A residue detection record tracing method for donkey-hide gelatin products

Abstract

This invention relates to the field of agricultural product quality and safety traceability technology, and particularly to a method for tracing pesticide residue detection records in donkey-hide gelatin products. The method includes: acquiring production traceability data containing information such as pesticide variety detection concentration and dynamic viscosity value; constructing a directed acyclic graph for production traceability starting with raw material batches, intermediate processes, and finished products; calculating the conduction retention rate based on the directed edge weights and the survival rate of each pesticide variety after physical degradation, and estimating the theoretical retention concentration; determining the retention correction coefficient for each pesticide variety based on the dynamic viscosity value and the logarithm of the octanol-water partition coefficient, and weighting and correcting it in conjunction with standard reference concentrations to obtain the equivalent risk retention amount; determining a reputation constraint value based on pesticide residue anomalies and the number of supplied batches, and then assessing the probability of risk attribution to generate targeted investigation instructions in reverse. This invention integrates thermal decay, colloidal retention, and supplier reputation to reduce the blindness of pesticide residue traceability investigations under mixing processes.

Description

Technical Field

[0001] This invention relates to the field of agricultural product quality and safety traceability technology, and in particular to a method for tracing the source of pesticide residue detection records in donkey-hide gelatin products. Background Technology

[0002] In large-scale donkey-hide gelatin production, the same batch of finished product is often formed by mixing raw materials from different suppliers and different batches of incoming materials through multiple feeding processes. Furthermore, the materials undergo prolonged heating and a high-viscosity colloidal state. When pesticide residues in the finished product are found to be excessive or abnormally high, quality inspectors need to work backwards from the finished product batch, traversing each intermediate node, to quickly identify the main batch posing a risk from a large number of candidate raw materials. This allows them to decide which batch of inventory to seal first, which supplier to investigate, and whether to expand the scope of finished product disposal.

[0003] Current practices often rely on simple sorting based on ingredient ratios, incoming inspection values, or suppliers' historical anomaly rates, failing to adequately consider the actual differences in pesticide residue retention during the processing. Easily degradable and recalcitrant pesticides are frequently categorized using approximate rules, and the encapsulation and adsorption of lipophilic pesticides by the high-viscosity gelatin matrix is ​​not fully reflected. Furthermore, if a supplier with fewer supply frequency experiences a single anomaly, their historical anomaly rate is easily amplified. These biases collectively alter the order in which different batches are investigated for anomalies in the finished product. For example, batches containing recalcitrant or highly residual pesticides, which actually contribute significantly, may be prioritized, while batches with fewer supply frequency and a seemingly higher anomaly rate are inspected first. Consequently, in the on-site handling of pesticide residue anomalies in multi-source mixed finished products, investigators are prone to first sealing and re-inspecting non-primary sources, while the truly hazardous raw materials may continue to be added or enter other finished products, leading to a prolonged inability to pinpoint the source of the anomaly and delaying production release, supply disruption, and recall decisions. Summary of the Invention

[0004] To address the technical problems in the traceability of pesticide residues in donkey-hide gelatin products, such as the lack of comprehensive consideration of differences in pesticide conduction along processing pathways, the retention effect of high-viscosity colloids, and the deviation of small sample supplier anomaly rates, which leads to distorted ranking of the risk contribution of source batches and difficulty in timely identification of major risk-causing batches, this invention provides a method for tracing pesticide residue detection records in donkey-hide gelatin products. This method includes steps S1-S4: S1. Obtain production traceability data; production traceability data includes the detection concentration of pesticide varieties, the logarithm of the octanol-water partition coefficient, the mass input, the dynamic viscosity value, and the number of abnormal pesticide residue batches and the total number of supplied batches from suppliers; construct a directed acyclic graph (DAG) for the production traceability of donkey-hide gelatin products. The DAG uses the source raw material batch as the starting node, the intermediate cooking and mixing process as the intermediate node, and the finished product batch as the ending node. Directed edges are established between adjacent nodes according to the material flow direction, with the corresponding mass input as the weight. Each starting node to the ending node has a unique flow path; S2. For each starting node, based on the weight of the directed edge and the survival rate of each pesticide variety after physical degradation at the intermediate node, calculate... Calculate the corresponding conduction retention rate; based on the detected concentration and conduction retention rate, calculate the theoretical retention concentration of each pesticide variety at the termination node; S3, based on the logarithm of the dynamic viscosity value and the octanol-water partition coefficient, determine the retention correction coefficient for each pesticide variety; based on the preset standard reference concentration and retention correction coefficient, weight and correct the theoretical retention concentration to obtain the equivalent risk retention amount at each starting node; S4, based on the number of batches with abnormal pesticide residues and the total number of supply batches corrected by the preset minimum sample size lower limit, determine the reputation constraint value; based on the equivalent risk retention amount and reputation constraint value, calculate the risk attribution probability at each starting node, determine high-risk batches based on the risk attribution probability, and generate targeted investigation instructions in reverse along the directed acyclic graph.

