Modeling And Optimization Of Food Safety Traceability And Process Risk Assessment

With the process of economic globalization is accelerating, food safety is not only a matter of life and health of consumers, but also related to social harmony and stability and healthy economic development. The international trade in food and agricultural products and the image of the country are also impacted by food safety problems. Food safety has become the focus of attention of the international community. Although China’s food safety condition has improved, the situation is still grim. It has become the core of the modern food industry management objectives to pursue, but also the governments to strengthen the main direction of the food safety supervision administration to identify, prevent, control and reduce possible hazards in the food chain from cultivation raw materials, processing, packaging, storage, transportation and sale to the consumption of food, trace harm source and track the flow of food products involved in safety problems.The traceability method based on tracking tags depends on the the process model of specific product categories. So it is difficult to implement traceability for different types of food chain processes, and the traceability model needs to be re-established. This makes traceability information systems lack compatibility and hinders the replication of these systems. Therefore there are a large number of information silos, and it is hared to gey a complete traceability information chain. However, the present process risk modeling approach ignored the importance of risk factors in the actual risk management, and therefore it can only carry out an efficient risk assessment. Identification and prevention of food safety risk can not be effectively carried out. It is of important theoretical and practical significance to study traceability modeling methods and process risk modeling integrated with risk factors.In this thesis, we first review the current research status and inadequation of food traceability and risk assessment modeling methods. Then we study general traceability modeling method and traceability under some special conditions. And we study the food safety process risk modeling approach and biotraceability. The main research works are summarized below.(1) The traceability modeling methods based on relationships of traceability units depend on specific food types and processes, which result in a lack of traceability information system reusability and system compatibility, and it is easy to form silos of traceability information. For the above problem, we propose a common traceability modeling method. According to the food traceability implementation elements, and from the perspective of information systems design and implementation, we abstract and classify the various processing steps, and define four universal traceability structures,"batch","position","trading unit" and "event". These structures form the expression of processing steps in the traceability model, which is conducive to the association of traceability information between different processing steps, and makes traceability easy and of general use. Then the universal traceability model may cover all types of food, and easy to implement traceability of a food chain and across food chains.(2) To solve the problem of traceability data missing, out of sync or incompleteness in food traceability, an intelligent traceability method is proposed which uses the turnaround time of traceability units to estimate units’flowing state and path. In this method, the turnaround time of a traceability unit between two nodes in the food chain network is seen as a continuous random variable, the turnaround time distribution is estimated by using historical data; then, a route estimation optimization model is established. The model uses the mathematical expectation of traceability units’turnaround time distribution between nodes as variables, and the minimum between the total time of the path and a given time as the optimization goal. And Monte Carlo simulation is used to simulate the time distribution of the flowing path to verify the feasibility of the estimated path. The simulation results show that, this method does not depend on the traceability unit synchronization information. The flow state of the traceability unit may be predicted in the food chain the network, as only the sent node is known. Given the sent node and the terminal node, the historical flow path of the traceability unit may be estimated.(3) In order to reduce the relationships between raw milk batches and product batches, a batch relationship model is established. First, raw milk batches are identified by using electronic tags and milk tanks because raw milk of different sources generally are filled into different tanks. Then the interval between two cleanings of the milk storage tank is seen as a research cycle to establish a batch relationship model. Combined with the production process data, the batch relationship model uses iterative equations to describe the batch relationship between batches of raw milk and products. When the number of batches increases, the iterative form may also avoid the exponential growth of the number of variables. By adjusting the relative order of time that raw milk mixed into the milk storage tank and that a product batch, the batch relationship number of raw milk and related products is reduced, thereby the complexity of the traceability drops. Lastly, an example of21raw milk batches and14product batches is given. The simulation shows that the complexity of traceability is significantly reduced above50%by the optimization model, which may significantly narrow the scope of recalls.(4) In the current food safety process risk modeling method, risk factors are not considered. So the assessment of the current defect risk of a microbiological hazard in food can only be carried out using the model. Thus we improved the food safety process risk modeling method. Varieties of risk factors such as operating environment, personnel, equipment and others may introduce microbiological hazards into foods. In order to effectively implement risk management, the quantitative microbiological risk assessment modeling method which uses the modular process risk model as a modeling framework is improved. In the improved method, risk factors are abstracted as a hazard transfer modular. Food safety risk factors include sources of raw materials, storage risk, operational personnel hygiene, and environmental contamination risks. Hazard transfer process describes the contamination frequency and the probability distribution of microbial quantity that introduced into a product by environment, operating personnel, equipment and other kinds of risk factors, which is mathematically described as if(0,0,P). If risk does not exist, the amount of microbial is zero. If risk exists, the probability distribution of microbial number is which can be discrete or continuous distribution. The proportion that transferred to a product can be described by function g(·) which is related with operating time, temperature, contact area and etc. Control process describes varieties of control measures may be taken in the production, such as the use of different disinfection, the implementation of different test frequencies. The prevention and control impact of a measure is described by the probability and quantity change of microbial introduced by risk factor in the model. And utility modular characterizes the consumption cost and gain of the control measure. Then by the following three steps a modular process risk model can be realized by Bayesian network.1) Make processes, materials mixing and partitioning, and processing parameters clear, find out risk factors that may introduce microbial into a product and affect the microbial dynamics;2) Select the appropriate basic processes, and determine the Bayesian network structure of risk model;3) Collect risk data, and determine the conditional probability of each node in the model by statistical analysis of risk data. Combined with predictive microbiology, the number and occurrence probability of microbiological hazards in each process is estimated using Bayesian inference. This risk model can be used to assess the risk situation and the impact of risk factors on food safety, and also can trace the source of microbial hazards. Add control processes to hazard transfer processes, and modify the conditional probability of the corresponding hazard transfer processes nodes, the effectiveness of one or several control measures can be verified. In the model, parameters of nodes can be adjusted, which makes it easy to assess the impact degree of risk factors and control measures on food product safety. Using the improved modeling method, the origin of microbiology hazard may be traced. And the compare of effectiveness and profit of different control measures can help managers to choose preferred one or control combination. Finally, we point out a few existing problems to be solved after summarizing the works done in this dissertation.