Safety management method and system applied to production and preparation of forsythia suspensa tea

Abstract

The present application relates to the technical field of safety management, in particular to a safety management method and system applied to the production and preparation of forsythia suspense tea, comprising: obtaining raw material image data of forsythia suspense tea and gas concentration data and temperature data of the frying process; based on the raw material image data, constructing a ratio feature vector of color saturation and brightness, performing vector matching on the ratio feature vector and a preset stem feature interval, and screening out a stem pixel set. The present application analyzes the ratio of color saturation and brightness of the raw material image, accurately screens the stem pixels and calculates the stem content characteristic value, and then maps it to the raw material quality influence coefficient. The coefficient is multiplied by the standard process risk threshold value to generate a dynamic threshold offset, so that the risk judgment threshold value can automatically decrease with the increase of the stem content in the raw material and appropriately increase with the decrease of the stem content, thereby improving the safety management precision of different batches of raw materials.

Description

Technical Field

[0001] This invention relates to the field of safety management technology, specifically to a safety management method and system applied to the production and preparation of Forsythia tea. Background Technology

[0002] Forsythia tea is a beverage with heat-clearing and detoxifying effects. The roasting process has a great impact on its quality. Roasting requires precise control of temperature and time to achieve the process goals of fixing, rolling and drying, while avoiding scorching or loss of aroma. Therefore, safety management is the key to ensuring product quality and production efficiency.

[0003] Currently, in the safety management of forsythia tea and similar tea processing, the mainstream method still mainly relies on experience-driven single-parameter control. The most common approach is to monitor the pot temperature based on contact temperature sensors and set a fixed alarm threshold. Once the temperature exceeds the preset value, the system triggers an audible and visual alarm or cuts off the heat source. In addition, some modern production lines have introduced simple image recognition technology to detect color changes in tea leaves to determine the degree of processing, but this is usually only used as an auxiliary reference and does not participate deeply in real-time control.

[0004] However, traditional methods use fixed temperature or gas thresholds for judgment, ignoring the significant impact of the stem content of the raw material on heat transfer. When the stem content of the raw material is high, the high thermal conductivity of the stem will lead to a surge in local thermal inertia. Even if the pot temperature does not reach the fixed threshold, the leaves may have already scorched. Conversely, when the stem content is low, excessive heat preservation may result in insufficient aroma. Summary of the Invention

[0005] To achieve the above objectives, the present invention provides the following technical solution: a safety management method applied to the production and preparation of Forsythia tea, comprising: Acquire raw material image data and gas concentration and temperature data during the roasting process of Forsythia tea; wherein, the gas concentration data includes at least the concentration values ​​of acetaldehyde, ethanol and ethyl acetate. Based on the raw material image data, a feature vector of the ratio of color saturation to brightness is constructed. The feature vector of the ratio is then matched with a preset stem feature interval to filter out the stem pixel set. The first area value of the pixel set of the stem and the second area value of the effective area of ​​the whole image are calculated, and a floating-point operation is performed to divide the two to obtain the stem content feature value. The stem content feature value is then mapped to the raw material quality influence coefficient. Based on gas concentration data, a feature vector of the concentration ratio of characteristic gas components is constructed; The process risk index is generated by multiplying the concentration ratio feature vector with the temperature data. The threshold for determining the process risk index is dynamically weighted and adjusted based on the raw material quality influence coefficient to obtain the target risk threshold. The process risk index is compared with the target risk threshold, and a graded control signal is generated based on the comparison result; the graded control signal is used to drive the actuator of the frying equipment.

[0006] Preferably, based on the raw material image data, a feature vector representing the ratio of color saturation to brightness is constructed. This feature vector is then matched with a preset stem feature interval to filter out a set of stem pixels, including: The raw material image data is converted to HSV color space, and the saturation channel value sequence and brightness channel value sequence of each pixel are extracted. Perform a pixel-by-pixel division operation between the saturation channel value sequence and the brightness channel value sequence to generate a one-dimensional ratio feature vector; Retrieve the stalk feature interval constructed based on physical optical properties, compare each element in the ratio feature vector with the upper and lower boundary values ​​of the stalk feature interval one by one, and generate a Boolean filtering mask. Boolean filtering masks are used to index and extract the raw material image data. The coordinates of pixels with true mask values ​​are retained, and pixels with false mask values ​​are removed to obtain the root pixel set.

[0007] Preferably, the first area value of the set of pixels containing the spurious part and the second area value of the effective region of the entire image are statistically analyzed, and a floating-point operation is performed to divide the two to obtain the spurious part content feature value, including: Extract the pixel coordinates from the pixel set of the stem, and accumulate the pixel coordinates to obtain the first area value representing the physical space occupied by the stem; The total number of pixels in the effective area of ​​the raw material image data is counted to obtain a second area value that represents the overall scale of the raw material. The first area value and the second area value are called, and a floating-point division operation is performed to generate a normalized stem content feature value; wherein, the stem content feature value represents the area ratio of stems in the raw material.

[0008] Preferably, mapping the stem content characteristic value to the raw material quality influence coefficient includes: Based on the physical difference between the high thermal conductivity of tea stem fibers and the high thermal resistance of leaf cells, a curve showing the relationship between stem content and thermal resistance was constructed. Input the stem content characteristic value as the horizontal axis into the stem content-thermal resistance interference relationship curve, and query the corresponding vertical axis value to obtain the raw material quality influence coefficient. Among them, the slope of the stem content-thermal resistance interference relationship curve increases monotonically with the increase of stem content, which is used to characterize the increase in heat transfer efficiency during stir-frying caused by the increase of stem content.

[0009] Preferably, based on gas concentration data, a feature vector of the concentration ratio of characteristic gas components is constructed, including: Based on the acetaldehyde concentration value and ethanol concentration value in the gas concentration data, the ratio of the two is calculated to generate the main characteristic ratio characterizing the coking risk. Using the principal feature ratio as the first element of the vector and the ethyl acetate concentration value in the gas concentration data as the second element of the vector, a two-dimensional concentration ratio feature vector is constructed; whereby the concentration ratio feature vector is used to characterize the caramelization ratio state of the current roasting stage.

