Food quality intelligent evaluation system

Abstract

The invention relates to the technical field of intelligent evaluation, in particular to an intelligent food quality evaluation system which comprises a gas path regulation and control module, a fluctuation recognition and aggregation module, a main cause mapping weight module, a risk score output module and a section risk calibration module. According to the invention, through the real-time adjustment of multi-region temperature control and a gas channel and the dynamic correction of structure response characteristic capture and component identification positions, the whole-course perception and hierarchical aggregation of gas behavior characteristics are realized, and the signal fluctuation and disturbance process is reconstructed through the main cause weight. The food state is represented as clear trend progression and risk partition under periodic change, signal range, stage switching and section refinement in the quality evolution process can be dynamically concluded, continuous, layered and traceable risk quantification basis is output for complex food samples, mixed components and sudden deterioration scenes, and the risk quantification effect is improved. And the sensitivity, meticulous property and data chain closed-loop capability of quality discrimination are enhanced.

Description

Technical Field

[0001] This invention relates to the field of intelligent evaluation technology, and in particular to an intelligent food quality evaluation system. Background Technology

[0002] Intelligent assessment involves a comprehensive information processing system that utilizes artificial intelligence, machine learning, and automated analysis methods to determine the state, quality, and predict trends of specific objects or processes. It is widely applied in various sub-fields such as industrial quality control, medical diagnosis, equipment failure prediction, and agricultural environmental monitoring. It is a comprehensive application field centered on logic and guided by evaluation objectives. Traditional intelligent food quality assessment systems refer to devices or systems used to determine and classify the physical, chemical, and biological indicators of food. They target non-destructive testing and comprehensive evaluation of food quality. Typically, electronic noses are used to acquire volatile organic compound signals, and principal component analysis is used to extract key feature variables. Then, a support vector machine model is used to complete sample classification and quality grade determination.

[0003] Existing technologies often employ static signal analysis and single feature focusing methods. When faced with gas component interactions, temperature changes, or stratified releases in food samples, they are often slow to respond to continuous changes, sudden disturbances, and main cause switching. The data chain lacks temporal features and multi-segment dynamic information, resulting in insufficient ability to distinguish abrupt changes, gradual changes, and detailed segmentation in the evaluation results. Risk segments are difficult to define accurately, and food quality trends are often unclear or anomalies are missed. Summary of the Invention

[0004] The purpose of this invention is to address the shortcomings of existing technologies by proposing an intelligent food quality assessment system.

[0005] To achieve the above objectives, the present invention adopts the following technical solution: a food quality intelligent assessment system, the system comprising:

[0006] The gas path control module is based on food samples. It uses an electronic nose to sense and analyze the flow path of the sample in the gas collection structure, adjusts the response position of the thermal control zone and the spectrum component, and performs calibration through dynamic signals to obtain dynamic channel response characteristics.

[0007] Based on the dynamic channel response characteristics, the fluctuation recognition and aggregation module analyzes the changes in time series signals, determines the concentration trends and change areas of continuous sampling points, extracts and aggregates the temporal features of the disturbance segment, and obtains the disturbance trajectory aggregation features.

[0008] The principal factor mapping weight module identifies the fluctuation differences of key gas factors under different time structures based on the aggregated features of the disturbance trajectory, adjusts the principal factor weight distribution, dynamically maps the periodicity of factor weights, and obtains the weight distribution mapping factor.

[0009] The risk scoring output module analyzes the trend performance of the dominant factor at each stage based on the weight distribution mapping factor, identifies the switching point, adjusts the risk score fluctuation, and obtains the stage risk change amount.

[0010] Based on the stage risk variation, the segment risk calibration module analyzes the scoring trajectory, identifies segments in the risk curve with continuous fluctuations, determines the time period of each segment and performs classification to obtain the continuous risk distribution interval.

[0011] The present invention improves upon the following: the dynamic channel response characteristics include channel flow velocity distribution, hot zone temperature change, and spectral signal characteristics; the disturbance trajectory aggregation characteristics include disturbance segment identification, time series characteristic parameters, and signal change trajectory; the weight distribution mapping factors include principal factor weight distribution, time series structure mapping parameters, and factor influence classification; the stage risk variation includes stage risk level, risk change magnitude, and stage change trend; and the continuous risk distribution interval includes segment risk classification, continuous interval time series, and interval identification number.

[0012] The present invention is improved in that the gas path control module includes:

[0013] The gas path identification submodule analyzes the flow path and concentration change trend of the gas released from the food sample at each cross section of the collection structure, compares the flow velocity changes and gas distribution characteristics at each cross section, determines the diffusion and convergence trajectory of the gas in the guide channel, and obtains the channel flow velocity distribution factor.

[0014] The hot zone division and adjustment submodule adjusts the spatial layout of the thermal control unit based on the channel velocity distribution factor, calculates the release area and release sequence of gas under each thermal control state, compares the changes in gas behavior under temperature control conditions, and obtains the hot zone temperature change sequence.

[0015] The spectrum response calibration submodule analyzes the influence of each hot zone on the gas release characteristics based on the temperature change sequence of the hot zone, compares the signal change trends of each response channel of the spectrum component, determines the change of the signal response center and the channel bandwidth adjustment requirements, optimizes the channel response parameter adjustment method, and obtains dynamic channel response characteristics.

[0016] The present invention is improved in that the fluctuation identification aggregation module includes:

[0017] Based on the dynamic channel response characteristics, the signal trend recognition submodule calculates the direction and rate of change of gas concentration at adjacent sampling points, determines the time interval of continuous change in the signal sequence, identifies continuous segments with consistent change direction and changing rate, and obtains the continuous change trend interval.

[0018] The disturbance section location submodule determines the location of signal rate abrupt change based on the continuous change trend interval, compares the degree of change in slope and rate of change of adjacent sampling points, analyzes the fluctuation direction and duration before and after the abrupt change point, determines the section boundary by the density of the abrupt change point distribution, and obtains the disturbance section identification index.

[0019] The temporal trajectory aggregation submodule calculates the start and end times, sampling frequency, and fluctuation direction of the disturbance segment based on the disturbance segment identifier index, determines the duration and intensity of change of each disturbance segment, optimizes the segment aggregation method of common time structure and repetitive fluctuation characteristics, and obtains the disturbance trajectory aggregation features.

[0020] The present invention is improved in that the principal factor mapping weight module includes:

[0021] The response interval extraction submodule analyzes the perturbation segment identifier and signal change trajectory based on the perturbation trajectory aggregation feature, compares the response changes of gas factors within fixed and sliding time windows, determines which time periods have response amplitudes higher than the average level of the same sequence, and obtains a set of active response intervals.

[0022] The fluctuation difference judgment submodule compares the gas factor response under each time window based on the set of active response intervals, calculates the response amplitude and rate of change of each segment, determines whether the amplitude change under each time window forms fluctuation, filters the dominant factors of response difference under multiple windows, and obtains the window difference response sequence.

[0023] The response weight adjustment submodule analyzes the average response amplitude and duration of the dominant factors under each time series characteristic based on the window difference response sequence, and obtains the weight distribution mapping factor of each dominant factor under each periodic structure.

[0024] The present invention is improved in that the risk scoring output module includes:

[0025] The factor performance analysis submodule analyzes the response characteristics in each stage based on the weight distribution mapping factor, compares the signal change direction and amplitude of the dominant factor in each stage, judges the response trend and fluctuation distribution of the dominant factor in each stage, and obtains the stage comparison factor group.

