Pet food production line monitoring system based on Internet of Things and batching device of pet food production line monitoring system

Abstract

The invention discloses a pet food production line monitoring system based on Internet of Things and a batching device thereof, and relates to the technical field of pet food production, and the pet food production line monitoring system comprises a data acquisition unit, a data transmission unit, a data management unit, a monitoring management unit and a quality management unit. According to the invention, the raw materials on the production line are identified through the raw material identification module, the characteristics and generation conditions of the raw materials are respectively collected through the characteristic sensing module and the condition sensing module, the batching optimization model is constructed through the model construction module, and the processing of the raw materials can be deeply monitored through the deep monitoring module. The analysis and optimization module is used for combining real-time data of raw material characteristics and production conditions with a pre-established ingredient optimization model, an optimal ingredient proportioning scheme is calculated, the proportion of raw materials is optimized and adjusted, then dynamic adjustment is performed through the dynamic adjustment module, and the ingredient accuracy and the proportioning effect are guaranteed.

Description

Technical Field

[0001] The present invention relates to the technical field of pet food production, and in particular to a pet food production line monitoring system and a batching device thereof based on the Internet of Things. Background Art

[0002] Pet food is food specially provided for pets to meet their nutritional needs. It mainly includes cat food, dog food, rabbit food, bird food, etc. Different types of pet food are designed according to the physiological characteristics and nutritional needs of different pets.

[0003] For example, Chinese patent CN115857461B discloses an online monitoring method and system for the production of piglet premix feed. This method generates a fault tracing operation array by combining feed production line observation data. Then, fault tracing operations in the fault tracing operation array are performed on the feed production line observation data to generate a material mixing fault node for the production monitoring partition in the feed production line observation data. The target material mixing fault node is parsed and generated by combining the feed production line observation data, the fault location field of the material mixing fault node, the fault location field of multiple fault node relationship networks, and the fault tracing operation array. Based on this, the fault point in the production monitoring partition is repaired.

[0004] In the above patent, although the problem of not considering the connectivity relationship of fault characteristics and affecting the accuracy of fault point repair is solved by repairing the fault points in the production monitoring partitions, the material ratio cannot be optimized and adjusted according to different raw material characteristics and production conditions when the materials are automatically mixed, which affects the accuracy and ratio effect of the ingredients. Summary of the Invention

[0005] The purpose of the present invention is to provide a pet food production line monitoring system and its batching device based on the Internet of Things, so as to solve the problem proposed in the above background technology that when materials are automatically mixed, the material ratio cannot be optimized and adjusted according to different raw material characteristics and production conditions, which affects the accuracy and batching effect of the ingredients.

[0006] To achieve the above-mentioned object, the present invention provides the following technical solutions: a pet food production line monitoring system based on the Internet of Things, comprising a data acquisition unit, a data transmission unit, a data management unit, a monitoring management unit and a quality management unit;

[0007] The data acquisition unit is used to collect the characteristic data of the current raw materials and the production condition data;

[0008] The data acquisition unit includes a raw material identification module, a characteristic perception module, and a condition perception module. The raw material identification module uses visual inspection equipment to identify the type of current raw materials; the characteristic perception module obtains data on the moisture, particle size, and main nutrients of the raw materials; and the condition perception module obtains temperature and humidity in the workshop and warehouse, as well as equipment operation data.

[0009] The monitoring management unit monitors the overall operation of the system and deeply integrates the system with production equipment;

[0010] The monitoring and management unit includes a monitoring interaction module, a deep monitoring module, an early warning and emergency module, an analysis and optimization module, and a dynamic adjustment module; the monitoring interaction module is used to link the control screen of the control terminal to view the characteristic data of the raw materials; the deep monitoring module dynamically tracks the raw material characteristics and production condition data to obtain data changes; when the early warning and emergency module detects that the raw material characteristics in the characteristic perception module exceed the preset range, it sends an early warning message to trigger the system's preset emergency mechanism; the analysis and optimization module calculates the optimal ingredient ratio scheme through real-time data combined with a pre-established ingredient optimization model; the dynamic adjustment module sends control instructions to the ingredient equipment according to the optimized ratio scheme in the analysis and optimization module to adjust the ingredient quantity of each raw material.

[0011] Preferably, the characteristic perception module obtains the moisture, particle size and main nutrient component data of the raw materials through moisture sensors, laser diffraction and machine vision combined technology and online near-infrared spectrometer. In the condition perception module, the equipment operation status is monitored through the motor intelligent monitoring module and the pressure flow detection system.

[0012] Preferably, when calculating the optimal ingredient ratio scheme in the analysis and optimization module, the following steps are also included:

[0013] A1. Data acquisition and preprocessing: The system continuously acquires real-time raw material characteristic data and inputs the data into the edge computing gateway for preliminary processing, removing outliers and interference information, and processing it into a format suitable for model input;

[0014] A2. Model input preparation: Organize and combine the pre-processed raw material characteristics and production condition data according to the format and dimensions required by the ingredient optimization model;

[0015] A3. Model Calling and Calculation: Calling the pre-established ingredient optimization model, using product quality standards as the objective function, and substituting the input real-time data as variables into the model's calculation logic to begin calculations and decision-making;

[0016] A4. Iterative Optimization Process: This process approaches the optimal solution through continuous iterative optimization. In each iteration, the model simulates the production results based on the current batching solution, obtains a feedback reward value, and adjusts the model's internal parameters based on the reward value.

