Digital production operation method of coal team

Abstract

The invention provides a digital production operation method of a coal team, which comprises the following steps: deploying various sensing elements in each area in a station to construct a monitoring system; cloud map information is obtained, and a robot and an unmanned aerial vehicle are arranged in the station according to the cloud map information to form an inspection network; collecting state data streams generated by the monitoring network and the inspection network in real time; performing image acquisition on team personnel and each production device to construct a database; training the regression model according to the database to generate a self-learning model; and when the self-learning model judges that the team personnel trigger illegal actions or the production equipment breaks down, the monitoring system gives an alarm. According to the invention, a monitoring system is built through multiple types of sensing elements, an inspection network is formed under the synergistic effect of the unmanned aerial vehicle and the robot so as to collect the state data flow of the whole production area, and the self-learning model is generated based on the database and the state data flow training, so that the safe operation of the whole production area is guided.

Description

Technical Field

[0001] The invention belongs to the technical field of digital production, and in particular relates to a digital production operation method for a coal team. Background Art

[0002] Traditional coal production monitoring mainly relies on manual inspections, which have problems such as long inspection cycles, low efficiency, and low accuracy. Manual inspections make it difficult to fully and comprehensively grasp the status of equipment in real time, and some potential faults may not be discovered in time, resulting in an increased risk of equipment failures and affecting the safe and stable operation of the power grid. In addition, production equipment failures often occur due to illegal operations by team members. For personnel control, traditional monitoring methods are extremely dependent on the carefulness of monitoring personnel, which often lags behind the time when the accident occurs. Summary of the invention

[0003] In order to solve the problem that the traditional monitoring method described in the background technology is extremely dependent on the carefulness of the monitoring personnel, which often lags behind the time when the accident occurs, the present invention proposes the following technical solutions:

[0004] A digital production operation method for a coal team, comprising:

[0005] Deploy multiple types of sensing elements in various areas of the station to build a monitoring system;

[0006] Obtain cloud map information and set up robots and drones in the station according to the cloud map information to form an inspection network;

[0007] Real-time collection of status data streams generated by monitoring networks and inspection networks;

[0008] Collect images of team members and each production equipment to build a database;

[0009] Training the regression model according to the database to generate a self-learning model;

[0010] When the self-learning model determines that the team member triggers an illegal action or the production equipment fails, the monitoring system issues an alarm.

[0011] The cloud map information is obtained by splicing the point cloud data of each target point in the UAV collection station after data registration.

[0012] Furthermore, after the point cloud data is stitched together, the intelligent center constructs a map model through a triangular meshing algorithm, simplifies the segmentation after texture mapping of the map model, and thus forms cloud map information.

[0013] Furthermore, when the regression model is performing data training, the intelligent center needs to preprocess the state data stream to generate sample data; after the self-learning model is generated, the self-learning model is updated and iterated according to the violation data in the database.

[0014] Furthermore, when the intelligent center pre-processes the state data stream, it is necessary to perform downsampling processing, frame reduction processing, noise reduction optimization, and frequency domain conversion on the state data stream to form the sample data.

[0015] Furthermore, the database stores image information of each team member and safe operation information of the production equipment; when the inspection network is running, the self-learning model performs action recognition and personnel matching on the team members in each site based on the image information, and the self-learning model performs fault judgment on the production equipment based on the safe operation information, so as to be visually displayed in real time in the map model.

[0016] Furthermore, when performing action recognition or personnel matching on the team members, the self-learning model extracts the features of the state data stream through a residual time-shift network to obtain a feature vector, and then calculates the similarity of the feature vector through a similarity algorithm to determine whether the team members have triggered an illegal action.

[0017] Beneficial effects: The present invention builds a monitoring system through multiple types of sensing elements, and forms an inspection network under the coordinated action of drones and robots to collect the status data stream of the entire production area, and generates a self-learning model based on the database and status data stream training, thereby guiding the safe operation of the entire production area. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] Figure 1 The present invention is a flowchart of a digital production and operation method for a coal team according to an embodiment of the present invention. DETAILED DESCRIPTION

[0019] In order to make the purpose, technical solution and advantages of the present application clearer, the present invention will be further described in detail below in conjunction with the accompanying drawings. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.

[0020] It should be understood that the terms "center", "up", "down", "front", "back", "left", "right", "vertical", "horizontal", "top", "bottom", "inside", "outside", etc., indicating the orientation or position relationship, are based on the orientation or position relationship shown in the drawings, and are only for the convenience of describing this patent and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be understood as a limitation on this patent.

[0021] Figure 1 The present invention is a flowchart of a digital production and operation method for a coal team according to an embodiment of the present invention.

