Chemical fine material digital twinborn planning display platform building method

Abstract

The invention relates to a chemical fine material digital twinborn planning display platform construction method, which can construct a millimeter-level precision three-dimensional model based on laser radar, unmanned aerial vehicle aerial photography and BIM data, and realize model dynamic update through AI image comparison. Deploying multiple types of sensors, carrying out spatio-temporal data fusion by adopting a distributed message queue, and finishing noise filtering and anomaly detection by combining edge computing nodes; and a YOLOv7 algorithm is used to identify video monitoring abnormal behaviors, and a risk thermodynamic diagram is generated through an LSTM model to predict equipment faults. And a gas leakage diffusion path is dynamically simulated based on a CFD algorithm and meteorological data, and a real-time evacuation route and a rescue resource scheduling scheme are generated. The method is suitable for dangerous chemical substance storage, equipment inspection and pollution diffusion prevention and control scenes.

Description

Technical Field

[0001] The present invention relates to a construction method, and in particular to a construction method for a digital twin planning and display platform for fine chemical materials. Background Art

[0002] For existing chemical industrial parks, their safety management mainly relies on a two-dimensional map system, which is completed based on the traditional two-dimensional mode combined with basic Internet of Things perception devices and manual inspections. It has the following defects:

[0003] 1. The traditional two-dimensional system cannot truly reflect the three-dimensional spatial structure of the park, resulting in incomplete safety risk analysis. For example, when a pipeline leaks, it is impossible to accurately locate the height and spatial diffusion path of the leakage point through a two-dimensional map.

[0004] 2. The existing system relies on manual inspections and regular data collection, and it is difficult to detect abnormal events (such as gas leaks, equipment failures) in a timely manner. For example, if a gas leak in a chemical industrial park is not timely warned due to sensor data transmission delay, it will cause environmental pollution accidents.

[0005] 3. Emergency plans are mostly static documents and lack dynamic simulation capabilities. It is difficult to quickly locate personnel and resources when an accident occurs. For example, in a fire accident in a certain park, due to the escape route not being updated in real time, the personnel evacuation will be chaotic.

[0006] 4. The data of each subsystem (such as video monitoring, environmental monitoring) is stored independently, lacking unified platform integration and linkage analysis. For example, if the video monitoring system and gas sensor data in a certain park are not linked, it is impossible to automatically associate the abnormal gas concentration alarm through the video screen.

[0007] 5. Traditional three-dimensional modeling relies on manual updates and cannot dynamically reflect changes in the physical environment of the park (such as new buildings, equipment relocation). For example, if the position model of a storage tank in a certain park is not updated in time, the emergency drill will not match the actual scenario.

[0008] In view of the above defects, the inventor actively conducts research and innovation in order to create a construction method for a digital twin planning and display platform for fine chemical materials, making it more valuable in the industry. Summary of the Invention

[0009] In order to solve the above technical problems, the purpose of the present invention is to provide a construction method for a digital twin planning and display platform for fine chemical materials.

[0010] A construction method for a digital twin planning and display platform for fine chemical materials according to the present invention includes the following steps:

[0011] Step 1: Generate millimeter-level precision 3D models of the park buildings, equipment and pipelines through LiDAR and drone aerial photography, build spatial structures in combination with BIM model data, and dynamically update the model by comparing real-time image data through AI algorithms; the AI ​​algorithm includes an open source framework consisting of PyTorch, TensorFlow, MMDetection, and OpenMMLab, performs point cloud processing through Open3D and PCL (Point Cloud Library), and uses public data sets such as ScanNet and Matterport3D for pre-training. In addition, data can be collected on-site in the park for corresponding fine-tuning processing to improve local adaptability.

[0012] Step 2: deploy several sensors to collect real-time data, perform timestamp alignment and spatial association processing through distributed message queues, and use the InfluxDB database to store structured data;

[0013] Step 3: Use the YOLOv7 algorithm to identify abnormal behaviors in video surveillance, and use the LSTM model to generate a risk heat map. Use time series data analysis to predict equipment failures.