[0017] A5. Plan evaluation and execution: After multiple rounds of iterative optimization, the model outputs a batching plan that optimizes the objective function. The system then conducts a final evaluation of the plan to check whether it meets all production constraints. If so, it is the optimal batching plan. The system adjusts the amount of each raw material according to the instructions in the plan.

[0018] Preferably, the data transmission unit includes a mode selection module, an edge computing module and a data upload module, and the mode selection module selects different transmission modes according to different communication areas;

[0019] The edge computing module performs preliminary processing, cleaning, and protocol conversion on the collected data through the built-in processor and edge computing software of the edge computing gateway;

[0020] The data upload module uploads the key data processed in the edge computing module to the cloud platform or local server.

[0021] Preferably, the data management unit includes a data storage module, a model building module and a server collaboration module, the data storage module is used to store raw material characteristic data, production condition data and historical ingredient data;

[0022] The model building module builds an ingredient optimization model based on the reinforcement learning algorithm;

[0023] The server collaboration module uses the local server as a backup and supplement to the cloud platform, synchronizing key data of the cloud platform in real time.

[0024] Preferably, in the model building module, when building the ingredient optimization model, the following contents are also included:

[0025] B1. Determine the state space: Real-time raw material characteristics, production conditions, and current batching progress information together constitute the model's state space;

[0026] B2. Define the action space and design the reward function: The action space refers to the decisions the model can take. In ingredient optimization, this is manifested as adjustments to the ratios of various raw materials. When the product quality of the ingredient solution generated by the model meets or exceeds the preset standard, a higher reward is given; if the product quality does not meet the standard, a lower reward or even a penalty is given.

[0027] B3. Select reinforcement learning algorithms: For discrete action spaces, use the Deep Q Network (DQN) and its extended algorithms; for continuous action spaces, use Deep Deterministic Policy Gradient (DDPG) and Proximal Policy Optimization (PPO);

[0028] B4. Model training and application: Use historical production data and real-time collected data to train the model in a simulated environment or actual production scenario. Use an independent test data set to verify the trained model and check whether the ingredient solution generated by the model meets the product quality standards and conduct a trial.

[0029] Preferably, the quality management unit includes a quality inspection module, a traceability and reinforcement module, and an analysis and improvement module. The quality inspection module is used to detect multiple data of the product and determine whether the product quality is qualified based on the inspection results;

[0030] The traceability and enhancement module is used to trace detailed information of raw materials used through product batch numbers;

[0031] The analysis and improvement module uses correlation analysis and cause-and-effect analysis methods to identify root causes and propose targeted improvement suggestions for quality issues.

[0032] The batching device of the pet food production line monitoring system based on the Internet of Things includes a support frame, a detection mechanism is arranged on the top of the support frame, the detection mechanism includes a hydraulic push rod, a guide frame and a detection disk, the bottom of the guide frame is fixedly connected to the support frame, the top of the guide frame is fixedly connected to the hydraulic push rod, one end of the hydraulic push rod passes through the guide frame and is fixedly connected to the detection disk, the outer ring surface of the detection disk is fixedly connected to a visual detection component, a moisture sensor component is arranged on one side of the visual detection component, the top of the moisture sensor component is fixedly connected to the outer ring surface of the detection disk, the bottom of the detection disk is fixedly connected to an analyzer probe, and the analyzer probe and the analyzer body are electrically connected.

[0033] Preferably, a support ring is fixedly connected to one side of the guide frame, the inner ring surface of the support ring is rotatably connected to the feeding tray, the outer ring surface of the feeding tray is fixedly connected to bevel gear four, one side of bevel gear four is meshed with bevel gear one, the middle part of bevel gear one is fixedly connected to motor one, motor one is fixedly connected to the support ring, and the outer ring surface of the feeding tray is fixedly connected to motor two, one end of the output shaft of motor two is fixedly connected to bevel gear two, one side of bevel gear two is meshed with bevel gear three, the shaft of bevel gear three passes through the side wall of the feeding tray and is fixedly connected to a blanking plate, and the blanking plate is rotatably connected to the inner wall of the feeding tray.

[0034] Preferably, a divider is inserted through the inside of the support frame, the outer wall of the divider is fixedly connected to the support frame, one side of the divider is fixedly connected to a control box, the control box is electrically connected to a hydraulic push rod 2, one end of the hydraulic push rod 2 is fixedly connected to the divider, the other end of the hydraulic push rod 2 is fixedly connected to an L-shaped connecting piece, one side of the connecting piece is fixedly connected to a baffle, one end of the baffle passes through the side wall of the divider and is slidably connected to the divider, one side of the support frame is fixedly connected to a recovery bucket, one side of the recovery bucket is fixedly connected to a collecting bucket, the bottom of the collecting bucket is fixedly connected to a spiral feeder, the outer ring surface of the spiral feeder is fixedly connected to a heating element, the bottom of the heating element is fixedly connected to a bottom plate, and one side of the bottom plate is fixedly connected to the support frame.