[0022] Reference Figure 1 According to an embodiment of the present invention, a digital production and operation method for a coal team includes:

[0023] S000. Deploy multiple types of sensing elements in various areas of the station to build a monitoring system.

[0024] Specifically, in this step, it is necessary to deploy various types of sensing elements in various areas within the coal station. These sensing elements cover various types, such as: temperature sensors for detecting temperature changes, pressure sensors for monitoring pressure values, gas sensors for sensing gas concentrations, vibration sensors for detecting equipment vibration amplitudes, as well as recording equipment and cameras. By placing these sensing elements in different areas in a targeted manner, an interconnected monitoring system is built. The intelligent center uses the monitoring system to conduct real-time and continuous monitoring of the physical parameters, environmental indicators, and equipment operating conditions of various production equipment in the station, providing a basic data source for subsequent production operation management and the construction of intelligent judgment models.

[0025] S010. Obtain cloud map information and set up robots and drones in the station according to the cloud map information to form an inspection network.

[0026] Specifically, in this step, drones equipped with high-precision positioning systems (such as GPS / RTK) are required as collection platforms to ensure that their own location information can be accurately obtained during flight. Each drone is equipped with a high-precision lidar scanner, and the data collection route of the drone is determined according to the scale, layout and equipment distribution of the coal station.

[0027] After collecting the cloud data of each target point, representative feature points are extracted from the collected point cloud data as the basis for data registration, such as the corners, edges, and landmarks of the equipment. Through the matching algorithm, the same-name feature point pairs are found in the point cloud data of different perspectives. Common matching algorithms include descriptor-based matching algorithms (such as SIFT, SURF, etc.) and geometric constraint-based matching algorithms (such as ICP algorithm and its improved algorithm). Substitute the matched feature point pairs into the mathematical model to calculate the relative transformation relationship (including translation, rotation, and translation) between the point cloud data of different perspectives, so as to achieve data registration. A certain number of ground control points are pre-set in the substation, and the three-dimensional coordinates of these control points are accurately measured by high-precision measuring instruments (such as total stations). Preferably, when the drone collects point cloud data, the position information of the ground control points in the point cloud data is obtained at the same time. The measured coordinates of the ground control points are matched with the coordinates of the corresponding points in the point cloud data, a coordinate transformation model is established, and the model parameters are solved by optimization algorithms such as the least squares method, so that the point cloud data is converted to a unified coordinate system to achieve high-precision data registration.

[0028] The registered point cloud data needs to be spliced ​​to integrate the local point cloud data into a complete point cloud model. The splicing algorithm fuses adjacent point cloud data based on the spatial relationship and feature similarity of the point cloud data, fills the gaps and overlapping areas between the data, and forms a continuous and complete point cloud representation. During the splicing process, the algorithm will seamlessly splice the point cloud data in adjacent areas according to the boundary characteristics and spatial continuity of the point cloud data in adjacent areas, ensuring the consistency of the point cloud model of the entire coal station in terms of geometry and spatial position, and providing basic data for building an accurate map model. Finally, the intelligent hub uses a triangular meshing algorithm to connect the points in the point cloud data into triangular patches to construct a map model.

[0029] During the construction process, the algorithm reasonably selects the connection method of triangles according to the distribution density and geometric shape of the point cloud data to ensure that the map model can accurately reflect the shape and surface characteristics of the target object. After the map model is built, the intelligent center uses texture mapping operations to map the image data previously obtained by oblique photography as texture information to the triangular mesh model, so that the map model has a realistic appearance and details. The texture mapping algorithm accurately attaches the color, texture and other information in the image to the surface of the map model based on the correspondence between the image data and the map model. The intelligent center obtains the corresponding cloud map information by calling the map model to guide the drone to conduct inspection flights.

[0030] Preferably, in order to improve the processing efficiency and availability of cloud map information, in this embodiment, after the map model is generated, the texture-mapped map model needs to be simplified and segmented. The simplification algorithm reduces the number of triangles in the map model, reduces the amount of data in the model without affecting the key features and overall shape, and maintains the accuracy of the model. The segmentation operation divides the map model into multiple meaningful areas or objects according to different application requirements and characteristics of the target object.

[0031] S020. Real-time collection of status data streams generated by the monitoring network and the inspection network.

[0032] S030. Collect images of team members and each production equipment to build a database.

[0033] Specifically, in this embodiment, it is necessary to collect images of each team member separately to store image data with facial feature information in the database. In addition, it is also necessary to collect images of production equipment to store the normal operation status of the production equipment in the database, thereby providing a basis for the subsequent intelligent center to judge the operation status of the production equipment. Among them, the image information collected by the production equipment is thermal imaging data and safety indicator data.