[0014] Step 4: Use computational fluid dynamics algorithms to simulate the diffusion path of gas leakage, adjust the diffusion model based on meteorological data, and dynamically plan the evacuation routes and rescue resource scheduling;

[0015] Step five: Build the platform backend to support a visual interface for VR / AR interaction, realize holographic display of equipment status, dynamic push of emergency plans, and multi-source data fusion analysis.

[0016] Furthermore, in the above-mentioned method for building a digital twin planning and display platform for chemical fine materials, in the step one, after the lidar scanning and drone aerial photography generate point cloud data, the three-dimensional data of complex indoor areas is supplemented by ground mobile scanning equipment, and refined modeling is performed in Blender software; the dynamic update can match the changes in the physical environment of the park, including the addition of new buildings and equipment relocation.

[0017] Furthermore, in the above-mentioned method for building a digital twin planning and display platform for chemical fine materials, in the step two, the MQTT protocol is used to transmit sensor data, and the edge computing node is cooperated to perform noise filtering and outlier elimination preprocessing, and lossless compression is used for low-frequency changing data and lossy compression is used for high-frequency data when storing data; the sensors include combustible gas sensors, toxic gas sensors, temperature and humidity sensors, and pressure sensors.

[0018] Furthermore, in the above-mentioned method for building a digital twin planning and display platform for chemical fine materials, in the step three, the YOLOv7 algorithm adapts to the edge device deployment through quantitative perception training, and identifies targets including flames, smoke, and equipment oil leakage, and triggers multi-level warnings in conjunction with sensor data; the generation of the risk heat map adopts a combination of historical data and real-time data.

[0019] Furthermore, in the above-mentioned method for building a digital twin planning and display platform for chemical fine materials, in step four, computational fluid dynamics are used to simulate the wind speed influencing factor and the smoke diffusion coefficient variable, and GPU accelerated calculations are performed through the NVIDIA PhysX engine to obtain gas leakage diffusion path simulation or smoke leakage diffusion path simulation, and the diffusion model is adjusted in combination with meteorological data. It can simulate the evacuation of personnel and the dispatch of rescue resources in emergency scenarios, and generate a dynamic plan including escape route navigation and fire resource list through simulation results.

[0020] Furthermore, in the above-mentioned method for building a digital twin planning and display platform for chemical fine materials, in the step five, the visualization interface displays the pressure curve of the corresponding equipment and the three-dimensional overlay information of the maintenance record through a Microsoft Holo Lens device or a VR helmet, and supports operation and query of voice commands.

[0021] Furthermore, the above-mentioned method for building a digital twin planning and display platform for chemical fine materials, in which a Neo4j graph database is used to build a device-space-event association network, supports compound queries based on geographic fences, including querying all associated sensors within a 5-meter radius of a specified device.

[0022] Furthermore, in the above-mentioned method for building a digital twin planning and display platform for chemical fine materials, dynamic texture mapping and detail level rendering of the model are realized through the Unity 3D engine, the model accuracy is automatically switched according to the user's viewing distance, and the real-time status parameters of the equipment are reflected through dynamic texture mapping.

[0023] Furthermore, in the above-mentioned method for building a digital twin planning and display platform for chemical fine materials, equipment failure prediction uses Fourier transform to analyze the vibration spectrum and establish a binding relationship between the equipment's unique UUID and geographic location coordinates.

[0024] Furthermore, in the above-mentioned method for building a digital twin planning and display platform for chemical fine materials, pollutant diffusion simulation is combined with meteorological data and fluid mechanics models, the prediction results are annotated on the three-dimensional scene and the sewage outlet valve is closed in conjunction, and pollution prevention and control recommendations are generated and pushed to the environmental protection department.

[0025] By means of the above scheme, the present invention has at least the following advantages:

[0026] 1. Possessing 3D space holographic perception and dynamic update capabilities. Through LiDAR, drone aerial photography and BIM data fusion, a millimeter-level precision 3D space model (such as storage tanks, pipeline interlayers) can be constructed, and dynamic updates can be achieved based on AI algorithms (such as CNN image comparison models), solving the problem that traditional 2D systems cannot reflect 3D structures. For example, when a pipeline leaks, the model can accurately locate the height of the leak point and the diffusion path, avoiding blind spots in manual inspections.

