A paeony cut flower preservation early warning method

Abstract

The present application relates to cut flower preservation and intelligent monitoring technical field, disclose a kind of peony cut flower preservation early warning method, including the construction offline dataset, obtain the time series data of volatile organic compound in multiple scenes;Establish adaptive multi-branch baseline model library containing standard decay trajectory and standard metabolic network correlation matrix evolution model;Collect initial data to build initial metabolic pathway state vector, calculate the Euclidean distance matching best baseline branch;Continuously collect real-time data, respectively calculate the state drift index relative to the best branch and the structural drift index based on correlation matrix;Weighted fusion is obtained to the comprehensive deterioration index, compared with multi-level early warning threshold output early warning signal and control instruction.The present application realizes the accurate monitoring and deterioration attribution of cut flower whole life cycle by state and structure double drift monitoring and adaptive baseline matching, solves the problem of early warning lag and poor environmental adaptability.

Description

Technical Field

[0001] This invention relates to the field of cut flower preservation and intelligent monitoring technology, specifically a method for early warning of peony cut flower preservation. Background Technology

[0002] Peony cut flowers, as a high-value cut flower variety, have vigorous post-harvest respiration and metabolism, making them sensitive to environmental fluctuations. They are prone to quality deterioration during post-harvest distribution, such as wilting, petal browning, and even rotting. To extend the vase life and ornamental value of cut flowers, real-time quality monitoring and early warning within the supply chain are crucial. Current monitoring methods mainly rely on passive recording of temperature and humidity data in the cold chain environment and regular random inspections of the cut flower's appearance. With the development of gas sensor technology, non-destructive monitoring technology based on volatile organic compound (VOC) detection is gradually being applied to the field of agricultural product preservation, assessing the physiological state of products by monitoring specific odor components.

[0003] However, existing monitoring technologies based on volatile organic compounds (VOCs) typically use the absolute concentration threshold of a single component as the basis for judgment. In the early stages of cut flower senescence or disease, the concentration fluctuations of VOCs are small and have not yet reached the threshold level to trigger an alarm. However, at this time, the synergistic relationship of the metabolic network within the cut flower has already changed. This method of judgment, which relies on a single concentration threshold, cannot identify structural abnormalities within the metabolic network, causing the warning signal to lag behind the actual irreversible physiological damage that has occurred in the cut flower, thus failing to achieve a truly meaningful early warning.

[0004] Furthermore, existing monitoring models are mostly based on fixed baseline standards under specific experimental conditions. However, in the actual cut flower supply chain, cut flowers undergo various complex environmental scenarios, including field harvesting, cold chain transportation, and temporary storage in warehouses, and the initial physiological quality of different batches of cut flowers varies. Using a uniform fixed baseline model to evaluate cut flowers at different logistics stages or with different initial states is prone to misjudgment of monitoring data due to differences in environmental background or individual sample differences, and lacks the ability to adaptively match the complex and ever-changing scenarios of the supply chain.

[0005] Meanwhile, existing monitoring systems typically only report the decline in cut flower quality, failing to differentiate the specific causes of this decline. Physical environmental stresses (such as vibration and temperature fluctuations), diseases caused by pathogen infection, and natural excessive aging induced by endogenous hormones all exhibit similar trends in single monitoring indicators. The lack of causal diagnostic capabilities for spoilage makes it impossible to determine the specific source of risk and hinders the implementation of targeted, differentiated control measures such as temperature regulation, sterilization, or ethylene inhibition, thus limiting the precision and effectiveness of preservation management. Summary of the Invention

[0006] To address the shortcomings of existing technologies, this invention provides a method for early warning of peony cut flower preservation. This method solves the problems of existing peony cut flower preservation monitoring technologies relying on single indicator thresholds or human sensory judgment, which result in delayed early warnings, inability to adapt to different logistics scenarios and environments, and difficulty in distinguishing between physiological aging and pathological infection.

[0007] To achieve the above objectives, the present invention provides the following technical solution: A method for early warning of freshness preservation of cut peony flowers includes the following steps: S1. Construct an offline dataset: Select peony cut flower samples and place them in multiple preset standard environmental scenarios. Use a volatile organic compound (VOC) collection device to collect full life cycle data in each scenario at predetermined time intervals to obtain VOC time series data for multiple scenarios. S2. Establish a baseline model library: Perform feature mapping and statistical analysis on the time series data of volatile organic compounds in the multiple scenarios, calculate the standard decay trajectory and standard metabolic network correlation matrix evolution model under each scenario, and construct an adaptive multi-branch baseline model library containing multiple baseline branches. S3. Obtain the initial state: At the start of online monitoring, collect the volatile organic compound data of the peony cut flowers to be monitored, and construct the initial metabolic pathway state vector of the batch of cut flowers based on the preset metabolic pathway classification rules. S4. Matching the baseline branch: Calculate the Euclidean distance between the initial metabolic pathway state vector and the initial node of each reference baseline branch in the adaptive multi-branch baseline model library, and select the branch with the smallest distance as the best matching baseline branch for this batch of cut flowers. S5. Calculate real-time drift: During subsequent monitoring, real-time data is continuously collected, and the state drift index of the real-time metabolic pathway state vector relative to the best-matching baseline branch is calculated, and the structural drift index of the real-time correlation matrix relative to the best-matching baseline branch is calculated. S6. Generate early warning decision: Perform weighted fusion calculation on the state drift index and the structural drift index to obtain a comprehensive deterioration index, compare the comprehensive deterioration index with the multi-level early warning threshold, and output the corresponding preservation early warning signal and control command.

