A high-temperature molten aluminum safety transportation early warning system

Abstract

This invention belongs to the field of intelligent transportation and discloses a high-temperature molten aluminum safety transportation early warning system, including a database module, an aluminum ladle breakage early warning module, a transport vehicle status early warning module, and a driver status early warning module. The database module further includes a data acquisition module and a data classification module; the aluminum ladle breakage early warning module includes a thermal imaging preprocessing module and a threshold segmentation breakage early warning module; the transport vehicle status early warning module includes a vehicle malfunction early warning module and a driving status early warning module; and the driver status early warning module includes a multi-type data preprocessing module, a multi-task model training module, and a status early warning module. This invention establishes a complete intelligent early warning system for the safe transportation of high-temperature molten aluminum, ensuring the safety of high-temperature molten aluminum transportation from multiple perspectives, including aluminum ladle lifting equipment, transport vehicles, and drivers.

Description

Technical Field

[0001] This invention belongs to the field of intelligent transportation, specifically relating to an early warning system for the safe transportation of molten aluminum at high temperatures. Background Technology

[0002] Transporting molten aluminum in high-temperature environments presents significant safety risks during loading, unloading, and transportation. Even minor negligence can severely threaten the lives of workers, cause substantial economic losses to companies, potentially trigger environmental pollution incidents, and even impact social stability. Specifically, during the lifting and transfer of molten aluminum ladles, the refractory brick lining is subjected to thermal stress, chemical erosion, and cold stress, leading to damage to the refractory material. To ensure safe production, it is crucial to promptly identify potential threats of damage to the ladle walls and bottom during molten aluminum transfer.

[0003] The transport of hazardous materials involves transport companies, vehicles, and drivers, and is characterized by long distances and cross-administrative regions. The vast majority of road transport accidents are caused by illegal transport and violations of regulations, including employee misconduct, transport violations, evasion of supervision, and issues with the health and condition of transport equipment.

[0004] Therefore, it is urgent to introduce advanced science and technology into the supervision of molten aluminum ladles, transport vehicles, and drivers in the transportation of molten aluminum at high temperatures, so as to promptly detect potential threats in the transportation of molten aluminum and improve the supervision capabilities of molten aluminum transportation. Summary of the Invention

[0005] In view of this, the purpose of this invention is to provide a high-temperature molten aluminum safety transportation early warning system, which aims to provide intelligent monitoring and early warning from multiple perspectives, including the status of molten aluminum ladle lifting equipment, transport vehicles, and drivers, to solve the problem of poor existing molten aluminum transportation supervision capabilities and difficulty in timely detection of potential threats.

[0006] The present invention solves the above-mentioned technical problems through the following technical means:

[0007] A high-temperature molten aluminum safety transportation early warning system includes a database module, an aluminum ladle breakage early warning module, a transport vehicle status early warning module, and a driver status early warning module;

[0008] The database module is used to collect various monitoring data from the aluminum liquid transportation early warning system, and to store the data in different databases according to the data type and data processing requirements.

[0009] The database module includes a data acquisition module and a data classification module. The data acquisition module is used to collect various monitoring data of high-temperature molten aluminum transportation, and the data classification module is used to classify the collected monitoring data according to data type. Different categories of data are stored in different databases.

[0010] The aluminum liquid ladle breakage early warning module is used to detect the breakage of the aluminum liquid transport ladle lifting equipment, to predict the risk of aluminum liquid leakage in advance, and to monitor and warn of the high temperature aluminum liquid ladle lifting equipment in real time, thereby preventing the aluminum liquid from being damaged by excessively high temperature.

[0011] The aluminum molten ladle lifting damage early warning module includes a thermal imaging image preprocessing module and a threshold segmentation damage early warning module. The thermal imaging image preprocessing module is used to crop and grayscale the thermal imaging image of the high-temperature aluminum molten ladle lifting equipment. The threshold segmentation damage early warning module is used to segment the grayscale thermal imaging image, automatically detect whether there are segmented areas exceeding the threshold, and promptly warn of damage to the ladle lifting equipment for segmented areas exceeding the threshold, so as to avoid serious hazards caused by aluminum molten leakage in the ladle lifting equipment.

