Grease detection data analysis system and method based on multi-source data

Abstract

The invention relates to the technical field of grease detection, in particular to a grease detection data analysis system and method based on multi-source data. The method comprises the following steps: acquiring grease multi-source detection data, and performing near infrared spectrum detection feature extraction according to the grease multi-source detection data so as to obtain infrared spectrum detection data; performing signal enhancement processing on the infrared spectrum detection data to generate enhanced infrared spectrum data; carrying out physicochemical index feature extraction based on the grease multi-source detection data so as to obtain physicochemical index data; performing detection anomaly analysis according to the physicochemical index data and the enhanced infrared spectrum data so as to obtain grease anomaly detection data; and performing type division on the grease anomaly detection data so as to obtain grease chemical component anomaly data and grease microorganism anomaly data. Based on the grease detection technology, the abnormal recognition rate and the quality evaluation accuracy of grease detection are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of oil detection, and particularly relates to an oil detection data analysis system and method based on multi-source data. Background Art

[0002] Traditional methods do not fully utilize the complementarity and complexity among multi-source oil data, resulting in relatively simple data processing steps. For example, they fail to deeply fuse multi-source data, or lack refined steps such as sufficient signal enhancement, noise filtering, and frequency band analysis when processing infrared spectrum data. This leads to low data analysis accuracy, especially in key steps such as detecting anomalies and evaluating quality. The anomaly detection of traditional methods generally only focuses on data from a single source and does not fully combine different types of data for multi-dimensional analysis. This limitation makes the detected anomaly information incomplete or not in-depth enough to reveal potential risks in oil quality. Traditional methods often ignore the comprehensiveness of oil chemical composition and microbial changes, or adopt overly simple detection means. This results in inaccurate prediction of oil quality changes, especially in the processes of oil aging, oxidation, etc., where subtle component or microbial anomalies cannot be detected in a timely manner. Summary of the Invention

[0003] Based on this, it is necessary for the present invention to provide an oil detection data analysis system and method based on multi-source data to solve at least one of the above technical problems.

[0004] To achieve the above object, an oil detection data analysis method based on multi-source data includes the following steps:

[0005] Step S1: Obtain multi-source oil detection data, and perform near-infrared spectrum detection feature extraction according to the multi-source oil detection data to obtain infrared spectrum detection data; perform signal enhancement processing on the infrared spectrum detection data to generate enhanced infrared spectrum data;

[0006] Step S2: Perform physical and chemical index feature extraction based on the multi-source oil detection data to obtain physical and chemical index data; perform detection anomaly analysis according to the physical and chemical index data and the enhanced infrared spectrum data to obtain oil anomaly detection data;

[0007] Step S3: Classify the oil anomaly detection data to obtain oil chemical composition anomaly data and oil microbial anomaly data; perform oxygen contact analysis according to the oil chemical composition anomaly data and the oil microbial anomaly data to obtain oxygen contact data;

[0008] Step S4: Analyze the oxidation reaction based on the oxygen contact data to obtain the oil oxidation reaction data; evaluate the oil aging according to the oil oxidation reaction data to obtain the oil aging degree data, and construct an oil aging evaluation model to obtain the oil aging evaluation model;

[0009] Step S5: Evaluate the quality of the multi-source detection data of the oil according to the oil aging evaluation model to obtain the oil quality detection and evaluation data.

[0010] The present invention can provide more comprehensive original data through the acquisition of multi-source detection data of oils and the extraction of near-infrared spectral features, enabling subsequent analysis to have higher basic data support. Enhancing the infrared spectral signal processing further improves the data quality. Especially when dealing with weak signals and noise interference, it can effectively enhance the signal recognition, thus providing higher precision for anomaly detection. By combining the physical and chemical index data with the enhanced infrared spectral data for anomaly analysis, the method can evaluate the anomalies of oils from multiple dimensions, making up for the limitation of traditional methods that only focus on a single data source, and improving the accuracy and comprehensiveness of anomaly detection. The classification of the types of oil anomaly data helps to further refine the analysis, especially the separate processing of chemical composition and microbial anomalies, making the identification of oil quality more accurate. At the same time, the analysis of oxygen contact provides key data support for subsequent oxidation reaction prediction, helping to reveal the subtle changes that occur during the aging process of oils. By analyzing the oxidation reaction data and combining with the oil aging evaluation, it is possible to more effectively predict the quality changes of oils during long-term use or storage. Finally, the establishment of the oil aging evaluation model can achieve accurate prediction and detection of oil quality, considering multiple factors comprehensively, avoiding the deficiencies of traditional methods that often rely on a single feature or data source for evaluation, and thus realizing more efficient and accurate oil quality evaluation.

[0011] Optionally, step S1 is specifically as follows:

[0012] Step S11: Obtain multi-source detection data of oils and extract near-infrared spectral detection features according to the multi-source detection data of oils to obtain infrared spectral detection data;

[0013] Step S12: Perform noise filtering on the infrared spectral detection data to obtain noise-reduced infrared spectral detection data;

[0014] Step S13: Smooth the spectrum according to the noise-reduced infrared spectral detection data to obtain smoothed infrared spectral detection data;

[0015] Step S14: Amplify the signal of the smoothed infrared spectral detection data to generate enhanced infrared spectral data.

[0016] The present invention provides high-quality basic data for subsequent analysis through the acquisition of multi-source detection data of oils and fats and the extraction of near-infrared spectral features. Through noise filtering processing, the method can effectively reduce the noise interference in the infrared spectral data, retain important signal components, thereby improving the clarity and reliability of the data, which lays a more solid foundation for subsequent analysis. The spectral smoothing step further optimizes the data, suppresses signal fluctuations, thus avoiding errors caused by excessive fluctuations, ensuring the stability of the signal, and enhancing the effectiveness of the signal in subsequent processing. The signal enhancement and amplification step ensures that weak or difficult-to-identify signals are highlighted by increasing the signal intensity, further improving the distinguishability of the data and the accuracy of analysis. Through this series of processing steps, the finally obtained enhanced infrared spectral data not only improves the signal quality but also optimizes the data accuracy in subsequent detection and analysis processes, thereby effectively improving the accuracy of key tasks such as oil and fat quality evaluation and anomaly detection, avoiding the limitations of rough data processing and single data source analysis in traditional methods, and comprehensively enhancing the comprehensive evaluation ability of oil and fat quality.

[0017] Optionally, step S2 is specifically as follows:

[0018] Step S21: Extract physical and chemical index features based on multi-source detection data of oils and fats to obtain physical and chemical index data;

[0019] Step S22: Conduct component anomaly analysis on the physical and chemical index data to obtain component anomaly data;

[0020] Step S23: Conduct frequency band anomaly analysis on the enhanced infrared spectral data to obtain frequency band anomaly data;

[0021] Step S24: Integrate oil and fat anomaly detection based on the component anomaly data and the frequency band anomaly data to obtain oil and fat anomaly detection data.

[0022] By comprehensively utilizing the advantages of multi-source data of oils and fats, the present invention realizes the accurate analysis of the quality of oils and fats. First, through the extraction of physicochemical index characteristics, the physicochemical property data of oils and fats are obtained, which provide detailed basic information for subsequent anomaly detection. The component anomaly analysis can effectively identify potential abnormal components in the oils and fats, timely detect quality changes caused by chemical reactions or microbial activities, and avoid the limitation of ignoring component changes in traditional methods. The frequency band anomaly analysis focuses on enhancing the frequency band characteristics in the infrared spectrum data, deeply excavating potential abnormal information, especially the weak anomalies existing in the infrared spectrum response frequency band of the oils and fats. Through the integrated analysis of the component anomaly data and the frequency band anomaly data, the quality problems existing in the oils and fats can be comprehensively identified, improving the accuracy and depth of anomaly detection. Finally, through the integration of the anomaly detection data of the oils and fats, the limitation of relying only on a single data source in traditional methods is avoided, realizing comprehensive anomaly identification from multiple dimensions and angles, thereby significantly improving the accuracy of the quality prediction and anomaly warning of the oils and fats. It improves the comprehensiveness, accuracy and real-time performance of the quality management of the oils and fats, and can more effectively reveal the potential quality risks of the oils and fats.

[0023] Optionally, step S22 is specifically as follows:

[0024] Step S221: Perform acid value anomaly detection on the physicochemical index data to obtain acid value anomaly data;

[0025] Step S222: Perform peroxide value anomaly detection on the physicochemical index data to obtain peroxide value anomaly data;

[0026] Step S223: Perform clustering of component anomaly indexes based on the acid value anomaly data and the peroxide value anomaly data to generate component anomaly indexes;

[0027] Step S224: Perform component anomaly extraction on the physicochemical index data according to the component anomaly indexes, so as to obtain component anomaly data.

[0028] Through the abnormal acid value detection, the present invention can timely identify the acidity change in oils and fats, which is crucial for predicting the oxidation process and spoilage risk of oils and fats. Especially under long-term storage or high-temperature conditions, the change in acid value indicates quality problems. The abnormal peroxide value detection can accurately identify the peroxide content in oils and fats. The generation of peroxides is an important sign of oil oxidation, and its change directly affects the shelf life and edible safety of oils and fats. By clustering the abnormal data of acid value and peroxide value, it is possible to effectively comprehensively evaluate the quality abnormality of oils and fats from multiple component perspectives, thereby improving the comprehensiveness and accuracy of abnormal detection. The aggregation and extraction steps of component abnormal indicators further improve the abnormal detection process. By combining different abnormal indicators, it enhances the sensitivity to the changes in the chemical composition of oils and fats and can reveal more detailed quality changes, avoiding the limitations of ignoring some potential problems in traditional methods. The advantage of the whole process is that by deeply integrating multi-source data, comprehensively analyzing various chemical components in oils and fats, it accurately identifies potential quality hazards, thereby enhancing the reliability and real-time nature of oil and fat quality detection and ensuring the quality consistency of oil and fat products in the market.

[0029] Optionally, step S23 is specifically as follows:

[0030] Step S231: Decompose the enhanced infrared spectrum data by frequency band to obtain the enhanced infrared spectrum data in the high-frequency region and the enhanced infrared spectrum data in the low-frequency region;

[0031] Step S232: Detect the peak and valley positions of the first derivative spectrum based on the enhanced infrared spectrum data in the high-frequency region to obtain the peak and valley position data of the high-frequency region spectrum;

[0032] Step S233: Separate the Lorentzian-fitted overlapping peaks based on the enhanced infrared spectrum data in the low-frequency region to obtain the overlapping peak data in the low-frequency region;

[0033] Step S234: Detect the peak position offset of the peak and valley position data of the high-frequency region spectrum to obtain the peak position offset data of the high-frequency region;

[0034] Step S235: Analyze the fitting residuals of the overlapping peak data in the low-frequency region to obtain the high-residual data in the low-frequency region;

[0035] Step S236: Integrate the frequency band anomalies according to the peak position offset data of the high-frequency region and the high-residual data of the low-frequency region to obtain the frequency band anomaly data.