[0010] Preferably, the concentration ratio feature vector is multiplied with temperature data to generate a process risk index, including: Extract the principal feature ratio from the concentration ratio feature vector, multiply it with the real-time collected pot temperature data, and generate a thermo-chemical reaction factor characterizing the intensity of oxidative coking. The ethyl acetate concentration value is extracted from the concentration ratio feature vector and multiplied with the pot temperature data to generate a heat-mass transfer factor characterizing the volatilization rate of aroma components. Based on the current technological focus of the roasting stage, risk weight coefficients are assigned to the thermo-chemical reaction factor and the heat-mass transfer factor respectively. The comprehensive process risk index is generated by multiplying the thermo-chemical reaction factor and the heat-mass transfer factor by their corresponding risk weight coefficients and then summing the results.

[0011] Preferably, the threshold for determining the process risk index is dynamically weighted and adjusted based on the raw material quality influence coefficient to obtain the target risk threshold, including: Retrieve the pre-stored standard process risk threshold, where the standard threshold is a fixed safety boundary without considering raw material fluctuations; The raw material quality influence coefficient is multiplied with the standard process risk threshold to generate a threshold offset that characterizes the degree of thermal inertia interference of the raw material. Based on the current operating time signal of the roasting equipment and the pre-stored stage-sensitivity mapping table, the threshold offset is weighted and corrected to generate a dynamic correction factor. The target risk threshold is obtained by subtracting the dynamic correction factor from the standard process risk threshold.

[0012] Preferably, based on the current operating time signal of the roasting equipment and a pre-stored stage-sensitivity mapping table, the threshold offset is weighted and corrected to generate a dynamic correction factor, including: The running time counter of the frying equipment is read, and based on the inherent physical timing characteristics of the forsythia tea frying process, the running time is rigidly divided into the fixation stage, the rolling stage, and the drying stage through a hardware comparator. Based on the current process stage, retrieve the pre-stored stage-sensitivity coefficient table to obtain the stage sensitivity coefficient; The stage sensitivity coefficient is multiplied by the threshold offset to generate a dynamic correction factor.

[0013] Preferably, the process risk index is compared with the target risk threshold, and a graded control signal is generated based on the comparison result, including: The process risk index is subtracted from the target risk threshold to obtain the real-time risk difference. The real-time risk difference is used to quantify the degree to which the current process state deviates from the safety boundary. The larger the difference, the higher the risk. Based on the magnitude of the real-time risk difference and the heat accumulation characteristics of the forsythia tea frying process, the risk status is rigidly divided into three levels: safe steady state, warning critical state, and dangerous runaway state. The dangerous runaway state is determined by the risk difference exceeding the preset heat accumulation rate limit, which is used to identify the risk of sudden scorching. Match the corresponding control strategy for the implementing agency based on the current risk status range.

[0014] A safety management system applied to the production and preparation of Forsythia suspensa tea, which is applicable to the aforementioned safety management methods applied to the production and preparation of Forsythia suspensa tea, including: The data acquisition module is used to acquire raw material image data of Forsythia tea and gas concentration data and temperature data during the frying process; wherein, the gas concentration data includes at least the concentration values ​​of acetaldehyde, ethanol and ethyl acetate. The image processing module is used to construct a feature vector of the ratio of color saturation to brightness based on the raw material image data, and to perform vector matching between the ratio feature vector and the preset stem feature interval to filter out the stem pixel set. The proportion quantization module is used to calculate the first area value of the pixel set of the stem and the second area value of the effective area of ​​the whole image, perform a floating-point operation to divide the two to obtain the stem content feature value, and map the stem content feature value to the raw material quality influence coefficient. The feature construction module is used to construct feature vectors of the concentration ratios of characteristic gas components based on gas concentration data; The risk calculation module is used to multiply the concentration ratio feature vector with temperature data to generate a process risk index. The threshold weighting module is used to dynamically adjust the judgment threshold of the process risk index based on the raw material quality influence coefficient to obtain the target risk threshold. The control output module is used to compare the process risk index with the target risk threshold and generate a graded control signal based on the comparison result; the graded control signal is used to drive the actuator of the frying equipment.

[0015] Compared with the prior art, the beneficial effects of the present invention are: This invention analyzes the ratio of color saturation to brightness in raw material images to accurately screen for stem pixels and calculate the stem content characteristic value, which is then mapped to a raw material quality influence coefficient. This coefficient is multiplied by a standard process risk threshold to generate a dynamic threshold offset. This allows the risk assessment threshold to automatically decrease as the stem content in the raw material increases and appropriately increase as the stem content decreases. This solves the problem of missed or false judgments caused by the inability of traditional fixed thresholds to adapt to the heterogeneity of raw materials, thus improving the accuracy of safety management for different batches of raw materials. This invention constructs a feature vector from the acetaldehyde / ethanol concentration ratio and the ethyl acetate concentration, and multiplies them with the real-time pot temperature to generate a thermo-chemical reaction factor and a heat-mass transfer factor. Finally, a weighted sum is obtained to obtain a process risk index, thereby upgrading single gas concentration monitoring to coupled analysis of temperature and concentration. This can quantify the severity of oxidation and coking and the rate of aroma volatilization, effectively identifying the potential scorching risk of low temperature and high concentration and the aroma loss risk of high temperature and low concentration during the management process. This overcomes the shortcomings of traditional single parameter monitoring, which cannot reflect the chemical reaction kinetics process. Based on the running time of the frying equipment, this invention rigidly divides the process into three stages: blanching, rolling, and drying. Different sensitivity adjustment factors are matched according to the risk sensitivity of each stage, and the threshold deviation is weighted and corrected. The system sensitivity is increased in the high-risk blanching stage, and a slight increase in the risk index can trigger a suppression signal. The sensitivity is reduced in the low-risk drying stage to avoid frequent start-ups and shutdowns of the equipment due to small fluctuations, thereby improving the stability and execution efficiency of the control system. Attached Figure Description

[0016] Figure 1 This is a schematic flowchart of the overall method in one embodiment of the present invention; Figure 2 This is a schematic diagram of the overall system architecture in one embodiment of the present invention.

[0017] In the diagram: 1. Data acquisition module; 2. Image processing module; 3. Proportion quantification module; 4. Feature construction module; 5. Risk calculation module; 6. Threshold weighting module; 7. Control output module. Detailed Implementation

[0018] The technical solutions of 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.