[0034] S040. Training the regression model according to the database to generate a self-learning model.

[0035] Specifically, when the intelligent center receives the state data stream, it is necessary to pre-process the state data stream to generate high-quality sample data to lay the foundation for model training. The preprocessing process includes: downsampling processing, frame reduction processing, noise reduction optimization, and frequency domain conversion. In the downsampling process, the intelligent center sets sampling rules based on factors such as the distribution law of the data, the characteristics of the time series, and the change trend of the data to select some data points from the original data stream, thereby reducing the amount of data while retaining the key features of the data. After the downsampling process, the intelligent center needs to perform frame reduction processing based on the dynamic change degree of the content corresponding to the state data stream, the distribution of key information, and the needs of subsequent model training. Then the intelligent center uses filtering algorithms and signal processing technology to identify and remove noise in the state data stream after frame reduction processing. Among them, the filtering methods used include but are not limited to: mean filtering, median filtering, Gaussian filtering, etc. Finally, the intelligent center converts the data stream from a time series to spectral information in the frequency domain through methods such as fast Fourier transform (FFT). After the above series of preprocessing steps, the intelligent center successfully converts the original state data stream into high-quality sample data, which will be used to train the regression model to improve the performance and predictive ability of the model.

[0036] After acquiring the sample data, the intelligent center constructs a self-learning model based on the sample data using the regression model as a framework. In this embodiment, the regression model uses a convolutional neural network (CNN)-related architecture. In other embodiments, the regression model can also be constructed using, including but not limited to, a recurrent neural network (RNN) and its variants such as a long short-term memory network (LSTM), a gated recurrent unit (GRU), or a hybrid architecture model.

[0037] During the construction of the self-learning model, the intelligent center will divide the labeled sample data into training set, validation set and test set according to a certain ratio. The training set data is input into the self-learning model to let the model learn the mapping relationship between the features in the sample and the labeled results. The model parameters are continuously adjusted using appropriate optimization algorithms (such as gradient descent method, Adam optimizer, etc.) to achieve the best performance of the self-learning model on the validation set. Finally, the test set is used to evaluate the generalization ability of the model to ensure that the model can accurately analyze and judge new and unprecedented audio and video information.

[0038] S050. When the self-learning model determines that the team member triggers an illegal action or the production equipment fails, the monitoring system issues an alarm.

[0039] Specifically, after the self-learning model is built, when receiving new status data streams, the intelligent center will input these real-time data into the self-learning model that has been built. The self-learning model will analyze the input status data stream based on its learned knowledge and pattern recognition capabilities, thereby operating the entire area. For example, during the production process in a coal mine, the model will determine whether the current equipment is operating normally and whether the workers' operations are in compliance with regulations. If there is an equipment failure or an operation violation, the self-learning model will coordinate with the monitoring system to issue an alarm.

[0040] Furthermore, the database stores rich and critical information, including image information of each team member and safe operation information of production equipment. These image information include image data of team members in various links and different angles of coal production operations, while safe operation information includes various parameter indicators, appearance morphological characteristics and previous fault case data of production equipment under normal operation. The intelligent center will correlate and analyze the received status data stream with the information in the memory database to compare whether there is a deviation between the output results of the current model and the actual production situation. If a deviation is found, it means that the model may need to be further optimized to adapt to the new situation. At this time, the newly acquired status data stream and the corresponding accurate production information will be re-incorporated into the sample data, and the model will be trained again, and the parameters of the model will be adjusted so that the model can continuously learn new patterns and features, update and iterate itself, thereby improving the model's adaptability to changing actual scenarios and monitoring accuracy.

[0041] When the inspection network is started and running, the self-learning model begins to play its intelligent analysis function. For the team members in the site, the self-learning model first uses the residual time-shift network to perform feature extraction operations on the state data stream. The residual time-shift network is based on the temporal characteristics of personnel actions and can effectively capture the changes in spatial and temporal characteristics of personnel during continuous actions. For example, it can accurately extract the movement trajectory of personnel's limb joints, the time interval of the action, and the transformation characteristics of body posture at different times, and integrate these features into a feature vector. Subsequently, the similarity algorithm is used to calculate the similarity between the feature vector and the standard action feature vector pre-stored in the database.

[0042] The database stores compliant action feature vector templates obtained through a large number of sample training and analysis. These templates represent the correct action patterns that personnel should have under different work tasks. By calculating the similarity between the feature vectors obtained in real time and these templates, the self-learning model can accurately determine whether the team members have triggered illegal actions. If the calculated similarity is lower than the preset threshold, it is determined that the person's actions deviate from the normal norms and there may be a risk of illegal operations. At this time, the self-learning model will issue a corresponding warning signal and display it in real time in a visual way in the map model so that the monitoring personnel can quickly locate the specific site location of the violator.