[0036] Through the refined processing of enhanced infrared spectral data, the present invention significantly improves the accuracy and meticulousness of frequency band anomaly detection. The frequency band decomposition step divides the infrared spectral data into a high-frequency region and a low-frequency region, enabling the data of each frequency band to be analyzed specifically, which provides a more accurate basis for subsequent signal processing. In the spectral peak and valley position detection step of the high-frequency region, through the first derivative analysis, the peak and valley positions in the spectrum are effectively identified, helping to capture subtle chemical composition changes. This is crucial for precisely locating the quality anomalies of oils and fats, especially in the subtle fluctuations involved in the high-frequency region. The separation of overlapping peaks in the low-frequency region uses the Lorentz fitting technique to effectively separate the overlapping spectral peaks, avoiding misjudgments in traditional methods, improving the accuracy of low-frequency region analysis, and ensuring the clarity and accuracy of low-frequency signals. For the detection of peak position shift in the high-frequency region, it can accurately identify the subtle changes in the peak positions in the spectrum, revealing the changes that occur in oils and fats under different storage or processing conditions, which helps to detect quality problems in advance. The fitting residual analysis in the low-frequency region can identify those signal components that cannot be fully fitted by analyzing the fitting error, and timely discover potential abnormal signals. This analysis method can effectively reveal the subtle changes in the oil and fat components, especially those not captured by conventional detection methods. Finally, the frequency band anomaly integration step integrates the high-frequency region peak position shift data and the low-frequency region high residual data, comprehensively improving the depth and breadth of oil and fat anomaly detection. It not only increases the detection dimension but also strengthens the complementarity between different data sources, thereby being able to more accurately identify the quality risks and potential problems of oils and fats, significantly improving the reliability and accuracy of oil and fat quality assessment.

[0037] Optionally, step S232 is specifically as follows:

[0038] Perform a first derivative calculation based on the enhanced infrared spectral data in the high-frequency region to obtain first derivative data;

[0039] Identify the zero-crossing points of the first derivative data to obtain zero-crossing point data;

[0040] Perform a second derivative calculation based on the zero-crossing point data to obtain second derivative data;

[0041] Perform division based on the second derivative data. If the second derivative data is greater than zero, it is judged as a peak position to obtain peak data; if the second derivative data is less than zero, it is judged as a valley position to obtain valley data;

[0042] Integrate the high-frequency region spectral peak and valley positions based on the peak data and the valley data to obtain the high-frequency region spectral peak and valley position data.

[0043] The present invention enhances infrared spectral data by gradually processing the high-frequency region, significantly improving the ability to identify abnormal oil quality. The first derivative calculation helps capture subtle changes in spectral data, especially in the high-frequency region, and can accurately reveal tiny fluctuations. By identifying the zero-crossing points in the first derivative data, the change turning points of the spectrum can be effectively located, helping to distinguish different features in the spectrum and providing a more accurate analysis basis. Further performing the second derivative calculation can strengthen the morphological analysis of the spectral curve, especially helping to identify those imperceptible change trends. When the second derivative is greater than zero, it is determined as the peak position, and when it is less than zero, it is determined as the valley position. This judgment method can accurately determine the peaks and valleys in the spectrum, thereby refining the analysis of spectral features and helping to capture the changes and abnormalities in the components of the oil. Finally, the integration of the peak and valley data provides a complete and accurate identification result for the peak-valley position data of the high-frequency region spectrum, and can better reveal the potential risks of oil quality. Through this precise peak-valley position identification and data integration process, the refinement degree of data analysis is enhanced, the accuracy of anomaly detection is improved, and stronger support is provided for the in-depth analysis of oil quality.

[0044] Optionally, step S233 is specifically as follows:

[0045] Partition the enhanced infrared spectral data in the low-frequency region to obtain the enhanced infrared spectral partition data in the low-frequency region;

[0046] Set the Lorentz function parameters for the enhanced infrared spectral partition data in the low-frequency region to obtain the Lorentz function parameters;

[0047] Perform single-peak interval fitting on the enhanced infrared spectral partition data in the low-frequency region according to the Lorentz function parameters to obtain the single-peak fitting data;

[0048] Perform multi-peak combination on the enhanced infrared spectral data in the low-frequency region according to the single-peak fitting data to obtain the overlapping peak data;

[0049] Perform fitting calculation on the overlapping peak data to obtain the overlapping peak data in the low-frequency region.

[0050] By partitioning the enhanced infrared spectral data in the low-frequency region, the present invention can effectively split complex spectral data into regions with different characteristics, which helps to extract more accurate analysis information from different frequency bands. On this basis, by setting the Lorentz function parameters and performing refined fitting on the data, it is possible to provide an in-depth understanding of the spectral curve morphology, especially in the complex low-frequency region, to help distinguish multiple potential component changes. Further fitting of the single-peak interval can accurately capture the single peak in the spectrum, avoid the influence of noise in complex data, and improve the accuracy of fitting. When faced with multiple overlapping peaks, a multi-peak combination method is used for processing, which can effectively extract independent component information from the overlapping spectra and reveal more detailed oil chemical characteristics. By performing fitting calculations on the overlapping peak data, overlapping peak data in the low-frequency region is obtained, which makes the spectral analysis more accurate and can effectively distinguish complex component changes and potential quality problems. It improves the accuracy and depth of data analysis, especially in the monitoring of oil quality, anomaly detection, and trend analysis of changes, providing more comprehensive and accurate support.

[0051] Optionally, step S3 is specifically as follows:

[0052] Step S31: Obtain standard data of oil components;

[0053] Step S32: Classify the oil anomaly detection data according to the standard data of oil components, so as to obtain oil chemical composition anomaly data and oil microorganism anomaly data;

[0054] Step S33: Screen oil samples based on the oil chemical composition anomaly data to obtain oil chemical composition anomaly samples;

[0055] Step S34: Measure the peroxide content of the oil chemical composition anomaly samples to obtain peroxide data;

[0056] Step S35: Screen oil samples based on the oil microorganism anomaly data to obtain oil microorganism anomaly samples;

[0057] Step S36: Measure the lipase activity of the oil microorganism anomaly samples to obtain lipase activity data;

[0058] Step S37: Evaluate the oxygen contact of the oil based on the peroxide data and the lipase activity data to obtain oxygen contact data.

[0059] The present invention provides a scientific basis for subsequent anomaly detection and classification by obtaining standard data of oil components, ensuring the accuracy and pertinence of the detection process. In the type classification step, by classifying oil anomaly detection data into two categories: chemical composition anomalies and microbial anomalies, the changes in oils can be analyzed more meticulously, avoiding the overly simple single analysis mode in traditional methods. Subsequently, based on the chemical composition anomaly data, oil samples are screened, enabling the precise identification of oil samples with abnormal chemical compositions and enhancing the accuracy of anomaly detection. Similarly, the screening step for microbial anomaly samples can identify potential microbial contamination or changes in oils, providing effective support for subsequent processing. Further, by measuring the peroxide value of samples with chemical composition anomalies, the obtained peroxide data helps analyze the degree of oil oxidation, thereby revealing the process and risks of oil aging. The lipase activity measurement helps reveal the microbial activities in oils and their impact on oil quality. Finally, based on the peroxide data and lipase activity data, an oxygen contact assessment is conducted to provide a quantitative analysis of the oil oxidation process, enabling the timely detection of changes in oil quality and potential risks. This improves the precision and comprehensiveness of oil quality monitoring, especially the detection effect during key change processes such as oil aging and oxidation, and helps reveal deeper quality risks.

[0060] Optionally, step S4 is specifically as follows:

[0061] Step S41: Calculate the oxidation reaction rate based on the oxygen contact data to obtain the oxidation reaction rate;

[0062] Step S42: Divide the oil samples based on the oxidation reaction rate to obtain high-oxidation-rate oil samples and low-oxidation-rate oil samples;

[0063] Step S43: Conduct a viscosity aging analysis on the high-oxidation-rate oil samples to generate viscosity aging data;

[0064] Step S44: Conduct a color aging assessment on the low-oxidation-rate oil samples to obtain color aging data;

[0065] Step S45: Evaluate the degree of oil aging based on the viscosity aging data and the color aging data to obtain oil aging degree data;

[0066] Step S46: Construct an oil aging assessment model based on the oil aging degree data to obtain the oil aging assessment model.

[0067] By calculating the oxidation reaction rate, the present invention can quantify the oxidation rate of oils and fats, providing a scientific basis for the subsequent classification of oil samples. Classifying oil samples based on the oxidation rate enables the separate treatment of oil samples with high and low oxidation rates, thereby allowing for a more detailed analysis of the oxidation processes of different types of oils and fats. Conducting viscosity aging analysis on oil samples with high oxidation rates helps to reveal the viscosity changes during the oxidation of oils and fats, providing important data for the physical property evaluation of oil quality. At the same time, performing color aging evaluation on samples with low oxidation rates can help identify the color changes during the oxidation of oils and fats, providing another dimension of quality evaluation indicators. Combining the viscosity aging data and color aging data for the evaluation of the degree of oil aging can comprehensively evaluate the overall aging status of oils and fats, thereby revealing their potential quality risks. Finally, constructing an oil aging evaluation model based on the oil aging degree data enables the quantification and prediction of the oil aging process, providing a scientific and operable evaluation tool for oil quality management. The comprehensive analysis of the changes in oils and fats during oxidation and aging avoids the overly single or simplified processing methods in traditional methods, improving the accuracy and reliability of oil quality detection.

