A sausage production whole-process key risk monitoring system and method based on a wireless sensor network and a score-weighted aggregation model

Abstract

This invention discloses a critical risk monitoring system and method for the entire sausage production process based on wireless sensor networks and a weighted aggregation model. The system includes a wireless acquisition module and a remote platform, where the remote platform includes modules for weight configuration, deviation scoring, alarms, and risk assessment. The wireless acquisition module sets up wireless acquisition devices for seven key stages: raw materials, grinding, chopping, adding auxiliary materials, steaming, metal handling, and the workshop environment. The weight configuration module uses the Delphi method combined with the analytic hierarchy process (AHP) to determine weights. The deviation scoring module sets interval thresholds and assigns scores based on SPC and critical limits. The alarm module implements tiered alarms according to R-value combination logic rules. The risk assessment module calculates the batch cumulative risk Q-value using a weighted aggregation model. This invention solves the problems of high latency from manual monitoring and the difficulty and cost of wired wiring, shifting the monitoring focus to cumulative risk throughout the entire process, and providing an effective means to improve the digitalization level of food safety management.

Description

Technical Field

[0001] This invention belongs to the field of food production and processing safety monitoring and Internet of Things technology, specifically relating to a risk monitoring system and method for the entire sausage production process based on wireless sensor networks and a weighted aggregation model. Background Technology

[0002] In the meat processing industry, traditional methods of monitoring production safety risks mainly rely on manual periodic sampling, paper records, or wired networks. This approach has the following problems in modern large-scale, highly automated meat processing lines:

[0003] (1) Manual sampling results in slow data feedback and requires continuous monitoring of multiple continuously changing values ​​and multiple thresholds, making it difficult to respond quickly and resulting in significant delays.

[0004] (2) Although wired network communication is fast and stable, it still has certain limitations, mainly in terms of poor flexibility and high installation and maintenance costs. Wired networks use cables to connect devices. If the cable between two devices breaks, communication between them will be interrupted, directly causing the entire communication network to fail. In meat processing, equipment used for grinding, chopping, and tumbling often rotates at high speeds, making wired equipment more susceptible to damage from mechanical forces. At the same time, as the number of factory equipment increases, installation and protection costs also increase.

[0005] (3) Lack of systematic risk assessment: The existing risk monitoring and alarm system for meat product production lacks a quantitative assessment of the cumulative effect of risks throughout the entire production process, and cannot determine the overall production compliance status of the production line.

[0006] Emerging technologies such as the Internet of Things (IoT) offer solutions, with Wireless Sensor Networks (WSNs) being a better option. Wireless communication technologies offer advantages such as large network size, high scalability, low power consumption, and high communication efficiency, enabling more flexible factory operations and reducing maintenance complexity and costs. Through IoT systems, factories can proactively monitor production processes, thereby improving their ability to identify and prevent food safety risks. Summary of the Invention

[0007] To overcome the shortcomings of existing technologies, this invention provides a risk monitoring system and method for meat product production based on wireless sensor networks, aiming to achieve digital, quantitative, and visual management of meat product production risks through wireless data acquisition devices and a weighted aggregation model.

[0008] To achieve the above objectives, the technical solution adopted by the present invention is as follows:

[0009] On the one hand, the present invention provides a key risk monitoring system for the entire sausage production process based on wireless sensor networks and a weighted aggregation model, which includes: a wireless acquisition module and a remote platform, wherein the remote platform includes a weight configuration module, a deviation scoring module, an alarm module, and a risk assessment module;

[0010] The wireless acquisition module is used to collect seven key risk monitoring indicators for sausage production, including: a raw material temperature wireless acquisition device for collecting raw material temperature information; a ground meat sample temperature wireless acquisition device for collecting ground meat sample temperature information; a chopped meat sample temperature wireless acquisition device for collecting meat sample temperature information at each stage of chopping; an auxiliary material addition amount wireless acquisition device installed on the weighing device for collecting auxiliary material addition amount information; a steaming environment temperature wireless acquisition device for collecting steaming temperature and time information; a metal detection wireless acquisition device installed on the metal detector at the end of the production line for collecting metal detection information; and a workshop environment temperature and humidity wireless acquisition device installed in the central area of ​​the main operating space of the workshop for collecting workshop environment temperature and humidity information. Each wireless acquisition device includes a microcontroller unit, a data acquisition unit, a wireless communication unit, and a power management unit. The wireless acquisition device reads the data collected by the data acquisition unit according to a preset sampling frequency or button operation and uploads it to the remote platform through the wireless communication unit.

[0011] Preferably, the data acquisition unit of the wireless data collector is configured with different front-end devices according to the type of monitored object and the requirements of actual operation:

[0012] The raw material feeding temperature wireless acquisition device uses a high-precision needle-type PT100 temperature sensor.

[0013] Both the ground meat sample temperature wireless acquisition device and the chopped meat sample temperature wireless acquisition device use the SA10AC infrared temperature sensor, and the output signal is RS485.

[0014] The auxiliary material addition acquisition device is equipped with a high-precision communication electronic scale, and the output signal is RS232.

[0015] The wireless temperature acquisition device for the steaming environment uses a KS-SHT5KTP patch temperature sensor, and the output signal is RS485.

[0016] The wireless metal detector is compatible with the HL4008 intelligent metal detector, with detection sensitivities of Fe≤Φ1.5mm; NoFe≤Φ2.0 mm; SuS≤Φ2.5 mm.

[0017] The workshop environment temperature and humidity acquisition device uses the KS-SHTE1KT temperature and humidity sensor, and the output signal is RS485.

[0018] Preferably, the power management unit of the wireless data collector adopts two power supply schemes according to the specific characteristics of the monitored object:

[0019] The raw material temperature wireless collector, the minced meat sample temperature wireless collector, and the chopped meat sample temperature wireless collector are powered by rechargeable lithium batteries.

[0020] The wireless data acquisition devices for auxiliary material addition, metal detection, cooking environment temperature, and workshop environment temperature and humidity are all powered by a DC12V 1A adapter.

[0021] The weight configuration module is used to store the weight coefficients Wi corresponding to the seven key risk monitoring indicators. Wi is the subjective and objective weight coefficient obtained by using the De Fel method and the Analytic Hierarchy Process (APH).