[0012] The transport vehicle status early warning module is used to monitor the vehicle's own status and driving status in real time, so as to promptly detect and warn of potential risks to the transport vehicle and ensure the vehicle's safety.

[0013] The transport vehicle status early warning module includes a vehicle self-fault early warning module and a driving status early warning module. The vehicle self-fault early warning module is used to monitor and warn of the transport vehicle's own status in real time, and the driving status early warning module is used to monitor and warn of the transport vehicle's operating status.

[0014] The driver status warning module is used to verify the driver's identity, monitor and warn of fatigue driving and dangerous driver behavior;

[0015] The driver status warning module includes a multi-type data preprocessing module, a multi-task model training module, and a status warning module. The multi-type data preprocessing module is used to preprocess multi-task learning data. The multi-task model training module uses a multi-task learning method to simultaneously train three computer vision detection tasks: identity verification, fatigue detection, and driver dangerous action recognition. The status warning module uses the trained multi-task learning model to verify the driver's identity in a timely manner when a driver's image is detected, and monitors fatigue driving and driver dangerous action recognition and warning in real time during the driver's driving.

[0016] Furthermore, the data classification module divides the data according to the real-time requirements and format, and then retrieves the data using different databases based on the data classification.

[0017] Furthermore, the aluminum molten ladle breakage early warning module realizes automatic detection and early warning of breakage of ladle lifting equipment through infrared thermal images and threshold-based segmentation methods.

[0018] Furthermore, the threshold-based segmentation method utilizes the pixel differences between the background and target regions in the image to determine the segmentation threshold;

[0019] Let the grayscale thermal imaging image be f(x,y), and the expression for threshold segmentation is:

[0020]

[0021] In the formula, f(x,y) is the gray value at pixel (x,y); T1, T2, ..., T n For different feature values, i.e., thresholds; P n These are the processed pixel values ​​used to segment the image; An image output according to a certain rule.

[0022] Furthermore, the vehicle's own fault early warning module monitors and issues early warnings about the transport vehicle's own status, including tire pressure, tire temperature, and vehicle maintenance status.

[0023] Among them, tire pressure and tire temperature are monitored and alerted by the installation of tire pressure and tire temperature detectors; vehicle maintenance status, that is, vehicle maintenance information, is submitted by the management personnel and the next maintenance mileage and maintenance time are set, and the data is stored on a remote server.

[0024] Furthermore, the driving status warning module monitors and issues warnings about the operating status of transport vehicles, including vehicle speeding, dangerous distance, exceeding the prescribed area, and vehicle blind spots.

[0025] Furthermore, the multi-task learning data preprocessed by the multi-class data preprocessing module includes data cropping, data augmentation, dataset partitioning, and multi-class label data setting.

[0026] Furthermore, the driver status warning module uses a multi-task learning scheme to jointly train driver identity verification, fatigue driving warning, and dangerous action recognition and warning;

[0027] The total loss function expression for the multi-task learning parameter hard-sharing scheme is:

[0028] L total =α1L1 + α2L2 + ... + α n L n

[0029] In the formula, L total The total loss function for multiple tasks; α n To assign weights to the loss functions of the corresponding sub-tasks, the loss functions L of multiple sub-tasks are trained simultaneously using a weighted approach. n ;

[0030] The objective function expression for the subtask model with multi-task parameter hard sharing is:

[0031]

[0032] In the formula, f nLet f be the objective function for the nth task; x be the training data; f g (x) represents a multi-task hard parameter layer; This is a local-specific layer for the nth task, used to adapt to the needs of each subtask in a special multitasking scenario;

[0033] Using a multi-task parameter hard-sharing scheme, the same parameters are shared between different detection tasks, which can effectively reduce model parameters and reduce model complexity. At the same time, joint multi-task training improves the model's learning ability.

