Method, device and equipment for detecting physico-chemical parameters of oil

Abstract

The application provides an oil physicochemical parameter detection method, device and equipment, relates to the technical field of spectral analysis and industrial equipment state monitoring, and the method comprises the following steps: obtaining hyperspectral data of to-be-detected oil; inputting the hyperspectral data into a pre-trained spectral decoupling network to output a base spectral component related to an oil brand and a common component feature vector irrelevant to the oil brand but related to an oil health state; and matching the common component feature vector with a pre-constructed universal reference library to determine a physicochemical parameter value of the to-be-detected oil. The method of the application embodiment realizes efficient and accurate detection of physicochemical parameters of multiple brands of oil, breaks the limitation of "one brand one model", effectively improves the accuracy of the detection result and the generalization ability of cross-brand adaptation, greatly reduces the deployment and maintenance cost of the detection system, and significantly improves the practicality of the multiple-brand oil state monitoring.

Description

Technical Field

[0001] This invention relates to the fields of spectral analysis and industrial equipment condition monitoring technology, and in particular to a method, apparatus and equipment for detecting the physicochemical parameters of oil. Background Technology

[0002] Hyperspectral imaging technology can capture the optical response of oil in a continuous narrow band. Combined with chemometric methods, it can be used to invert various physicochemical properties of oil, such as viscosity, acid value, and metal content.

[0003] In related technologies, testing oils requires building a separate testing model for each brand. Standard spectral data for that brand under various conditions must be collected beforehand and entered into a dedicated reference library. Otherwise, spectral differences between brands can lead to mismatches and inversion errors, resulting in a "one brand, one model" dilemma. With hundreds or even thousands of lubricant brands and models on the market, building and maintaining a dedicated reference library for each brand under all conditions presents extremely high implementation difficulties in data acquisition, sample calibration, and system updates, making it almost impractical. Furthermore, differences in base oils and additives among different brands result in different intrinsic spectra, leading to insufficient generalization ability of existing testing algorithms and an inability to adapt to cross-brand testing needs. Summary of the Invention

[0004] This invention provides a method, apparatus, and device for detecting the physicochemical parameters of oil. Through a pre-trained spectral decoupling network, it accurately separates the common component feature vectors, which are strongly correlated with the health status of the oil and the brand-related base spectral components, from the hyperspectral data of the oil under test. This eliminates interference caused by brand differences at the source, enabling focused oil condition information. Simultaneously, it matches the common component feature vectors of the oil under test with a pre-built universal reference library, achieving efficient and accurate detection of the physicochemical parameters of oils from multiple brands. This breaks the limitation of "one model per brand," effectively improving the accuracy of the detection results and the generalization ability across brands, significantly reducing the deployment and maintenance costs of the detection system, and significantly enhancing the practicality of multi-brand oil condition monitoring.

[0005] This invention provides a method for detecting the physicochemical parameters of oil, comprising the following steps: Acquire hyperspectral data of the oil to be tested; The hyperspectral data is input into a pre-trained spectral decoupling network, which outputs the base spectral components related to the oil brand and the common component feature vectors that are unrelated to the oil brand but related to the health status of the oil. The common component feature vectors are matched with a pre-built general reference library to determine the physicochemical parameter values ​​of the oil to be tested; the general reference library contains common component feature vectors and corresponding physicochemical parameter values ​​of oil samples from multiple different brands.

[0006] According to the present invention, an oil physicochemical parameter detection method is provided, wherein the spectral decoupling network is obtained based on a hybrid loss function trained thereon, the hybrid loss function comprising: The first loss function is used to ensure the accuracy of hyperspectral data reconstruction. The second loss function is used to minimize the brand information in the common component feature vector; The third loss function is used to ensure that the feature vector of common components is strongly correlated with the health status of the oil.

[0007] According to the present invention, a method for detecting the physicochemical parameters of oil is provided, wherein the common component feature vector is used to characterize the spectral features corresponding to water, wear metal particles, oxidation products and external contaminants in the oil.

[0008] According to a method for detecting the physicochemical parameters of an oil according to the present invention, the method matches the common component feature vector with a pre-constructed general reference library to determine the physicochemical parameter values ​​of the oil to be tested, including: The distance between the common component feature vector and the common component feature vector of each oil sample in the general reference library is determined. The K nearest reference entries are selected, and a weighted calculation is performed based on the distance weight to obtain the physicochemical parameter values ​​of the oil to be tested.

