An IoT-based konjac food safety detection and analysis system

Abstract

The present invention relates to the technical field of food safety detection, and discloses a konjac food safety detection and analysis system based on the Internet of Things, including: an edge intelligent perception layer, a pollution field dynamic reconstruction engine, and a lightweight blockchain data layer. The system collects microbial metabolite and surface component information in real time through a multi-modal sensor array, combines a dynamic threshold adaptive algorithm, and intelligently adjusts the detection sensitivity according to different processing stages, improving the recognition accuracy of dynamic pollution. The pollution field dynamic reconstruction engine uses spatio-temporal correlation modeling and transfer learning algorithms to form a continuous pollution concentration field and predict the diffusion trend. At the same time, the self-healing logic chain can reconstruct the pollution field when local nodes are abnormal to ensure uninterrupted monitoring. The blockchain double-chain structure cooperates with the intelligent contract mechanism to achieve a dynamic balance between the privacy protection of enterprise self-inspection data and the verification of regulatory traceability.

Description

Technical Field

[0001] The present invention relates to a konjac food safety detection and analysis system based on the Internet of Things, and belongs to the technical field of food safety detection. Background Art

[0002] During the konjac processing, microbial contamination has dynamic change characteristics. Traditional food safety detection methods usually adopt a sampling inspection mode at fixed time intervals and record the detection results through a centralized platform. Its basic principle is to sample at specific process nodes and statically analyze the contamination indicators. Such methods have the advantages of simple operation and low implementation cost in a conventional environment. However, in scenarios where the real-time contamination change in the konjac processing line is intense and the spatio-temporal diffusion is rapid, the existing solutions gradually show certain limitations.

[0003] Specifically, the following problems generally exist in the prior art: 1. It is difficult to capture the pollution peak in real time: Due to the time delay in sampling inspection, the sharp fluctuations of microbial contamination within a short period are often missed, resulting in an underestimated pollution risk. 2. The credibility of data is in doubt: The data on the centralized platform is vulnerable to tampering or human intervention, and there is no effective mutual trust mechanism between the self-inspection data of enterprises and the supervision sampling data. 3. The detection parameters are rigid and the adaptability is insufficient: The existing systems generally adopt fixed threshold settings and cannot flexibly adjust the detection standards according to different stages and different raw material characteristics in konjac processing, resulting in a relatively high false alarm rate.

[0004] In order to alleviate the above problems, some solutions in the industry attempt to introduce Internet of Things sensors and cloud data management platforms to improve the detection response speed by increasing the sampling frequency and centralized processing. However, such methods still face new challenges. On the one hand, the data of edge nodes needs to go through complex transmission and processing when uploaded to the cloud, resulting in the accumulation of response delays. On the other hand, the data security and privacy risks caused by over-reliance on the central server reduce the willingness of enterprises to cooperate and it is difficult to comprehensively guarantee the supervision transparency. In addition, even if the multi-sensor fusion detection technology is introduced, due to the lack of dynamic threshold adjustment and pollution diffusion trend prediction mechanisms, the overall system still relies on local sampling data and cannot accurately depict the continuous change of the pollution field, and the fault tolerance ability of the system for abnormal nodes is also insufficient.

[0005] Therefore, how to capture the pollution state in real time, credibly and adaptively throughout the entire konjac processing process and maintain the monitoring continuity in case of node failures or data anomalies has become the technical problem to be solved by the present invention. Summary of the Invention

[0006] The present invention provides a konjac food safety detection and analysis system based on the Internet of Things, and its main purpose is to solve the problems of lagging real-time performance, insufficient data credibility and poor adaptability of detection parameters.

[0007] To achieve the above object, the present invention provides a konjac food safety detection and analysis system based on the Internet of Things, which system includes:

[0008] An edge intelligent perception layer, disposed on the konjac processing production line, including a multi-modal sensor array and a dynamic threshold adaptive algorithm; the multi-modal sensor array is used to collect multi-dimensional parameter information in the konjac processing environment in real time, at least including the concentration of microbial metabolites and spectral data of the konjac surface components; the dynamic threshold adaptive algorithm is configured to adjust the microbial detection threshold range adapted to this stage in real time according to different processing stages of the konjac. Among them, if the current processing stage is the fine processing stage , then the upper limit of the microbial detection threshold satisfies the relationship: , where is the upper limit of the microbial detection threshold in the primary processing stage, is a preset threshold reduction coefficient for the fine processing stage;

[0009] A lightweight blockchain data layer, including a private chain, a public chain and a smart contract trigger mechanism; the private chain is used to store the enterprise self-inspection data from the edge intelligent perception layer and only open the access permission to authorized nodes; the public chain is used to cross-chain synchronize the key detection data screened in the private chain, at least including the konjac batch number and the pollution index status, to the supervision platform to achieve data verifiability; the smart contract trigger mechanism is configured to automatically generate a deposit record in the blockchain data layer and send an alarm notification to the preset upstream and downstream enterprise nodes when the pollution degree detected by the edge intelligent perception layer exceeds the dynamic threshold corresponding to the current processing stage;

[0010] A pollution field dynamic reconstruction engine, configured to receive and process the multi-dimensional parameter information from the edge intelligent perception layer, and construct a dynamic pollution field model reflecting the environmental pollution distribution and time-varying of the konjac processing; this dynamic pollution field model maps the data collected by sensors at different positions to the corresponding timestamps through a spatio-temporal correlation modeling method to form a three-dimensional pollution concentration distribution, and predicts the potential path of pollution diffusion based on historical pollution data using a transfer learning algorithm; and, this pollution field dynamic reconstruction engine also includes a self-healing logic chain, which is used to perform interpolation calculations based on the data collected by adjacent normal nodes when detecting that a single or part of the sensors fails, and reconstruct the local pollution field information to ensure that the system can still maintain effective pollution monitoring and risk prediction capabilities in the case of partial sensing data loss.