[0034] Beneficial effects:

[0035] 1. This invention provides a complete intelligent early warning system for the safe transportation of molten aluminum at high temperatures. It enables comprehensive monitoring and early warning of the status of molten aluminum lifting equipment, transport vehicles, and drivers, effectively improving the regulatory capabilities for the transportation of molten aluminum at high temperatures.

[0036] 2. This invention utilizes thermal imaging technology and a threshold segmentation scheme to automatically monitor and warn of damage to the ladle lifting device, providing automatic and reliable detection and early warning capabilities for aluminum liquid ladle lifting equipment.

[0037] 3. This invention utilizes a multi-task learning algorithm to simultaneously complete driver identity verification, fatigue detection, and driver dangerous behavior warning. Through hard parameter sharing and collaborative training of related tasks, it improves model performance while reducing model complexity and ensures the accuracy of driver status warning.

[0038] Other advantages, objectives, and features of the invention will be set forth in part in the description which follows, and in part will be apparent to those skilled in the art from the following examination, or may be learned from practice of the invention. The objectives and other advantages of the invention can be realized and obtained through the following description. Attached Figure Description

[0039] Figure 1 This is a flowchart of a high-temperature molten aluminum safety transportation early warning system according to the present invention;

[0040] Figure 2 This invention provides a multi-task learning model framework for a high-temperature molten aluminum safety transportation early warning system.

[0041] Figure 3 This invention relates to a database module in a high-temperature molten aluminum safety transportation early warning system. Detailed Implementation

[0042] The following specific embodiments illustrate the implementation of the present invention. Those skilled in the art can understand the advantages and effects of the present invention from the content disclosed in this specification. It should be noted that the illustrations provided in the following embodiments are for illustrative purposes only and represent schematic diagrams, not actual pictures, and should not be construed as limiting the present invention. In order to better illustrate the embodiments of the present invention, some components in the figures may be omitted, enlarged, or reduced, and do not represent the actual product size; it is understandable for those skilled in the art that some well-known structures and their descriptions may be omitted in the figures.

[0043] In the figures of this invention, the same or similar reference numerals correspond to the same or similar components. In the description of this invention, it should be understood that if terms such as "upper," "lower," "left," "right," "front," and "rear" indicate the orientation or positional relationship based on the orientation or positional relationship shown in the figure, they are only for the convenience of describing this invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, the terms used to describe positional relationships in the figures are only for illustrative purposes and should not be construed as limiting this invention. For those skilled in the art, the specific meaning of the above-mentioned terms can be understood according to the specific circumstances.

[0044] like Figure 1 As shown, the present invention provides a high-temperature molten aluminum safe transportation early warning system, including a database module, an aluminum ladle breakage early warning module, a transport vehicle status early warning module, and a driver status early warning module;

[0045] The database module is used to collect various monitoring data from the aluminum liquid transportation early warning system, and uses different databases to store the data according to the data type and data processing requirements, thereby reducing data storage costs while meeting the requirements of efficient data access.

[0046] The database module includes a data acquisition module and a data classification module. The data acquisition module is used to collect various monitoring data of high-temperature molten aluminum transportation, and the data classification module is used to classify the collected monitoring data according to data type. Different categories of data are stored in different databases. The data classification module divides the data according to the real-time requirements and format (structured data and unstructured data), and then uses different databases to retrieve the data according to the data classification.

[0047] The aluminum molten ladle breakage early warning module is used to detect the breakage of aluminum molten transport ladle lifting equipment, to predict the risk of aluminum molten leakage in advance, and to perform real-time temperature monitoring and early warning of high-temperature aluminum molten ladle lifting equipment, thereby preventing the ladle lifting equipment from being damaged by excessively high aluminum molten temperature.