[0035] Compared with the prior art, the present invention has the following beneficial effects:

[0036] 1. In the present invention, the raw materials on the production line are identified by the raw material identification module, the characteristics and production conditions of the raw materials are collected by the characteristic perception module and the condition perception module respectively, the ingredient optimization model is constructed by the model construction module, and the processing of the raw materials can be deeply monitored by the in-depth monitoring module. When the raw materials are proportioned, the analysis and optimization module is used to combine the real-time data of the raw material characteristics and production conditions with the pre-established ingredient optimization model to calculate the optimal ingredient ratio scheme, optimize and adjust the raw material ratio, and then dynamically adjust it through the dynamic adjustment module to ensure the accuracy and proportioning effect of the ingredients;

[0037] 2. In the present invention, the monitoring interaction module can interact with the system's monitoring, and the early warning emergency module can timely detect abnormal conditions of data, sensors or equipment and promptly alarm. After the alarm, the corresponding emergency mechanism of the system can be triggered to ensure the stable operation of the system and equipment. The quality of the product can be tested through the quality inspection module, and multiple parameters of the product after production can be obtained in time. When there are abnormalities in the data, the traceability and reinforcement module can be used to trace and reinforce it, and the quality problems can be analyzed and improved through the analysis and improvement module, thereby ensuring the product quality of pet food production. BRIEF DESCRIPTION OF THE DRAWINGS

[0038] Figure 1 This is a schematic diagram of the three-dimensional structure of the pet food production line monitoring system based on the Internet of Things of the present invention;

[0039] Figure 2 This is a flow chart of the pet food production line monitoring system based on the Internet of Things of the present invention;

[0040] Figure 3 This is a schematic diagram of the installation structure of the spiral feeder of the batching device of the pet food production line monitoring system based on the Internet of Things of the present invention;

[0041] Figure 4 This is a schematic diagram of the control box structure installation of the batching device of the pet food production line monitoring system based on the Internet of Things of the present invention;

[0042] Figure 5 This is a schematic diagram of the installation of the baffle structure of the batching device of the pet food production line monitoring system based on the Internet of Things of the present invention;

[0043] Figure 6 This is a schematic diagram of the installation structure of the material tray of the batching device of the pet food production line monitoring system based on the Internet of Things of the present invention;

[0044] Figure 7This is a schematic diagram of the installation structure of the blanking plate of the batching device of the pet food production line monitoring system based on the Internet of Things of the present invention.

[0045] In the picture:

[0046] 1. Data acquisition unit; 11. Raw material identification module; 12. Characteristic perception module; 13. Condition perception module; 2. Data transmission unit; 21. Mode selection module; 22. Edge computing module; 23. Data upload module; 3. Data management unit; 31. Data storage module; 32. Model building module; 33. Server collaboration module; 4. Monitoring management unit; 41. Monitoring interaction module; 42. In-depth monitoring module; 43. Early warning and emergency module; 44. Analysis and optimization module; 45. Dynamic adjustment module; 5. Quality management unit; 51. Quality inspection module; 52. Traceability and reinforcement module; 53. Analysis and improvement module

[0047] 6. Detection mechanism; 61. Hydraulic push rod 1; 62. Guide frame; 63. Motor 1; 64. Bevel gear 1; 65. Bevel gear 2; 651. Bevel gear 3; 652. Unloading plate; 66. Motor 2; 67. Bevel gear 4; 68. Support ring; 681. Material holding tray; 69. Detection tray; 691. Visual detection component; 692. Analyzer probe; 693. Moisture sensor component; 7. Support frame; 71. Collection bucket; 72. Recovery bucket; 73. Distributor; 731. Hydraulic push rod 2; 732. Connector; 733. Baffle; 74. Control box; 8. Screw loader; 81. Heating element; 9. Analyzer body. DETAILED DESCRIPTION

[0048] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the implementation regulations described are only part of the embodiments of the present invention, not all of the embodiments. All other embodiments obtained by ordinary technicians in this field based on the embodiments of the present invention without making any creative efforts shall fall within the scope of protection of the present invention.

[0049] Example 1: Reference Figure 1-Figure 2 As shown: A pet food production line monitoring system based on the Internet of Things includes a data acquisition unit 1, a data transmission unit 2, a data management unit 3, a monitoring management unit 4 and a quality management unit 5; the data acquisition unit 1 is used to collect the characteristic data and production condition data of the current raw materials; the data acquisition unit 1 includes a raw material identification module 11, a characteristic perception module 12 and a condition perception module 13.

[0050] The raw material identification module 11 identifies the type of the current raw material through visual detection equipment.

[0051] The characteristic perception module 12 obtains the moisture, particle size and main nutrient component data of the raw materials through moisture sensors, laser particle size analyzers combined with machine vision technology and online near-infrared spectrometers. It deploys multiple types of moisture sensors in raw material storage warehouses, transportation pipelines and key processing nodes, establishes a moisture sensor matrix, accurately measures the moisture content, sets dynamic thresholds for the moisture content of different raw materials, and immediately issues an alarm once the range is exceeded. It uses a combination of laser diffraction and machine vision technology to monitor the particle size of solid raw materials in real time and analyze the size and shape distribution of the particles. The online near-infrared spectrometer detects the raw materials. When it detects that the protein content of the bean raw materials in the raw materials is higher than the standard value, the system automatically marks and starts the ratio optimization process to ensure the nutritional balance of the final product.

[0052] The condition perception module 13 densely deploys temperature and humidity sensors, air quality sensors in production workshops, raw material warehouses and key processing areas, and installs intelligent monitoring modules on motors such as mixers and conveying equipment, integrating current, voltage, speed and vibration sensors. The module transmits data to the edge computing gateway through Bluetooth low-power technology, and analyzes the motor operation status in real time. When the motor current rises abnormally, it combines the vibration data to determine whether the equipment load is too large or whether there is a mechanical failure. If there is an abnormality in the equipment, the system immediately issues an early warning. Pressure sensors and flow sensors are installed in key locations such as the ingredient pipeline and reactor to accurately control the flow and ensure the accuracy of the ingredient quantity. Through the joint analysis of pressure and flow data, the flow status of the material can be grasped in real time.