[0011] Preferably, the multimodal sensor array includes a microfluidic biosensor and a hyperspectral imaging module. The microfluidic biosensor is used to detect the concentration change of microbial metabolites, and the hyperspectral imaging module is used to analyze the spectral characteristics of the konjac surface components in real time. The data collected by both are fused to more comprehensively evaluate the quality and safety status of konjac.

[0012] Preferably, for the dynamic threshold adaptive algorithm, according to the different characteristics of the cleaning stage, slicing stage, and drying stage in the konjac processing flow, corresponding microbial detection threshold ranges are preset, and the current processing stage is automatically identified during the system operation, and the corresponding threshold range is called for pollution determination, and the response time for threshold switching is less than or equal to fifty milliseconds.

[0013] Preferably, the lightweight blockchain data layer adopts a double-chain structure. When the key sampling inspection data stored on the public chain is synchronized to the supervision platform, the data hash value method is used to achieve the verifiability of the data by the supervision department, and at the same time, the content of the original enterprise self-inspection data is hidden.

[0014] Preferably, when the self-healing logic chain reconstructs the pollution field information, a weighted interpolation algorithm is adopted, and the weight coefficients of adjacent nodes are dynamically determined based on their spatial distance from the faulty node and the similarity of historical data.

[0015] Preferably, when the smart contract trigger mechanism sends an alarm notification, it not only includes the information of excessive pollution, but also includes the information of the potential pollution diffusion path predicted by the pollution field dynamic reconstruction engine, so that relevant enterprise nodes can take more targeted preventive measures.

[0016] Preferably, the edge intelligent perception layer performs local preprocessing on the collected original sensor data, including noise filtering, data calibration, and feature extraction, to reduce the amount of data transmitted to the pollution field dynamic reconstruction engine and the processing delay.

[0017] Preferably, the system further includes a human-computer interaction interface, which is used to display the current konjac processing environmental pollution field model, pollution alarm information, and blockchain evidence storage status in real time, and allows authorized users to perform parameter configuration and system monitoring.

[0018] Preferably, between the edge intelligent perception layer and the lightweight blockchain data layer, and between the lightweight blockchain data layer and the pollution field dynamic reconstruction engine, a wireless communication protocol is used for data transmission. Among them, the data transmission rate of the wireless communication protocol needs to satisfy: where is the minimum rate threshold to ensure real-time data transmission.

[0019] Compared with the problems in the background technology, the beneficial effects of the present invention are:

[0020] 1. During the konjac processing, the microbial contamination characteristics caused by different stages are significantly different. The traditional fixed threshold detection method is often unable to adapt to the dynamic changes of the process flow, resulting in frequent misjudgments in pollution monitoring. The present invention introduces a multimodal sensor array in the edge intelligent perception layer to achieve real-time collaborative perception of changes in the concentration of microbial metabolites and surface components, and combines the dynamic threshold adaptive algorithm to spontaneously adjust the detection sensitivity according to the process characteristics of the current processing stage, thereby forming a set of adaptive, closed-loop pollution identification mechanism within the system, which not only avoids the pollution omission or false alarm problems caused by stage mismatch under traditional static detection, but also naturally guides the sensor output data to be optimized on demand after local preprocessing, effectively compressing the data pressure of the subsequent pollution field reconstruction engine, so that the entire system can still maintain the collaborative ability of accurate perception and low-latency response under complex working conditions.

[0021] 2. In view of the uncertain characteristics of spatial dynamic evolution and local sudden changes in pollution diffusion in the konjac processing production line, the present invention innovatively constructs a dynamic reconstruction engine for the pollution field. By modeling the multi-dimensional parameter information collected by the sensor in time and space, a three-dimensional pollution concentration distribution with continuous evolution is naturally generated. At the same time, the system introduces a transfer learning algorithm, which can autonomously deduce the potential pollution diffusion path in the current environment based on the historical pollution diffusion pattern. In order to cope with the problem of missing pollution field information caused by the failure of some sensor nodes, the dynamic reconstruction engine of the pollution field further integrates a self-healing logic chain, and uses the spatial position relationship and historical similarity data of adjacent normal nodes to dynamically reconstruct the local pollution field, and realize the steady-state monitoring of the system under local degradation of the sensor network. This series of mechanisms are linked to shift the identification of pollution risks from isolated sampling to continuous field modeling, significantly improving the system integrity and engineering availability of pollution diffusion prediction.