[0008] Preferably, in step S1, the preset standard environment scenarios cover typical working conditions in the supply chain, specifically including: an ideal control group scenario simulating optimal warehousing conditions, serving as a benchmark for physiological aging; a transportation stress group scenario simulating temperature fluctuations and mechanical vibrations during cold chain transportation, corresponding to the volatile characteristics under physical damage; a high-humidity susceptibility group scenario simulating fungal diseases, corresponding to the metabolic signals in the early stages of pathogen infection; and an ethylene-induced group scenario simulating the accelerated aging process caused by ethylene pollution, corresponding to abnormal aging caused by exogenous hormone pollution.

[0009] Preferably, in step S2, the method for feature mapping of the multi-scenario volatile organic compound time-series data is as follows: the identified volatile organic compounds are mapped to preset metabolic function categories to obtain metabolic function category mapping relationships; the metabolic function categories are divided based on plant physiological mechanisms and include at least aroma synthesis metabolic pathways characterizing cut flower variety characteristics and openness, stress response metabolic pathways characterizing the degree of physical damage or stress to cut flowers, and pathogen interaction metabolic pathways characterizing whether cut flowers are infected by pathogens; based on the mapping relationships, a metabolic pathway state vector characterizing the metabolic intensity at a specific time is generated through aggregation calculation.

[0010] Preferably, in step S2, the process of calculating the standard decay trajectory under each scenario is as follows: statistically averaging the repeated sample data to reduce the influence of individual sample differences, calculating the standard state points of the scenario at each discrete time point, and connecting them in time sequence to form a continuous trajectory reflecting the physiological evolution law of cut flowers under that scenario.

[0011] Preferably, in step S2, the method for calculating the evolution model of the standard metabolic network correlation matrix under each scenario is as follows: set a sliding time window, calculate the Pearson correlation coefficient between different volatile organic compounds for each time point on the standard decay trajectory, and construct a standard correlation matrix; this matrix reflects the synergistic or antagonistic relationship between different metabolites under a specific physiological stage, thereby generating a model describing the evolution of the metabolic network topology over time.

[0012] Preferably, in step S4, the selection logic of the optimal matching baseline branch is as follows: by calculating the Euclidean distance between the initial state of the cut flower to be monitored and the starting point of each scene baseline in the library, the category of the initial physiological state to which the current cut flower belongs is determined, so as to realize the adaptive matching of the monitoring benchmark and avoid misjudgment caused by a single fixed baseline.

[0013] Preferably, in step S5, the calculation of the state drift index simultaneously quantifies the difference in numerical deviation and trend: calculating the state divergence between the real-time metabolic pathway state vector and the best-matching baseline branch at the corresponding standard state point; and the rate divergence between the rate of change vector and the theoretical rate of change vector; and finally, weighted fusion of the two.

[0014] Preferably, in step S5, the calculation of the structural drift index focuses on monitoring the stability of the metabolic network: a sliding window is constructed based on real-time data and a real-time correlation matrix is ​​calculated. By calculating the Frobenius norm distance between this matrix and the corresponding standard matrix in the standard metabolic network correlation matrix evolution model, the degree of change in the synergistic relationship within the metabolic network is quantified. This index can detect metabolic abnormalities caused by disease or stress and provide early warning when the concentration of a single component does not exceed the standard.

[0015] Preferably, in step S6, the process of generating early warning decisions includes: using the comprehensive deterioration index calculation formula to fuse state drift and structure drift information to obtain a comprehensive deterioration index; comparing the index with a first threshold and a second threshold respectively; wherein, the first threshold corresponds to the state where the physiological indicators of cut flowers deviate from the normal range but no irreversible damage has occurred; the second threshold corresponds to the state where cut flowers have obvious lesions or excessive aging.

[0016] Preferably, in step S6, the output control instruction includes an attribution diagnosis mechanism: when an early warning is triggered, the contribution rate of three metabolic pathways—aroma synthesis, stress response, and pathogen interaction—to the overall state drift is calculated; the dominant factor of current spoilage is determined based on the one with the largest contribution rate, or whether a structural imbalance of the metabolic network has occurred is determined based on the proportion of the structural drift index; and a diagnostic status label containing information on physical environmental stress, disease infection risk, or excessive aging is output to provide specific guidance for subsequent preservation interventions.

[0017] This invention provides a method for early warning of freshness preservation of cut peony flowers. It has the following beneficial effects: 1. This invention overcomes the lag in traditional single concentration threshold judgment by introducing a dual monitoring mechanism of structural drift index and state drift index. By calculating the Frobenius norm distance between the real-time correlation matrix and the standard metabolic network model, it can quantify the small changes in the cooperative relationship within the metabolic network. Structural abnormalities in the metabolic network can be detected in the early stage before the concentration of volatile organic compounds changes drastically, thereby achieving early warning of the risk of cut flower spoilage.

[0018] 2. This invention constructs an adaptive baseline model library that includes multiple scenarios such as ideal control, transportation stress, high humidity susceptibility to disease, and ethylene induction. Based on the initial metabolic pathway state vector, it performs optimal baseline branch matching, which solves the problem that fixed baseline models cannot adapt to the differences in complex supply chain environments. It can automatically match the most suitable standard decay trajectory according to the initial state of the cut flowers to be monitored, eliminate misjudgments caused by differences in environmental background or initial quality, and improve the accuracy of early warning in different logistics and warehousing scenarios.

