An online monitoring and management system for quality of puffed food based on artificial intelligence

Abstract

This invention provides an AI-based online monitoring and management system for the quality of puffed foods, relating to the field of big data analysis and mining. It includes a data acquisition module, a data preprocessing module, and an AI model training module. The data acquisition module collects product testing data, equipment operation and maintenance data, and market feedback data, and stores them on a big data platform. The product testing data includes image data and physicochemical data of raw materials and finished products. This AI-based online monitoring and management system for puffed foods, through its data acquisition module, collects testing data, equipment operation and maintenance data, and market feedback data. Based on big data, it trains an AI analysis model to achieve intelligent monitoring and adjustment of puffed food quality, reducing reliance on manual labor and improving the production efficiency and quality of puffed foods.

Description

Technical Field

[0001] This invention relates to the field of big data analysis and mining, and in particular to an online monitoring and management system for the quality of puffed food based on artificial intelligence. Background Technology

[0002] The quality of puffed products is mainly affected by the quality of raw materials and the production process. Puffed foods made primarily from rice, corn, wheat, oats, and soybeans through extrusion puffing are rich in dietary fiber, nutritionally balanced, convenient to eat, and improve the texture and taste of the product, thus benefiting health. During production, various interfering factors, such as changes in the physicochemical properties of raw materials, machine wear, and environmental temperature variations, can cause deviations in process parameters, leading to changes in the quality and performance of the produced puffed foods. How to intelligently adjust process parameters online based on changes in the quality and performance of puffed foods to maintain stable quality and performance is a challenge in existing extrusion puffing production technologies.

[0003] In the production of puffed food, the setting and adjustment of the puffing process are mainly based on the experimental settings of on-site employees. For example, the degree of gelatinization is used as the evaluation criterion (response value). Data is obtained through single-factor experiments, multi-factor orthogonal experiments, etc., and then data analysis is performed. Whether the parameters are appropriate is related to whether the employees have sufficient experience and the level of their analytical skills.

[0004] Therefore, it is necessary to provide an artificial intelligence-based online monitoring and management system for the quality of puffed food to solve the above-mentioned technical problems. Summary of the Invention

[0005] This invention provides an online monitoring and management system for the quality of puffed food based on artificial intelligence, which solves the problem that in the existing puffed food production process, the setting and adjustment of the puffing process mainly rely on the experimental settings of on-site employees, which requires a high level of skill from the employees.

[0006] To solve the above-mentioned technical problems, the present invention provides an artificial intelligence-based online monitoring and management system for the quality of puffed food, comprising:

[0007] The data acquisition module is used to collect product testing data, equipment operation and maintenance data, and market feedback data, and store them in a big data platform. The product testing data includes image data and physicochemical data of raw materials and finished products.

[0008] The data preprocessing module is used to clean, standardize, and perform feature engineering preprocessing on the data acquired by the data acquisition module.

[0009] The AI ​​model training module takes the data processed by the data preprocessing module and feeds it into the corresponding model to train the AI ​​analysis model. The AI ​​analysis module includes an appearance defect recognition model, a process parameter prediction model, and an anomaly early warning model.

[0010] The appearance defect identification model determines whether a product is qualified based on its appearance. The process parameter prediction model predicts the product's quality indicators based on the product's physicochemical data and equipment operating data. The anomaly early warning model predicts impending quality defects or equipment anomalies based on real-time data.

[0011] Preferably, the appearance defect recognition model is trained using the YOLO model, the process parameter prediction model is trained using the LightGBM model, and the anomaly early warning model is trained using the LSTM model.

[0012] Preferably, the data acquisition module includes an image acquisition device, which is used to acquire image data of the product, identify unqualified products based on the acquired image data using the appearance defect recognition model, and remove unqualified products using a rejection device.

[0013] Preferably, the image acquisition device includes a mounting plate, a rotation device, a tilting device, a camera, a mounting bracket, and a cleaning device;

[0014] The rotating device is mounted on the mounting plate, the pitching device is detachably mounted on the output end of the rotating device via a positioning plate, and the camera is mounted on the output end of the pitching device.

[0015] One end of the fixing frame is detachably connected to the mounting plate via a fixing member;

[0016] The cleaning device includes an assembly tube, a fan-shaped nozzle, and a water inlet pipe. The assembly tube is installed at the other end of the mounting bracket, the fan-shaped nozzle is installed at one end of the assembly tube and is tilted above the camera, and the water inlet pipe is installed at the other end of the assembly tube.

[0017] Preferably, the rotating device includes a rotary motor and a mounting shaft. The rotary motor is mounted on the mounting plate, and the mounting shaft is mounted on the output end of the rotary motor. A square groove is provided on the top of the mounting shaft.

[0018] The pitch device includes a mounting frame, a pitch motor, and a square shaft. The square shaft is installed at the bottom of the mounting frame, the pitch motor is horizontally installed on the mounting frame, the camera is installed at the output end of the pitch motor, the square shaft is inserted into the square slot, and slots are provided on both sides of the square shaft.

[0019] The positioning disc is sleeved on the mounting shaft and connected to the mounting shaft via a sliding key. Positioning elements are provided on both sides of the inner wall of the positioning disc, and the positioning end of the positioning element passes through the through hole and engages with the slot.

[0020] Preferably, the fixing frame includes a U-shaped sleeve, a nut, two connecting arms, and multiple positioning arms. The U-shaped sleeve is fitted onto the mounting plate, the nut is installed at the bottom of the U-shaped sleeve, one end of each of the two connecting arms is symmetrically slidably mounted on the U-shaped sleeve, and the other end of each of the two connecting arms is provided with a clamp. The multiple positioning arms are mounted on the connecting arms and located on both sides of the U-shaped sleeve.

[0021] The U-shaped sleeve has a through-hole for assembly, which is aligned with the nut. The mounting plate has a mounting hole corresponding to the through-hole.

[0022] The assembly tube is installed between the two clamps.