[0068] Optionally, the present specification also provides an oil detection data analysis system based on multi-source data for performing the oil detection data analysis method based on multi-source data as described above. The oil detection data analysis system based on multi-source data includes:

[0069] An infrared spectrum signal enhancement module for obtaining multi-source oil detection data, extracting near-infrared spectrum detection features based on the multi-source oil detection data to obtain infrared spectrum detection data, and performing signal enhancement processing on the infrared spectrum detection data to generate enhanced infrared spectrum data;

[0070] An oil anomaly detection module for extracting physical and chemical index features based on the multi-source oil detection data to obtain physical and chemical index data, and performing detection anomaly analysis based on the physical and chemical index data and the enhanced infrared spectrum data to obtain oil anomaly detection data;

[0071] An oxygen contact analysis module for classifying the oil anomaly detection data to obtain oil chemical composition anomaly data and oil microorganism anomaly data, and performing oxygen contact analysis based on the oil chemical composition anomaly data and the oil microorganism anomaly data to obtain oxygen contact data;

[0072] An oil aging evaluation module for performing oxidation reaction analysis based on the oxygen contact data to obtain oil oxidation reaction data, performing oil aging evaluation based on the oil oxidation reaction data to obtain oil aging degree data, and constructing an oil aging evaluation model to obtain an oil aging evaluation model;

[0073] A quality inspection and evaluation module for performing quality inspection and evaluation on multi-source detection data of grease according to a grease aging evaluation model, so as to obtain grease quality inspection and evaluation data.

[0074] A grease detection data analysis system based on multi-source data according to the present invention can implement any grease detection data analysis method based on multi-source data of the present invention, and is used as a medium for coordinating operations and signal transmission between various modules to complete the grease detection data analysis method based on multi-source data. The internal modules of the system cooperate with each other, thereby improving the abnormal recognition rate and quality evaluation accuracy of grease detection. Brief Description of the Drawings

[0075] By reading the detailed description of the non-restrictive embodiments with reference to the following drawings, other features, objects, and advantages of the present invention will become more apparent:

[0076] Figure 1 It is a schematic flow chart of the steps of the grease detection data analysis method based on multi-source data of the present invention;

[0077] Figure 2 It is a detailed schematic flow chart of step S1 in the present invention;

[0078] Figure 3 It is a detailed schematic flow chart of step S2 in the present invention;

[0079] The realization, functional characteristics, and advantages of the object of the present invention will be further described in conjunction with the embodiments with reference to the drawings. Detailed Embodiments

[0080] The technical method of the present invention will be clearly and completely described below with reference to the drawings. Obviously, the described embodiments are some, but not all, of the embodiments of the present invention. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative efforts fall within the scope of protection of the present invention.

[0081] In addition, the drawings are only schematic diagrams of the present invention and are not necessarily drawn to scale. The same reference numerals in the drawings represent the same or similar parts, and thus their repeated description will be omitted. Some of the block diagrams shown in the drawings are functional entities and do not necessarily correspond to physically or logically independent entities. The functional entities can be implemented in software form, or in one or more hardware modules or integrated circuits, or in different networks and / or processor methods and / or microcontroller methods.

[0082] It should be understood that although terms such as "first", "second", etc. may be used herein to describe various units, these units should not be limited by these terms. These terms are only used to distinguish one unit from another. For example, without departing from the scope of the exemplary embodiments, the first unit may be referred to as the second unit, and similarly the second unit may be referred to as the first unit. The term "and / or" used herein includes any and all combinations of one or more of the listed associated items.

[0083] To achieve the above object, please refer to Figures 1 to 3 , the present invention provides a method for analyzing grease detection data based on multi-source data, and the method includes the following steps:

[0084] Step S1: Obtain multi-source grease detection data, and perform near-infrared spectroscopy detection feature extraction based on the multi-source grease detection data to obtain infrared spectroscopy detection data; perform signal enhancement processing on the infrared spectroscopy detection data to generate enhanced infrared spectroscopy data;

[0085] In this embodiment, various detection data of a grease sample are obtained through a multi-source grease detection system, and these data include but are not limited to near-infrared spectroscopy data, physical and chemical index data, microbial indexes, etc. For the near-infrared spectroscopy data, a high-precision near-infrared spectrometer is used for data acquisition, and a light source with a wavelength range between 1000 nm and 2500 nm is usually used to ensure coverage of the key absorption bands of grease components. The acquired data is presented in the form of a spectrogram, and its amplitude is proportional to the absorbance of the sample components. The acquired original spectral data needs to be subjected to feature extraction. Using typical spectral analysis methods, such as principal component analysis (PCA) or partial least squares (PLS), key feature information such as peak position, wavelength interval, absorbance and other parameters are extracted from the original spectrum to form preliminary infrared spectroscopy data. Next, for the acquired infrared spectroscopy data, an adaptive filter is used for noise removal. The filter parameters set in this process include sampling frequency, filter type (such as high-pass or low-pass filter), and a threshold of 1.0 is set to remove noise signals with an amplitude lower than this threshold. This process generates noise-reduced infrared spectroscopy data. Next, spectral smoothing is performed on the noise-reduced spectral data, and the Savitzky-Golay smoothing filter algorithm is used for spectral smoothing. The window size is set to 5 points, and the smoothing order is set to 3 to remove high-frequency noise and obtain smoothed infrared spectroscopy data. Finally, signal enhancement amplification is performed. A high-pass filter is used to attenuate the low-frequency components and enhance the high-frequency part of the signal. A gain coefficient with a standard deviation of 2.0 is used to adjust the signal amplitude to generate enhanced infrared spectroscopy data.

[0086] Step S2: Extract the physical and chemical index features based on the multi-source detection data of the oil and fat, so as to obtain the physical and chemical index data; perform detection anomaly analysis based on the physical and chemical index data and the enhanced infrared spectrum data, so as to obtain the oil and fat anomaly detection data;

[0087] In this embodiment, the physical and chemical indexes of the oil and fat are extracted through the multi-source detection data of the oil and fat, mainly including acid value, peroxide value, refractive index, etc. Use an acid value detector to analyze the acidity of the oil and fat sample, use a peroxide value detector to measure the peroxide content of the oil and fat, and measure the refractive index of the sample through a refractometer. The detection value of each physical and chemical index is set based on its standard range. The normal range of the acid value is 0.5 to 2.0, and the standard range of the peroxide value is 0 to 10 meq / kg. Then, use the extracted physical and chemical index data and the previously obtained enhanced infrared spectrum data for anomaly detection. Anomaly detection is determined by setting the threshold of each index. For example, if the acid value exceeds 2.5, it is regarded as abnormal, and if the peroxide value exceeds 12 meq / kg, it is abnormal. In this step, by combining the characteristics of the enhanced spectrum data (such as the intensity and position of specific spectral peaks), comparing with the physical and chemical indexes, and using clustering analysis techniques (such as the K-means algorithm) to preliminarily classify the abnormal data. After detecting the abnormal data, generate the oil and fat anomaly detection data as the input for the subsequent steps.

[0088] Step S3: Classify the oil and fat anomaly detection data to obtain the oil and fat chemical composition anomaly data and the oil and fat microorganism anomaly data; perform oxygen contact analysis based on the oil and fat chemical composition anomaly data and the oil and fat microorganism anomaly data to obtain the oxygen contact data;

[0089] In this embodiment, it is necessary to classify the oil and fat anomaly detection data. Based on the chemical composition and microorganism analysis results of the oil and fat, first extract the chemical composition anomaly data. By analyzing the main components in the oil and fat, such as fatty acids, fatty alcohols, etc., set the normal range of their contents. For example, the normal range of the fatty acid content is 35% to 45%. Data outside this range is regarded as chemical composition anomaly. Microorganism anomaly data is extracted through the analysis of bacteria or fungi in the oil and fat. Use the culture medium analysis method to detect the types and quantities of microorganisms in the oil and fat, and set that the total number of bacteria exceeding 1000 CFU / mL is abnormal. Perform oxygen contact analysis on the chemical composition anomaly data and the microorganism anomaly data. According to the relationship between the oxygen contact amount of the oil and fat and the environmental oxygen concentration, use an oxygen concentration sensor to monitor the oxygen exposure of the oil and fat sample within a certain period of time to obtain the oxygen contact data. By quantitatively analyzing the contact time of the oil and fat sample with air under specific conditions and the change of oxygen concentration, obtain the oxygen contact amount data. These data will provide a basis for the subsequent oxidation reaction evaluation of the oil and fat.

[0090] Step S4: Analyze the oxidation reaction based on the oxygen contact data to obtain the oil oxidation reaction data; evaluate the oil aging based on the oil oxidation reaction data to obtain the oil aging degree data, and construct an oil aging evaluation model to obtain the oil aging evaluation model;

[0091] In this embodiment, the oxidation reaction rate is calculated according to the oxygen contact data. The Arrhenius equation is used to calculate the oxidation reaction rate of the oil at different temperatures and oxygen concentrations. The initial value of the reaction rate constant is set to 0.05 min^-1, the temperature is set at 25 °C, and the oxygen concentration is 21%. The oxidation reaction rate is a key indicator for judging the degree of oil oxidation. Based on the oxidation reaction rate, the oil samples are divided. For samples with a reaction rate greater than 0.1 min^-1, they are classified as high oxidation rate oil samples; for samples with a reaction rate less than 0.05 min^-1, they are classified as low oxidation rate oil samples. The aging analysis is carried out on the high and low oxidation rate oil samples respectively. For high oxidation rate samples, their aging degree is evaluated by measuring the viscosity change. The viscometer is used to measure the viscosity change of the oil at different time points, and the threshold of the viscosity change rate is set to 5% / hour. For low oxidation rate samples, the color change analysis is carried out. The colorimeter is used to measure the color difference of the samples at different time points, and a color difference change of more than 1.5 is set as the aging sign. According to the viscosity aging data and the color aging data, the aging degree of the oil is evaluated. By the weighted average method, combining the aging scores of the two indicators, the comprehensive aging degree of the oil is obtained, and the linear regression method is used to construct an oil aging evaluation model with the regression coefficient set to 1.0 to obtain the oil aging evaluation model.

[0092] Step S5: Perform quality inspection and evaluation on the multi-source detection data of the oil according to the oil aging evaluation model to obtain the oil quality inspection and evaluation data.