[0048] The aluminum molten ladle lifting damage early warning module includes a thermal imaging image preprocessing module and a threshold segmentation damage early warning module. The thermal imaging image preprocessing module is used to crop and grayscale the thermal imaging image of the high-temperature aluminum molten ladle lifting equipment. The threshold segmentation damage early warning module is used to segment the grayscale thermal imaging image, automatically detect whether there are segmented areas that exceed the threshold, and promptly warn of damage to the ladle lifting equipment for segmented areas that exceed the threshold, so as to avoid serious hazards caused by aluminum molten leakage in the ladle lifting equipment.

[0049] The aluminum molten ladle breakage early warning module realizes automatic detection and early warning of breakage of ladle lifting equipment through infrared thermal images and threshold-based segmentation methods;

[0050] Threshold-based segmentation methods use the pixel differences between the background and target regions in an image to determine the segmentation threshold;

[0051] Let the grayscale thermal imaging image be f(x,y), and the expression for threshold segmentation is:

[0052]

[0053] In the formula, f(x,y) is the gray value at pixel (x,y); T1, T2, ..., T n For different feature values, i.e., thresholds; P n These are the processed pixel values ​​used to segment the image; This is an image output according to a certain rule. The method divides the image into regions by determining an optimal threshold. Since the surface temperature of the bag-lifting equipment is generally within a certain range, the threshold segmentation method can automatically partition the grayscale image of the bag wall and bottom. Partitions exceeding a certain threshold are judged as damaged or potentially damaged, providing timely warnings of potential risks to the bag-lifting equipment.

[0054] The transport vehicle status early warning module is used to monitor the vehicle's own status and driving status in real time, so as to promptly detect and warn of potential risks to the transport vehicle and ensure its safety.

[0055] The transport vehicle status early warning module includes a vehicle self-fault early warning module and a driving status early warning module. The vehicle self-fault early warning module is used to monitor and warn of the transport vehicle's own status in real time, while the driving status early warning module is used to monitor and warn of the transport vehicle's operating status.

[0056] The vehicle self-fault early warning module monitors and issues early warnings regarding the transport vehicle's own status, including tire pressure, tire temperature, and vehicle maintenance status. Tire pressure and temperature are monitored and alerted via additional tire pressure and temperature detectors; any abnormal data detected is promptly notified to the driver and remote management personnel. Vehicle maintenance status, i.e., vehicle repair information, is submitted by management personnel who set the next repair mileage and repair time; this data is stored on a remote server. The driving status early warning module monitors and issues early warnings regarding the transport vehicle's operating status, including speeding, dangerous following distances, exceeding designated areas, and blind spots. Various monitors promptly detect and issue early warnings for potential vehicle hazards, ensuring transport vehicle safety. The driving status early warning module only monitors and issues early warnings while the vehicle is in motion, reducing unnecessary resource consumption.

[0057] The driver status warning module is used to verify the driver's identity and monitor and warn of fatigue driving and dangerous driver behavior;

[0058] The driver status warning module comprises a multi-class data preprocessing module, a multi-task model training module, and a status warning module. The multi-class data preprocessing module preprocesses the multi-task learning data, including data cropping, data augmentation, dataset partitioning, and multi-class labeling. The multi-task model training module uses a multi-task learning method to simultaneously train three computer vision detection tasks: identity verification, fatigue detection, and driver dangerous action recognition. The status warning module, through the trained multi-task learning model, promptly verifies the driver's identity upon detecting a driver's image and monitors for fatigue driving and identifies and warns of driver dangerous actions in real time during driving.

[0059] The driver status warning module uses a multi-task learning scheme to jointly train driver identity verification, fatigue driving warning, and dangerous action recognition and warning.