[0023] Preferably, the fastener includes a threaded pin and a U-shaped bracket. The U-shaped bracket is sleeved on the threaded pin, and the threaded pin is threadedly connected to the nut through the assembly hole and the mounting hole. The U-shaped bracket is sleeved on the positioning arms located on both sides of the U-shaped bracket.

[0024] Preferably, the circumferential side of the positioning disk is provided with a first toothed surface, and the end cap of the threaded pin is provided with a plurality of second toothed surfaces. When the bottom end of the threaded pin abuts against the nut, the first toothed surface and the second toothed surface are aligned, and the positioning element is located below the through hole.

[0025] Preferably, the AI-based online monitoring and management system for puffed food quality further includes a wiping assembly. The wiping assembly includes a protective box, a drive component, a connecting block, an elastic telescopic rod, a wiping roller, and a sealing block. The protective box is installed on the connecting arm, the connecting block is slidably installed inside the protective box, the elastic telescopic rod connects the connecting block and the wiping roller, one end of the sealing block is elastically installed at the opening of the protective box, and the sealing block is used to block the opening of the protective box. The drive component is used to drive the connecting block to move so that the wiping roller moves out of the protective box.

[0026] Preferably, the driving component is a driving arm, one end of which passes through the side wall of the protective box through a strip hole and is connected to the connecting block. The other end of the driving component is provided with a third tooth surface. The first tooth surface forms an arc gear with the positioning disk. The third tooth surface is flush with the height of the positioning disk. The wiping assembly also includes a second elastic component. One end of the second elastic component is installed inside the protective box, and the other end contacts the connecting block.

[0027] Compared with related technologies, the artificial intelligence-based online monitoring and management system for puffed food quality provided by this invention has the following beneficial effects:

[0028] This invention provides an online monitoring and management system for the quality of puffed food based on artificial intelligence. It collects testing data, equipment operation and maintenance data, and market feedback data of puffed food by setting up a data acquisition module. Based on big data, it trains corresponding models to generate AI analysis models, including appearance defect recognition models, process parameter prediction models, and anomaly early warning models.

[0029] The appearance defect recognition model uses product appearance data collected by surveillance cameras to quickly determine whether the product is qualified, eliminating the need for manual inspection based on experience and ensuring product quality.

[0030] The process parameter prediction model predicts product quality indicators (such as moisture content, oil content, and degree of expansion) based on the collected physicochemical data of the product (such as moisture, oil, and acidity) and equipment operating data (such as the collected extruder temperature / pressure, fryer oil temperature / frying time, and conveyor belt speed).

[0031] The anomaly warning model predicts impending quality defects or equipment malfunctions based on real-time data detected by various sensors, such as excessively high oil temperature leading to scorch marks.

[0032] By training AI analysis models with big data, the quality of puffed foods can be intelligently monitored and adjusted, reducing reliance on manual labor and improving the production efficiency and quality of puffed foods. Attached Figure Description

[0033] Figure 1 A block diagram illustrating the composition of the artificial intelligence-based online monitoring and management system for puffed food quality provided by this invention;

[0034] Figure 2 This is a schematic diagram of the structure of the image acquisition device provided by the present invention;

[0035] Figure 3 for Figure 2 The diagram shows the structural design of the mounting frame and fasteners of the image acquisition device.

[0036] Figure 4 A diagram illustrating the usage state of the rotating device provided by the present invention, wherein, Figure 4 (a) is a schematic diagram showing the state in which the rotating device drives the positioning disk to rotate the threaded pin. Figure 4 (b) is a schematic diagram of the rotating device used to adjust the camera's orientation;

[0037] Figure 5 The schematic diagram of the installation principle of the fastener provided by the present invention, wherein, Figure 5 Image (a) shows a schematic diagram of the threaded pin abutting against the nut. Figure 5 (b) is a schematic diagram of the threaded connection between the threaded pin and the nut;

[0038] Figure 6 A schematic diagram of the wiping assembly provided by the present invention assembled on a fixing frame;

[0039] Figure 7 A schematic diagram of the wiping assembly provided by the present invention;

[0040] Figure 8 A schematic diagram showing the state of the wiping roller being removed from the protective box according to the present invention;

[0041] Figure 9 This is a schematic diagram showing the camera in a clean state, as provided by the present invention.

[0042] Numbering on the map:

[0043] 1. Mounting plate; 11. Mounting holes; 12. Rectangular holes;

[0044] 2. Rotating device; 21. Rotary motor; 22. Mounting shaft;

[0045] 221. Keyway; 222. Square groove; 223. Through hole;

[0046] 3. Pitch device; 31. Mounting bracket; 32. Pitch motor; 33. Square shaft;

[0047] 331. Card slot;

[0048] 4. Camera;

[0049] 5. Positioning plate; 51. Key block; 52. First tooth surface; 53. Positioning component;

[0050] 531. Block; 532. First elastic element;

[0051] 6. Fixing bracket; 61. U-shaped sleeve; 62. Connecting shaft; 63. Connecting arm; 64. Positioning arm; 65. Nut; 611. Assembly hole; 631. Clamp;

[0052] 7. Fastener; 71. Threaded pin; 72. U-shaped bracket; 73. Second tooth surface;

[0053] 8. Cleaning device; 81. Assembly pipe; 82. Fan-shaped nozzle; 83. Water inlet pipe;

[0054] 9. Wiping assembly; 91. Protective box; 92. Drive component; 93. Connecting block; 94. Elastic telescopic rod; 95. Wiping roller; 96. Second elastic component; 97. Sealing block;

[0055] 911, strip-shaped hole; 921, third tooth surface. Detailed Implementation

[0056] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of the present invention, and not all of them. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0057] This invention provides an online monitoring and management system for the quality of puffed food based on artificial intelligence.