[0022] The deviation scoring module, for each key risk monitoring indicator, sets normal, warning, and danger zones based on statistical process control (SPC) and key limits. For danger zone limits, the more suitable range between μ ± 3σ and the key limit is always selected as the final threshold. For warning zone limits, based on μ ± 2σ, when statistical fluctuations intensify and cause the value to reach or even exceed the key limit, the warning zone limit is further narrowed to 90% of the key limit to ensure the effectiveness of the warning interval. The warning zone should be included within the danger zone. The monitoring data from each wireless data collector is scored for deviation according to the defined interval limits. Specifically, the following is an explanation:

[0023] The normal range is: (LWL, UWL)

[0024] The warning interval is: (LCL, LWL]∪[UWL, UCL)

[0025] The danger zone is defined as: (-∞, LCL] ∪ [UCL, +∞)

[0026] The values ​​of the first lower threshold LWL, the second lower threshold LCL, the first upper threshold UWL, and the second upper threshold UCL are selected based on the following specific circumstances:

[0027] If μ + 2σ ≤ CL H ≤ μ + 3σ, UWL =μ + 2σ, UCL = CL H ;

[0028] If CL H ≤ μ + 2σ,UWL = CL H - |10% CL H |, UCL = CL H ;

[0029] If CL H ≥ μ + 3σ or no definite CL H When, UWL =μ + 2σ, UCL =μ + 3σ;

[0030] If μ - 2σ ≥ CL L ≥ μ - 3σ, LWL =μ - 2σ, LCL= CL L

[0031] If CL L ≥ μ - 2σ, LWL =CL L +10% CL L |, LCL = CL L ;

[0032] If CL L ≤ μ - 3σ or no definite CL L When, LWL =μ - 2σ, LCL =μ - 3σ;

[0033] Among them, CL H CL represents the critical upper limit value. L This indicates the critical lower limit value.

[0034] The normal zone is assigned a score of Ri=0, the warning zone is assigned a score of Ri=1, and the danger zone is assigned a score of Ri=2.

[0035] The alarm module generates production alarms based on preset R-value combination logic rules, which include at least three levels: If any CCP risk monitoring indicator shows Rccp=2, a level one (red) alarm is triggered; If no red warning is issued, any OPRP risk monitoring indicator R... OPRP If Ri = 2 or three indicators accumulate to 1, a Level 2 (yellow) warning is triggered; if all indicators Ri ≤ 1 and the cumulative number of Ri ≤ 2, the production batch is in a normal (green) state. Among them, the amount of auxiliary materials added, cooking temperature and time, metal detection, and workshop ambient temperature and humidity belong to CCP; the raw material temperature, minced meat sample temperature, and meat sample temperature at each stage of chopping belong to OPRP.

[0036] The risk assessment module calculates the batch cumulative risk value based on the assigned scores Ri and weights Wi of the real-time collected data of each key risk monitoring indicator. The higher the Q value, the further the production system deviates from steady state, and the greater the degree of high risk tendency.

[0037] Preferably, the weighting coefficients Wi are obtained using the De Fel method and the Analytic Hierarchy Process (APH). Specifically, experts in the fields of meat processing and food safety quality control are invited.

[0038] First, the importance of the seven key risk monitoring indicators in controlling food safety risks during sausage production was scored.

[0039] Then, the experts compared the relative importance of each indicator pairwise, constructed a judgment matrix, calculated the eigenvectors, and performed a consistency test to obtain the initial weights.

[0040] Finally, after multiple rounds of expert feedback and adjustments, the weight values ​​for each key risk monitoring indicator were determined.

[0041] In addition to the aforementioned modules (weight configuration module, deviation scoring module, alarm module, and risk assessment module), the remote platform also includes four display modules: a production dashboard module, an equipment management module, a historical data module, and a quality management module. Specifically:

[0042] The production screen module is used to visualize the 3D models of production equipment and wireless data acquisition devices, and to display the real-time monitoring data uploaded by each wireless data acquisition module, as well as the results of the alarm module and risk assessment module.

[0043] The device management module is used to manage the wireless collectors and set independent risk limit ranges for each wireless collector based on the threshold values ​​of each interval determined by the deviation scoring module.

[0044] The historical data module is used to store and retrieve historical monitoring data;

[0045] The quality management module is used to perform statistical analysis and visualization of product quality, including but not limited to statistics on the number of good and defective products, trends in the yield rate, and statistics on the classification of causes of defective products.

[0046] On the other hand, the present invention also provides a method for monitoring key risks throughout the sausage production process based on wireless sensor networks and a weighted aggregation model, comprising the following steps:

[0047] Step S1: Multi-source data acquisition. Real-time data is collected using wireless data acquisition devices at seven key risk monitoring stages: raw material receiving, grinding, chopping, auxiliary material addition, cooking, metal detection, and workshop environment.

[0048] Step S2: Establish risk indicator weights. Expert evaluations are solicited using the Delphi method, a judgment matrix is ​​constructed, and the relative risk weights Wi for each monitoring link are calculated using the Analytic Hierarchy Process (AHP), ensuring that the consistency ratio CR of the weight judgment matrix is ​​less than 0.1.

[0049] Step S3: Dual Threshold Setting. Combining the statistical process control (SPC) results and critical limits (CL) of historical production data, set interval thresholds for normal, warning, and danger zones for each indicator.

[0050] Step S4: Deviation Scoring. Assign a corresponding risk score Ri based on the range in which each risk monitoring data falls: if the monitoring data is in the normal range, then Ri = 0; if the monitoring data is in the warning range, then Ri = 1; if the monitoring data is in the danger range, then Ri = 2.

[0051] Step S5: Hierarchical alarm: Trigger production alarms based on preset R-value combinational logic rules.

[0052] Step S6: Calculation of batch cumulative risk Q-value and judgment of risk trend. Based on the scoring results Ri and corresponding weights Wi of each indicator, the risk is calculated using the formula... Calculate the batch cumulative risk value Q, and judge the production line status based on the trend of Q value changes.

