Big data-driven organic tea oil quality monitoring and early warning method

Abstract

The present invention discloses a big data-driven organic tea oil quality monitoring and early warning method, which relates to the technical field of tea oil quality assessment, and includes the following steps: real-time collection of tea oil odor components by electronic nose technology, and inputting the collected odor data into a data management system. The present invention dynamically calculates the odor change coefficient, odor level correlation index, and time series fluctuation index through a machine learning model, and can accurately capture the initial stage of odor change, dynamically adjust the sensor response time, and improve the detection accuracy of abnormal odor change periods. This big data-driven system improves the accuracy and real-time performance of tea oil quality monitoring, effectively reduces human errors, reduces the risk of quality problems, ensures consumer safety, and enhances corporate brand reputation.

Description

Technical Field

[0001] The present invention relates to the technical field of tea oil quality assessment, and in particular to a big data driven organic tea oil quality monitoring and early warning method. Background Art

[0002] Big data-driven organic tea oil quality monitoring and early warning refers to the use of big data technology to monitor the quality of tea oil in real time by collecting, storing and analyzing a large amount of data related to organic tea oil, and automatically issuing early warning signals when abnormalities or potential risks occur, so that corresponding measures can be taken. Specifically, this method includes obtaining a large amount of data (such as temperature and humidity, chemical composition, sensory characteristics, etc.) from the production, processing, storage and transportation of tea oil, and processing this information through big data analysis technology (such as data mining, machine learning, etc.) to build a tea oil quality monitoring model. On this basis, the system can evaluate the quality status of tea oil in real time, predict possible quality problems (such as oxidation, contamination, etc.), and promptly issue alarms to relevant personnel when potential quality fluctuations or risks are found, thereby optimizing management processes, improving the quality control level of tea oil products, and ensuring consumer safety.

[0003] The prior art has the following deficiencies:

[0004] When the existing technology monitors the aroma of tea oil in storage, the electronic nose usually presets the sensor response time based on historical data to ensure that the changes in the smell of tea oil can be captured in time. Although this response time set based on historical data can meet most monitoring needs, the odor components of tea oil are usually composed of a variety of volatile organic compounds, and these components may be complicatedly superimposed and interfered during the deterioration process. In particular, when new odor components (such as oxides) overlap with the original aroma components, if the sensor response time is not adjusted in time to cope with these changes, a series of serious consequences may occur. The complex odor changes of volatile organic compounds in tea oil are an important sign of the deterioration process. If the electronic nose fails to accurately identify these changes, the oxidation, rancidity or other quality problems of tea oil may be ignored, which will lead to expired oil entering the market and affecting product quality. After consumers use spoiled tea oil, they may suffer from health risks such as poisoning, indigestion, skin allergies, and even food safety incidents in severe cases. If companies fail to monitor the quality of tea oil in a timely manner, they may face serious consequences such as quality complaints, product recalls, and loss of brand reputation. Therefore, timely and accurate detection of odor changes is crucial to ensuring tea oil quality, consumer safety and corporate reputation.

[0005] The above information disclosed in this Background section is only for enhancement of understanding of the background of the present disclosure and therefore it may contain information that does not constitute the prior art that is already known to one of ordinary skill in the art. Summary of the invention

[0006] The purpose of the present invention is to provide a big data-driven organic tea oil quality monitoring and early warning method. The present invention dynamically calculates the odor change coefficient, the odor level correlation index and the time series fluctuation index through a machine learning model, which can accurately capture the initial stage of odor change, dynamically adjust the sensor response time, and improve the detection accuracy of the abnormal period of odor change. This big data-driven system improves the accuracy and real-time performance of tea oil quality monitoring, effectively reduces manual errors, reduces the risk of quality problems, ensures consumer safety, and enhances corporate brand reputation to solve the problems in the above-mentioned background technology.

[0007] In order to achieve the above object, the present invention provides the following technical solution: a big data driven organic tea oil quality monitoring and early warning method, comprising the following steps:

[0008] The odor components of tea oil are collected in real time through electronic nose technology, and the collected odor data are input into the data management system;

[0009] The collected tea oil odor data is used to construct an analysis set, and key features reflecting odor changes are extracted from it;

[0010] The extracted key features are analyzed and processed within the detection window, and the analyzed feature data are input into a pre-trained machine learning model, and the machine learning model is used to predict the quality change of tea oil;

[0011] Based on the prediction results of the machine learning model, the odor change process of tea oil is divided into a stable odor characteristic period and an abnormal odor characteristic period;

[0012] During the odor characteristic stabilization period, the odor of tea oil changes steadily and meets the normal quality standards. Odor monitoring continues according to the sensor response time preset by the electronic nose to ensure that the quality of tea oil remains within the normal range;

[0013] For periods of abnormal odor characteristics, the response time of the sensor is dynamically extended based on the odor change characteristics evaluated by the machine learning model to ensure that the changes in tea oil odor are fully analyzed and the subtle changes in odor are accurately captured.

[0014] Preferably, key features reflecting odor changes are extracted from the analysis set, and the extracted features include the correlation between different odor components and the time series fluctuation of odor changes. Under the detection window, the correlation between different odor components and the time series fluctuation of odor changes are analyzed and processed to generate an odor hierarchical correlation index and an odor time series fluctuation index, respectively. The odor hierarchical correlation index quantifies the degree of correlation between different odor components, the interaction between various volatile organic compounds and their concentration changes during the deterioration of tea oil; the odor time series fluctuation index quantifies the fluctuation of tea oil odor changes over time. During the deterioration of tea oil, the changes in odor components often manifest as irregular and nonlinear fluctuations.

[0015] Preferably, after obtaining the odor level association index and the odor time series fluctuation index generated after analyzing and processing the extracted key features, the odor level association index and the odor time series fluctuation index are input into a pre-learned machine learning model, and the odor variation coefficient is generated by the machine learning model, and the quality change of tea oil is predicted by the odor variation coefficient.