[0058] Please refer to the following: Figure 1 In one embodiment of the present invention, the artificial intelligence-based online monitoring and management system for puffed food quality includes:

[0059] The data acquisition module is used to collect product testing data, equipment operation and maintenance data, and market feedback data, and store them in a big data platform. The product testing data includes image data and physicochemical data of raw materials and finished products.

[0060] The data preprocessing module is used to clean, standardize, and perform feature engineering preprocessing on the data acquired by the data acquisition module.

[0061] The AI ​​model training module takes the data processed by the data preprocessing module and feeds it into the corresponding model to train the AI ​​analysis model. The AI ​​analysis module includes an appearance defect recognition model, a process parameter prediction model, and an anomaly early warning model.

[0062] The appearance defect identification model determines whether a product is qualified based on its appearance. The process parameter prediction model predicts the product's quality indicators based on the product's physicochemical data and equipment operating data. The anomaly early warning model predicts impending quality defects or equipment anomalies based on real-time data.

[0063] By setting up a data acquisition module to collect testing data of puffed food, equipment operation and maintenance data, and market feedback data, and using big data to train corresponding models to generate AI analysis models, including appearance defect recognition models, process parameter prediction models, and anomaly early warning models.

[0064] The appearance defect recognition model uses product appearance data collected by surveillance cameras to quickly determine whether the product is qualified, eliminating the need for manual inspection based on experience and ensuring product quality.

[0065] For example, it can identify scorched spots, damage, foreign objects, uneven color, etc. in potato chips / shrimp chips, and output "pass or fail" and the type of defect, etc.

[0066] The process parameter prediction model predicts product quality indicators (such as moisture content, oil content, and degree of expansion) based on the collected physicochemical data of the product (such as moisture, oil content, and acidity) and equipment operating data (such as the temperature / pressure of the extruder, the oil temperature / frying time of the fryer, and the speed of the conveyor belt). For example, if the raw material moisture is too high and the extrusion pressure of the extruder is insufficient, resulting in insufficient expansion, the AI ​​analysis model will automatically adjust the extrusion pressure of the extruder and the moisture content of the raw material, and increase the drying time, etc.

[0067] The anomaly warning model predicts impending quality defects or equipment malfunctions based on real-time data detected by various sensors, such as excessively high oil temperature leading to scorch marks.

[0068] By training AI analysis models with big data, the quality of puffed foods can be intelligently monitored and adjusted, reducing reliance on manual labor and improving the production efficiency and quality of puffed foods.

[0069] The AI ​​intelligent analysis module can also reduce machine wear and product quality. For example, by changing the screw speed setting while fixing the settings of other process parameters, extruded products under different screw speed parameter conditions can be produced. Quality parameters are measured separately, and a neural network model between quality parameters and screw speed is established. During production, the screw speed parameter setting is periodically and intelligently adjusted in a timely manner according to the changes in the quality parameters of the extruded products, based on the trained neural network model. The screw speed parameter setting is continuously optimized during production to reduce or eliminate the impact of interference factors such as machine wear and changes in ambient temperature, thereby achieving the goal of controlling the stability of extruded product quality.

[0070] By setting multiple sets of extrusion puffing process parameters, puffed products under different process conditions are obtained. Quality parameters are measured and evaluated separately, resulting in measured values ​​and quality evaluation scores. The set values ​​of the extrusion puffing process parameters are calculated using the maximum value method and nonlinear weighting methods. The key feature is that the calculation of the extrusion puffing process parameter set values ​​is based on the central value of the process parameters obtained from the optimal quality evaluation, while also taking into account other process parameter values. The degree of consideration is mainly related to the distance between each process parameter value and the central value of the process parameters, and the magnitude of the measured quality parameter values. Furthermore, it can be adjusted by a nonlinear factor (extended parameter) that can change the magnitude of the nonlinear influence.

[0071] The data collection includes cross-production line and cross-factory data: production data from different factories and different equipment models (such as oil temperature-quality data of fryers in Factory A, and pressure-expansion data of extrusion puffing machines in Factory B).

[0072] Data across time dimensions: historical production data (including data from peak or off-peak seasons, different raw material batches, and different environmental conditions);

[0073] Cross-dimensional data linkage: raw material testing data (such as moisture and starch content of corn flour), equipment operation and maintenance data (such as motor speed and bearing temperature), market feedback data (such as quality issues reported by consumers), etc.

[0074] This massive amount of data is stored and managed through big data platforms (such as Hadoop and Spark), providing AI analysis models with training samples covering various abnormal scenarios. This avoids the model being applicable only to a single production line and enables it to generalize across different scenarios. By using process parameter prediction models, the optimal process parameters for new products can be quickly matched based on historical data (such as the temperature / time combination for adding baked potato chips), shortening the trial production cycle of new products.

[0075] This includes data cleaning: removing invalid data and correcting deviation values ​​(such as sudden changes in the value of the oil temperature sensor).

[0076] Data standardization: unify the data format of different devices and different factory areas (e.g., unify the resolution of different cameras).

[0077] Feature engineering: Extracting key features from massive amounts of data (such as extracting features like average color and scorch area ratio from potato chip images, and extracting features like temperature fluctuation amplitude from process parameters).

[0078] In this embodiment, the appearance defect recognition model is trained using the YOLO model, the process parameter prediction model is trained using the LightGBM model, and the anomaly early warning model is trained using the LSTM model.

[0079] Training the appearance defect recognition model includes the following steps:

[0080] S11. Model initialization: Load YOLOv8-nano pre-trained weights (pre-trained based on the COCO dataset), transfer learning (to reduce sample size requirements, especially in scenarios with few defective samples);

[0081] S12, Training Configuration (Adapted to the characteristics of puffed food images):

[0082] Batch size: Adjust according to GPU memory, preferably 16 / 32 (small batch training, improve defect recognition accuracy);

[0083] Learning rate: Initial learning rate 0.001, decayed using cosine annealing strategy (to avoid overfitting in the later stages);

[0084] Training epochs: 20-30 epochs, early stopping strategy (stop if val_loss does not decrease for 3 consecutive epochs to avoid overfitting);

[0085] Key optimizations: For cases with "few defective samples" (e.g., foreign object samples account for only 1%), use "data augmentation" (image rotation, flipping, adding noise) + "focal loss function" (to balance positive and negative samples).