[0093] In this embodiment, multi-source detection data of oils and fats, such as infrared spectral data obtained by near-infrared spectroscopy, physicochemical index data (such as acid value, peroxide value, etc.), and microbial detection data, are input into the established oil and fat aging evaluation model. The aging evaluation model adopts the regression analysis method, and the model includes the influencing factors and their regression coefficients of key quality indicators such as oxidation degree, color, and viscosity. These regression coefficients were obtained through regression analysis based on a large amount of oil and fat sample data in previous experiments, ensuring the accuracy and stability of the model. When evaluating the quality of oils and fats, first, the regression coefficients in the model are used to calculate various quality indicators of the oils and fats. For example, the oxidation degree is evaluated by factors such as the oxidation reaction rate and peroxide value, the color evaluation is based on spectral data and color difference analysis, and the viscosity change is determined according to the viscosity data in the physicochemical indicators. For the evaluation of each quality indicator, specific standard thresholds are set. For example, for the oxidation degree, a standard is set that the oxidation degree not exceeding 40% is a qualified product; for the viscosity, a change range within 10% is high-quality oil, and exceeding this threshold will affect the use value of the oil. In addition, the color index can be set that the color difference does not exceed a certain range to meet the quality requirements. Based on these standards, the model analyzes by comparing the actually measured data with the set thresholds to obtain the qualification of each indicator. Finally, the evaluation results of each indicator are integrated to form a comprehensive oil and fat quality detection and evaluation data. These evaluation results include the oxidation degree, color, and viscosity change of the oils and fats, and their quality grades are determined according to the set standards. These oil and fat quality detection and evaluation data will be fed back to the production line or quality monitoring system for further analysis by operators or quality control personnel. These feedback data help to optimize the production process, adjust production conditions or change the use of raw materials, so as to ensure the stable quality of the final product and meet the standard requirements.

[0094] Optionally, step S1 is specifically as follows:

[0095] Step S11: Obtain multi-source detection data of oils and fats, and perform near-infrared spectral detection feature extraction based on the multi-source detection data of oils and fats to obtain infrared spectral detection data;

[0096] In this embodiment, multi-source detection data of grease need to be collected, which usually includes data obtained through different detection methods (such as near-infrared spectroscopy, physical and chemical index analysis, microbial analysis, etc.). In this step, the key is to extract the near-infrared spectroscopy detection data. In specific operations, a near-infrared spectroscopy instrument is used to scan the grease sample and record its reflection spectrum or transmission spectrum. The frequency range is usually set to 700nm to 2500nm. Through the sampling function of the spectrometer, the grease sample is scanned step by step at a set step size (such as 1nm or 2nm) to obtain the original infrared spectroscopy data. During this process, parameters such as the light source intensity and detector gain of the spectrometer need to be set properly to ensure that the measured spectroscopy data has sufficient resolution and accuracy. The obtained spectroscopy data will be used as the basic data for subsequent analysis.

[0097] Step S12: Perform noise filtering on the infrared spectroscopy detection data to obtain noise-reduced infrared spectroscopy detection data;

[0098] In this embodiment, noise filtering is performed on the infrared spectroscopy detection data to reduce the interference of environmental noise and instrument noise. Common methods include Gaussian filtering or median filtering. In specific operations, a suitable filtering window is selected, such as a 5-point or 7-point window, and the filtering parameters are set according to the noise characteristics of the data. When performing Gaussian filtering, the standard deviation σ value is determined. Common values range from 0.5 to 2, depending on the volatility of the data. By filtering the original infrared spectroscopy data, overly sharp noise points are removed, and the main spectroscopy signals are retained, thereby obtaining noise-reduced infrared spectroscopy data. This process requires careful control of the filtering intensity to avoid signal loss caused by over-filtering. Finally, the obtained data is smoother and more stable, providing a basis for subsequent spectroscopy smoothing.

[0099] Step S13: Perform spectroscopy smoothing based on the noise-reduced infrared spectroscopy detection data to obtain smoothed infrared spectroscopy detection data;

[0100] In this embodiment, a spectroscopy smoothing algorithm is used to smooth the noise-reduced infrared spectroscopy data to reduce random fluctuations in the data. Common spectroscopy smoothing methods include the Savitzky-Golay smoothing method and the moving average method. Taking the Savitzky-Golay smoothing method as an example, first, the sliding window size is set, generally taking values from 5 to 15 data points, and a suitable window length is determined according to the characteristics of the data. Then, a suitable polynomial order (usually 2nd or 3rd order) is selected to smooth the spectroscopy data. During the smoothing process, each data point within the window is estimated through polynomial fitting, and the volatility of the smoothed spectroscopy data is reduced compared to the original data. This step requires smoothing while ensuring signal characteristics and avoiding distortion, so strict selection of smoothing parameters (such as window size and polynomial order) is required during implementation.

[0101] Step S14: Perform signal enhancement and amplification on the smoothed infrared spectroscopy detection data to generate enhanced infrared spectroscopy data.

[0102] In this embodiment, an amplification algorithm or an enhancement method based on baseline correction is used for operation. Common signal enhancement techniques include baseline correction and amplitude amplification. During the baseline correction process, the baseline is estimated through the lowest point or the maximum point, and then a relative translation of the baseline is performed to ensure that the intensity change of the spectral signal can more clearly reflect the spectral characteristics. When performing amplitude amplification, a suitable gain coefficient (such as 1.5 times, 2 times, or a coefficient deduced from experimental data) is selected, and each data point of the smoothed data is multiplied by the gain coefficient to amplify the intensity of the signal. During the amplification process, it is necessary to avoid over-amplification of the signal resulting in non-linear distortion or feature blurring. Therefore, the enhancement parameters (such as the gain coefficient) should be adjusted according to the actual experimental data to ensure that the enhanced signal can still reflect the true spectral characteristics of the oil. Finally, the obtained enhanced infrared spectroscopy data will have a stronger signal intensity, facilitating subsequent analysis and identification.

[0103] Optionally, step S2 is specifically as follows:

[0104] Step S21: Extract the physical and chemical index characteristics based on the multi-source detection data of the oil to obtain the physical and chemical index data;

[0105] In this embodiment, through multi-source detection of the oil sample, including near-infrared spectroscopy, chemical analysis, physical detection, etc., the physical and chemical index characteristics of the oil are extracted. Specifically, during the operation, first collect the original multi-source data of the oil, including but not limited to physical and chemical properties such as the acid value, peroxide value, viscosity, and density of the oil. Use chemical analysis instruments such as pH meters, viscometers, and colorimeters to process the oil sample and record the corresponding values. For example, use a pH meter to measure the acid value of the oil, and set the standard value to 0.5 mg KOH / g. When the acid value exceeds this standard value, it indicates a relatively high degree of acidification of the oil. When performing density measurement, use a standard densitometer and calibrate it according to environmental parameters such as temperature and humidity. Each value of these physical and chemical indexes should be detected within the range that meets the industry standards, and appropriate thresholds should be set according to the experiment to extract the physical and chemical index data of the oil.

[0106] Step S22: Perform component anomaly analysis on the physical and chemical index data to obtain component anomaly data;

[0107] In this embodiment, statistical analysis methods (such as outlier detection and box plot analysis) are used to preprocess the physicochemical index data of oils and fats. During implementation, reasonable upper and lower limit thresholds are set (for example, the acid value standard is 0.5 mg KOH / g, and if this value is exceeded, it is considered that the oil and fat are abnormally acidified). Then, the obtained physicochemical data are input into an outlier detection algorithm for processing. For example, the 3σ rule is used (that is, points where the data are greater than or less than 3 times the standard deviation of the mean are regarded as outliers), and it is analyzed whether each physicochemical index conforms to the expected range. For values that do not meet the standard, they are determined as component abnormal data. In addition, in combination with the normal distribution of historical oil and fat samples, an outlier detection method based on clustering (such as K-means or DBSCAN) can be used to further classify the data and identify outlier points to obtain component abnormal data.

[0108] Step S23: Perform frequency band anomaly analysis on the enhanced infrared spectrum data to obtain frequency band anomaly data;

[0109] In this embodiment, the enhanced infrared spectrum data are divided into multiple frequency bands (such as 1000 - 1500 nm, 1500 - 2000 nm, etc.), and statistical analysis of the data is performed within each frequency band. By calculating parameters such as the mean and standard deviation of each frequency band, the normal fluctuation range of each frequency band is set. For example, for an absorption peak in a certain frequency band, if its amplitude exceeds the set threshold (such as 3 times the standard deviation threshold), then this frequency band is regarded as abnormal. Then, Fourier transform is used to perform spectral analysis on the infrared spectrum data to check whether there are significant deviations or distortions in the frequency band. If the signal intensity of a certain frequency band is abnormally high or low, it will be determined that this frequency band is abnormal to obtain frequency band anomaly data. In the frequency band anomaly analysis, reasonable frequency band widths and frequency ranges also need to be set to more accurately locate and analyze abnormal fluctuations.

[0110] Step S24: Perform integration of oil and fat anomaly detection based on the component abnormal data and the frequency band abnormal data to obtain oil and fat anomaly detection data.

[0111] In this embodiment, the component anomaly data and the frequency band anomaly data are aligned according to the time series or the sample identifier to ensure the synchronization of the component anomalies and the frequency band anomaly data of the same oil sample. Then, a data fusion method is used to comprehensively analyze the two types of anomaly data, such as the weighted average method, logistic regression, etc. For example, the weights of the component anomaly data and the frequency band anomaly data are set. Usually, a higher weight (such as 0.7) is given to the frequency band anomaly data, while the weight of the component anomaly data is lower (such as 0.3). Then, the two are weighted and fused to generate comprehensive oil anomaly detection data. If the fused data exceeds the set anomaly threshold (such as 0.8), it is determined as abnormal oil. This process requires the use of accurate threshold setting (such as 0.8 as the anomaly determination threshold) to ensure the accurate identification of anomaly data. Finally, the integrated oil anomaly detection data will be used as an important basis for oil quality control and production process optimization.

[0112] Optionally, step S22 is specifically as follows:

[0113] Step S221: Perform acid value anomaly detection on the physicochemical index data to obtain acid value anomaly data;

[0114] In this embodiment, acid value anomaly detection is performed on the physicochemical index data of the oil. Using the standard acid value determination method, the acid value of the oil sample is measured by an automated acidimeter, and the measured acid value is recorded. The normal range of the acid value is set to 0.1 - 0.5 mg KOH / g. All measured acid values are compared with the set normal range. If the acid value exceeds 0.5 mg KOH / g, it is considered that the acid value of the sample is abnormal. To improve the accuracy, the three-sigma method (3σ rule) is adopted. The mean and standard deviation of the acid value data of all oil samples are calculated. If the acid value of a certain sample differs from the mean by more than 3 times the standard deviation, it is also marked as abnormal. These abnormal data are output as acid value anomaly data for subsequent processing.

[0115] Step S222: Perform peroxide value anomaly detection on the physicochemical index data to obtain peroxide value anomaly data;

[0116] In this embodiment, the peroxide value, as an important indicator of the degree of oil oxidation, is measured for the oil sample using a peroxide determination kit through chemical analysis, and its specific value is recorded. The normal range of the peroxide value is set to 0 - 10 meq / kg, and a peroxide value outside this range is considered abnormal. In specific implementation, first, the peroxide value of the sample is measured to obtain the value of each sample, and it is compared with the set range. If it exceeds 10 meq / kg or is lower than 0 meq / kg, it is considered abnormal. In the detection, the deviation of the peroxide value is detected in combination with the three - standard - deviation method (3σ rule). If a data point deviates from the mean by more than 3 times the standard deviation, it is also marked as abnormal. These abnormal data are output as peroxide - value abnormal data for further analysis.