[0053] Beneficial effects

[0054] The present invention has the following beneficial effects:

[0055] (1) High flexibility, strong scalability and controllable cost: The use of wireless collectors to build a wireless sensor network eliminates the need for large-scale wiring modifications to existing production lines, minimizes construction impact, reduces equipment modification costs, and results in relatively low overall investment. The construction impact is small, enabling rapid deployment and flexible expansion, which is convenient for large-scale promotion and application.

[0056] (2) Simple operation: By assigning continuous data to discrete 0, 1, and 2, the cognitive load of on-site operators is reduced, thereby improving the response speed.

[0057] (3) Coverage of key risks across the entire chain: covering the core physical, chemical and microbiological risk control points from raw materials to finished products.

[0058] (4) Systematic and quantitative risk assessment: The scores Ri of the seven key risk monitoring indicators are aggregated into the quantitative score Q value to achieve comprehensive and real-time monitoring of risks throughout the entire production process, providing managers with a global basis for production control judgment. Attached Figure Description

[0059] Figure 1 This is a schematic diagram of the system architecture of a key risk monitoring system for the entire sausage production process based on wireless sensor networks and a weighted aggregation model, as described in this invention.

[0060] Figure 2 This is a flowchart illustrating a method for monitoring key risks throughout the sausage production process based on wireless sensor networks and a weighted aggregation model, as described in this invention.

[0061] Figure 3This is a schematic diagram showing the results of the environmental collector metering performance test in the wireless collector module of Embodiment 1 of the present invention.

[0062] Figure 4 This is a schematic diagram of the results of the risk indicator weighting module in Embodiment 1 of the present invention. Detailed Implementation

[0063] The present invention will now be described in detail with reference to the accompanying drawings and specific embodiments. It should be noted that the following embodiments are for illustrative purposes only and do not constitute a limitation on the scope of protection of the present invention.

[0064] Example 1

[0065] This embodiment provides a key risk monitoring system for sausage production based on wireless sensor networks and a weighted aggregation model. For example... Figure 1 As shown, the system includes a wireless acquisition module and a remote platform. The remote platform includes a weight configuration module, a deviation scoring module, an alarm module, a risk assessment module, a production screen module, an equipment management module, a historical data module, and a quality management module.

[0066] The wireless data acquisition module includes wireless data collectors specifically designed for the seven key risks in sausage production. These include: a wireless data collector for raw material temperature; a wireless data collector for minced meat sample temperature; a wireless data collector for chopped meat sample temperature at each stage of chopping; a wireless data collector for auxiliary material addition amount on the weighing device; a wireless data collector for cooking environment temperature and time; a wireless data collector for metal detection on the metal detector at the end of the production line; and a wireless data collector for workshop environmental temperature and humidity in the central area of ​​the main operating space. All wireless data collectors were designed and manufactured by the inventor, including general hardware structure design and embedded program development. Data from each wireless data collector is uploaded and stored on a remote platform via Wi-Fi, providing a foundation for subsequent data analysis. The HACCP positioning and control hazard types for the seven key risk indicators are shown in Table 1.

[0067] Table 1 7 Key Risk Indicators

[0068]

[0069] The general hardware structure of the wireless data collector includes a microcontroller unit, a data acquisition unit, a wireless communication unit, and a power management unit.

[0070] The microcontroller unit uniformly uses the STM32F103C8T6 microcontroller as the central core processing unit for each wireless data acquisition unit, significantly reducing production and maintenance costs while achieving efficient reuse of the underlying firmware. The STM32 series chips offer good hardware and software compatibility, especially during subsequent system upgrades that require replacing the chip with one of the same series with one featuring more pins; the existing code can run directly without significant modification, making porting convenient. The built-in acquisition control program manages sensor data reading, data preprocessing, Wi-Fi transmission and reception, OLED interface refresh, and button indicator light operation responses.

[0071] The data acquisition unit, the wireless collector, is equipped with different sensors according to the type of monitoring object and the requirements of actual operation.

[0072] The raw material feeding temperature wireless acquisition device uses a high-precision needle-type PT100 temperature sensor with a measurement range of -20-200℃.

[0073] The wireless temperature acquisition devices for meat samples during grinding and chopping processes both use SA10AC infrared temperature sensors, with an output signal of RS485 and a measurement range of -20 to 100℃.

[0074] The wireless temperature acquisition device for the steaming environment uses a KS-SHT5KTP patch temperature sensor with a measurement range of -40-125℃ and an accuracy of ±1℃. The sensor adopts intelligent computer self-calibration technology, has an IP65 waterproof rating, and uses RS485 for communication.

[0075] The workshop environment temperature and humidity acquisition module uses the KS-SHTE1KT temperature and humidity sensor, with a measurement range of -40-125℃ & 0-100%, an accuracy of ±1℃ & ±1.5%RH, and a communication method of RS485.

[0076] The metal detection and acquisition module uses the HL4008 intelligent metal detector developed by Shanghai Biancheng Electronics Technology Co., Ltd., with a detection sensitivity of Fe≤Φ1.5 mm; NoFe≤Φ2.0 mm; SuS≤Φ2.5 mm. This detection sensitivity meets the requirements for detecting ferrous, non-ferrous metals, and stainless steel in sausage products. The metal detector is equipped with an audible and visual alarm and a rejector. When metal is detected, an audible and visual alarm is triggered simultaneously, and unqualified products are rejected by the rejector at the end of the equipment.

[0077] The wireless communication unit uses a BW16 dual-band Wi-Fi and Bluetooth combined module. The microcontroller controls the BW16 module via USART to achieve Wi-Fi connection. When Wi-Fi is successfully connected, the u23 indicator light is green, indicating that data can be uploaded to the server; if not connected, the light is red.

[0078] The power management unit employs two power supply schemes based on the specific characteristics of the monitored objects: a rechargeable lithium battery powers the wireless acquisition devices for raw material feeding temperature and the temperature of minced and chopped meat samples; while a DC 12V 1A adapter powers the devices for auxiliary material addition, metal detection, cooking environment temperature, and workshop ambient temperature and humidity. To ensure the stable operation of the microcontroller and each module, a multi-stage step-down and voltage-regulating circuit is designed for each module.

[0079] The embedded program of the wireless data collector is embedded in the microcontroller of each data collector, including:

[0080] Main program: The main program consists of two parts: an initialization module and a loop control module.