[0086] S13 Training Process Monitoring: Real-time observation of two core metrics: training set loss (decreasing trend) and validation set mAP (average accuracy, target ≥ 0.95, defect identification accuracy ≥ 99%).

[0087] Training a process parameter prediction model includes the following steps:

[0088] S21. Feature Input: Screen the core features after preprocessing (such as raw material moisture, extrusion temperature, frying time, and oil temperature), and remove irrelevant features (such as workshop humidity).

[0089] S22 model training:

[0090] Initialize the LightGBM model and set the core parameters (num_leaves=31, learning_rate=0.01).

[0091] Five-fold cross-validation was used (to avoid overfitting). The training objective was: the difference between the predicted and actual values ​​(MAE) ≤ 0.5% (e.g., the prediction error of oil content ≤ 0.5%).

[0092] S23. Model Interpretation: After training, the importance ranking of output features (such as "frying time" having the highest weight on oil content) is provided as a basis for factory process optimization.

[0093] Training an anomaly warning model includes the following steps:

[0094] S31. Time-series data processing: Organize continuous equipment parameters and quality data into a "sliding window sequence" (e.g., window size = 10 seconds, sliding once every 1 second), and label it as "normal / abnormal" (abnormal = quality defects appear in the following 10 seconds).

[0095] S32, Model Training:

[0096] Initialize the LSTM model (hidden units = 64, layers = 2) to avoid slow inference due to an overly deep model;

[0097] Training objective: Early warning accuracy ≥ 98%, early warning delay ≤ 50ms (meeting the real-time intervention requirements of the production line);

[0098] Key optimizations: To address the issue of "too many false warnings," adjust the threshold (e.g., only output a warning if the prediction probability is ≥0.9), and filter out invalid warnings by combining production logic (e.g., do not issue a warning if the oil temperature fluctuates briefly but does not exceed the threshold).

[0099] The subsequent steps involve optimizing and adjusting the parameters of each model, validating each model, and finally deploying the models.

[0100] In this embodiment, the data acquisition module includes an image acquisition device, which is used to acquire image data of the product, identify unqualified products based on the acquired image data using the appearance defect recognition model, and remove unqualified products using a rejection device.

[0101] Image acquisition devices can collect image data of puffed food in real time, providing a basis for AI analysis models to determine whether the products are qualified. Unqualified products can be automatically removed by rejection devices.

[0102] The rejection device can be a robotic arm, etc.

[0103] Please see Figure 2 As an optional embodiment, the image acquisition device includes a mounting plate 1, a rotation device 2, a pitch device 3, a camera 4, a mounting bracket 6, and a cleaning device 8.

[0104] The rotating device 2 is mounted on the mounting plate 1, the pitching device 3 is detachably mounted on the output end of the rotating device 2 via the positioning plate 5, and the camera 4 is mounted on the output end of the pitching device 3;

[0105] One end of the fixing frame 6 is detachably connected to the mounting plate 1 via a fixing member 7;

[0106] The cleaning device 8 includes an assembly pipe 81, a fan-shaped nozzle 82, and a water inlet pipe 83. The assembly pipe 81 is installed at the other end of the fixing frame 6, the fan-shaped nozzle 82 is installed at one end of the assembly pipe 81 and is tilted above the camera 4, and the water inlet pipe 83 is installed at the other end of the assembly pipe 81.

[0107] Because the monitoring equipment monitors the extrusion molding of puffed food, during the high-temperature extrusion molding process, steam carries a large amount of tiny food powder (mostly in the micrometer range) to form "dust mist," which adheres to the lens of the monitoring equipment. To ensure the normal use of the monitoring equipment, it is necessary to clean it regularly.

[0108] In this implementation, by setting up the rotation device 2 and the tilt device 3, the monitoring angle of the surveillance camera 4 can be easily adjusted, and can be flexibly adjusted according to needs;

[0109] When the lens of camera 4 needs to be cleaned, the rotating device 2 rotates camera 4 horizontally toward the fan-shaped nozzle 82, and then the tilting device 3 adjusts camera 4 toward the fan-shaped nozzle 82. The fan-shaped nozzle 82 sprays cleaning fluid onto the lens of camera 4. At the same time, the tilting device 3 adjusts the tilt angle of camera 4 repeatedly within a certain range so that the cleaning fluid sprayed by the fan-shaped nozzle 82 can be evenly applied to the lens of camera 4 to ensure cleaning effect. Through automatic cleaning, camera 4 can continue to work normally.

[0110] In this embodiment, the cleaning device 8 also includes a storage tank, a water pump, pipes, a wastewater tank, a return pipe, and a receiving box (not shown). The input end of the water pump is connected to the storage tank through a pipe, and the output end of the water pump is connected to the inlet pipe 83. The storage tank stores cleaning liquid and can be added.

[0111] The receiving box is installed on the fixed frame 6 and is located below the fan-shaped nozzle 82 and the cleaning end of the camera 4 during the cleaning process. The return pipe connects the receiving box and the wastewater tank. The receiving box receives the liquid after cleaning and flows into the wastewater tank through the return pipe.

[0112] Wastewater tanks and storage tanks are installed on the ground or other locations.

[0113] Please see Figure 2 and Figure 4 As an optional embodiment, the rotating device 2 includes a rotating motor 21 and a mounting shaft 22. The rotating motor 21 is mounted on the mounting plate 1, and the mounting shaft 22 is mounted on the output end of the rotating motor 21. A square groove 222 is provided on the top of the mounting shaft 22.