[0117] Step S223: Perform clustering of component abnormality indicators based on the acid - value abnormal data and the peroxide - value abnormal data to generate component abnormality indicators;

[0118] In this embodiment, first, the acid - value abnormal data and the peroxide - value abnormal data are subjected to clustering analysis. The K - means clustering algorithm is used. By summarizing these two sets of data, the relationships and distributions in different characteristic dimensions are analyzed. The data of the acid value and the peroxide value are input into the K - means algorithm, and the number of clusters is set to 2 (representing normal and abnormal categories). The K - means algorithm will classify all data based on the Euclidean distance. During the clustering process, if the acid value and the peroxide value of a data point simultaneously exceed the set abnormal threshold, it is classified as the abnormal category. For the abnormal category obtained after clustering, relevant indicators of component abnormality, such as the combined effect of the acid value and the peroxide value, are further analyzed to generate component abnormality indicators. These component abnormality indicators can serve as the basis for subsequent extraction of component abnormalities.

[0119] Step S224: Extract component abnormalities from the physical and chemical index data according to the component abnormality indicators to obtain component - abnormality data.

[0120] In this embodiment, the generated component abnormality indicators are used as screening conditions to extract the sample data related to abnormal components from the original physical and chemical data. In specific implementation, the acid value and the peroxide value of each sample are further verified. If these values simultaneously meet the threshold conditions of the component abnormality indicators (for example, the acid value exceeds 0.5 mg KOH / g and the peroxide value exceeds 10 meq / kg), then this sample is marked as component - abnormality data. At the same time, in combination with other physical and chemical indicators (such as viscosity, density, etc.), a multivariate anomaly detection algorithm (such as PCA analysis) is used to further confirm whether there are other abnormal patterns. If the aggregated values of these physical and chemical indicators also exceed the normal range, they are marked as component - abnormality data. Finally, the oil sample data that meet the component abnormalities are output for subsequent analysis.

[0121] Optionally, step S23 is specifically as follows:

[0122] Step S231: Decompose the enhanced infrared spectral data into frequency bands to obtain high-frequency region enhanced infrared spectral data and low-frequency region enhanced infrared spectral data;

[0123] In this embodiment, the enhanced infrared spectral data is decomposed into frequency bands. The overall spectral data is decomposed into multiple frequency bands using the Fourier transform method, and is divided into a high-frequency region and a low-frequency region according to the required frequency band range. Specifically, when implementing, the cut-off frequency of the high-frequency region is set to 2000 cm -1 , and the cut-off frequency of the low-frequency region is set to 500 cm -1 . The components of each frequency band are calculated through Fourier transform, and the high-frequency information in the original enhanced infrared spectral data is extracted as high-frequency region enhanced infrared spectral data, and the low-frequency information is extracted as low-frequency region enhanced infrared spectral data. To ensure the accuracy of the decomposition process, an appropriate window function (such as a Hanning window) is used for smoothing to reduce the phenomenon of frequency leakage. Finally, the generated high-frequency region and low-frequency region spectral data will be used as the input for subsequent analysis respectively.

[0124] Step S232: Detect the peak and valley positions of the first derivative spectrum based on the high-frequency region enhanced infrared spectral data to obtain high-frequency region spectral peak and valley position data;

[0125] In this embodiment, the first derivative data is obtained by calculating the intensity difference between two adjacent wavelength points in the spectral data. This process is based on the intensity values of the spectral data at each wavelength position, and the first derivative is calculated by dividing the intensity difference between adjacent wavelength points by the wavelength interval. In this way, the derivative data reflecting the spectral change rate can be obtained. After the first derivative calculation is completed, a standard peak and valley detection algorithm is used to identify the peaks and valleys in the spectrum. The peak and valley detection locates by analyzing the sign change of the first derivative: when the derivative value changes from positive to negative, it indicates that the current position is a peak; when the derivative value changes from negative to positive, it indicates that the current position is a valley. During the detection process, a threshold for peak and valley recognition is set to ensure that only the peaks and valleys with obvious intensity differences are effectively marked. For example, the peak intensity difference is set to be greater than 0.02 as an obvious peak. Finally, the marked high-frequency region spectral peak and valley position data is output for subsequent analysis.

[0126] Step S233: Separate the Lorentzian-fitted overlapping peaks based on the low-frequency region enhanced infrared spectral data to obtain low-frequency region overlapping peak data;

[0127] In this embodiment, the Lorentz function is used for preliminary fitting. The Lorentz function is used to describe the shape of a single peak in the spectrum and is mainly controlled by three parameters: the height of the peak, the position of the peak, and the full width at half maximum (FWHM). The height of the peak represents the intensity of the peak, the peak position determines the wavelength of the spectral peak, and the FWHM controls the shape or width of the peak. For multiple overlapping peaks, the multiple Lorentz fitting method is used to independently fit each overlapping peak one by one. This process adjusts the three fitting parameters of each peak to make the fitting result as close as possible to the actually observed spectral data. When dealing with overlapping peaks, first, a peak spacing threshold (e.g., 0.5 cm -1 ) is set to determine whether there are overlapping peaks. If the wavelength positions of two peaks are very close and the spacing is less than the set threshold, they are considered overlapping peaks. Next, by optimizing the parameters in the fitting process, the height, position, and FWHM of the peak are adjusted to gradually approach the actual spectral data. Finally, information such as the position and intensity of each peak is extracted to generate the processed overlapping peak data in the low-frequency region, and these data are used as the basis for subsequent analysis.

[0128] Step S234: Perform peak position offset detection on the peak valley position data in the high-frequency region spectrum to obtain the high-frequency region peak position offset data;

[0129] In this embodiment, a threshold range is set. For example, if the peak position offset exceeds 0.5 cm -1 , it is considered an offset. By calculating the difference in peak positions between two adjacent measurements, if the difference exceeds the set threshold, it is considered that the peak position has shifted. The least squares method can be used to fit the curve and compare it with the original data to calculate the offset amount of the peak position. All peak positions with offset amounts exceeding the set threshold are marked as high-frequency region peak position offset data and output for subsequent analysis.

[0130] Step S235: Perform fitting residual analysis on the overlapping peak data in the low-frequency region to obtain the high-residual data in the low-frequency region;

[0131] In this embodiment, the Lorentz fitting result obtained in step S233 is used and applied to the overlapping peak data in the low-frequency region. The curve obtained from the fitting process is compared with the original data to evaluate the fitting accuracy. Then, the residual between each spectral data point and the fitting curve is calculated. The residual refers to the difference between the original data and the fitting curve. To determine which data points belong to the high-residual data, a residual threshold is set. In specific operations, when the residual exceeds the set threshold (e.g., 0.1), the point is marked as a high-residual point. This threshold can be adjusted according to the characteristics of the actual data to accurately identify the points with large fitting errors. By analyzing the residuals of all data points, those points with large residual values can be screened out. These points represent the parts where the fitting model fails to accurately reflect the actual data. Finally, these high-residual data are output and provided as reference data for the subsequent frequency band anomaly integration step.

[0132] Step S236: Perform band anomaly integration based on the peak position offset data in the high-frequency region and the high residual data in the low-frequency region, so as to obtain band anomaly data.

[0133] In this embodiment, it is necessary to fuse the peak position offset data in the high-frequency region and the high residual data in the low-frequency region. To ensure the accuracy of anomaly detection, a comprehensive band anomaly threshold is set, and this threshold determines whether there is a band anomaly through the combined action of two indicators. For the peak position offset data in the high-frequency region and the high residual data in the low-frequency region, the following rules are set: when the peak position offset value in the high-frequency region exceeds 0.5 cm -1 , and the high residual value in the low-frequency region is greater than 0.1, it is considered that the data in this band is abnormal. The thresholds 0.5 cm -1 and 0.1 are set according to the experimental data and the characteristics of spectral analysis, and can be fine-tuned according to the actual situation of the data set in actual use. Then, compare the peak position offset in the high-frequency region and the high residual in the low-frequency region of all data points one by one to check whether each data point meets the above threshold conditions. If a certain data point exceeds the set threshold in both of these two indicators, then this data point is marked as band anomaly data. All data points in the entire data set are subjected to such comparison and screening. Finally, the selected band anomaly data is output, providing a basis for subsequent oil anomaly detection and analysis.

[0134] Optionally, step S232 is specifically:

[0135] Perform first-order derivative calculation based on the enhanced infrared spectral data in the high-frequency region to obtain first-order derivative data;

[0136] In this embodiment, obtain the enhanced infrared spectral data in the high-frequency region, determine the wavelength range of the data and the corresponding intensity values. Use the first-order derivative algorithm to calculate the spectral data point by point. By calculating the intensity difference between adjacent data points and dividing it by the wavelength interval, the first-order derivative value is obtained. This process provides a basis for subsequent peak-valley identification by determining the change trend of the spectral data. When calculating, it is necessary to ensure that the wavelength interval is about 0.1 cm -1 to ensure the calculation accuracy, and the finally generated first-order derivative data will be used as the input for the next zero-crossing point identification.

[0137] Perform zero-crossing point identification on the first-order derivative data to obtain zero-crossing point data;

[0138] In this embodiment, the obtained first derivative data is analyzed to identify its zero-crossing points. A zero-crossing point refers to a point where the first derivative changes from a positive value to a negative value or from a negative value to a positive value. By traversing the first derivative data and judging its sign change, when the first derivative value changes from a positive value to a negative value or from a negative value to a positive value, this point is recorded as a zero-crossing point. At this time, a threshold is set to control the amplitude of the signal change. For example, 0.02 is set as the threshold. When the change in the derivative value exceeds this threshold, this point is considered a valid zero-crossing point. By this method, the key positions in the spectrum are accurately identified for use in the subsequent second derivative calculation.

[0139] Perform a second derivative calculation based on the zero-crossing point data to obtain second derivative data;

[0140] In this embodiment, based on the zero-crossing point data, the second derivative is further calculated. The second derivative can be obtained by performing a derivative calculation on the first derivative data again. For each first derivative point, calculate the difference between its adjacent points and divide by the wavelength interval to obtain the second derivative data. During the implementation process, the calculation results need to be smoothed to reduce the influence of noise. For example, a smoothing window can be used for filtering, and the window size is set to 5 data points, which can ensure the stability and reliability of the calculated second derivative data. Finally, the obtained second derivative data will be used for the determination of peak and valley positions.