[0081] Information Acquisition Program: The raw material temperature acquisition program uses the ADC1 channel to read the weak voltage signal across the PT100 sensor every 10ms. A median filtering algorithm is used to process the voltage signal to improve data accuracy. Based on this voltage signal, a resistance value is calculated. The program uses a binary search method (tem_find) to quickly find the temperature corresponding to the current resistance in the standard PT100 resistance value calibration table array r_t[]. The initial measured temperature is linearly corrected using calibration parameters stored in the Flash memory chip.

[0082] The wireless temperature acquisition device for minced and chopped meat samples uses sensors that communicate with the microcontroller via an RS485 line. After receiving a Modbus RTU query command, the sensor sends the monitoring data back to the microcontroller through USART1. Upon receiving and confirming the data, the microcontroller uses a shift algorithm to concatenate the high and low bits of data in the data reception buffer to obtain the raw temperature measurement value. Infrared sensors are highly sensitive, and the measured values ​​are prone to fluctuations. To obtain stable temperature measurements, the program employs a sliding window filtering algorithm.

[0083] The auxiliary material addition wireless data acquisition device obtains the weight by receiving serial port messages sent by the electronic scale. The microcontroller does not need to actively send query commands, but instead uses the USART3 interface to listen to the serial port data in real time, and adopts an asynchronous reception method of "serial port idle interrupt + DMA".

[0084] The two wireless data acquisition devices for cooking temperature and workshop environment use a unified architecture in their data acquisition program design. The temperature and humidity sensors used in the data acquisition devices communicate with the microcontroller via RS485 lines. The microcontroller communicates digitally with the sensors via the USART2 serial port, strictly adhering to the Modbus RTU protocol's polling and response mechanism. The system uses timer TIM2 to trigger a sensor reading task every 10 seconds.

[0085] Wireless communication program: The wireless acquisition devices for raw material temperature, meat sample temperature during grinding and chopping all employ a 5-point temperature acquisition method, with the system automatically calculating the average and actively reporting the data. Its microcontroller interacts with the Wi-Fi module via USART2 and AT commands. Operationally, personnel collect temperature data sequentially at 5 different locations on the meat sample using buttons. After the 5-point data acquisition is complete, the program automatically calculates the average value t_value and triggers a transmission flag.

[0086] The four types of wireless data collectors—for auxiliary material addition, cooking environment temperature and humidity, workshop environment temperature and humidity, and metal detection—interact with the BW16 wireless module via the USART1 interface. The triggering logic for data reporting differs among the four types of collectors.

[0087] The auxiliary material addition data acquisition system employs an event-triggered mechanism. When the acquisition module detects the serial port message sent by the "Confirm" button on the electronic scale, the wireless communication module immediately calls a key function to complete the protocol packetization of the data and upload it to the cloud.

[0088] The cooking and workshop environment data acquisition unit uses a TIM3 counter to report temperature and humidity measurements to a remote platform every 10 seconds. The reported data is in a 25-byte long message format, which includes the device identification, function code, temperature and humidity values, and a CRC16 checksum to ensure data integrity.

[0089] The metal detection data collector uses an abnormal alarm triggering mechanism. When the device detects metal, or when a staff member presses the send button, the program begins to package the data and report it to the remote platform.

[0090] All seven types of wireless data collectors use the TCP protocol to transmit monitoring data to the remote platform (server address: 47.103.79.173, port: 6060). Upon successful system connection, a registration packet is automatically sent. To maintain a long-term connection and monitor network status, the system sends a heartbeat message every 60 seconds.

[0091] OLED screen driver and button operation program: The wireless acquisition devices for raw material temperature, meat sample temperature during grinding and chopping are consistent in both the OLED screen driver and button operation program. The system refreshes the screen every 20ms, displaying the current measured temperature value in real time. To facilitate the operation of 5-point temperature measurement, the UI interface uses the status variable ui (range 1-5) to inform the operator of the current temperature measurement sequence number, and uses a checkmark to indicate whether the data of that temperature measurement point has been locked and recorded. The program is designed with an abnormal display function: when the temperature exceeds the sensor's measurement range or there is an internal disconnection in the equipment, the screen display will trigger a clear screen and display an abnormal symbol "---" to prevent operator misinterpretation. The three wireless acquisition devices for cooking environment temperature, workshop environment temperature and humidity, and metal detection send drawing text commands to the OLED screen through the USART3 interface. To improve data readability, the program uses blue (color code 61) for numerical rendering and uses DCV24 commands to precisely control the display position of the data on the screen coordinate axis.

[0092] In this example, the metrological performance of seven types of wireless data collectors was evaluated, and their communication reliability and real-time performance were tested. Packet loss rate and transmission time were used as evaluation indicators for communication reliability and real-time performance.

[0093] In this example, the metrological performance results are as follows: Figure 3 As shown in Tables 2 and 3, within the test range of -5℃ to 20℃, the absolute errors of the three wireless data acquisition devices for raw materials, grinding, and chopping were all ≤0.5℃, meeting the temperature monitoring requirements in the meat processing industry. The standard deviation of each data acquisition device was between 0.07 and 0.32℃, demonstrating good stability.

[0094] Under a constant temperature setting of 37℃, the temperature measurements from both the cooking environment and the workshop environment data acquisition devices remained stable, with the measured temperature curves nearly horizontal, and absolute errors of 0.2℃ and 0.1℃, respectively. Under a constant humidity setting of 48%, the humidity measurement curve from the workshop environment wireless data acquisition device showed minimal fluctuation, with a maximum absolute error of 1.5%, both lower than the maximum error of conventional humidity monitoring equipment. The results indicate that the wireless data acquisition devices for both environmental parameters possess strong anti-interference capabilities and good data continuity, meeting the requirements for real-time, high-frequency, and long-term monitoring of environmental temperature and humidity during production.

[0095] For three standard metal test blocks of FEΦ1.2mm, SUSΦ2.5mm, and NONFEΦ2.0mm, the metal detection collector consistently recorded 50 counts throughout the entire testing cycle, with zero missed counts and a 100% consistency rate. The results demonstrate that the metal detection collector can accurately capture the trigger signals from the underlying metal detector, meeting the real-time monitoring requirements for key physical hazard control points during meat processing.