[0053] The data transmission unit 2 includes a mode selection module 21, an edge computing module 22, and a data upload module 23. The mode selection module 21 selects different transmission modes according to different communication areas. In the local area of ​​the equipment, BLE technology is used to transmit data. Inside the production workshop, a high-speed and stable Wi-Fi6 network is built to transmit data. For areas that are far away and have difficult signal coverage, such as raw material warehouses and remote corners of the factory, LoRaWAN networks are deployed to achieve data transmission;

[0054] The edge computing module 22 performs preliminary processing, cleaning, and protocol conversion on various types of collected data through the edge computing gateway. The gateway has a built-in ARM architecture processor and OpenEdge to perform real-time data analysis, perform anomaly detection on device operation data, and filter invalid data;

[0055] The data upload module 23 uploads the key data processed in the edge computing module 22 to the cloud platform or local server.

[0056] The data management unit 3 includes a data storage module 31, a model building module 32 and a server collaboration module 33. The data storage module 31 adopts a distributed cloud database, combining HBase and Cassandra to store massive raw material characteristic data, production condition data and historical ingredient data.

[0057] The model building module 32 uses deep neural networks to extract and analyze the characteristics of raw materials and production conditions data, predicts their impact on product quality, and builds an ingredient optimization model based on the reinforcement learning algorithm. It takes product quality standards as the objective function and combines real-time raw material characteristics and production conditions data. Through continuous iterative optimization, it automatically generates the optimal ingredient ratio plan. When the raw material characteristics or production conditions change, the model quickly adjusts the ratio plan to ensure that product quality always meets the standards.

[0058] When building an ingredient optimization model, the following are specifically included:

[0059] 1. Determine the state space: Real-time raw material characteristics such as moisture content, particle size, nutritional content, production environment temperature and humidity, equipment operating parameters, and current batching progress information are combined to form the model's state space;

[0060] 2. Define the action space and design the reward function: The action space refers to the decisions the model can take. In ingredient optimization, this is manifested as adjustments to the ratios of various raw materials. When the product quality of the ingredient solution generated by the model meets or exceeds the preset standard, a higher reward is given; if the product quality does not meet the standard, a lower reward or even a penalty is given.

[0061] 3. Select a reinforcement learning algorithm: For discrete action spaces, the Deep Q-Network (DQN) and its extended algorithms can be used. For continuous action spaces, deep deterministic policy gradient (DDPG) and proximal policy optimization (PPO) algorithms can be used. The DQN algorithm approximates the Q-value function by constructing a deep neural network. The Q-value represents the expected reward for taking a certain action in a certain state. The network input is the characteristics of the state space, and the output is the Q-value corresponding to each action. During training, the model continuously interacts with the environment and updates the network parameters based on reward feedback to improve its ability to select the optimal action.

[0062] 4. Model training and optimization: Use historical production data and real-time collected data to train the model in a simulated environment or actual production scenarios. The model selects actions based on the current state, observes the new state and the reward obtained, and continuously updates the Q-value function. Use an independent test data set to verify the trained model and check whether the ingredient solution generated by the model meets product quality standards. Conduct a small-scale trial in actual production, collect feedback data from actual production, and further fine-tune the model.

[0063] The server collaboration module 33 uses the local server as a backup and supplement to the cloud platform, synchronizing key data of the cloud platform in real time. When a network failure or cloud service anomaly occurs, the local server immediately takes over the data storage and processing tasks to ensure the continuity of production monitoring.

[0064] The monitoring management unit 4 monitors the overall operation of the system and deeply integrates the system with the production equipment. The monitoring management unit 4 includes a monitoring interaction module 41, a deep monitoring module 42, an early warning and emergency module 43, an analysis and optimization module 44 and a dynamic adjustment module 45.

[0065] The monitoring center of the monitoring interaction module 41 is equipped with an ultra-high-definition large screen and a multi-screen linkage system, which uses a graphical interface to display raw material characteristics, real-time data of production conditions and the dynamics of the batching process. Operators can view the current moisture, particle size, ingredient content of various raw materials, as well as environmental temperature and humidity, equipment operating parameters and other information. The interface uses real-time data update technology to update data in real time.

[0066] The in-depth monitoring module 42 continuously and dynamically monitors the moisture, particle size, nutritional content and other characteristics of various raw materials, and constructs a detailed raw material characteristic file, which records the current characteristic data and historical change trends of the raw materials; for new batches of raw materials, the system automatically starts a comprehensive detection process, quickly generates a raw material characteristic report, and conducts a multi-dimensional comparison with historical data and standard parameters to evaluate the quality and applicability of raw materials from multiple angles. Through real-time and in-depth analysis of production condition data, data mining and machine learning algorithms are used to timely discover potential risks. Using historical data of equipment operating parameters, an equipment failure prediction model is established to predict possible equipment failures in advance. When potential risks are detected, an early warning signal is promptly sent to the early warning emergency module 43.

[0067] When the early warning emergency module 43 detects that the raw material characteristics exceed the preset range, the production conditions are abnormal, or the ingredient deviation may affect the product quality, it will immediately send early warning information to relevant personnel through various methods such as sound and light alarms, text messages, emails, APP push, etc. The early warning system is classified according to the risk level, divided into level one, level two and level three. The higher the level, the more urgent the situation. When the level three warning is reached, it indicates an emergency situation that seriously affects product quality or production safety. The system immediately activates the set emergency response plan to remind staff to check and repair in time.