[0019] Example 1, please refer to Figure 1 This invention provides a technical solution: a safety management method applied to the production and preparation of Forsythia tea, comprising: S1. Obtain raw material image data of Forsythia tea and gas concentration and temperature data during the frying process; wherein, the gas concentration data includes at least the acetaldehyde concentration value, ethanol concentration value and ethyl acetate concentration value; S2. Based on the raw material image data, construct a feature vector of the ratio of color saturation to brightness, and perform vector matching between the ratio feature vector and the preset stem feature interval to filter out the stem pixel set. S3. Calculate the first area value of the pixel set of the stem and the second area value of the effective area of ​​the whole image, perform a floating-point operation to divide the two to obtain the stem content feature value, and map the stem content feature value to the raw material quality influence coefficient. S4. Based on gas concentration data, construct a feature vector of the concentration ratio of characteristic gas components; S5. Multiply the concentration ratio feature vector with the temperature data to generate the process risk index; S6. The threshold for determining the process risk index is dynamically weighted and adjusted based on the raw material quality influence coefficient to obtain the target risk threshold. S7. Compare the process risk index with the target risk threshold, and generate a graded control signal based on the comparison result; wherein, the graded control signal is used to drive the actuator of the frying equipment.

[0020] It should be noted that raw material image data refers to digital image information obtained by photographing fresh forsythia tea leaves using image acquisition equipment. This data includes the color, texture, and morphological characteristics of the raw materials and is usually stored in the form of a pixel matrix. For example, before the roasting process begins, an industrial camera installed above the feed inlet photographs the fresh forsythia leaves on the conveyor belt, obtaining a bitmap image containing three channels: red, green, and blue. This image is the raw material image data. The color saturation to brightness ratio feature vector refers to a single-dimensional numerical sequence obtained by converting the raw material image data from the red-green-blue color space to the hue-saturation-brightness color space, extracting the saturation and brightness components of each pixel, and dividing the two. For example, for a pixel in the image, its saturation value is 120 and its brightness value is 200, and dividing the two gives 0.6. Repeating this operation for all pixels in the image finally yields a sequence of hundreds of thousands of decimals between 0.3 and 0.9, which is the ratio feature vector. The stem feature range refers to the range of saturation to brightness ratio values ​​that are pre-measured and stored based on the differences in physical and optical properties between the stems and leaves of Forsythia tea. This range is used to filter out pixels belonging to the stems from the ratio feature vector. For example, experimental measurements have shown that the saturation to brightness ratio of Forsythia tea stems is usually concentrated between 0.65 and 0.85, while the ratio of leaves is usually less than 0.5. Therefore, 0.65 to 0.85 is set as the stem feature range. The stem pixel set refers to the set of pixel coordinates retained after filtering the raw material image data using the stem feature range. The stem content feature value refers to the normalized value obtained by performing a floating-point division operation between the first area value and the second area value, used to characterize the area ratio of stems in the raw material; for example, dividing 3000 by 10000 yields 0.3, which is the stem content feature value, indicating that the stems account for 30% of the raw material in the current batch; the raw material quality influence coefficient refers to the correction coefficient obtained by querying the preset stem content-thermal resistance interference relationship curve based on the stem content feature value, used to characterize the degree of influence of the thermal conductivity of the raw material on the frying process; the characteristic gas component concentration ratio feature vector refers to a two-dimensional numerical vector constructed by selecting the concentration values ​​of specific gases generated during the frying process and performing ratio calculations or direct extraction, used to characterize the current chemical reaction state; the process risk index refers to the comprehensive value obtained by multiplying the elements in the concentration ratio feature vector with the real-time pot temperature data, assigning different weights, and then summing them, used to quantify the degree of danger in the current frying process; the graded control signal refers to the set of instructions generated based on the comparison results of the process risk index and the target risk threshold, used to drive the action of the frying equipment actuator.

[0021] In an optional embodiment, based on the raw material image data, a feature vector representing the ratio of color saturation to brightness is constructed. This feature vector is then matched with a preset stem feature interval to filter out a set of stem pixels, including: The raw material image data is converted to HSV color space, and the saturation channel value sequence and brightness channel value sequence of each pixel are extracted. It's important to note that HSV color space conversion refers to the image processing procedure of converting image data in the red, green, and blue color space into three channel components: hue, saturation, and lightness. For example, a 1920x1080 RGB image of forsythia tea raw materials, after color space conversion, will generate three independent grayscale images: one for the saturation channel and another for the lightness channel. Each pixel in the image corresponds to a value from 0 to 255. The saturation and lightness channel value sequences refer to the sets of saturation and lightness values ​​extracted from all pixels in the HSV color space converted image, respectively. These values ​​are arranged in a matrix, forming two one-dimensional numerical sequences. For example, for a small image containing 100 pixels, extracting the saturation channel yields a sequence of 100 values, such as 120, 130, and 145; extracting the lightness channel yields another sequence of 100 values, such as 200, 195, and 180. Perform a pixel-by-pixel division operation between the saturation channel value sequence and the brightness channel value sequence to generate a one-dimensional ratio feature vector; It's important to note that pixel-wise division refers to the process of dividing the saturation channel value sequence by two values ​​at the same position in the brightness channel value sequence. This operation is performed independently for each pixel in the image, generating a new numerical sequence. For example, the first pixel in the top left corner of the image has a saturation value of 120 and a brightness value of 200, so dividing them gives 0.6; the second pixel has a saturation value of 150 and a brightness value of 180, so dividing them gives approximately 0.83. Repeating this operation for all pixels in the entire image results in a new sequence consisting of hundreds of thousands of decimals. A single-dimensional ratio feature vector is a sequence of data generated through pixel-wise division, containing only a single numerical dimension. Each element in this vector represents the ratio of the saturation to the brightness of the corresponding pixel, used for subsequent feature matching. For example, after division, the original two-dimensional image data is transformed into a one-dimensional list of values ​​containing the ratio results of all pixels in the image, such as 0.6, 0.83, 0.45, etc. This list is the ratio feature vector. Retrieve the stalk feature interval constructed based on physical optical properties, compare each element in the ratio feature vector with the upper and lower boundary values ​​of the stalk feature interval one by one, and generate a Boolean filtering mask. It should be noted that the stem feature interval constructed based on physical optical properties refers to the pre-measured and stored range of saturation to brightness ratios, utilizing the physical differences in light reflection and absorption characteristics between the stems and leaves of Forsythia tea. Since the stems are typically darker in color and have a coarser texture, their ratios tend to concentrate within a specific range. For example, through measurements of numerous samples, it was found that the saturation to brightness ratio of Forsythia tea stems is mainly distributed between 0.65 and 0.85, while the ratio for leaves is usually less than 0.5. Therefore, the closed interval of 0.65 to 0.85 is set as the stem feature interval. The step-by-step comparison refers to the logical judgment process of comparing each element in the ratio feature vector individually with the lower and upper limits of the stem feature interval. This process checks sequentially... The function checks whether each value in the vector falls within a preset interval. For example, for the first value 0.7 in the ratio feature vector, comparing it with 0.65 reveals it is greater than the lower limit, and comparing it with 0.85 reveals it is less than the upper limit, thus it is determined to fall within the interval. For the value 0.4, it is determined to be less than the lower limit and not fall within the interval. The Boolean filtering mask is a judgment matrix composed of logical true and logical false values ​​generated based on the results of each comparison. The size of this mask is consistent with the number of pixels in the original image and is used to mark whether each pixel belongs to the stem. For example, after comparison, the position corresponding to the pixel belonging to the stem is marked as logical true, and the position corresponding to the pixel not belonging to the stem is marked as logical false, ultimately generating a sequence composed of true or false values ​​with the same number of pixels as the image.