[0016] Preferably, the odor change coefficient generated after analyzing the extracted key features is compared with a preset odor change coefficient reference threshold value to divide the odor change process of tea oil, and the division steps are as follows:

[0017] If the odor change coefficient is greater than or equal to a preset odor change coefficient reference threshold, the odor change process of the tea oil is classified as an odor characteristic abnormal period;

[0018] If the odor change coefficient is less than a preset odor change coefficient reference threshold, the odor change process of the tea oil is divided into an odor characteristic stable period.

[0019] Preferably, based on the smell change characteristics evaluated by the machine learning model, the specific steps of dynamically extending the response time of the sensor to ensure full analysis of the smell changes of tea oil and accurately capture the subtle changes in the smell changes are as follows:

[0020] Once the odor abnormality period is identified, the odor change coefficient The value of is used to dynamically adjust the sensor's response time. The calculation expression is as follows: , where is the adjusted sensor response time, is the preset sensor response time, is the preset reference threshold value of the odor variation coefficient, is the adjustment coefficient;

[0021] During the odor abnormality period, the extended response time depends not only on the odor variation coefficient, but also on the variation characteristics between odor components. By analyzing the variation of volatile compounds in the odor time series, the volatile compound variation index is introduced to quantify the fluctuation degree of the odor time series. The calculation expression is as follows: , where is the volatile compound variation index, It's time point The measured value of the odor components at the moment, is the time step index, is the total number of time steps;

[0022] Based on the adjusted sensor response time and Volatile Compound Variation Index , generate the final response time adjustment, the calculation expression is as follows: , where is the response time adjustment, Is the weighted coefficient, which is used to balance the odor variation coefficient and Volatile Compound Variation Index The degree of influence on the adjustment of sensor response time;

[0023] Adjust the response time by With preset sensor response time Add together to get the final sensor response time, the calculation expression is as follows: , where is the final adjusted response time, which indicates the final sensor response time after adjustment during the abnormal odor period.

[0024] Preferably, in the detection window, the correlation between different odor components is analyzed and processed to generate the odor level correlation index in the following specific steps:

[0025] Under the detection window, the electronic nose technology is used to collect the odor component data of tea oil, and based on the acquired data, the concentration matrix of the tea oil odor components is generated. , ,in, It's time point Moment, The concentration of the odor components, is the total amount of tea oil odor components;

[0026] By calculating the correlation between odor components, the interaction relationship between different odor components is revealed. The calculation expression is as follows: , where is the odor level association value, indicating the time point Moment, Odor components and The correlation between the odor components, It's time point Moment, The concentration of the odor components, It is Odor components and Odor components at time points The relevance of time, It is Odor components at time points The individual probability distributions at time instants, It is Odor components at time points The individual probability distributions at time instants, It is Odor components and Odor components at time points The joint probability distribution of time;

[0027] Based on the odor level association value The hierarchical correlation of each odor component in tea oil was evaluated and the odor hierarchical correlation index was generated. The calculation expression is as follows: , where is the odor hierarchy association index, It is Odor components and The weight of the odor components, is the Euclidean distance between odor components, is the attenuation factor.

[0028] Preferably, within the detection window, the specific steps of analyzing and processing the time series fluctuation of the odor change and generating the odor time series fluctuation index are as follows:

[0029] In the detection window, the odor component data collected by the electronic nose technology exists in the form of a time series, where each data point represents the odor concentration at each time point. To ensure the accuracy of subsequent analysis, the data needs to be smoothed, denoised, and time-aligned;

[0030] Calculate the rate of change of odor components and capture the fluctuation characteristics of odor. The calculation expression is as follows: , where is the rate of change of the odor components, It's time point The odor concentration value at the moment, It's the previous time point The odor concentration value at the moment, is the unit time span;

[0031] In order to evaluate the long-term fluctuation trend of odor changes, the regularity of the time series is calculated and the regularity is quantified by calculating the fractal dimension of the time series. The expression is as follows: , where is the regularity index, is the odor concentration value, i.e., the time point The odor concentration value at the moment, Refers to A point in time, is the total number of time points, is an index that controls local variation;

[0032] Rate of change based on odor components and regularity index Generate the odor time series volatility index, the calculation expression is as follows: , where is the odor time series volatility index, and Both are adjustment parameters, which respectively control the change rate of odor components and the influence weight of regularity index in the odor time series fluctuation index. It is a point in time.

[0033] In the above technical solution, the technical effects and advantages provided by the present invention are:

[0034] Traditional electronic nose technology relies on the sensor response time preset by historical data, but this method may not be able to respond to the complex superposition and interference of odor components in a timely manner during the change of tea oil odor, especially in the early stage of tea oil deterioration. When tea oil is oxidized, rancid or has other quality problems, the new odor components overlap with the original odor components, which may cause the correlation between the odor components to change, making it difficult to capture subtle signs of odor changes. The present invention can accurately judge the early stage of odor changes and respond in time, accurately capture the abnormal period of odor changes, and effectively avoid tea oil quality problems from being ignored by dynamically calculating the odor change coefficient, odor hierarchy correlation index, and odor time series fluctuation index evaluated based on a machine learning model. By dynamically adjusting the response time of the electronic nose sensor, it is possible to better analyze complex odor components during the abnormal period of odor characteristics, ensure detection accuracy, and avoid quality problems caused by response delays from entering the market.

[0035] The present invention can maintain stable monitoring during the stable period of the tea oil's odor characteristics by real-time monitoring and intelligent evaluation of tea oil's odor changes, ensuring that the tea oil quality meets normal standards, and more accurately analyze the odor changes by dynamically adjusting the sensor response time during the abnormal period. Through this big data-driven monitoring system, companies can obtain accurate feedback on tea oil quality in real time, greatly improving the company's quality control efficiency. This intelligent monitoring can not only reduce manual operation errors, but also detect potential quality problems in a timely manner, effectively avoiding product recalls, complaints, and food safety incidents, thereby greatly improving the company's product quality management level, reducing risks, and strengthening the company's brand reputation to ensure the health, safety, and interests of consumers. BRIEF DESCRIPTION OF THE DRAWINGS

[0036] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in the present invention. For ordinary technicians in this field, other drawings can also be obtained based on these drawings.