[0019] 3. This invention establishes a mapping relationship between metabolic function categories and calculates the deviation contribution rate of each metabolic pathway, thereby achieving attribution diagnosis of the causes of cut flower deterioration. By quantifying the contribution ratio of three types of metabolic pathways—aroma synthesis, stress response, and pathogen interaction—to the overall state drift, it can clearly distinguish the quality decline caused by physical environmental stress, fungal disease infection, or excessive physiological aging, and output corresponding diagnostic status labels. This provides a decision-making basis for subsequent targeted preservation control measures such as temperature regulation, sterilization treatment, or ethylene inhibition. Attached Figure Description

[0020] Figure 1 This is a schematic diagram of the overall architecture of a peony cut flower preservation early warning system according to an embodiment of the present invention; Figure 2 This is a flowchart of a peony cut flower preservation early warning method according to an embodiment of the present invention; Figure 3 This is a flowchart illustrating the early warning decision-making and attribution diagnosis in one embodiment of the present invention.

[0021] Among them, 10 is a volatile organic compound (VOC) collection device; 20 is a data transmission network; 30 is a data processing terminal; 301 is an offline model building module; 302 is a real-time data receiving module; 303 is a baseline adaptive matching module; 304 is a drift calculation module; and 305 is an early warning decision module. Detailed Implementation

[0022] The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0023] See attached document Figure 1 This invention provides a method for early warning of peony cut flower preservation, which is implemented based on a specific hardware system architecture. The system mainly includes a volatile organic compound (VOC) collection device 10, a data transmission network 20, and a data processing terminal 30.

[0024] A volatile organic compound (VOC) collection device 10 is deployed in the actual storage or transportation environment of peony cut flowers to acquire gas data released by the cut flowers. In this embodiment, the VOC collection device 10 is a headspace gas chromatography-ion mobility spectrometry (HS-GC-IMS) system. This device can separate and detect the mixture of VOCs released by the cut flowers, outputting three-dimensional spectral data including retention time, drift time, and signal intensity. A data transmission network 20 connects the VOC collection device 10 to the data processing terminal 30, and is responsible for transmitting the collected raw spectral data or preprocessed data to the data processing terminal 30 in real time.

[0025] The data processing terminal 30 is a computing platform that executes the core algorithm of this invention. It is internally configured with a memory and a processor. The memory stores a computer program, and the processor executes the computer program to implement the early warning method. Logically, the data processing terminal 30 includes an offline model construction module 301, a real-time data receiving module 302, a baseline adaptive matching module 303, a drift calculation module 304, and an early warning decision module 305.

[0026] Specifically, the offline model building module 301 processes pre-collected multi-scenario experimental data, builds and stores an adaptive multi-branch baseline model library, and is configured to execute subsequent steps S1 and S2. The real-time data receiving module 302 receives real-time monitoring data transmitted by the volatile organic compound (VOC) collection device 10 and converts it into a metabolic pathway state vector, and is configured to execute the data acquisition portion of subsequent steps S3 and S5. The baseline adaptive matching module 303 retrieves the most suitable standard baseline from the adaptive multi-branch baseline model library based on the state vector at the initial monitoring time, and is configured to execute subsequent step S4. The drift calculation module 304 calculates the state drift index and structural drift index of the real-time data relative to the selected standard baseline, and is configured to execute the core calculation portion of subsequent step S5. The early warning decision module 305 generates graded early warning signals and control instructions based on the comparison between the comprehensive calculation results and preset thresholds, and is configured to execute subsequent step S6. The above modules work together to realize the preservation early warning process of the present invention.

[0027] See attached document Figure 2 The method steps shown in this flowchart are based on the above. Figure 1 The system architecture shown, particularly the specific execution through the data processing terminal 30 calling its internal functional modules, provides a method for early warning of peony cut flower preservation, comprising the following steps: S1, Construct an offline dataset: Select peony cut flower samples and place them in multiple preset standard environmental scenarios. Use a volatile organic compound (VOC) collection device to collect full life cycle data in each scenario at predetermined time intervals to obtain VOC time series data for multiple scenarios. S2, Establish a baseline model library: Perform feature mapping and statistical analysis on time series data of volatile organic compounds in multiple scenarios, calculate the standard decay trajectory and standard correlation matrix evolution model under each scenario, and construct an adaptive multi-branch baseline model library containing multiple baseline branches. S3, Obtain the initial state: At the start of online monitoring, collect the volatile organic compound data of the peony cut flowers to be monitored, and construct the initial metabolic pathway state vector of the batch of cut flowers based on the preset metabolic pathway classification rules. S4, Matching baseline branch: Calculate the Euclidean distance between the initial metabolic pathway state vector and the initial node of each reference baseline branch (i.e., the scene corresponding to the normal or acceptable physiological state) in the adaptive multi-branch baseline model library, and select the branch with the smallest distance as the best matching baseline branch for this batch of cut flowers. S5, Calculate real-time drift: During subsequent monitoring, real-time data is continuously collected, and the state divergence and rate divergence of the real-time metabolic pathway state vector relative to the best-matching baseline branch are calculated to obtain the state drift index, and the structural drift index of the real-time correlation feature relative to the best-matching baseline branch is calculated. S6, Generate Early Warning Decision: The state drift index and the structural drift index are weighted and fused to obtain a comprehensive deterioration index. This index is then compared with multi-level early warning thresholds, and corresponding preservation early warning signals and control instructions are output.

[0028] To enable those skilled in the art to better understand the technical solution of the present invention, the implementation principles and processes of the above steps will be explained in detail below with reference to specific technical details.