[0141] Perform a division based on the second derivative data. If the second derivative data is greater than zero, it is judged as a peak position to obtain peak data; if the second derivative data is less than zero, it is judged as a valley position to obtain valley data;

[0142] In this embodiment, a division is performed based on the second derivative data. For each second derivative data point, if the second derivative value of this point is greater than zero, this point is a peak position; if the second derivative value of this point is less than zero, this point is a valley position. For the points with a second derivative value of zero, they can be ignored or marked as boundary points. To ensure the accuracy of peak and valley determination, a threshold needs to be set, and the threshold is 0.05 to ensure that the calculated peaks and valleys have an obvious change trend. By this method, the positions of peaks and valleys can be clearly determined, providing a basis for the integration of peak and valley positions in the high-frequency region spectrum in the subsequent steps.

[0143] Perform an integration of the peak and valley positions in the high-frequency region spectrum based on the peak data and the valley data to obtain the peak and valley position data in the high-frequency region spectrum.

[0144] In this embodiment, the data at all peak positions and valley positions are combined into a single data set and sorted according to the order of wavelengths. Then, by combining adjacent data points before and after, the order and positions of the peaks and valleys are ensured to be reasonable. During the integration process, if there are misjudged peak and valley positions, they can be corrected through further smoothing processing or re-determination. Finally, the generated spectral peak and valley position data in the high-frequency region will be used for subsequent analysis and applications.

[0145] Optionally, step S233 is specifically as follows:

[0146] Partition the enhanced infrared spectral data in the low-frequency region to obtain the partitioned data of the enhanced infrared spectrum in the low-frequency region;

[0147] In this embodiment, the enhanced infrared spectral data in the low-frequency region is obtained, and the data is divided into several intervals according to the wavelength range. To ensure the scientific nature of the partitioning, the wavelength interval is set to 10 cm -1 , for example, the low-frequency region data is divided into multiple sub-intervals with a step size of 10 cm -1 . For each interval, the corresponding spectral data is extracted, including the intensity value and the wavelength value. During the implementation process, data analysis tools (such as the MATLAB or Python's NumPy library) are used to gradually process the original spectral data and partition it according to the set wavelength range to ensure that the data in each partition is representative. Finally, the spectral data of each partition is output as the input for subsequent processing.

[0148] Set the Lorentz function parameters for the partitioned data of the enhanced infrared spectrum in the low-frequency region to obtain the Lorentz function parameters;

[0149] In this embodiment, for the spectral data of each partition in the low-frequency region, the Lorentz function is used to fit it, and the initial parameters of the Lorentz function are first set. These parameters include the height of the peak, the position of the peak, and the full width at half maximum, which represent the intensity of the peak, the wavelength position, and the width respectively. During the implementation, initial values are set. For example, the height of the peak is set to the maximum spectral intensity, the position of the peak is set to the central position of the spectrum, and the full width at half maximum is set to 20 cm -1 . These parameters are optimized by the least squares method to make the fitting result as close as possible to the actual spectral data. During the optimization process, an optimization threshold (such as the fitting error being less than 0.05) needs to be set to ensure the fitting accuracy. The goal of this step is to obtain accurate Lorentz function parameters for each partition, providing a basis for subsequent peak fitting and overlapping peak processing.

[0150] Perform single-peak interval fitting on the partitioned data of the enhanced infrared spectrum in the low-frequency region according to the Lorentz function parameters to obtain the single-peak fitting data;

[0151] In this embodiment, the spectral data of each partition is matched with the Lorentz function, and the parameters of the Lorentz function are adjusted using the least squares fitting method to minimize the difference between the fitting curve and the actual spectral data. During the fitting process, the fitting error tolerance is set to 0.02 to ensure the accuracy of the fitting result. For each partition, if the fitting result error is lower than this tolerance, then this partition is considered a single-peak interval, and the corresponding single-peak fitting data is obtained. These data include the peak position, intensity, and full width at half maximum after fitting, and finally these single-peak fitting data are output as the input for subsequent multi-peak fitting.

[0152] Based on the single-peak fitting data, the enhanced infrared spectral data in the low-frequency region is subjected to multi-peak combination to obtain overlapping peak data;

[0153] In this embodiment, the single-peak fitting data obtained in step S3 is jointly analyzed with other existing peaks. For the spectral data in each low-frequency partition, by calculating the distance between each peak and its adjacent peak, it is judged whether there is an overlap. The peak distance threshold is set to 0.5 cm -1 , and when the distance between two peaks is less than this threshold, these two peaks are considered overlapping peaks. During the implementation process, a suitable algorithm (such as calculating the distance between adjacent peaks point by point and comparing) is used to identify overlapping peaks. Through multi-peak joint fitting, multiple overlapping peaks are combined into one fitting result to obtain overlapping peak data. For each overlapping peak, its position, intensity, and width parameters are extracted and these data are output for subsequent processing.

[0154] Fitting calculations are performed on the overlapping peak data to obtain the overlapping peak data in the low-frequency region.

[0155] In this embodiment, the initial parameters of each overlapping peak are set, including the positions, intensities, and full widths at half maximum of multiple peaks. Then, the multiple least squares fitting algorithm is adopted to optimize the fitting result by adjusting these initial parameters. In this process, the non-linear least squares algorithm needs to be used to more accurately fit the situation of multiple overlapping peaks. For each overlapping peak, the error tolerance during the fitting calculation is set to 0.01 to ensure that the final fitting curve is close to the original spectral data. Through this process, the overlapping peak data in the low-frequency region obtained contains the accurate position and intensity information of each peak, and these data will be used as the basis for subsequent analysis.

[0156] Optionally, step S3 is specifically:

[0157] Step S31: Obtain the standard data of grease components;

[0158] In this embodiment, standard oil samples from known sources are collected, including different types of oil components. Using chemical analysis methods such as gas chromatography-mass spectrometry (GC-MS) technology, these standard oil samples are analyzed in detail. The standard oil component data includes the types and contents of fatty acids, as well as other key chemical components (such as peroxides, vitamin E, etc.). These data need to be obtained through laboratory analysis, and the precision of the analysis equipment is required to reach the μg level to ensure the accuracy of the data. During the analysis process, each sample is subjected to at least three repeated experiments to ensure the consistency and reliability of the data. Finally, all the analysis results are sorted into a standard data set of oil components and stored in the form of a table or a database.

[0159] Step S32: Classify the abnormal oil detection data according to the standard data of oil components, so as to obtain abnormal data of oil chemical components and abnormal data of oil microorganisms;

[0160] In this embodiment, according to the standard oil component data, the normal component range of the oil is set (such as the standard content range of fatty acids, the allowable concentration of peroxides, etc.). Using these standard data, the chemical components in the oil sample are analyzed, and the oil is tested by chemical analysis equipment (such as high performance liquid chromatography HPLC or ultraviolet-visible spectrophotometer UV-Vis). If the oil components exceed the set normal range, it is determined as abnormal chemical components. At the same time, through microbiological analysis methods (such as cultivation method or PCR detection method), the microbial content in the oil is detected, such as bacteria, fungi, etc. If the number of microorganisms in the oil exceeds the normal standard, it is determined as microbial abnormality. Finally, based on the chemical component and microbial data, the oil samples are classified into two categories: abnormal data of oil chemical components and abnormal data of oil microorganisms.

[0161] Step S33: Screen the oil samples based on the abnormal data of oil chemical components to obtain abnormal samples of oil chemical components;

[0162] In this embodiment, a threshold range is set. For each chemical component, such as the proportion of fatty acids, the concentration of peroxides, etc., a threshold value exceeding the normal value is set (for example, when the peroxide concentration exceeds 0.5 mmol / kg, it is considered abnormal). The analysis data of the oil samples is compared with these threshold values. If the component data in the oil sample exceeds the threshold range, then this sample is considered an abnormal sample of chemical components. To ensure the accuracy of the screening process, each sample needs to be tested independently three times to determine the consistency of the results. Finally, the abnormal samples of oil chemical components are screened out for subsequent analysis.

[0163] Step S34: Determine the peroxides in the abnormal samples of oil chemical components to obtain peroxide data;

[0164] In this embodiment, samples to be subjected to peroxide determination are screened out from oil and fat samples with abnormal chemical components. After the samples are selected, the peroxide concentration in the oil and fat is determined by classical chemical determination methods (such as the iodometric method or the ferric chloride method). Taking the iodometric method as an example, first, a certain mass of the oil and fat sample (for example, 5 g) is taken, and a known volume of solvent (such as dichloromethane) is added. The peroxide in the oil and fat is extracted into the solution by ultrasonic oscillation. Then, a potassium iodide solution with a certain concentration is added, and the peroxide reacts with the potassium iodide solution to generate iodide ions. Next, a standard sodium thiosulfate solution (Na2S2O3) is used for titration until the color of the solution changes from yellow to colorless. Record the volume of the sodium thiosulfate solution consumed (in milliliters), and calculate the peroxide concentration according to the formula: Peroxide concentration (mmol / L) = Volume of sodium thiosulfate solution consumed (mL) × Concentration of sodium thiosulfate (mol / L) / Mass of oil and fat sample (g). Each sample is subjected to at least three experiments to ensure the reliability and accuracy of the measurement results, and the measurement error is controlled within ±0.01 mmol / L. In addition, the ferric chloride method can also be used. First, the oil and fat sample is mixed with the solvent and the peroxide is extracted. Then, a ferric chloride solution with a known concentration is added. The peroxide reacts with the ferric chloride to form a red complex. The absorbance of the solution is measured using a UV-Vis spectrophotometer at a specific wavelength (such as 510 nm), and the concentration data of the peroxide is obtained by comparing with the standard solution according to the Beer-Lambert law. Finally, all the peroxide concentration data will be recorded in a data table for subsequent analysis.