[0005] This invention integrates production traceability data, including information such as detection concentration and logarithmic value of octanol-water partition coefficient, and establishes a directed acyclic graph for the traceability of donkey-hide gelatin products. Based on this, the theoretical retention concentration from each starting node to the ending node is calculated. Then, the retention correction coefficient is derived by combining the dynamic viscosity value to obtain the equivalent risk retention amount. Finally, the reputation constraint value obtained after lower limit correction is integrated to calculate the probability of risk attribution to generate targeted investigation instructions. Considering that the raw materials in donkey-hide gelatin production undergo multiple boiling and dispersion processes, simple terminal detection is difficult to trace back to the source. This method eliminates the interference of purely physical transfer at intermediate nodes and reduces the residual assessment distortion caused by colloidal matrix encapsulation by evaluating the spatial steric hindrance and molecular affinity of high-viscosity fluids. It integrates objective physical evolution with supplier quality performance, reduces the source location error in multi-source mixed flow scenarios, and reduces the production capacity loss caused by complete line shutdown.

[0006] Preferably, the step of calculating the corresponding conduction retention rate based on the weight of the directed edge and the survival rate of each pesticide variety after physical degradation at the intermediate node includes: dividing the weight of the directed edge by the total mass input of all material batches to the corresponding intermediate node to obtain the physical dilution; multiplying the physical dilution by the survival rate of each pesticide variety after physical degradation at the intermediate node to obtain the single-node retention rate of each pesticide variety; and multiplying the single-node retention rates of all intermediate nodes along the material flow direction to obtain the conduction retention rate of each pesticide variety.

[0007] This invention divides the weight of the directed edge by the total mass input of all material batches to obtain the physical dilution. It then combines the survival rate of the intermediate node after physical degradation to deduce the single-node retention rate and thus the conduction retention rate. Considering the frequent mixing of multiple batches of raw materials in the continuous production of donkey-hide gelatin, this conversion separates the actual contamination ratio of a single starting node to the finished product, eliminates the interference of concentration decay caused by non-source factors, ensures that the evaluation benchmark is consistent with the objective material conduction state, and reduces the misjudgment of responsibility caused by large differences in mixing ratio.

[0008] Preferably, the step of calculating the theoretical retention concentration of each pesticide variety to the termination node based on the detected concentration and the conduction retention rate includes: multiplying the detected concentration of each pesticide variety by the conduction retention rate of the corresponding flow path from the starting node to the termination node to obtain the theoretical retention concentration of the corresponding pesticide variety.

[0009] Preferably, determining the retention correction coefficient for each pesticide variety based on the dynamic viscosity value and the logarithm of the octanol-water partition coefficient includes: calculating the ratio of the maximum dynamic viscosity value at each intermediate node of the corresponding flow path to the standard reference viscosity constant to obtain the viscosity ratio; for each pesticide variety, multiplying the viscosity ratio, the preset molecular affinity scaling scalar, and the logarithm of the octanol-water partition coefficient of the corresponding pesticide variety, and then adding the preset reference constant to obtain the retention correction coefficient for each pesticide variety.

[0010] Preferably, the step of weighting and correcting the theoretical retention concentration based on the preset standard reference concentration and retention correction coefficient to obtain the equivalent risk retention amount at each starting point includes: dividing the theoretical retention concentration of each pesticide variety by the corresponding standard reference concentration to obtain the relative hazard of each pesticide variety; and multiplying the retention correction coefficient of each pesticide variety by the relative hazard and summing the results to obtain the equivalent risk retention amount at each starting point.

[0011] This invention derives the relative hazard level by comparing the theoretical retention concentration with the standard reference concentration, and derives the equivalent risk retention level by combining the retention correction coefficient. Considering that the toxicity levels of various pesticides vary greatly, the direct summation of concentrations cannot objectively reflect the degree of risk. This conversion completes the mapping from absolute physical concentration to dimensionless risk multiple, enabling multiple trace pesticide residues to be evaluated under a unified safety standard, reducing the probability of missing high-risk batches caused by low concentrations of a single detected variety but overall toxicity accumulation.

[0012] Preferably, determining the reputation constraint value based on the number of batches with abnormal pesticide residues and the total number of supplied batches corrected by a preset minimum sample size limit includes: selecting the maximum value between the total number of supplied batches and the preset minimum sample size constraint constant as the correction denominator base; dividing the number of batches with abnormal pesticide residues by the correction denominator base to obtain the correction defect rate; and multiplying the correction defect rate by the reputation constraint coefficient and adding the benchmark constant to obtain the reputation constraint value.

[0013] This invention selects the maximum value between the total number of supply batches and the minimum sample size constraint constant as the base of the correction denominator to calculate the correction defect rate and thus obtain the reputation constraint value. Considering the objective phenomenon that small and medium-sized suppliers are easily affected by a single detection event due to their low supply frequency, resulting in abnormally high frequency, this mechanism makes the reputation estimation of small batch suppliers tend to be stable through lower limit correction, reduces the excessive deviation of the evaluation results caused by occasional fluctuations of small samples, and balances the statistical reliability of the historical quality performance of suppliers of different sizes.

[0014] Preferably, the step of calculating the risk attribution probability of each starting node based on the equivalent risk retention and reputation constraint value includes: calculating the comprehensive risk score of each starting node based on the equivalent risk retention and reputation constraint value; dividing the comprehensive risk score of each starting node by the sum of the comprehensive risk scores of all starting nodes associated with the same termination node to obtain the risk attribution probability of each starting node.

[0015] This invention calculates a comprehensive risk-causing score based on equivalent risk retention and reputation constraint value, and calculates the risk-causing attribution probability based on the sum of the associated comprehensive risk-causing scores. Considering that absolute scores are difficult to intuitively reflect the exact pollution proportion of a specific starting node in the entire finished product, this normalization mechanism maps isolated risks to relative responsibility shares within the same constituent system, ensuring that starting nodes from different sources have comparability at the same order of magnitude in exceedance events, and providing a rigorous reference benchmark for performing retrospective investigations along directed acyclic graphs.