[0029] See attached document Figure 2 In step S1, the process of constructing the offline dataset is specifically implemented by establishing a multi-dimensional environmental scenario library to simulate various environmental conditions encountered by peony cut flowers during actual circulation. This step aims to obtain the evolution patterns of volatile organic compounds under different physiological and pathological states, providing comprehensive data support for the subsequent construction of baseline models.

[0030] Step S1 specifically includes the following sub-steps: S11, Establish a multi-dimensional environment scenario library. Pre-set scenarios in a controlled laboratory environment. A standardized experimental scenario covering the entire process from ideal preservation to extreme deterioration. Specifically, it includes: Scenario 1 is set as the ideal control group, simulating the best storage conditions, with the ambient temperature controlled in the constant temperature range of 3℃-5℃, the relative humidity controlled in the range of 85%-90%, and the ethylene concentration in the ambient gas below 0.01ppm. Scenario 2 is set as the transportation stress group, simulating environmental fluctuations during cold chain transportation. The temperature is set to fluctuate periodically in a sinusoidal pattern between 4℃ and 10℃, and a vibration table is used to apply continuous mechanical vibration at a frequency of 5-10Hz to simulate mechanical damage and respiratory heat accumulation caused by road bumps. Scenario 3 is set as a high-humidity susceptible group, with the relative humidity maintained above 95% to simulate an extreme high-humidity environment that easily induces fungal diseases such as gray mold; Scenario 4 is set as the ethylene-induced group, in which exogenous ethylene gas with a concentration of 0.1-1.0 ppm is introduced into the environment to simulate the accelerated aging process caused by ethylene pollution.

[0031] S12, Perform time-series data acquisition. Place peony cut flower samples with consistent physiological states in the above scenarios, from the time of collection... Start by setting the sampling time interval. (For example, every 4 to 6 hours), repeated samples from each scenario are continuously analyzed using headspace gas chromatography-ion mobility spectrometry (HS-GC-IMS). For headspace sampling parameters, those skilled in the art can set the incubation temperature (e.g., 40℃-60℃) and incubation time (e.g., 10-20 minutes) according to the volatility characteristics of peony cut flowers to ensure the effective release of volatile components. The collection process continues until the samples exhibit obvious wilting, browning, or rotting, thus losing their ornamental value, thereby obtaining time-series data on volatile organic compounds throughout their entire life cycle for each scenario.

[0032] In step S2, the offline model construction module 301 receives the raw data collected in step S1 and performs feature extraction and vectorization mapping to construct metabolic pathway state vectors. The raw HS-GC-IMS data contains thousands of three-dimensional pixels corresponding to their retention and drift times, which are computationally intensive to process directly and lack biological interpretability. Therefore, it is necessary to transform them into metabolic pathway state vectors with clear physiological significance.

[0033] Step S2 specifically includes the following sub-steps: S21, Qualitative Analysis and Classification of Volatile Organic Compounds. The offline model building module 301 utilizes built-in volatile organic compound databases (such as the NIST and IMS databases) to perform qualitative analysis on the characteristic peaks in the acquired spectra, identifying... Specific volatile organic compounds. Subsequently, the offline model construction module 301, based on plant physiological metabolic pathways, identifies the... Volatile organic compounds are mapped to preset metabolic function categories.

[0034] The metabolic function categories include at least: aroma synthesis metabolic pathways, corresponding to terpenes (such as linalool and geraniol), esters and aromatic compounds, characterizing the varietal characteristics and openness of cut flowers; The stress response metabolic pathways correspond to membrane lipid peroxidation products (such as hexanal and trans-2-hexenal) and anaerobic respiration products (such as ethanol), characterizing the degree of physical damage, chilling injury or water stress suffered by cut flowers. The pathogen-pathogen interaction metabolic pathway corresponds to fungal or bacterial specific metabolites (such as 3-octanone, 1-octen-3-ol and other characteristic volatiles of fungi), which characterize whether cut flowers are infected by pathogens.

[0035] S22, Constructing Metabolic Pathway State Vectors. Based on the above classification results, the concentration values ​​(or peak volume, peak intensity) of volatile organic compounds of the same category are vectorized and assembled according to a predetermined sorting rule to construct a multidimensional state vector. The time step is calculated using the definition formula of the metabolic pathway state vector. The metabolic pathway state vector S(t) is defined as follows: ; In the formula, Indicates at time The metabolic pathway state vector, which is a multidimensional column vector or row vector containing the concentration data of all selected volatile organic compounds, is used to mathematically represent the comprehensive physiological state of the cut flower at that moment. The vector construction operator is used to concatenate and combine the internal sub-vectors in a predetermined order. Indicates time The aroma component vector is not a single value, but rather a combination of factors attributed to aroma synthesis and metabolism pathways. Concentration values ​​of volatile organic compounds The sub-vectors formed by sequential arrangement are used to quantify the degree to which the quality of floral fragrance is maintained; Indicates time The stress response component vector, which is composed of components belonging to the stress response metabolic pathway. The subvector is formed by arranging the concentration values ​​of various volatile organic compounds in sequence. When cut flowers are subjected to temperature fluctuations or mechanical damage, the modulus values ​​of each element within this component will increase significantly. Indicates time The pathogen marker component vector, which is attributed to pathogen-interacting metabolic pathways. The subvector is formed by sequentially arranging the concentration values ​​of various volatile organic compounds. In a healthy state, the values ​​of each element within this component tend to be zero, while in the early stages of disease, the value of this component exhibits a non-linear increase. This indicates other metabolic pathway components that can be expanded according to actual testing needs, such as components related to the ethylene synthesis pathway.