[0022] Boolean filtering masks are used to index and extract the raw material image data. The coordinates of pixels with true mask values ​​are retained, and pixels with false mask values ​​are removed to obtain the root pixel set. It should be noted that index extraction refers to the process of using a Boolean filter mask as an index to filter out the coordinates of specific pixels from the original image data. This operation determines whether to retain or discard pixels based on the logical value of the mask. The coordinates of pixels with a true mask value refer to the original image pixel position information corresponding to the elements marked as logically true in the Boolean filter mask. These coordinates constitute the specific location set of the pixel in the image. For example, in the mask sequence, if the 100th element is true, the 101st element is false, and the 102nd element is true, then the row and column coordinates of the pixels corresponding to the 100th and 102nd elements in the image are extracted. Pixels with a false mask value refer to... In a Boolean filtering mask, the original image pixel position information corresponding to elements marked as logically false will be discarded by the system and will not participate in subsequent calculations. For example, if all elements in the mask corresponding to the leaf position are false, the system will ignore these coordinates during extraction and only retain the stalk coordinates marked as true. The stalk pixel set refers to the sum of all pixel coordinates that are finally retained through the index extraction operation. This set only contains pixels that are determined to be stalks and is used to calculate the stalk content. For example, in a 10,000-pixel image, after filtering, 3,000 pixels have a mask value of true, and the coordinates of these 3,000 pixels together form the stalk pixel set.

[0023] In an optional embodiment, the first area value of the set of pixels containing the lenticular portion and the second area value of the effective region of the entire image are statistically analyzed, and a floating-point operation is performed to divide the two to obtain the lenticular portion feature value, including: Extract the pixel coordinates from the pixel set of the stem, and accumulate the pixel coordinates to obtain the first area value representing the physical space occupied by the stem; It should be noted that the first area value refers to the total number of pixels contained in the root pixel set. This value directly represents the total number of pixel units occupied by the root in the digital image. For example, when the system identifies the root pixel set, it traverses all the coordinate points in the set and counts that a total of 3,000 pixels are marked as root, then the first area value is 3,000. The total number of pixels in the effective area of ​​the raw material image data is counted to obtain a second area value that represents the overall scale of the raw material. It should be noted that the second area value refers to the total number of all pixels within the effective detection area in the raw material image. The effective area is determined by a pre-calibrated image mask template, which covers the expected tea leaf dropping area on the conveyor belt, thereby removing black background interference at the image edges. Alternatively, the effective area can also be obtained by segmenting the raw material image using a background subtraction algorithm based on a color threshold, i.e., setting a green component threshold and retaining only pixels that match the color characteristics of tea leaves. For example, in an image with a resolution of 1920 x 1080, if the black background at the edges is removed through preprocessing, and only 10,000 pixels containing tea leaves in the middle are retained as the analysis object, then the second area value is 10,000. The first area value and the second area value are called, and a floating-point division operation is performed to generate a normalized stem content feature value; wherein, the stem content feature value represents the area ratio of stems in the raw material; It should be noted that floating-point division refers to a division method that retains the decimal part during the calculation process to ensure the accuracy of the result; normalization refers to mapping the calculation result to the interval between 0 and 1 so that it can intuitively reflect the proportional relationship; for example, dividing the first area value of 3000 by the second area value of 10000, after performing a floating-point operation, yields 0.3. This 0.3 is the stem content feature value, indicating that the stem occupies 30% of the area in the current raw material image.

[0024] In an optional embodiment, mapping the stem content characteristic value to a raw material quality influence coefficient includes: Based on the physical difference between the high thermal conductivity of tea stem fibers and the high thermal resistance of leaf cells, a curve showing the relationship between stem content and thermal resistance was constructed. It should be noted that the stem content-thermal resistance interference curve refers to a pre-determined and stored curve used to describe the relationship between the stem content in the raw material and the degree of heat transfer resistance. This curve is generated through polynomial fitting. Specifically, under experimental conditions, the thermal conductivity of Forsythia tea samples with stem contents of 10%, 20%, 30%, 40%, 50%, and 60% is measured under a constant heat source. The stem content is used as the abscissa, and the thermal conductivity is used as the ordinate. A curve with a slope that varies with stem content is obtained by fitting using the least squares method. The equation of the curve that monotonically increases with the increase of stem content is stored. The shape of this curve is that as the value of the horizontal axis increases, the value of the vertical axis increases at an increasingly faster rate. For example, in a laboratory environment, the heating rate of Forsythia tea raw materials with stem content of 20%, 30%, and 40% is measured under the same heat source. It is found that for every 10% increase in stem content, the increase in heat transfer efficiency is not a fixed value, but gradually increases. Connecting these measurement points into a line yields a curve with a gradually steepening slope. This curve is the stem content-thermal resistance interference relationship curve. Input the stem content characteristic value as the horizontal axis into the stem content-thermal resistance interference relationship curve, and query the corresponding vertical axis value to obtain the raw material quality influence coefficient. It should be noted that the raw material quality impact coefficient refers to the corresponding value found on the preset relationship curve based on the current stem content characteristic value of the raw material. This value is used to quantify the degree of influence of the physical structure of the raw material itself on the heat conduction speed during the frying process. The larger the coefficient value, the faster the internal heat transfer of the raw material, and the more likely local overheating will occur. For example, if the calculated stem content characteristic value is 30%, and it is used as the abscissa to be vertically projected on the relationship curve, the corresponding point on the curve is found, and the ordinate value of the point is read as 1.3. Then the raw material quality impact coefficient is 1.3. This means that compared with the standard raw material, the heat transfer efficiency of the current batch of raw material has increased by 30%, and subsequent risk assessment needs to be tightened based on this coefficient. Among them, the slope of the stem content-thermal resistance interference relationship curve increases monotonically with the increase of stem content, which is used to characterize the increase in heat transfer efficiency during stir-frying caused by the increase of stem content. It should be noted that the monotonically increasing slope means that the steepness of the curve continuously increases with the increase of the stem content, reflecting that the higher the stem content, the more significant the reduction effect on the overall thermal resistance. That is, the increase in thermal efficiency brought about by the increase of a unit of stem content is more obvious in the high stem content stage than in the low stem content stage. For example, in the range where the stem content increases from 10% to 20%, the thermal resistance interference coefficient may only increase by 0.1; while in the range where the stem content increases from 40% to 50%, the same 10% increment may lead to an increase of 0.3 in the thermal resistance interference coefficient. This continuously increasing rate of increase intuitively reflects the dominant role of the high thermal conductivity of the stem fiber on the overall thermal field at high concentrations.

