High-throughput detection intelligent early warning management and control system for medicinal and edible homologous contaminants

Abstract

The application discloses a medicine and food homologous pollutant high-throughput detection intelligent early warning management and control system and relates to the field of medicine and food homologous pollutant detection.The system comprises a binding module, a detection module and the like.The binding module is used for inputting raw material origin identification and processing batch information of a medicine and food homologous sample to be detected and binding the information with a sample physical carrier.The detection module is used for completing concentration determination of heavy metals, pesticide residues and mycotoxins in a same detection process in parallel and outputting original detection values of various pollutants.The application can realize accurate and efficient supervision of medicine and food homologous products by synchronously detecting multiple types of pollutants at one time, tracing and encoding batch information of origins, generating a sample risk fingerprint based on pollutant concentration, establishing a pollution determination standard by combining historical fingerprints, identifying various types of pollution in a double-dimension mode, dynamically adjusting early warning levels, directionally pushing disposal instructions and continuously optimizing determination accuracy.

Description

Technical Field

[0001] This invention relates to the field of detection technology for contaminants that are both food and medicine, specifically to a high-throughput intelligent early warning and control system for contaminants that are both food and medicine. Background Technology

[0002] Food and medicinal substances possess dual attributes, being consumed in large quantities daily, and their safety directly impacts public health. During cultivation, environmental factors such as soil and water sources may introduce heavy metal contamination like lead and cadmium; during processing, improper temperature and humidity control can easily lead to the growth of fungal toxins, and pesticide residues are also a significant concern. These pollutants are cumulative and have synergistic toxicity, meaning even low-dose, long-term exposure can pose health risks. With the continuous growth in the types and usage of food and medicinal products, regulatory authorities are imposing increasingly stringent quality and safety requirements, necessitating efficient and rapid testing and risk management methods.

[0003] The invention patent application with application number 202510352916.1 discloses a traceability detection system and method based on food safety supervision. The application aims to identify abnormal food through chromatography and mass spectrometry analysis, and then conduct traceability analysis and early warning for abnormal links.

[0004] However, existing high-throughput testing equipment mostly focuses on outputting raw test values ​​in batches. The test data is disconnected from traceability information such as the origin of raw materials and processing batches. It is impossible to automatically identify pollution risks and trigger early warnings based on batch test results. The lack of data interaction channels between high-throughput testing and traceability management has, to some extent, restricted the effective transformation of test data into risk decision-making.

[0005] To address this, we have proposed a high-throughput intelligent early warning and control system for pollutants that originate from both food and medicine. Summary of the Invention

[0006] To address the aforementioned shortcomings of existing technologies, this invention provides a high-throughput intelligent early warning and control system for the detection of contaminants that are both food and medicine, which can effectively solve the problems of existing technologies.

[0007] To achieve the above objectives, the present invention is implemented through the following technical solutions; This invention discloses a high-throughput intelligent early warning and control system for contaminants that are both food and medicine, comprising: The system comprises the following modules: a binding module for inputting the origin and batch information of the raw materials of the food-medicine homology sample to be tested, and binding the information to the physical carrier of the sample; a detection module for performing parallel concentration measurements of multiple pollutants, including heavy metals, pesticide residues, and mycotoxins, in the same detection process, and outputting the original detection values ​​of each pollutant; a fingerprint generation module for acquiring the original detection values ​​of various pollutants output by the detection module, combining the original detection values ​​of various pollutants of the same sample into a multi-dimensional vector, and using the multi-dimensional vector as the risk fingerprint of the sample; a fingerprint comparison module for receiving the risk fingerprint output by the fingerprint generation module and matching the risk fingerprint with each risk pattern in the risk pattern library; an early warning module for receiving the matching results output by the fingerprint comparison module, generating early warning information containing risk type, origin, and batch information based on the risk pattern corresponding to the matching results; and an output module for acquiring the early warning information generated by the early warning module, generating and outputting control and disposal instructions based on the risk type, origin identification, and batch information in the early warning information. The binding module is interconnected with a detection module via a wireless network. The detection module is interconnected with a fingerprint generation module via a wireless network. The fingerprint generation module is interconnected with a fingerprint comparison module via a wireless network. The fingerprint comparison module is interconnected with an early warning module via a wireless network. The early warning module is interconnected with an output module via a wireless network.

[0008] Furthermore, during the binding module's operation phase, a unique traceability code is generated, which is composed of the origin code segment, the processing batch code segment, and the sample sequence code segment concatenated sequentially. The raw material origin identification includes geographical location information of the origin and information of the planting and production entity; the processing batch information includes batch production time information and processing parameters. The binding module associates and stores the raw material origin identifier, processing batch information and the unique traceability code in the built-in traceability database of the binding module; The binding module attaches a unique traceability code to the physical carrier of the sample in the form of a machine-recognizable identifier. All data generated in each stage of the system process, including detection, fingerprint generation, comparison, early warning, and control, are automatically attached to the corresponding unique traceability code.