[0027] 2. It can realize real-time linkage and intelligent early warning of multi-source data. It uses distributed message queues (RocketMQ) for timestamp alignment and spatial association, and combines edge computing nodes (such as noise filtering and 3σ outlier removal) to realize real-time linkage of sensor data (methane concentration, temperature) and video monitoring (YOLOv7 flame recognition). For example, when the methane concentration exceeds the threshold, the video screen is automatically triggered to locate the leak source, solving the problem of response delay caused by the isolation of subsystem data.

[0028] 3. Dynamic emergency plan generation and simulation can be performed. Based on CFD algorithms (such as ANSYS Fluent), the gas leakage diffusion path is simulated, the evacuation route is dynamically adjusted in combination with meteorological data, and the historical case library is automatically matched to generate rescue resource scheduling plans (such as fire hydrant locations and personnel evacuation navigation). For example, in a fire scene, the escape route can be updated in real time to avoid the lag of traditional static plans.

[0029] 4. It is easy to carry out integrated management. The Neo4j graph database is used to build an associated network and support compound queries based on geographic fences.

[0030] 5. Satisfy the compatibility of high-precision models and low-load rendering. Use the Unity 3D engine to achieve dynamic texture mapping and LOD (Level of Detail) rendering, automatically switch model accuracy according to the user's viewing distance (such as the display of the temperature gradient on the surface of the tank), and reduce the computing load through GPU acceleration to avoid the jamming problem caused by the large amount of data in traditional 3D systems.

[0031] 6. It can improve environmental supervision and pollution prevention and control. It can predict the diffusion path of pollutants based on fluid mechanics models and link the valve control of sewage outlets.

[0032] The above description is only an overview of the technical solution of the present invention. In order to more clearly understand the technical means of the present invention and implement it according to the contents of the specification, the following is a detailed description of the preferred embodiments of the present invention in conjunction with the accompanying drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0033] Figure 1It is a structural and functional schematic diagram of the digital twin planning and display platform for fine chemical materials built by the present invention. Specific embodiments

[0034] The following combines the drawings and embodiments to further describe in detail the specific embodiments of the present invention. The following embodiments are used to illustrate the present invention, but are not used to limit the scope of the present invention.

[0035] Such as Figure 1 A method for building a digital twin planning and display platform for fine chemical materials, characterized by including the following steps:

[0036] Step 1: Generate a millimeter-level accurate 3D model of the park buildings, equipment and pipelines through LiDAR (Light Detection and Ranging) and UAV aerial photography, construct a spatial structure in combination with BIM model data, and realize dynamic update of the model by comparing real-time image data through AI algorithms. During implementation, after the LiDAR scan and UAV aerial photography generate point cloud data, the ground mobile scanning equipment is used to supplement the 3D data of complex indoor areas such as pipe galleries and inside equipment. Also, refined modeling of buildings and pipelines and other objects is carried out in Blender software. At the same time, the dynamic update of the model can match the physical environment changes of the park, including new buildings and equipment relocation. In this way, corresponding adjustments are made in a timely manner when the physical environment changes. Considering the accuracy and convenience of aerial photography data, the DJI M300 RTK UAV can be preferably used. Furthermore, the AI algorithms adopted include open source frameworks composed of PyTorch, TensorFlow, MMDetection, OpenMMLab, etc. Point cloud processing can be carried out through Open3D and PCL (Point Cloud Library), and public data sets such as ScanNet and Matterport3D are used for pre-training. In order to improve local adaptability, data can be collected on-site in the park for corresponding fine-tuning processing.

