Method for carrying out risk early warning analysis on food-borne food safety by simulating propagation and diffusion of heavy metals in environment through complex network

Abstract

The invention discloses a method for carrying out risk early warning analysis on food-borne food safety by simulating propagation and diffusion of heavy metal pollution in an environment through a complex network, and the method comprises the steps: building an atmosphere-soil pollutant interaction complex network model through the complex network, and considering the propagation dynamics of heavy metals between nodes; comprising propagation behaviors such as propagation in the atmosphere and soil, atmospheric settlement and migration in the soil, a network topology structure and a propagation dynamic model between nodes are combined to form an atmosphere-soil pollutant interaction complex network model, effective combination of physical driving and data driving is realized, and more accurate and reliable risk early warning is provided. According to the method, key nodes are recognized through network topology structure analysis, attention is given preferentially in risk management and control, node food safety risks are quantified through diet exposure assessment, uncertainty caused by dependence on qualitative analysis or rough estimation is avoided, and a decision basis is provided for development of targeted food safety risk prevention and control work.

Description

Technical Field

[0001] The present invention relates to the field of food safety, and specifically is a method for conducting risk early warning analysis on food-borne food safety by constructing an atmosphere-soil pollutant interaction complex network model through a complex network to simulate the propagation and diffusion of pollutants in the environment. Background Art

[0002] In recent years, pollution problems at the top of the food supply chain have become serious, and food safety problems have occurred frequently. According to the National Land Pollution Survey Bulletin, 19.4% of cultivated land in my country has excessive heavy metal pollution. The content of heavy metals in the soil is an important indicator for environmental monitoring. The accumulation of heavy metals in the soil may lead to the introduction of heavy metals into the food chain and pose a health risk to humans. Studies have shown that heavy metals such as cadmium, chromium, arsenic and lead have been listed as clear carcinogens by the International Agency for Research on Cancer (IARC). In addition to the risk of cancer, increased concentrations of these heavy metals may also significantly increase the risk of kidney, bone and nervous system diseases, and are closely related to the occurrence of chronic diseases and cardiovascular diseases. In order to better ensure human health, it is crucial to take proactive measures such as risk warnings.

[0003] Food risk early warning refers to the use of prediction models to timely identify potential risks and hazards of food and prevent them from evolving into major problems. Currently, the early warning methods for food safety are mainly divided into qualitative analysis and quantitative data prediction methods.

[0004] Qualitative analysis methods are mainly used to draw warning conclusions from policy and theoretical analysis. This method is very common in food control research, analyzing participants' attitudes towards food safety risks, and analyzing participants' understanding of food safety issues. However, the data used for analysis mainly comes from questionnaires, reports, etc., and cannot be used to accurately build a prediction model.

[0005] The quantitative data prediction method is an early warning method based on data analysis. Researchers analyze food testing data, including time series and regression analysis results in daily monitoring networks, and analyze historical data to predict potential problems. However, this method mostly focuses on using data-driven methods such as machine learning to predict potential risks by analyzing historical data, and often ignores the physical behavior of pollutants and their propagation mechanisms in the environment, resulting in limited prediction accuracy and reliability in certain scenarios.

[0006] In contrast, the complex network approach can more comprehensively consider the propagation characteristics of pollutants in the environment, especially the physical model of pollutant diffusion and propagation. By building a model based on complex networks and combining the propagation dynamics of pollutants in different environmental media, it is possible to effectively combine physical drive and data drive, providing more accurate and reliable risk warnings.

[0007] However, the propagation behavior of pollutants in the environment is affected by many factors, and the propagation process is both time-dependent and highly spatially heterogeneous. In addition, the propagation path of pollutants usually exhibits complex network characteristics of nonlinear coupling and multi-dimensional interaction, which is difficult to accurately describe using traditional methods.

[0008] Therefore, there is an urgent need to find a method that can accurately simulate the diffusion behavior of heavy metals in the environment in order to effectively and accurately conduct risk warning analysis of foodborne food safety. Summary of the invention

[0009] In view of the above problems, in order to further effectively analyze the propagation and diffusion behavior of heavy metals in the environment, the present invention proposes a method for risk warning analysis of foodborne food safety by constructing an atmosphere-soil pollutant interaction complex network model through a complex network to simulate the propagation and diffusion of heavy metal pollution in the environment. The propagation and diffusion behavior model of heavy metals in the environment is established through a complex network, and the propagation and diffusion behavior of heavy metals in the environment is simulated by constructing a propagation dynamics model between nodes. The propagation and diffusion behavior of heavy metals in the environment is analyzed more accurately to achieve a more accurate and reliable risk warning.