[0026] The risk curve identification submodule identifies the continuous response characteristics of the dominant factor in time based on the stage comparison factor group, analyzes the change in the response direction of the dominant factor between each stage, judges the change trajectory of the dominant factor between adjacent stages, optimizes the identification criteria of the dominant factor switching interval, and obtains the switching interval sequence set.

[0027] The scoring fluctuation adjustment submodule, based on the switching segment sequence set, determines the changing trend of the dominant factor response trajectory before and after the switching, calculates the intensity and rhythm of scoring fluctuations in the difference segment, and uses the following formula:

[0028] ;

[0029] By obtaining the response amplitude of the scoring rhythm, the change in risk at each stage can be obtained. Indicates the first Differences in the scoring rhythm response of the sections Indicates the dominant factor in the th Average response magnitude in the phase prior to segment handover Indicates the dominant factor in the th Average response magnitude after segment handover Indicates the first Concentration of spectral response changes in the segment Indicates the first The degree of coupling of the phased response of the section Indicates the first segment dominant factor and the first The joint response offset of the three co-occurring factors.

[0030] The present invention is improved in that the section risk identification module includes:

[0031] The scoring trajectory analysis submodule calculates the rate of change of scoring amplitude between consecutive sampling points based on the stage risk change amount, determines the time interval of continuous change in the scoring trajectory, identifies fluctuation segments with continuous amplitude change trends and different rates, and obtains the scoring amplitude fluctuation segment.

[0032] The risk segment identification submodule determines the start and end positions of each segment on the time axis of the scoring curve based on the scoring amplitude fluctuation segment, compares the period and trend of scoring fluctuation within each segment, identifies risk segments with continuous scoring changes, and obtains the stage scoring time sequence index.

[0033] The continuous interval division submodule analyzes the temporal connection relationship of each scoring segment based on the stage scoring time sequence index, adjusts the segment boundary method, compares the continuity of the scoring trajectory within each segment, constructs a combined distribution of scoring fluctuation segments of the same type, and obtains a continuous risk distribution interval.

[0034] The present invention is improved in that the gas collection structure refers to the sealed transmission channel formed between the food sample and the detection chamber, the time series signal is the response data sequence continuously output by the sensor within the sampling period, and the continuous sampling points represent the data points recorded by the sensor at fixed intervals within the same sampling period.

[0035] Compared with the prior art, the advantages and positive effects of the present invention are as follows:

[0036] In this invention, through multi-region temperature control and real-time adjustment of gas channels, combined with structural response feature capture and dynamic correction of component identification positions, the full-process perception and hierarchical aggregation of gas behavior characteristics are achieved. Signal fluctuations and disturbances are reconstructed through principal weights, so that the food state exhibits a clear trend progression and risk zoning under periodic changes. The signal range, stage switching, and segment refinement in the quality evolution process can all be dynamically summarized. For complex food samples, mixed components, and sudden spoilage scenarios, continuous, hierarchical, and traceable risk quantification data is output, enhancing the sensitivity, detail, and data chain closed-loop capability of quality judgment. Attached Figure Description

[0037] Figure 1 This is a system flowchart of the present invention;

[0038] Figure 2 This is a flowchart of the gas path control module in this invention;

[0039] Figure 3 This is a flowchart of the fluctuation recognition and aggregation module in this invention;

[0040] Figure 4 This is a flowchart of the principal cause mapping weight module in this invention;

[0041] Figure 5 This is a flowchart of the risk scoring output module in this invention;

[0042] Figure 6 This is a flowchart of the section risk assessment module in this invention. Detailed Implementation

[0043] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0044] In the description of this invention, it should be understood that the terms "length," "width," "upper," "lower," "front," "rear," "left," "right," "vertical," "horizontal," "top," "bottom," "inner," and "outer," etc., indicating orientation or positional relationships, are based on the orientation or positional relationships shown in the accompanying drawings and are only for the convenience of describing the invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation, and therefore should not be construed as a limitation of the invention. Furthermore, in the description of this invention, "a plurality of" means two or more, unless otherwise explicitly specified.

[0045] Example

[0046] Please see Figure 1This invention provides a technical solution: a food quality intelligent assessment system comprising:

[0047] The gas path control module is based on food samples. It uses an electronic nose to analyze the gas movement path of the sample in each stage of the gas acquisition structure, optimizes the thermal control zone division, changes the gas release environment by adjusting heating and cooling conditions, judges the structural response characteristics of the adsorbent gas release area, adjusts the gas component response position corresponding to the spectrum component, and performs dynamic calibration based on the response signal to obtain dynamic channel response characteristics.

[0048] The fluctuation identification and aggregation module analyzes the time series signal within the acquisition period based on the dynamic channel response characteristics. By judging the concentration change trend between continuous sampling points, it identifies the continuous change segment with large fluctuation amplitude, judges the abrupt change point in the signal of each segment, optimizes the feature identification process of the disturbance segment, and aggregates the time series information of the disturbance segment to obtain the disturbance trajectory aggregation feature.

[0049] The principal factor mapping weight module analyzes the response range of each gas factor based on the perturbation trajectory aggregation characteristics, identifies the fluctuation differences of the dominant factor under multiple time structures, judges the parameter change trend caused by the periodic change of the dominant factor, adjusts the response distribution weight of each dominant factor in the differential period, and obtains the weight distribution mapping factor.

[0050] The risk scoring output module is based on the weight distribution mapping factor. It compares the performance characteristics of the dominant factor in each stage. By analyzing the changing trend of the dominant factor on the risk curve, it determines the point of change when the dominant factor switches. It then adjusts the response amplitude of the risk score with the point of change to obtain the stage risk change.

[0051] The segment risk labeling module analyzes the stage scoring trajectory based on the stage risk variation, identifies risk segments with continuous fluctuations in the scoring curve, determines the time period of the risk segment, optimizes the risk segment division process, and obtains the continuous risk distribution interval.

[0052] Dynamic channel response characteristics include channel flow velocity distribution, hot zone temperature changes, and spectral signal characteristics; disturbance trajectory aggregation characteristics include disturbance segment identification, time series characteristic parameters, and signal change trajectory; weight distribution mapping factors include principal factor weight distribution, time series structure mapping parameters, and factor influence classification; stage risk variation includes stage risk level, risk change magnitude, and stage change trend; continuous risk distribution interval includes segment risk classification, continuous interval time series, and interval identification number.

[0053] In the gas path control module, the gas acquisition structure refers to the sealed transmission channel formed between the food sample and the detection chamber, used to guide the volatile gases released from the sample into the electronic nose system; the gas movement path refers to the entire process of gas molecules being released from the sample surface within the acquisition structure, passing through the guide channel, and reaching the detection site; the thermal control zone division method refers to the formation of different temperature zones within the acquisition chamber by setting heating or cooling units, used to induce the stratified release of different active gases; the gas release environment refers to the combined conditions such as temperature, humidity, and flow rate of the gas during release, which directly affects the stability and rate of gas release; the adsorbed gas release region refers to the specific spatial region where gas release is concentrated due to delayed gas release caused by surface adsorption; the structural response characteristics are used to describe the physical response of the acquisition structure under temperature changes, such as changes in diffusion rate or flow resistance; the spectrum component is a spectral analysis device for detecting gas components, used to resolve the characteristic bands corresponding to different gases; the gas component response position refers to the center position of the response channel corresponding to different gas characteristic signals in the spectrum component; dynamic calibration refers to automatically adjusting the detection channel parameters based on the real-time response signal to maintain the accuracy of data output.