[0016] Preferably, the step of calculating the comprehensive risk-causing score for each starting node based on the equivalent risk retention amount and the credit constraint value includes: multiplying the equivalent risk retention amount and the credit constraint value to obtain the comprehensive risk-causing score.

[0017] Preferably, the step of determining high-risk batches based on the risk attribution probability and generating targeted investigation instructions in reverse along the directed acyclic graph includes: sorting the risk attribution probabilities of each starting node in descending order; marking the starting nodes whose risk attribution probabilities are greater than a preset risk threshold as high-risk batches; and extracting the flow path nodes of the high-risk batches in reverse along the directed acyclic graph to generate targeted investigation instructions.

[0018] Preferably, the method further includes dynamically adjusting the preset risk threshold: when the number of starting nodes constituting the finished product batch is greater than a set node number threshold, the preset risk threshold is lowered; when the number of starting nodes constituting the finished product batch is less than a set node number threshold, the preset risk threshold is raised.

[0019] The beneficial effects of this invention are as follows: This invention establishes a directed acyclic graph for the traceability of donkey-hide gelatin products and calculates the theoretical retention concentration. Considering the reality that raw materials undergo multiple dispersed cooking processes, it eliminates purely physical transfer interference in the process network and solves the tracking gap problem caused by frequent cross-mixing in continuous batch processes. This invention obtains the equivalent risk retention amount based on the logarithmic values ​​of dynamic viscosity and octanol-water partition coefficient. Considering the steric hindrance of fluids and the lipophilic adsorption law of molecules, it compensates for the thermal escape obstruction error caused by the retention of polymer colloidal matrix, achieving a unified scale assessment of different mixed pesticide residues. This invention integrates the equivalent risk retention amount with the reputation constraint value corrected to the lower limit to calculate the probability of risk attribution. Considering the statistical limitations of the supply frequency of small and medium-sized suppliers, it reduces the assessment bias caused by occasional fluctuations in small samples, achieving targeted traceability of responsibility under complex production traceability topologies. Attached Figure Description

[0020] Figure 1 A flowchart illustrating a method for recording and tracing pesticide residues in donkey-hide gelatin products, provided in an embodiment of the present invention; Figure 2 This is a schematic diagram of the topological structure of the directed acyclic graph for tracing the production of donkey-hide gelatin products provided in this embodiment of the invention. Detailed Implementation

[0021] This invention provides a method for tracing and recording pesticide residues in donkey-hide gelatin products, such as... Figure 1 As shown, the method includes steps S1-S4: Step S1: Obtain production traceability data and construct a directed acyclic graph.

[0022] It should be noted that the production process of donkey-hide gelatin products has a unidirectional flow characteristic, and multiple batches of raw materials are mixed and separated in different processes. Therefore, this invention preferably uses a directed acyclic graph (DAG) to construct this data structure. A DAG is a classic topological graph theory model that represents the unidirectional dependencies between elements through a set of nodes and directed edges, and does not contain any closed loops. This invention chooses a DAG precisely because it possesses both the ability to describe unidirectional flow and the ability to express complex topological relationships in the multi-source mixing and traceability scenario of donkey-hide gelatin products. It can accurately express the flow dependencies between source raw material batches, intermediate cooking and mixing process nodes, and finished product batches, while ensuring that the material traceability path does not have logical loops. This guarantees the stable analysis of the convergence relationship of multi-source raw materials and avoids reverse confusion caused by multi-level flow.

[0023] Specifically, this invention constructs a directed acyclic graph for the traceability of donkey-hide gelatin product production, acquiring the batch input records of raw materials at the source, the mixing ratio data of each process node, and the filing information of finished product batches during the production process; constructing a node set by taking each batch of raw materials as the starting node, the intermediate boiling and mixing processes as intermediate nodes, and the finished product batch as the ending node; establishing directed edges according to the material flow direction, and using the mass input of each batch of raw materials in the mixing process as the weight of the corresponding directed edge, thus generating a directed acyclic graph for the traceability of donkey-hide gelatin product production.

[0024] It should be noted that constructing a high-fidelity traceability map to support the subsequent numerical calculation of pesticide residue transmission models requires not only topological information but also the joint input of multi-dimensional production and quality inspection data. Specifically, background pesticide residue data serves as the initial boundary condition for the transmission model calculation, determining the initial concentration of pesticide varieties in each source raw material; process operation data records thermal processing intensity parameters, providing a direct basis for calculating the degree of pesticide thermal degradation; material topological data carries information on mixing ratios and mass inputs, forming the basis for the allocation of directed edge weights and the estimation of batch component proportions within the graph structure; material property parameters reflect the physical encapsulation effect of the colloidal matrix and the affinity binding characteristics of pesticide molecules, used to correct the matrix effect for the thermal degradation coefficient in an ideal solution system; and historical data on abnormal pesticide residues from suppliers provides a priori probability reference for attribution analysis at the end of the traceability chain. Based on this, this invention collects the above five types of data.

[0025] Specifically, regarding the baseline data on pesticide residues, a liquid chromatography-mass spectrometry (LC-MS) instrument deployed at the raw material warehousing quality inspection stage is used to simultaneously detect multiple pesticide residues in representative samples at each starting point. The total number of pesticide varieties detected in each batch and the detection concentration of each pesticide variety are recorded. The baseline data on pesticide residues includes the total number of pesticide varieties corresponding to each starting point, pesticide variety number, variety name, and corresponding detection concentration.