[0022] 3. In food safety data management, it is necessary to ensure the privacy of the company's self-inspection data and to achieve effective verification of the authenticity of the data by the regulatory authorities. Traditional centralized platforms are difficult to take both into account. The present invention constructs a lightweight blockchain data layer that combines private chains and public chains to form a data path that clearly separates internal enterprise data from external regulatory needs; at the same time, in conjunction with the smart contract trigger mechanism, when the pollution detection exceeds the dynamic threshold, the system automatically generates a record of evidence and issues an early warning to upstream and downstream enterprise nodes, avoiding information delays caused by manual intervention lags. Through this on-chain automated trusted process, not only is data verifiability and traceability transparency enhanced, but the traditional data island problem is naturally avoided in terms of technical path, so that the food safety monitoring system achieves a dynamic balance between privacy protection and public supervision that can be implemented in the engineering.

[0023] 4. In the face of the problem that the pollution load in konjac processing changes rapidly with time and process stages, the present invention, through the dynamic threshold adaptive mechanism of the edge intelligent perception layer and combining the prediction of the pollution diffusion trend of the pollution field dynamic reconstruction engine, can not only issue real-time alarms when the pollution index is initially exceeded, but also actively push preventive adjustment suggestions based on the predicted diffusion path before the pollution reaches the key links. This mechanism avoids the one-way logic of the traditional system that only triggers alarms relying on the current pollution value, forming a complete closed-loop control process of perception, prediction, early warning, and intervention, thereby significantly enhancing the early response ability to sudden pollution events and the foresight of local risk control at the system level. Considering the complex environment of the konjac processing production line and the fact that the sensing nodes are vulnerable to factors such as mechanical disturbances and humid and hot environments, resulting in data anomalies, the present invention designs a local preprocessing module in the edge intelligent perception layer to perform noise filtering and data calibration in real time. At the same time, it cooperates with the self-healing logic chain to automatically compensate and reconstruct based on the data of adjacent nodes when data anomalies occur, and ensures the continuity of the data link through a low-latency wireless communication mechanism. This combined design not only enhances the overall fault tolerance and robustness of the system, but also enables the pollution monitoring and early warning functions to self-correct and recover on the edge side, avoiding the risk of global monitoring interruption caused by single-point failures and improving the reliable operation ability of the konjac food processing production line in the actual industrial environment. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] Figure 1 It is a schematic diagram of the lightweight blockchain data layer structure of the present invention;

[0025] Figure 2 It is a flowchart of pollution exceeding the standard warning and evidence storage of the present invention;

[0026] Figure 3 It is a data processing flowchart of the edge intelligent perception layer of the present invention.

[0027] The implementation, functional features, and advantages of the present invention will be further described with reference to the embodiments and the accompanying drawings. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0028] It should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.

[0029] The embodiments of the present application provide an Internet of Things-based konjac food safety detection and analysis system, which includes:

[0030] An edge intelligent perception layer, arranged on the konjac processing production line, including a multi-modal sensor array and a dynamic threshold adaptive algorithm; the multi-modal sensor array is used to collect multi-dimensional parameter information in the konjac processing environment in real time, at least including the concentration of microbial metabolites and spectral data of the konjac surface components; the dynamic threshold adaptive algorithm is configured to according to different processing stages of the konjac Adjust the microbial detection threshold range adapted to this stage in real time. Among them, if the current processing stage is the fine processing stage Then the upper limit of the microbial detection threshold Satisfies the relational expression: Among them Is the upper limit of the microbial detection threshold in the primary processing stage, Is the preset threshold reduction coefficient for the fine processing stage;

[0031] The lightweight blockchain data layer includes a private chain, a public chain, and a smart contract trigger mechanism; the private chain is used to store the enterprise self-inspection data from the edge intelligent perception layer and only opens the access permission to authorized nodes; the public chain is used to cross-chain synchronize the selected key detection data in the private chain, at least including the konjac batch number and the pollution index status, to the supervision platform to achieve data verifiability; the smart contract trigger mechanism is configured to automatically generate an evidence record in the blockchain data layer and send an alarm notification to the preset upstream and downstream enterprise nodes when the pollution degree detected by the edge intelligent perception layer exceeds the dynamic threshold corresponding to the current processing stage;

[0032] The pollution field dynamic reconstruction engine is configured to receive and process multi-dimensional parameter information from the edge intelligent perception layer, and construct a dynamic pollution field model reflecting the environmental pollution distribution and time-varying of konjac processing; this dynamic pollution field model uses a spatio-temporal correlation modeling method to map the data collected by sensors at different positions with the corresponding timestamps to form a three-dimensional pollution concentration distribution, and predicts the potential path of pollution diffusion based on historical pollution data using a transfer learning algorithm; and, this pollution field dynamic reconstruction engine also includes a self-healing logic chain, which is used to perform interpolation calculations based on the data collected by adjacent normal nodes when a single or part of the sensors are detected to fail, and reconstruct the local pollution field information to ensure that the system can still maintain effective pollution monitoring and risk prediction capabilities in the case of partial sensor data loss.