[0165] Step S35: Screen oil and fat samples based on the abnormal data of oil and fat microorganisms to obtain oil and fat microorganism abnormal samples;

[0166] In this embodiment, the normal content range of microorganisms in the oil and fat is set. The number of bacteria should not exceed 10 2 CFU / g, and the number of fungi should not exceed 10 2 CFU / g. According to this standard, each oil and fat sample is subjected to microorganism detection by microbiological analysis methods. Commonly used detection methods include the cultivation method and the PCR detection method. If the cultivation method is used, first, the oil and fat sample is diluted and spread on an appropriate culture medium (such as nutrient agar medium), and then the culture medium is placed in a suitable temperature condition for 24 to 48 hours. After cultivation, the number of microorganisms is evaluated by counting the bacterial or fungal colony-forming units (CFU) growing on the sample. If the number of microorganisms exceeds the set threshold (such as 10 2If the number of microorganisms in the sample is greater than 10⁶ CFU / g, then the sample is determined to be a microorganism abnormal sample. If the PCR detection method is used, first extract the DNA from the oil sample, then use specific primer pairs to amplify the genes of bacteria or fungi, and then analyze the number of microorganisms by gel electrophoresis or real-time quantitative PCR. Through the above methods, all oil samples are detected, and the detected results are compared with the standard data to screen out the oil samples with abnormal microorganism numbers. At least three repeated experiments are required for the detection of each sample to ensure the reliability and accuracy of the results. Finally, all oil samples that meet the microorganism abnormality standard will be screened out and prepared for the next step of lipase activity determination.

[0167] Step S36: Determine the lipase activity of the oil microorganism abnormal sample to obtain lipase activity data;

[0168] In this embodiment, select the oil microorganism abnormal samples that have been screened out and determine their lipase activities. The determination method uses the standard colorimetric method or fluorescence method, and determines the activity level by measuring the rate of the lipase-catalyzed reaction in the sample. In the colorimetric method, first mix the oil sample with the reaction solution containing the lipase substrate. The reaction solution contains a suitable substrate such as oil or emulsified oil. When the lipase acts on the substrate, corresponding reaction products will be produced. These products will cause a change in the color of the solution. Use a spectrophotometer to measure the absorbance of the color change of the solution, and calculate the lipase activity by comparing the change rate of the absorbance. The fluorescence method uses the intensity of the fluorescence signal generated after the lipase-catalyzed reaction to quantitatively analyze the lipase activity. In the experiment, the measurement unit of lipase activity is set as U / g, that is, the lipase activity unit per gram of the oil sample. During the test, to ensure the reliability and accuracy of the results, each sample needs to be repeated at least three times. Before each experiment, calibrate with a lipase standard solution of known concentration to ensure the test accuracy, and the goal is to ensure that the measurement error does not exceed ±0.1 U / g. After the experiment, the lipase activity data of all oil samples will be recorded in a data table as the data basis for subsequent analysis.

[0169] Step S37: Conduct an oil oxygen contact assessment based on the peroxide data and the lipase activity data to obtain oxygen contact data.

[0170] In this embodiment, an oxygen contact assessment model is set up. The establishment of the model is based on the mutual relationship between the peroxide concentration and the lipase activity. Through research, it is found that the increase in peroxide concentration is usually closely related to the time of oil exposure to oxygen and the oxygen concentration, while the lipase activity is somewhat related to the oxidative stability of the oil. Therefore, an oxygen contact threshold is set. According to the experimental data, if the peroxide concentration exceeds 0.5 mmol / kg and the lipase activity is lower than 1 U / g, it is considered that the oxygen contact of the oil is excessive. Excessive oxygen contact means that the oil has undergone excessive oxidation during storage or use, resulting in a decrease in the quality of the oil. Then, by combining the peroxide concentration data and the lipase activity data of all samples and comparing them one by one, the oxygen contact situation of each sample is analyzed. During this process, the accuracy and integrity of the data are ensured, and all samples need to be experimented at least three times to verify the results. Through comparative analysis, the oil samples with abnormal oxygen contact are screened out, and their corresponding oxygen contact data are recorded. These data will be used as the basis for subsequent oil quality assessment and treatment, helping to identify the oil samples that need to optimize the storage or treatment methods.

[0171] Optionally, step S4 is specifically as follows:

[0172] Step S41: Calculate the oxidation reaction rate based on the oxygen contact data to obtain the oxidation reaction rate;

[0173] In this embodiment, the oxidation reaction rate is calculated based on the oxygen contact data of the oil samples. To perform this calculation, first, the oxygen contact time and the change data of the peroxide concentration of each sample under specific conditions are collected. The oxygen contact time is determined by the period of the set oxidation reaction experiment, usually detected within 30 minutes to 48 hours. The change in peroxide concentration is used as the basis for measuring the oxidation reaction rate, and the change threshold of the peroxide concentration is set to 0.2 mmol / L·h. The change in the peroxide concentration of the sample is measured by spectrophotometry at a specific wavelength (such as 300 nm), and the change rate of the peroxide concentration per unit time, that is, the oxidation reaction rate, is calculated. The oxidation reaction rate is set to mmol / kg·h, and this rate is calculated based on the oxygen contact data. Finally, the oxidation reaction rate of each oil sample is obtained, providing reference data for the subsequent steps.

[0174] Step S42: Divide the oil samples based on the oxidation reaction rate to obtain high-oxidation-rate oil samples and low-oxidation-rate oil samples;

[0175] In this embodiment, based on the oxidation reaction rate data obtained in step S41, all oil samples are divided into high-oxidation-rate oil samples and low-oxidation-rate oil samples. The threshold value of the oxidation reaction rate is set at 0.5 mmol / kg·h. Among all samples, oil samples with an oxidation reaction rate greater than this value are classified as high-oxidation-rate oil samples, and those with a rate lower than this value are classified as low-oxidation-rate oil samples. This threshold value is obtained through experimental data analysis and can effectively distinguish oil samples with higher and lower oxidation rates. Through this classification process, it can provide a sample basis for the viscosity aging analysis and color aging evaluation in the subsequent steps. The grouping of each sample needs to be verified through repeated experiments to ensure the accuracy of the classification, and the oxidation rate data of each sample should be recorded.

[0176] Step S43: Conduct a viscosity aging analysis on the high-oxidation-rate oil samples to generate viscosity aging data;

[0177] In this embodiment, high-oxidation-rate oil samples are selected and a viscosity aging analysis is performed using a rotational viscometer. The experimental temperature is set at 50 °C and the humidity at 65% to simulate the aging process of oil in a high-temperature environment. The viscosity of each sample is measured, with the test time set at 10 hours and the viscosity value measured every 1 hour. By recording the viscosity changes measured each time, the viscosity growth rate is determined, and the viscosity change value per hour is calculated. The threshold value of the viscosity change is set at 0.1 Pa·s. If the viscosity increase exceeds this value, the data at this point is recorded. The measurement of each sample needs to be carried out in at least three independent experiments to ensure the accuracy and reliability of the experimental results. By analyzing the viscosity aging data of the high-oxidation-rate oil samples, it provides data support for the subsequent aging degree evaluation.

[0178] Step S44: Conduct a color aging evaluation on the low-oxidation-rate oil samples to obtain color aging data;

[0179] In this embodiment, low-oxidation-rate oil samples are selected and the color change is measured using a spectrophotometer. The experimental conditions are set as follows: the light intensity is 1000 lux, the light source is a white LED lamp, and the test wavelength range is 400 - 700 nm. The color change of the sample is measured every 1 hour, and the change in the color absorption peak value of the sample in the visible light region is recorded. The threshold value of the color change is set at 0.03 units. If the color change exceeds this value during the experiment, it is considered that the sample has undergone significant color aging. The measurement of each low-oxidation-rate oil sample needs to be carried out in at least three experiments to ensure the stability and accuracy of the data. Finally, the color aging data of each sample is recorded for use in the subsequent oil aging evaluation.

[0180] Step S45: Evaluate the oil aging degree based on the viscosity aging data and the color aging data, thereby obtaining the oil aging degree data;

[0181] In this embodiment, by combining the viscosity aging data obtained in step S43 and the color aging data obtained in step S44, the aging degree of the oil is evaluated by the weighted average method. The weight of the viscosity aging data is set to 0.7, and the weight of the color aging data is set to 0.3. After combining the two, the comprehensive aging index of each oil sample is calculated. For each oil sample, it is classified according to its comprehensive aging index. The standard for classifying the aging degree of the oil is set as follows: samples with a comprehensive aging index between 0.1 and 0.2 are slightly aged, samples between 0.2 and 0.3 are moderately aged, and samples exceeding 0.3 are severely aged. The aging degree data of all samples need to be recorded and output for subsequent oil quality control and analysis.

[0182] Step S46: Construct an oil aging evaluation model based on the oil aging degree data to obtain the oil aging evaluation model.

[0183] In this embodiment, the oil aging degree data obtained in step S45 is sorted out. The viscosity aging data and the color aging data are used as independent variables, and the oil aging degree data is used as the dependent variable. The viscosity aging data includes the viscosity change rate (unit: Pa·s / h), the color aging data includes the color absorption change rate (unit: optical density unit / h), and the aging degree data is the comprehensive aging index (unitless). To ensure the accuracy of the model and the comparability of the data, all data is normalized using the maximum-minimum normalization method. The formula is: normalized value = (original value - minimum value) / (maximum value - minimum value), ensuring that the independent variables and the dependent variable have the same numerical range. Then, the least squares method is used for regression analysis. The normalized viscosity aging data and color aging data are input into the regression analysis model, and the regression coefficients between each independent variable and the dependent variable are calculated. The model fitting degree requirement is set as R 2The value is greater than 0.95 to determine whether the accuracy of the model meets the standard. If the model fitting degree is lower than the requirement, the data set is readjusted or interaction terms are added for further optimization to ensure that the regression equation can accurately reflect the linear relationship of the data. After completing the regression analysis, the final regression coefficients are extracted to construct an oil aging evaluation equation, and the equation form is: Y = aX1 + bX2 + c; where Y is the aging degree data, X1 is the viscosity aging data, X2 is the color aging data, a and b are the regression coefficients, and c is the constant term. Through this model, the aging degree of a new oil sample can be predicted based on the viscosity and color aging data of the oil sample. The application of the model needs to further verify its prediction accuracy. For this purpose, a part of the oil samples not involved in the model construction are selected as the test set, the viscosity and color aging data of the samples are input, the predicted aging degree is calculated using the evaluation model, and compared with the actual data to calculate the prediction error, ensuring that the mean value of the prediction error is less than 5%. Finally, the model is encapsulated to form a usable oil quality evaluation tool to support subsequent analysis and quality control work.