[0037] Figure 1 This is a method flow chart of the big data driven organic tea oil quality monitoring and early warning method of the present invention. DETAILED DESCRIPTION

[0038] Example embodiments will now be described more fully with reference to the accompanying drawings. However, example embodiments can be implemented in a variety of forms and should not be construed as limited to the examples set forth herein; rather, these example embodiments are provided so that the description of the present disclosure will be more comprehensive and complete, and the concept of the example embodiments will be fully conveyed to those skilled in the art.

[0039] The present invention provides Figure 1 The big data driven organic tea oil quality monitoring and early warning method shown includes the following steps:

[0040] The odor components of tea oil are collected in real time through electronic nose technology, and the collected odor data are input into the data management system;

[0041] When the electronic nose technology is used to analyze the odor components of tea oil in real time, the tea oil odor is first sampled and detected based on the preset sensor response time of the electronic nose. The sensor collects volatile organic compounds (VOCs) in the tea oil odor at appropriate time intervals based on historical data and the set response time. These compounds may reflect the quality, storage status and deterioration process of the tea oil. The sensor continuously monitors the odor changes during this response time and generates corresponding odor data. The collected odor data is then input into the data management system in real time for further analysis and processing. In this way, the electronic nose can efficiently and real-time capture the subtle changes in the odor of tea oil, and provide basic data support for subsequent quality assessment, prediction and alarm.

[0042] The collected tea oil odor data is used to construct an analysis set, and key features reflecting odor changes are extracted from it;

[0043] The specific steps of constructing an analysis set of the collected tea oil odor data include: first, preliminarily cleaning and screening the tea oil odor data collected in real time, and removing noise data and outliers to ensure the accuracy and reliability of the data. Next, based on the main components and change patterns of tea oil odor, characteristic parameters related to quality changes, such as the concentration, odor intensity, and frequency characteristics of specific volatile organic compounds (VOCs), are selected to construct an analysis set containing these key information. In this process, the data also needs to be standardized and normalized for subsequent analysis and comparison. Finally, the odor data from different time periods are integrated to form a comprehensive analysis data set.

[0044] The purpose of establishing an analysis set is to provide an orderly and structured data foundation to facilitate subsequent feature extraction and machine learning model training. Through the analysis set, a large amount of odor information can be effectively integrated, and key features that can reflect the changes in the odor of tea oil can be extracted from it. These features are of great value for subsequent quality monitoring, trend prediction, and anomaly detection. Through in-depth analysis of these features, the system can identify quality changes and potential deterioration risks of tea oil, provide a scientific basis for subsequent quality assessment and early warning, and ensure the stability and safety of tea oil quality.

[0045] The extracted key features are analyzed and processed within the detection window, and the analyzed feature data are input into a pre-trained machine learning model, and the machine learning model is used to predict the quality change of tea oil;

[0046] Key features reflecting odor changes are extracted from the analysis set. The extracted features include the correlation between different odor components and the time series fluctuation of odor changes. Under the detection window, the correlation between different odor components and the time series fluctuation of odor changes are analyzed and processed to generate the odor hierarchical correlation index and odor time series fluctuation index respectively. The odor hierarchical correlation index quantifies the degree of correlation between different odor components, the interaction between various volatile organic compounds (VOCs) and their concentration changes during the deterioration of tea oil; the odor time series fluctuation index quantifies the fluctuation of tea oil odor changes over time. During the deterioration of tea oil, the changes in odor components often manifest as irregular and nonlinear fluctuations.

[0047] After obtaining the odor level association index and the odor time series fluctuation index generated by analyzing and processing the extracted key features, the odor level association index and the odor time series fluctuation index are input into the pre-learned machine learning model, and the odor change coefficient is generated by the machine learning model, and the quality change of tea oil is predicted by the odor change coefficient;

[0048] A pre-learned machine learning model usually refers to a model that has been trained and optimized on a data set, a process called "model training". During this training process, the model gradually learns and adjusts the model's parameters through input features in a large amount of historical data (such as the concentration of odor components, hierarchical correlation, time series fluctuations, etc.) and known output labels (such as the state of tea oil quality, whether it has deteriorated, etc.), so that it can accurately predict or classify new, unseen data. In the tea oil odor monitoring system, the pre-learned machine learning model "understands" the laws of tea oil odor changes by learning a large amount of labeled odor data (including tea oil odor data in normal and abnormal periods), and can effectively predict future odor data based on these laws.

[0049] For tea oil quality monitoring, the training process not only includes the identification of various odor components, but also involves the analysis of complex features such as odor hierarchical correlation and time series volatility. By processing these features, the machine learning model can "learn" how to distinguish normal odor changes from abnormal odor changes. For example, when tea oil enters the stage of quality problems such as oxidation and rancidity, the change in odor is not only manifested as an increase or decrease in the concentration of certain specific odor components, but may also involve complex interactions between components (such as the production of oxides and the overlap of aroma components). These patterns of changes are closely related to the decline in tea oil quality. Through the analysis and learning of historical data, the machine learning model can identify these changing patterns and provide accurate predictions for quality changes in tea oil.

[0050] The core value of a pre-learned machine learning model is that it can process large amounts of complex input data and automatically extract useful information from it. For the tea oil quality monitoring system, the input features include the concentration of each odor component in the tea oil, the hierarchical correlation index (OHI) between different components, and the odor time series volatility index (TOVI). These input features reflect the changing characteristics of the odor, and by training on historical data sets, the model learns how to identify patterns of odor changes under these features. When new data enters the system, the model can generate an odor change coefficient (OCCI) by processing these input features. This coefficient can reflect the current odor change trend of the tea oil and determine whether there is a quality problem based on the preset threshold.