[0060] The total loss function expression for the multi-task learning parameter hard-skill solution is:

[0061] L total =α1L1 + α2L2 + ... + α n L n

[0062] In the formula, L total The total loss function for multiple tasks; α n To assign weights to the loss functions of the corresponding sub-tasks, the loss functions L of multiple sub-tasks are trained simultaneously and in a weighted manner. n ;

[0063] The objective function expression for the subtask model with multi-task parameter hard sharing is:

[0064]

[0065] In the formula, f nLet f be the objective function for the nth task; x be the training data; f g (x) represents a multi-task hard parameter layer; This is a local-specific layer for the nth task, used to adapt to the needs of each subtask in a special multitasking scenario;

[0066] Using a multi-task parameter hard-sharing scheme, the same parameters are shared between different detection tasks, which can effectively reduce model parameters and reduce model complexity. At the same time, joint multi-task training improves the model's learning ability.

[0067] As a preferred embodiment, the driver status warning module uses a multi-task learning scheme to jointly train driver identity verification, fatigue driving warning, and dangerous action recognition and warning. The multi-task model is as follows: Figure 2 As shown, multiple subtasks share a large number of model parameters through hard parameters, which reduces model complexity.

[0068] The total loss function expression for the multi-task learning parameter hard-sharing scheme is:

[0069] L total = 0.3L1 + 0.3L2 + 0.4L3

[0070] L total The total loss function for multiple tasks is defined as follows: the weights of the loss functions for identity verification, fatigue driving warning, and dangerous action recognition tasks are 0.3, 0.3, and 0.4, respectively. Multiple sub-tasks are trained simultaneously using a weighted approach.

[0071] The objective function expression for the subtask model with multi-task parameter hard sharing is:

[0072]

[0073] f n Let f be the objective function for the nth task, and x be the training data; g (x) is a multi-task hard parameter layer, using the classic MobileNet_V2 model; This is a local-specific layer for the nth task, used to adapt to the needs of each subtask in a special multitasking scenario; each subtask's local layer The model maps basic linearized network layers to corresponding task classifications. A multi-task parameter hard-sharing scheme is used, allowing different detection tasks to share the same parameters, effectively reducing model parameters and complexity. Simultaneously, joint multi-task training enhances the model's learning ability.

[0074] As a preferred embodiment, the database module categorizes the collected data according to data requirements. For example... Figure 3As shown, based on real-time requirements, real-time acquired data is stored and used in an HBase database to ensure high availability and real-time access capabilities. Examples include real-time video, real-time location data, and real-time alert information. MySQL is used to store important metadata or structured data for complex relational queries, such as license plate numbers, driver names, and vehicle maintenance information. Historical data is also periodically archived to MySQL to reduce data storage costs and leverage MySQL's ACID properties to ensure data consistency and reliability. Historical data is transferred from the HBase database to the MySQL database every four hours for storage.

[0075] In a preferred embodiment, infrared thermal image acquisition uses an infrared imager with a fixed temperature value for imaging. The temperature of the acquired images is fixed between 0-400℃, ensuring a correspondence between the temperature of each sample and the imaging pixels. The acquired infrared thermal images are cropped to 1024×1024 pixels and then converted to grayscale images with pixels between 0-255. The normal range for the infrared thermal image temperature value of the ladle lifting equipment is set to 200-300℃. Images below 200℃ are displayed as background, areas between 300-350℃ are considered general risk, and areas above 350℃ are considered high risk, providing timely warnings to prevent aluminum molten metal leakage due to damage to the ladle lifting equipment. The corresponding grayscale image pixel thresholds are T1=120, T2=190, T3=210. After data segmentation, the pixel values ​​are P0=0, P1=50, P2=100, P3=200.

[0076] In a preferred embodiment, the vehicle status warning module is connected to an onboard industrial control computer via a tire pressure monitoring system, a tire temperature detector, an onboard GPS positioning system, and a laser rangefinder. As shown in Table 1, tire pressure, tire temperature, distance to other vehicles, vehicle speed, and blind spot information are first uploaded to the onboard industrial control computer, which promptly issues voice warnings to the driver. The information is also uploaded to a remote server via a 5G communication box to save historical warning information, ensuring vehicle safety and providing historical data accessibility. For vehicle exceeding designated areas, the industrial control computer sends the vehicle's location latitude and longitude to the remote server via 5G communication to monitor whether the vehicle has exceeded the designated driving area, providing timely warnings to ensure the safety of the transport vehicle's route.