[0068] The analysis and optimization module 44 calculates the best ingredient ratio scheme through real-time data combined with a pre-established ingredient optimization model. On the premise of meeting the nutritional requirements of the product, it optimizes the ingredient scheme in combination with the price fluctuations of the raw material market, thereby reducing production costs while ensuring product quality. When there are slight changes in the raw material characteristics or production conditions, the model can quickly adjust the ratio scheme to ensure that product quality is not affected.

[0069] When building an ingredient optimization model, the following are specifically included:

[0070] 1. Data collection and preprocessing: The system continuously acquires real-time raw material characteristic data, such as moisture content, particle size, and nutritional content, as well as production condition data, including ambient temperature and humidity, and equipment operating parameters. After inputting the collected data into the edge computing gateway, it undergoes preliminary cleaning, denoising, and format conversion to remove outliers and interference information, and processes it into a format suitable for model input to ensure data accuracy and consistency.

[0071] 2. Model input preparation: The pre-processed raw material characteristics and production condition data are organized and combined according to the format and dimensions required by the batching optimization model. If the model uses multi-dimensional vectors as input, the various data types must be integrated into vectors of the corresponding dimensions, including the various characteristic values ​​of different raw materials as well as the key parameter values ​​of the production environment and equipment status. This ensures that the input data fully covers all key factors affecting the batching and accurately reflects the current production status.

[0072] 3. Model call and calculation: Call the pre-established ingredient optimization model. This model uses product quality standards as the objective function and substitutes the input real-time data as variables into the model's calculation logic. The model is based on intelligent algorithm mechanisms such as reinforcement learning algorithms. Through the policy network or value network formed by learning historical data, it begins to perform calculations and decisions. The policy network in the deep reinforcement learning model will output a series of possible ingredient adjustment actions and their corresponding probability distributions based on the current state.

[0073] 4. Iterative optimization process: The model approaches the optimal solution through continuous iterative optimization. In each iteration, the model simulates the results of the current batching solution in production and obtains a feedback reward value. This reward value reflects the degree of conformity of the solution with product quality standards and its comprehensive impact on factors such as production efficiency and cost. The model adjusts internal parameters based on the reward value, such as the weight of the neural network model, to improve the next batching solution output, so that the reward value continues to increase, thereby continuously optimizing the solution.

[0074] 5. Scheme evaluation and execution: After multiple rounds of iterative optimization, the model outputs a batching scheme that optimizes the objective function. The system conducts a final evaluation of the scheme to check whether it meets all production constraints, such as equipment capacity restrictions, raw material inventory restrictions, etc. If so, the scheme is determined to be the optimal batching ratio scheme; if not, the model will continue to adjust, or the system will issue a prompt to inform the operator that manual intervention is required to adjust some parameters and then recalculate to determine the optimal batching ratio scheme. The system will immediately output it to the batching equipment control system. The control system adjusts the batching amount of each raw material according to the instructions in the scheme, such as controlling the opening of the batching valve, the speed of the conveying equipment, etc., and starts to execute the new batching process. During the execution process, the actual batching situation is monitored in real time through sensors to ensure consistency with the optimization scheme.

[0075] The dynamic adjustment module 45 sends control instructions to the batching equipment according to the optimized ratio scheme in the analysis and optimization module 44, and adjusts the batching quantity of each raw material. During the batching process, the actual batching situation is fed back in real time through high-precision weight sensors, flow sensors, etc., and the weighing and unloading are realized, and the ratio is compared with the preset ratio in real time. If there is a deviation, the system will immediately make dynamic adjustments, and fine-tune the batching quantity by precisely controlling the valve opening, motor speed and other parameters of the batching equipment to ensure that the batching accuracy is controlled within an extremely small error range.

[0076] After each batch of production is completed, the system comprehensively collects product quality inspection data, conducts in-depth comparisons between the actual product quality and the expected standards, uses data analysis tools to analyze the optimization effects, and further improves the ingredient optimization model based on the analysis results to form a more complete closed-loop feedback optimization mechanism, continuously improving ingredient accuracy and product quality.

[0077] The quality management unit 5 includes a quality inspection module 51, a traceability and reinforcement module 52, and an analysis and improvement module 53. The quality inspection module 51 detects the moisture content, nutritional components, microbial indicators, and physical properties of the product after mixing, drying, and disinfection, and determines whether the product quality is qualified. If the product fails, it enters the processing equipment for secondary processing. If it passes, it is directly output;

[0078] The traceability and reinforcement module 52 keeps a comprehensive and detailed record of the production process of each batch of pet food, from raw material procurement, warehousing, inspection, ingredients preparation, processing, packaging to finished product delivery. The key data of each link is completely and accurately stored by the system, and detailed information of the raw materials used can be traced back through the product batch number.

[0079] The analysis and improvement module 53 combines quality inspection data with production process data, and uses data mining algorithms and statistical process control (SPC) to deeply explore the factors affecting product quality. For frequently occurring quality problems, the system uses correlation analysis and causal analysis methods to find the root cause, propose targeted improvement suggestions for quality problems, and track the implementation effect of improvement measures.

[0080] Example 2: According to Figure 3-Figure 7 As shown, the batching device of the pet food production line monitoring system based on the Internet of Things includes a support frame 7, a detection mechanism 6 is arranged on the top of the support frame 7, the detection mechanism 6 includes a hydraulic push rod 61, a guide frame 62 and a detection disk 69, the bottom of the guide frame 62 is fixedly connected to the support frame 7, the top of the guide frame 62 is fixedly connected to the hydraulic push rod 61, one end of the hydraulic push rod 61 passes through the guide frame 62 and is fixedly connected to the detection disk 69, the outer ring surface of the detection disk 69 is fixedly connected to a visual detection component 691, a moisture sensor component 693 is arranged on one side of the visual detection component 691, the top of the moisture sensor component 693 is fixedly connected to the outer ring surface of the detection disk 69, the bottom of the detection disk 69 is fixedly connected to an analyzer probe 692, and the analyzer probe 692 is electrically connected to the analyzer body 9.