[0009] Furthermore, the detection module includes a preprocessing unit, a multi-channel parallel detection unit, and a timing synchronization control unit; The pretreatment unit performs a single pretreatment operation on the sample to be tested to obtain a test solution that can be adapted to the detection of multiple types of pollutants at the same time, and simultaneously delivers the test solution to each detection channel of the multi-channel parallel detection unit; each detection channel corresponds to different pollutant targets in heavy metals, pesticide residues, and mycotoxins, and the signals generated by each detection channel do not interfere with each other; The timing synchronization control unit controls all detection channels to synchronously complete signal acquisition, data processing and raw detection value output within the same detection cycle. All output raw detection values ​​correspond one-to-one with the unique traceability code of the sample to be tested.

[0010] Furthermore, during the operation of the fingerprint generation module, the original detection values ​​of various pollutants output by the detection module are converted into nonlinear risk quantification, and then the converted values ​​are arranged into a multi-dimensional vector according to a preset fixed order of pollutant categories. This multi-dimensional vector is used as the risk fingerprint of the sample. Wherein, after the nonlinear risk quantification transformation, the risk fingerprint dimension value corresponding to the i-th type of pollutant is: ; In the formula: This represents the original detection concentration value of pollutant of type i; This represents the statutory safety limit value corresponding to the i-th type of pollutant; This represents the chronic risk growth coefficient in the low concentration range. This represents the acute risk growth coefficient within the range exceeding the standard, and > >0.

[0011] Furthermore, the risk pattern library pre-stores several sets of independent risk patterns. Each set of risk patterns includes a corresponding feature fingerprint vector, risk type label, risk level identifier, typical pollution causes, and control direction guidance. The feature fingerprint vector of each risk pattern is calculated by clustering centers from the risk fingerprints of multiple batches of historical positive samples under the corresponding pollution scenario. Each set of risk patterns corresponds to a unique pollution cause type and risk level. The feature fingerprint vectors of different risk patterns have distinguishable dimensional distribution characteristics.

[0012] Furthermore, the matching operation in the fingerprint comparison module follows the following rules: Extract the dimensions with a value greater than 1 from the risk fingerprint to be matched as the set of dominant dimensions to be matched, and extract the dimensions with a value greater than 1 from the feature fingerprint vector of the corresponding risk pattern as the set of dominant dimensions of the pattern. Let t be the total number of dimensions in the union of two sets, and s be the number of dimensions in the intersection of two sets. Calculate the order matching coefficient of two sets. Matching coefficient with amplitude Final matching degree ; Where the order matching coefficient The ratio of the number of intersection dimensions *s* to the total number of union dimensions *t*; magnitude matching coefficient. ; In the formula: The dimension value of the risk fingerprint to be matched corresponds to the i-th dimension within the intersection; The dimension value of the feature fingerprint vector corresponding to the i-th dimension within the intersection; The fingerprint comparison module selects the risk pattern with the highest matching degree and a value not lower than the preset matching threshold as the final matching result.

[0013] Furthermore, the early warning module is pre-set with multiple early warning levels corresponding one-to-one with risk levels. After obtaining the matching results output by the fingerprint comparison module, the early warning module first uses the risk level identifier corresponding to the matched risk pattern as the basic early warning level, and then adjusts the early warning level up or down according to the comprehensive risk characterization value calculated based on the values ​​of each dimension of the risk fingerprint to be matched. It also generates a pollution cause prediction by combining the typical pollution causes corresponding to the matched risk pattern, and finally generates early warning information that includes risk type, early warning level, pollution cause prediction, raw material origin identifier, processing batch information and unique traceability code.

[0014] Furthermore, the output module is pre-set with a set of control and response strategies corresponding to each warning level and each risk type; After obtaining the early warning information generated by the early warning module, the output module retrieves the corresponding control and disposal strategy based on the early warning level and risk type in the early warning information. Simultaneously, it applies the raw material origin identification, processing batch information and unique traceability code to generate targeted control and disposal instructions. The control and disposal instructions are pushed to the production entity terminal, processing enterprise terminal and regulatory department terminal in the corresponding origin, and the traceability verification process of the corresponding batch of products is triggered simultaneously.

[0015] Furthermore, the fingerprint comparison module synchronously acquires the execution feedback data corresponding to the control and disposal instructions generated by the output module, as well as the verification and detection data output by the detection module for the corresponding batch of samples. It supplements the risk fingerprints corresponding to the positive samples confirmed by verification into the sample set of the corresponding risk pattern in the risk pattern library. The fingerprint comparison module performs cluster center calculation on the sample set of each risk pattern according to a preset period, completes the iterative update of the corresponding feature fingerprint vector, and dynamically adjusts the preset matching threshold according to the accuracy of historical matching results.

[0016] Compared with the known prior art, the technical solution provided by this invention has the following beneficial effects: This invention connects the origin and batch information throughout the entire process through a unique traceability code, enabling full traceability of each sample data. Based on the conversion detection values, it constructs a unique risk fingerprint, which can objectively distinguish the differences in acute and chronic risks brought by different pollutants. By using historical positive samples to form a standardized risk pattern library, it can accurately identify the risk characteristics corresponding to various pollution causes. By calculating the matching degree through the dominant dimension of ordinal position and amplitude, it can not only lock highly similar known pollution scenarios, but also quickly identify unknown abnormal samples without corresponding patterns. Attached Figure Description

[0017] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the accompanying drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are merely some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without any creative effort.