[0033] Preferably, the multi-modal sensor array includes a microfluidic biosensor and a hyperspectral imaging module. The microfluidic biosensor is used to detect the concentration change of microbial metabolites, and the hyperspectral imaging module is used to analyze the spectral characteristics of the konjac surface in real time. The data collected by both are fused to more comprehensively evaluate the quality and safety status of konjac.

[0034] Preferably, the dynamic threshold adaptive algorithm presets corresponding microbial detection threshold ranges according to the different characteristics of the cleaning stage, slicing stage, and drying stage in the konjac processing flow, automatically identifies the current processing stage during the system operation, calls the corresponding threshold range for pollution determination, and the response time for threshold switching is less than or equal to fifty milliseconds.

[0035] Preferably, the lightweight blockchain data layer adopts a double-chain structure. When the key sampling inspection data stored on the public chain is synchronized to the supervision platform, the data hash value method is used to achieve the verifiability of the data by the supervision department, while hiding the original content of the enterprise self-inspection data.

[0036] Preferably, the pollution field dynamic reconstruction engine uses a transfer learning algorithm to predict the possible range and trend of pollution diffusion in the current processing environment based on the correlation between different environmental parameters and pollution diffusion patterns during the historical production process.

[0037] Preferably, when the self-healing logic chain reconstructs the pollution field information, a weighted interpolation algorithm is adopted, and the weight coefficients of adjacent nodes are dynamically determined based on their spatial distance from the faulty node and the similarity of historical data.

[0038] Preferably, when the intelligent contract trigger mechanism sends an alarm notification, it not only includes the pollution exceeding standard information, but also includes the information on the potential pollution diffusion path predicted by the pollution field dynamic reconstruction engine, so that relevant enterprise nodes can take more targeted preventive measures.

[0039] Preferably, the edge intelligent sensing layer performs local preprocessing on the collected original sensor data, including noise filtering, data calibration, and feature extraction, to reduce the amount of data transmitted to the pollution field dynamic reconstruction engine and the processing delay.

[0040] Preferably, the system further includes a human-computer interaction interface for real-time displaying the current konjac processing environmental pollution field model, pollution alarm information, and blockchain evidence storage status, and allowing authorized users to perform parameter configuration and system monitoring.

[0041] Preferably, between the edge intelligent sensing layer and the lightweight blockchain data layer, and between the lightweight blockchain data layer and the pollution field dynamic reconstruction engine, a wireless communication protocol is used for data transmission. Among them, the data transmission rate of the wireless communication protocol needs to satisfy: where is the minimum rate threshold to ensure real-time data transmission.

[0042] Example 1: In this example, in the construction of the edge intelligent sensing layer, a microfluidic biosensor module and a hyperspectral imaging module are deployed. The former collects the concentration of microbial metabolites, and the latter collects the spectral characteristics of the konjac surface components. After the two types of data are preliminarily denoised and normalized by a preset data processing unit, they are locally fused into a feature vector and sent to the pollution determination algorithm model. To ensure the adaptability of the system to different processing stages, the system adopts a dynamic threshold adjustment logic based on the processing stage in the microbial detection threshold formula where the variable The upper limit of the microbial data distribution from the initial processing stage of history is obtained by comprehensively estimating the sample mean and manually setting the safety factor; the parameter is an empirical adjustment factor, which is set in the system initialization configuration according to the konjac variety, equipment type and process mode.

[0043] The call mechanism of the dynamic threshold is based on the real-time feedback of the processing stage recognition module. When the sensor acquisition frequency is less than the 100ms level, the system judges the current stage in each round of loop and loads the upper threshold corresponding to this stage in real time for pollution over-limit judgment, which can effectively avoid false alarms caused by stage mismatch. In the dynamic reconstruction engine of the pollution field, in order to construct a continuous pollution concentration distribution, a three-dimensional data structure based on multi-sensor data and timestamps is adopted. The system performs interpolation modeling based on this data set and introduces a pollution diffusion trend prediction model; this model combines the historical pollution field evolution trajectory and the current sampling state, and establishes a potential space path prediction of pollution concentration change through transfer learning technology. In the case of single or partial sensor node failures, the system calls the self-healing logic chain mechanism to compensate and reconstruct the pollution field information; specifically, the spatial distance between adjacent nodes can be calculated through the three-dimensional coordinates set during deployment, and the historical data similarity is measured by the Euclidean distance based on the pollution concentration change trend within a short-period window.

[0044] When the pollution determination result exceeds the current threshold range, the system sends a deposit request to the blockchain data layer through the intelligent contract trigger mechanism. The preset trigger condition in the intelligent contract module is that when two consecutive over-limit values are determined as pollution events within any monitoring period, the system automatically generates a pollution record with a timestamp and batch number to the private chain and synchronizes its hash digest to the public chain; at the same time, in order to enhance the fault tolerance and data credibility of the system, the private chain in the double-chain structure stores the original detection data, and the public chain only synchronizes the hash and key event information to ensure both regulatory verifiability and data privacy. A lightweight cross-chain protocol is adopted between the two chains, and the data update period is set by the system itself to ensure the regulatory response timeliness.