[0026] Based on the process operation data, the heating time and process temperature of each batch of material at each node are read using heating kettle timers and temperature sensors distributed at each intermediate node. The process operation data includes the heating time and process temperature at each intermediate node.

[0027] For material topology data, the process flow records of the production execution system are used to extract the set of all intermediate nodes that each starting node passes through to reach the ending node. The total quantity of all material batches input to each intermediate node is collected by the workshop material feeding and barcode scanning terminal. The mass input of each material batch input to the current intermediate node is obtained in real time by the hopper weighing sensor at each intermediate node. Combined with the feeding rule data provided by the formula ratio system, the mass input of materials including the components of each starting node to each intermediate node is obtained by reverse conversion through mass conservation. The material topology data includes the complete flow path node set of each starting node, the mass input of each material batch at each intermediate node, and the mass input of each component of each starting node after reverse conversion.

[0028] Regarding material properties, for each starting node, the dynamic viscosity data of the batch of material is collected in real time along the flow path from the starting node to the ending node using online rotational viscometers distributed at each intermediate node. Simultaneously, a list of all pesticide varieties detected during the incoming quality inspection at each starting node is retrieved, and the octanol-water partition coefficient logarithmic value is obtained from a standard chemical property database based on the variety name and number. The material properties include the dynamic viscosity data at each intermediate node and the octanol-water partition coefficient logarithmic value for each pesticide variety.

[0029] For the supplier's historical data on abnormal pesticide residues, the supply chain management system is used to extract the total number of batches supplied by the supplier corresponding to each starting node in the past year and the number of batches with abnormal pesticide residues detected. The supplier's historical data on abnormal pesticide residues includes the total number of batches supplied by the supplier corresponding to each starting node and the number of batches with abnormal pesticide residues.

[0030] like Figure 2 As shown, this diagram is a topological schematic of a directed acyclic graph (DAG) for tracing the production of donkey-hide gelatin products. The DAG in the diagram consists of a set of nodes and a set of directed edges. The set of nodes includes three types: starting nodes, intermediate nodes, and ending nodes, corresponding to the source raw material batches, the boiling and mixing processes, and the finished product batches, respectively. The source raw material batches are represented in the diagram as batch one to batch four. The directed edges are established according to the flow direction of materials in actual production. Directed edges with arrows indicate that materials flow from the previous process node to the next process node. Multiple directed edges converging into the same intermediate node indicate that multiple batches of raw materials are mixed at that point. Branching off from the intermediate node downstream indicates that materials continue to flow. Each starting node eventually converges into the ending node through a series and parallel combination of several intermediate nodes. Figure 2The system also displays the mass input of each batch of raw materials in the mixing process. The input directed edges of each mixing process node are labeled with numerical values. The system also labels the collection and input relationships of five types of data: background pesticide residue data, process operation data, material topology data, material property parameters, and supplier pesticide residue anomaly historical data, respectively, for the liquid chromatography-mass spectrometry, temperature sensor, hopper weighing sensor, online rotational viscometer, and supply chain management system. All of the above data are input into the directed acyclic graph construction module to assign complete attribute information to each node and directed edge.

[0031] Thus, a directed acyclic graph and production traceability data for donkey-hide gelatin products were obtained.

[0032] Step S2: Calculate the theoretical residual concentration of each pesticide variety at the termination point.

[0033] It should be noted that in the heat processing of donkey-hide gelatin products, if the system is assumed to be an ideal free solution system, pesticide residues at the initial stage will undergo thermal degradation within a specific heating time. Since the thermal stability of different pesticide varieties varies significantly, in order to accurately track residue levels, it is necessary to assign an independent thermal degradation coefficient to each specific pesticide variety detected at the initial stage, perform cumulative multiplication along a directed acyclic graph, and establish a calculation model for the theoretical residual concentration of each pesticide variety.

[0034] It should be further noted that the degradation behavior of pesticides during thermal processing is recognized in environmental chemistry and food science as following first-order reaction kinetics, and its mathematical expression is: ,in The degradation rate constant is For reaction time, It is through time After the pesticide residue concentration, This represents the initial pesticide residue concentration. This fundamental physical model originates from the integral form of the Arrhenius equation under isothermal conditions, i.e. ,in It is activation energy. It is the product of thermodynamic temperature and the gas constant. This refers to the pre-factor; therefore, an expression for the change of residual concentration over time can be derived. Although, strictly speaking, the specific physical thermal degradation coefficient should be written as a function of temperature as the independent variable if it changes with temperature, in actual donkey-hide gelatin processing, the heating temperature of a single boiling process is kept constant. Therefore, using the specific physical thermal degradation coefficient at this constant temperature as the empirical rate constant for this process is a standard and universally accepted engineering simplification practice.

[0035] First, a basic parameter set must be acquired and established. It is important to note that to ensure the first-order reaction kinetic model can accurately execute in actual production traceability and reflect the true physical degradation laws, core calculation parameters consistent with the actual physical environment of the process must be accurately acquired beforehand. Directly using general parameters to replace data from specific intermediate nodes and complex mixing states will fail to accurately reflect the actual dilution and thermal degradation of materials under specific process formulations and temperatures, leading to significant deviations in the final residual estimate.