[0018] Figure 1 A schematic diagram of a high-throughput intelligent early warning and control system for contaminants from both food and medicine sources; Figure 2 This is a schematic diagram of the parallel determination process for multiple pollutants in the detection module. Detailed Implementation

[0019] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, 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, not all, of the embodiments of the present invention. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection of the present invention.

[0020] The present invention will be further described below with reference to embodiments.

[0021] Example: The high-throughput intelligent early warning and control system for food and drug contaminants in this embodiment, such as Figure 1 As shown, it includes: The binding module is used to input the origin identification and processing batch information of the raw materials of the food and medicine homology samples to be tested, and to bind the information to the physical carrier of the sample; During the binding module's runtime phase, a unique traceability code is generated, which is composed of the origin code segment, processing batch code segment, and sample sequence code segment concatenated sequentially. The raw material origin identification includes geographical location information of the origin and information on the planting and production entities, while the processing batch information includes batch production time information and processing parameters. The binding module associates and stores the raw material origin identifier, processing batch information and unique traceability code in the built-in traceability database of the binding module; The binding module attaches the unique traceability code to the physical carrier of the sample in the form of a machine-recognizable identifier. All data generated in each stage of the system process, including detection, fingerprint generation, comparison, early warning and control, are automatically attached to the corresponding unique traceability code. The detection module is used to perform parallel determination of the concentration of multiple pollutants such as heavy metals, pesticide residues, and mycotoxins in the same detection process, and output the original detection values ​​of each pollutant. The detection module includes a preprocessing unit, a multi-channel parallel detection unit, and a timing synchronization control unit; The pretreatment unit performs a single pretreatment operation on the sample to be tested to obtain a test solution that can be adapted to the detection of multiple types of pollutants at the same time, and simultaneously delivers the test solution to each detection channel of the multi-channel parallel detection unit; each detection channel corresponds to different pollutant targets in heavy metals, pesticide residues, and mycotoxins, and the signals generated by each detection channel do not interfere with each other; The timing synchronization control unit controls all detection channels to synchronously complete signal acquisition, data processing and raw detection value output within the same detection cycle. All output raw detection values ​​correspond one-to-one with the unique traceability code of the sample to be tested. Specifically, a single preprocessing operation is as follows: First, the sample to be tested is homogenized and pulverized, and then a composite extraction solvent is added for constant temperature oscillation extraction. Heavy metals, pesticide residues and fungal toxins in the sample are extracted simultaneously. Then, the sample is purified by a mixed solid phase extraction column to remove matrix pigments and co-extract impurities. Finally, the sample is diluted to volume and filtered through a microporous membrane to obtain the test solution. In this example, the composite extraction solvent is a mixture of acetonitrile and a 1% (v / v) formic acid aqueous solution; the isothermal shaking extraction temperature is 30-50℃, the shaking frequency is 150-250 rpm, and the extraction time is 20-40 minutes; the mixed solid-phase extraction column is a composite column packed with a mixture of octadecyl bonded silica gel and a strong cation exchange adsorbent; the volume is adjusted using the composite extraction solvent; and the pore size of the microporous filter membrane is 0.22 micrometers. The heavy metal detection channel uses the electrochemical voltammetry detection principle, the pesticide residue detection channel uses the surface-enhanced Raman spectroscopy detection principle, and the mycotoxin detection channel uses the time-resolved fluorescence immunoassay principle. The three channels complete the detection based on the response signals of different physical types to ensure no signal overlap. The heavy metal detection channel employs a working electrode, such as a glassy carbon electrode modified with a bismuth film, in conjunction with a silver / silver chloride reference electrode and a platinum wire counter electrode. It uses square-wave anodic stripping voltammetry with a scanning potential range of -1.4 V to 0.6 V, a square-wave frequency of 25 Hz, a square-wave amplitude of 25 mV, and a potential increment of 4 mV. The pesticide residue detection channel uses gold nanoparticle sol as the surface-enhanced Raman spectroscopy substrate. The gold nanoparticles have a particle size of 55 ± 5 nm, the excitation laser wavelength is 785 nm, the laser power is 50 mW, and the integration time is 10 seconds. The mycotoxin detection channel uses a time-resolved fluorescence immunochromatographic strip with europium (Eu) as the fluorescent label. 3+The nanospheres were labeled with an excitation wavelength of 365 nm and an emission wavelength of 615 nm. The signal was acquired 15 minutes after the test strip had reacted. The data processing stage follows the following rules: With a unified clock reference, each detection channel is allocated an independent signal acquisition time slot and a parallel calculation thread. The original response signal acquisition of each channel is triggered synchronously. Each detection channel independently completes the calculation by calling its own pre-calibrated signal-concentration calibration curve. During the calculation process, the data of each channel are isolated from each other. Finally, the original detection values ​​of all pollutants are output at the same time. The fingerprint generation module is used to obtain the original detection values ​​of various pollutants output by the detection module, and combine the original detection values ​​of various pollutants in the same sample into a multi-dimensional vector, which is used as the risk fingerprint of the sample. During the fingerprint generation module operation phase, the original detection values ​​of various pollutants output by the detection module are transformed into nonlinear risk quantification. Then, the transformed values ​​are arranged into a multi-dimensional vector according to the preset fixed order of pollutant categories. This multi-dimensional vector is used as the risk fingerprint of the sample. After nonlinear risk quantification transformation, the risk fingerprint dimension value corresponding to the i-th type of pollutant is: ; In the formula: This represents the original detection concentration value of pollutant of type i; This represents the statutory safety limit value corresponding to the i-th type of pollutant; The chronic risk growth coefficient in the low concentration range is determined by the chronic toxicity level of this type of pollutant. The acute risk growth coefficient for the exceedance range is determined by the acute toxicity level of the pollutant, and... > >0; The above formula divides the pollutant into two concentration ranges: those not exceeding the standard and those exceeding the standard, and calculates the risk value separately. For the low concentration range, the chronic risk coefficient is calculated by exponentiation of the ratio of the measured concentration of the pollutant to the safety limit, which is consistent with the cumulative chronic harm caused by long-term low-dose intake of pollutants. For the exceeding range, the acute risk coefficient increment corresponding to the exceedance