[0045] Example 2: This example combines Figures 1 to 3 to illustrate the konjac food safety detection and analysis system based on the Internet of Things, as Figure 1 shown Figure 1Shows the structure of the lightweight blockchain data layer, which includes a private chain and a public chain, as well as a smart contract trigger mechanism; the edge intelligent perception layer outputs self-check data to the lightweight blockchain data layer, and the private chain for generating evidence records stores the data in the private chain, which is used to store enterprise self-check data and conduct authorized access. When the edge intelligent perception layer detects the trigger of a pollution exceedance signal, it will trigger the smart contract trigger mechanism; the smart contract trigger mechanism will send an alarm notification containing pollution information and a predicted path to upstream and downstream enterprise nodes. At the same time, the public chain for screening key data hash values will synchronize the data hash values to the public chain, which is used to store the hash values of key detection data to achieve data verifiability. In addition, the data hash values in the public chain will also be synchronized to the supervision platform for supervision.

[0046] As Figure 2 shown, when the edge intelligent perception layer detects that the pollution level exceeds the dynamic threshold of the current stage, it will send pollution exceedance data and related self-check information to the lightweight blockchain data layer and trigger the smart contract to request the generation of evidence records. After the lightweight blockchain data layer completes the private chain for evidence records, the smart contract will send an alarm notification (including pollution information and predicted diffusion path) to upstream and downstream enterprise nodes. The lightweight blockchain data layer will also request to synchronize the hash values of key data. After the synchronization of the public chain is completed, the upstream and downstream enterprise nodes will receive the alarm and take preventive measures.

[0047] As Figure 3 shown, first, real-time data is collected, specifically through a multi-modal sensor array. Among them, the microfluidic biosensor collects the concentration of microbial metabolites, and the hyperspectral imaging module collects spectral data of surface components; after the collected data is fused, local preprocessing (noise filtering, data calibration, feature extraction) is performed. Then, using the dynamic threshold adaptive algorithm, the current processing stage is identified to determine whether it is the cleaning stage, the slicing stage, or the drying stage, and the cleaning stage threshold, the slicing stage threshold, or the drying stage threshold is called respectively, and then pollution determination is performed to judge whether it exceeds the threshold. If so, the alarm and evidence storage process is triggered; if not, the data is determined to be normal. Finally, the processed data is output to the engine and the blockchain.

[0048] Example 3: In the edge intelligent perception layer, the dynamic threshold adaptive mechanism takes the processing stage parameter as the core input variable, and automatically loads the upper limit of the microbial detection threshold corresponding to this stage by identifying the current process stage of the production line for pollution level judgment. In the specific process, The recognition module is based on the combination of multi-modal data features, including environmental temperature and humidity, processing equipment status, and raw material position coding information. It uses preset logical rules to complete stage classification and uses the results as index parameters to call the threshold setting function. For For the calculation of, the system calls the following relational expressions:

[0049] ,

[0050] Among them, is the upper limit of the detection threshold for the primary processing stage, which is derived from the microbial index data corresponding to the stage in the enterprise's historical production records. This data obtains its upper quantile value through statistical processing and is corrected by combining a specific safety factor as the setting basis; is a correction factor related to the characteristics of the current processing stage, which is set by the system during initialization according to the konjac variety, production line equipment type, and historical pollution frequency. It belongs to the category of configuration parameters, and the system allows technicians to adjust it as needed during the deployment stage to achieve better pollution determination sensitivity, which are all extended implementation methods known to those of ordinary skill in the art.

[0051] When the dynamic threshold is actually called in the system, combined with the sensor sampling period, stage recognition is performed before each data acquisition cycle and the current threshold is dynamically updated. Since the sampling frequency does not exceed 100 ms and the threshold switching response time is controlled within 50 ms, the system can still achieve timely and accurate pollution determination in a production line environment with high-frequency dynamic changes, avoiding false alarms or missed alarms. In the pollution field dynamic reconstruction engine part, to support the integrity of the input data of the pollution trend prediction model, the historical pollution data similarity between nodes is constructed from the microbial concentration change rate sequence within a short-period sliding window, and the similarity is measured by the Euclidean distance, which is used for the calculation of the weight coefficient in the weighted interpolation algorithm. The weight parameters are all automatically calculated by the system during operation without manual intervention, ensuring the automation and robustness of the compensation logic chain.

[0052] To further clarify the blockchain trigger mechanism and the judgment logic of the smart contract, the system performs pollution determination in units of each monitoring cycle. If the detection values in two consecutive cycles both exceed the loaded in the current stage, it is regarded as a pollution event, and the smart contract is started to generate a private chain deposit record and synchronize the summary to the public chain through a preset cross-chain channel. Subsequently, the contract module automatically pushes warning information to the upstream and downstream enterprise nodes; this trigger mechanism logically avoids false alarms caused by single-point fluctuations and improves the timeliness of data uploading and the accuracy of event response. At the same time, to support the real-time nature of pollution data transmission, the minimum threshold value of the wireless communication protocol transmission rate It is preset by the system evaluation model and calculated by backtracking based on the size of the sensor data packet, the sampling frequency, and the maximum tolerable delay; when the system initializes, it conducts a rate configuration check on all nodes. If the communication ability of a certain node fails to meet the conditions, it will be marked as a non-conforming node and will not participate in the data input of the pollution field reconstruction model.