[0036] Specifically, for each intermediate process temperature, a control sample containing pesticide standards and a simulated donkey-hide gelatin matrix was prepared. This control sample consisted of a base liquid in which donkey-hide gelatin raw materials were mixed according to a standard ratio and reached the equivalent viscosity and component state of that process temperature. The sample was then placed in a constant-temperature reactor set at the intermediate process temperature for thermal degradation experiments. Residual concentrations were sampled and detected at preset time points, and the degradation was analyzed according to a first-order reaction kinetic curve. Perform fitting; where, This is the initial pesticide residue concentration. It is through time After the pesticide residue concentration, The reaction time is used. The specific physical-thermal degradation coefficients of each pesticide variety at this node temperature are obtained through fitting, and the calibration results are pre-written into the standard process parameter library. Secondly, for the mixed material, the mass input of specific starting node components is inferred through a traceable directed acyclic graph: based on the feed formulation ratio at each node and the mixing weight of each upstream batch, the converted mass of each starting node component in the input material is determined.

[0037] Then, the theoretical residual concentration of each pesticide variety at each starting node to the ending node is calculated. It should be noted that in the actual continuous production scenario of donkey-hide gelatin products, the final product is often formed by the complex mixing of multiple starting nodes from different sources through a directed acyclic graph and multiple heating processes. Relying solely on terminal testing data at the finished product end will make it impossible to trace the specific source of pollution once pesticide residues are found to exceed standards. Because the physical dilution ratios and specific heating times experienced by materials at different starting nodes vary significantly along their flow paths, a simple linear mean allocation cannot accurately reflect the true pollution contribution of each starting node to the finished product. Therefore, to achieve accurate targeted risk attribution, it is necessary to independently calculate the theoretical residual concentration of each pesticide variety at each starting node along its unique flow path to the ending node. This isolates the independent residual contribution of each pesticide variety at a single starting node under the dual physical effects of dilution and degradation, providing a concentration table with granular detail down to the individual variety level for subsequent pesticide-specific toxicity weighting and normalization summarization.

[0038] Based on the above logic, for the directed acyclic graph of donkey-hide gelatin product production traceability, starting from the first... The first starting node The theoretical residual concentration of each pesticide variety at the termination point satisfies the following relationship: ; in, It is the first The first starting node The theoretical residual concentration of each pesticide variety at the termination point; It is the first When the first node is entered into the database, the first The detection concentration of each pesticide variety; It is the multiplication operator; It is the first The set of all intermediate nodes from the starting node to the ending node; It is the flow path The first in One intermediate node; Is it including the first The material input from the starting node is sent to the intermediate node. Quality input quantity; It is input to intermediate nodes The total quantity of all material batches; It is the first Each material batch is entered into the intermediate node. Quality input quantity; It is the first The pesticide varieties are at the intermediate node The specific physical and thermal degradation coefficient; It is an intermediate node The duration of heating.

[0039] In this relation, Reflects the material at the current node The degree of physical dilution increases monotonically as the mass input of a specific material increases. This reflects the theoretical decay rate of a specific pesticide at a given temperature; as the heating time is extended, this term decreases monotonically. Reflecting the percentage of survivors after physical degradation, over time Increase or degradation coefficient As it increases, this term monotonically decreases. This reflects the cumulative effect of dilution and degradation in a single process. This term monotonically increases as the physical dilution level at a single node increases. The product of the cumulative effects of dilution and degradation in a single process along the entire path reflects the composite transport retention rate along multiple nodes in the directed acyclic graph. This term monotonically decreases as the number of nodes in the transport path increases. The product of the detected concentration and the aforementioned composite transport retention rate reflects the theoretical retention concentration of the current pesticide variety transported to the finished product. This term monotonically increases as the detected concentration increases. The above relationships output the independent theoretical retention concentration of a single pesticide variety, maintaining the separation between pesticide varieties and providing input for subsequent weighted toxicity calculations and aggregation for each variety.

[0040] Thus, the theoretical retention concentration of each pesticide variety from the starting node to the ending node was obtained.

[0041] Step S3: Calculate the equivalent risk retention of each starting node.

[0042] It should be noted that during the actual processing of donkey-hide gelatin, the dynamic viscosity of the material changes drastically as collagen continues to hydrolyze and concentrate, forming a complex colloidal matrix. This high-viscosity colloidal matrix produces a physical encapsulation effect, hindering the thermal escape and diffusion of pesticide molecules.

[0043] It should be further noted that the toxicity limits of different pesticide varieties vary greatly. Directly adding the theoretical retention concentrations of each variety would lack toxicological basis and fail to accurately reflect the actual risk contribution of each variety. Furthermore, a single-source batch often contains multiple pesticide residues. Performing independent high-order differential continuous calculations of the matrix adsorption effect for each pesticide would lead to computational overload and easily introduce cumulative errors. Therefore, this invention applies toxicity weighting and retention correction to the theoretical retention concentrations of each pesticide variety, and then summarizes them into an equivalent risk retention amount that reflects the overall risk level of a single batch.

[0044] Based on the above logic, the first The equivalent risk retention of each starting node satisfies the following relationship: ; in, It is the first The equivalent risk retention of each starting node; It is the first The first starting node The theoretical residual concentration of each pesticide variety at the termination point; It is the first The total number of pesticide varieties in each starting node; It is the first The standard reference concentration for each pesticide variety is preferably taken as the maximum residue limit of the corresponding pesticide variety in the food safety standard. It is the first Each starting node in the flow path The maximum value of the dynamic viscosity data of each intermediate node is denoted as the maximum dynamic viscosity value; It is the standard reference viscosity constant, and the preferred value is the absolute value of the dynamic viscosity of pure water at room temperature; It is a molecular affinity scaling scalar; It is the first The first starting node Logarithmic values ​​of octanol-water partition coefficients for each pesticide variety.