is added to the basic value to distinguish the difference in the harm growth of chronic and acute toxicity. At the same time, two growth coefficients are set according to the pollutant toxicology data and the acute coefficient value is fixed to be larger. Unlike the simple processing method of uniform linear conversion, it can accurately amplify the risk weight of pollutants exceeding the standard, so that the risk difference of samples with different toxicity and different pollution levels is intuitively reflected in the fingerprint vector dimension value. , All values ​​are determined using a tiered approach: chronic toxicity is classified into multiple levels, from weak to strong, based on the acceptable daily intake of pollutants. The higher the level of chronic toxicity, the greater the corresponding... The larger the value of ; Acute toxicity is classified into several levels from low to high based on the oral median lethal dose (LD50) of the pollutant. The higher the acute toxicity level, the greater the corresponding... The larger the value of ; The two growth coefficients for the same pollutant are matched one-to-one with its toxicity level, and the acute risk growth coefficient is always greater than the chronic risk growth coefficient. The fingerprint comparison module is used to receive the risk fingerprints output by the fingerprint generation module and match the risk fingerprints with each risk pattern in the risk pattern library; The risk pattern library pre-stores several independent risk patterns. Each risk pattern includes a corresponding feature fingerprint vector, risk type label, risk level identifier, typical pollution causes and control direction guidance. The feature fingerprint vector of each risk pattern is calculated by clustering centers from the risk fingerprints of multiple batches of historical positive samples under the corresponding pollution scenario. Each risk pattern corresponds to a unique pollution cause type and risk level. The feature fingerprint vectors of different risk patterns have distinguishable dimensional distribution characteristics. Among them, the risk pattern library was pre-constructed based on the detection data of multiple sets of historical food and medicine homology samples; Historical positive samples refer to samples that have been confirmed by laboratory confirmatory testing to have a concentration of at least one type of pollutant exceeding the statutory safety limit value for the corresponding category in the GB2762, GB2763 and GB5009 series standards. The confirmatory testing is carried out independently using national standard testing methods and is independent of the test results of the detection module of this invention. Historical samples were sourced from at least three major production areas of medicinal and edible plantation bases and at least five processing enterprises, with a time span of at least 12 months. Each risk pattern corresponds to a sample set containing risk fingerprint data of no fewer than 20 batches of independent positive samples, with no overlap in origin or processing batch among the sample batches; Cluster center calculation takes the risk fingerprint of multiple batches of historical positive samples under the same pollution cause type as the input sample set, uses an unsupervised clustering algorithm to perform clustering operation on the input sample set, and takes the cluster center vector output by the operation as the feature fingerprint vector corresponding to the risk pattern. Distinguishing dimensional distribution features refer to the differences in the types of typical pollutants exceeding the standard corresponding to different risk patterns. The dimensional position, number of dimensions, and numerical range of high-value dimensions with a dimension value greater than 1 in the corresponding feature fingerprint vector all show significant identifiable differences. Based on these differences in dimensional distribution, the pollution cause type and risk level corresponding to different risk patterns can be distinguished. The matching operation in the fingerprint comparison module follows the following rules: Extract the dimensions with a value greater than 1 from the risk fingerprint to be matched as the set of dominant dimensions to be matched, and extract the dimensions with a value greater than 1 from the feature fingerprint vector of the corresponding risk pattern as the set of dominant dimensions of the pattern. Let t be the total number of dimensions in the union of two sets, and s be the number of dimensions in the intersection of two sets. Calculate the order matching coefficient of two sets. Matching coefficient with amplitude Final matching degree ; Where the order matching coefficient The ratio of the number of intersection dimensions *s* to the total number of union dimensions *t*; magnitude matching coefficient. ; The above formula constructs an exponential decay operation by using the mean of the logarithmic difference between the risk values ​​of the fingerprint to be tested and the standard feature fingerprint under the intersection dimension. The logarithmic transformation is used to eliminate the matching interference caused by the difference in the magnitude of the pollutant values. The absolute value is taken to capture the degree of deviation of the two types of fingerprints in the same contamination dimension. The mean processing balances the influence of the common deviation of multiple types of pollutants. The exponential function can convert the magnitude of the deviation into a matching score in the range of 0 to 1. The larger the deviation, the lower the matching score. Unlike the traditional matching algorithm that simply compares the numerical difference, it can simultaneously take into account the relative multiple difference of pollutant risk and the synergistic deviation of multiple pollutants, and accurately quantify the degree of fit of the contamination severity of the two types of fingerprints. In the formula: The dimension value of the risk fingerprint to be matched corresponds to the i-th dimension within the intersection; The dimension value of the feature fingerprint vector corresponding to the i-th dimension within the intersection; The fingerprint comparison module selects the risk pattern with the highest matching degree and a value not lower than the preset matching threshold as the final matching result; If the matching degree of all risk patterns is lower than the preset matching threshold, it is determined that no known risk pattern is matched. The fingerprint comparison module generates an abnormal risk mark, stores the risk fingerprint to be matched in the risk sample library to be reviewed, and outputs an abnormal signal of non-matching to the early warning module. The fingerprint comparison module synchronously acquires the execution feedback data corresponding to the control and disposal instructions generated by the output module, as well as the verification test data output by the detection module for the corresponding batch of samples. It supplements the risk fingerprints corresponding to the positive samples confirmed by verification into the sample set of the corresponding risk pattern in the risk pattern library. The fingerprint comparison module performs cluster center calculation on the sample set of each risk pattern according to a preset period, completes the iterative update of the corresponding feature fingerprint vector, and dynamically adjusts the preset matching threshold according to the accuracy of historical matching results. The cluster center calculation uses the K-Means mean clustering algorithm, which takes the arithmetic mean of the values ​​of each dimension of all risk fingerprint vectors in the same risk pattern sample set as the updated feature fingerprint vector. The accuracy of historical matching results is represented by the ratio of the number of samples whose matching judgment results are consistent with the review test results under the same risk mode to the total number of matching samples under that mode. When the accuracy is lower than the set lower limit, the matching threshold is increased; when the accuracy is higher than the set upper limit, the matching threshold is decreased. The initial value of the matching threshold is