[0114] The pitch device 3 includes a mounting frame 31, a pitch motor 32, and a square shaft 33. The square shaft 33 is installed at the bottom of the mounting frame 31, the pitch motor 32 is horizontally installed on the mounting frame 31, the camera 4 is installed at the output end of the pitch motor 32, the square shaft 33 is inserted into the square groove 222, and slots 331 are provided on both sides of the square shaft 33.

[0115] The positioning disk 5 is sleeved on the mounting shaft 22 and is connected to the mounting shaft 22 by a sliding key. Positioning elements 53 are provided on both sides of the inner wall of the positioning disk 5. The positioning end of the positioning element 53 passes through the through hole 223 and engages with the slot 331.

[0116] When the angle of camera 4 needs to be adjusted, the rotary motor 21 drives the square shaft 33 to rotate through the mounting shaft 22, thereby driving the camera 4 to rotate horizontally through the mounting bracket 31 and the pitch motor 32 in sequence. When adjusting the pitch, the pitch motor 32 drives the camera 4 to rotate in pitch.

[0117] By making the pitch device 3 and the rotation device 2 detachable, when installing the entire monitoring equipment, the mounting plate 1 is first installed in the corresponding position. The mounting plate 1 has a mounting part on one side, and the mounting part has fixing holes for installation with bolts, etc. The rotation device 2 and the mounting plate 1 are pre-installed as a whole. After the mounting plate 1 is installed, the camera 4 is installed together with the pitch device 3 and the rotation device 2, that is, the square shaft 33 is inserted into the square groove 222, and then the positioning plate 5 is moved up to lock in place, simplifying the installation operation.

[0118] The mounting bracket 31 is a U-shaped frame. The mounting part at the bottom of the camera 4 is rotatably mounted inside the U-shaped frame via a rotating shaft. The pitch motor 32 is mounted on one side of the U-shaped frame, and its output end passes through one side of the U-shaped frame and is connected to the mounting part of the camera 4.

[0119] As another optional approach in this embodiment, mounting plates can also be provided at the bottom of the mounting shaft 22 and the bottom of the mounting bracket 31, with fixing holes provided on both mounting plates, and subsequent assembly can be carried out using bolts and nuts.

[0120] Please see Figure 2 and Figure 3 As an optional embodiment, the fixing frame 6 includes a U-shaped sleeve 61, a nut 65, two connecting arms 63, and multiple positioning arms 64. The U-shaped sleeve 61 is fitted onto the mounting plate 1, the nut 65 is installed at the bottom of the U-shaped sleeve 61, one end of each of the two connecting arms 63 is symmetrically slidably mounted on the U-shaped sleeve 61, and the other end of each of the two connecting arms 63 is provided with a clamp 631. The multiple positioning arms 64 are mounted on the connecting arms 63 and are located on both sides of the U-shaped sleeve 61.

[0121] The U-shaped sleeve 61 has a through-hole 611, which is aligned with the nut 65. The mounting plate 1 has a mounting hole 11 corresponding to the assembly hole 611.

[0122] The assembly tube 81 is installed between the two clamps 631.

[0123] When installing the mounting bracket 6, the U-shaped sleeve 61 is fitted onto one side of the mounting plate 1, and the nut 65 and the assembly hole 611 are aligned with the mounting hole 11. Then, the assembly tube 81 in the cleaning device 8 is placed between the two clamps 631, and the two connecting arms 63 are closed so that the two clamps 631 clamp the assembly tube 81. Then, the fastener 7 is threadedly connected to the nut 65 through the assembly hole 611 and the mounting hole 11. At the same time, the U-shaped sleeve 61 and the mounting plate 1 are installed, and the positioning arm 64 is limited. The positioning arm 64 limits the connecting arm 63, that is, limits the clamp 631, so as to realize the installation of the nozzle part in the cleaning device 8 and simplify the installation operation of the nozzle part in the cleaning device 8.

[0124] In one embodiment, a protrusion is provided in the middle of the side of the U-shaped sleeve 61 away from the mounting plate 1, and a connecting shaft 62 is installed on both sides of the protrusion. An end cap is provided at the end of the connecting shaft 62, and two connecting arms 63 are sleeved on the connecting shaft 62 and located on both sides of the protrusion, so as to realize the sliding connection between the connecting arms 63 and the U-shaped sleeve 61.

[0125] In another embodiment, a slide rail can also be installed on the side of the U-shaped sleeve 61 away from the mounting plate 1, with the U-shaped sleeve 61 extending from both ends of the slide rail, and the connecting arm 63 sleeved on the slide rail to form a sliding assembly.

[0126] As another optional method in this embodiment, a connecting arm 63 can be provided without a positioning arm 64. One end of the connecting arm 63 is directly connected to the U-shaped sleeve 61, and one end of the two clamps 631 is rotatably installed on the other end of the connecting arm 63. Holes are opened on the other end of the two clamps 631. After the two clamps 631 are sleeved on the assembly tube 81, they are locked and fixed by bolts and nuts.

[0127] Please see Figure 3 As an optional embodiment, the fastener 7 includes a threaded pin 71 and a U-shaped bracket 72. The U-shaped bracket 72 is sleeved on the threaded pin 71. The threaded pin 71 is threadedly connected to the nut 65 through the assembly hole 611 and the mounting hole 11. The U-shaped bracket 72 is sleeved on the positioning arms 64 located on both sides of the U-shaped bracket 61.

[0128] The U-shaped bracket 72 and the threaded pin 71 can rotate relative to each other.

[0129] When the threaded pin 71 and the nut 65 are connected by threads, the U-shaped bracket 72 is fitted onto the positioning arms 64 located on both sides of the U-shaped sleeve 61, thereby realizing the installation of the U-shaped sleeve 61 and the mounting plate 1, and simultaneously locking and limiting the assembly tube 81, that is, simultaneously realizing the installation of the nozzle part in the cleaning device 8, simplifying the installation.

[0130] When there are two positioning arms 64, they are located on both sides of the U-shaped sleeve 61 and above the mounting plate 1. At this time, the U-shaped frame 72 is sleeved on the two positioning arms 64 and abuts against the mounting plate 1.