[0096] Table 2 Temperature measurement results of meat samples using wireless data acquisition device

[0097]

[0098] Table 3 Metal Detection Test Results

[0099]

[0100] In this example, the communication stability and real-time performance evaluation results are shown in Tables 4 and 5. No packet loss occurred in the data transmission of any of the wireless data collectors during the test period, meeting the requirements for stable transmission of the monitoring system. In the wireless router environment, the average data transmission time of the data collectors was 56.20 ± 8.34 ms, with a maximum of 72 ms during the test period, indicating smooth transmission. In the mobile hotspot environment, the data transmission time increased to 81.10 ± 16.34 ms, with a maximum of 111 ms. The results show that the data transmission speed of the data collectors in both different wireless network environments meets the needs of the risk monitoring system, enabling timely risk response. In comparison, the data collectors in the wireless router environment exhibit lower and more stable latency.

[0101] Table 4. Test results of remote data transmission for each wireless data collector

[0102]

[0103] Table 5. Test results of data transmission latency under different wireless network environments

[0104]

[0105] The weight configuration module is used to store the weight coefficients Wi corresponding to each key risk monitoring indicator; Wi is the weight coefficient obtained by using the De Fel method and the Analytic Hierarchy Process (APH).

[0106] This example invited 10 experts in the field of meat processing and food safety quality control to first score the importance of seven key risk monitoring indicators in controlling food safety risks during sausage production. In the second round, the experts were required to compare the relative importance of each indicator pairwise, constructing a judgment matrix using a 1-9 scale.

[0107] In this embodiment, the expert scores for the importance of each risk monitoring indicator are as follows: Figure 4As shown in (A) and Table 6, the results indicate that the experts' scores on the importance of each indicator show a good central tendency, with the average scores for each indicator ranging from 6.40 to 9.10. The highest scores were for steaming temperature and time (9.10 ± 0.88), while the lowest score was for minced meat sample temperature (6.40 ± 0.97). The coefficients of variation for all indicators were less than 0.25, and no extreme outliers were found in the box plots. This indicates that the expert group had a high degree of consistency and coordination in their understanding of the importance of each risk indicator, allowing the survey to be stopped without multiple rounds of expert consultation, thus ensuring the reliability of the data.

[0108] Table 6. Statistical Results of Importance Scores for the 7 Major Indicators

[0109]

[0110] In this embodiment, based on the pairwise comparison and scoring results of the relative importance of each indicator, the Analytic Hierarchy Process (AHP) is used to calculate the weight of each risk indicator, and the weight distribution is as follows. Figure 4 As shown in (B), the consistency test results are presented in Table 7. The results show that the maximum eigenvalue of the judgment matrix is ​​λmax = 7.3429, the consistency index CI = 0.0571, and the consistency ratio CR = 0.0420. CR < 0.1, indicating that the judgment matrix passes the consistency test and the weight allocation results are reliable. After normalization, the risk weight coefficients for each risk monitoring indicator are: raw material temperature 5.11%, minced meat sample temperature 4.69%, meat sample temperature at each chopping stage 4.47%, auxiliary material addition amount 19.90%, cooking temperature and time 34.98%, metal monitoring 20.99%, and workshop temperature and humidity 9.87%. Once determined, the weights remain unchanged during operation unless reassessed due to significant changes in the production process.

[0111] Table 7 Consistency Test Results of the Comprehensive Judgment Matrix

[0112]

[0113] In this embodiment, to verify the consistency between the expert consensus and the mathematical model, Spearman rank correlation analysis was performed on the importance score and weight coefficient. The results are as follows: Figure 4 As shown in (C), the distribution trend of each data point along the diagonal is obvious. The Spearman rank correlation coefficient rs=0.991, p<0.001, indicating that the importance score is significantly positively correlated with the AHP weight. The importance ranking of the indicators obtained by the two methods is highly consistent, which further verifies the credibility of the results.

[0114] The deviation scoring module sets threshold ranges for each key risk monitoring indicator by combining statistical process control (SPC) and HACCP key limits.

[0115] In this embodiment, the data range of the indicator is divided into three zones, each corresponding to a different scoring value R. i The details are shown in Table 8.

[0116] Table 8. R-value Assignment Rules

[0117]

[0118] In practical applications, when the calculation results of SPC conflict with industry regulations, or when there are no clear and unified industry or process regulations for this indicator, this embodiment uses the following fusion rules to determine the limits for each interval:

[0119] Safety first principle: If the range defined by μ±3σ calculated based on historical data is more stringent than CL, or if there is no clear CL, in order to maximize food safety, the more restrictive range of μ±3σ shall be selected as the limit value of the danger zone.

[0120] Regulatory bottom line principle: If statistical fluctuations are large enough to cause μ±2σ to exceed the regulatory bottom line specified by CL, the warning zone limit will be further tightened on the basis of CL, and modified to CLs×90% to ensure the operability of the warning zone.

[0121] Specifically:

[0122] The normal range is: (LWL, UWL)

[0123] The warning interval is: (LCL, LWL]∪[UWL, UCL)

[0124] The danger zone is defined as: (-∞, LCL] ∪ [UCL, +∞)

[0125] The values ​​of the first lower threshold LWL, the second lower threshold LCL, the first upper threshold UWL, and the second upper threshold UCL are selected based on the following specific circumstances:

[0126] If μ + 2σ≤CL H ≤μ + 3σ, UWL =μ + 2σ, UCL = CL H ;

[0127] If CL H ≤ μ + 2σ,UWL = CL H - |10% CL H |, UCL = CL H ;

[0128] If CL H ≥ μ + 3σ or no definite CL H When, UWL =μ + 2σ, UCL =μ + 3σ;

[0129] If μ - 2σ ≥ CL L ≥ μ - 3σ, LWL =μ - 2σ, LCL= CL L

[0130] If CL L ≥ μ - 2σ, LWL =CL L +10% CL L |, LCL = CL L ;

[0131] If CL L ≤ μ - 3σ or no definite CL L When, LWL =μ - 2σ, LCL =μ - 3σ;

[0132] Among them, CL H CL represents the critical upper limit value. L This indicates the critical lower limit value.

[0133] In this embodiment, the process of setting the limit values ​​for each interval is explained in detail, taking the temperature monitoring index of meat samples at each stage of chopping as an example. According to the process specifications, the final sample output temperature at this stage should be <12℃.