[0036] Through the above mapping, high-dimensional, unstructured chromatographic data is transformed into low-dimensional, structured vectors with clear biological semantics. Furthermore, it preserves the independent variation information between different volatile organic compounds, providing a unified mathematical foundation for subsequent baseline modeling and metabolic network correlation analysis for different environmental scenarios. For the same time point... If there are multiple duplicate samples, the metabolic pathway state vector corresponding to each sample is calculated separately to form the sample set data.

[0037] See attached document Figure 2 After completing the construction of the multi-dimensional environment scene library and data vectorization, this embodiment continues to execute the modeling process in step S2 to build an adaptive multi-branch baseline model library.

[0038] Step S2 specifically includes the following sub-steps: S23, calculate the standard decay trajectory. For each preset scenario in the adaptive multi-branch baseline model library (where...) To obtain statistically representative patterns of physiological state evolution under these environmental conditions, statistical averaging was performed on repeated sample data from this scenario. Specifically, for the scenario... Each discrete time point in The central trend at that moment is calculated using the standard state point calculation formula. The standard state point calculation formula is: ; In the formula, Indicates the first The scene at a certain point in time The standard state point, whose physical meaning is the average metabolic state of the cut flower population in this scenario, has the same dimension as the state vector of a single sample. This represents the total number of biological replicates in this scenario, set during implementation. The value (e.g.) To meet the statistical significance requirement; Indicates the sample index from to Perform cumulative calculation; Indicates the first In the scenario, the first A number of duplicate samples at time point The metabolic pathway state vector.

[0039] The offline model building module 301 connects the standard state points at each time point according to the time series sequence. Generate the standard decay trajectory for this scenario. The trajectory data is then stored in the adaptive multi-branch baseline model library.

[0040] S24, construct a standard metabolic network correlation matrix evolution model. To provide a structural reference with temporal dynamics for subsequent online monitoring, the offline model construction module 301 calculates the coordinated changes of different volatile organic compounds over time under standard conditions. For the first... In this scenario, based on the standard decay trajectory obtained in step S23... Set the sliding time window length to match that of online monitoring. For each time point on the trajectory (satisfy Construct a standard correlation matrix with dimensions P×P. Calculate the i-th element in the matrix using the standard sliding window correlation formula. Line 1 The elements of the column. The standard sliding window correlation formula is: ; In the formula, Indicates at a point in time The first in the standard decay trajectory Volatile organic compounds and the first Volatile organic compounds in a time window The Pearson correlation coefficient within the range is used to characterize the dynamic synergistic pattern of two metabolites during a specific environmental evolution process; This indicates the length of the sliding time window, and its value is consistent with the window length set in the online monitoring in the subsequent step S5; Represents the discrete-time variables within the sliding window; Indicates the first Standard decay trajectory for each scenario The corresponding number in the middle Concentration values ​​of various volatile organic compounds; Indicates the first The arithmetic mean concentration of a certain volatile organic compound within the current standard sliding window; and They represent the first The concentration components of each volatile organic compound and their arithmetic mean concentration within the current standard sliding window; This represents the arithmetic square root operation.

[0041] By iterating through all valid time points, the offline model building module 301 generates a standard correlation matrix evolution model. Offline model building module 301 and standard decay trajectory Together they form the baseline branch data of the offline model library.

[0042] In step S3, the initial metabolic state is acquired. This is performed when the batch of peony cut flowers to be monitored enters the monitoring environment. The volatile organic compound (VOC) collection device collects VOC data for this batch, and the data processing terminal generates the initial metabolic pathway state vector of this batch of cut flowers according to the mapping rules in step S22. .

[0043] In step S4, intelligent baseline branch selection based on the initial state is performed. Given the objective differences in the quality of cut flowers entering the warehouse from different batches, the matching degree between the initial state vector and the starting points of each baseline branch in the model library is calculated to determine the most suitable reference standard. The index of the best-matching baseline branch is determined using the optimal baseline selection formula. The formula for selecting the optimal baseline is: ; In the formula, The index number representing the selected best baseline branch; This indicates the minimum value parameter operation, which selects the index that minimizes the objective function value. ; This represents the actual metabolic pathway state vector of the batch to be monitored at the start of monitoring. Represents the first in the adaptive multi-branch baseline model library The standard state points of each baseline branch at the initial time; This represents the Euclidean norm operation, used to quantify the linear distance between two multidimensional state vectors in the feature space.

[0044] The processor calculates the index Locate the corresponding standard decay trajectory from the model library. (Right now ) and standard correlation matrix evolution model (Right now This will serve as the exclusive dynamic benchmark for this monitoring batch.

[0045] See attached document Figure 2 In the online real-time early warning stage, after determining the exclusive dynamic benchmark, this embodiment executes step S5 to perform multi-dimensional drift calculation on the real-time collected data.

[0046] Step S5 is implemented as follows: exist At each subsequent sampling time The real-time data receiving module 302 receives volatile organic compound detection data and generates a real-time pathway state vector. The drift calculation module 304 calls the optimal matching baseline branch locked by the baseline adaptive matching module 303 in step S4 above to obtain the standard state point at the corresponding time. To quantify the absolute deviation of the current physiological state from the ideal evolutionary path, the drift calculation module 304 uses the state divergence calculation formula to calculate the state divergence. The formula for calculating state divergence is: ; In the formula: Indicates time The state divergence, in its physical sense, is the Euclidean distance between the current metabolic composition of the cut flower to be monitored and the ideal preservation state of the batch. The larger the value, the further the physiological quality of the cut flower deviates from the ideal benchmark. Indicates time The real-time path state vector; Indicates based on index The selected standard decay trajectory at time Standard state point; This represents Euclidean norm operations.