[0077] Table 1. Warning Rules of the Transport Vehicle Status Warning Module

[0078] Serial Number Alarm Types Triggering conditions Voice broadcast 1 Tire pressure abnormality warning Tire pressure below / above set value abnormal tire pressure 2 Too close following warning HWM time less than 0.6 seconds Please maintain a safe distance. 3 Excessive speed warning Vehicle speed exceeding 45km / h Please pay attention to speed. 4 Right blind spot warning Right blind spot obstruction Please keep away from vehicles. 5 Warning of deviation from the designated area Vehicle GPS positioning deviates from the designated area Please pay attention to the driving route.

[0079] As a preferred embodiment, after the vehicle is in operation, the driver status warning module verifies the driver's identity, detects the driver's fatigue state and dangerous actions, and promptly warns of potential risks. The specific steps are as follows:

[0080] (1) Data collection and preprocessing: The vehicle-mounted camera records driver videos, which are then uploaded to the Qineng Aluminum Equipment Room server database via a 5G communication box. The database uses HBase to provide flexible video storage. Data preprocessing involves cropping each frame of the video to 112×112 pixels, and then performing geometric transformations on a portion of the dataset for data augmentation. Flipping and rotation operations are used to change pixel positions and increase training samples for certain label types. For the same frame image, for three detection methods—identity verification, fatigue detection, and dangerous action recognition—the model training labels are the driver's name corresponding to the image, fatigue status label, and dangerous action category label, respectively. 80% of the dataset is divided into the training set, and 20% into the test set. The detailed pseudocode statements for the driver status warning module based on Python software are as follows:

[0081] Part 1: Data Preprocessing

[0082] ## Describing the data

[0083] df.describe(image)

[0084] ## Crop image size

[0085] import torchvision

[0086] crop_obj=torchvision.transforms.CenterCrop((112,112))

[0087] image = crop_obj(image)

[0088] ##Contrast Enhancement

[0089] import cv2

[0090] b, g, r = cv2.split(image)

[0091] clahe=cv2.createCLAHE(clipLimit=2,tileGridSize=(8,8))

[0092] image_clahe=cv2.merge([clahe.apply(b),clahe.apply(g),clahe.apply(r)])

[0093] ##Data Normalization

[0094] from torchvision import transforms as T Norm_=T.Normalize((0.485,0.456,0.406),(0.5,0.5,0.5))

[0095] image = Norm_(image_clahe)

[0096] ## Oversampling: Corresponding image processing library skimage

[0097] from skimage.transform import resampleresampled_image=resample(image,output_shape=target_shape,order=1)

[0098] ## Segmenting the Dataset (Three Detection Tasks Corresponding to Three Labels)

[0099] dftrain,dftest=train test split(dfdata,train size=0.8,random state=38)

[0100] Xtrain,Y1train,Y2train,Y3train=dftrain.drop(label_col,axis=1),dftrain[lable_col]

[0101] Data_train = cb.pool(data = Xtrain, label = Y1train, Y2train, Y3train, cat_features = cat_cols) Data_test = cb.pool(data = Xtrain, label = Y1train, Y2train, Y3train, cat_features = cat_cols) Part Two: Model Training

[0102] ## Setting parameters

[0103] lr = 0.0001

[0104] epoch = 200

[0105] ## Model Training

[0106] for epoch in range(epochs):

[0107] for i, (img, y1, y2, y3) in enumerate(Data_train):

[0108] img, y1, y2, y3 = img.to(device), y1.to(device), y2.to(device), y3.to(device)

[0109] y1_pred, y2_pred, y3_pred = model(img)

[0110] loss1 = nn.CrossEntropyLoss()

[0111] loss2 = nn.BCELoss()

[0112] [[ID=2D]]loss3 = nn.CrossEntropyLoss()