[0054] In the fluctuation recognition aggregation module, the time series signal is the continuous response data sequence output by the sensor within the sampling period, used to describe the change of gas concentration over time; continuous sampling points represent the data points recorded by the sensor at fixed intervals within the same sampling period; the concentration change trend refers to the direction and magnitude of the increase or decrease in gas concentration between adjacent sampling points; fluctuation amplitude is used to measure the overall intensity of signal change over a period of time; continuous change segment refers to a signal segment in the time series where the direction of gas concentration change remains consistent; abrupt change points within the signal are key points where the rate of change of gas signal in the time series suddenly increases or the direction reverses; the disturbance segment feature recognition process describes how to locate and identify the disturbance signal region by comparing the slope and rate of change; the disturbance segment is the signal interval containing the abrupt change point and adjacent sampling points; the temporal information refers to the time arrangement, duration, and periodicity of sampling points within the disturbance segment, used to describe the rhythm of gas dynamic changes.

[0055] In the principal factor mapping weight module, the gas factor response interval refers to the time period during which each gas component exhibits an active response in the time series; the dominant factor represents the key gas component that plays a major role in the changes of food gases in the current detection cycle; the multi-time structure is a method of parallel analysis of signals using multiple time windows such as fixed time periods and sliding time periods; the fluctuation difference reflects the difference in the intensity or frequency of the dominant factor's response under different time windows; the parameter change trend refers to the direction and pattern of the change of the dominant factor's response parameters over time during the detection process; and the response distribution weight refers to the proportion of influence allocated to each factor after comprehensive analysis of each gas factor, used to calculate the overall risk contribution.

[0056] In the risk scoring output module, each stage represents a different time phase in the testing cycle of the food sample, such as the initial, active, and stable stages; the performance characteristics describe the response pattern and fluctuation characteristics of the dominant factor in each stage; the risk curve is a risk change trend graph generated based on the response and weight of the dominant factor; the change trend refers to the direction and rate of change of the risk curve over time, used to reflect the dynamic changes in quality risk; the point of change when a switch occurs is a key time node when the dominant factor or risk stage changes; the risk score is a quantitative indicator obtained by calculating the response intensity and weight ratio of the dominant factor; the response amplitude describes the fluctuation range of the risk score before and after the point of change, used to measure the intensity of risk change.

[0057] In the segment risk calibration module, the stage scoring trajectory is a continuous change curve formed by the risk scores of different stages in chronological order; the continuous fluctuation amplitude change is used to measure the persistence and intensity of the changes in the scoring trajectory within adjacent time periods; the segmentation process refers to the entire process of segmenting, risk classification and numbering and archiving after identifying risk segments in the scoring trajectory, which is used to establish a risk distribution model.

[0058] Please see Figure 2 The gas path control module includes:

[0059] The gas path identification submodule analyzes the flow path and concentration change trend of the gas released from the food sample at each cross section of the collection structure, compares the flow velocity changes and gas distribution characteristics at each cross section, determines the diffusion and convergence trajectory of the gas in the guide channel, and obtains the channel flow velocity distribution factor.

[0060] Samples such as cooked meat, jams, and fermented yogurt are placed in a sealed collection chamber. A gas collection structure is constructed from the sample surface to the sensor, and multiple detection sections are divided within the structure, for example, at distances of 5 mm, 10 mm, and 15 mm from the sample. Each section is equipped with a detector for gas flow rate and gas concentration. The sensor continuously collects and records the instantaneous flow rate and target gas component concentration at each section, with a detection time interval of once per second, covering a period of 60 seconds. During data acquisition, the flow rate difference at the same section at adjacent time points is compared to determine the flow rate trend. For example, if the flow rate difference between the current and previous moments is greater than 0.08 m / s, it is marked as an accelerating phase; if it is less than -0.08 m / s, it is marked as a decelerating phase; and if the flow rate fluctuation remains within ±0.02 m / s, it is considered a stable flow phase. To determine the concentration trend, a multi-channel gas detector is used to record the peak concentration of each gas type. The direction of concentration change of the same gas component between different cross sections is analyzed. For example, the concentration of ethanol released from the sample is 3.1 ppm at 5 mm, 2.3 ppm at 10 mm, and 1.6 ppm at 15 mm, indicating a decreasing trend from the sample surface outwards. If the flow rate data also shows an accelerating trend, then the channel segment can be confirmed as a diffusion area. If the concentration changes in an increasing direction while the flow rate tends to slow down, then gas convergence is presumed. In the experiment, when testing samples of cured bacon, five cross-sections were set up, and the average flow velocities were measured to be 0.28, 0.33, 0.31, 0.29, and 0.35 meters per second, respectively. By calculating the fluctuation range and direction of the flow velocity difference between adjacent cross-sections, the degree of non-uniformity of gas flow at different locations in the entire channel was obtained. This degree of non-uniformity can be used to describe the gas diffusion and convergence path. After combining the flow velocity difference results of multiple cross-sections and sampling time, the overall performance of the gas diffusion and convergence trend in the guide channel was obtained, thus obtaining the channel flow velocity distribution factor as the basis for subsequent temperature control unit configuration.

[0061] The hot zone division and adjustment submodule adjusts the spatial layout of the thermal control unit based on the channel velocity distribution factor, calculates the release area and release sequence of gas under each thermal control state, compares the changes in gas behavior under temperature control conditions, and obtains the hot zone temperature change sequence.

[0062] Based on the aforementioned channel velocity distribution factor, the thermal control unit layout is set. First, the gas channel is divided into several preset thermal control zones, and heating or cooling units are preferentially deployed in areas with significant velocity changes. For example, if the velocity difference in a certain section exceeds 0.05 meters per second, that section is selected as a candidate thermal control node area. Then, the zone is further divided into upper and lower three-layer temperature control zones according to the deployment height (e.g., 5 mm, 10 mm, 15 mm). After deploying thermal control elements in each layer, the heating elements are set to 35°C and the cooling elements to 15°C through real-time temperature control. The standard area is maintained at approximately 25°C. Then, the sample gas release process is initiated, and the temperature control is adjusted accordingly. The system records the types of released gases, their concentration changes, and peak times to analyze the impact of temperature control on gas behavior. For example, in the detection of salted egg yolk samples, the peak time for ammonia release was 18 seconds at 15℃, 14 seconds at 25℃, and 9 seconds at 35℃, with response amplitudes of 1.0, 1.4, and 2.1, respectively. By comparing the data, it can be seen that gas release is more rapid and the response intensity is higher under heating temperature control, indicating that high temperature promotes active gas release. The system also records the peak time, concentration range, and signal attenuation rate under different temperature controls to obtain the correlation sequence between temperature regulation and release behavior. This sequence uses the actual measured temperature points, gas response delays, and amplitudes as components. Through sorting and mapping, the performance under each thermal control state is organized into a continuous record, forming a thermal zone temperature change sequence. In actual experiments, for example, in the detection results of a fermented fruit and vegetable sample under three temperature settings, the concentration difference of high-frequency volatiles such as ethanol is more than 1.8, and the response time difference is more than 10 seconds. The difference becomes the basis for thermal zone adjustment. The gas response behavior under temperature control is arranged in sequence, and the corresponding thermal zone temperature change sequence under the sample is output to guide the subsequent dynamic calibration of the spectral channel parameters.