[0043] For production equipment, the self-learning model judges its faults based on the stored safe operation information. It analyzes the real-time collected equipment operating parameters, appearance images and other status data streams, and compares them with the normal operation characteristic data of the equipment in the database. For example, the temperature, vibration frequency, wear degree and other parameters of key equipment components are monitored. When these parameters exceed the normal range or the image display equipment shows signs of abnormal deformation, damage and other signs, the self-learning model combines the fault case data and characteristic patterns in the safe operation information to determine whether the equipment has failed. Once the equipment is determined to be faulty, the location of the faulty equipment, the type of fault and other information will also be displayed in real time in a visual manner in the map model, providing strong support for timely repair and maintenance, thereby ensuring the safe, efficient and orderly production operations of the coal team, and realizing accurate monitoring and intelligent decision-making of digital production operation management.

[0044] In summary, the present invention builds a monitoring system through multiple types of sensing elements, forms a patrol network under the coordinated action of drones and robots, collects the status data stream of the entire production area, and generates a self-learning model based on the database and status data stream training, thereby guiding the safe operation of the entire production area.

[0045] The foregoing describes certain embodiments of the invention. Other embodiments are within the scope of the following claims.

[0046] The terms "exemplary," "example," and the like used throughout this specification mean "used as an example, instance, or illustration" and do not mean "preferred" or "advantageous" over other embodiments. The detailed description includes specific details for the purpose of providing an understanding of the described techniques. However, these techniques can be implemented without these specific details. In some instances, in order to avoid obscuring the concepts of the described embodiments, well-known structures and devices are shown in block diagram form.

[0047] The optional implementation modes of the embodiments of the present invention are described in detail above in conjunction with the accompanying drawings. However, the embodiments of the present invention are not limited to the specific details in the above implementation modes. Within the technical concept of the embodiments of the present invention, various simple modifications can be made to the technical solutions of the embodiments of the present invention, and these simple modifications all belong to the protection scope of the embodiments of the present invention.

[0048] The above description of the contents of this specification is provided to enable any person of ordinary skill in the art to implement or use the contents of this specification. Various modifications to the contents of this specification will be apparent to those of ordinary skill in the art, and the general principles defined herein may be applied to other variations without departing from the scope of protection of the contents of this specification. Therefore, the contents of this specification are not limited to the examples and designs described herein, but are consistent with the widest range of principles and novel features disclosed herein.

Claims

1. A digital production and operation method for a coal team, characterized in that: include: Deploy multiple types of sensing elements in various areas of the station to build a monitoring system; Obtain cloud map information and set up robots and drones in the station according to the cloud map information to form an inspection network; Real-time collection of status data streams generated by monitoring networks and inspection networks; Collect images of team members and each production equipment to build a database; Training the regression model according to the database to generate a self-learning model; When the self-learning model determines that the team member triggers an illegal action or the production equipment fails, the monitoring system issues an alarm.

2. According to the digital production and operation method of a coal team as described in claim 1, it is characterized in that: The cloud map information is obtained by splicing the point cloud data of each target point in the drone collection station after data alignment.

3. According to the digital production and operation method of a coal team as described in claim 2, it is characterized in that: After the point cloud data is spliced, the intelligent center also needs to build a map model through a triangular meshing algorithm, simplify the segmentation after texture mapping of the map model, and thus form cloud map information.

4. According to the digital production and operation method of a coal team as described in claim 3, it is characterized in that: When the regression model is performing data training, the intelligent center needs to pre-process the state data stream to generate sample data; after the self-learning model is generated, the self-learning model is updated and iterated according to the violation data in the database.

5. According to the digital production and operation method of a coal team as described in claim 4, it is characterized in that: When the intelligent center pre-processes the state data stream, it needs to perform downsampling, frame reduction, noise reduction optimization, and frequency domain conversion on the state data stream to form the sample data.

6. According to the digital production and operation method of a coal team as described in claim 4, it is characterized in that: The database stores image information of each team member and safe operation information of the production equipment; when the inspection network is running, the self-learning model performs action recognition and personnel matching of the team members in each site based on the image information, and the self-learning model performs fault judgment on the production equipment based on the safe operation information, so as to display it visually in real time in the map model.

7. According to the digital production and operation method of a coal team as described in claim 4, it is characterized in that: When performing action recognition or personnel matching on the team members, the self-learning model extracts the features of the state data stream through a residual time-shift network to obtain a feature vector, and then calculates the similarity of the feature vector through a similarity algorithm to determine whether the team members have triggered an illegal action.

Images