[0025] In an optional embodiment, based on gas concentration data, a feature vector of concentration ratios of characteristic gas components is constructed, including: Based on the acetaldehyde and ethanol concentration values ​​in the gas concentration data, a ratio calculation is performed to divide the two values ​​to generate a main characteristic ratio characterizing the risk of coking. Acetaldehyde is an intermediate product of the thermal degradation of sugars, and ethanol is a derivative of cyanobacterium. The ratio of the two increases with the increase of the roasting temperature and is used to quantify the current degree of oxidative coking. It should be noted that the principal characteristic ratio is the quotient obtained by dividing the acetaldehyde concentration by the ethanol concentration. This quotient is used to quantify the tendency of tea leaves to undergo non-enzymatic browning and produce charred products under heat. For example, when the gas sensor detects an acetaldehyde concentration of 10 ppm and an ethanol concentration of 50 ppm, dividing 10 by 50 yields 0.2. This 0.2 is the principal characteristic ratio. The smaller the value, the milder the chemical reaction; the larger the value, the closer the tea leaves are to a charred state. Using the principal feature ratio as the first element of the vector and the ethyl acetate concentration value in the gas concentration data as the second element of the vector, a two-dimensional concentration ratio feature vector is constructed; whereby the concentration ratio feature vector is used to characterize the caramelization ratio state of the current roasting stage. It should be noted that the concentration ratio feature vector refers to a set of data composed of two independent numerical parameters in a fixed order. The first element carries information on the risk of chemical reaction, and the second element carries information on the retention of aroma substances. This two-dimensional structure can simultaneously monitor negative and positive reactions during the roasting process, avoiding misjudgments caused by a single indicator. For example, if the calculated principal feature ratio is 0.2 and the detected ethyl acetate concentration is 5 ppm, then these two values ​​are combined into an ordered pair, with the first digit being 0.2 and the second digit being 5. This ordered pair is the concentration ratio feature vector. The system can use this vector to know that the current roasting state is characterized by low charring risk and moderate aroma volatilization, thereby maintaining the existing heating power.

[0026] In an optional embodiment, the concentration ratio feature vector is multiplied with temperature data to generate a process risk index, including: Extract the principal feature ratio from the concentration ratio feature vector, multiply it with the real-time collected pot temperature data, and generate a thermo-chemical reaction factor characterizing the intensity of oxidative coking. It should be noted that the thermo-chemical reaction factor is the product obtained by directly multiplying the principal characteristic ratio, which represents the risk of coking, by the pot temperature, which represents the thermal intensity. This value is used to quantify the intensity of chemical reactions under high-temperature conditions. Temperature amplifies the rate of chemical reactions, making this factor more accurate in reflecting the potential risk of charring than a single concentration. For example, if the principal characteristic ratio is 0.2 and the real-time pot temperature is 180 degrees Celsius, multiplying 0.2 by 180 yields 36, which is the thermo-chemical reaction factor. If the temperature rises to 200 degrees Celsius, the ratio remains the same, but the factor becomes 40. The increase in value directly reflects the increase in the risk of coking. The ethyl acetate concentration value is extracted from the concentration ratio feature vector and multiplied with the pot temperature data to generate a heat-mass transfer factor characterizing the volatilization rate of aroma components. It should be noted that the heat-mass transfer factor is the product obtained by directly multiplying the ethyl acetate value representing the aroma concentration by the pot temperature. This value is used to quantify the driving force of heat energy on the volatilization of aroma substances. High temperatures accelerate the dissipation of aroma components, and this factor can characterize the difficulty of aroma preservation at the current temperature. For example, if the ethyl acetate concentration is 5 ppm and the pot temperature is 180 degrees Celsius, multiplying 5 by 180 gives 900, which is the heat-mass transfer factor. If the temperature is too high and the aroma is lost too quickly, this factor will increase significantly, indicating that the temperature needs to be lowered to preserve the aroma. Based on the current technological focus of the roasting stage, risk weight coefficients are assigned to the thermo-chemical reaction factor and the heat-mass transfer factor respectively. It should be noted that the risk weighting coefficient refers to a pre-set numerical ratio based on the different safety and quality requirements at different stages of the roasting process. This coefficient is used to adjust the proportion of the two factors in the final risk assessment to ensure that the assessment results meet the current process objectives. For example, in the blanching stage, preventing scorching is the primary task, so the weighting coefficient of the thermo-chemical reaction factor is set to 0.6, and the weighting coefficient of the heat-mass transfer factor is set to 0.4. In the drying stage, in order to enhance the aroma, the weighting of the heat-mass transfer factor may be adjusted to 0.6, and the weighting of the thermo-chemical reaction factor may be adjusted to 0.4. The comprehensive process risk index is generated by multiplying the thermo-chemical reaction factor and the heat-mass transfer factor by their corresponding risk weight coefficients and then summing the results. It should be noted that the process risk index is the final value obtained by adding the weighted thermo-chemical reaction factor and the weighted heat-mass transfer factor. This index is a comprehensive indicator that covers both coking risk and aroma loss risk. For example, assuming the thermo-chemical reaction factor is 36 with a weighting coefficient of 0.6, it is calculated to be 21.6; the heat-mass transfer factor is 900 with a weighting coefficient of 0.4, it is calculated to be 360. Adding 21.6 and 360 gives 381.6, which is the process risk index. The system uses the magnitude of this value to determine whether the current process is within a safe range.