[0063] The spectrum response calibration submodule analyzes the influence of each hot zone on gas release characteristics based on the temperature change sequence of the hot zone, compares the signal change trends of each response channel of the spectrum component, judges the change of the signal response center and the channel bandwidth adjustment requirements, optimizes the channel response parameter adjustment method, and obtains dynamic channel response characteristics.

[0064] The response results of each spectral channel under different thermal control conditions were retrieved one by one. The shift of the main peak band of the target gas in each response channel under temperature control changes was compared. First, the peak value of the dominant absorption band of the sample gas by the sensor was extracted in each temperature control region. For example, the spectral center was at 4.20 micrometers at 35℃, while it was 4.32 micrometers at 15℃, showing a shift in the center wavelength towards longer wavelengths. By comparing the difference in the main peak position under different temperature control conditions, it was determined whether the current spectral channel center setting was matched. If the change range exceeded 0.1 micrometers, it was recorded as needing channel center adjustment; if the change was less than 0.03 micrometers, it was recorded as not needing adjustment. Then, the maximum response amplitude of the same gas component under different thermal control conditions was compared. For example, the signal strength of ammonia was 2.3 at high temperature and 1.4 at low temperature. A change amplitude exceeding 60% was classified as a high-fluctuation response intensity group. Then, it is determined whether the channel bandwidth should be expanded to cover the drift range of the response peak. For example, if the original bandwidth is ±0.1 micrometers, and the band drift exceeds this range, the bandwidth should be expanded to ±0.2 micrometers. The newly added channels are set and numbered. At the same time, the original spectral response center is compared with the adjusted response center, and the adjustment magnitude and direction are recorded. In actual sample detection, such as the spectral response of ethanol released from fermented beverages having a main peak wavelength difference of ±0.15 micrometers under different temperature control conditions, the corresponding spectral channel center is adjusted to the median value and the channel bandwidth is expanded to adapt to the response drift. The corrected channel configuration is then used as the standard setting in the current environment. The current channel center band position, signal peak value, response delay time, and bandwidth configuration parameters are output to form dynamic channel response characteristics. This feature set is used to reflect the actual detection capability and response position configuration of the current spectrum component under real-time sampling conditions.

[0065] Please see Figure 3 The fluctuation recognition aggregation module includes:

[0066] The signal trend recognition submodule calculates the direction and rate of change of gas concentration at adjacent sampling points based on the dynamic channel response characteristics, determines the time interval of continuous change in the signal sequence, identifies continuous segments with consistent change direction and changing rate, and obtains the continuous change trend interval.

[0067] After the concentration values ​​are recorded at 1-second intervals within the acquisition period to form a sequence of data, the difference between adjacent sampling points is processed to obtain the direction of gas concentration change. If the concentration at the current point minus the concentration at the previous point is greater than 0, it is marked as an increase; otherwise, it is marked as a decrease. If the absolute value of the difference is less than 0.05 ppm, it is marked as no significant change. Segments with the same continuous direction of change are clustered and labeled. At the same time, the rate of change is calculated as the difference between adjacent concentrations divided by the time interval. The entire sequence is traversed to obtain each segment with consistent direction and its corresponding rate. In the experiment of milk spoilage samples, the ethanol concentration sequence is [0.5, 0.8, 1.1, 1.5, 1.8, 2.0, 1.9, 1.7, 1.5] ppm, collected once per second. The concentration increases continuously from the 1st to the 6th second and decreases from the 7th to the 9th second. Two continuous change segments are formed by direction determination, and the change rates are recorded as 0.3, 0.3, and 0.4 ppm, respectively. The rates were 0.3, 0.2, 0.2 and -0.1, -0.2, -0.2 ppm / s. The rates of each sampling point within each continuous variation segment were then interpolated to extract the rate change trend. If the rate fluctuation amplitude exceeded 0.2 ppm / s within a single segment, it was identified as a non-uniform speed segment; if the fluctuation was below 0.05 ppm / s, it was considered a steady-state variation segment. This method refines the identification of parts of the signal sequence where the change direction is consistent but the rate is different. In the above samples, the rate changes from 1 to 4 seconds were 0.3, 0.3, 0.4, and 0.3 respectively, with a fluctuation range of 0.1 ppm / s. Therefore, this segment was marked as a relatively consistent rate region. From 5 to 6 seconds, the rate changed from 0.2 ppm / s to 0.2 ppm / s with zero fluctuation, and was identified as a steady-state rising segment. By comprehensively processing the change direction and rate characteristics of all adjacent sampling points, the continuous change trend intervals of each segment were obtained.

[0068] The disturbance section location submodule determines the location of signal rate abrupt changes based on the continuous change trend interval, compares the degree of change in the slope and rate of changes of adjacent sampling points, analyzes the fluctuation direction and duration before and after the abrupt change point, determines the section boundary by the density of the abrupt change point distribution, and obtains the disturbance section identification index.

[0069] The slope of adjacent sampling points in the sequence is calculated, and the difference between the rate at each point and the rate at the previous sampling point is obtained and compared with the rate value itself. If the absolute value of the difference is greater than 0.3 ppm / s, it is judged as a mutation position; if it is less than 0.1 ppm / s, the trend is considered stable. Then, the direction of the fluctuation before and after is combined to determine whether there is a reversal before and after the mutation. If the rate changes from positive to negative or from negative to positive, it is classified as an inversion mutation. For example, in a gas release sequence of a fermented soybean product, the ammonia concentration rises from 2.0 to 3.5 ppm in 5 seconds and then suddenly drops to 2.6 ppm, and then rises to 2.9 ppm in a short time. From the 5th to the 6th second, the rate changes from 0.3 to -0.9 ppm / s, with a difference of -1.2, which exceeds the mutation judgment threshold of 0.3 ppm / s. Therefore, this point is marked as a mutation point. The duration of fluctuations before and after the mutation point is analyzed. If the total fluctuation period before and after the mutation exceeds 5 seconds and the direction is reversed before and after the mutation, it is constructed as an independent perturbation candidate segment. The positions of each candidate perturbation point in the sequence are compared. If three or more mutation points appear consecutively in a certain segment and the distance between adjacent mutation points is less than 2 seconds, the segment is determined to be a mutation-dense segment. The start and end boundary positions of the segment are then determined and recorded. In the above soybean product sample, the mutation density occurs between the 5th and 8th seconds, including two inversion mutation points and one point of drastic rate change. The start time of this segment is recorded as the 5th second and the end time as the 8th second, and it is numbered D-01. This segment is marked as a perturbation segment and a perturbation segment identification index is established.

[0070] The temporal trajectory aggregation submodule calculates the start and end times, sampling frequency, and fluctuation direction of the disturbance segment based on the disturbance segment identifier index, determines the duration and intensity of change of each disturbance segment, optimizes the segment aggregation method for common time structure and repetitive fluctuation characteristics, and obtains the disturbance trajectory aggregation features.