[0010] Compared with the traditional method, the method of the present invention has the following beneficial effects:

[0011] ① The present invention utilizes complex networks to construct an atmosphere-soil pollutant interaction complex network model to conduct risk warning analysis on foodborne food safety. The use of complex networks can more comprehensively consider the propagation characteristics of pollutants in the environment and more clearly capture the propagation paths of heavy metals in the environment.

[0012] ② The present invention utilizes complex networks to construct an atmosphere-soil pollutant interaction complex network model to conduct risk warning analysis on foodborne food safety, constructs a dynamic model of heavy metal pollution transmission between nodes, and realizes an effective combination of physical drive and data drive.

[0013] ③ The present invention uses complex networks to construct an atmosphere-soil pollutant interactive complex network model to conduct risk warning analysis on foodborne food safety. By analyzing the characteristics of the complex network topology structure, key nodes are identified and given priority attention in risk management and control, providing a decision-making basis for carrying out targeted food safety risk prevention and control work. BRIEF DESCRIPTION OF THE DRAWINGS

[0014] Figure 1 The present invention is a flow chart of constructing an atmosphere-soil pollutant interaction complex network model through a complex network.

[0015] Figure 2 This is the node concentration conservation diagram of the soil heavy metal pollution migration-propagation network model of the present invention.

[0016] Figure 3 This is a node concentration conservation diagram of the atmospheric heavy metal pollution deposition-propagation network model of the present invention.

[0017] Figure 4 This is a complex network structure diagram of the interaction between air and soil pollutants in an embodiment of the present invention.

[0018] Figure 5 This is a performance diagram of the model prediction of an embodiment of the present invention.

[0019] Figure 6 This is a graph of concentration changes predicted by the model of an embodiment of the present invention.

[0020] Figure 7 This is a risk assessment result diagram of an embodiment of the present invention. DETAILED DESCRIPTION

[0021] The present invention will be further described in detail below with reference to the accompanying drawings and embodiments.

[0022] The present invention proposes a method for conducting risk warning analysis on food safety by constructing an atmosphere-soil pollutant interactive complex network model through a complex network to simulate the propagation and diffusion of heavy metal pollution in the environment. The method is described in two parts. The first part is how to construct an atmosphere-soil heavy metal pollution interactive complex network model, and the second part is to conduct risk warning analysis on food safety by applying the atmosphere-soil heavy metal pollution interactive complex network model constructed by the present invention.

[0023] Part I: Constructing a complex network model of air-soil heavy metal pollution interactions

[0024] In the present invention, Figure 1 This is a flowchart for constructing an interactive complex network model of air-soil heavy metal pollution. Specifically, the model is divided into two parts: soil pollutant degradation-propagation network model and air pollutant deposition-propagation network model. It simulates the propagation and diffusion process of pollutants such as heavy metals and pesticide residues in the atmosphere and soil. The specific construction steps are:

[0025] Construction step A: constructing a soil heavy metal pollution migration-propagation network model;

[0026] In the present invention, the soil heavy metal pollution degradation-propagation network model is defined as G = (E, V, θ), where the set V = {v1, v2, v3, ... v n} is a node set, with regional centers or monitoring stations as nodes; set E = {e1, e2, e3, …e m} is the set of edges between nodes. Bordering regions or adjacent monitoring stations are connected as edges; the set θ={θ1,θ2,θ3,…θ n} is the pollutant concentration of each node. Figure 2 As shown in Figure 1, a general node receives pollutants θ from adjacent nodes whose heavy metal pollutant concentration is higher than its own. in (t), and diffuse to the neighboring nodes with lower heavy metal pollutant concentration. Some leaf nodes, that is, nodes with only one edge, only transmit or receive to other neighboring nodes. represents the concentration increment caused by atmospheric deposition, For the migration of heavy metal pollutants in the soil, the mass balance of general nodes and leaf nodes can be described as:

[0027]

[0028] θ(t) is the heavy metal concentration of the node at the current moment.