[0047] At the same time, the processor calls the previous sampling time cached in memory. The metabolic state data is used to calculate the rate vector of change, which is then compared with the theoretical rate of change of a standard trajectory to capture the kinetic anomalies in the physiological changes of cut flowers. The rate divergence is calculated using the rate divergence calculation formula. The formula for calculating the rate divergence is: ; In the formula, Indicates time The rate divergence is used to quantify the difference in acceleration between the physiological decay rate of cut flowers and the standard model. This represents the incremental vector of the real-time path state vector within the sampling interval; This represents the theoretical increment vector of the standard decay trajectory within the same time period; Indicates the time interval between two consecutive samples; The overall representation is a vector of the rate of change of metabolic state as monitored in real time; The overall representation is the vector of metabolic state change rates expected by the standard model.

[0048] After obtaining the state divergence and velocity divergence, the processor performs a weighted fusion of the two. The time step is calculated using the state drift exponent calculation formula. State drift index The formula for calculating the state drift index is: ; In the formula, Indicates time The state drift index is a dimensionless scalar. The weighting coefficients represent the state divergence and range from 0 to 1. The weighting coefficients representing the rate divergence range from 0 to 1 and satisfy the following conditions: ; The normalization factor representing the state divergence takes values ​​corresponding to the baseline branch during the offline modeling phase. The statistical standard deviation of the state divergence of all samples is used to eliminate the influence of dimensions and achieve standardization. The normalization factor representing the rate divergence takes values ​​corresponding to the baseline branch during the offline modeling phase. The statistical standard deviation of the rate divergence of all samples.

[0049] Through the above calculation logic, the system generates information containing both position and velocity deviations. This index serves as the primary reference for determining the warning level in subsequent steps.

[0050] See attached document Figure 2 After calculating the state drift index Based on this, in step S5 of this embodiment, the structural drift index is further executed. The calculation of this index is used to monitor the stability of the synergistic relationships among components within the metabolic network, in order to identify early anomalies where the concentration of volatile organic compounds has not yet significantly exceeded the standard, but structural changes have already occurred in the metabolic regulatory mechanisms.

[0051] The specific implementation of step S5 regarding the calculation of the structural drift index is as follows: To extract the dynamic correlation features of real-time monitoring data, the drift calculation module 304 constructs a data buffer with a length of [missing information]. A time-sliding window. At the current sampling time. The drift calculation module 304 extracts the drift within this time window (i.e., from time t). At that time Historical metabolic data sequences of the monitored [data]. Volatile organic compounds and the first We calculate the dynamic correlation between volatile organic compounds and their time-series correlations. A real-time correlation matrix is ​​calculated using a real-time sliding window correlation formula. elements in The formula for the correlation of the real-time sliding window is: ; In the formula, Indicates at time , No. Volatile organic compounds and the first Volatile organic compounds in a time window The Pearson correlation coefficient within the matrix constitutes the real-time correlation matrix. The Line number Column elements; This represents the length of the number of samples in the sliding time window, and its value is set according to the sampling frequency (e.g., setting...). to (), to cover a time period sufficient to reflect short-term metabolic trends; Represents the discrete-time variables within the sliding window, traversing from... arrive At each sampling moment; Represents the state vector of the actual metabolic pathway The corresponding number in the middle Concentration values ​​of various volatile organic compounds; Indicates the first The arithmetic mean concentration of the volatile organic compounds within the current sliding window, i.e. ; and They represent the first Volatile organic compounds at any time The concentration component values ​​and the arithmetic mean concentration value within the current sliding window; This represents the arithmetic square root operation.

[0052] Based on the above calculations, the drift calculation module 304 constructs a dimension of Real-time correlation matrix Subsequently, the drift calculation module 304 calls the standard correlation matrix evolution model locked in step S4 to extract the current time step. Corresponding standard correlation matrix To quantify the overall difference between the actual metabolic network structure and the ideal reference network structure, the structural drift index was calculated using the structural drift index calculation formula. The formula for calculating the structural drift index is: ; In the formula, Indicates time The structural drift index quantifies the degree of deviation of the internal connection patterns of the real-time metabolic network of cut flowers from the standard pattern. This represents the real-time correlation matrix calculated based on real-time sliding window data. This represents the standard correlation matrix obtained based on the best-matching baseline branch; The Frobenius norm operation of a matrix is ​​used to calculate the combined Euclidean distance between two matrices over all elements. This indicates the total number of types of volatile organic compounds being monitored; This indicates a double summation operation, covering the matrix from row 1, column 1 to column 2. Line number All elements of the column; Represents the first [item] in the real-time correlation matrix. Line number The elements of the column reflect the current dynamic coupling relationships; Represents the first in the standard correlation matrix Line number The elements of the column, derived from the statistical regularities observed during the offline modeling phase, reflect the standard coupling relationships expected in this physiological stage. The calculated... Similar to the aforementioned state drift index Together, they constitute the basis for multi-dimensional early warning judgment. When When the value increases significantly, even At a low level, the processor also determined that the cut flower was in the early stage of metabolic disorder, indicating that its internal regulatory mechanism had deviated from the steady state.

[0053] See attached document Figure 3The state drift index is calculated in step S5. and structural drift index Next, the early warning decision module 305 executes step S6, namely, comprehensive early warning decision and attribution diagnosis. The early warning decision module 305 integrates multi-dimensional monitoring data into a single decision indicator, and when the judgment result triggers the early warning threshold, it analyzes the dominant physiological cause of the abnormality.