[0045] In this relationship, for each pesticide variety, This reflects the additional residue ratio resulting from the affinity entrainment of specific pesticide molecules by the adhesive. This term increases monotonically as the maximum dynamic viscosity or the logarithm of the octanol-water partition coefficient increases. This is the retention correction factor for this pesticide variety; This represents the relative hazard of a single pesticide variety, completing the dimensionless conversion from absolute concentration to risk multiple. This term monotonically increases when the theoretical retention concentration increases or the legal limit decreases. Multiplying the two values ​​reflects the actual hazard contribution of the pesticide variety after being amplified by retention in complex matrices. This term monotonically increases when the retention correction coefficient or relative hazard increases. Summing the actual hazard contributions of all pesticide varieties yields the equivalent risk retention at the starting point, reflecting the total risk equivalent released from a single batch of mixed pesticide residues at the source to the terminal. This term monotonically increases when the hazard contribution of various pesticides increases or the number of detected types increases.

[0046] The construction of this relationship originates from a combined deduction of the thermodynamic diffusion-limited theory of polymer solutions and the principle of "like dissolves like". In the actual process of preparing donkey-hide gelatin, as water evaporates and collagen macromolecules crosslink, the matrix transforms from a Newtonian fluid to a non-Newtonian fluid, and the dense spatial network structure formed inside generates real physical steric hindrance for free pesticide molecules. Therefore, the ratio of the maximum dynamic viscosity to the reference viscosity is constructed as a physical encapsulation constraint condition. At the same time, since the stronger the lipophilicity of pesticide molecules, the more stable their van der Waals forces and hydrophobic interactions with the hydrophobic cavities of proteins, making them more difficult to volatilize with water vapor, the logarithmic value of the octanol-water partition coefficient of each pesticide variety is introduced as a chemical adsorption constraint condition, so that the retention correction is implemented at the specific molecular level. Based on this, the theoretical retention concentration was dimensionlessly processed using the maximum residue limit of food safety standards for each pesticide variety as the denominator, so that pesticides with different toxicity levels could be risk-summed on the same scale. In addition, the positive compensation bias brought about by both physical and chemical factors was added to evaluate the phenomenon of pesticide residue escape obstruction caused by complex colloidal matrix environment, and corrected the limitations of relying solely on temperature and time for decay calculation.

[0047] It should be noted that molecular affinity scaling is a scalar. The value needs to be set according to the purification level of the raw materials at the starting point of the donkey-hide gelatin production and the depth of impurity removal in the intermediate processes: for raw material batches that have not been finely filtered and have a high content of colloidal impurities, the matrix has a strong non-specific adsorption effect on pesticides, and the value can be appropriately increased. If we assume The value was set to 0.015 to improve the sensitivity of the retention increment calculation; for raw material batches that have undergone multi-stage centrifugation purification and have relatively simple matrix components, the macromolecular encapsulation effect is relatively weakened, and the value can be appropriately reduced. If we assume The value is set to 0.008 to avoid over-amplifying the correction factor. This is to ensure the non-negativity of the equivalent risk retention and the physical meaning of the retention correction. Take a value greater than or equal to 0; for the first... The first starting node For each pesticide variety, the retention correction factor should meet the following requirements. ;when hour, The values ​​should be set by combining the maximum dynamic viscosity and the octanol-water partition coefficient, and should meet the following requirements. .when When taking consistent values ​​across different starting points and pesticide varieties within the same finished product batch, It should be less than all that satisfy The minimum value among the upper limits corresponding to the pesticide varieties. Under the constraint that the retention correction coefficient is positive, in this embodiment, The preferred value is 0.01.

[0048] At this point, the equivalent risk retention of each starting node has been obtained.

[0049] Step S4: Calculate the probability of risk attribution and generate targeted investigation instructions.

[0050] It should be noted that a single concentration contribution value cannot fully reflect the true attribution of supply chain quality responsibility. When using the historical frequency of supplier anomalies to assess quality reputation, small and medium-sized suppliers are prone to having abnormally high frequencies due to their smaller total number of supply batches, such as supplying only once with only one anomaly. Therefore, a minimum sample size constraint constant needs to be introduced to reconstruct the historical reputation index, combining the equivalent risk retention with the reputation constraint value to form a comprehensive risk-causing score, thereby outputting the probability of risk attribution.

[0051] First, calculate the comprehensive risk score for each individual starting point. It should be noted that the equivalent risk retention already incorporates multiple factors such as physical dilution, thermal degradation, colloidal retention, and toxicity differences, but it does not yet consider the supplier's historical quality reputation. Even if different suppliers have the same equivalent risk retention, their actual risk levels can vary significantly due to differences in their past pesticide residue detection records. Therefore, it is necessary to overlay the supplier's historical abnormal frequency reputation constraint onto the equivalent risk retention to obtain the comprehensive risk score, which serves as the basis for subsequent liability determination.