[0029] θ0 is the initial heavy metal concentration of the node.

[0030] Due to the complex nonlinear propagation patterns of heavy metal pollutants in the environment, it is difficult to describe their propagation and diffusion in the environment using a single propagation dynamics model.

[0031] In the present invention, the diffusion behavior of heavy metals in soil is simulated by the combined effect of groundwater molecular diffusion and diffusion. Due to the uneven concentration of heavy metals in soil and groundwater, heavy metals always spread from high concentration areas to low concentration areas until a uniform concentration is reached. This process is called molecular diffusion and is expressed by Fick's second law.

[0032]

[0033] D x represents the diffusion coefficient in the horizontal direction.

[0034] D y represents the diffusion coefficient in the vertical direction.

[0035] In addition to molecular diffusion, the diffusion of heavy metal pollution caused by groundwater flow is also a major way of spreading in the soil. Assuming v(v x ,v y ) is the velocity of the fluid, θ i is the concentration value of node i at a certain moment, then d t The material increment in the x and y directions during time is,

[0036]

[0037] Combined with the law of conservation of mass, the concentration value changes with time as

[0038]

[0039] vx Indicates the diffusion velocity in the horizontal and vertical directions.

[0040] v y Indicates the diffusion velocity in the horizontal and vertical directions.

[0041] In addition, heavy metal pollution in the soil can be transferred to crops through the root system in the contaminated soil, leading to crop pollution. The decrease in heavy metal concentration in the soil over time is represented by first-order kinetics.

[0042]

[0043] k represents the heavy metal transfer factor, which indicates the extent of heavy metal migration from soil to crops.

[0044] Assuming that the diffusion coefficient and diffusion velocity are consistent in the horizontal and vertical directions, the propagation dynamics model of pollutants in the soil pollutant migration-propagation network model is:

[0045]

[0046] Construction step B, constructing the atmospheric heavy metal pollution deposition-propagation network model;

[0047] In the present invention, the atmospheric heavy metal pollution deposition-propagation network model is constructed as a directed network in combination with wind direction, and the average wind direction is used as the basis for judging whether the nodes are connected. In the network, the source node is the first node in the average wind direction, representing the initial release point of heavy metal pollutants. Due to the effect of wind force, the source node continuously transports and diffuses heavy metal pollutants to subsequent adjacent nodes, forming its propagation path in the network. Figure 3 As shown in Figure 1, in this process, the subsequent adjacent nodes act as intermediate nodes, similar to the general nodes in the soil heavy metal pollution migration-propagation network model, which not only receive the heavy metal pollutant concentration from the previous node, but also continue to diffuse to the subsequent adjacent nodes. out (t) represents the concentration of heavy metal pollutants reduced by atmospheric deposition. The mass conservation of the source node and the subsequent nodes can be expressed as,

[0048] <![CDATA[C(t)=C0-C out (t)-ψ out (t)]]> (8) <![CDATA[C(t)=C0+C in (t)-C out (t)-ψ out (t)]]> (9)

[0049] C(t) is the heavy metal concentration of the node at the current moment.

[0050] C0 is the initial heavy metal concentration of the node.

[0051] In the present invention, a Gaussian plume model is used to simulate the propagation and diffusion behavior between nodes in an atmospheric pollutant deposition-propagation network, and the pollutant concentration C(x, y) of any node downwind can be expressed by the Gaussian plume formula.

[0052]

[0053] Q is the static source strength, that is, the pollutant concentration at the previous node in the average wind direction.

[0054] σ y is the diffusion coefficient perpendicular to the wind direction.

[0055] y0 is the lateral position of the subsequent node relative to the previous node.

[0056] The pollutant concentration at a node with multiple source strengths can be expressed as,

[0057]

[0058] A ij is the network adjacency matrix.

[0059] In the present invention, the interaction between the soil pollutant degradation-propagation network model and the atmospheric pollutant deposition-propagation network model is represented by the deposition of pollutants in the atmosphere. Atmospheric deposition refers to the process in which pollutants suspended in the atmosphere settle to the ground or other surfaces due to gravity. In the pollutant diffusion model, deposition can significantly affect the concentration of pollutants in the soil. Generally, atmospheric deposition can be described by the deposition flux F,

[0060] <![CDATA[F=C(t)×V s ]]> (12)

[0061] V s is the deposition rate of pollutants in the atmosphere.