[0009] According to the present invention, a method for detecting the physicochemical parameters of oil is provided, wherein the reference entries cover the complete state range of oil from brand new to deteriorated and failed.

[0010] According to the present invention, a method for detecting the physicochemical parameters of oil is provided, the method further includes: When the confidence level of the matching between the common component feature vector of the oil to be tested and the general reference library is lower than a preset threshold, the oil to be tested is determined to be a new brand oil, and hyperspectral data of at least two different health states of the new brand oil are collected. The spectral decoupling network is used to extract common component feature vectors from the hyperspectral data; The common component feature vectors and their corresponding physicochemical parameter values ​​are added as new reference entries to the general reference library.

[0011] According to the present invention, the method for detecting the physicochemical parameters of oil is provided in which the confidence level is calculated based on the distance between the common component feature vector of the oil to be tested and the common component feature vector of the nearest reference entry in the general reference library, and the anomaly degree of the corresponding base spectral component of the oil to be tested; wherein, the smaller the distance and the lower the anomaly degree of the base spectral component, the higher the matching confidence level.

[0012] The present invention also provides an oil physicochemical parameter detection device, comprising the following modules: The acquisition module is used to acquire hyperspectral data of the oil to be tested; The decoupling module is used to input the hyperspectral data into a pre-trained spectral decoupling network and output the base spectral components related to the oil brand and the common component feature vectors that are unrelated to the oil brand but related to the health status of the oil. The detection module is used to match the common component feature vector with a pre-built general reference library to determine the physicochemical parameter values ​​of the oil to be tested; the general reference library contains common component feature vectors and corresponding physicochemical parameter values ​​of oil samples from multiple different brands.

[0013] The present invention also provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the oil physicochemical parameter detection method as described above.

[0014] The present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the oil physicochemical parameter detection method as described above.

[0015] The present invention also provides a computer program product, including a computer program that, when executed by a processor, implements the oil physicochemical parameter detection method as described above.

[0016] This invention provides a method, apparatus, and device for detecting the physicochemical parameters of oil. Through a pre-trained spectral decoupling network, it accurately separates the common component feature vectors, which are strongly correlated with the health status of the oil and the brand-related base spectral components, from the hyperspectral data of the oil under test. This eliminates interference caused by brand differences at the source, enabling focused oil condition information. Simultaneously, it matches the common component feature vectors of the oil under test with a pre-built universal reference library, achieving efficient and accurate detection of the physicochemical parameters of oils from multiple brands. This breaks the limitation of "one model per brand," effectively improving the accuracy of the detection results and the generalization ability across brands, significantly reducing the deployment and maintenance costs of the detection system, and significantly enhancing the practicality of multi-brand oil condition monitoring. Attached Figure Description

[0017] To more clearly illustrate the technical solutions in this invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0018] Figure 1 This is a schematic flowchart of the oil physicochemical parameter detection method provided by the present invention.

[0019] Figure 2 This is a schematic diagram of the structure of the oil physicochemical parameter detection device provided by the present invention.

[0020] Figure 3 This is a schematic diagram of the structure of the electronic device provided by the present invention. Detailed Implementation

[0021] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of this invention. All other embodiments obtained by those skilled in the art based on the embodiments of this invention without creative effort are within the scope of protection of this invention.

[0022] The following is combined with Figures 1 to 3 The present invention describes a method, apparatus, and equipment for detecting the physicochemical parameters of oil.

[0023] Figure 1 This is a schematic flowchart of the oil physicochemical parameter detection method provided by the present invention, as shown below. Figure 1 As shown, the method includes the following: Step 101: Obtain the hyperspectral data of the oil to be tested.

[0024] Specifically, in this embodiment of the application, hyperspectral data of the oil to be tested is first acquired using a hyperspectral acquisition device.

[0025] Step 102: Input the hyperspectral data into the pre-trained spectral decoupling network and output the base spectral components related to the oil brand, as well as the common component feature vectors that are unrelated to the oil brand but related to the health status of the oil.

[0026] Specifically, after acquiring the hyperspectral data of the oil to be tested, this application inputs the hyperspectral data of the oil to be tested into a pre-trained spectral decoupling network, and outputs the base spectral components related to the oil brand, as well as the common component feature vectors that are unrelated to the oil brand but related to the health status of the oil. This achieves precise decoupling of spectral features and effectively eliminates interference information caused by brand differences.

