be correlated with its property data by selecting an appropriate chemometric multivariate correction method.

[0098] As an implementation method, the method for establishing a standardized data set includes: using an infrared spectrometer to collect standard spectral data of the main components to be tested in oil products, and establishing a standard sample spectral database including a standard spectral data set through sample optimization and multiple collections. In order to maintain the scientificity and purity of the standard spectral data, an intelligent discrimination method can be used to benchmark the physical and chemical detection indicators, and abnormal data can be screened out for elimination. Abnormal screening methods such as Mahalanobis distance discrimination method, spectral peak abnormality discrimination method, spectral residual ratio method, etc., to ensure the consistency of the standard spectral data of a single component.

[0099] As an implementation method, the method of near infrared spectrum preprocessing includes: mainly eliminating interference information such as noise and background information in standard spectrum data, improving the spectrum signal-to-noise ratio, and retaining effective spectrum information for modeling. The preprocessing operation of near infrared spectrum is mainly for the processing of useless information such as noise information and background information in the spectrum, eliminating the influence of useless information, and retaining more useful information for modeling. Commonly used NIR spectrum preprocessing methods include baseline correction, normalization, derivative, smoothing, standard overall variable transformation and multivariate scattering correction. Each processing method has its advantages and disadvantages. For example, differential processing can eliminate background interference, but it will introduce certain noise and need to be denoised again. Therefore, this embodiment combines multiple spectrum preprocessing methods to achieve better processing effects.

[0100] In some implementations, the spectral features include characteristic spectral segments and characteristic wavelength variables, the characteristic spectral segments are screened and determined by one or more methods selected from the group consisting of interval partial least squares, joint interval partial least squares, and moving window partial least squares, and the characteristic wavelength variables are screened and determined by one or more methods selected from the group consisting of correlation coefficient method, genetic algorithm, ant colony algorithm, random forest, and continuous projection algorithm.

[0101] Specifically, the method of spectral feature analysis includes: performing spectral feature analysis based on standard spectral data, extracting effective spectral features within the full coverage spectral range, and removing redundant information. There are various methods for optimizing spectral feature information, including methods for selecting characteristic spectral segments (i.e., wavelength intervals) and methods for screening characteristic wavelength variables. Among them, commonly used methods for selecting characteristic spectral segments include in-interval partial least squares (iPLS), joint interval partial least squares (SiPLS), and moving window partial least squares (MWPLS). The characteristics of these methods are that the entire working wavelength range is first divided into multiple continuous or discontinuous bands, and then for each spectrum, the PLS model is used to predict the properties of the same validation set data sample, and the root mean square error of the prediction results is calculated. The one or more spectrum segments corresponding to the minimum root mean square error are selected as the final characteristic spectrum segments. These methods can be used alone or in combination to achieve better results. Commonly used characteristic wavelength variable screening methods can include correlation coefficient method, genetic algorithm (GA), ant colony algorithm (ACO), random forest (RF) and continuous projection algorithm (SPA), etc. The characteristics of these methods are to use corresponding algorithms to extract characteristic wavelength variables that best characterize sample properties from the entire working wavelength range. Since the algorithms used in each method are different, the operation complexity is also different, and the final model prediction effect is also different. Among them, the correlation coefficient method is to calculate the correlation between each wavelength of the spectrum and the sample property. The larger the correlation coefficient, the more useful information it covers. The wavelength corresponding to the maximum correlation is selected as the characteristic wavelength variable. This method has the characteristics of simple algorithm and quick use. The selected characteristic wavelength variable can meet the requirements of model prediction accuracy. Relatively speaking, the implementation of the genetic algorithm is relatively complex. The basic principle is to refer to the natural selection and genetic mechanism of the biological world. Through operator operations such as selection, exchange and mutation, the variables with better objective function values ​​are retained and the poorer variables are eliminated. The selected variables are characteristic wavelength variables. Although the method is complex to operate, the selected variables optimize the detection model and enhance the prediction ability of the model. The present invention selects the optimal model for feature selection according to the spectral characteristics of the oil sample set to be tested. Therefore, through spectral feature analysis, the spectral features used for model prediction are determined.