[0134] Statistical analysis: Using the corresponding wireless data acquisition device, production data of 30 batches of meat samples from the three stages of chopping and mixing were collected. The mean temperature of the meat samples in the first stage of chopping and mixing was calculated to be μ=3.84℃, with a standard deviation σ=0.61℃; the mean temperature of the meat samples in the second stage of chopping and mixing was μ=8.87℃, with a standard deviation σ=0.51℃; and the mean temperature of the meat samples in the third stage of chopping and mixing was μ=10.68℃, with a standard deviation σ=0.57℃.

[0135] Three-interval limit settings:

[0136] Temperature of meat sample during the first stage of chopping and mixing:

[0137] Warning zone settings: The lower limit of the warning zone is μ-2σ=2.63℃, and the upper limit is μ+2σ=5.06℃. Considering that in actual operation, the data precision of the wireless data collector is retained to one decimal place, the lower limit of the warning zone is 2.6℃, and the upper limit is 5.1℃. If the temperature falls within this range, Ri = 1.

[0138] Hazard zone setting: Based on process specifications, the lower limit of the hazardous zone is set at 2.0℃, and the upper limit at 6.0℃. Temperatures falling within this range are classified as Ri = 2.

[0139] Meat sample temperature during the second stage of chopping and mixing:

[0140] Warning zone settings: The lower limit of the warning zone is μ-2σ=7.85℃, and the upper limit is μ+2σ=9.88℃. Considering that in actual operation, the data precision of the wireless data collector is retained to one decimal place, the lower limit of the warning zone is 9.9℃, and the upper limit is 7.9℃. If the temperature falls within this range, Ri = 1.

[0141] Hazard zone setting: Comparing the process specifications with μ±3σ, where μ±3σ is less than the process specifications, the lower limit of the hazard zone is set as μ-3σ=7.34℃, and the upper limit as μ+3σ=10.39℃. Considering that in actual operation, the data from the wireless data acquisition device is retained to one decimal place, the lower limit of the warning zone is 7.3℃, and the upper limit is 10.4℃. If the temperature falls within this range, Ri = 2.

[0142] Meat sample temperature during the third stage of chopping and mixing:

[0143] Warning zone settings: The lower limit of the warning zone is μ-2σ=9.53℃, and the upper limit is μ+2σ=11.84℃. Considering that in actual operation, the data precision of the wireless data collector is retained to one decimal place, the lower limit of the warning zone is 9.5℃, and the upper limit is 11.8℃. If the temperature falls within this range, Ri = 1.

[0144] Hazard zone setting: According to relevant regulations, the final sample output temperature in this step should be <12℃, but no clear lower limit is specified. Comparing μ ± 3σ with the process upper limit of 12℃, μ + 3σ = 12.39℃ (greater than 12℃). Therefore, the lower limit of the hazard zone is set as μ - 3σ = 8.97℃, and the upper limit of the hazard zone is taken as the process-specified value of 12.0℃. Considering that in actual operation, the data accuracy of the wireless acquisition device is retained to one decimal place, the lower limit of the warning zone is 9.0℃. If the temperature falls within this range, Ri = 2.

[0145] The limit values ​​for the remaining indicators are set in the same way as in the example above. They are all based on the SPC calculation results of historical data and the HACCP critical limits or process specifications to determine the specific boundaries of the normal zone, warning zone and danger zone.

[0146] Metal detection and excipient addition are unique. Metal detection is a count-based indicator and a critical control point, a key step in controlling the introduction of physical hazards. Therefore, this indicator does not have a warning zone, only a normal zone (0 items) and a danger zone (≥1 item). Excipient addition mainly involves nitrites. The logic for setting this range is based on process standards and national standard limits, rather than SPC rules. The warning zone limit is set based on the maximum permissible physical error of the weighing device, in addition to the set production formula amount. In this embodiment, a secondary electronic balance is used as the weighing device, and the maximum permissible physical error of the device under the corresponding load is ±0.05 g.

[0147] In this embodiment, the scoring rule is a fixed logic that is not adjusted with changes in process parameters, ensuring the comparability of risk scores between batches.

[0148] The alarm module triggers the corresponding alarm level based on a preset R-value combination logic rule. The combination logic rule includes at least three levels, as shown in Table 9.

[0149] Table 9 Alarm Logic Based on R-value Combinational Logic Rules

[0150]

[0151] The risk assessment module calculates the cumulative risk Q value for each production batch based on the R-value score Ri and corresponding weight Wi of the real-time monitoring data of each indicator. The formula is as follows:

[0152]

[0153] in:

[0154] i: Represents the monitoring link number (i=1 to 7, corresponding to raw material temperature, grinding, chopping, auxiliary materials, cooking, metal inspection, and environment).

[0155] Wi: The weight of the risk indicator for the i-th stage determined by the Deffer method-AHP, satisfying...

[0156] Ri: The real-time risk score for the i-th stage. Its value depends on the range in which the indicator falls.

[0157] The data is in the normal range: Ri=0;

[0158] Data exceeds the warning zone threshold: Ri=1;

[0159] The data exceeds the danger zone threshold range: Ri=2.

[0160] In this embodiment, the Q value quantifies the "compliance level" of the entire production batch, reflecting the overall risk level of the batch. Within the same alarm color level, a higher Q value indicates that the risk is more concentrated in core processes, i.e., high-weight processes, thus assisting management in determining the order of batch checks. The system displays the trend of Q value changes with batches on the display interface as a line graph, forming an "overall production trend curve." When the Q value shows an upward trend for three consecutive batches, the system issues a "risk breakthrough probability increased" alert, suggesting that management check the production line status. The preset can be configured according to different meat product categories or enterprise quality objectives.

[0161] In this embodiment, the Q value does not have the authority to trigger or change the predetermined alarm color (level), thereby avoiding logical conflicts caused by the overlap of multiple indicator value thresholds.

[0162] The remote platform is used to receive, process, and display real-time monitoring data uploaded by various wireless data collectors. In addition to the aforementioned modules, the platform also includes four major display and management modules: a production dashboard module, an equipment management module, a historical data module, and a quality management module.