set to 0.75; the lower limit of accuracy is set to 85%, and the upper limit is set to 95%; when the accuracy of 50 consecutive matching samples is lower than 85%, the matching threshold is increased by 0.05; when the accuracy of 50 consecutive matching samples is higher than 95%, the matching threshold is decreased by 0.05; the adjusted matching threshold is no lower than 0.60 and no higher than 0.90. The early warning module is used to receive the matching results output by the fingerprint comparison module, and generate early warning information containing risk type, place of origin and batch information based on the risk pattern corresponding to the matching results. The early warning module has a multi-level early warning system that corresponds one-to-one with the risk level. After obtaining the matching results output by the fingerprint comparison module, the early warning module first uses the risk level identifier corresponding to the matched risk pattern as the basic level of the early warning. Then, it adjusts the early warning level by floating up or down according to the comprehensive risk characterization value calculated based on the values ​​of each dimension of the risk fingerprint to be matched. It also generates a pollution cause prediction by combining the typical pollution causes corresponding to the matched risk pattern. Finally, it generates early warning information that includes risk type, early warning level, pollution cause prediction, raw material origin identifier, processing batch information and unique traceability code. The comprehensive risk characterization value is obtained by power-means calculation based on all dimensions of the risk fingerprint to be matched. ; In the formula: R is the comprehensive risk characterization value corresponding to the sample to be matched; is the value of the i-th dimension of the risk fingerprint to be matched; n is the total number of dimensions of the risk fingerprint; p is the preset risk extreme value sensitivity coefficient and satisfies p>1. Its value is positively correlated with the regulatory strictness level and the comprehensive toxicity level of pollutants in the application scenario. The higher the regulatory strictness level and the higher the comprehensive toxicity level of pollutants, the larger the value of p. The above formula uses a power-means algorithm based on multi-dimensional risk values ​​to calculate the overall risk level of the sample. It introduces an extreme value sensitivity coefficient greater than 1 to amplify the impact of high-risk pollutants on the overall risk result. This aligns with the regulatory needs of prioritizing the control of highly toxic and excessive pollutants in regulatory scenarios. The coefficient value is adjusted synchronously with the degree of regulatory strictness and the comprehensive toxicity of the pollutants. Unlike the calculation method of ordinary arithmetic average, which dilutes the risk contribution of pollutants exceeding the standard, this formula can highlight the overall safety hazards brought about by single or multiple severely excessive pollutants, thereby providing a quantitative basis for the dynamic adjustment of the warning level that fits the actual regulatory logic. The early warning module has a multi-level floating threshold range corresponding to the comprehensive risk characterization value. The comprehensive risk characterization value is positively correlated with the risk level. The early warning module adjusts the basic early warning level up or down according to the threshold range into which the comprehensive risk characterization value falls. The warning levels are divided into four levels, from lowest to highest: Level IV (Attention), Level III (Alert), Level II (Warning), and Level I (Emergency). The fluctuation threshold range of the comprehensive risk characterization value R is as follows: When R≤1.5, the warning level will not be adjusted and will remain at the basic warning level. When 1.5 < R ≤ 3.0, the warning level is raised by one level; When 3.0 < R ≤ 5.0, the warning level is raised by two levels; When R > 5.0, the warning level is raised by three levels, but the highest level after the upgrade shall not exceed Level I; If the basic warning level is Level IV, and R ≤ 1.5, then Level IV will remain unchanged and will not be downgraded. The output module is used to acquire the early warning information generated by the early warning module, and generate and output control and disposal instructions based on the risk type, place of origin and batch information in the early warning information. The output module has a set of control and response strategies pre-set for each warning level and risk type; After obtaining the warning information generated by the warning module, the output module retrieves the corresponding control and disposal strategy based on the warning level and risk type in the warning information. Simultaneously, it uses the raw material origin identification, processing batch information and unique traceability code to generate targeted control and disposal instructions. The control and disposal instructions are pushed to the production entity terminal, processing enterprise terminal and regulatory department terminal in the corresponding origin, and the traceability verification process of the corresponding batch of products is triggered simultaneously. In the targeted control and disposal instruction generation stage, the output module first extracts the built-in basic disposal rules, disposal execution time limit and disposal verification requirements of the retrieved control and disposal strategy, and then embeds the risk type, warning level and pollution cause prediction contained in the early warning information as the basis for disposal into the instruction content. At the same time, it matches the production entity and regulatory authority information of the corresponding location based on the raw material origin identification, locks the production node, inventory range and circulation range of the corresponding batch of products based on the processing batch information, and generates a verifiable traceability identifier that is bound to the instruction one by one with the unique traceability code. Finally, it generates differentiated targeted control and disposal instructions that are adapted to the production entity terminal, processing enterprise terminal and regulatory department terminal respectively. Example: The instructions pushed to the production entity's terminal include the risk type, warning level, pollution cause prediction, corresponding place of origin identification, and a batch list of all positive samples from that place of origin in the past 6 months. The instructions require the production entity to submit a self-inspection report of the place of origin environment within 24 hours and suspend the harvesting and delivery of raw materials from that place of origin. The instructions pushed to the processing enterprise's terminal include the risk type, warning level, pollution cause prediction, corresponding processing batch information, and the origin of the raw materials in that batch. The instructions require the processing enterprise to isolate and seal the finished products and work-in-process in the same batch within 48 hours and to initiate internal traceability investigation. The instruction pushed to the regulatory authority's terminal includes all the aforementioned information, as well as a unique traceability code, a time limit for disposal execution, disposal verification requirements, and preliminary traceability information on the place of origin and the distribution range of the batch of products. The instruction requires the regulatory authority to complete the on-site verification and upload the verification conclusion within 72 hours. The binding module is connected to the detection module via a wireless network. The detection module is connected to the fingerprint generation module via a wireless network. The fingerprint generation module is connected to the fingerprint comparison module via a wireless network. The fingerprint comparison module is connected to the early warning module via a wireless network. The early warning module is connected to the output module via a wireless network.