[0062] In summary, the propagation dynamics model of pollutants in the atmospheric pollutant deposition-propagation network model is:

[0063]

[0064] Construction step C, using optimization algorithms to optimize unknown parameters in propagation dynamics;

[0065] Since parameters such as the diffusion coefficient and diffusion velocity in the inter-node propagation dynamics model are unknown, these parameters are often difficult to measure directly or obtain through traditional methods in practical applications.

[0066] In the present invention, the meta-learning optimization algorithm of Model-Agnostic Meta-Learning (MAML) is adopted to optimize the unknown parameters in the complex network model through multi-task learning.

[0067] In the present invention, the pollutant concentration distribution of each node in each time period is first set as an independent task. Task T tIt is defined as predicting the pollutant concentration in the next time period given the pollutant concentration distribution of each node in the previous time period as input.

[0068] Second, initialize the model parameters θ, for each task T t , learn from the training data and update the model parameters through gradients (i.e., inner loop iterations) to adapt to the task.

[0069] The updated parameter is θ t '

[0070]

[0071] α is the learning rate of the inner loop,

[0072] is the loss function for the training data

[0073] After the inner loop is updated, in task T t Evaluate the performance of the model on the test data and calculate the test loss And update the model's initialization parameters θ through the test loss of all tasks

[0074]

[0075] β is the learning rate of the outer loop.

[0076] The gradient of the test loss is calculated for all tasks.

[0077] The goal of MAML is to optimize the initial model parameters θ so that the model can quickly adapt to new tasks through a small amount of gradient updates and achieve better prediction results. The optimization goal is expressed as:

[0078]

[0079] MAML learns shared knowledge in multi-task training, which improves the model's predictive ability in different pollutant propagation scenarios, thus providing strong support for the accurate prediction of pollutant concentrations.

[0080] Construction step D, constructing a complex network model of air-soil pollutant interactions;

[0081] In the present invention, the soil pollutant migration-propagation network model and the atmospheric pollutant deposition-propagation network model are connected through the deposition effect of pollutants in the atmosphere to obtain an atmosphere-soil pollutant interaction complex network model.

[0082] Constructing step E, analyzing the key nodes according to the properties of the complex network topology;

[0083] In this invention, the topological characteristics of complex networks are parameters used to describe their structural characteristics. This paper analyzes the importance of nodes in the network and the degree of mutual influence. Statistics such as degree centrality, clustering coefficient, betweenness centrality, and closeness centrality are introduced to characterize the structural characteristics of nodes.

[0084] 1) Degree Centrality

[0085] The degree of a node in a network indicates the number of edges connected to the node, and is further divided into out-degree and in-degree in a directed network. For a simple graph, the degree is the number of adjacent nodes. The larger the node degree, the higher its correlation with other nodes and the greater its importance. Degree distribution is used in complex network theory to describe the probability distribution of node degrees in a network. It indicates the probability or frequency of nodes of different degrees appearing in a network. Degree distribution is one of the important characteristics of a network, which can help us understand the topological structure and network properties of the network.

[0086]

[0087] k i is the degree value of node i.

[0088] N is the total number of nodes in the network.

[0089] 2) Betweenness Centrality

[0090] Node betweenness refers to the number of shortest paths through a node in a network. Betweenness centrality measures the "mediating" role of a node between other node pairs, that is, the frequency with which the node appears in the shortest paths of other node pairs. The higher the betweenness centrality of a node, the stronger its "bridge" role in the network and the more important its mediating role in information dissemination.

[0091]

[0092] is the number of paths that contain node i in the shortest path from node s to node t.

[0093] g st is the number of shortest paths from node s to node t

[0094] 3) Closeness Centrality

[0095] Closeness centrality measures the average distance from a node to all other nodes in the network. It emphasizes the accessibility of the node in the network. The higher the closeness centrality, the faster the node can access all other nodes in the network, and the higher the efficiency of information dissemination.

[0096]

[0097] d ijis the shortest path length from node i to node j.

[0098] 4) Eigenvector centrality

[0099] Eigenvector centrality not only considers the number of direct neighbors of a node, but also the indirect influence through neighbor connections. Specifically, eigenvector centrality is obtained by solving the main eigenvector of the network adjacency matrix, that is, finding a vector in which the score of each node is the weighted sum of the scores of all its neighbors, and this vector is the eigenvector corresponding to the maximum eigenvalue of the adjacency matrix. The calculation formula is as follows,

[0100] A·x=λ·x (20)

[0101] A is the adjacency matrix of the network.