[0051] This prediction method based on machine learning has strong adaptability and accuracy. Especially when tea oil enters an abnormal period of odor characteristics, the system can judge the abnormality of the odor change through the pre-learned model. Through the prediction results of the model, the production line can adjust the storage conditions of the tea oil in time or take measures to prevent the tea oil from further deteriorating, thereby effectively ensuring the quality of the tea oil and the safety of consumers. Unlike the traditional single sensor method, the machine learning model can achieve higher accuracy and reliability of quality prediction by integrating multiple odor components and their interrelationships, and as the amount of data increases, the prediction accuracy of the model will continue to improve.

[0052] When the correlation between different odor components fluctuates greatly, it usually indicates that the smell of tea oil has changed abnormally. This fluctuation reflects that the concentration relationship of volatile organic compounds (VOCs) in tea oil has changed significantly, which means that the odor components of tea oil are undergoing complex superposition and transformation. Under normal circumstances, the odor components of tea oil are relatively stable, and various volatile organic compounds (such as aromatic hydrocarbons, esters, etc.) show a certain correlation and work synergistically in a specific proportion to form the unique aroma of tea oil. However, during the deterioration of tea oil, with the occurrence of oxidation, rancidity or mold, new odor components (such as oxides, acidic substances, etc.) will gradually be generated and interact with the original aroma components. The generation of these new components and their overlap with the original components will cause drastic changes in the structure of the tea oil odor, thereby destroying the originally stable odor correlation. At this time, the proportional relationship between the originally closely related odor components fluctuates greatly, which is manifested as a sharp decline in correlation or irregular fluctuations, becoming an important signal that the tea oil has entered the deterioration stage, and quality assessment and control measures need to be taken immediately.

[0053] Under the detection window, the correlation between different odor components is analyzed and processed to generate the odor hierarchical correlation index. The specific steps are as follows:

[0054] Under the detection window, the electronic nose technology is used to collect the odor component data of tea oil, and based on the acquired data, the concentration matrix of the tea oil odor components is generated. , ,in, It's time point Moment, The concentration of the odor components, is the total amount of tea oil odor components;

[0055] The concentration matrix of tea oil odor components is used to comprehensively record and analyze the changes in the concentration of various volatile organic compounds (VOCs) in tea oil over time. Each matrix element represents the concentration value of a specific odor component at a specific time point. By organizing and storing these concentration values ​​by time and odor component, a multidimensional data structure is formed. Each row in the matrix usually corresponds to a different odor component, and each column corresponds to a different time point. Through this concentration matrix, the dynamic changes of tea oil odor components can be systematically tracked, providing basic data for subsequent odor change analysis and quality prediction.

[0056] The role of the concentration matrix is ​​that it provides raw data support for the quantitative analysis of odor changes, allowing for a systematic analysis of the correlation and volatility between various odor components. In addition, the concentration matrix can serve as the basis for input data to machine learning models, helping to identify abnormal odor changes by extracting time series features of odor component concentrations, and thus assessing the quality status of tea oil. Ultimately, this matrix structure not only helps monitor tea oil quality in real time, but also provides strong data support for predicting its shelf life and detecting possible quality issues.

[0057] By calculating the correlation between odor components, the interaction relationship between different odor components is revealed. The calculation expression is as follows:

[0058] , where is the odor level association value, indicating the time point Moment, Odor components and The correlation between the odor components, It's time point Moment, The concentration of the odor components, It is Odor components and Odor components at time points The relevance of time, It is Odor components at time points The individual probability distributions at time instants, It is Odor components at time points The individual probability distributions at time instants, It is Odor components and Odor components at time points The joint probability distribution of time;

[0059] The odor hierarchy correlation value refers to a quantitative indicator of the relationship between multiple odor components. It reveals how odor components interact, influence each other, and their changing trends by analyzing the correlation between the concentrations of different odor components. Specifically, the odor hierarchy correlation value evaluates the synergistic change pattern between odor components at different time points or conditions by mathematically modeling the concentration data of odor components. Its function is to effectively capture the interaction of odor components, especially in the process of quality change of tea oil, with the occurrence of changes such as oxidation and rancidity, the odor components may undergo complex superposition. Through the odor hierarchy correlation value, the system can identify whether there is abnormal interaction between odor components, so as to promptly discover potential quality problems or abnormal changes, and provide an important basis for subsequent quality control and adjustment.

[0060] A single probability distribution is the probability distribution of a random variable under all possible outcomes, ignoring the influence of other related variables. It represents the behavior or characteristics of a single variable, independent of the existence of other variables. For example, in odor component analysis, a single probability distribution can be used to describe the probability of occurrence of each odor component without considering the existence of other odor components.

[0061] Joint probability distribution refers to the probability distribution of multiple random variables together, which describes the joint behavior or relationship between these variables. In odor component analysis, joint probability distribution can be used to describe the probability of multiple odor components appearing at the same time, reflecting the interaction and correlation between different components.

[0062] Individual probability distributions are used to understand the characteristics of a single variable, such as the concentration trend of a single odor component, to help identify its range of variation and normal fluctuations; while joint probability distributions are used to reveal the relationships between multiple variables, to help understand how odor components change together, and whether there are complex interactions or superpositions, thereby effectively identifying the overall pattern of odor changes, especially when multiple factors change, such as the superposition of volatile organic compounds in tea oil.

[0063] Based on the odor level association value

[0064] The hierarchical correlation of each odor component in tea oil was evaluated and the odor hierarchical correlation index was generated. The calculation expression is as follows: , where is the odor hierarchy association index, It is Odor components and The weights of the odor components indicate their relative importance. is the Euclidean distance between odor components, is the attenuation factor.

[0065] The Euclidean distance between odor components is a metric used to measure the "distance" between two odor components in the feature space. In odor analysis, each odor component can be regarded as a multidimensional feature vector, where each dimension represents the concentration or attribute of the odor component under different conditions. The Euclidean distance quantifies the similarity between the two odor components by calculating the straight-line distance between them in the feature space. Specifically, the smaller the Euclidean distance, the smaller the difference between the two odor components in the feature space, which means that their odor characteristics are more similar; the larger the Euclidean distance, the greater the difference between the odor components, which means that their odor characteristics are more significantly different. In the odor monitoring of tea oil, the Euclidean distance can be used to evaluate the similarity between different odor components and help identify whether the odor change is caused by the change of certain components. For example, if the concentration change of certain volatile organic compounds causes a large change in their position in the feature space (i.e., the Euclidean distance increases), it may mean that the odor of the tea oil has undergone an abnormal change, such as oxidation or rancidity. By calculating the Euclidean distance, the change pattern of different odor components can be quickly detected in the multidimensional feature space, helping to identify potential quality problems in a timely manner.