[0022] In this embodiment, the system starts operating from sample registration and binding. Staff enter information such as sample origin, planting entity, production time, and processing technology. The system automatically generates a unique traceability code that integrates origin, batch, and sample serial number. The relevant information is stored in the database, and the code is affixed to the sample as an identifiable label. All data generated by subsequent testing, early warning, and control will be associated with this code.

[0023] After the samples are bound, they enter the testing stage. They undergo a unified pretreatment process, including crushing, simultaneous extraction, solid-phase extraction purification, and volume-adjusting filtration, to prepare the test solution, which is then diverted to three independent testing channels. Heavy metals, pesticide residues, and mycotoxins are detected using different principles, and the signals do not interfere with each other. The entire system collects signals synchronously, calculates concentrations separately, and outputs all original test data of pollutants in the same cycle. The data and traceability codes correspond one-to-one.

[0024] The concentration data obtained from the detection is transmitted to the fingerprint generation stage. The system compares the national standard limits for various pollutants with the chronic and acute toxicity coefficients of the pollutants themselves and performs piecewise nonlinear conversion to transform the simple concentration into a risk value. Then, it combines the values ​​in a fixed order to form a multidimensional vector, which serves as the unique risk fingerprint of the sample.

[0025] After generating the risk fingerprint, the system proceeds to the comparison stage. The system retrieves a pre-built risk pattern library, in which standard fingerprints for various pollution scenarios are derived from the clustering of historical positive samples. The program filters out the dimensions that exceed the standard in the two fingerprints, calculates the matching degree based on the number of overlapping dimensions and the deviation of risk values, and selects the risk type with the highest matching degree. If all matching degrees fail to meet the standard, the sample is marked as an unknown abnormal sample and stored separately for further review.

[0026] The matching results are pushed to the early warning stage. First, the matched risk level is used as the base level. Then, the comprehensive risk value is calculated by combining all the risk values ​​of the sample. The early warning level is adjusted up or down according to the preset threshold. At the same time, the corresponding pollution cause judgment is given. The risk type, early warning level, place of origin and batch, and traceability code are summarized to form a complete early warning message.

[0027] When the early warning information is sent to the output disposal stage, the system retrieves the corresponding control plan based on the early warning level and pollution type, and generates disposal instructions that distinguish between the production end, processing end, and supervision end by combining the information on the place of origin, distribution and inventory. The instructions are then pushed to the corresponding terminals simultaneously, and the traceability and verification of the batch of products is automatically initiated.