[0027] Step 103: Match the common component feature vector with the pre-built general reference library to determine the physicochemical parameter values ​​of the oil to be tested; the general reference library contains the common component feature vectors and corresponding physicochemical parameter values ​​of oil samples from multiple different brands.

[0028] Specifically, in this embodiment, a pre-built universal reference library stores common component feature vectors for multiple oil samples from different brands, and each feature vector is associated with a corresponding physicochemical parameter value of the oil. After decoupling the hyperspectral data of the oil to be tested through a spectral decoupling network to obtain common component feature vectors that are independent of the oil brand and only associated with the health status of the oil, these common component feature vectors can be matched with the universal reference library to accurately deduce the physicochemical parameter values ​​of the oil to be tested. This process enables universal detection of physicochemical parameters of oils from multiple brands, breaking the limitation of "one model per brand," effectively improving the accuracy of detection and cross-brand generalization ability, and significantly reducing the deployment and maintenance costs of the detection system.

[0029] For example, this application uses a pre-trained spectral decoupling network to extract the hyperspectral signal of the oil. Decoupling is achieved by the superposition of three core components: in, The brand's base spectrum is determined by the type of base oil, refining process, and basic additive system. It is a characteristic of the brand's identity, but it is weakly correlated with changes in the state of the oil, such as contamination and deterioration. This section describes the characteristic spectra of common components. It includes spectral variations caused by state-related components such as water, abrasive metal particles, oxidation products, and external contaminants. These characteristics are cross-brand and strongly correlated with oil health. This application uses noise from the spectrum of the target spectrum. Inhibition Enhance and extract Then utilize public, brand-agnostic resources. The characteristics are used to perform state inversion to obtain the physicochemical parameter values ​​of the oil to be tested.

[0030] The method described in the above embodiments, through a pre-trained spectral decoupling network, accurately separates the common component feature vectors that are strongly correlated with the health status of the oil from the hyperspectral data of the oil to be tested. This eliminates the interference caused by brand differences at the source and achieves the focus of oil status information. At the same time, the common component feature vectors of the oil to be tested are matched with a pre-built general reference library, enabling efficient and accurate detection of physicochemical parameters of oils from multiple brands. This breaks the limitation of "one model per brand," effectively improves the accuracy of the detection results and the generalization ability of cross-brand adaptation, significantly reduces the deployment and maintenance costs of the detection system, and significantly improves the practicality of multi-brand oil status monitoring.

[0031] In some embodiments, the spectral decoupling network is trained based on a hybrid loss function, which includes: The first loss function is used to ensure the accuracy of hyperspectral data reconstruction. The second loss function is used to minimize the brand information in the common component feature vector; The third loss function is used to ensure that the feature vector of common components is strongly correlated with the health status of the oil.

[0032] Specifically, in this embodiment of the application, to achieve accurate decoupling of the hyperspectral data of the oil to be tested, a spectral decoupling network is trained based on a hybrid loss function. Optionally, the hybrid loss function includes a first loss function, a second loss function, and a third loss function. The first loss function minimizes the error between the hyperspectral data of the oil to be tested and the decoupled brand-based spectral components and common component feature vectors, ensuring that the two types of components after decoupling can completely restore the original spectral information. This guarantees the accuracy of hyperspectral data reconstruction and the integrity and accuracy of the brand-based spectral and common component feature vectors, preventing information gaps caused by decoupling from affecting subsequent detection accuracy. Optionally, the second loss function minimizes the brand information in the common component feature vectors until the classifier cannot distinguish the brand category from the common component feature vectors. This effectively eliminates brand interference in the common component feature vectors, solves the generalization problem caused by the intrinsic spectral differences of different brands of oil, ensures the comparability of the state characteristics of cross-brand oils, and provides unified feature dimension support for the matching logic of a general reference library. Optionally, the third loss function minimizes the error between the predicted physicochemical parameter values ​​based on the common component feature vector and the actual physicochemical parameter values, so that the common component feature vector can accurately encode information related to the health status of the oil. This ensures that the common component feature vector focuses on core information strongly related to the health status of the oil, such as water content, wear metals, and oxidation products, and avoids the inclusion of irrelevant and redundant features in the common component feature vector, thereby ensuring that the common component feature vector has the ability to effectively invert the physicochemical parameters of the oil.