[0053] Example 4: In this example, for instance, the historical production data of konjac processing enterprises can be first called, the concentration data sequence of microbial metabolites of typical batches in the primary processing stage can be extracted, and the upper quantile value of the samples in this stage can be calculated using the distribution fitting method as the basic basis for variables ; Considering the differences in processing environments, equipment loads, and konjac varieties among different enterprises, this quantile value is corrected by combining with a safety factor set by experience to ensure that it can reflect the reasonable detection boundary in the primary processing stage.

[0054] For the threshold correction coefficient , during the system deployment process, it is assigned based on the fine processing process parameters (such as heat treatment temperature, treatment time) adopted by the target production line and elements such as the raw material source through preset rules or empirical models. This coefficient aims to reflect the differences in pollution tolerance in different processing stages. During actual operation, the system calls in real-time the threshold determination logic for pollution level judgment. The dynamic update of the threshold is driven by the processing stage identification module, which integrates information such as temperature and humidity sensor data, equipment operation status, and feedback from production line control nodes, and uses a logical discrimination algorithm to automatically determine the current stage. This mechanism ensures that even in a highly dynamic scenario with rapid switching of processing links, the pollution determination system still has the ability to match stages, avoiding misjudgment or missed detection caused by static thresholds.

[0055] For the pollution field dynamic reconstruction engine part, to improve the reconstruction accuracy in the context of missing data, the system introduces a weighted interpolation mechanism, and its weight factor consists of two parts: one is the three-dimensional spatial position relationship of the sensing nodes, and the other is the similarity measure of historical short-period pollution data. Specifically, the spatial distance between adjacent nodes and the target fault node is calculated through the initial deployment coordinates, and the closer the distance, the higher the weight; at the same time, the system extracts the pollution concentration change trend of each node within a set time window (such as the past 10 minutes) and calculates the Euclidean distance between it and the historical data of the fault node as the basis for similarity weight evaluation. After the two weights are combined, an interpolation weight matrix is formed to estimate the local pollution concentration when a node fails, ensuring that the reconstructed pollution field has continuity and reasonableness, and all belong to the extended implementation methods known to those of ordinary skill in the art.

[0056] In addition, when the original sensor data collected by the edge intelligent perception layer is preprocessed locally, it includes three key steps: noise filtering uses a method combining moving average and median filtering, data calibration corrects the deviation with reference to a pre-calibrated reference sample, and feature extraction uses methods such as principal component analysis (PCA) to perform dimensionality reduction on multi-source data, improving the expression efficiency and discriminative power of the feature vector. This feature vector is then input into the pollution determination model, and the model determines the pollution level according to the dynamic threshold loaded in the current stage. Judge the pollution level.

[0057] Wireless communication rate The evaluation model of the wireless communication rate is executed during system initialization, and its calculation logic is based on the coupling relationship among the single-packet size, the system sampling period, and the tolerated transmission delay; the system first reads the upper limit of the number of bytes of the data packet for each type of sensor, and then combines the set sampling period and the maximum data delay defined by the system to inversely deduce the lower rate limit. When a certain node cannot meet this rate requirement, the data it collects is not included in the pollution field modeling to avoid introducing delay errors by low-speed nodes. At the same time, to improve the robustness of the pollution event trigger logic, the system uses a continuous determination method within a period to execute the smart contract conditions. That is, when the data exceeds the current threshold twice continuously within a monitoring period, the system triggers the contract execution process, generates a pollution certificate with a batch number and a timestamp, and synchronizes its summary data to the public chain. This mechanism effectively avoids false triggers caused by accidental fluctuations and improves the certainty of pollution event recognition.

[0058] Example 5: First, as the front end of the system, the core function of the edge intelligent perception layer is to capture pollution information in the konjac processing environment in real time and multi-dimensionally. It continuously collects the concentration of microbial metabolites and spectral data of the konjac surface components through a multi-modal sensor array; these original data are preprocessed locally through noise filtering, data calibration, and feature extraction. This process not only reduces the burden of subsequent data transmission and processing, but more importantly, through feature extraction, the heterogeneous data of the original sensors is converted into a unified feature vector, providing a standardized input for subsequent pollution determination and pollution field modeling. The key data after local preprocessing is transmitted to the pollution field dynamic reconstruction engine and the lightweight blockchain data layer through a reliable wireless communication protocol. The selection of the wireless communication protocol needs to meet the rate requirements for real-time data transmission. , where the determination fully considers the sensor data packet size, the acquisition frequency, and the system's tolerance for response delay, ensuring the smoothness and efficiency of the data link, and all belong to the extended implementation methods known to those of ordinary skill in the art.