[0028] Meanwhile, the comparison module collects processing execution feedback and re-examination data, adds confirmed positive sample fingerprints to the risk pattern library, and regularly re-clusters and updates standard fingerprints; at the same time, it dynamically adjusts the matching judgment threshold based on past matching accuracy to continuously improve recognition accuracy.

[0029] In the above embodiments, the system can simultaneously measure multiple types of pollutants through integrated pre-processing and multi-channel synchronous detection. It relies on exclusive traceability codes to bind the place of origin and batch information, uses quantitative formulas to distinguish the acute and chronic risks of pollutants, and combines historical detection data to form standardized pollution judgment criteria. The dual-dimensional matching method can identify known pollution problems and capture unknown abnormal samples. It can flexibly adjust the warning level based on the comprehensive risk value, accurately predict the cause of pollution, and push targeted disposal instructions to the production, processing, and regulatory ends simultaneously. It can also continuously update the judgment criteria to optimize the identification accuracy, which facilitates the rapid completion of sample screening and precise control of local batches on-site.

[0030] See Figure 2 As shown in the diagram, this is an example of the operational flow and content between the detection module and the fingerprint generation module: The detection module receives the sample and its associated information from the binding module. Within the same detection process, it simultaneously determines the concentrations of multiple pollutants via heavy metal detection channel A, pesticide residue detection channel B, and mycotoxin detection channel C. Specifically, heavy metal detection channel A simultaneously measures elements such as lead, cadmium, arsenic, and mercury; pesticide residue detection channel B simultaneously measures multiple pesticides including organochlorines, organophosphates, pyrethroids, and carbamates; and mycotoxin detection channel C simultaneously measures toxins such as aflatoxin B1, ochratoxin A, zearalenone, and deoxynivalenol. The three detection channels operate in parallel, each outputting the raw detection values ​​for its corresponding pollutant category for use by the fingerprint generation module.

[0031] In summary, this embodiment of the system uses a unique traceability code to link the entire process information of origin and batch, enabling full traceability of each sample data. Based on conversion detection values, it constructs a unique risk fingerprint, which can objectively distinguish the differences in acute and chronic risks brought by different pollutants. By clustering historical positive samples to form a standardized risk pattern library, it can accurately identify the risk characteristics corresponding to various pollution causes. By calculating the matching degree through both the dominant dimension of ordinal position and amplitude, it can not only lock highly similar known pollution scenarios but also quickly identify unknown abnormal samples without corresponding patterns. Simultaneously, the system dynamically calibrates the early warning level using comprehensive risk characterization values, ensuring that the warning level matches the actual severity of sample pollution. It also configures tiered and corresponding control and disposal instructions for output, facilitating regulatory personnel to quickly implement targeted control measures based on origin and batch according to the early warning information. Furthermore, it can dynamically optimize the matching judgment criteria based on historical matching accuracy, continuously improving the stability and reliability of pollution identification and early warning judgment.

[0032] The above embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions will not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A high-throughput intelligent early warning and control system for contaminants originating from both food and medicine, characterized in that, include: The binding module is used to input the origin identification and processing batch information of the raw materials of the food and medicine homology samples to be tested, and to bind the information to the physical carrier of the samples; The detection module is used to perform parallel determination of the concentration of multiple pollutants such as heavy metals, pesticide residues, and mycotoxins in the same detection process, and output the original detection values ​​of each pollutant. The fingerprint generation module is used to obtain the original detection values ​​of various pollutants output by the detection module, combine the original detection values ​​of various pollutants of the same sample into a multi-dimensional vector, and use the multi-dimensional vector as the risk fingerprint of the sample. The fingerprint comparison module is used to receive the risk fingerprint output by the fingerprint generation module and match the risk fingerprint with each risk pattern in the risk pattern library; The early warning module is used to receive the matching results output by the fingerprint comparison module, and generate early warning information containing risk type, place of origin and batch information based on the risk pattern corresponding to the matching results. The output module is used to acquire the early warning information generated by the early warning module, and generate and output control and disposal instructions based on the risk type, place of origin and batch information in the early warning information.

2. The high-throughput intelligent early warning and control system for food and drug homology pollutants according to claim 1, characterized in that, During the operation phase of the binding module, a unique traceability code is generated, which is composed of the origin code segment, the processing batch code segment, and the sample sequence code segment concatenated in sequence. The raw material origin identification includes geographical location information of the origin and information of the planting and production entity; the processing batch information includes batch production time information and processing parameters. The binding module associates and stores the raw material origin identifier, processing batch information and the unique traceability code in the built-in traceability database of the binding module; The binding module attaches a unique traceability code to the physical carrier of the sample in the form of a machine-recognizable identifier. All data generated in each stage of the system process, including detection, fingerprint generation, comparison, early warning, and control, are automatically attached to the corresponding unique traceability code.

3. The high-throughput intelligent early warning and control system for contaminants from both food and medicine sources according to claim 1, characterized in that, The detection module includes a preprocessing unit, a multi-channel parallel detection unit, and a timing synchronization control unit; The pretreatment unit performs a single pretreatment operation on the sample to be tested to obtain a test solution that can be adapted to the detection of multiple types of pollutants at the same time, and simultaneously delivers the test solution to each detection channel of the multi-channel parallel detection unit; each detection channel corresponds to different pollutant targets in heavy metals, pesticide residues, and mycotoxins, and the signals generated by each detection channel do not interfere with each other; The timing synchronization control unit controls all detection channels to synchronously complete signal acquisition, data processing and raw detection value output within the same detection cycle. All output raw detection values ​​correspond one-to-one with the unique traceability code of the sample to be tested.