[0066] The attenuation factor is a parameter used to describe the change in the strength of association between odor components over time or distance. Its function is to simulate the influence of time continuity or physical dispersion of odor components during the change process by reducing the strength of association. Specifically, the attenuation factor can control the influence between odor components to gradually weaken over time, or decrease with the increase of spatial distance, thereby accurately describing the change in the relationship between odor components at different time or spatial scales.

[0067] Under the detection window, the larger the odor hierarchy correlation index performance value generated after analyzing the correlation between different odor components, it usually indicates that the odor of tea oil has undergone abnormal changes. Under normal conditions, the odor components of tea oil have a relatively stable correlation, and the concentration changes between different volatile organic compounds (VOCs) show a certain regularity and synergy. However, during the deterioration of tea oil, especially when quality problems such as oxidation, rancidity or mold occur, new odor components (such as oxides, acidic substances, etc.) will gradually be generated and overlap with the original odor components. The emergence of these new components breaks the stable correlation between the original odor components, making the relationship between them more complex and irregular, resulting in an increase in the odor hierarchy correlation index. On the contrary, when the odor hierarchy correlation index is small or remains stable, it means that the relationship between the odor components of tea oil has not changed significantly, indicating that the odor of tea oil has not undergone abnormal changes.

[0068] The time series fluctuations of odor changes show irregular and nonlinear fluctuations, which usually indicate abnormal changes in the odor of tea oil, especially in the process of tea oil deterioration. When the volatile organic compounds (VOCs) in tea oil are oxidized, rancid or moldy, the changes in its odor components often show complex superposition and interference, which is different from the normal, linear odor changes. In the deterioration stage, the newly generated odor components (such as oxides, sourness, etc.) will interact with the original aroma components, resulting in nonlinear fluctuations in the concentration, component ratio and change rate of the odor. For example, the generation of oxides may react with certain aroma components in tea oil, causing the originally stable odor components to suddenly disappear or the concentration to fluctuate violently, and these changes cannot be described by a simple linear model. In this process, the changes in odor components are not only complex and changeable, but also the speed, intensity and order of change are extremely difficult to predict, resulting in irregular time series fluctuations. Therefore, the nonlinear fluctuations of the time series are an important sign of abnormal changes in the odor of tea oil, which can reflect the complex interactions of volatile organic compounds in tea oil and the formation of new odor components.

[0069] Under the detection window, the time series fluctuation of odor changes is analyzed and processed, and the specific steps of generating the odor time series fluctuation index are as follows:

[0070] In the detection window, the odor component data collected by the electronic nose technology exists in the form of a time series, where each data point represents the odor concentration at each time point. To ensure the accuracy of subsequent analysis, the data needs to be smoothed, denoised, and time-aligned; the rate of change of the odor component is calculated to capture the fluctuation characteristics of the odor. The calculation expression is as follows: , where is the rate of change of the odor components, It's time point The odor concentration value at the moment, It's the previous time point The odor concentration value at the moment, It is the unit time span, which refers to the time interval used to measure the change of odor during the analysis process. It is used to normalize the rate of change of odor to ensure that the odor changes in different time periods can be compared and evaluated within a unified time frame.

[0071] Calculating the rate of change of odor components and capturing the fluctuation characteristics of odor are mainly to accurately identify and monitor the dynamic changes of tea oil odor, especially during the deterioration of tea oil. As the storage time of tea oil increases, the volatile organic compounds (VOCs) in the odor components will undergo complex changes, especially when oxidation, rancidity or other quality problems occur, the odor components may fluctuate irregularly and nonlinearly. By calculating the rate of change of odor components, the rate and amplitude of odor fluctuations can be revealed, helping to identify possible abnormal changes in odor in a timely manner. Capturing these fluctuation characteristics helps to distinguish key odor change signals from normal changes and abnormal changes, thereby providing an important basis for subsequent quality assessment and early warning. This process can not only optimize the response time of the detection sensor, but also effectively prevent deteriorated tea oil from entering the market, ensuring the quality of tea oil and consumer safety.

[0072] In order to evaluate the long-term fluctuation trend of odor changes, the regularity of the time series is calculated and the regularity is quantified by calculating the fractal dimension of the time series. The expression is as follows: , where is the regularity index, is the odor concentration value, i.e., the time point The odor concentration value at the moment, Refers to A point in time, is the total number of time points, It is an index that controls local changes and is used to adjust the sensitivity when calculating regularity;

[0073] The "fractal dimension" of a time series is an indicator used to quantify the regularity of a time series, which reflects the complexity and irregularity of the time series. The concept of fractal dimension originates from fractal geometry, and refers to whether the details of a time series present similar patterns at different scales. When a time series has regularity, its patterns and structures at different scales are similar, but the details may be finer or coarser. By calculating the fractal dimension, we can quantify this regularity and reveal the degree of fluctuation and irregularity of the series over time. If a time series has a high fractal dimension, it means that it has strong complexity and irregularity; conversely, it means that the series is relatively stable or regular. Therefore, the use of the fractal dimension of a time series can effectively capture the nonlinear characteristics and complex dynamics hidden in the time series, especially for nonlinear systems whose change patterns cannot be described by traditional statistical methods (such as mean and variance), such as the complex fluctuation patterns in the changes in tea oil odor. Fractal dimension provides a way to quantitatively analyze such complex dynamics, which helps to identify potential anomalies or changing trends in advance.