[0102] x is the eigenvector centrality vector of the node.

[0103] λ is the maximum eigenvalue of the matrix.

[0104] In the present invention, a multi-dimensional analysis of the importance of the node structure is performed according to various network topology parameters to identify key nodes. By accurately predicting the changes in pollutant concentrations in the areas where these key nodes are located, strong data support can be provided for food safety management, and the risk of diffusion and transmission of pollutants can be effectively prevented and controlled.

[0105] Step F is constructed to quantify the risk of each node using dietary exposure assessment and conduct early warning analysis;

[0106] In the present invention, an environmental health risk assessment model is used to evaluate the health exposure risk of pollutants at each node to the human body. The method evaluates the health risk caused by consumption according to the exposure assessment.

[0107] Exposure parameters are basic parameters used to describe the characteristics and behaviors of human exposure to environmental media, and are key factors in determining the accuracy of environmental health risk assessment. Environmental health risk assessment is divided into two categories: carcinogen risk assessment (health risk assessment of pollutants without thresholds) and non-carcinogen risk assessment (health risk assessment of pollutants with thresholds). Both are based on the accurate evaluation of human exposure doses of pollutants. When the concentration of pollutants in the environmental medium is accurately quantified, the closer the exposure parameter value is to the actual exposure status of the target population, the more accurate the evaluation results of the exposure dose will be, and the more accurate the results of the environmental health risk assessment will be.

[0108] The risk assessment of non-carcinogens and carcinogens is shown in formula (21-22):

[0109]

[0110] RfD is the oral dose of the hazard, mg / (kg·d).

[0111] Q is the human carcinogenicity coefficient estimated based on population data, [mg / (kg·d)] -1 .

[0112] ADD is the average daily exposure dose of pollutants, mg / (kg·d).

[0113] The calculation method for the average daily exposure dose of human body to pollutants in food through oral administration is shown in formula (23):

[0114]

[0115] BW is the average body weight of residents, which is set to 60 kg.

[0116] EF is the hazard exposure frequency, which is set to 365 days.

[0117] ED is the exposure period, which is set to 60 years.

[0118] AT is the average exposure time of the hazard, which is the exposure frequency * exposure years.

[0119] IR i,j is the frequency of oral intake, kg / d;

[0120] C i,j is the 95th percentile of the content of heavy metal pollutants in province i, mg / kg. Part II: Application of the complex network model of air-soil heavy metal pollution interaction

[0121] Application

[0122] Taking the 2019 heavy metal-cadmium monitoring data of Hunan Province, a typical province, as an example, a complex network model of atmosphere-soil heavy metal pollution interaction was constructed to simulate the spread of heavy metal pollution in the environment and conduct risk warning analysis on foodborne food safety.

[0123] Step 1: Construct a complex network model of atmospheric-soil heavy metal pollution interaction in a typical region;

[0124] Taking the city centers of various cities in Hunan Province as nodes, the 95th percentile of heavy metal-cadmium detection concentrations in various cities as the node heavy metal concentrations in the network model, and the bordering regions as edges, a soil heavy metal pollution migration-propagation network model was constructed.

[0125] The fluid diffusion law, Fick's second law and the first-order kinetic model are used as the inter-node propagation kinetic model in the soil heavy metal pollution migration-propagation network model.

[0126] Secondly, another major source of heavy metals in the soil is the concentration increment brought by atmospheric deposition. Based on this, an atmospheric heavy metal pollution deposition-propagation network model is constructed. Combined with wind direction, it is constructed into a directed network. The average wind direction is used as the basis for judging whether the nodes are connected. The Gaussian plume model is used to simulate the propagation and diffusion behavior between nodes in the atmospheric heavy metal pollution deposition-propagation network model. The atmospheric deposition model is used to represent the interaction between the soil pollutant degradation-propagation network model and the atmospheric pollutant deposition-propagation network model. The combination of the two constitutes the complex network model of atmospheric-soil heavy metal pollution interaction in Hunan Province, such as Figure 4 shown.