[0071] The system retrieves all segment numbers and their start and end times from the disturbance segment identifier index. For each disturbance segment, it extracts the corresponding time span, sampling frequency, and fluctuation direction for classification. For example, segment D-01 starts at second 5, ends at second 8, has a sampling frequency of 1Hz, and exhibits a fluctuation direction of first rising, then falling, then rising again, classifying it as a multiple-reversal disturbance. The duration of this segment is recorded as 3 seconds, with two fluctuation direction changes occurring at second 6 and second 7. The difference between the maximum and minimum fluctuation amplitudes is calculated to determine the change intensity as 1.0 ppm. All disturbance segments are grouped according to the number of fluctuation direction changes: single-reversal group, multiple-reversal group, and continuous unidirectional group. This further identifies disturbances that repeatedly occur in multiple samples. For dynamic time structure, if different samples all show the same type of disturbance segment within a similar time period (error less than 2 seconds), it is determined to be a common time structure. For example, if both chicken and fish samples show a disturbance segment with a fluctuation amplitude greater than 1.5 ppm and a fluctuation direction switching twice from the 10th to the 14th second, this segment is classified into a common disturbance type and labeled as aggregation category A1. Segment merging is performed on disturbance segments of the same category to form a unified aggregation structure feature set. In this example, the start time of the aggregated disturbance trajectory is taken as the minimum value of the start time of each segment, and the end time is taken as the maximum value. The aggregation feature time period is determined to be from the 10th to the 14th second, the type is A1, and the repetition rate within multiple samples is more than 70%. This feature is recorded and output as the disturbance trajectory aggregation feature.

[0072] Please see Figure 4 The principal cause mapping weight module includes:

[0073] The response interval extraction submodule analyzes the perturbation segment identifier and signal change trajectory based on the perturbation trajectory aggregation feature, compares the response changes of gas factors in fixed and sliding time windows, determines which time periods have response amplitudes higher than the average level of the same sequence, and obtains the set of active response intervals.

[0074] The signal change trajectory corresponding to each perturbation segment is called to obtain the concentration data column sampled per second. Then, two types of analysis intervals are set in each perturbation segment: fixed window and sliding window. The fixed window length is set to 5 seconds, and the sliding window step size is set to 1 second. The concentration data contained in the current window are summarized in each window. The difference between the maximum and minimum values ​​in the window is calculated as the response amplitude. Then, this amplitude value is compared with the average concentration response amplitude of the entire perturbation trajectory. If the amplitude value is greater than 1.2 times the average value, the window is determined to be an active response segment. For example, in the detection of salted egg yolk samples, the ammonia response concentration fluctuates between the 10th and 25th seconds. The concentration differences are calculated to be 1.1, 2.5, and 0.8 ppm in the set fixed windows

[1015] ,

[1520] , and

[2025] , respectively. The entire perturbation trajectory is relatively stable. The average difference is 1.2 ppm. The value of window

[1520] is 1.2 times higher than the average, i.e., 2.5 > 1.44, which is determined to be an active response segment. At the same time, a sliding window is used to further refine the response. The difference in window

[1419] is 2.7 ppm, which also meets the judgment condition. Therefore, the interval

[1419] is listed as an active sub-segment. In the actual processing, the response activity judgment standard is set as a response amplitude threshold that is greater than 1.2 times the average amplitude. This threshold is set according to the standard amplitude variation range of the dominant component (such as ammonia) in the stable release stage in the test sample. The set value comes from the 95th percentile fluctuation value of historical sample data as a reference. The sampling points in the active window are further marked, and the time periods that continuously meet the conditions are merged into a set of active response intervals to obtain a set of response segments.

[0075] The fluctuation difference judgment submodule compares the gas factor response under each time window based on the set of active response intervals, calculates the response amplitude and rate of change of each segment, determines whether the amplitude change under each time window forms fluctuations, filters the dominant factors of response difference under multiple windows, and obtains the window difference response sequence.

[0076] The response characteristics of different gas factors under various time windows were analyzed one by one. For each active interval, the concentration data sequence was extracted according to the gas component. First, the difference between the maximum and minimum values ​​in the current segment was calculated as the amplitude value. Then, the concentration difference between adjacent sampling points was divided by the time interval to obtain the rate of change sequence. The difference between the maximum and minimum rates of change was then calculated to determine the fluctuation. If the difference exceeded 0.3 ppm / s, it was marked as having significant fluctuation. For example, in the fermented tofu sample, the concentration of ethanol in the interval of 30-40 seconds was [2.5, 2.9, 3.2, 2.8, 2.6, 3.1, 2.7, 2.4, 2.6, 2.5] ppm, and the rate of change between adjacent differences was 0.4, 0.3, -0.4, -0.2, 0.5, -0.4, -0.3, 0.2, -0.1 ppm / s, with a maximum of 0.5. The value is -0.4, and the difference is 0.9, which is greater than the set threshold of 0.3. Therefore, this segment is determined to be a fluctuating region. Then, the response amplitude and rate of change values ​​corresponding to different gas factors in all active intervals are sorted, and the three components with the strongest changes are extracted as the dominant factors of the current time window. In the example above, the amplitudes of ethanol, ammonia and sulfur dioxide in the same segment are 1.5, 2.1 and 0.7 ppm, respectively, and the rates of change are 0.5, 0.8 and 0.2 ppm / s, respectively. Therefore, ethanol and ammonia are determined to be the dominant response factors. They are numbered and marked with the current time window number W-04. At the same time, the response attributes of each factor are recorded. The same calculation is performed in all time windows. The numbers and performance indicators of all dominant factors in different time periods are collected to form a list of dominant response factors sorted by time. This list is output as the window difference response sequence.

[0077] The response weight adjustment submodule analyzes the average response amplitude and duration of the dominant factor under various time-series characteristics based on the window difference response sequence, using the following formula:

[0078] ;

[0079] The weight distribution mapping factors of each dominant factor under each periodic structure are obtained, where, Indicates dominant factor In periodic structure The weighted distribution mapping factor is used to quantify the proportion of the factor's response contribution within a given period. Indicates dominant factor In the cycle The change in response amplitude within a given period is calculated by the difference between the maximum and minimum response amplitudes of this factor during that period. Indicates dominant factor In the cycle The average response amplitude within the period is obtained by taking the arithmetic mean of all response data within that period. Indicates dominant factor The duration of the disturbance segment, i.e., the time span during which the factor remains active within a complete disturbance cycle. This represents the average length of the disturbance interval for all dominant factors. It is the average of the duration of each dominant factor and is used to measure the difference between the duration of a single factor disturbance and the overall trend.

[0080] The weighted distribution mapping factor refers to a proportional parameter that quantitatively reflects the contribution of each dominant factor to the overall dynamic response in the current period, after taking into account the changes in response amplitude, average response level, and differences in the time structure of the disturbance segment for each dominant factor under different periodic structures. It is a value used to measure the "influence" or "importance" of each dominant gas factor in different periods, expressing the magnitude of the factor's contribution to the overall risk change or system state in different time periods.

[0081] The numerator of the formula is used to integrate the response intensity of the factor, while the denominator introduces the absolute value of the perturbation time difference to balance its response time structure shift, thus constructing a unified dimensional contribution index of the dominant factor to the periodic response. Taking three dominant factors as an example, the original data are given below:

[0082] The maximum response is 0.47, the minimum response is 0.32, the average response is 0.401, and the duration of the disturbance is 8 seconds.

[0083] The maximum response is 0.62, the minimum response is 0.55, the average response is 0.581, and the duration of the disturbance is 10 seconds.

[0084] The maximum response is 0.5, the minimum response is 0.45, the average response is 0.473, and the duration of the disturbance is 12 seconds.