[0074] The regularity index is an indicator used to measure the degree of regularity or repeatability in time series data. It reflects whether the data presents a stable, predictable pattern or periodic fluctuations within a certain time frame. The index usually determines its regularity by analyzing the trend, periodic changes and self-similarity in the data. If the regularity index is high, it means that the time series presents obvious regularity, the changes are relatively stable and easy to predict, which usually indicates that the system is in a normal or healthy state; on the contrary, a low regularity index may indicate that the time series shows more randomness or irregular fluctuations, which is usually a signal of abnormality or problems in the system. For the monitoring of tea oil odor, the regularity index can help detect the stability of odor changes, and then identify whether the tea oil is in a state of deterioration or abnormal quality. If the odor changes lose their regularity, it may mean that oxidation, rancidity or other quality problems have occurred in the tea oil.

[0075] Rate of change based on odor components and regularity index Generate the odor time series volatility index, the calculation expression is as follows: , where is the odor time series volatility index, and Both are adjustment parameters, which respectively control the change rate of odor components and the influence weight of regularity index in the odor time series fluctuation index. It is a point in time.

[0076] Under the detection window, the larger the value of the odor time series fluctuation index generated after analyzing and processing the time series fluctuation of odor changes, the greater the value, which means that the relationship between the odor components has changed significantly, and new odor components (such as oxides, sourness, etc.) may appear superimposed on the original odor. On the contrary, if the fluctuation index is small, it means that the odor changes relatively smoothly, which is consistent with the normal storage state of tea oil. At this time, the quality of tea oil does not change significantly, and the odor does not fluctuate abnormally.

[0077] The machine learning model is not limited here, and can realize the association index of odor levels and the odor time series volatility index Comprehensive analysis to generate odor variation coefficient The machine learning model can be used. In order to realize the technical solution of the present invention, the present invention provides a specific implementation method;

[0078] Odor variation coefficient The generation formula is as follows: , where , Odor Hierarchy Relation Index and the odor time series volatility index The preset scaling factor of , Both are greater than 0.

[0079] From the calculation expression of the odor variation coefficient, it can be seen that the larger the odor hierarchy correlation index expression value generated after analyzing and processing the correlation between different odor components under the detection window, the larger the odor time series fluctuation index expression value generated after analyzing and processing the time series fluctuation of odor changes, and the larger the odor variation coefficient expression value generated after analyzing the extracted key features under the detection window, indicating that during the deterioration process, new odor components (such as oxides, sour taste, etc.) may be superimposed and interfered with the original aroma components, causing the odor change to become more complicated. Otherwise, it means that the odor component changes are stable and the tea oil odor has not changed abnormally.

[0080] The preset proportionality coefficient refers to the pre-set value in the machine learning model or mathematical formula used to weight the odor level correlation index. and the odor time series volatility index The constant value of the contribution. These coefficients and Used to measure the coefficient of change of two indicators in odor In other words, the preset proportionality coefficient is calculated by the odor level correlation index and the odor time series volatility index Different weights are assigned to adjust their influence in the final result. Their values ​​are usually determined by experimental data, domain knowledge, or the training process of machine learning algorithms to ensure that the odor variation coefficient It can accurately reflect the actual situation of tea oil odor changes. In this formula, and Both are greater than 0, indicating that these two indicators have a positive impact on the odor change coefficient, and their influence is controlled by the size of these coefficients.

[0081] Based on the prediction results of the machine learning model, the odor change process of tea oil is divided into a stable odor characteristic period and an abnormal odor characteristic period;

[0082] The odor change coefficient generated after analyzing the extracted key features is compared with the preset odor change coefficient reference threshold, and the odor change process of tea oil is divided. The division steps are as follows:

[0083] If the odor change coefficient is greater than or equal to a preset odor change coefficient reference threshold, the odor change process of the tea oil is classified as an odor characteristic abnormal period;

[0084] If the odor change coefficient is less than a preset odor change coefficient reference threshold, the odor change process of the tea oil is classified as an odor characteristic stable period;

[0085] The stable period of odor characteristics refers to the period during which the odor components of tea oil remain relatively constant during storage, without obvious changes or fluctuations, and the correlation between odor components and time series fluctuations are relatively stable, meeting normal quality standards; the abnormal period of odor characteristics refers to the period during which the odor components of tea oil undergo complex superposition and interference during the deterioration process, resulting in obvious changes in the odor.

[0086] During the odor characteristic stabilization period, the odor of tea oil changes steadily and meets the normal quality standards. Odor monitoring continues according to the sensor response time preset by the electronic nose to ensure that the quality of tea oil remains within the normal range;

[0087] During the odor characteristic stabilization period, the odor of tea oil changes steadily and meets normal quality standards. Odor monitoring continues according to the sensor response time preset by the electronic nose to ensure that the quality of tea oil remains within the normal range. The core purpose of this practice is to prevent any quality problems from further deteriorating by timely detecting possible minor fluctuations or initial odor changes through continuous monitoring of odor changes in tea oil. Since the odor changes of tea oil are usually gradual, through regular inspections and real-time monitoring of odor changes, the system can immediately capture early abnormal signals of tea oil quality and take appropriate preventive measures. This monitoring method not only helps maintain the stable quality of tea oil, but also reduces potential quality risks and production losses, ensuring the safety and satisfaction of consumers when using tea oil.

[0088] In the abnormal period of odor characteristics, based on the odor change characteristics evaluated by the machine learning model, the sensor's response time is dynamically extended to ensure that the changes in tea oil odor are fully analyzed and the subtle changes in odor changes are accurately captured;

[0089] Based on the odor change characteristics evaluated by the machine learning model, the specific steps of dynamically extending the sensor's response time to ensure full analysis of tea oil odor changes and accurately capture subtle changes in odor changes are as follows:

[0090] Once the odor abnormality period is identified, the odor change coefficient The value of is used to dynamically adjust the sensor's response time. The calculation expression is as follows: , where is the adjusted sensor response time, is the preset sensor response time, is the preset reference threshold value of the odor variation coefficient, is the adjustment coefficient;

[0091] Adjustment coefficient It plays a key role in the formula for dynamically adjusting the sensor response time, which determines the odor change coefficient Reference threshold value of odor variation coefficient The difference between the two factors affects the sensitivity of the sensor response time. The size of directly affects the speed and degree of the system's response to odor changes. If the value is larger, the system will be more sensitive to odor changes, that is, the difference between the odor change coefficient and the reference threshold will cause the sensor response time to be significantly prolonged, ensuring that small changes can be accurately captured at the early stage of odor change; on the contrary, if If the value is smaller, the system will react more slowly to the odor change, and the adjustment range of the sensor response time will be smaller. This adjustment coefficient allows the system to flexibly adjust the response time according to the complexity of the actual odor change, thereby optimizing the accuracy and sensitivity of tea oil quality monitoring.