[0131] When there are four positioning arms 64, two U-shaped sleeves 61 are in each position, fitting snugly against the top and bottom of the mounting plate 1. Figure 2 and Figure 5 Rectangular holes 12 are provided on both sides of the mounting plate 1 and the U-shaped frame 72. The two ends of the U-shaped frame 72 are respectively connected through the two rectangular holes 12 to limit the two positioning arms 64 on each side.

[0132] As another optional method in this embodiment, abutment protrusions can be provided at both ends of the bottom of the end cap of the threaded pin 71. When the threaded pin 71 is tightened with the nut 65, the two abutment protrusions abut against the positioning arms 64 on both sides of the U-shaped sleeve 61, thereby limiting the positioning arms 64, that is, limiting the two connecting arms 63 and the clamp 631.

[0133] Please see Figure 2 and Figure 3 In a preferred embodiment, the circumferential side of the positioning disk 5 is provided with a first tooth surface 52, and the end cap of the threaded pin 71 is provided with a plurality of second tooth surfaces 73. When the bottom end of the threaded pin 71 abuts against the nut 65, the first tooth surface 52 and the second tooth surface 73 are connected, and the positioning member 53 is located below the through hole 223.

[0134] By providing a first toothed surface 52 on the circumferential side of the positioning plate 5 and a second toothed surface 73 on the end cap of the threaded pin 71, when installing the entire monitoring equipment, after installing the mounting plate 1, first install the nozzle part of the cleaning device 8. After the bottom end of the threaded pin 71 passes through the mounting hole 11 and the assembly hole 611 and abuts against the nut 65, the second toothed surface 73 engages with the first toothed surface 52. Figure 4 In the middle (a); at this time, the staff only needs to hold the two connecting arms 63 to clamp the assembly tube 81 with the two clamps 631. For the threaded connection between the threaded pin 71 and the nut 65, the mounting shaft 22 is driven by the rotary motor 21 to drive the positioning plate 5 to rotate, thereby using the action of the first tooth surface 52 and the second tooth surface 73 to drive the threaded pin 71 to rotate, thus realizing the threaded connection, thereby further simplifying the operation of the staff.

[0135] The thickness of the positioning disc 5 is greater than the thread height of the nut 65, so that after the threaded pin 71 and the nut 65 are safely threaded, the first tooth surface 52 and the second tooth surface 73 still have a meshing part.

[0136] After completing the installation of the nozzle part of the cleaning device 8 and the fixed frame 6, the camera 4 is installed together with the tilting device 3 and the rotating device 2. That is, the square shaft 33 is inserted into the large square groove 222. At this time, the slot 331 is aligned with the through hole 223. Then, the positioning plate 5 is moved up so that the positioning part 53 is aligned with the through hole 223, so that the positioning end of the positioning part 53 is fixed to the slot 331.

[0137] The rotary motor 21 can be controlled by setting a switch or a compatible APP to drive the positioning disk 5 to rotate, thereby installing the threaded pin 71. The subsequent orientation adjustment is made through the background control terminal.

[0138] Among them, such as Figure 2 and Figure 4 A keyway 221 is symmetrically provided on the mounting shaft 22, and key blocks 51 are symmetrically provided on the inner wall of the positioning disk 5. The key blocks 51 slide into the keyway 221 to form a sliding key connection.

[0139] In one embodiment, the positioning member 53 includes a first elastic member 532 and a locking block 531, wherein the inner wall of the positioning disk 5 is provided with an installation cavity by the key block 51, the locking block 531 is slidably installed in the installation cavity, and is connected to the installation cavity through the first elastic member 532;

[0140] When the locking block 531 is aligned with the through hole 223, the locking block 531 abuts against the mounting shaft 22, and the first elastic element 532 is in a compressed state. The locking block 531 abuts against the mounting shaft 22, increasing friction, so that the positioning disk 5 can be more stably held on the mounting shaft 22, and the first tooth surface 52 meshes with the second tooth surface 73.

[0141] When the locking block 531 is aligned with the through hole 223, the elastic action of the first elastic member 532 pushes the locking block 531 through the through hole 223 and inserts it into the slot 331 to achieve the limiting.

[0142] Preferably, there are two positioning elements 53, which are set to correspond to the two sliding keys.

[0143] The first elastic element 532 is an elastic component such as a spring or a leaf spring;

[0144] The positioning component 53 also includes a pull rope. One end of the pull rope is connected to the locking block 531, and the other end passes through the first elastic member 532 and then through the positioning disk 5. When it is necessary to disassemble the camera 4, the locking block 531 can be moved out of the slot 331 by pulling the pull rope, and then it can be disassembled.

[0145] In another embodiment, the positioning element 53 can also be a bolt. An installation tube is provided on the positioning plate 5, and the installation tube is sleeved on the installation shaft 22. A threaded hole is opened on the installation tube, and the bolt is threadedly connected to the installation tube. The connection is achieved by turning the bolt into the slot 331.

[0146] Please see Figure 6 and Figure 7 As a preferred embodiment, the AI-based online monitoring and management system for puffed food quality further includes a wiping component 9. The wiping component 9 includes a protective box 91, a driving component 92, a connecting block 93, an elastic telescopic rod 94, a wiping roller 95, and a sealing block 97. The protective box 91 is installed on the connecting arm 63, and the connecting block 93 is slidably installed inside the protective box 91. The elastic telescopic rod 94 connects the connecting block 93 and the wiping roller 95. One end of the sealing block 97 is elastically installed at the opening of the protective box 91, and the sealing block 97 is used to block the opening of the protective box 91. The driving component 92 is used to drive the connecting block 93 to move so that the wiping roller 95 moves out of the protective box 91.