[0027] In an optional embodiment, the threshold for determining the process risk index is dynamically weighted and adjusted based on the raw material quality influence coefficient to obtain the target risk threshold, including: Retrieve the pre-stored standard process risk threshold, where the standard threshold is a fixed safety boundary without considering raw material fluctuations; It should be noted that the standard process risk threshold refers to the upper limit of the process risk index obtained through historical data statistics under ideal raw material conditions. This value serves as the system's default baseline and does not change with changes in raw material characteristics. For example, before the system leaves the factory, it conducts a large number of standard raw material experiments and sets the process risk index to be dangerous if it exceeds 300. This 300 is the standard process risk threshold. Regardless of the current stem content of the raw material, the system defaults to using this value as the initial judgment benchmark. The raw material quality influence coefficient is multiplied with the standard process risk threshold to generate a threshold offset that characterizes the degree of thermal inertia interference of the raw material. It should be noted that the threshold offset is the value obtained by multiplying the raw material quality influence coefficient by the standard process risk threshold. It is used to quantify the degree of interference of the raw material's physical properties with the standard safety boundary. This value represents the threshold adjustment range required due to changes in the thermal conductivity of the raw material. For example, if the raw material quality influence coefficient is 0.2 and the standard process risk threshold is 300, multiplying the two gives 60. This 60 is the threshold offset, indicating that due to the high stem content leading to accelerated heat transfer, the safety boundary needs to be adjusted by 60 units relative to the standard value. Based on the current operating time signal of the roasting equipment and the pre-stored stage-sensitivity mapping table, the threshold offset is weighted and corrected to generate a dynamic correction factor. It should be noted that the dynamic correction factor is a value obtained by weighting the threshold offset based on the sensitivity of the current process stage. It is used to reflect the difference in risk tolerance at different roasting stages. For example, in the high-risk blanching stage, the system needs more sensitive control, so the threshold offset is multiplied by a coefficient greater than 1 to amplify it. In the low-risk drying stage, it is multiplied by a coefficient less than 1 to reduce it. The final value obtained after stage weighting is the dynamic correction factor. The target risk threshold is obtained by subtracting the dynamic correction factor from the standard process risk threshold. It should be noted that the target risk threshold refers to the final result of adding or subtracting the standard process risk threshold and the dynamic correction factor. This value is the real-time boundary used to determine whether the process risk index exceeds the standard under the current operating conditions. For example, when the standard process risk threshold is 300 and the dynamic correction factor is 90, the difference between the two is 210. This 210 is the target risk threshold, which means that under the current raw material and process stage, an alarm will be triggered once the process risk index exceeds 210.

[0028] In an optional embodiment, based on the current operating time signal of the frying equipment and a pre-stored stage-sensitivity mapping table, a weighted correction is performed on the threshold offset to generate a dynamic correction factor, including: The running time counter of the frying equipment is read, and based on the inherent physical timing characteristics of the forsythia tea frying process, the running time is rigidly divided into the fixation stage, the rolling stage, and the drying stage through a hardware comparator. It should be noted that rigid division refers to the operation of forcibly dividing the continuous frying process into different stages according to pre-set time nodes. Each stage has an independent control strategy, and the switching between stages does not rely on fuzzy judgment, but is based on precise time counting. For example, after the frying equipment is started, the first 5 minutes are defined as the blanching stage, the 5th to 15th minutes are the kneading stage, and the 15th minute onwards is the drying stage. The system directly determines which specific time period it is in by reading the timer value. Based on the current process stage, retrieve the pre-stored stage-sensitivity coefficient table to obtain the stage sensitivity coefficient; It should be noted that the stage-sensitivity coefficient table is a set of parameter lookup tables pre-stored in the control system's memory. These tables define the sensitivity of different time periods during the forsythia tea processing to process risks. The table is indexed by the name of the process stage and contains the corresponding sensitivity value. The value represents the intensity of the stage's response to changes in risk. For example, in the forsythia tea processing, the fixation stage requires rapid heating to deactivate enzymes, making it extremely sensitive to temperature and gas changes; therefore, a high sensitivity value, such as 1.5, is set. The rolling stage focuses on shaping, with moderate sensitivity, and is set to 1.0. The drying stage mainly involves moisture evaporation, allowing for some fluctuations, and has lower sensitivity, set to 0.8. The system establishes a mapping relationship between these values ​​and the corresponding time periods (e.g., 0 to 5 minutes, 5 to 15 minutes, and after 15 minutes) and stores this information, thus forming the stage-sensitivity coefficient table. When the equipment reaches the 3rd minute, the system identifies it as the fixation stage and automatically looks up the value 1.5 from the table as the stage sensitivity coefficient. The stage sensitivity coefficient and the threshold offset are multiplied to generate a dynamic correction factor; It should be noted that the multiplication operation here refers to scaling the coefficient representing the stage sensitivity with the threshold offset representing the degree of raw material interference to obtain a correction amount that meets the control requirements of the current stage. For example, if the threshold offset is 60 and the stage sensitivity coefficient found in the current blanching stage is 1.5, multiplying 60 by 1.5 gives 90, which is the dynamic correction factor used to adjust the standard threshold in the future.