[0054] Step S6 specifically includes the following sub-steps: S61, calculate the Comprehensive Deterioration Index (CDI). The early warning decision module 305 employs a weighted fusion strategy, combining the indicators representing the degree of numerical deviation... With characterizing the degree of network structural anomalies The data is merged to generate a comprehensive index for final classification. The comprehensive deterioration index calculation formula is used to calculate the time step. of The formula for calculating the comprehensive deterioration index is: ; In the formula, Indicates time The comprehensive deterioration index indicates that the higher the value, the greater the overall quality risk of the cut flowers. This represents the weighting coefficient of the early warning strategy, with a value ranging from 0.1 to 0.9 (e.g., set to 0.6). This coefficient is stored in the system configuration parameters. If the monitoring scenario focuses on acute damage caused by drastic environmental changes, then the coefficient should be increased. Value; if the focus is on monitoring latent deterioration caused by disease infection, then adjust the value lower. value; This represents the state drift index calculated in step S5; This represents the structural drift index calculated in step S5; The structural drift reference threshold is specifically defined as the statistical average or lower limit of the structural drift index of all samples marked as "critical deterioration" during the offline modeling phase, which is used as the normalization denominator to unify the dimensions.

[0055] S62, Execute the graded early warning determination. The early warning decision module 305 calls a set of preset graded thresholds from the memory. , calculate The current warning level is determined by comparing it with a threshold: when When the system is deemed safe, it maintains its regular monitoring frequency. when When the system determines the level to be a yellow alert, it automatically increases the sampling frequency (e.g., by increasing the sampling frequency to a higher level). Reduce by half), and mark the batch as a key focus; when When the emergency level is determined to be red, the system triggers an audible and visual alarm signal and sends a handling instruction to the management terminal.

[0056] S63, Perform graded attribution diagnosis. When the system determines the result to be a yellow warning level or a red emergency level, the warning decision module 305 calculates the deviation contribution of each metabolic pathway to locate the specific cause of the deterioration. Based on the metabolic function categories defined in step S21 (including aroma synthesis, stress response, and pathogen interaction), the warning decision module 305 identifies the real-time metabolic pathway state vector. Belongs to the first A collection of components of a metabolic pathway The formula for calculating the contribution rate of pathway deviation is used to calculate the first Contribution rate of metabolic pathways The formula for calculating the contribution rate of pathway deviation is: ; In the formula, Indicates at time , No. The percentage contribution of metabolic pathways (e.g., stress response pathways) to overall state drift; Indicates belonging to the first A set of volatile organic compound indexes for metabolic pathways; Represents a set Sum all volatile organic compounds in the solution; The first element in the real-time metabolic pathway state vector represents the... Concentration values ​​of volatile organic compounds; The standard state point vector representing the best-matching baseline branch The corresponding number in the middle Standard concentrations of various volatile organic compounds; This indicates the total number of volatile organic compounds monitored; This represents the summation operation, where For the traversal index of volatile organic compounds, from Traverse to total This formula represents the expression for all monitored volatile organic compounds at time [time value missing]. The absolute deviations are summed up and used as the normalized benchmark for calculating the contribution rate.

[0057] The early warning decision module 305 calculates the contribution rate of each pathway. The system calculates the algebraic deviation direction of the real-time concentration of volatile organic compounds (VOCs) from the standard concentration in each pathway, outputs corresponding diagnostic status labels, and sends warning signals and control commands through the output interface of the data processing terminal 30: if the contribution rate of the stress response pathway... If the contribution rate of the pathogen interaction pathway is the highest among all pathways, a diagnostic label of "physical environmental stress" will be output, suggesting an inspection of cold chain equipment or packaging stability; if the contribution rate of the pathogen interaction pathway is the highest... The largest, or structural drift index, among all pathways. right If the weighted contribution of the aroma synthesis pathway exceeds a preset proportion, a diagnostic label of "risk of disease infection" will be output, indicating that the batch of cut flowers should be isolated and pathologically tested; if the contribution rate of the aroma synthesis pathway... It is the largest among all pathways, and the sum of the algebraic deviations of volatile organic compound concentrations within this pathway. If the value is negative, a diagnostic label of "over-aging" will be output, indicating that the product should be prioritized for sale or downgraded.

Claims

1. A method for early warning of freshness preservation of cut peony flowers, characterized in that, Includes the following steps: S1. Construct an offline dataset: Select peony cut flower samples and place them in multiple preset standard environmental scenarios. Use a volatile organic compound (VOC) collection device to collect full life cycle data in each scenario at predetermined time intervals to obtain VOC time series data for multiple scenarios. S2. Establish a baseline model library: Perform feature mapping and statistical analysis on the time series data of volatile organic compounds in the multiple scenarios, calculate the standard decay trajectory and standard metabolic network correlation matrix evolution model under each scenario, and construct an adaptive multi-branch baseline model library containing multiple baseline branches. S3. Obtain the initial state: At the start of online monitoring, collect the volatile organic compound data of the peony cut flowers to be monitored, and construct the initial metabolic pathway state vector of the batch of cut flowers based on the preset metabolic pathway classification rules. S4. Matching the baseline branch: Calculate the Euclidean distance between the initial metabolic pathway state vector and the initial node of each reference baseline branch in the adaptive multi-branch baseline model library, and select the branch with the smallest distance as the best matching baseline branch for this batch of cut flowers. S5. Calculate real-time drift: During subsequent monitoring, real-time data is continuously collected, and the state drift index of the real-time metabolic pathway state vector relative to the best-matching baseline branch is calculated, and the structural drift index of the real-time correlation matrix relative to the best-matching baseline branch is calculated. S6. Generate early warning decision: Perform weighted fusion calculation on the state drift index and the structural drift index to obtain a comprehensive deterioration index, compare the comprehensive deterioration index with the multi-level early warning threshold, and output the corresponding preservation early warning signal and control command.