[0059] In the dynamic threshold adaptive algorithm, according to the different processing stages of konjac , the upper limit of the threshold for microbial detection is dynamically adjusted , and the identification of the processing stage does not rely on simple preset time points, but is automatically determined by fusing multi-modal data features collected by the edge intelligent perception layer, which includes but is not limited to environmental temperature and humidity, the operating state of processing equipment, and the position coding information of raw materials on the production line. For example, the system adopts a logical discrimination algorithm. Based on these real-time perceived parameter combinations, it accurately judges the current specific processing stage, such as cleaning, slicing, or drying. Once the current stage is determined, the system will immediately load the threshold reduction coefficient preset for this stage , and the upper limit of the microbial detection threshold is calculated according to the relationship: , where represents the upper limit of the microbial detection threshold in the primary processing stage. Its value can be obtained, for example, from the statistical analysis of the distribution of microbial index data of typical batches in the primary processing stage during the historical production process of konjac processing enterprises. These are all extended implementation methods known to those of ordinary skill in the art. By estimating the upper quantile value of historical data and correcting it by combining an empirical safety factor considering differences in different enterprises, equipment, and konjac varieties, an value that reflects both the objective pollution level and has a certain safety margin is determined. The threshold reduction coefficient is an empirical adjustment factor. Its setting comprehensively considers factors such as the technological characteristics of the current processing stage (such as heat treatment temperature, time, etc. in the fine processing stage) and historical pollution frequencies, aiming to reflect the differences in the tolerance of different processing stages to microbial pollution. The system allows technicians to optimize according to the actual situation during deployment to achieve the best pollution determination sensitivity. The dynamic update of the threshold is accompanied by the real-time identification of the processing stage, ensuring timely and accurate pollution determination in a rapidly changing production environment and effectively avoiding false alarms or missed alarms caused by using a fixed threshold.

[0060] Once a pollution exceeding signal is detected by the edge intelligent perception layer, when the detection values in two consecutive monitoring cycles both exceed the , it is considered a pollution event. At this time, the system will automatically trigger the smart contract mechanism and send a proof request to the lightweight blockchain data layer. After receiving the proof request, the smart contract module inside the lightweight blockchain data layer will execute the corresponding logic according to the preset trigger conditions (such as continuous over-limit judgment). The key information of the pollution event, including the time of occurrence, batch number, pollution index status, etc., will be generated as a proof record and stored in the private chain. The private chain is mainly used to store the company's self-inspection data and open access rights to authorized nodes (usually internal or designated auditors of the company) to protect the privacy of corporate data. At the same time, in order to meet the needs of supervision and traceability, the key test data in the private chain (at least including the konjac batch number and pollution index status) will be screened and its data hash value will be synchronized to the public chain. The public chain, as the basis for data verification, is open to the supervision platform and other relevant parties (such as upstream and downstream enterprise nodes), but only the hash summary of the data is exposed instead of the original data content, thereby ensuring the credibility and traceability of the data while taking into account the privacy of corporate data. The two chains synchronize data through a lightweight cross-chain protocol, and the data update cycle can be customized by the system according to the requirements of regulatory response time. After completing the evidence storage, the smart contract will automatically send an alarm notification to the preset upstream and downstream enterprise nodes. The notification not only contains information on excessive pollution, but also integrates the potential pollution diffusion path information predicted by the dynamic reconstruction engine of the pollution site. This on-chain automated process avoids the delay of human intervention and improves the timeliness and accuracy of information transmission.

[0061] The pollution field dynamic reconstruction engine plays the role of pollution situation awareness and prediction. It receives multi-dimensional parameter information from the edge intelligent sensing layer. Through spatio-temporal correlation modeling methods, it maps the data collected by sensors at different locations to the corresponding timestamps, constructs a three-dimensional dynamic pollution concentration field model reflecting the environmental pollution distribution of konjac processing and its changes over time. Spatio-temporal correlation modeling aims to capture the distribution characteristics of pollution in spatial positions and its evolution law over time. Based on historical pollution data, the engine uses transfer learning algorithms to predict the potential paths and trends of pollution diffusion. The application of transfer learning enables the system to draw on pollution diffusion patterns learned in different environments or historical batches, improving the accuracy of prediction in the current processing environment. To address the problem of data loss caused by sensor failures, the pollution field dynamic reconstruction engine integrates a self-healing logic chain. When it detects that a single or some sensor nodes fail, the self-healing logic chain performs interpolation calculations based on the data collected by adjacent normal nodes to reconstruct local pollution field information. During the interpolation process, a weighted interpolation algorithm is used. The weight coefficients of adjacent nodes are not fixed, but are dynamically determined based on their spatial distance from the failed node and the similarity of historical data. Nodes with a closer spatial distance are usually assigned higher weights, and nodes with a high similarity of historical data also indicate that their current state is more valuable for compensating the failed node. This dynamic weighted interpolation mechanism ensures that the system can still maintain effective pollution monitoring and risk prediction capabilities in the case of partial sensing data loss.

[0062] Finally, the human-computer interaction interface of the system provides users with intuitive monitoring and management tools. Through this interface, authorized users can view the current environmental pollution field model of konjac processing in real time, receive pollution warning information, and understand the blockchain evidence storage status; the interface also provides a parameter configuration function, allowing users to adjust some operating parameters of the system, such as the initial setting of threshold coefficients or the configuration of the recipients of warning notifications, but these adjustments need to be carried out under strict permission control to ensure the stability and security of the system. Through the organic cooperation of each module, the entire system forms a closed loop from pollution perception, dynamic determination, situation prediction to trusted evidence storage and intelligent warning, effectively improving the real-time performance, credibility and adaptive ability of konjac food safety detection and analysis, and all belong to the extended implementation methods known to those of ordinary skill in the art.