[0085] Average duration of perturbations for all dominant factors Seconds, the above parameters are normalized, among which response amplitude parameters ( , ) and disturbance time parameters ( , Using maximum value normalization, the corresponding values ​​after normalization are as follows:

[0086] : , , , ;

[0087] : , , , ;

[0088] : , , , ;

[0089] Substituting the normalized data into the formula for calculation, we obtain:

[0090] right :

[0091] ;

[0092] right :

[0093] ;

[0094] right :

[0095] ;

[0096] Based on the stability requirements of the periodic dynamic response structure

[0097] set up The evaluation benchmark range is as follows:

[0098] when When the factor is defined as the "primary response factor", it means that its response amplitude changes significantly and the average response intensity is high, indicating that the factor has a dominant contribution to the disturbance change in the current cycle.

[0099] when When the factor is defined as a “moderate response factor”, it indicates that it has certain differences in response magnitude or time structure, but the overall impact is at a moderate level.

[0100] when When the factor is classified as a "weak response factor", it means that its disturbance response is weak or its periodic influence is small, and it does not constitute a significant dynamic disturbance driving term.

[0101] Based on the calculation results, of , of All of them fall within the "primary response factor" range. of Falling into the "medium response factor" range, the calculated results not only clarify the level of influence of the dominant factor in the current cycle, but also provide a categorizable and callable quantitative basis for the output weight distribution mapping factor.

[0102] Please see Figure 5 The risk scoring output module includes:

[0103] The factor performance analysis submodule is based on the weight distribution mapping factor, analyzes the response characteristics in each stage, compares the signal change direction and magnitude of the dominant factor in each stage, judges the response trend and fluctuation distribution of the dominant factor in each stage, and obtains the stage comparison factor group.

[0104] Gas response data within each stage were extracted and a stage index sequence was constructed. The sample detection period was first divided into three time periods: initial stage, active stage, and stable stage, labeled P1, P2, and P3 respectively. The response intensity and direction of the dominant gas component within each corresponding time period were extracted. The direction of the response was determined based on the positive or negative value of the concentration change between adjacent sampling points. A continuous upward trend was recorded as a positive response trend, and a downward trend as a negative response trend. If frequent reversals occurred within a stage, and the number of reversals exceeded three, it was marked as a fluctuating response. For example, in the active stage P2, the ammonia concentration showed an increase-decrease-increase change within 30 seconds, with two reversals, which did not meet the fluctuating response condition. Ethanol, however, showed five reversals within the same stage, and was marked as a fluctuating response gas. The difference between the maximum response amplitude of each dominant factor and the average amplitude of other factors within the stage was calculated. If the difference was greater than 0.8 ppm, it was determined to be a strong influencing factor; if it was less than 0.3 ppm, it was marked as a weak influencing factor. The amplitude difference threshold set here is derived from the median value of the amplitude difference within the 90% confidence interval of the training samples. In actual operation, for example, in the fermented beverage sample, the initial stage ammonia response amplitude is 0.6 ppm, ethanol is 1.8 ppm, and sulfur dioxide is 0.7 ppm. The ethanol amplitude is higher than the mean (1.03) to 0.77, but does not exceed the 0.8 threshold, so it is still marked as a general influencing factor. In the active stage, ammonia reaches 3.1 ppm, exceeding the mean 1.9 ppm to 1.2 ppm, and is marked as a strong influencing factor. Based on the above judgment, the direction, intensity and fluctuation type of the dominant factor in each stage are recorded, and a stage factor feature table containing factors such as factor number, response trend, maximum amplitude, number of reversals, and influence level is constructed. The various response parameters of the same dominant factor in different stages are compared to determine whether its performance in different stages is consistent or changes. The dominant factor combination with differences in response trend, response amplitude and fluctuation structure among the main influencing factors in each stage is output as the stage comparison factor group.

[0105] The risk curve identification submodule identifies the continuous response characteristics of the dominant factor in time based on the stage comparison factor group, analyzes the change in the response direction of the dominant factor between each stage, judges the change trajectory of the dominant factor between adjacent stages, optimizes the identification criteria of the dominant factor switching interval, and obtains the switching interval sequence set.

[0106] A temporal continuous response feature index of the dominant factors was established. The complete concentration response sequences of each dominant factor within the detection period were retrieved and sliced ​​according to stage boundaries. The change in response direction between adjacent stages was analyzed. If the previous stage showed a positive response and the next stage a negative response, or vice versa, and the change occurred within 5 seconds before or after the stage switch, it was recorded as a direction switch. If the response direction was consistent in both stages but the response amplitude changed by more than 1.5 ppm, it was also recorded as a significant change. By calculating the mean and slope changes of the response of each dominant factor within 10 seconds before and after the stage switch, a significant inflection point was identified in its trajectory. If the slope change was greater than 0.2 ppm / s, it was recorded as an inflection point segment. For example, in a cooked chicken sample, the ethanol response trend was decreasing in the initial stage and increasing in the active stage. The change occurred at the 29-second stage. The concentration change before and after the switch point was 1.4 to 2.7 ppm, with a slope of 0.13 ppm / s, which did not reach the set slope change threshold and was not recorded as a switching segment. However, the response trend of ammonia changed from rising to falling at the switch point from active to stable phase, with a slope change of 0.28 ppm / s, and was marked as a switching segment. All marked switching segments were then numbered and the start and end times, response factor numbers, change direction and magnitude were recorded. Segments with two or more dominant factors changing simultaneously were marked as composite switching segments and assigned a priority weight of 1.5. Other segments with single factor changes were assigned a weight of 1.0. The weight values ​​here are derived from the proportion of the influence of composite change segments on the change in the total risk score in historical samples. Subsequently, based on the time position, factor characteristics and change magnitude of all identified switching segments, the segments were aggregated and sorted by time to form a switching segment sequence set.

[0107] The score fluctuation adjustment submodule, based on the switching segment sequence set, determines the changing trend of the dominant factor response trajectory before and after the switching, calculates the intensity and rhythm of score fluctuations in the difference segment, and uses the following formula:

[0108] ;

[0109] By obtaining the response amplitude of the scoring rhythm, the change in risk at each stage can be obtained. Indicates the first Differences in the scoring rhythm response of the sections Indicates the dominant factor in the th Average response magnitude in the phase prior to segment handover Indicates the dominant factor in the th Average response magnitude after segment handover Indicates the first Concentration of spectral response changes in the segment Indicates the first The degree of coupling of the phased response of the section Indicates the first segment dominant factor and the first The joint response offset of the three co-occurring factors;

[0110] The rating rhythm response amplitude reflects the degree of change in the response intensity of the dominant factor in a certain segment due to the difference in response state before and after the switch, as well as the concentrated change in the spectral response. After normalization and adjustment, the amplitude is calculated by combining the average response amplitude of the dominant factor before and after the switch with the concentration of the spectral response change. It is then normalized by combining the stage response coupling degree and the joint offset of the accompanying factors to quantify the rating change amplitude and overall volatility of the response rhythm of the dominant factor in the switch segment.