[0029] In an optional embodiment, the process risk index is compared with a target risk threshold, and a graded control signal is generated based on the comparison result, including: The process risk index is subtracted from the target risk threshold to obtain the real-time risk difference. The real-time risk difference is used to quantify the degree to which the current process state deviates from the safety boundary. The larger the difference, the higher the risk. It should be noted that the real-time risk difference is the result obtained by subtracting the target risk threshold from the specific value of the process risk index. This difference can intuitively reflect the remaining space between the current risk and the safety red line. When the result is negative, it means that it is within the safety line, and when it is positive, it means that it has exceeded the boundary. For example, if the calculated process risk index is 280 and the current target risk threshold is 240, subtracting 240 from 280 gives 40. This 40 is the real-time risk difference. A positive number indicates that the current risk has exceeded the safety boundary. The larger the value, the more it has exceeded the boundary, and the more urgent the risk of scorching. Based on the magnitude of the real-time risk difference and the heat accumulation characteristics of the forsythia tea frying process, the risk status is rigidly divided into three levels: safe steady state, warning critical state, and dangerous runaway state. The dangerous runaway state is determined by the risk difference exceeding the preset heat accumulation rate limit, which is used to identify the risk of sudden scorching. It should be noted that the three-level intervals refer to three consecutive numerical ranges artificially set based on the magnitude of the real-time risk difference, corresponding to three different safety levels in the process. A safe steady state represents complete controllability, a warning threshold represents approaching the edge of loss of control, and a dangerous runaway represents irreversible quality damage caused by excessively rapid heat accumulation. The heat accumulation rate limit is a constant pre-calculated based on the specific heat capacity and mass of Forsythia tea. The specific value can be determined through small-scale experiments. For example, the preset heat accumulation rate limit corresponds to a risk difference of 100. For example, the system presets a safe steady state interval where the risk difference is less than 50, a warning threshold interval where the risk difference is between 50 and 100, and a dangerous runaway interval where the risk difference is greater than 100. When the real-time risk difference is 30, the system determines it to be in a safe steady state; when the difference rises to 70, the system determines it to be entering the warning threshold; if the difference instantly spikes to 150, exceeding the heat accumulation rate limit, the system determines it to be in dangerous runaway. Match the corresponding control strategy of the implementing agency based on the current risk status range; It should be noted that the actuator control strategy refers to a pre-written set of equipment action instructions for different risk levels. The instructions include specific operations such as adjusting the heating power, changing the exhaust frequency, or triggering emergency unloading. The 20% reduction in heating tube power specifically refers to reducing the output duty cycle of the PID controller by 20 percentage points, and the high-speed operation of the exhaust fan specifically refers to adjusting the fan drive voltage to 100% of the rated voltage. For example, when the system determines that it is in a safe steady-state range, a maintenance strategy is matched, instructing the heating tube to maintain the current power and the exhaust fan to run at low speed. When it is determined that it has entered the warning critical range, an intervention strategy is matched, instructing the heating tube power to be reduced by 20% and the exhaust fan to run at high speed. When it is determined that it is in a dangerous out-of-control range, a stop-loss strategy is matched, instructing the immediate cutting off of the heating power and opening the unloading door to urgently discharge the tea leaves.

[0030] Example 2, please refer to Figure 2 This invention provides a technical solution: a safety management system applied to the production and preparation of Forsythia suspensa tea, which is applicable to the aforementioned safety management method applied to the production and preparation of Forsythia suspensa tea, comprising: Data acquisition module 1 is used to acquire raw material image data of forsythia tea and gas concentration data and temperature data during the frying process; wherein, the gas concentration data includes at least acetaldehyde concentration value, ethanol concentration value and ethyl acetate concentration value; Image processing module 2 is used to construct a feature vector of the ratio of color saturation to brightness based on the raw material image data, and to perform vector matching between the ratio feature vector and the preset stem feature interval to filter out the stem pixel set; The proportion quantization module 3 is used to calculate the first area value of the pixel set of the stem and the second area value of the effective area of ​​the whole image, perform a floating-point operation to divide the two to obtain the stem content feature value, and map the stem content feature value to the raw material quality influence coefficient. Feature construction module 4 is used to construct feature vectors of concentration ratios of characteristic gas components based on gas concentration data; Risk calculation module 5 is used to multiply the concentration ratio feature vector with temperature data to generate a process risk index; Threshold weighting module 6 is used to dynamically weight and adjust the judgment threshold of the process risk index based on the raw material quality influence coefficient to obtain the target risk threshold; The control output module 7 is used to compare the process risk index with the target risk threshold and generate a graded control signal based on the comparison result; the graded control signal is used to drive the actuator of the frying equipment.

[0031] The embodiments of the present invention have been described in detail above with reference to the accompanying drawings. However, the present invention is not limited thereto. Various changes can be made within the scope of knowledge possessed by those skilled in the art without departing from the spirit of the present invention.

Claims

1. A safety management method applied to the production and preparation of Forsythia tea, characterized in that, include: Acquire raw material image data and gas concentration and temperature data during the roasting process of Forsythia tea; wherein, the gas concentration data includes at least the concentration values ​​of acetaldehyde, ethanol and ethyl acetate. Based on the raw material image data, a feature vector of the ratio of color saturation to brightness is constructed. The feature vector of the ratio is then matched with a preset stem feature interval to filter out the stem pixel set. The first area value of the pixel set of the stem and the second area value of the effective area of ​​the whole image are calculated, and a floating-point operation is performed to divide the two to obtain the stem content feature value. The stem content feature value is then mapped to the raw material quality influence coefficient. Based on gas concentration data, a feature vector of the concentration ratio of characteristic gas components is constructed; The process risk index is generated by multiplying the concentration ratio feature vector with the temperature data. The threshold for determining the process risk index is dynamically weighted and adjusted based on the raw material quality influence coefficient to obtain the target risk threshold. The process risk index is compared with the target risk threshold, and a graded control signal is generated based on the comparison result; the graded control signal is used to drive the actuator of the frying equipment.

2. The safety management method applied to the production and preparation of Forsythia tea according to claim 1, characterized in that, Based on the raw material image data, a feature vector representing the ratio of color saturation to brightness is constructed. This feature vector is then matched with a preset stem feature interval to filter out a set of stem pixels, including: The raw material image data is converted to HSV color space, and the saturation channel value sequence and brightness channel value sequence of each pixel are extracted. Perform a pixel-by-pixel division operation between the saturation channel value sequence and the brightness channel value sequence to generate a one-dimensional ratio feature vector; Retrieve the stalk feature interval constructed based on physical optical properties, compare each element in the ratio feature vector with the upper and lower boundary values ​​of the stalk feature interval one by one, and generate a Boolean filtering mask. Boolean filtering masks are used to index and extract the raw material image data. The coordinates of pixels with true mask values ​​are retained, and pixels with false mask values ​​are removed to obtain the root pixel set.