[0033] For example, in the embodiments of this application, the spectral decoupling network is constructed and trained in the following manner: Design a deep convolutional neural network as a spectral decoupling network. Its input is the original hyperspectral curve S(λ) or its preprocessed data, and its output is the brand-based estimated spectrum. and common component eigenvectors , This represents a low-dimensional vector that encodes state information. Optionally, the spectral decoupling network is trained using a large-scale oil spectral dataset containing multiple brands and states. Its loss function includes loss function 1, loss function 2, and loss function 3, where loss function 1 is the reconstruction loss, used to ensure... +Decoder( It can reconstruct the input spectrum S(λ); the loss function 2 is the brand adversarial loss, which is achieved through... Then a brand classifier is added and trained through a gradient inversion layer, so that... The text avoids including brand information as much as possible; loss function 3 is a state regression loss, which trains a regression model using known, partially labeled physicochemical values, such as iron content, from... The regressor to the target parameter ensures It contains valid state information. That is, this application ensures that the decoupled brand base spectrum and common component feature vector can completely restore the original hyperspectral information through reconstruction loss, avoiding the loss of key optical information and laying the data foundation for subsequent detection; brand adversarial loss accurately removes brand interference in the common component feature vector through gradient inversion and game with the brand classifier, making the features comparable across brands, fundamentally solving the industry dilemma of "one model for one brand"; state regression loss forces the common component feature vector to focus on the core health state information such as oil moisture and worn metal, improving the effectiveness of physicochemical parameter inversion.

[0034] The method described above trains the spectral decoupling network using a hybrid loss function, enabling the trained spectral decoupling network to accurately separate brand-related base spectral components from common component feature vectors that are strongly correlated with the health status of the oil. This significantly improves detection accuracy and cross-brand generalization ability, and greatly reduces the deployment and maintenance costs of multi-brand oil detection systems.

[0035] In some embodiments, matching the common component feature vectors with a pre-built universal reference library to determine the physicochemical parameter values ​​of the oil to be tested includes: Determine the distance between the common component feature vector and the common component feature vector of each oil sample in the general reference library, select the K nearest reference entries, and perform weighted calculation based on the distance weight to obtain the physicochemical parameter values ​​of the oil to be tested.

[0036] Specifically, in this embodiment, the distance between the common component feature vector of the oil to be tested and the common component feature vectors of all oil samples in the general reference library is first calculated one by one. Optionally, the distance can be calculated using common methods such as Euclidean distance or cosine distance. Then, all calculated vector space distances are sorted in ascending order, and the top K reference entries with the smallest distances are selected. The physicochemical parameter values ​​of the K reference entries are then weighted and summed to obtain the physicochemical parameter values ​​of the oil to be tested, such as metal content and acid value. In other words, this application significantly improves the accuracy of physicochemical parameter inversion through precise feature matching and weighted inversion, effectively ensuring the reliability of the test results, completely breaking the limitation of "one model per brand," and achieving efficient and accurate detection of physicochemical parameters of multi-brand oils.

[0037] For example, in this embodiment of the application, a reference library is constructed to store the common component feature vector and its corresponding precise physicochemical parameter values ​​for each reference sample. These reference samples may come from different brands but cover a wide range of states. Optionally, for the oil to be tested, its common component feature vector is extracted using a trained spectral decoupling network. And calculate the common component feature vector of the oil to be tested. With all in the reference library The parameters of the oil under test are derived by using a distance-weighted average based on the physicochemical values ​​of the nearest reference sample, according to the vector space distance.

[0038] The method described in the above embodiments significantly improves the accuracy of physicochemical parameter inversion through precise feature matching and weighted inversion, effectively ensuring the reliability of the test results, completely breaking the limitation of "one model for one brand", and realizing efficient and accurate detection of physicochemical parameters of oils from multiple brands.

[0039] In some embodiments, the reference entries cover the full range of states of the oil, from brand new to deteriorated and failed.