[0147] By setting up the wiping assembly 9, when cleaning the lens of the camera 4, the driving component 92 pushes the connecting block 93, and the connecting block 93 pushes the wiping roller 95 through the elastic telescopic rod 94 to push the sealing block 97 out of the protective box 91, as shown. Figure 9 During the cleaning process, the pitch motor 32 drives the camera 4 to pitch and rotate within a certain range, making full contact with the cleaning liquid. At the same time, the lens of the camera 4 interacts with the wiping roller 95, which wipes away the dirt and other contaminants attached to the lens of the camera 4, improving the cleaning effect, especially for cameras 4 used in frying environments.

[0148] When the camera 4 tilts and rotates within a certain range and interacts with the wiping roller 95, the wiping roller 95 drives the elastic telescopic rod 94 to extend and retract adaptively to meet the distance requirements.

[0149] After cleaning is completed, the drive component 92 moves the wiping roller 95 into the protective box 91, and the sealing block 97 seals the opening of the protective box 91, thereby protecting the wiping roller 95 and preventing dust and other particles from adhering to the wiping roller 95 and affecting the cleaning effect.

[0150] The wiping roller 95 is preferably rotatably mounted on the telescopic end of the elastic telescopic rod 94.

[0151] Preferably, there are two elastic telescopic rods 94. Each elastic telescopic rod 94 includes a mounting cylinder, a spring and a connecting rod. The connecting cylinder is mounted on the connecting block 93. One end of the connecting rod extends into the mounting cylinder and is connected to the mounting cylinder by the spring. The other end passes through the mounting cylinder. The wiping roller 95 is rotatably mounted between the two connecting rods.

[0152] The wiping roller 95 includes a central roller and a cleaning cotton sleeve or rubber sleeve. The cleaning cotton sleeve or rubber sleeve is fitted and fixed on the central roller, and the central roller is rotatably mounted between two connecting rods.

[0153] In one embodiment, the sealing block 97 can be a rubber block, with the top of the rubber block connected to the top of the protective box 91, sealing the elastic potential energy of the rubber block so that the sealing block 97 can be automatically closed.

[0154] In another embodiment, the sealing block 97 is made of a rigid material. The top of the sealing block 97 is rotatably connected to the top of the protective box 91 via a pivot. A torsion spring is provided on the pivot. One end of the torsion spring is connected to the protective box 91, and the other end is connected to the sealing block 97. The sealing block 97 is automatically closed by utilizing the elastic potential energy of the torsion spring.

[0155] In one embodiment, the driving component 92 is an electric push rod, the electric push cylinder is mounted on the protective box 91, and the output end of the electric push cylinder is connected to the connecting block 93.

[0156] Please see Figure 6 and Figure 8 In another embodiment, the driving member 92 is a driving arm. One end of the driving member 92 passes through the side wall of the protective box 91 through the strip hole 911 and is connected to the connecting block 93. The other end of the driving member 92 is provided with a third tooth surface 921. The first tooth surface 52 and the positioning disk 5 form an arc gear. The third tooth surface 921 is flush with the height of the positioning disk 5. The wiping assembly 9 also includes a second elastic member 96. One end of the second elastic member 96 is installed in the protective box 91, and the other end is in contact with the connecting block 93.

[0157] When the lens of camera 4 needs to be cleaned, as the rotating device 2 drives camera 4 toward the fan-shaped nozzle 82, the positioning disk 5 rotates with the mounting shaft 22. The first tooth surface 52 on the positioning disk 5 interacts with the third tooth surface 921 on the driving component 92, driving the driving component 92 to move. The driving component 92 then drives the wiping roller 95 out of the protective box 91 through the connecting block 93 and the elastic telescopic rod 94, thereby enabling the subsequent cleaning of the lens of camera 4.

[0158] After cleaning, the rotary motor 21 rotates in the opposite direction, and the drive component 92 drives the wiping roller 95 to move into the protective box 91 through the connecting block 93 and the elastic telescopic rod 94 in sequence; no additional drive equipment such as electric push rod is required.

[0159] Thus, in this embodiment, the rotating device 2 can drive the threaded pin 71 to install the fixing frame 6 and the mounting plate 1 and the nozzle part in the cleaning device 8 in one state, and can adjust the position of the camera 4 in another state. When switching from the first state to the second state, the camera 4 can be installed, and the first tooth surface 52 on the positioning plate 5 can be aligned with the third tooth surface 921 on the driving member 92, in preparation for the subsequent driving of the wiping roller 95 to move out of the protective box 91.

[0160] In this embodiment, the positioning disk 5 and the first tooth surface 52 form an arc gear, such as... Figure 6 That is, the upper part of the positioning disk 5 is not equipped with teeth, so that when the angle of the camera 4 is adjusted within a certain angle adjustment range, the drive component 92 will not drive the wiping roller 95 to move out.

[0161] By setting a second elastic element 96, each time the first tooth surface 52 on the positioning disk 5 separates from the third tooth surface 921 on the driving element 92, the driving element 92 can be kept in the same position, so that when the positioning disk 5 rotates again, the first tooth surface 52 will act with the third tooth surface 921 again.

[0162] When the second elastic element 96 extends naturally, the wiping roller 95 abuts against the sealing block 97.

[0163] The second elastic element 96 is an elastic component such as a spring or leaf spring.

[0164] Preferably, a sealing plate (as shown in the figure) is installed on the drive member 92. The sealing plate blocks the strip hole 911. When the drive member 92 moves the wiping roller 95 out, the sealing plate moves along with it, thereby preventing dust from entering the protective box 91 through the strip hole 911 and adhering to the wiping roller 95.

[0165] In this embodiment, during the process of driving the camera 4 toward the fan-shaped nozzle 82, the rotating device 2 can first adjust the camera 4 upward to a preset elevation angle, and after it is facing the fan-shaped nozzle 82, it rotates downward so that the lens of the camera 4 interacts with the wiping roller 95. At this time, the bottom of the lens of the camera 4 is aligned with the fan-shaped nozzle 82.