3. The safety management method applied to the production and preparation of Forsythia tea according to claim 2, characterized in that, The first area value of the pixel set containing the spurious spurious spurs and the second area value of the effective region of the entire image are calculated, and a floating-point operation is performed to divide the two to obtain the spurious ... Extract the pixel coordinates from the pixel set of the stem, and accumulate the pixel coordinates to obtain the first area value representing the physical space occupied by the stem; The total number of pixels in the effective area of ​​the raw material image data is counted to obtain a second area value that represents the overall scale of the raw material. The first area value and the second area value are called, and a floating-point division operation is performed to generate a normalized stem content feature value; wherein, the stem content feature value represents the area ratio of stems in the raw material.

4. The safety management method applied to the production and preparation of Forsythia tea according to claim 3, characterized in that, Mapping the stem content characteristic value to the raw material quality influence coefficient, including: Based on the physical difference between the high thermal conductivity of tea stem fibers and the high thermal resistance of leaf cells, a curve showing the relationship between stem content and thermal resistance was constructed. Input the stem content characteristic value as the horizontal axis into the stem content-thermal resistance interference relationship curve, and query the corresponding vertical axis value to obtain the raw material quality influence coefficient. Among them, the slope of the stem content-thermal resistance interference relationship curve increases monotonically with the increase of stem content, which is used to characterize the increase in heat transfer efficiency during stir-frying caused by the increase of stem content.

5. The safety management method applied to the production and preparation of Forsythia tea according to claim 4, characterized in that, Based on gas concentration data, a feature vector of the concentration ratio of characteristic gas components is constructed, including: Based on the acetaldehyde concentration value and ethanol concentration value in the gas concentration data, the ratio of the two is calculated to generate the main characteristic ratio characterizing the coking risk. Using the principal feature ratio as the first element of the vector and the ethyl acetate concentration value in the gas concentration data as the second element of the vector, a two-dimensional concentration ratio feature vector is constructed; whereby the concentration ratio feature vector is used to characterize the caramelization ratio state of the current roasting stage.

6. The safety management method applied to the production and preparation of Forsythia tea according to claim 5, characterized in that, The process risk index is generated by multiplying the concentration ratio feature vector with the temperature data, including: Extract the principal feature ratio from the concentration ratio feature vector, multiply it with the real-time collected pot temperature data, and generate a thermo-chemical reaction factor characterizing the intensity of oxidative coking. The ethyl acetate concentration value is extracted from the concentration ratio feature vector and multiplied with the pot temperature data to generate a heat-mass transfer factor characterizing the volatilization rate of aroma components. Based on the current technological focus of the roasting stage, risk weight coefficients are assigned to the thermo-chemical reaction factor and the heat-mass transfer factor respectively. The comprehensive process risk index is generated by multiplying the thermo-chemical reaction factor and the heat-mass transfer factor by their corresponding risk weight coefficients and then summing the results.

7. The safety management method applied to the production and preparation of Forsythia tea according to claim 6, characterized in that, The threshold for determining the process risk index is dynamically weighted and adjusted based on the raw material quality impact coefficient to obtain the target risk threshold, which includes: Retrieve the pre-stored standard process risk threshold, where the standard threshold is a fixed safety boundary without considering raw material fluctuations; The raw material quality influence coefficient is multiplied with the standard process risk threshold to generate a threshold offset that characterizes the degree of thermal inertia interference of the raw material. Based on the current operating time signal of the roasting equipment and the pre-stored stage-sensitivity mapping table, the threshold offset is weighted and corrected to generate a dynamic correction factor. The target risk threshold is obtained by subtracting the dynamic correction factor from the standard process risk threshold.

8. The safety management method applied to the production and preparation of Forsythia tea according to claim 7, characterized in that, Based on the current operating time signal of the roasting equipment and the pre-stored stage-sensitivity mapping table, the threshold offset is weighted and corrected to generate a dynamic correction factor, including: The running time counter of the frying equipment is read, and based on the inherent physical timing characteristics of the forsythia tea frying process, the running time is rigidly divided into the fixation stage, the rolling stage, and the drying stage through a hardware comparator. Based on the current process stage, retrieve the pre-stored stage-sensitivity coefficient table to obtain the stage sensitivity coefficient; The stage sensitivity coefficient is multiplied by the threshold offset to generate a dynamic correction factor.

9. The safety management method applied to the production and preparation of Forsythia tea according to claim 8, characterized in that, The process risk index is compared with the target risk threshold, and a graded control signal is generated based on the comparison results, including: The process risk index is subtracted from the target risk threshold to obtain the real-time risk difference. The real-time risk difference is used to quantify the degree to which the current process state deviates from the safety boundary. The larger the difference, the higher the risk. Based on the magnitude of the real-time risk difference and the heat accumulation characteristics of the forsythia tea frying process, the risk status is rigidly divided into three levels: safe steady state, warning critical state, and dangerous runaway state. The dangerous runaway state is determined by the risk difference exceeding the preset heat accumulation rate limit, which is used to identify the risk of sudden scorching. Match the corresponding control strategy for the implementing agency based on the current risk status range.

10. A safety management system applied to the production and preparation of Forsythia tea, which is applicable to the safety management method applied to the production and preparation of Forsythia tea as described in any one of claims 1-9, characterized in that, include: The data acquisition module is used to acquire raw material image data of Forsythia tea and gas concentration data and temperature data during the frying process; wherein, the gas concentration data includes at least the concentration values ​​of acetaldehyde, ethanol and ethyl acetate. The image processing module is used to construct a feature vector of the ratio of color saturation to brightness based on the raw material image data, and to perform vector matching between the ratio feature vector and the preset stem feature interval to filter out the stem pixel set. The proportion quantization module is used to calculate the first area value of the pixel set of the stem and the second area value of the effective area of ​​the whole image, perform a floating-point operation to divide the two to obtain the stem content feature value, and map the stem content feature value to the raw material quality influence coefficient. The feature construction module is used to construct feature vectors of the concentration ratios of characteristic gas components based on gas concentration data; The risk calculation module is used to multiply the concentration ratio feature vector with temperature data to generate a process risk index. The threshold weighting module is used to dynamically adjust the judgment threshold of the process risk index based on the raw material quality influence coefficient to obtain the target risk threshold. The control output module is used to compare the process risk index with the target risk threshold and generate a graded control signal based on the comparison result; the graded control signal is used to drive the actuator of the frying equipment.

Images