[0081] A support ring 68 is fixedly connected to one side of the guide frame 62, and the inner ring surface of the support ring 68 is rotatably connected to the feeding tray 681, and the outer ring surface of the feeding tray 681 is fixedly connected to bevel gear four 67, and one side of bevel gear four 67 is engaged with bevel gear one 64, and the middle part of bevel gear one 64 is fixedly connected to motor one 63, and motor one 63 is fixedly connected to the support ring 68, and the outer ring surface of the feeding tray 681 is fixedly connected to motor two 66, and one end of the output shaft of motor two 66 is fixedly connected to bevel gear two 65, and one side of bevel gear two 65 is engaged with bevel gear three 651, and the shaft of bevel gear three 651 passes through the side wall of the feeding tray 681 and is fixedly connected to the blanking plate 652, and the blanking plate 652 is rotatably connected to the inner wall of the feeding tray 681.

[0082] A divider 73 is inserted through the inside of the support frame 7, and the outer wall of the divider 73 is fixedly connected to the support frame 7. One side of the divider 73 is fixedly connected to a control box 74, and the control box 74 is electrically connected to a hydraulic push rod 2 731. One end of the hydraulic push rod 2 731 is fixedly connected to the divider 73, and the other end of the hydraulic push rod 2 731 is fixedly connected to an L-shaped connecting piece 732. One side of the connecting piece 732 is fixedly connected to a baffle 733, and one end of the baffle 733 penetrates the side wall of the divider 73 and is slidably connected to the divider 73. One side of the support frame 7 is fixedly connected to a recovery bucket 72, and one side of the recovery bucket 72 is fixedly connected to a collecting bucket 71. The bottom of the collecting bucket 71 is fixedly connected to a screw feeder 8, and the outer ring surface of the screw feeder 8 is fixedly connected to a heating element 81. The bottom of the heating element 81 is fixedly connected to a bottom plate, and one side of the bottom plate is fixedly connected to the support frame 7.

[0083] In this embodiment, the guide frame 62 can be used to support the hydraulic push rod 1 61, and the use height of the detection plate 69 and the moisture sensor assembly 693 can be adjusted by the hydraulic push rod 1 61. The visual detection assembly 691 facilitates visual identification of the raw materials and distinguishes each raw material. The moisture sensor assembly 693 can detect the moisture content of the raw materials. The detection plate 69 and the analyzer body 9 can assist in detecting the main nutritional component data of the raw materials and assist in obtaining the characteristics of various raw materials. The bevel gear 1 64 and the bevel gear 4 67 can drive the rotation of the material tray 681 inside the support ring 68. When the material tray 681 rotates, the support ring 68 It can provide support to ensure the stability of the material tray 681 during rotation. Bevel gear 2 65 and bevel gear 3 651 can be used to drive the flipping of the discharge plate 652, thereby realizing the discharge of raw materials after detection. The baffle 733 can control the flow of raw materials inside the distributor 73, change the discharge direction of the raw materials, and make the raw materials enter the collection bucket 71 or the recovery bucket 72. The spiral feeder 8 can accelerate the discharge of raw materials in the collection bucket 71 and send it to the subsequent processing equipment. The spiral feeder 8 can be heated by the heating element 81, thereby assisting in preheating the raw materials inside the spiral feeder 8, increasing the temperature of the raw materials, and facilitating the subsequent drying of the raw materials.

[0084] The method of use and working principle of this device: First, the raw material data of pet food production on the production line is collected through the data acquisition unit 1. When collecting data, the condition perception module 13 is used to obtain external conditions, and the characteristic perception module 12 is used to obtain the characteristics of the raw materials. When the characteristics of the raw materials are obtained, the raw materials on the production line are sampled, and the particle size of the raw materials is obtained by combining the laser particle size analyzer with machine vision technology. Then the sample is sent to the material tray 681, and the visual detection component 691 visually identifies the type of raw materials. After identification, the control terminal can obtain the identification result, and then use the external control terminal to drive the hydraulic push rod 61 to start, and the hydraulic push rod 61 drives the detection tray 69 to move downward. , insert the moisture sensor assembly 693 into the raw material, and the analyzer probe 692 is located above the raw material. At this time, the moisture sensor assembly 693 detects the moisture of the raw material, and the motor 1 63 drives the bevel gear 1 64 to rotate, and the bevel gear 1 64 drives the bevel gear 4 67 and the material tray 681 to rotate. While rotating, the analyzer probe 692 can obtain the main nutritional component data of the raw material, and the test results can be displayed on the display screen of the analyzer body 9. The motor 2 66 drives the bevel gear 2 65 to rotate, and the bevel gear 2 65 drives the bevel gear 3 651 and the blanking plate 652 to rotate, so that the blanking plate 652 no longer blocks the bottom of the inner cavity of the material tray 681. At this time, the raw materials after detection can enter the distributor 73;