[0040] Specifically, in constructing the universal reference library, this application selects multiple oils from different brands as reference sample sources. For each brand, samples covering the entire life cycle of the oil are collected, including new, slightly deteriorated, moderately deteriorated, severely deteriorated, and deteriorated to failure states. Hyperspectral data of these samples are collected, and a pre-trained spectral decoupling network is used to extract brand-independent common component feature vectors. Simultaneously, accurate physicochemical parameter values ​​corresponding to the sample are obtained through standard testing methods. This effectively constructs the universal reference library, enabling the test oil to select the K nearest reference entries within the corresponding state range from the universal reference library, regardless of whether it is new, deteriorating, or nearing failure. This provides targeted reference for weighted calculations, effectively eliminates blind spots in state matching, enhances the accuracy and stability of physicochemical parameter inversion, and further improves the reliability of the test results.

[0041] The method described in the above embodiments eliminates the matching blind spot of the oil under test by allowing the reference entries of the universal reference library to cover the entire life cycle of the oil from a brand new state to a deteriorated and failed state. This ensures that no matter what health state the oil under test is in, the reference entries within the corresponding state range can be selected from the reference library, thereby enhancing the accuracy and stability of the physicochemical parameter inversion and effectively improving the reliability of the test results.

[0042] In some embodiments, when the confidence level of the matching between the common component feature vector of the oil to be tested and the general reference library is lower than a preset threshold, the oil to be tested is determined to be a new brand oil, and hyperspectral data of at least two different health states of the new brand oil are collected. A spectral decoupling network is used to extract common component feature vectors from hyperspectral data. The common component feature vectors and their corresponding physicochemical parameter values ​​are added as new reference entries to the general reference library.

[0043] Specifically, after extracting the common component feature vector of the oil to be tested, this application can calculate the matching confidence level between the vector and all reference entries in the general reference library. If the confidence level is lower than a preset threshold, the oil is determined to be a new brand oil not covered by the general reference library, triggering the new brand adaptation process. Optionally, at least two samples of the new brand oil in different health states can be collected, such as brand new state, moderately deteriorated state, or slightly deteriorated and severely deteriorated state, and hyperspectral data of each sample can be obtained through a hyperspectral acquisition device.

[0044] Optionally, the hyperspectral data of the collected samples from various states of the new brand can be input into a pre-trained spectral decoupling network to obtain the base spectral components related to the new brand, as well as the common component feature vectors that are unrelated to the brand and only associated with the health status. Further, precise physicochemical parameter values ​​corresponding to each sample of the new brand, such as metal content, acid value, and viscosity, are obtained through standard testing methods. The common component feature vector and corresponding physicochemical parameter values ​​of each sample are then used as new reference entries and added to a general reference library to complete the registration of reference entries for the new brand's oil. This significantly reduces the adaptation cost for new brands, overcomes the limitation of "one model per brand," and improves the reliability and accuracy of the testing results.

[0045] The method described in the above embodiments can accurately identify new brand oils not covered by the general reference library by judging the matching confidence of the common component feature vector of the oil to be tested with the general reference library. By extracting the common component feature vectors corresponding to at least two samples of the new brand oil in different health states, the reference entries of the new brand oil can be registered, which greatly reduces the time and cost of adapting to new brands, breaks through the limitation of "one brand, one model", significantly reduces the deployment and long-term maintenance costs of multi-brand oil detection systems, and improves the reliability and accuracy of detection results.

[0046] In some embodiments, the confidence level is calculated by weighting the distance between the common component feature vector of the oil to be tested and the common component feature vector of the nearest reference entry in the general reference library, and the anomaly of the corresponding base spectral component of the oil to be tested; wherein, the smaller the distance and the lower the anomaly of the base spectral component, the higher the matching confidence level.

[0047] Specifically, in this embodiment, the confidence level is calculated by weighting the distance between the common component feature vector of the oil to be tested and the common component feature vector of the nearest reference entry in the general reference library, and the anomaly of the corresponding substrate spectral component of the oil to be tested. This effectively ensures the accuracy of identifying new brand oils and avoids misjudgment or missed judgment. Optionally, in this embodiment, the smaller the distance between the common component feature vector of the oil to be tested and the common component feature vector of the nearest reference entry in the general reference library, the more accurately the oil to be tested can be detected by the reference library; the larger the distance, the more likely the oil to be tested is a new brand that is not yet covered. Optionally, the anomaly of the corresponding substrate spectral component of the oil to be tested is the degree of difference between the corresponding substrate spectral component of the oil to be tested and the substrate spectrum of existing oil brands. The lower the anomaly, the closer the brand characteristics of the oil to be tested are to existing brands; conversely, the higher the anomaly, the more likely it is a new brand oil.