[0052] Based on the above logic, the overall risk score satisfies the following relationship: ; in, It is the first The overall risk score of each starting node; It is the first The equivalent risk retention of each starting node; It is the historical abnormal reputation constraint coefficient; It is the first Each starting node corresponds to the number of batches of the supplier that were found to have abnormal pesticide residues in the past year. This refers to the total number of batches supplied by the supplier over the past year; It is a preset minimum sample size constraint constant; It is a function that takes the maximum value.

[0053] In this relation, Reflects the denominator base after forced sample lower limit correction, when the actual number of deliveries... When the threshold is exceeded, it increases synchronously with the threshold; otherwise, it remains constant. Reflecting the corrected defect rate after eliminating small sample bias, when the number of abnormal batches... It increases monotonically as it increases. The weight bias, which reflects the overall historical performance, increases monotonically as the frequency of constraints increases. The overall relationship reflects the overall risk score after supplier reputation weighting. As the equivalent risk retention or reputation constraint value increases, the overall risk score increases monotonically.

[0054] The construction of this relation stems from a joint consideration of statistical reliability and physical risk equivalent in supplier quality reputation assessment. When the total number of batches supplied by a supplier is small, its anomaly frequency is easily affected by extreme samples and deviates from the true quality level. Therefore, a minimum sample size constraint constant is introduced to lower-limit the denominator of the frequency calculation, making the frequency estimation of small-batch suppliers more conservative and avoiding over-constraint due to accidental anomalies. At the same time, the historical anomaly frequency should work synergistically with the risk equivalent of the current batch. Suppliers with higher anomaly frequencies are more likely to have batches with potential quality problems. Therefore, the frequency term is used as a constraint factor to weight and amplify the equivalent risk retention. The constraint strength is determined by the coefficient. control.

[0055] It should be noted that the historical anomaly reputation constraint coefficient and minimum sample size constraint constant The value needs to be set based on the statistical characteristics of the actual supply chain quality data. The weighted constraint strength used to control the frequency of historical anomalies on the equivalent risk retention is typically between 0.3 and 0.8. A larger value can be used when the supplier's historical anomalies are highly indicative of quality risk, and a smaller value can be used when the equivalent risk retention itself has sufficient explanatory power. In this embodiment... The preferred value is 0.5. The lower bound clamping is used to calculate the anomaly frequency. Physically, this refers to the minimum sample size that can statistically reliably estimate the true anomaly rate of suppliers. If the value is too small, the sample correction effect will be insignificant; if the value is too large, it may excessively suppress the true anomaly signals of small and medium-sized suppliers. It is typically set to an integer between 5 and 15. In this embodiment... The preferred value is 10.

[0056] Based on this, the probability of risk attribution for each starting point is calculated. It should be noted that the absolute value of the overall risk score for a single batch only reflects its own overall risk potential. However, the truly effective reference for judging whether a batch is high-risk within the overall composition of finished product batches is the proportion of that batch's score in the sum of all relevant batch scores. Only by placing batch scores within a global context for relative comparison can the liability proportion of each batch in the final product be reflected, providing a unified decision-making benchmark for subsequent risk ranking.

[0057] Specifically, the overall risk-causing score of each starting node is divided by the sum of the overall risk-causing scores of all starting nodes constituting the finished product batch. The resulting ratio is the risk-causing probability of each starting node, and the sum of the risk-causing probabilities of all starting nodes is 1.

[0058] Finally, all calculated attribution probabilities are arranged in descending order, and batches with probabilities exceeding a preset risk threshold are marked as high-risk batches. Targeted screening instructions are then generated in reverse along the directed acyclic graph. The preset risk threshold value needs to be set based on the number of starting nodes involved in the finished product batch and the actual judgment criteria for pesticide residue testing: when the number of starting nodes constituting the finished product batch is large, such as exceeding 20, the average attribution probability of each batch naturally decreases, and the threshold can be appropriately lowered to 0.2 to avoid missing high-risk batches due to an excessively large denominator; when the number of starting nodes is small, such as less than 10, the probability distribution of each batch is relatively concentrated, and the threshold can be appropriately raised to 0.3 to avoid over-marking and thus dispersing screening resources. Furthermore, for pesticide varieties with high toxicity or high detection frequency, the risk threshold can be temporarily lowered to 0.2 for the corresponding batches to improve screening sensitivity; for varieties with low toxicity or few previous detection records, the threshold can be raised to 0.3 to reduce the false alarm rate. In this embodiment, considering the range of typical starting points involved in the production of donkey-hide gelatin products and the requirements for pesticide residue risk control, the risk threshold is preferably 0.25. The risk threshold ranges from 0 to 1.

[0059] At this point, the probability of risk attribution for each starting node and the targeted investigation instructions corresponding to high-risk batches have been obtained.

[0060] The above are all preferred embodiments of the present invention and are not intended to limit the scope of protection of the present invention. Therefore, all equivalent changes made in accordance with the structure, shape and principle of the present invention should be covered within the scope of protection of the present invention.