[0085] The data acquired by the data acquisition unit 1 can be transmitted through the data transmission unit 2, and the processed valid data can be uploaded to the data management unit 3. The data storage module 31 is used to store the valid data, and the ingredient optimization model is constructed through the model construction module 32 to automatically produce the best ingredient ratio solution. In the event of network failure or server abnormality, the data is backed up and supplemented through the server collaboration module 33, so that the monitoring management unit 4 can quickly obtain data, monitor the entire system through the monitoring management unit 4, and implement monitoring interaction through the monitoring interaction module 41. The raw material and equipment information can be viewed in real time, and the in-depth monitoring module 42 is used for continuous dynamic monitoring to record the current characteristic data and change trends of the raw materials. If there is an abnormality in the raw material characteristics or equipment operating parameters, the early warning module 43 will issue an early warning, and the corresponding emergency treatment plan will be activated to process the raw materials or adjust the equipment;

[0086] The raw materials are judged to be qualified according to the test results. If qualified, the hydraulic push rod 2 731 on the right side drives the connecting piece 732 and the baffle 733 to move, open the channel on the right side of the distributor 73, and allow the raw materials to enter the recovery bucket 72 for collection, and then be sent to the external weighing equipment through the recovery bucket 72 for unloading after weighing. When the raw materials are unqualified, the hydraulic push rod 2 731 on the left side drives the connecting piece 732 and the baffle 733 to move, open the channel on the left side of the distributor 73, and allow the raw materials to enter the collection bucket 71, and enter the feeding port of the spiral feeder 8 through the collection bucket 71. The auger inside the spiral feeder 8 drives the raw materials to move to the discharge port, and enters the processing equipment connected to the discharge port to process the raw materials. The corresponding optimized proportioning scheme is generated by the analysis and optimization module 44 in combination with the preset proportioning optimization model, and then the proportions of various raw materials and the operating parameters of the equipment are dynamically adjusted by the dynamic adjustment module 45 to complete the proportions of various raw materials.

[0087] The quality inspection module 51 is used to inspect the quality of the raw materials after mixing and stirring, the traceability and reinforcement module 52 is used to trace and reinforce the products, and then the analysis and improvement module 53 is used to analyze the inspection data, and the existing quality problems are improved according to the analysis results.

[0088] Although the present invention has been described in detail with reference to the aforementioned embodiments, it is still possible for those skilled in the art to modify the technical solutions described in the aforementioned embodiments, or to make equivalent substitutions for some of the technical features therein. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. The IoT-based pet food production line monitoring system is characterized by: It includes a data acquisition unit (1), a data transmission unit (2), a data management unit (3), a monitoring management unit (4) and a quality management unit (5); The data acquisition unit (1) is used to collect characteristic data and production condition data of the current raw materials, and includes a raw material identification module (11), a characteristic perception module (12) and a condition perception module (13). The raw material identification module (11) identifies the type of the current raw materials through visual detection equipment; the characteristic perception module (12) obtains the moisture, particle size and main nutritional component data of the raw materials; and the condition perception module (13) obtains the temperature and humidity in the workshop and warehouse and the equipment operation data. The monitoring management unit (4) monitors the overall operation of the system and deeply integrates the system with the production equipment. It includes a monitoring interaction module (41), a deep monitoring module (42), an early warning and emergency module (43), an analysis and optimization module (44) and a dynamic adjustment module (45); the monitoring interaction module (41) is used to link the control screen of the control terminal to view the characteristic data of the raw materials; the deep monitoring module (42) dynamically tracks the raw material characteristics and production condition data to obtain data changes; when the early warning and emergency module (43) detects that the raw material characteristics in the characteristic perception module (12) exceed the preset range, it sends an early warning message to trigger the system's preset emergency mechanism; the analysis and optimization module (44) calculates the optimal ingredient ratio scheme by combining real-time data with a pre-established ingredient optimization model; the dynamic adjustment module (45) sends control instructions to the ingredient equipment according to the optimized ratio scheme in the analysis and optimization module (44) to adjust the ingredient quantity of each raw material.

2. The IoT-based pet food production line monitoring system according to claim 1, characterized in that: The characteristic perception module (12) obtains the moisture, particle size and main nutrient content data of the raw materials through a moisture sensor, a combination of laser diffraction and machine vision technology, and an online near-infrared spectrometer. In the condition perception module (13), the equipment operation status is monitored through a motor intelligent monitoring module and a pressure and flow detection system.

3. The IoT-based pet food production line monitoring system according to claim 1, characterized in that: When calculating the optimal ingredient ratio scheme in the analysis and optimization module (44), the following steps are specifically included: A1. Data acquisition and preprocessing: The system continuously acquires real-time raw material characteristic data and inputs the data into the edge computing gateway for preliminary processing, removing outliers and interference information, and processing it into a format suitable for model input; A2. Model input preparation: Organize and combine the pre-processed raw material characteristics and production condition data according to the format and dimensions required by the ingredient optimization model; A3. Model Calling and Calculation: Calling the pre-established ingredient optimization model, using product quality standards as the objective function, and substituting the input real-time data as variables into the model's calculation logic to begin calculations and decision-making; A4. Iterative Optimization Process: This process approaches the optimal solution through continuous iterative optimization. In each iteration, the model simulates the production results based on the current batching solution, obtains a feedback reward value, and adjusts the model's internal parameters based on the reward value. A5. Plan evaluation and execution: After multiple rounds of iterative optimization, the model outputs a batching plan that optimizes the objective function. The system then conducts a final evaluation of the plan to check whether it meets all production constraints. If so, it is the optimal batching plan. The system adjusts the amount of each raw material according to the instructions in the plan.