[0102] After constructing a standard sample spectral database and screening and determining the spectral features, an oil composition inversion model can be established.

[0103] As an implementation method, a method for establishing an oil composition inversion model includes the following five steps:

[0104] Step 1, obtain standard spectral data samples from the standard sample spectral database, perform data cleaning and random sampling, and establish training sets and test sets;

[0105] Step 2, preprocessing the training set, performing spectral data processing after the preprocessing, extracting the spectral data vector of each sample, and then extracting the above-screened spectral features from the spectral data vector;

[0106] Step 3: Input the spectral features into the trainer model. The trainer can select known model architectures such as linear discriminant analysis model, feedforward neural network, random forest, etc., take the oil component information known in the training set as the output result, obtain the key node parameters of the model through machine learning of the spectral features, and build a classification model;

[0107] Step 4: Input the test set data into the classification model to verify the recognition results, and calculate the standard deviation of the recognition results of the test set components;

[0108] Step 5: If the standard deviation is higher than expected, repeat steps 1-4. During the repetition, the model can be optimized by changing the structure of the training set and the test set in step 1, changing the number of machine learning iterations in step 3, etc., so as to meet the model prediction accuracy requirements and finally obtain the oil composition inversion model.

[0109] It should be noted that the current modeling methods can be divided into linear modeling and nonlinear modeling. Among them, the Partial Least Squares (PLS) method is a typical representative of linear modeling, and nonlinear modeling methods include artificial neural networks (ANN) and support vector machines (SVM). PLS and ANN correction methods have their own advantages and disadvantages. The PLS method is suitable for objects with good linear properties, while the ANN method is suitable for predicting objects with strong nonlinearity.

[0110] In the detection technology based on near-infrared spectroscopy, establishing an oil component inversion model with high prediction accuracy and high credibility is a key link. This is not only about the choice of modeling method, but also has a direct relationship with the spectral information involved in modeling. Therefore, if you want to improve the prediction accuracy and credibility of the oil component inversion model, you need to start from the root, that is, when selecting the standard spectral data samples involved in modeling, on the one hand, take certain measures to eliminate the influence of all interference and improve the spectral signal-to-noise ratio as much as possible, which is called "false removal" operation; on the other hand, use appropriate methods to effectively extract useful information and retain as much spectral feature information as possible for modeling, which is called "true preservation" operation. Among them, the operation of reducing the influence of interference usually includes filtering denoising, eliminating background interference, etc. The preprocessing methods mentioned above, such as baseline correction, normalization, differentiation and wavelet transform, all have the ability to "remove false". In addition, since the sample temperature will also affect the spectrum, the increase in temperature will lead to problems such as decreased absorbance and spectral peak shift. In order to eliminate the influence of temperature, standard spectral data samples collected at different temperatures can be used to establish a mixed temperature correction set to establish a correction model. Both methods are effective in eliminating the influence of temperature and can improve the prediction accuracy of the oil composition inversion model. In addition, the commonly used method for extracting useful information is principal component analysis, which can not only effectively extract useful information, but also reduce the data dimension, greatly shorten the data processing time, and improve the real-time prediction. This method has good prediction accuracy and model adaptability. Therefore, the spectral features can be reduced in dimension by principal component analysis.

[0111] Therefore, in some embodiments, the preprocessing of the near-infrared absorption spectrum image data includes:

[0112] The near-infrared absorption spectrum image data is corrected by a correction model; wherein the correction model is trained based on spectrum image data samples collected at different temperatures.

[0113] This device uses an oil composition inversion model to predict the composition of the oil product to be tested, has a fast detection speed, and can achieve real-time detection. In addition, this method can improve the accuracy of the test, and the use of full spectrum information (spectral information of all spectral bands within the range of 1100-1700nm) for prediction can reduce the background prediction error rate by half, and has strong universality. This method can quickly detect the key properties (i.e., ingredients or components) of the oil sample to be tested, provide qualitative and quantitative light oil component indicators, density, viscosity, flash point and other physical and chemical indicators, and can also identify jet fuel antistatic additives, anti-icing additives (specific identification indicators can be constructed according to the corresponding model according to the identification requirements), etc., and ensure that the test results have a certain correlation with the standard method.