[0163] The production dashboard module displays 3D models of key processing equipment and various wireless data collectors, and can dynamically display monitoring data from each wireless collector. When a user clicks on a device icon, an information box pops up displaying the latest monitoring data for that device. This module includes historical alert records, such as specific alarm color levels, alarm reasons, corresponding batch Q-values, and alarm times.

[0164] The device management module centrally manages all wireless data collectors, allowing users to view each device's serial number, name, and online status through its interface. The platform allows users to set independent risk threshold ranges for different data collectors. This setting is based on the dual-threshold rule determined by the deviation scoring module, thus establishing the warning and danger zone boundaries for each control point.

[0165] The historical monitoring data module stores all monitoring data. Users can query historical data recorded by each wireless data collector on this interface, with data trends displayed visually as line graphs. The queried historical data can be exported as a spreadsheet file for subsequent data analysis.

[0166] The quality management module is primarily used to analyze the overall production status of the product. The platform displays the current number of good and defective products on the interface, showing the yield rate trend over the past 7 days in the form of a line graph. The interface also includes a pie chart to statistically analyze the specific reasons for defective products, such as deviations in auxiliary material addition, detection of metal foreign objects, and equipment malfunctions.

[0167] Example 2

[0168] This embodiment provides a critical risk monitoring system for the entire sausage production process based on wireless sensor networks and a weighted aggregation model. The method is implemented using the system described in Embodiment 1. Figure 2 As shown, the specific execution steps are as follows:

[0169] Step S1: Multi-source data acquisition

[0170] A wireless sensor network comprised of seven wireless data acquisition devices collects process parameters for each key risk indicator in real time, following the data acquisition logic corresponding to each indicator or a preset sampling frequency. The data is then uploaded to a remote platform via a wireless communication module. The data frame format includes: data acquisition device ID, timestamp, and monitoring value.

[0171] Step S2: Determining Risk Weights

[0172] The Delphi method combined with the analytic hierarchy process (AHP) was used to determine the risk weights Wi of each key risk monitoring indicator. Specifically, the expert group first compared the importance of each indicator pairwise, constructed a judgment matrix, calculated the eigenvectors, and performed a consistency check to obtain the initial weights. After multiple rounds of expert feedback and adjustments, the weight coefficients of the seven key risk monitoring indicators were finally determined. Once determined, these weight coefficients remained unchanged during system operation.

[0173] Step S3: Dual Threshold Setting

[0174] For each risk monitoring indicator, the mean μ and standard deviation σ of its historical data are calculated based on statistical process control, and a key upper limit value CL for that indicator is introduced. H and / or the critical lower limit value CL L Determine the limits for each interval based on the upper and lower sides respectively:

[0175] Danger zone: μ+3σ and CL are taken from the upper limit side. H For the smaller one, the lower limit is taken as μ-3σ and CL. L The larger of the two; if there is no corresponding CL on a certain side, then μ±3σ is taken on that side;

[0176] Warning zone: The upper and lower limits are defaulted to μ ± 2σ; if μ + 2σ < CL H Or μ - 2σ > CL L If so, the side boundary is tightened to 90% (upper limit) or 110% (lower limit) of CL.

[0177] Normal zone: The area defined by the warning zone.

[0178] Once the above boundary limits are determined, they are used in step S4.

[0179] Step S4: Deviation Assignment

[0180] Based on the real-time monitoring data of various risk indicators for the current production batch, a corresponding R value is assigned according to the range in which they fall. If it is in the normal zone, the score is Ri=0; if it is in the warning zone, the score is Ri=1; if it is in the danger zone, the score is Ri=2.

[0181] Step S5: Tiered Alarm

[0182] The compliance of the production process is judged based on preset R-value combination logic rules. If any CCP indicator R... CCP =2, triggering a Level 1 (red) alarm, indicating a serious breach of process limits in the production process; if not under red conditions, any OPRP indicator R OPRPIf all indicators Ri = 2 or the cumulative number of indicators Ri = 1 is 1, a Level 2 (yellow) alarm is triggered, indicating that the production process is approaching the compliance boundary; if all indicators Ri ≤ 1 and the number of indicators with Ri = 1 is ≤ 2, then the production batch is in a normal state and within the acceptable range.

[0183] Step S6: Calculation of Batch Cumulative Risk Value and Judgment of Risk Trend

[0184] Calculate the batch cumulative risk value When multiple production batches are detected at the same alarm level (e.g., all triggering a yellow alert), the Q values ​​of each batch are compared to guide managers to prioritize intervention for batches with higher Q values. When the Q value shows an upward trend for three consecutive batches, the platform issues a "Risk Breakthrough Possibility Increased" alert.

[0185] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the above embodiments do not limit the scope of protection of the present invention in any way, and all technical solutions obtained by equivalent substitution or other means fall within the scope of protection of the present invention. Parts not covered in this invention are the same as or can be implemented using existing technology.

Claims

1. A critical risk monitoring system for the entire sausage production process based on wireless sensor networks and a weighted aggregation model, characterized in that, The system includes: a wireless data acquisition module and a remote platform. The remote platform includes a weight configuration module, a deviation scoring module, an alarm module, and a risk assessment module. Specifically: The wireless acquisition module is used to collect seven key risk monitoring indicators for sausage production, including: a raw material temperature wireless acquisition device for collecting raw material temperature information; a ground meat sample temperature wireless acquisition device for collecting ground meat sample temperature information; a chopped meat sample temperature wireless acquisition device for collecting meat sample temperature information at each stage of chopping; an auxiliary material addition amount wireless acquisition device installed on the weighing device for collecting auxiliary material addition amount information; a steaming environment temperature wireless acquisition device for collecting steaming temperature and time information; a metal detection wireless acquisition device installed on the metal detector at the end of the production line for collecting metal detection information; and a workshop environment temperature and humidity wireless acquisition device installed in the central area of ​​the main operating space of the workshop for collecting workshop environment temperature and humidity information. Each wireless acquisition device includes a microcontroller unit, a data acquisition unit, a wireless communication unit, and a power management unit. The wireless acquisition device reads the data collected by the data acquisition unit according to a preset sampling frequency or button operation and uploads it to the remote platform through the wireless communication unit. The weight configuration module is used to store the weight coefficients Wi corresponding to the seven key risk monitoring indicators. The deviation scoring module assigns a risk score Ri to each key risk monitoring indicator based on the threshold values ​​of the normal zone, warning zone, and danger zone corresponding to that indicator, where Ri represents the risk score of the i-th key risk monitoring indicator. The alarm module generates alarms based on preset R-value combination logic rules; The risk assessment module calculates the batch cumulative risk value based on the risk score Ri and its weight Wi corresponding to the real-time monitoring data of each key risk monitoring indicator. .