2. The method for early warning of peony cut flower preservation according to claim 1, characterized in that, In step S1, the preset standard environment scenario includes at least the following: An ideal control group scenario simulating optimal warehousing conditions; Simulate transportation stress scenarios involving temperature fluctuations and mechanical vibrations during cold chain transportation; Simulate a high-humidity, susceptible disease group scenario that easily induces fungal diseases; An ethylene-induced group scenario simulating the accelerated aging process caused by ethylene pollution.

3. The method for early warning of peony cut flower preservation according to claim 1, characterized in that, In step S2, the feature mapping of the multi-scenario volatile organic compound time-series data specifically includes: The identified volatile organic compounds are mapped to preset metabolic function categories to obtain the metabolic function category mapping relationship. The metabolic function categories include at least the aroma synthesis metabolic pathway, the stress response metabolic pathway, and the pathogen interaction metabolic pathway. Based on the aforementioned metabolic function category mapping relationship, the metabolic pathway state vector at a specified time is calculated using the metabolic pathway state vector definition formula.

4. The method for early warning of peony cut flower preservation according to claim 3, characterized in that, In step S2, calculating the standard decay trajectory for each scenario specifically includes: The repeated sample data of each scenario in the multi-scenario volatile organic compound time series data are statistically averaged, and the standard state point of the scenario at each discrete time point is calculated using the standard state point calculation formula. By connecting the standard state points in time sequence, a standard decay trajectory in the scenario is generated.

5. The method for early warning of peony cut flower preservation according to claim 4, characterized in that, In step S2, the calculation of the standard metabolic network correlation matrix evolution model for each scenario specifically includes: A sliding time window is set, and for each time point on the standard decay trajectory, the Pearson correlation coefficient between different volatile organic compounds is calculated using the standard sliding window correlation formula, and a standard correlation matrix is ​​constructed. Based on the aforementioned standard correlation matrix, a standard metabolic network correlation matrix evolution model is generated.

6. The method for early warning of peony cut flower preservation according to claim 1, characterized in that, In step S4, calculating the Euclidean distance between the initial metabolic pathway state vector and the initial nodes of each reference baseline branch in the adaptive multi-branch baseline model library, and selecting the branch with the smallest distance as the best matching baseline branch for this batch of cut flowers, specifically includes: The Euclidean distance between the initial metabolic pathway state vector and the standard state points of each reference baseline branch in the adaptive multi-branch baseline model library at the initial time is calculated using the optimal baseline selection formula. The minimum value is determined based on the Euclidean distance, and the branch index corresponding to the minimum value is determined as the best matching baseline branch.

7. The method for early warning of peony cut flower preservation according to claim 1, characterized in that, In step S5, calculating the state drift index of the real-time metabolic pathway state vector relative to the best-matched baseline branch specifically includes: The state divergence between the real-time metabolic pathway state vector and the best-matched baseline branch at the corresponding time standard state point is calculated using the state divergence calculation formula. The rate divergence between the rate of change vector of the real-time metabolic pathway state vector and the theoretical rate of change vector of the best-matched baseline branch is calculated using the rate divergence calculation formula. The state divergence and the velocity divergence are weighted and fused using the state drift index calculation formula to obtain the state drift index.

8. The method for early warning of peony cut flower preservation according to claim 5, characterized in that, In step S5, calculating the structural drift index of the real-time correlation matrix relative to the best-matching baseline branch specifically includes: A real-time sliding time window is constructed based on real-time acquired data, and the real-time correlation matrix is ​​calculated using the real-time sliding window correlation formula. The Frobenius norm distance between the real-time correlation matrix and the standard correlation matrix at the corresponding time in the standard metabolic network correlation matrix evolution model is calculated using the structural drift index calculation formula, thus obtaining the structural drift index.

9. The method for early warning of peony cut flower preservation according to claim 1, characterized in that, In step S6, the generation of early warning decisions specifically includes: The state drift index and the structural drift index are weighted and fused using the comprehensive metamorphic index calculation formula to obtain the comprehensive metamorphic index; The comprehensive deterioration index is compared with the first threshold for determining the yellow warning level and the second threshold for determining the red emergency level, respectively, and a freshness warning signal is output. The first threshold is a preset value that characterizes the physiological indicators of cut flowers as deviating from the normal range but without irreversible damage, and the second threshold is a preset value that characterizes the cut flowers as having obvious lesions or excessive aging.

10. A method for early warning of peony cut flower preservation according to claim 3, characterized in that, In step S6, the output of the corresponding preservation warning signal and control command specifically includes: When the judgment result triggers the warning level, the contribution rates of the aroma synthesis metabolic pathway, the stress response metabolic pathway, and the pathogen interaction metabolic pathway to the overall state drift are calculated using the pathway deviation contribution rate calculation formula. Based on the contribution rate of the overall state drift, the metabolic pathway category corresponding to the maximum value or the weighted contribution ratio based on the structural drift index is selected, and a diagnostic state label containing physical environmental stress, disease infection risk or excessive aging is output as a regulatory instruction.

Images