4. The intelligent early warning and control system for high-throughput detection of food and medicine homologous pollutants according to claim 1, characterized in that, During the operation of the fingerprint generation module, the original detection values ​​of various pollutants output by the detection module are converted into nonlinear risk quantification, and then the converted values ​​are arranged into a multi-dimensional vector according to the preset fixed order of pollutant categories. This multi-dimensional vector is used as the risk fingerprint of the sample. Wherein, after the nonlinear risk quantification transformation, the risk fingerprint dimension value corresponding to the i-th type of pollutant is: ； In the formula: This represents the original detection concentration value of pollutant of type i; This represents the statutory safety limit value corresponding to the i-th type of pollutant; This represents the chronic risk growth coefficient in the low concentration range. This represents the acute risk growth coefficient within the exceeded range, and > >

0.

5. The intelligent early warning and control system for high-throughput detection of food and medicine homologous pollutants according to claim 1, characterized in that, The risk pattern library pre-stores several independent risk patterns. Each risk pattern includes a corresponding feature fingerprint vector, risk type label, risk level identifier, typical pollution causes, and control direction guidance. The feature fingerprint vector of each risk pattern is calculated by clustering centers from the risk fingerprints of multiple batches of historical positive samples under the corresponding pollution scenario. Each risk pattern corresponds to a unique pollution cause type and risk level. The feature fingerprint vectors of different risk patterns have distinguishable dimensional distribution characteristics.

6. The intelligent early warning and control system for high-throughput detection of food and drug homologous pollutants according to claim 1, characterized in that, The matching operation in the fingerprint comparison module follows the following: Extract the dimensions with a value greater than 1 from the risk fingerprint to be matched as the set of dominant dimensions to be matched, and extract the dimensions with a value greater than 1 from the feature fingerprint vector of the corresponding risk pattern as the set of dominant dimensions of the pattern. Let t be the total number of dimensions in the union of two sets, and s be the number of dimensions in the intersection of two sets. Calculate the order matching coefficient of two sets. Matching coefficient with amplitude Final matching degree ; Where the order matching coefficient The ratio of the number of intersection dimensions *s* to the total number of union dimensions *t*; magnitude matching coefficient. ; In the formula: The dimension value of the risk fingerprint to be matched corresponds to the i-th dimension within the intersection; The dimension value of the feature fingerprint vector corresponding to the i-th dimension within the intersection; The fingerprint comparison module selects the risk pattern with the highest matching degree and a value not lower than the preset matching threshold as the final matching result.

7. The intelligent early warning and control system for high-throughput detection of food and medicine homologous pollutants according to claim 1, characterized in that, The early warning module has a multi-level early warning hierarchy that corresponds one-to-one with the risk level. After obtaining the matching result output by the fingerprint comparison module, the early warning module first uses the risk level identifier corresponding to the matched risk pattern as the basic level of the early warning. Then, it adjusts the early warning level by floating up or down according to the comprehensive risk characterization value calculated based on the values ​​of each dimension of the risk fingerprint to be matched. It generates a pollution cause prediction by combining the typical pollution causes corresponding to the matched risk pattern. Finally, it generates early warning information that includes risk type, early warning level, pollution cause prediction, raw material origin identifier, processing batch information and unique traceability code.

8. The intelligent early warning and control system for high-throughput detection of food and medicine homologous pollutants according to claim 1, characterized in that, The output module has a set of control and handling strategies corresponding to each warning level and each risk type. After obtaining the early warning information generated by the early warning module, the output module retrieves the corresponding control and disposal strategy based on the early warning level and risk type in the early warning information. Simultaneously, it applies the raw material origin identification, processing batch information and unique traceability code to generate targeted control and disposal instructions. The control and disposal instructions are pushed to the production entity terminal, processing enterprise terminal and regulatory department terminal in the corresponding origin, and the traceability verification process of the corresponding batch of products is triggered simultaneously.

9. The intelligent early warning and control system for high-throughput detection of food and medicine homologous pollutants according to claim 6, characterized in that, The fingerprint comparison module synchronously acquires the execution feedback data corresponding to the control and disposal instructions generated by the output module, as well as the verification test data output by the detection module for the corresponding batch of samples. It supplements the risk fingerprints corresponding to the positive samples confirmed by verification into the sample set of the corresponding risk pattern in the risk pattern library. The fingerprint comparison module performs cluster center calculation on the sample set of each risk pattern according to a preset period, completes the iterative update of the corresponding feature fingerprint vector, and dynamically adjusts the preset matching threshold according to the accuracy of historical matching results.

10. The intelligent early warning and control system for high-throughput detection of food and medicine homologous pollutants according to claim 1, characterized in that, The binding module is interconnected with a detection module via a wireless network. The detection module is interconnected with a fingerprint generation module via a wireless network. The fingerprint generation module is interconnected with a fingerprint comparison module via a wireless network. The fingerprint comparison module is interconnected with an early warning module via a wireless network. The early warning module is interconnected with an output module via a wireless network.

Images