[0111] Collect dominant factors in the first The original response magnitude data before and after the segment handover, with the average response before the handover being... The average response after the switch is The two are transformed using the minimum-maximum normalization method. , Subsequently, the concentration of spectral response changes in this segment was obtained, and the original concentration was calculated as follows: After normalization, it becomes Next, by calculating the synchronization rate of the stage signals, the coupling degree of the stage response between the dominant factor and the overall response is obtained, with the original value being... Keeping the dimensionless form unchanged after normalization, the joint response offsets of the dominant factor and three accompanying factors in this segment were then collected. The original offsets were as follows: , , After minimum-maximum normalization, they are respectively , , Then, the three normalized offsets are summed to obtain Substitute into the formula to calculate the absolute difference term:

[0112] ;

[0113] Multiply The product term is obtained as Performing the square root operation on the product yields... ;

[0114] Next, calculate the denominator. The rating rhythm response amplitude was calculated as follows:

[0115] ;

[0116] Scoring rhythm response difference The numerical space in which it exists is divided into five intervals according to the empirical distribution results of the scoring model, among which:

[0117] when The period is defined as a period with no significant change, indicating that the response state of the dominant factor is approximately continuous before and after the switching phase, and the fluctuation does not cause interference.

[0118] when When this period is defined as a weak variation segment, it indicates that the dominant factor has a small discontinuous response, but does not cause a disturbance to the scoring structure.

[0119] when The period is defined as the intermediate fluctuation segment, which indicates that a slight inconsistency in response occurs during the switching of the dominant factor, resulting in local perturbations in the scoring rhythm.

[0120] when This period is defined as a period of strong change, characterized by a structural abrupt change in the response of the dominant factor that affects the continuity of the scoring rhythm.

[0121] when When the time is defined as an abnormal fluctuation segment, the scoring status in this segment is severely disturbed, accompanied by interference peaks or spectrum mismatch.

[0122] The current calculation result is This is located in the middle to high position of the fourth segment of strong change, indicating that there are typical characteristics of the switching and abrupt change of the dominant factor in this segment. The rhythm response of the scoring structure at this stage has produced a nonlinear amplitude jump. The results show that the change in the scoring rhythm in this segment has exceeded the tolerance threshold of continuous scoring, and has the conditions for stage risk weight adjustment, which directly drives the stage risk change to produce amplitude correction behavior at this position.

[0123] Please see Figure 6 The section risk assessment module includes:

[0124] The scoring trajectory analysis submodule calculates the rate of change of scoring amplitude between consecutive sampling points based on the stage risk variation, determines the time interval of continuous change in the scoring trajectory, identifies fluctuation segments with continuous amplitude change trends and different rates, and obtains the scoring amplitude fluctuation segment.

[0125] Extract the continuous score point sequence obtained at 1-second intervals throughout the entire sampling period, and calculate the score difference between any two adjacent sampling points to obtain the score amplitude change rate sequence for each second. Then, perform a traversal analysis on this rate sequence to determine whether the score change within any consecutive time period shows a consistent trend. If the change rate direction of 5 or more consecutive points is consistent (i.e., all are positive or all are negative), then this segment is marked as a continuous change segment. For example, in the initial detection stage of a baking sample, the score value rises from 3.1 to 5.4 from the 10th to the 20th second, with single-point changes of +0.2, +0.3, +0.4, +0.5, +0.3, +0.2, +0.1, +0.2, +0.1, and +0.2, all of which are positive values, satisfying the continuous change condition, and is therefore determined to be a continuous change segment. This segment represents an upward trend. Within this segment, a rate difference determination is performed, calculating the rate change difference between adjacent rates. For example, the difference is +0.4 in the 3rd second and +0.5 in the 4th second, with a difference of +0.1. If this difference is greater than the preset fluctuation rate benchmark value of 0.12 or less than -0.12, the point is marked as a rate fluctuation inflection point. If there are at least two or more rate change inflection points in a certain segment, and their interval is less than 5 seconds, then the segment is further marked as a rate fluctuation segment. The fluctuation rate benchmark value is set to 0.12 based on the common fluctuation range of score changes in the sample, and is derived from the statistical results of the 85th percentile value of the historical data of the sample scores in a single cycle. All segments that meet the above conditions, with continuous score changes and significant rate fluctuations, are extracted as score amplitude fluctuation segments.

[0126] The risk segment identification submodule determines the start and end positions of each segment on the time axis of the scoring curve based on the score amplitude fluctuation segment, compares the period and trend of score fluctuation within each segment, identifies risk segments with continuous score changes, and obtains the stage score time series index.

[0127] The start and end times of each segment are read sequentially. The complete fluctuation cycle length and the difference between extreme scores within the cycle are calculated based on the scoring sequence within each segment. Segments with fluctuation cycle lengths less than 5 seconds or greater than 30 seconds are excluded. Normal segments with scoring change cycles between 5 and 30 seconds are retained and labeled as ascending, descending, or alternating types based on the direction of the scoring change trend. For example, in a fermented protein beverage detection sequence, the scoring fluctuation from second 35 to second 50 shows +0.3, +0.5, +0.2, -0.1, -0.3, -0.2, +0.2, +0.3, +0.4, -0.2, -0.1, -0.3, -0.5, +0.1, +0.3, which is a typical alternating fluctuation. If the number of reversals exceeds 3, it is marked as an alternating risk segment. Then, the score change trend in each risk segment is compared to see if it is consistent with the previous or next stage. If the previous and next segments are rising while the current segment is falling, it is recorded as an abnormal trend segment. If the trend continues in the same direction, it is marked as a trend continuation segment. Then, the time position of the maximum and minimum values ​​of each score within the fluctuation cycle is counted. If the distance between the two is less than 20% of the cycle length, the segment is marked as a rapid change segment. If it is greater than 60%, it is marked as a slow change segment. Through the above classification process, the score fluctuation segment is refined into three characteristic parameters: trend type, fluctuation rate, and change cycle. Each segment is assigned a segment number and a type label to form a complete stage score time series index.

[0128] The continuous interval division submodule analyzes the temporal connection relationship of each scoring segment based on the stage scoring time sequence index, adjusts the segment boundary method, compares the continuity of the scoring trajectory within each segment, constructs the combined distribution of the same type of scoring fluctuation segments, and obtains the continuous risk distribution interval.

[0129] The time boundaries between adjacent segments are extracted one by one, and a segment connection map is established. The difference between the end time of each pair of adjacent segments and the start time of the next segment is calculated. If the difference is less than or equal to 3 seconds, the two segments are merged into a continuous connected segment, and the new start and end time range and the average score trajectory after merging are recorded. If the difference is greater than 3 seconds but the scoring trends of the two segments are consistent, and the difference in the rate of change of scores is less than 0.1, it is judged as having a continuous trend but not temporal continuity. Logical merging is used to retain their respective boundaries, but they are classified into the same category in the analysis. The continuity of the score trajectory within each segment is then used as the core judgment parameter to analyze each... If there is a sudden jump between the end of a segment's scoring point and the starting value of the next segment's scoring point, and the jump is greater than the scoring threshold of 0.8, it is marked as a discontinuity point. This scoring jump threshold is calculated from the median of the maximum scoring jump amplitude in the training samples. If no scoring jump occurs in three or more consecutive segments, the trend type is consistent, and the difference in the amplitude of the scoring trajectory fluctuation is less than 0.3, then these three segments are constructed as a combination of scoring fluctuation segments of the same type, and the starting segment number is used as the combination sequence number. The average amplitude, average trend direction, and average cycle length of the combined scoring trajectory are all recorded in the combined segment parameter table to form a continuous risk distribution interval.