Claims

1. A method for recording and tracing pesticide residues in donkey-hide gelatin products, characterized in that, include: S1. Obtain production traceability data; production traceability data includes the detection concentration of pesticide varieties, the logarithm of the octanol-water partition coefficient, the mass input, the dynamic viscosity value, and the number of abnormal pesticide residue batches and the total number of supplied batches from suppliers; construct a directed acyclic graph (DAG) for the production traceability of donkey-hide gelatin products. The DAG uses the source raw material batch as the starting node, the intermediate cooking and mixing process as the intermediate node, and the finished product batch as the ending node. Directed edges are established between adjacent nodes according to the material flow direction, with the corresponding mass input as the weight. Each starting node to the ending node has a unique flow path; S2. For each starting node, based on the weight of the directed edge and the survival rate of each pesticide variety after physical degradation at the intermediate node, calculate... Calculate the corresponding conduction retention rate; based on the detected concentration and conduction retention rate, calculate the theoretical retention concentration of each pesticide variety at the termination node; S3, based on the logarithm of the dynamic viscosity value and the octanol-water partition coefficient, determine the retention correction coefficient for each pesticide variety; based on the preset standard reference concentration and retention correction coefficient, weight and correct the theoretical retention concentration to obtain the equivalent risk retention amount at each starting node; S4, based on the number of batches with abnormal pesticide residues and the total number of supply batches corrected by the preset minimum sample size lower limit, determine the reputation constraint value; based on the equivalent risk retention amount and reputation constraint value, calculate the risk attribution probability at each starting node, determine high-risk batches based on the risk attribution probability, and generate targeted investigation instructions in reverse along the directed acyclic graph.

2. The method for recording and tracing pesticide residues in donkey-hide gelatin products according to claim 1, characterized in that, The calculation of the corresponding transport retention rate based on the weights of directed edges and the survival rate of each pesticide variety after physical degradation at the intermediate node includes: Divide the weight of the directed edge by the total mass input of all material batches to the corresponding intermediate node to obtain the physical dilution. The physical dilution rate is multiplied by the survival rate of each pesticide variety after physical degradation at the intermediate node to obtain the single-node retention rate of each pesticide variety. The single-node retention rate of each pesticide variety is obtained by multiplying the single-node retention rate of all intermediate nodes along the material flow direction.

3. The method for recording and tracing pesticide residues in donkey-hide gelatin products according to claim 1, characterized in that, The calculation of the theoretical retention concentration of each pesticide variety at the termination point based on the detected concentration and the conduction retention rate includes: The theoretical retention concentration of each pesticide variety is obtained by multiplying the detected concentration by the retention rate of the corresponding flow path from the starting node to the ending node.

4. The method for recording and tracing pesticide residues in donkey-hide gelatin products according to claim 1, characterized in that, The determination of retention correction coefficients for each pesticide variety based on the logarithm of dynamic viscosity and octanol-water partition coefficient includes: The viscosity ratio is obtained by calculating the ratio of the maximum value of the dynamic viscosity at each intermediate node of the corresponding flow path to the standard reference viscosity constant. For each pesticide variety, the retention correction coefficient is obtained by multiplying the viscosity ratio, the preset molecular affinity scaling scalar value, and the logarithmic value of the octanol-water partition coefficient of the corresponding pesticide variety, and then adding the preset reference constant.

5. The method for recording and tracing pesticide residues in donkey-hide gelatin products according to claim 1, characterized in that, The theoretical retention concentration is weighted and corrected based on a preset standard reference concentration and a retention correction coefficient to obtain the equivalent risk retention at each starting node, including: The relative hazard level of each pesticide variety is obtained by dividing the theoretical residual concentration of each pesticide variety by the corresponding standard reference concentration. The equivalent risk retention at each starting point is obtained by multiplying the retention correction coefficient of each pesticide variety by its relative hazard level and then summing the results.

6. The method for recording and tracing pesticide residues in donkey-hide gelatin products according to claim 1, characterized in that, The determination of the reputation constraint value based on the number of batches with abnormal pesticide residues and the total number of supplied batches adjusted by a preset minimum sample size limit includes: The maximum value between the total number of supply batches and the preset minimum sample size constraint constant is selected as the base of the corrected denominator; Divide the number of batches with abnormal pesticide residues by the base of the correction denominator to obtain the corrected defect rate; The credit constraint value is obtained by multiplying the corrected defect rate by the credit constraint coefficient and then adding the benchmark constant.

7. The method for recording and tracing pesticide residues in donkey-hide gelatin products according to claim 1, characterized in that, The calculation of the risk attribution probability for each starting node based on the equivalent risk retention and reputation constraint value includes: Based on the equivalent risk retention and credit constraint value, calculate the comprehensive risk score for each starting node; The risk attribution probability of each starting node is obtained by dividing the overall risk score of each starting node by the sum of the overall risk scores of all starting nodes associated with the same ending node.

8. The method for tracing and recording pesticide residues in donkey-hide gelatin products according to claim 7, characterized in that, The calculation of the comprehensive risk score for each starting node based on the equivalent risk retention and credit constraint value includes: Multiplying the equivalent risk retention amount by the credit constraint value yields the comprehensive risk score.

9. The method for recording and tracing pesticide residues in donkey-hide gelatin products according to claim 1, characterized in that, The process of identifying high-risk batches based on risk attribution probability and generating targeted investigation instructions in reverse along a directed acyclic graph includes: Arrange the risk attribution probabilities of each starting node in descending order; The starting node whose risk attribution probability is greater than a preset risk threshold is marked as a high-risk batch; Extract the flow path nodes of high-risk batches in reverse along the directed acyclic graph to generate targeted investigation instructions.

10. The method for recording and tracing pesticide residues in donkey-hide gelatin products according to claim 9, characterized in that, The method also includes dynamically adjusting the preset risk threshold: When the number of starting nodes constituting the finished product batch is greater than the set node number threshold, the preset risk threshold is lowered. When the number of starting nodes constituting the finished product batch is less than the set node number threshold, the preset risk threshold is increased.

Images