[0092] During the odor abnormality period, the extended response time depends not only on the odor variation coefficient, but also on the variation characteristics between odor components. By analyzing the variation of volatile compounds in the odor time series, the volatile compound variation index is introduced to quantify the fluctuation degree of the odor time series. The calculation expression is as follows: , where is the volatile compound variation index, It's time point The measured value of the odor components at the moment, is the time step index, i.e. from the time point Starting from the moment, it gradually extends to the future time point. is the total number of time steps, representing the time from point The number of time steps extending backward from the moment indicates the time range to be considered when calculating the volatile compound variation index;

[0093] Based on the adjusted sensor response time and Volatile Compound Variation Index , generate the final response time adjustment, the calculation expression is as follows: , where is the response time adjustment, Is the weighted coefficient, which is used to balance the odor variation coefficient and Volatile Compound Variation Index The influence degree on the sensor response time adjustment,the weighting coefficient not only plays a role in the formula for adjusting the sensor response time,,but also indirectly reflects the fluctuation degree of volatile compounds.,By adjusting the value of the weighting coefficient, the influence of the correlation and fluctuation degree between different,odor components on the sensor response time adjustment,can be indirectly controlled;

[0094] Adjust the response time With preset sensor response time Add together to get the final sensor response time, the calculation expression is as follows: , where is the final adjusted response time, which indicates the final sensor response time after adjustment during the abnormal odor period.

[0095] During the abnormal odor characteristic period, the odor of tea oil changes significantly, usually due to quality problems such as oxidation, rancidity, and mold. At this stage, the volatile organic compounds (VOCs) in tea oil may undergo complex superposition, and new odor components (such as oxides, sourness, etc.) may appear, making the odor change more complex and subtle. The sensor response time preset by traditional electronic nose technology is usually based on historical experience or fixed standards, which may not be enough to capture these small but critical odor changes. In order to more accurately identify these abnormal changes, the sensor response time must be dynamically adjusted.

[0096] By evaluating the characteristics of odor changes through machine learning models, the system can identify the trends and fluctuations of odor changes, and thus determine whether the odor characteristics have entered an abnormal period. Once abnormal characteristics of odor changes are detected, the system will dynamically extend the response time of the sensor based on the evaluation results of the machine learning model. This extension not only provides a longer data collection window, but also allows the sensor to have enough time to capture subtle odor fluctuations, especially those changes that only appear when odor components are superimposed.

[0097] Through this dynamic adjustment, the system can optimize the accuracy of capturing odor changes in real time, ensuring that the early signs of quality problems can be accurately identified during abnormal periods when the odor components are complex and intertwined, so as to provide timely warnings and interventions. This process not only improves the system's sensitivity to subtle odor changes, but also ensures the reliability of tea oil quality, avoiding quality problems and potential health risks that may be caused by inaccurate odor detection.

[0098] Traditional electronic nose technology relies on the sensor response time preset by historical data, but this method may not be able to respond to the complex superposition and interference of odor components in a timely manner during the change of tea oil odor, especially in the early stage of tea oil deterioration. When tea oil is oxidized, rancid or has other quality problems, the new odor components overlap with the original odor components, which may cause the correlation between the odor components to change, making it difficult to capture subtle signs of odor changes. The present invention can accurately judge the early stage of odor changes and respond in time, accurately capture the abnormal period of odor changes, and effectively avoid tea oil quality problems from being ignored by dynamically calculating the odor change coefficient, odor hierarchy correlation index, and odor time series fluctuation index evaluated based on a machine learning model. By dynamically adjusting the response time of the electronic nose sensor, it is possible to better analyze complex odor components during the abnormal period of odor characteristics, ensure detection accuracy, and avoid quality problems caused by response delays from entering the market.

[0099] The present invention can maintain stable monitoring during the stable period of the tea oil's odor characteristics by real-time monitoring and intelligent evaluation of tea oil's odor changes, ensuring that the tea oil quality meets normal standards, and more accurately analyze the odor changes by dynamically adjusting the sensor response time during the abnormal period. Through this big data-driven monitoring system, companies can obtain accurate feedback on tea oil quality in real time, greatly improving the company's quality control efficiency. This intelligent monitoring can not only reduce manual operation errors, but also detect potential quality problems in a timely manner, effectively avoiding product recalls, complaints, and food safety incidents, thereby greatly improving the company's product quality management level, reducing risks, and strengthening the company's brand reputation to ensure the health, safety, and interests of consumers.

[0100] The above description is only by way of illustration of certain exemplary embodiments of the present invention. It is undoubted that those skilled in the art can modify the described embodiments in various ways without departing from the spirit and scope of the present invention. Therefore, the above drawings and descriptions are illustrative in nature and should not be construed as limiting the scope of protection of the claims of the present invention.