[0048] For example, in the embodiments of this application, while inverting the physicochemical parameters of the oil, the common component feature vector of the oil to be tested is used as a basis. Common component eigenvectors with nearest neighbor reference entries in the general reference library The distance, and the substrate spectrum of the brand corresponding to the oil being tested. The system outputs a confidence score for anomalies. When the confidence score of a certain oil sample is detected to be lower than a preset threshold, the system prompts that "brand registration" is required. By collecting the spectra of a few (e.g., 2-3) standard samples of the new brand in known states, such as the spectra of a brand new oil sample and a moderately contaminated oil sample, the system extracts the brand's basic features using a pre-trained spectral decoupling network. The common components are stored in the brand base library. Simultaneously, using transfer learning or domain adaptation techniques, some network layers of the common feature extraction network are fine-tuned to more accurately decouple the spectral data of the new brand's oil. Finally, the feature vectors of these new samples are... Physicochemical values ​​are added as new reference entries to the general reference library. This process eliminates the need to retrain the entire model or collect full-state spectral data for the brand, achieving a paradigm shift from "one brand, one model" to "adaptive new brand using a small number of samples".

[0049] The method described above determines the confidence level based on the distance between the common component feature vector of the oil to be tested and the common component feature vector of the nearest reference entry in the general reference library, as well as the anomaly of the substrate spectral component corresponding to the oil to be tested. This effectively ensures the accuracy of identifying new brand oils, avoids misjudgment or omission, significantly reduces the deployment and long-term maintenance costs of multi-brand oil detection systems, and ensures the reliability and stability of the detection results.

[0050] The following describes the oil physicochemical parameter detection device provided by the present invention. The oil physicochemical parameter detection device described below can be referred to in correspondence with the oil physicochemical parameter detection method described above. Figure 2 As shown, the oil physicochemical parameter detection device includes: The acquisition module 210 is used to acquire the hyperspectral data of the oil to be tested; The decoupling module 220 is used to input hyperspectral data into a pre-trained spectral decoupling network and output the base spectral components related to the oil brand and the common component feature vectors that are unrelated to the oil brand but related to the health status of the oil. The detection module 230 is used to match the common component feature vector with a pre-built general reference library to determine the physicochemical parameter values ​​of the oil to be tested; the general reference library contains the common component feature vectors and corresponding physicochemical parameter values ​​of oil samples from multiple different brands.

[0051] Figure 3 An example is a schematic diagram of the physical structure of an electronic device, such as... Figure 3 As shown, the electronic device may include a processor 310, a communication interface 320, a memory 330, and a communication bus 340. The processor 310, communication interface 320, and memory 330 communicate with each other via the communication bus 340. The processor 310 can call logical instructions in the memory 330 to execute a method for detecting the physicochemical parameters of an oil. This method includes: acquiring hyperspectral data of the oil to be tested; inputting the hyperspectral data into a pre-trained spectral decoupling network, outputting base spectral components related to the oil brand, and common component feature vectors unrelated to the oil brand but related to the oil's health status; matching the common component feature vectors with a pre-built general reference library to determine the physicochemical parameter values ​​of the oil to be tested; the general reference library contains common component feature vectors and corresponding physicochemical parameter values ​​of oil samples from multiple different brands.

[0052] Furthermore, the logical instructions in the aforementioned memory 330 can be implemented as software functional units and, when sold or used as independent products, can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, essentially, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0053] On the other hand, the present invention also provides a computer program product, which includes a computer program that can be stored on a non-transitory computer-readable storage medium. When the computer program is executed by a processor, the computer can execute the oil physicochemical parameter detection method provided by the above methods. The method includes: acquiring hyperspectral data of the oil to be tested; inputting the hyperspectral data into a pre-trained spectral decoupling network, and outputting a base spectral component related to the oil brand and a common component feature vector unrelated to the oil brand but related to the health status of the oil; matching the common component feature vector with a pre-constructed general reference library to determine the physicochemical parameter values ​​of the oil to be tested; the general reference library contains common component feature vectors and corresponding physicochemical parameter values ​​of oil samples from multiple different brands.