[0127] The experimental data sampling time span is from April to October 2019, and the concentration data are counted monthly. The concentration distribution data of each month is used to predict the concentration distribution of the next month as an independent task, with the data from April to September as the training set and the data from October as the test set. The MAML optimization algorithm is used to optimize the unknown parameters in the complex network model through multi-task learning.

[0128] Figure 5 In order to use the trained model to predict the concentration distribution and true value distribution of various regions in Jiangsu Province in October 2019, it can be seen from the figure that the model prediction value is consistent with the true value distribution as a whole, which can better reflect the distribution characteristics of heavy metals in the region.

[0129] Figure 7 To predict the concentration distribution in various regions of Hunan Province in the next 11-12 months, it can be seen from the figure that the overall heavy metal concentration in Hunan Province is low, and the heavy metal pollution concentration in most areas such as Changsha and Zhangjiajie is on a downward trend. Changde, Yiyang, Xiangtan and other regions have shown a slight upward trend. Although the heavy metal pollution concentration in Xiangxi and Zhangjiajie has shown a downward trend, it is still at a high level compared with other regions. For these areas, targeted governance measures should be taken, such as strengthening the supervision of heavily polluting industries, optimizing agricultural fertilization and pesticide use strategies, etc., in order to formulate a more scientific regional prevention and control strategy to ensure that heavy metal pollution in the province is effectively controlled and achieves long-term improvement.

[0130] Step 2: Identify the regions where key nodes are located in the complex network model;

[0131] According to the complex network analysis, the centrality results of each node are shown in Table 1: In terms of degree centrality, Huaihua, Yiyang, Changsha and Hengyang performed relatively high, indicating that these nodes have the largest number of connections with other nodes in the network and have strong direct connection capabilities. In betweenness centrality, Huaihua is significantly higher than other nodes, indicating that it plays an important bridge role in the network and is a key node for many shortest paths; in terms of proximity centrality, Huaihua also has an advantage, reflecting that it has the shortest average distance to other nodes and can quickly access the entire network; in eigenvector centrality, Changsha, Yiyang and Huaihua rank high, showing that they not only have high connectivity, but also have strong correlations with other important nodes, reflecting their core position in the network.

[0132] Combined with the centrality analysis, it can be seen that pollution control in important nodes such as Huaihua, Yiyang, and Changsha is crucial. By strengthening pollution control in these key node areas, the risk of pollutant spread can be effectively reduced and agricultural and food safety in surrounding areas can be guaranteed. At the same time, timely monitoring and taking necessary pollution prevention and control measures in areas where pollution sources are more concentrated will have a positive impact on ensuring food safety and regional environmental quality.

[0133] Table 1: Centrality index of each node

[0134]

[0135]

[0136] Step 3: Use dietary exposure assessment to quantify the risks at each node and conduct early warning analysis;

[0137] The method for determining the risk assessment score is the five-point scale method. It was proposed by American psychologist Likert in 1932 to assess the attitude or opinion of respondents towards a statement or question. This rating method standardizes the respondents' responses into five levels and assigns numbers within the range of 1-5.

[0138] According to the above five-point scale method, the risk assessment results are divided into five levels: Ⅰ, Ⅱ, Ⅲ, Ⅳ, Ⅴ, as shown in Table 2.

[0139] Table 2: Risk level classification table

[0140]

[0141] As early as 2017, the World Health Organization's International Agency for Research on Cancer announced that heavy metals - cadmium and its compounds were classified as Class I carcinogens, which seriously endanger human health. The severity level of the hazard was calculated through food health risk assessment. Among them, the carcinogenic intensity coefficient Q of heavy metal cadmium is 6.1 (kg·d) / mg, and the frequency of oral ingestion of soil IR is 0.00015kg / d. According to the model constructed in step 1, the concentration distribution of various regions in Hunan Province in the future time period was predicted to conduct dietary exposure assessment. The dietary exposure assessment results and levels of various regions in Hunan Province are shown in Table 3. The results show that the overall risk of various regions in Hunan Province is low, and the risks of Xiangxi, Zhangjiajie and other regions are higher than other regions. Quantifying risks through dietary exposure assessment can provide a scientific basis to help identify potential health hazards, formulate reasonable food safety policies, and play a key role in reducing food safety risks and protecting public health.