[0063] For those skilled in the art, it is obvious that the present invention is not limited to the details of the above exemplary embodiments, and can be implemented in other specific forms without departing from the spirit or basic characteristics of the present invention.

[0064] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit them. Although the present invention has been described in detail with reference to the preferred embodiments, those of ordinary skill in the art should understand that the technical solutions of the present invention can be modified or equivalently replaced without departing from the spirit and scope of the technical solutions of the present invention.

Claims

1. An IoT-based konjac food safety detection and analysis system, characterized in that, The system includes: The edge intelligent perception layer is set up in the konjac processing production line and includes a multi-modal sensor array and a dynamic threshold adaptive algorithm; the multi-modal sensor array is used to collect multi-dimensional parameter information in the konjac processing environment in real time, at least including the concentration of microbial metabolites and spectral data of the konjac surface components; the dynamic threshold adaptive algorithm is configured to , adjust the microbial detection threshold range adapted to this stage in real time according to the different processing stages of the konjac. Among them, if the current processing stage is the fine processing stage , then the upper limit of the microbial detection threshold satisfies the relationship: , where is the upper limit of the microbial detection threshold in the primary processing stage, is the preset threshold reduction coefficient in the fine processing stage; A lightweight blockchain data layer, which includes a private chain, a public chain, and a smart contract trigger mechanism. The private chain is used to store the enterprise self-inspection data from the edge intelligent perception layer and only open the access permission to authorized nodes. The public chain is used to cross-chain synchronize the selected key detection data in the private chain, at least including the konjac batch number and the pollution index status, to the supervision platform to achieve the verifiability of the data. The smart contract trigger mechanism is configured to automatically generate an evidence record in the blockchain data layer and send an alarm notification to the preset upstream and downstream enterprise nodes when the pollution degree detected by the edge intelligent perception layer exceeds the dynamic threshold corresponding to the current processing stage. A dynamic pollution field reconstruction engine, which is configured to receive and process multi-dimensional parameter information from the edge intelligent perception layer and construct a dynamic pollution field model reflecting the environmental pollution distribution and time-varying of konjac processing. Through a spatio-temporal correlation modeling method, the dynamic pollution field model maps the data collected by sensors at different positions with the corresponding timestamps to form a three-dimensional pollution concentration distribution, and predicts the potential path of pollution diffusion based on historical pollution data using a transfer learning algorithm. In addition, the dynamic pollution field reconstruction engine also includes a self-healing logic chain, which is used to perform interpolation calculations based on the data collected by adjacent normal nodes when a single or part of the sensors are detected to be faulty, and reconstruct the local pollution field information to ensure that the system can still maintain effective pollution monitoring and risk prediction capabilities in the case of partial sensor data loss.

2. The IoT-based konjac food safety detection and analysis system according to claim 1, characterized in that, The multi-modal sensor array includes a microfluidic biosensor and a hyperspectral imaging module. The microfluidic biosensor is used to detect the concentration change of microbial metabolites, and the hyperspectral imaging module is used to analyze the spectral characteristics of the konjac surface composition in real time.

3. The konjac food safety detection and analysis system based on the Internet of Things according to claim 1, characterized in that, A dynamic threshold adaptive algorithm presets corresponding microbial detection threshold ranges according to the different characteristics of the cleaning stage, slicing stage, and drying stage in the konjac processing process, automatically identifies the current processing stage during the system operation, calls the corresponding threshold range for pollution determination, and the response time of threshold switching is less than or equal to 50 milliseconds.

4. The konjac food safety detection and analysis system based on the Internet of Things according to claim 1, characterized in that, The lightweight blockchain data layer adopts a double-chain structure. When the randomly inspected key data stored on the public chain is synchronized to the supervision platform, it uses the data hash value method to achieve the verifiability of the data by the supervision department, while hiding the content of the original enterprise self-inspection data.

5. The IoT-based konjac food safety detection and analysis system according to claim 4, characterized in that, When sending an alarm notification, the smart contract trigger mechanism not only includes the pollution exceeding standard information, but also includes the potential pollution diffusion path information predicted based on the dynamic pollution field reconstruction engine.

6. The IoT-based konjac food safety detection and analysis system according to claim 5, characterized in that, The edge intelligent perception layer performs local preprocessing on the collected original sensor data, including noise filtering, data calibration, and feature extraction.

7. The konjac food safety detection and analysis system based on the Internet of Things according to claim 1, characterized in that, The system also includes a human-computer interaction interface, which is used to display the current konjac processing environmental pollution field model, pollution alarm information, and blockchain evidence status in real time, and allows authorized users to perform parameter configuration and system monitoring.

8. The IoT-based konjac food safety detection and analysis system according to claim 7, characterized in that, Between the edge intelligent perception layer and the lightweight blockchain data layer, and between the lightweight blockchain data layer and the pollution field dynamic reconstruction engine, a wireless communication protocol is used for data transmission. Among them, the data transmission rate of the wireless communication protocol shall meet the requirements that , where is the minimum rate threshold to ensure real-time data transmission.

Images