[0166] The above description is merely an embodiment of the present invention and does not limit the patent scope of the present invention. Any equivalent structural or procedural transformations made based on the content of the present invention specification and drawings, or direct or indirect applications in other related technical fields, are similarly included within the patent protection scope of the present invention.

Claims

1. An online monitoring and management system for the quality of puffed food based on artificial intelligence, characterized in that, include: The data acquisition module is used to collect product testing data, equipment operation and maintenance data, and market feedback data, and store them in a big data platform. The product testing data includes image data and physicochemical data of raw materials and finished products. The data preprocessing module is used to clean, standardize, and perform feature engineering preprocessing on the data acquired by the data acquisition module. The AI ​​model training module takes the data processed by the data preprocessing module and feeds it into the corresponding model to train the AI ​​analysis model. The AI ​​analysis module includes an appearance defect recognition model, a process parameter prediction model, and an anomaly early warning model. The appearance defect identification model determines whether a product is qualified based on its appearance. The process parameter prediction model predicts the product's quality indicators based on the product's physicochemical data and equipment operating data. The anomaly early warning model predicts impending quality defects or equipment anomalies based on real-time data.

2. The artificial intelligence-based online monitoring and management system for puffed food quality according to claim 1, characterized in that, The appearance defect identification model is trained using the YOLO model, the process parameter prediction model is trained using the LightGBM model, and the anomaly early warning model is trained using the LSTM model.

3. The artificial intelligence-based online monitoring and management system for puffed food quality according to claim 1, characterized in that, The data acquisition module includes an image acquisition device, which is used to acquire image data of the product. Based on the acquired image data, the appearance defect recognition model identifies unqualified products, and the unqualified products are removed by a rejection device.

4. The artificial intelligence-based online monitoring and management system for puffed food quality according to claim 3, characterized in that, The image acquisition device includes a mounting plate, a rotating device, a tilting device, a camera, a mounting bracket, and a cleaning device; The rotating device is mounted on the mounting plate, the pitching device is detachably mounted on the output end of the rotating device via a positioning plate, and the camera is mounted on the output end of the pitching device. One end of the fixing frame is detachably connected to the mounting plate via a fixing member; The cleaning device includes an assembly tube, a fan-shaped nozzle, and a water inlet pipe. The assembly tube is installed at the other end of the mounting bracket, the fan-shaped nozzle is installed at one end of the assembly tube and is tilted above the camera, and the water inlet pipe is installed at the other end of the assembly tube.

5. The artificial intelligence-based online monitoring and management system for puffed food quality according to claim 4, characterized in that, The rotating device includes a rotary motor and a mounting shaft. The rotary motor is mounted on the mounting plate, and the mounting shaft is mounted on the output end of the rotary motor. A square groove is provided on the top of the mounting shaft. The pitch device includes a mounting frame, a pitch motor, and a square shaft. The square shaft is installed at the bottom of the mounting frame, the pitch motor is horizontally installed on the mounting frame, the camera is installed at the output end of the pitch motor, the square shaft is inserted into the square slot, and slots are provided on both sides of the square shaft. The positioning disc is sleeved on the mounting shaft and connected to the mounting shaft via a sliding key. Positioning elements are provided on both sides of the inner wall of the positioning disc, and the positioning end of the positioning element passes through the through hole and engages with the slot.

6. The artificial intelligence-based online monitoring and management system for puffed food quality according to claim 5, characterized in that, The fixing frame includes a U-shaped sleeve, a nut, two connecting arms, and multiple positioning arms. The U-shaped sleeve is fitted onto the mounting plate, the nut is installed at the bottom of the U-shaped sleeve, one end of each of the two connecting arms is symmetrically slidably mounted on the U-shaped sleeve, and the other end of each of the two connecting arms is provided with a clamp. The multiple positioning arms are mounted on the connecting arms and located on both sides of the U-shaped sleeve. The U-shaped sleeve has a through-hole for assembly, which is aligned with the nut. The mounting plate has a mounting hole corresponding to the through-hole. The assembly tube is installed between the two clamps.

7. The artificial intelligence-based online monitoring and management system for puffed food quality according to claim 6, characterized in that, The fastener includes a threaded pin and a U-shaped bracket. The U-shaped bracket is sleeved on the threaded pin. The threaded pin is threadedly connected to the nut through the assembly hole and the mounting hole. The U-shaped bracket is sleeved on the positioning arms located on both sides of the U-shaped bracket.

8. The artificial intelligence-based online monitoring and management system for puffed food quality according to claim 7, characterized in that, The circumferential side of the positioning disk is provided with a first toothed surface, and the end cap of the threaded pin is provided with a plurality of second toothed surfaces. When the bottom end of the threaded pin abuts against the nut, the first toothed surface and the second toothed surface are aligned, and the positioning element is located below the through hole.

9. The artificial intelligence-based online monitoring and management system for puffed food quality according to claim 8, characterized in that, The AI-based online monitoring and management system for puffed food quality also includes a wiping assembly. The wiping assembly includes a protective box, a drive component, a connecting block, an elastic telescopic rod, a wiping roller, and a sealing block. The protective box is installed on the connecting arm, the connecting block is slidably installed inside the protective box, the elastic telescopic rod connects the connecting block and the wiping roller, one end of the sealing block is elastically installed at the opening of the protective box, and the sealing block is used to seal the opening of the protective box. The drive component is used to drive the connecting block to move so that the wiping roller moves out of the protective box.

10. The artificial intelligence-based online monitoring and management system for puffed food quality according to claim 9, characterized in that, The driving component is a driving arm. One end of the driving component passes through the side wall of the protective box through a strip hole and is connected to the connecting block. The other end of the driving component is provided with a third tooth surface. The first tooth surface forms an arc gear with the positioning disk. The third tooth surface is flush with the height of the positioning disk. The wiping assembly also includes a second elastic element. One end of the second elastic element is installed inside the protective box, and the other end contacts the connecting block.

Images