[0113] ## Multi-task Loss Function Weight Allocation

[0114] loss = 0.3 * loss1(y1_pred, y1) + 0.3 * loss2(y2_pred, y2) + 0.4 * loss3(y3_pred, y3)

[0115] loss.backward()

[0116] ## Model Evaluation

[0117] [[ID=D9]]from sklearn.metrics import accuracy_score

[0118] for img, y1, y2, y3 in enumerate(Data_test):

[0119] img, y1, y2, y3 = img.to(device), y1.to(device), y2.to(device), y3.to(device)

[0120] y1_pred, y2_pred, y3_pred = model(img)

[0121] accuracy_identity=accuracy_score(y1_pred,y1)

[0122] accuracy_fatigue=accuracy_score(y2_pred,y2)

[0123] accuracy_action=accuracy_score(y3_pred,y3)

[0124] print(f'Identity Accuracy:{accuracy_identity},Fatigue Accuracy:{accuracy_fatigue},Action Accuracy:{accuracy_action}')

[0125] ##Part Three: Status Warning Module

[0126] ## Load the training model

[0127] model = Mutitask_Model()

[0128] model.load_state_dict(torch.load('Mutitask_model.pth'))

[0129] ## Get Real-time Video

[0130] cap = cv2.VideoCapture(0)

[0131] while True:

[0132] #Read a frame of image

[0133] ret,frame = cap.read()

[0134] #Data Preprocessing

[0135] img=data_preprocessing(frame)

[0136] #Real-time feedback on driver status based on model detection results, and warnings of dangerous situations.

[0137] y1_pred,y2_pred,y3_pred=model(img)

[0138] The driver status early warning module includes early warning rules as shown in Table 2, including identity verification early warning, fatigue detection early warning (divided into yawning and eye closing warnings), and driver dangerous action recognition warnings (including phone calls, smoking, looking around, looking down, and not wearing seat belts). The driver status early warning module simultaneously monitors the driver's identity, fatigue status, and dangerous actions through a multi-task model, improving driver supervision capabilities in the molten aluminum transportation system.

[0139] Table 2 Warning Rules for Driver Status Warning Module

[0140]

[0141]

[0142] The above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the present invention, and all such modifications and substitutions should be covered within the scope of the claims of the present invention. Technical aspects, shapes, and structures not described in detail in this invention are all well-known technologies.

Claims

1. A high-temperature molten aluminum safety transportation early warning system, characterized in that: This includes a database module, an aluminum molten ladle breakage warning module, a transport vehicle status warning module, and a driver status warning module; The database module is used to collect various monitoring data from the aluminum liquid transportation early warning system, and to store the data in different databases according to the data type and data processing requirements. The database module includes a data acquisition module and a data classification module. The data acquisition module is used to collect various monitoring data of high-temperature molten aluminum transportation, and the data classification module is used to classify the collected monitoring data according to data type. Different categories of data are stored in different databases. The aluminum liquid ladle breakage early warning module is used to detect the breakage of the aluminum liquid transport ladle lifting equipment, to predict the risk of aluminum liquid leakage in advance, and to monitor and warn of the high temperature aluminum liquid ladle lifting equipment in real time, thereby preventing the aluminum liquid from being damaged by excessively high temperature. The aluminum molten ladle lifting damage early warning module includes a thermal imaging image preprocessing module and a threshold segmentation damage early warning module. The thermal imaging image preprocessing module is used to crop and grayscale the thermal imaging image of the high-temperature aluminum molten ladle lifting equipment. The threshold segmentation damage early warning module is used to segment the grayscale thermal imaging image, automatically detect whether there are segmented areas exceeding the threshold, and promptly warn of damage to the ladle lifting equipment for segmented areas exceeding the threshold, so as to avoid serious hazards caused by aluminum molten leakage in the ladle lifting equipment. The transport vehicle status early warning module is used to monitor the vehicle's own status and driving status in real time, so as to promptly detect and warn of potential risks to the transport vehicle and ensure the vehicle's safety. The transport vehicle status early warning module includes a vehicle self-fault early warning module and a driving status early warning module. The vehicle self-fault early warning module is used to monitor and warn of the transport vehicle's own status in real time, and the driving status early warning module is used to monitor and warn of the transport vehicle's operating status. The driver status warning module is used to verify the driver's identity, monitor and warn of fatigue driving and dangerous driver behavior; The driver status warning module includes a multi-type data preprocessing module, a multi-task model training module, and a status warning module. The multi-type data preprocessing module is used to preprocess multi-task learning data. The multi-task model training module uses a multi-task learning method to simultaneously train three computer vision detection tasks: identity verification, fatigue detection, and driver dangerous action recognition. The status warning module uses the trained multi-task learning model to verify the driver's identity in a timely manner when a driver's image is detected, and monitors fatigue driving and driver dangerous action recognition and warning in real time during the driver's driving.