[0184] Optionally, this specification also provides an oil detection data analysis system based on multi-source data for performing the oil detection data analysis method based on multi-source data as described above. The oil detection data analysis system based on multi-source data includes:

[0185] An infrared spectrum signal enhancement module, configured to obtain multi-source oil detection data, perform near-infrared spectrum detection feature extraction based on the multi-source oil detection data to obtain infrared spectrum detection data; perform signal enhancement processing on the infrared spectrum detection data to generate enhanced infrared spectrum data;

[0186] An oil anomaly detection module, configured to perform physical and chemical index feature extraction based on the multi-source oil detection data to obtain physical and chemical index data; perform detection anomaly analysis based on the physical and chemical index data and the enhanced infrared spectrum data to obtain oil anomaly detection data;

[0187] An oxygen contact analysis module, configured to classify the oil anomaly detection data to obtain oil chemical composition anomaly data and oil microorganism anomaly data; perform oxygen contact analysis based on the oil chemical composition anomaly data and the oil microorganism anomaly data to obtain oxygen contact data;

[0188] An oil aging evaluation module, configured to perform oxidation reaction analysis based on the oxygen contact data to obtain oil oxidation reaction data; perform oil aging evaluation based on the oil oxidation reaction data to obtain oil aging degree data, and construct an oil aging evaluation model to obtain an oil aging evaluation model;

[0189] A quality inspection and evaluation module is used to perform quality inspection and evaluation on multi-source detection data of grease according to a grease aging evaluation model, so as to obtain grease quality inspection and evaluation data.

[0190] A grease detection data analysis system based on multi-source data according to the present invention. This system can implement any one of the grease detection data analysis methods based on multi-source data of the present invention. It is a medium for combining the operations and signal transmissions between each module to complete the grease detection data analysis method based on multi-source data. The internal modules of the system cooperate with each other, thereby improving the abnormal recognition rate and quality evaluation accuracy of grease detection.

[0191] Therefore, from any perspective, the embodiments should be regarded as exemplary and non-restrictive. The scope of the present invention is defined by the appended claims rather than the above description. Therefore, all changes falling within the meaning and scope of the equivalent elements of the application document are intended to be included in the present invention.

[0192] The above are only the specific embodiments of the present invention, enabling those skilled in the art to understand or implement the present invention. Various modifications to these embodiments will be obvious to those skilled in the art. The general principles defined herein can be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention will not be limited to these embodiments shown herein, but rather to the widest scope consistent with the principles and novel features invented herein.

Claims

1. A method for analyzing oil and fat detection data based on multi-source data, characterized in that: The following steps are involved: Step S1: acquiring oil multi-source detection data, and performing near infrared spectrum detection feature extraction based on the oil multi-source detection data, thereby obtaining infrared spectrum detection data; Performing signal enhancement processing on infrared spectrum detection data to generate enhanced infrared spectrum data; Step S2: extracting physical and chemical index features based on oil multi-source detection data to obtain physical and chemical index data; Perform abnormality analysis based on physical and chemical index data and enhanced infrared spectrum data to obtain abnormal oil and fat detection data; Step S3: classifying the oil abnormality detection data into types, thereby obtaining oil chemical component abnormality data and oil microbial abnormality data; Oxygen contact analysis is performed based on abnormal data of oil chemical composition and abnormal data of oil microorganisms to obtain oxygen contact data; Step S4: performing oxidation reaction analysis based on the oxygen contact data to obtain oil oxidation reaction data; performing oil aging assessment based on the oil oxidation reaction data to obtain oil aging degree data, and constructing an oil aging assessment model to obtain an oil aging assessment model; Step S5: Perform quality detection and evaluation on the oil multi-source detection data according to the oil aging evaluation model, so as to obtain oil quality detection and evaluation data.

2. The grease detection data analysis method based on multi-source data according to claim 1, characterized in that: Step S1 is specifically as follows: Step S11: acquiring oil multi-source detection data, and performing near infrared spectrum detection feature extraction based on the oil multi-source detection data, thereby obtaining infrared spectrum detection data; Step S12: performing noise filtering on the infrared spectrum detection data to obtain noise-reduced infrared spectrum detection data; Step S13: performing spectrum smoothing according to the de-noised infrared spectrum detection data, thereby obtaining infrared spectrum smoothed detection data; Step S14: performing signal enhancement and amplification on the infrared spectrum smoothing detection data to generate enhanced infrared spectrum data.

3. The grease detection data analysis method based on multi-source data according to claim 1 is characterized in that: Step S2 is specifically as follows: Step S21: extracting physical and chemical index features based on oil multi-source detection data, thereby obtaining physical and chemical index data; Step S22: performing component abnormality analysis on the physical and chemical index data to obtain component abnormality data; Step S23: performing frequency band anomaly analysis on the enhanced infrared spectrum data to obtain frequency band anomaly data; Step S24: performing grease abnormality detection integration according to the component abnormality data and the frequency band abnormality data, thereby obtaining grease abnormality detection data.

4. The method for analyzing grease detection data based on multi-source data according to claim 3, characterized in that: Step S22 is specifically as follows: Step S221: Perform acid value abnormality detection on the physical and chemical index data to obtain acid value abnormality data; Step S222: performing peroxide value abnormality detection on the physical and chemical index data to obtain peroxide value abnormality data; Step S223: clustering the component abnormality indexes according to the abnormal acid value data and the abnormal peroxide value data to generate the component abnormality indexes; Step S224: extracting component abnormality from the physical and chemical index data according to the component abnormality index, thereby obtaining component abnormality data.

5. The method for analyzing grease detection data based on multi-source data according to claim 3, characterized in that: Step S23 is specifically as follows: Step S231: performing frequency band decomposition on the enhanced infrared spectrum data, thereby obtaining high-frequency enhanced infrared spectrum data and low-frequency enhanced infrared spectrum data; Step S232: performing first-order derivative spectrum peak and valley detection based on the high-frequency enhanced infrared spectrum data to obtain high-frequency spectrum peak and valley data; Step S233: performing Lorentz fitting overlapping peak separation based on the low-frequency enhanced infrared spectrum data to obtain low-frequency overlapping peak data; Step S234: performing peak position shift detection on the high-frequency region spectrum peak and valley position data, thereby obtaining high-frequency region peak position shift data; Step S235: performing fitting residual analysis on the overlapping peak data in the low frequency region, thereby obtaining high residual data in the low frequency region; Step S236: performing frequency band anomaly integration according to the high frequency region peak offset data and the low frequency region high residual data, thereby obtaining frequency band anomaly data.

6. The method for analyzing grease detection data based on multi-source data according to claim 5, characterized in that: Step S232 is specifically as follows: Perform first-order derivative calculation based on high-frequency enhanced infrared spectrum data to obtain first-order derivative data; Performing zero crossing point identification on the first-order derivative data to obtain zero crossing point data; Perform second-order derivative calculation according to zero-crossing point data to obtain second-order derivative data; Divide according to the second-order derivative data, if the second-order derivative data is greater than zero, it is judged to be the peak position, and the peak data is obtained; If the second-order derivative data is less than zero, it is judged to be the valley position and the valley data is obtained; The peak and valley positions of the high-frequency spectrum are integrated according to the peak data and valley value data, so as to obtain the peak and valley position data of the high-frequency spectrum.

7. The method for analyzing grease detection data based on multi-source data according to claim 5, characterized in that: Step S233 is specifically as follows: Partitioning is performed based on the low-frequency enhanced infrared spectrum data, thereby obtaining low-frequency enhanced infrared spectrum partitioning data; The parameters of the Lorentz function are set for the low-frequency enhanced infrared spectrum partition data, thereby obtaining the parameters of the Lorentz function; The single-peak interval fitting is performed on the low-frequency enhanced infrared spectrum partition data according to the Lorentz function parameters, thereby obtaining the single-peak fitting data; The multi-peak combination of low-frequency enhanced infrared spectrum data is performed according to the single-peak fitting data to obtain overlapping peak data; The overlapping peak data are fitted and calculated to obtain the overlapping peak data in the low-frequency region.

8. The method for analyzing grease detection data based on multi-source data according to claim 1, characterized in that: Step S3 is specifically as follows: Step S31: Obtaining standard data of oil and fat composition; Step S32: classifying the oil abnormality detection data according to the oil component standard data, thereby obtaining oil chemical component abnormality data and oil microbial abnormality data; Step S33: screening oil samples based on the oil chemical component abnormality data to obtain oil chemical component abnormality samples; Step S34: performing peroxide determination on the oil and fat chemical composition abnormal sample to obtain peroxide data; Step S35: screening oil samples based on the oil microorganism abnormality data to obtain oil microorganism abnormality samples; Step S36: measuring the lipase activity of the oil microorganism abnormal sample to obtain lipase activity data; Step S37: performing oil and fat oxygen contact assessment based on the peroxide data and the lipase activity data to obtain oxygen contact data.

9. The method for analyzing grease detection data based on multi-source data according to claim 1, characterized in that: Step S4 is specifically as follows: Step S41: Calculating the oxidation reaction rate according to the oxygen contact data to obtain the oxidation reaction rate; Step S42: dividing the oil sample based on the oxidation reaction rate, thereby obtaining the oil sample with high oxidation rate and the oil sample with low oxidation rate; Step S43: performing viscosity aging analysis on the high oxidation rate oil sample to generate viscosity aging data; Step S44: performing color aging evaluation on the low oxidation rate oil sample to obtain color aging data; Step S45: evaluating the grease aging degree according to the viscosity aging data and the color aging data, thereby obtaining grease aging degree data; Step S46: constructing a grease aging assessment model according to the grease aging degree data to obtain a grease aging assessment model.

10. A grease detection data analysis system based on multi-source data, characterized in that: Used to execute the grease detection data analysis method based on multi-source data as claimed in claim 1, the grease detection data analysis system based on multi-source data comprises: The infrared spectrum signal enhancement module is used to obtain the oil multi-source detection data, and perform near-infrared spectrum detection feature extraction based on the oil multi-source detection data, so as to obtain infrared spectrum detection data; perform signal enhancement processing on the infrared spectrum detection data to generate enhanced infrared spectrum data; The oil anomaly detection module is used to extract physical and chemical index features based on oil multi-source detection data to obtain physical and chemical index data; perform detection anomaly analysis based on physical and chemical index data and enhanced infrared spectrum data to obtain oil anomaly detection data; The oxygen contact analysis module is used to classify the oil abnormality detection data into types, so as to obtain oil chemical composition abnormality data and oil microbial abnormality data; oxygen contact analysis is performed based on the oil chemical composition abnormality data and oil microbial abnormality data, so as to obtain oxygen contact data; The oil aging assessment module is used to perform oxidation reaction analysis based on oxygen contact data to obtain oil oxidation reaction data; perform oil aging assessment based on the oil oxidation reaction data to obtain oil aging degree data, and construct an oil aging assessment model to obtain an oil aging assessment model; The quality detection and evaluation module is used to perform quality detection and evaluation on the oil multi-source detection data according to the oil aging evaluation model, so as to obtain oil quality detection and evaluation data.

Images