4. The IoT-based pet food production line monitoring system according to claim 1, characterized in that: The data transmission unit (2) includes a mode selection module (21), an edge computing module (22) and a data upload module (23), wherein the mode selection module (21) selects different transmission modes according to different communication areas; The edge computing module (22) performs preliminary processing, cleaning and protocol conversion on the collected data through the built-in processor and edge computing software of the edge computing gateway; The data upload module (23) uploads the key data processed in the edge computing module (22) to the cloud platform or the local server.

5. The IoT-based pet food production line monitoring system according to claim 1, characterized in that: The data management unit (3) includes a data storage module (31), a model building module (32) and a server collaboration module (33), wherein the data storage module (31) is used to store raw material characteristic data, production condition data and historical ingredient data; The model building module (32) builds an ingredient optimization model based on a reinforcement learning algorithm; The server collaboration module (33) uses the local server as a backup and supplement for the cloud platform, and synchronizes key data of the cloud platform in real time.

6. The IoT-based pet food production line monitoring system according to claim 5, characterized in that: In the model building module (32), when building the ingredient optimization model, the following contents are specifically included: B1. Determine the state space: Real-time raw material characteristics, production conditions, and current batching progress information together constitute the model's state space; B2. Define the action space and design the reward function: The action space refers to the decisions the model can take. In ingredient optimization, this is manifested as adjustments to the ratios of various raw materials. When the product quality of the ingredient solution generated by the model meets or exceeds the preset standard, a higher reward is given; if the product quality does not meet the standard, a lower reward or even a penalty is given. B3. Select reinforcement learning algorithms: For discrete action spaces, use the Deep Q Network (DQN) and its extended algorithms; for continuous action spaces, use Deep Deterministic Policy Gradient (DDPG) and Proximal Policy Optimization (PPO); B4. Model training and application: Use historical production data and real-time collected data to train the model in a simulated environment or actual production scenario. Use an independent test data set to verify the trained model and check whether the ingredient solution generated by the model meets the product quality standards and conduct a trial.

7. The IoT-based pet food production line monitoring system according to claim 1, characterized in that: The quality management unit (5) includes a quality inspection module (51), a traceability and reinforcement module (52), and an analysis and improvement module (53). The quality inspection module (51) is used to inspect multiple data of the product and determine whether the product quality is qualified based on the inspection results. The traceability and reinforcement module (52) is used to trace the detailed information of the raw materials used through the product batch number; The analysis and improvement module (53) uses correlation analysis and causal analysis methods to find the root cause and propose targeted improvement suggestions for quality problems.

8. The ingredient dispensing device of the pet food production line monitoring system based on the Internet of Things is characterized by: The pet food production line monitoring system based on the Internet of Things according to any one of claims 1 to 7 is used, comprising a support frame (7), a detection mechanism (6) is arranged on the top of the support frame (7), the detection mechanism (6) comprises a hydraulic push rod (61), a guide frame (62) and a detection disk (69), the bottom of the guide frame (62) is fixedly connected to the support frame (7), the top of the guide frame (62) is fixedly connected to the hydraulic push rod (61), one end of the hydraulic push rod (61) passes through the guide frame (62) and is fixedly connected to the detection disk (69), the outer ring surface of the detection disk (69) is fixedly connected to a visual detection component (691), a moisture sensor component (693) is arranged on one side of the visual detection component (691), the top of the moisture sensor component (693) is fixedly connected to the outer ring surface of the detection disk (69), the bottom of the detection disk (69) is fixedly connected to an analyzer probe (692), and the analyzer probe (692) is electrically connected to the analyzer body (9).

9. The batching device of the pet food production line monitoring system based on the Internet of Things according to claim 8, characterized in that: A support ring (68) is fixedly connected to one side of the guide frame (62), and the inner ring surface of the support ring (68) is rotatably connected to the material receiving tray (681). The outer ring surface of the material receiving tray (681) is fixedly connected to a bevel gear four (67), and one side of the bevel gear four (67) is meshed with a bevel gear one (64). The middle part of the bevel gear one (64) is fixedly connected to a motor one (63). The motor one (63) is fixedly connected to the support ring (68), and the outer ring surface of the material receiving tray (681) is fixedly connected to a motor two (66). One end of the output shaft of the motor two (66) is fixedly connected to a bevel gear two (65). One side of the bevel gear two (65) is meshed with a bevel gear three (651). The shaft of the bevel gear three (651) passes through the side wall of the material receiving tray (681) and is fixedly connected to a blanking plate (652). The blanking plate (652) is rotatably connected to the inner wall of the material receiving tray (681).

10. The batching device of the pet food production line monitoring system based on the Internet of Things according to claim 8, characterized in that: A material distributor (73) is inserted through the support frame (7), and the outer wall of the material distributor (73) is fixedly connected to the support frame (7). One side of the material distributor (73) is fixedly connected to a control box (74), and the control box (74) is electrically connected to a second hydraulic push rod (731). One end of the second hydraulic push rod (731) is fixedly connected to the material distributor (73), and the other end of the second hydraulic push rod (731) is fixedly connected to an L-shaped connecting piece (732). One side of the connecting piece (732) is fixedly connected to a baffle (7 33), one end of the baffle (733) penetrates the side wall of the distributor (73) and is slidably connected to the distributor (73), one side of the support frame (7) is fixedly connected to a recovery bucket (72), one side of the recovery bucket (72) is fixedly connected to a collection bucket (71), the bottom of the collection bucket (71) is fixedly connected to a spiral feeder (8), the outer ring surface of the spiral feeder (8) is fixedly connected to a heating element (81), the bottom of the heating element (81) is fixedly connected to a bottom plate, and one side of the bottom plate is fixedly connected to the support frame (7).

Images