[0130] The above are merely preferred embodiments of the present invention and are not intended to limit the present invention in any other way. Any person skilled in the art may make changes or modifications to the above-disclosed technical content to create equivalent embodiments that can be applied to other fields. However, any simple modifications, equivalent changes, and modifications made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the protection scope of the present invention.

Claims

1. A food quality intelligent evaluation system, characterized in that, The system comprises: The gas path regulation module uses an electronic nose to perceive and analyze the flow path of the sample in the gas collection structure based on the food sample, adjusts the response position of the heat control area and the spectrum component, performs calibration through dynamic signals, and obtains dynamic channel response characteristics; The fluctuation recognition and aggregation module analyzes the time series signal changes based on the dynamic channel response characteristics, identifies the concentration trend and change area of the continuous sampling points, extracts and aggregates the time sequence characteristics of the disturbance section, and obtains disturbance trajectory aggregation characteristics; The main factor mapping weight module identifies the fluctuation difference of the key gas factor under each time structure based on the disturbance trajectory aggregation characteristics, adjusts the main factor weight distribution, dynamically maps the periodicity of the factor weight, and obtains the weight distribution mapping factor; The risk score output module analyzes the trend performance of the dominant factor at each stage based on the weight distribution mapping factor, identifies the switching change point, adjusts the risk score fluctuation, and obtains the stage risk variation amount; The section risk calibration module analyzes the score trajectory based on the stage risk variation amount, identifies the section with a continuous fluctuation amplitude change in the risk curve, judges the time period of each section and performs classification, and obtains the continuous risk distribution interval.

2. The food quality intelligent evaluation system according to claim 1, characterized in that, The dynamic channel response characteristics include channel flow rate distribution, heat zone temperature change, and spectrum signal characteristics. The disturbance trajectory aggregation characteristics include disturbance section identification, time sequence characteristic parameters, and signal change trajectory. The weight distribution mapping factor includes main factor weight distribution, time sequence structure mapping parameters, and factor influence classification. The stage risk variation amount includes stage risk level, risk change amplitude, and stage variation trend. The continuous risk distribution interval includes section risk classification, continuous interval time sequence, and interval identification number.

3. The food quality intelligent evaluation system according to claim 1, characterized in that, The gas path regulation module comprises: The gas path recognition submodule analyzes the flow path and concentration change trend of the sample released gas in each section of the collection structure based on the food sample, compares the flow rate change and gas distribution characteristics of each section, judges the diffusion and convergence trajectory of the gas in the flow channel, and obtains the channel flow rate distribution factor; The heat zone division adjustment submodule adjusts the spatial layout of the heat control unit based on the channel flow rate distribution factor, calculates the release area and release time sequence of the gas under each heat control state, compares the changes of the gas behavior under the temperature control condition, and obtains the heat zone temperature change sequence; The spectrum response calibration submodule analyzes the influence of each heat zone on the gas release characteristics based on the heat zone temperature change sequence, compares the signal change trend of each response channel of the spectrum component, judges the change of the signal response center and the channel bandwidth adjustment demand, optimizes the channel response parameter adjustment mode, and obtains the dynamic channel response characteristics.

4. The food quality intelligent evaluation system according to claim 1, characterized in that, The fluctuation recognition and aggregation module comprises: The signal trend recognition submodule calculates the change direction and change rate of the gas concentration of adjacent sampling points based on the dynamic channel response characteristics, judges the time interval of the continuous change in the signal sequence, identifies the continuous section with consistent change direction and rate variation, and obtains the continuous change trend interval; The disturbance section positioning submodule determines the position of signal rate mutation based on the continuous change trend interval, compares the change slope and rate change degree of adjacent sampling points, analyzes the fluctuation direction and duration period before and after the mutation point, determines the section boundary through the density of the mutation point distribution, and obtains a disturbance section identification index; The time sequence trajectory aggregation submodule calculates the start and end time, sampling frequency and fluctuation direction corresponding to the disturbance section based on the disturbance section identification index, judges the duration period and change intensity of each disturbance section, optimizes the section aggregation mode of common time structure and repeated fluctuation characteristics, and obtains a disturbance trajectory aggregation feature.

5. The food quality intelligent evaluation system according to claim 1, characterized in that, The main factor mapping weight module includes: The response interval extraction submodule analyzes the disturbance section identification and signal change trajectory based on the disturbance trajectory aggregation feature, compares the response change of the gas factor in the fixed and sliding time windows, judges which time period has a response amplitude higher than the average level of the same sequence, and obtains a response active interval set; The fluctuation difference judgment submodule compares the gas factor response in each time window based on the response active interval set, calculates the response amplitude and change rate of each section, judges whether the amplitude change in each time window forms a fluctuation, screens the dominant factor of the response difference in multiple windows, and obtains a window difference response sequence; The response weight adjustment submodule analyzes the average response amplitude and duration of the dominant factor under each time sequence feature based on the window difference response sequence, and obtains a weight distribution mapping factor of each dominant factor under each periodic structure.

6. The food quality intelligent evaluation system according to claim 1, characterized in that, The risk score output module includes: The factor performance analysis submodule analyzes the response characteristics in each stage based on the weight distribution mapping factor, compares the signal change direction and amplitude of the dominant factor in each stage, judges the response trend and fluctuation distribution of the dominant factor in each stage, and obtains a stage comparison factor group; The risk curve identification submodule identifies the continuous response characteristics of the dominant factor in time sequence based on the stage comparison factor group, analyzes the response direction change of the dominant factor between stages, judges the change trajectory of the dominant factor between adjacent stages, optimizes the identification standard of the switching interval of the dominant factor, and obtains a switching section sequence set; The scoring fluctuation adjustment submodule judges the change trend of the response trajectory of the dominant factor before and after switching based on the switching section sequence set, calculates the intensity and rhythm of the scoring fluctuation in the difference section, and uses the formula: ； obtaining a response amplitude of the score rhythm, obtaining a risk variation amount of the stage, wherein, a response amplitude of the score rhythm of the first segment, a response amplitude of the score rhythm of the first segment, a response amplitude of the score rhythm of the first segment, a response amplitude of the score rhythm of the first segment, a response amplitude of the score rhythm of the first segment, a response amplitude of the score rhythm of the first segment, a response amplitude of the score rhythm of the first 7. The food quality intelligent evaluation system according to claim 1, characterized in that, The section risk calibration module includes: The scoring trajectory analysis submodule calculates the scoring amplitude change rate between consecutive sampling points based on the stage risk variation amount, judges the time interval of continuous change in the scoring trajectory, identifies the fluctuation segment with continuous amplitude change trend and rate difference, and obtains a scoring amplitude fluctuation section; The risk section identification submodule judges the start and end position of each section on the scoring curve time axis based on the scoring amplitude fluctuation section, compares the period and change trend of the scoring fluctuation in each section, identifies the risk segment with continuous scoring variation characteristics, and obtains a stage scoring time sequence index; The continuous interval division sub-module analyzes the time sequence connection relationship of each score section based on the stage score time sequence index, adjusts the section boundary mode, compares the continuity of the score track in each section, constructs the combined distribution of the same type score fluctuation section, and obtains the continuous risk distribution interval.

8. The food quality intelligent evaluation system according to claim 1, characterized in that, The gas collection structure refers to a sealed transmission channel formed between the food sample and the detection cavity, the time sequence signal is a response data sequence continuously output by the sensor in a sampling period, and the continuous sampling points represent each data point recorded by the sensor at a fixed interval in the same sampling period.

Images