[0114] Key properties of gasoline may include, but are not limited to, research octane number, benzene content, aromatics content, olefin content, oxygen content, and methanol content. Key properties of diesel may include, but are not limited to, polycyclic aromatic hydrocarbons content, cetane number, and cetane index.

[0115] It should be noted here that a personalized oil composition inversion model can be established according to the indicator identification requirements to achieve the output of the predicted value of the specific indicator input data.

[0116] In the description of the aforementioned embodiments, the description with reference to the terms "one embodiment", "some embodiments", "example", "specific example", or "some examples" etc. means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present application. In this specification, the schematic representations of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described may be combined in any one or more embodiments or examples in a suitable manner. In addition, those skilled in the art may combine and combine the different embodiments or examples described in this specification and the features of the different embodiments or examples, without contradiction.

[0117] In addition, the terms "first" and "second" are used for descriptive purposes only and should not be understood as indicating or implying relative importance or implicitly indicating the number of the indicated technical features. Therefore, the features defined as "first" and "second" may explicitly or implicitly include at least one of the features. In the description of this application, the meaning of "plurality" is at least two, such as two, three, etc., unless otherwise clearly and specifically defined.

[0118] Any process or method description in a flowchart or otherwise described herein may be understood to represent a module, fragment or portion of code comprising one or more executable instructions for implementing the steps of a custom logical function or process, and the scope of the preferred embodiments of the present application includes alternative implementations in which functions may not be performed in the order shown or discussed, including performing functions in a substantially simultaneous manner or in the reverse order depending on the functions involved, which should be understood by technicians in the technical field to which the embodiments of the present application belong.

[0119] The logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as an ordered list of executable instructions for implementing logical functions, which can be embodied in any computer-readable medium for use by an instruction execution system, apparatus or device (such as a computer-based system, a system including a processor or other system that can fetch instructions from an instruction execution system, apparatus or device and execute instructions), or used in combination with these instruction execution systems, apparatuses or devices.

[0120] It should be understood that the various parts of the present application can be implemented by hardware, software, firmware or a combination thereof. In the above-mentioned embodiments, multiple steps or methods can be implemented by software or firmware stored in a memory and executed by a suitable instruction execution system. For example, if implemented by hardware, as in another embodiment, it can be implemented by any one of the following technologies known in the art or their combination: a discrete logic circuit having a logic gate circuit for implementing a logic function for a data signal, a dedicated integrated circuit having a suitable combination of logic gate circuits, a programmable gate array (PGA), a field programmable gate array (FPGA), etc.

[0121] A person skilled in the art may understand that all or part of the steps in the method for implementing the above-mentioned embodiment may be completed by instructing related hardware through a program, and the program may be stored in a computer-readable storage medium, which, when executed, includes one or a combination of the steps of the method embodiment.

[0122] In addition, each functional unit in each embodiment of the present application may be integrated into a processing module, or each unit may exist physically separately, or two or more units may be integrated into one module. The above-mentioned integrated module may be implemented in the form of hardware or in the form of a software functional module. If the integrated module is implemented in the form of a software functional module and sold or used as an independent product, it may also be stored in a computer-readable storage medium.

[0123] The storage medium mentioned above may be a read-only memory, a magnetic disk or an optical disk, etc. Although the embodiments of the present application have been shown and described above, it can be understood that the above embodiments are exemplary and cannot be understood as limiting the present application. A person of ordinary skill in the art may change, modify, replace and modify the above embodiments within the scope of the present application.

Claims

1. A near infrared spectrum detection method for oil products, characterized in that: The following steps are involved: Irradiate the oil sample to be tested by near infrared light; Performing multi-spectral spectroscopic processing on the light signal transmitted by the oil sample to be tested to obtain a multi-spectral transmission signal; sensing the multi-spectral transmission signal and generating image data; Based on the image data, obtaining spectral features; The spectral features are input into an oil composition inversion model to obtain oil composition information of the oil sample to be tested; the oil composition inversion model is trained based on training samples obtained from near-infrared absorption spectrum image data of different light oils and corresponding oil composition information.