Claims

1. A big data-driven organic tea oil quality monitoring and early warning method, characterized in that: The following steps are involved: The odor components of tea oil are collected in real time through electronic nose technology, and the collected odor data are input into the data management system; The collected tea oil odor data is used to construct an analysis set, and key features reflecting odor changes are extracted from it; The extracted key features are analyzed and processed within the detection window, and the analyzed feature data are input into a pre-trained machine learning model, and the machine learning model is used to predict the quality change of tea oil; Based on the prediction results of the machine learning model, the odor change process of tea oil is divided into a stable odor characteristic period and an abnormal odor characteristic period; During the odor characteristic stabilization period, the odor of tea oil changes steadily and meets the normal quality standards. Odor monitoring continues according to the sensor response time preset by the electronic nose to ensure that the quality of tea oil remains within the normal range; In the abnormal period of odor characteristics, based on the odor change characteristics evaluated by the machine learning model, the sensor's response time is dynamically extended to ensure that the changes in tea oil odor are fully analyzed and the subtle changes in odor changes are accurately captured; Based on the odor change characteristics evaluated by the machine learning model, the specific steps of dynamically extending the sensor's response time to ensure full analysis of tea oil odor changes and accurately capture subtle changes in odor changes are as follows: Once the odor abnormality period is identified, the odor change coefficient The value of is used to dynamically adjust the sensor's response time. The calculation expression is as follows: , where is the adjusted sensor response time, is the preset sensor response time, is the preset reference threshold value of the odor variation coefficient, is the adjustment coefficient; During the odor abnormality period, the extended response time depends not only on the odor variation coefficient, but also on the variation characteristics between odor components. By analyzing the variation of volatile compounds in the odor time series, the volatile compound variation index is introduced to quantify the fluctuation degree of the odor time series. The calculation expression is as follows: , where is the volatile compound variation index, It's time point The measured value of the odor components at the moment, is the time step index, is the total number of time steps; Based on the adjusted sensor response time and Volatile Compound Variation Index , generate the final response time adjustment, the calculation expression is as follows: , where is the response time adjustment, Is the weighted coefficient, which is used to balance the odor variation coefficient and Volatile Compound Variation Index The degree of influence on the adjustment of sensor response time; Adjust the response time by With preset sensor response time Add together to get the final sensor response time, the calculation expression is as follows: , where is the final adjusted response time, which indicates the final sensor response time after adjustment during the abnormal odor period.

2. The big data driven organic tea oil quality monitoring and early warning method according to claim 1 is characterized by: The key features reflecting the changes in odor are extracted from the analysis set. The extracted features include the correlation between different odor components and the time series fluctuations of odor changes. Under the detection window, the correlation between different odor components and the time series fluctuations of odor changes are analyzed and processed to generate the odor hierarchical correlation index and the odor time series fluctuation index respectively. The odor hierarchical correlation index quantifies the degree of correlation between different odor components, the interaction between various volatile organic compounds and their concentration changes during the deterioration of tea oil; the odor time series fluctuation index quantifies the fluctuation of tea oil odor changes over time. During the deterioration of tea oil, the changes in odor components often manifest as irregular and nonlinear fluctuations.

3. The big data driven organic tea oil quality monitoring and early warning method according to claim 2 is characterized in that: After obtaining the odor hierarchy association index and odor time series fluctuation index generated after analyzing and processing the extracted key features, the odor hierarchy association index and odor time series fluctuation index are input into the pre-learned machine learning model, and the odor variation coefficient is generated by the machine learning model. The quality change of tea oil is predicted by the odor variation coefficient.

4. The big data driven organic tea oil quality monitoring and early warning method according to claim 3 is characterized by: The odor change coefficient generated after analyzing the extracted key features is compared with the preset odor change coefficient reference threshold, and the odor change process of tea oil is divided. The division steps are as follows: If the odor change coefficient is greater than or equal to a preset odor change coefficient reference threshold, the odor change process of the tea oil is classified as an odor characteristic abnormal period; If the odor change coefficient is less than a preset odor change coefficient reference threshold, the odor change process of the tea oil is divided into an odor characteristic stable period.

5. The big data driven organic tea oil quality monitoring and early warning method according to claim 2 is characterized in that: Under the detection window, the correlation between different odor components is analyzed and processed, and the specific steps of generating the odor hierarchical correlation index are as follows: Under the detection window, the odor component data of tea oil is collected using electronic nose technology, and the concentration matrix of tea oil odor components is generated based on the acquired data. , ,in, It's time point Moment, The concentration of the odor components, is the total amount of tea oil odor components; By calculating the correlation between odor components, the interaction relationship between different odor components is revealed. The calculation expression is as follows: , where is the odor level association value, indicating the time point Moment, Odor components and The correlation between the odor components, It's time point Moment, The concentration of the odor components, It is Odor components and Odor components at time points The relevance of time, It is Odor components at time points The individual probability distributions at time instants, It is Odor components at time points The individual probability distributions at time instants, It is Odor components and Odor components at time points The joint probability distribution of time; Based on the odor level association value The hierarchical correlation of each odor component in tea oil was evaluated and the odor hierarchical correlation index was generated. The calculation expression is as follows: , where is the odor hierarchy association index, It is Odor components and The weight of the odor components, is the Euclidean distance between odor components, is the attenuation factor.

6. The big data driven organic tea oil quality monitoring and early warning method according to claim 2, characterized in that: Under the detection window, the time series fluctuation of odor changes is analyzed and processed, and the specific steps of generating the odor time series fluctuation index are as follows: In the detection window, the odor component data collected by the electronic nose technology exists in the form of a time series, where each data point represents the odor concentration at each time point. To ensure the accuracy of subsequent analysis, the data needs to be smoothed, denoised, and time-aligned; Calculate the rate of change of odor components and capture the fluctuation characteristics of odor. The calculation expression is as follows: , where is the rate of change of the odor components, It's time point The odor concentration value at the moment, It's the previous time point The odor concentration value at the moment, is the unit time span; In order to evaluate the long-term fluctuation trend of odor changes, the regularity of the time series is calculated and the regularity is quantified by calculating the fractal dimension of the time series. The expression is as follows: , where is the regularity index, is the odor concentration value, i.e., the time point The odor concentration value at the moment, Refers to A point in time, is the total number of time points, is an index that controls local variation; Rate of change based on odor components and regularity index Generate the odor time series volatility index, the calculation expression is as follows: , where is the odor time series volatility index, and Both are adjustment parameters, which respectively control the change rate of odor components and the influence weight of regularity index in the odor time series fluctuation index. It is a point in time.

Images