[0054] In another aspect, the present invention also provides a non-transitory computer-readable storage medium storing a computer program thereon. When executed by a processor, the computer program implements the oil physicochemical parameter detection method provided by the above methods. The method includes: acquiring hyperspectral data of the oil to be tested; inputting the hyperspectral data into a pre-trained spectral decoupling network, outputting a base spectral component related to the oil brand, and a common component feature vector unrelated to the oil brand but related to the health status of the oil; matching the common component feature vector with a pre-constructed general reference library to determine the physicochemical parameter values ​​of the oil to be tested; the general reference library contains common component feature vectors and corresponding physicochemical parameter values ​​of oil samples from multiple different brands.

[0055] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Those skilled in the art can understand and implement this without any creative effort.

[0056] Through the above description of the embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus necessary general-purpose hardware platforms, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solutions, in essence or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods described in the various embodiments or some parts of the embodiments.

[0057] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for detecting the physicochemical parameters of oil, characterized in that, include: Acquire hyperspectral data of the oil to be tested; The hyperspectral data is input into a pre-trained spectral decoupling network, which outputs the base spectral components related to the oil brand and the common component feature vectors that are unrelated to the oil brand but related to the health status of the oil. The common component feature vectors are matched with a pre-built general reference library to determine the physicochemical parameter values ​​of the oil to be tested; the general reference library contains common component feature vectors and corresponding physicochemical parameter values ​​of oil samples from multiple different brands. The spectral decoupling network is trained based on a hybrid loss function, which includes a first loss function that minimizes the error between the hyperspectral data and the decoupled brand base spectral components and common component feature vectors to ensure the accuracy of hyperspectral data reconstruction. The second loss function minimizes the brand information in the common component feature vector by connecting a brand classifier after the common component feature vector and training it through a gradient reversal layer. The third loss function minimizes the error between the predicted physicochemical parameter values ​​based on the common component feature vector and the actual physicochemical parameter values, ensuring that the common component feature vector is strongly correlated with the health status of the oil. The method further includes: When the confidence level of the matching between the common component feature vector of the oil to be tested and the general reference library is lower than a preset threshold, the oil to be tested is determined to be a new brand oil, and hyperspectral data of at least two different health states of the new brand oil are collected. The spectral decoupling network is used to extract common component feature vectors from the hyperspectral data; The common component feature vectors and their corresponding physicochemical parameter values ​​are added as new reference entries to the general reference library; The matching confidence is calculated by weighting the distance between the common component feature vector of the oil to be tested and the common component feature vector of the nearest reference entry in the general reference library, and the anomaly of the corresponding base spectral component of the oil to be tested; wherein, the smaller the distance and the lower the anomaly of the base spectral component, the higher the matching confidence.

2. The method for detecting the physicochemical parameters of oil according to claim 1, characterized in that, The physicochemical parameter values ​​of the oil to be tested are determined by matching the feature vector of the common components with a pre-built general reference library, including: The distance between the common component feature vector and the common component feature vector of each oil sample in the general reference library is determined. The K nearest reference entries are selected, and a weighted calculation is performed based on the distance weight to obtain the physicochemical parameter values ​​of the oil to be tested.

3. The method for detecting the physicochemical parameters of oil according to claim 2, characterized in that, The reference entries cover the complete range of states of the oil, from brand new to deteriorated and failed.

4. An oil physicochemical parameter detection device, used to implement the oil physicochemical parameter detection method as described in any one of claims 1 to 3, characterized in that, include: The acquisition module is used to acquire hyperspectral data of the oil to be tested; The decoupling module is used to input the hyperspectral data into a pre-trained spectral decoupling network and output the base spectral components related to the oil brand and the common component feature vectors that are unrelated to the oil brand but related to the health status of the oil. The detection module is used to match the common component feature vector with a pre-built general reference library to determine the physicochemical parameter values ​​of the oil to be tested; the general reference library contains common component feature vectors and corresponding physicochemical parameter values ​​of oil samples from multiple different brands.

5. An electronic device comprising a memory, a processor, and a computer program stored in the memory and running on the processor, characterized in that, When the processor executes the computer program, it implements the method for detecting the physical and chemical parameters of oil as described in any one of claims 1 to 3.

6. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the method for detecting the physical and chemical parameters of oil as described in any one of claims 1 to 3.

Images