2. A near infrared spectrum detection device for oil products, characterized in that: include: A sample pool, used to contain the oil sample to be tested; A near-infrared light source, the near-infrared light source is used to emit a near-infrared signal to irradiate the oil sample to be tested in the sample pool; A linear infrared filter, wherein the linear infrared filter is a step-type thin film FP cavity filter, and the step-type thin film FP cavity filter is used to perform multi-spectral spectroscopic processing on the light signal transmitted when the near-infrared light source irradiates the oil sample to be tested, so as to obtain a multi-spectral transmission signal; A linear camera, the linear camera comprising a linear infrared image sensor and a camera circuit board, the incident window of the linear infrared image sensor being coupled and aligned with the stepped thin film FP cavity filter through a structural member, the linear infrared image sensor being used to sense the multi-spectral transmission signal and generate image data through the camera circuit board; The circuit part is used to drive the near-infrared light source and obtain the component information of the oil sample to be tested based on the image data and the oil component inversion model.

3. The oil product near infrared spectrum detection device according to claim 2, characterized in that: The method for manufacturing the stepped thin film FP cavity filter comprises: A photolithography-etching process is performed on the light-transmitting layer of the filter structure through a photolithography mask with 128 equally spaced intervals, so as to obtain a stepped thin-film FP cavity filter including 128 stepped structures.

4. The oil product near infrared spectrum detection device according to claim 2, characterized in that: The stepped thin film FP cavity filter includes a stepped structure formed by alternating high refractive index materials and low refractive index materials, the high refractive index material is one of Ta2O5, Ti3O5, TiO2, α-Si, Si3N4 and Nb2O5, and the low refractive index material is one of SiO2, MgF2 and Al2O3.

5. The oil product near-infrared spectrum detection device according to claim 2, characterized in that: The structural member includes a frame with a central opening, a support plate and a bracket. The bracket is arranged parallel to the frame through the support plate perpendicular to the frame. The bracket installs the linear infrared filter through a slot thereon and is fixed by a slot cover. The linear infrared image sensor is arranged on the support plate at a position between the frame and the bracket, and the end of the frame away from the bracket is connected to the camera circuit board.

6. The oil product near infrared spectrum detection device according to claim 2, characterized in that: The oil product near-infrared spectrum detection device also includes a portable handheld housing, and the sample pool, near-infrared light source, linear infrared filter, linear camera and circuit part are all fixed in the portable handheld housing.

7. The oil product near-infrared spectrum detection device according to claim 2, characterized in that: When the circuit part obtains the component information of the oil sample to be tested based on the image data and the oil composition inversion model; Used for: Preprocessing and feature extraction are performed on the image data to obtain spectral features; The spectral characteristics are input into an oil composition inversion model to obtain oil composition information of the oil sample to be tested.

8. The oil product near-infrared spectrum detection device according to claim 7, characterized in that: The method for constructing the oil composition inversion model comprises: Obtain near-infrared absorption spectrum image data of different light oils and corresponding oil composition information; Preprocessing the near-infrared absorption spectrum image data and extracting data vectors to obtain vector data of the absorption spectrum image data; Extracting spectral features based on the determined spectral feature types and the vector data of the absorption spectrum image data; The spectral characteristics of the light oils of different grades and the corresponding oil composition information are used as the model input and model output of the initial model respectively, and the initial model is trained to obtain an oil composition inversion model.

9. The oil product near-infrared spectrum detection device according to claim 7, characterized in that: The spectral characteristics include characteristic spectral segments and characteristic wavelength variables. The characteristic spectral segments are screened and determined by one or more methods selected from the group consisting of interval partial least squares, joint interval partial least squares and moving window partial least squares. The characteristic wavelength variables are screened and determined by one or more methods selected from the group consisting of correlation coefficient method, genetic algorithm, ant colony algorithm, random forest and continuous projection algorithm.

10. The oil product near infrared spectrum detection device according to claim 7, characterized in that: The preprocessing of the near-infrared absorption spectrum image data comprises: The near-infrared absorption spectrum image data is corrected by a correction model; wherein the correction model is trained based on spectrum image data samples collected at different temperatures.

Images