[0142] Table 3: Predicted concentration assessment results and risk levels in various regions of Hunan Province

[0143]

[0144] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the above embodiments, ordinary technicians in the relevant field should understand that the specific implementation methods of the present invention can still be modified or replaced by equivalents. Any modification or equivalent replacement that does not depart from the spirit and scope of the present invention should be covered within the scope of protection of the claims of the present invention.

Claims

1. A method for risk early warning analysis of food safety by constructing an atmosphere-soil pollutant interaction complex network model through complex networks to simulate the spread of heavy metal pollution in the environment, characterized in that The construction of the complex network model of air-soil pollutant interactions includes the following steps: Construction step A: constructing a soil heavy metal pollution migration-propagation network model; In the present invention, the soil pollutant degradation-propagation network model is defined as G = (E, V, θ), where the set V = {v1, v2, v3, ... v n } is a node set, with regional centers or monitoring stations as nodes; set E = {e1, e2, e3, …e m } is the set of edges between nodes. Bordering regions or adjacent monitoring stations are connected as edges; the set θ={θ1,θ2,θ3,…θ n } is the pollutant concentration at each node. In the present invention, Fick's second law and fluid diffusion law are used as the heavy metal diffusion model between nodes in the soil pollutant degradation-propagation network model, and the first-order kinetic model is used to simulate the migration behavior of heavy metals in the node itself. The two are combined to form the propagation kinetic model between nodes in the soil pollutant degradation-propagation network model. The formula is as follows: k represents the heavy metal transfer factor. v represents the diffusion velocity in the horizontal and vertical directions. D represents the diffusion coefficient in the horizontal and vertical directions. Construction step B, constructing the atmospheric heavy metal pollution deposition-propagation network model; In the present invention, the atmospheric heavy metal pollution deposition-propagation network model is constructed as a directed network in combination with wind direction, and the average wind direction is used as the basis for judging whether the nodes are connected. In the network, the source node is the first node in the average wind direction, representing the initial release point of heavy metal pollutants. Due to the effect of wind force, the source node continuously transports and diffuses heavy metal pollutants to subsequent adjacent nodes, forming its propagation path in the network. In the present invention, the Gaussian plume model is used to simulate the propagation and diffusion behavior between nodes in the atmospheric pollutant deposition-propagation network. The interaction between the soil pollutant degradation-propagation network model and the atmospheric pollutant deposition-propagation network model is represented by the deposition of pollutants in the atmosphere. Atmospheric deposition refers to the process in which pollutants suspended in the atmosphere settle to the ground or other surfaces due to gravity. The combination of the two constitutes a propagation dynamics model between nodes in the atmospheric heavy metal pollution deposition-propagation network model. The formula is as follows, Q i is the static source strength, that is, the pollutant concentration of the previous node in the average wind direction. σ y is the diffusion coefficient perpendicular to the wind direction. y 0,i The horizontal position of the subsequent node relative to the previous node A ij is the network adjacency matrix. V s is the deposition rate of pollutants in the atmosphere. Construction step C, using optimization algorithms to optimize unknown parameters in propagation dynamics; In the present invention, the meta-learning optimization algorithm of Model-Agnostic Meta-Learning (MAML) is adopted to optimize the unknown parameters in the complex network model through multi-task learning. In the present invention, the pollutant concentration distribution of each node in each time period is first set as an independent task. Task T t It is defined as predicting the pollutant concentration in the next time period given the pollutant concentration distribution of each node in the previous time period as input. Second, initialize the model parameters θ, for each task T t , learn from the training data, and update the model parameters through gradients (i.e., inner loop iterations) to adapt to the task. The updated parameters are θ t ' α is the learning rate of the inner loop, is the loss function for the training data After the inner loop is updated, in task T t Evaluate the performance of the model on the test data and calculate the test loss And update the model's initialization parameters θ through the test loss of all tasks β is the learning rate of the outer loop. The gradient of the test loss is calculated for all tasks. The goal of MAML is to optimize the initial model parameters θ so that the model can quickly adapt to new tasks through a small amount of gradient updates and achieve better prediction results. The optimization goal is expressed as: Construction step D, constructing a complex network model of air-soil pollutant interactions; In the present invention, the soil pollutant migration-propagation network model and the atmospheric pollutant deposition-propagation network model are connected through the deposition effect of pollutants in the atmosphere to obtain an atmosphere-soil pollutant interaction complex network model. Constructing step E, analyzing the key nodes according to the properties of the complex network topology; In the present invention, a multi-dimensional analysis of the importance of the node structure is performed according to various network topology parameters to identify key nodes. By accurately predicting the changes in pollutant concentrations in the areas where these key nodes are located, strong data support can be provided for food safety management, and the risk of diffusion and transmission of pollutants can be effectively prevented and controlled. Construct step F, use dietary exposure assessment to conduct risk warning analysis for each node; In the present invention, an environmental health risk assessment model is used to evaluate the health exposure risk of pollutants at each node to the human body. The method evaluates the health risk caused by consumption according to the exposure assessment. The risk assessment of non-carcinogens and carcinogens is shown in formula (4-5): RfD is the oral dose of the hazard, mg / (kg·d). Q is the human carcinogenicity coefficient estimated based on population data, [mg / (kg·d)] -1 . ADD is the average daily exposure dose of pollutants, mg / (kg·d). The calculation method for the average daily exposure dose of human body to pollutants in food through oral administration is shown in formula (21): BW is the average body weight of residents, which is set to 60 kg. EF is the hazard exposure frequency, which is set to 365 days. The ED exposure period is set to 60 years. AT is the average exposure time of the hazard, which is the exposure frequency * exposure years. IR i,j is the average daily consumption of hazardous substance j in province i, kg / d; C i,j It is the 95th percentile of the content of heavy metal pollutant j in province i, mg / kg.