2. The system according to claim 1, characterized in that, The weighting coefficients Wi are obtained using the De Fel method and the Analytic Hierarchy Process (APH).

3. The system according to claim 2, characterized in that, The specific method for obtaining the weighting coefficient Wi is as follows: experts in the fields of meat processing and food safety quality control are invited. First, the importance of the seven key risk monitoring indicators in controlling food safety risks during sausage production was scored. Then, the experts compared the relative importance of each indicator pairwise, constructed a judgment matrix, calculated the eigenvectors, and performed a consistency test to obtain the initial weights. Finally, after multiple rounds of expert feedback and adjustments, the weight values ​​for each key risk monitoring indicator were determined.

4. The system according to claim 1, characterized in that, In the deviation scoring module, for each risk monitoring indicator, the mean μ and standard deviation σ of its historical data are calculated based on statistical process control, and each interval is set in conjunction with the key limit CL, specifically as follows: The normal range is: (LWL, UWL) The warning interval is: (LCL, LWL]∪[UWL, UCL) The danger zone is defined as: (-∞, LCL] ∪ [UCL, +∞) The values ​​of the first lower threshold LWL, the second lower threshold LCL, the first upper threshold UWL, and the second upper threshold UCL are selected based on the following specific circumstances: If μ + 2σ ≤ CL H ≤ μ + 3σ,UWL = μ + 2σ,UCL = CL H ; If CL H ≤ μ + 2σ,UWL = CL H -∣10% CL H ∣,UCL = CL H ; If CL H ≥ μ + 3σ or no definite CL H When, UWL =μ + 2σ, UCL =μ + 3σ; If - 2σ ≥ CL L ≥ μ - 3σ,LWL =μ - 2σ,LCL= CL L If CL L ≥ μ - 2σ,LWL =CL L +∣10% CL L ∣,LCL = CL L ; If CL L ≤ μ - 3σ or no definite CL L When, LWL =μ - 2σ, LCL =μ - 3σ; Among them, CL H CL represents the critical upper limit value. L This indicates the critical lower limit value.

5. The system according to claim 1, characterized in that, The normal zone is assigned a score of Ri=0, the warning zone is assigned a score of Ri=1, and the danger zone is assigned a score of Ri=2.

6. The system according to claim 5, characterized in that, The R-value combinational logic rules in the alarm module include at least three levels: If any CCP risk monitoring indicator shows Ri=2, a Level 1 alarm will be triggered; Under the premise of not being a Level 1 warning, a Level 2 warning is triggered if any OPRP risk monitoring indicator Ri=2 or if the cumulative three indicators Ri≥1. If all indicators Ri≤1 and the cumulative number of Ri=1 is not greater than 2, then the production batch process is in a normal state. Among them, the amount of auxiliary materials added, cooking temperature and time, metal detection, and workshop environmental temperature and humidity belong to CCP; the raw material temperature, minced meat sample temperature, and meat sample temperature at each stage of chopping belong to OPRP.

7. The system according to claim 1, characterized in that, The data acquisition unit of the wireless data collector is configured with different front-end devices according to the type of monitored object and the requirements of actual operation: The raw material feeding temperature wireless acquisition device uses a high-precision needle-type PT100 temperature sensor. Both the ground meat sample temperature wireless acquisition device and the chopped meat sample temperature wireless acquisition device use the SA10AC infrared temperature sensor, and the output signal is RS485. The auxiliary material addition acquisition device is equipped with a high-precision communication electronic scale, and the output signal is RS232. The wireless temperature acquisition device for the steaming environment uses a KS-SHT5KTP patch temperature sensor, and the output signal is RS485. The wireless metal detector is compatible with the HL4008 intelligent metal detector, with detection sensitivities of Fe≤Φ1.5 mm; NoFe≤Φ2.0 mm; SuS≤Φ2.5 mm. The workshop environment temperature and humidity acquisition device uses the KS-SHTE1KT temperature and humidity sensor, and the output signal is RS485.

8. The system according to claim 1, characterized in that, The power management unit of the wireless data collector adopts two power supply schemes based on the specific characteristics of the monitored object: The raw material temperature wireless collector, the minced meat sample temperature wireless collector, and the chopped meat sample temperature wireless collector are powered by rechargeable lithium batteries. The wireless data acquisition devices for auxiliary material addition, metal detection, cooking environment temperature, and workshop environment temperature and humidity are all powered by a DC12V 1A adapter.

9. The system according to claim 1, characterized in that, The risk assessment module issues a risk warning if the Q value shows an upward trend for three consecutive batches. The higher the Q value, the further the production system deviates from the steady state, and the higher the degree of high risk tendency.

10. A method for monitoring key risks throughout the sausage production process based on wireless sensor networks and a weighted aggregation model, characterized in that, The specific steps are as follows: Step S1: Multi-source data acquisition. Real-time data collection of seven key risk monitoring indicators for sausage production is conducted via a wireless acquisition module. Step S2: Determine the weight coefficient Wi corresponding to each key risk monitoring indicator, and ensure that the consistency ratio CR of the weight judgment matrix is ​​less than 0.1; Step S3: Dual Threshold Setting. Combining the statistical process control (SPC) results of historical production data with the critical limit CL, set the threshold ranges for the normal zone, warning zone, and danger zone for each critical risk monitoring indicator; Step S4: Deviation Scoring. Based on the interval in which the real-time monitoring data of each indicator falls, deviation scores are assigned. Step S5: Tiered Alarms. Based on preset R-value combination logic rules, trigger the corresponding level of production alarms; Step S6: Calculation of batch cumulative risk Q-value and judgment of risk trend. Based on the scoring results Ri and corresponding weights Wi of each indicator, the risk is calculated using the formula... Calculate the batch cumulative risk value Q, and judge the production line risk status based on the trend of Q value changes.

Images