2. The high-temperature molten aluminum safety transportation early warning system according to claim 1, characterized in that: The data classification module divides the data according to the real-time requirements and format, and then retrieves the data using different databases based on the data classification.

3. The high-temperature molten aluminum safety transportation early warning system according to claim 1, characterized in that: The aluminum molten ladle breakage early warning module uses infrared thermal images and a threshold-based segmentation method to automatically detect and warn of breakage of the ladle lifting equipment.

4. The high-temperature molten aluminum safety transportation early warning system according to claim 3, characterized in that: The threshold-based segmentation method uses the pixel difference between the background and target regions in the image to determine the segmentation threshold; Let the grayscale thermal imaging image be f(x,y), and the expression for threshold segmentation is: In the formula, f(x,y) is the gray value at pixel (x,y); T1, T2, ..., T n For different feature values, i.e., thresholds; P n These are the processed pixel values ​​used to segment the image; An image output according to a certain rule.

5. The high-temperature molten aluminum safety transportation early warning system according to claim 1, characterized in that: The vehicle fault early warning module monitors and warns about the status of the transport vehicle itself, including tire pressure, tire temperature and vehicle maintenance status. Among them, tire pressure and tire temperature are monitored and alerted by the installation of tire pressure and tire temperature detectors; vehicle maintenance status, that is, vehicle maintenance information, is submitted by the management personnel and the next maintenance mileage and maintenance time are set, and the data is stored on a remote server.

6. The high-temperature molten aluminum safety transportation early warning system according to claim 1, characterized in that: The driving status early warning module monitors and issues early warnings about the operating status of transport vehicles, including vehicle speeding, dangerous distance, exceeding the prescribed area, and vehicle blind spots.

7. The high-temperature molten aluminum safety transportation early warning system according to claim 1, characterized in that: The multi-class data preprocessing module preprocesses multi-task learning data, including data cropping, data augmentation, dataset partitioning, and multi-class label data setting.

8. The high-temperature molten aluminum safety transportation early warning system according to claim 1, characterized in that: The driver status warning module uses a multi-task learning scheme to jointly train driver identity verification, fatigue driving warning, and dangerous action recognition and warning. The total loss function expression for the multi-task learning parameter hard-sharing scheme is: L total =α1L1+α2L2+…+α n L n In the formula, L total The total loss function for multiple tasks; α n To assign weights to the loss functions of the corresponding sub-tasks, the loss functions L of multiple sub-tasks are trained simultaneously using a weighted approach. n ; The objective function expression for the subtask model with multi-task parameter hard sharing is: In the formula, f n Let f be the objective function for the nth task; x be the training data; f g (x) is a multi-task hard parameter layer; f l n (x) is a local-specific layer for the nth task, used to adapt to the needs of each subtask in a special multitasking scenario; Using a multi-task parameter hard-sharing scheme, the same parameters are shared between different detection tasks, which can effectively reduce model parameters and reduce model complexity. At the same time, joint multi-task training improves the model's learning ability.

Images