2. A method for risk early warning analysis of typical areas using the complex network model of air-soil heavy metal pollution interaction, characterized by There are the following steps: Step 1: Construct a complex network model of atmospheric-soil heavy metal pollution interaction in a typical region; According to the construction steps AB in the first part, a complex network model of air-soil heavy metal pollution interaction in the typical province of Hunan was constructed, as shown in Figure 4. The experimental data sampling time span is from April to October 2019, and the concentration data are counted monthly. The concentration distribution data of each month is used to predict the concentration distribution of the next month as an independent task, with the data from April to September as the training set and the data from October as the test set. The unknown parameters in the complex network model are optimized according to the construction step C. Figure 5 shows the concentration distribution and true value distribution predicted by the trained model for various regions in Jiangsu Province in October 2019. It can be seen from the figure that the model prediction value is consistent with the true value distribution as a whole, which can better reflect the distribution characteristics of heavy metals in the region. Figure 7 shows the predicted concentration distribution of various regions in Hunan Province in November and December. It can be seen from the figure that the overall heavy metal concentration in Hunan Province is low, and the heavy metal pollution concentration in most regions such as Changsha and Zhangjiajie is on a downward trend. The heavy metal pollution concentration in Changde, Yiyang, Xiangtan and other regions is on a small upward trend. Although the heavy metal pollution concentration in Xiangxi and Zhangjiajie is on a downward trend, it is still at a high level compared with other regions. Step 2: Identify the regions where key nodes are located in the complex network model; According to the construction step E, the key regional nodes in the complex network are identified. The centrality calculation results of each node are shown in Table 1. In terms of degree centrality, Huaihua, Yiyang, Changsha and Hengyang perform relatively well, indicating that these nodes have the largest number of connections with other nodes in the network and have strong direct connection capabilities. In betweenness centrality, Huaihua is significantly higher than other nodes, indicating that it plays an important bridge role in the network and is a key node for many shortest paths; in terms of proximity centrality, Huaihua also has an advantage, reflecting that it has the shortest average distance to other nodes and can quickly access the entire network; in eigenvector centrality, Changsha, Yiyang and Huaihua rank high, showing that they not only have high connectivity, but also have strong correlations with other important nodes, reflecting their core position in the network. Step 3: Use dietary exposure assessment to quantify the risks at each node and conduct early warning analysis; According to the prediction of the heavy metal concentration distribution in the future time period by the complex network model in step F, the risk assessment was carried out and the levels were divided according to the five-point scale method. The results showed that the overall risk in various regions of Hunan Province was low, and the risks in Xiangxi, Zhangjiajie and other regions were higher than other regions. Quantifying the risk through dietary exposure assessment can provide a scientific basis to help identify potential health hazards, formulate reasonable food safety policies, and play a key role in reducing food safety risks and protecting public health.

Images