[0037] Step 2: deploy several sensors to collect real-time data, perform timestamp alignment and spatial correlation processing through distributed message queues (such as RocketMQ), and use the InfluxDB database to store structured data. In this way, fast query and data analysis can be supported. Specifically, the MQTT protocol is used to transmit sensor data, and the edge computing nodes are used for preprocessing of noise filtering and outlier removal (such as the 3σ principle). When storing data, lossless compression is used for low-frequency change data and lossy compression is used for high-frequency data. During implementation, lossless compression uses algorithms such as LZMA (Lempel-Ziv-Markov chain algorithm), Zstandard or Deflate to ensure data accuracy. Lossy compression uses DCT (discrete cosine transform) encoding or Delta encoding combined with quantization technology to achieve compression effect, which is suitable for high-frequency and high-redundancy sensor data.

[0038] At the same time, in order to meet the need for expansion of sensing content involved in the field of chemical fine materials, the sensors used in the present invention include combustible gas sensors (methane, ethane), toxic gas sensors (hydrogen sulfide, carbon monoxide), temperature and humidity sensors, and pressure sensors. Of course, the layout of corresponding sensors can be carried out according to the type of field equipment, not limited to the above categories of sensors. During implementation, the deployed InfluxDB database can be used to store real-time data of sensors, supporting fast queries by time range, such as "methane concentration in the past hour".

[0039] Step three: Based on the YOLOv7 algorithm, identify abnormal behaviors in video surveillance, and generate risk heat maps in combination with the LSTM model. Through time series data analysis of various states including equipment vibration, it is easy to predict equipment failure. During the implementation period, the YOLOv7 algorithm can be used to quantify perception training to adapt to edge device deployment. The identified targets include flames, smoke, and equipment oil leakage, and they are linked with sensor data to trigger multi-level warnings. At the same time, the generation of risk heat maps combines historical data with real-time data to improve the accuracy and authenticity of the content displayed in the risk heat maps. In addition, the YOLOv7 algorithm can be trained using the PyTorch framework, combined with quantized perception training (QAT) to optimize the model size and inference speed to adapt to edge device deployment scenarios with limited computing power.

[0040] During the implementation, the YOLOv7 model can be trained by using the PyTorch framework. For example, 100,000 chemical scene images containing abnormal conditions such as flames, smoke, and leaks are annotated. After that, the computing resource usage is reduced through Quantized Aware Training (QAT) to adapt to edge device deployment. In addition, multi-level warning rules can be defined based on abnormal conditions. For example, methane concentration >5%LEL triggers a yellow warning, >10%LEL triggers a red warning and closes nearby valves.

[0041] Step 4: Use the computational fluid dynamics (CFD) algorithm to simulate the gas leakage diffusion path and adjust the diffusion model in combination with meteorological data. In this way, the evacuation route of personnel and the scheduling of rescue resources can be dynamically planned. Specifically, through computational fluid dynamics, the influence factors of wind speed and the variables of smoke diffusion coefficient are simulated, and GPU acceleration calculation is performed through the NVIDIA PhysX engine. Thereby, the simulation of the gas leakage diffusion path or the smoke leakage diffusion path is obtained. By adjusting the diffusion model in combination with meteorological data, the evacuation of personnel and the scheduling of rescue resources in emergency scenarios such as fires and leaks can be simulated. And through the simulation results, a dynamic plan including escape route navigation and a list of fire fighting resources can be generated. Thus, based on the knowledge graph, historical similar cases can be automatically matched, and a dynamic plan including escape routes, rescue resources, and communication links can be generated.

[0042] Moreover, the fluid dynamics algorithm involved in the present invention can be based on the CUDA architecture, and open source frameworks such as OpenFOAM or NVIDIA Flex that support GPU parallel computing can be synchronously adopted to achieve high-performance simulation of the gas and smoke diffusion behaviors in complex flow fields.

[0043] Step 5: Build the platform backend to support a visualization interface for VR / AR interaction, which is convenient for enhancing the user experience and decision-making efficiency. It can realize the holographic display of device status, the dynamic push of emergency plans, and the fusion analysis of multi-source data. During the implementation, the visualization interface displays the pressure curve and three-dimensional superimposed information of the maintenance record of the corresponding device through a Microsoft Holo Lens device or a VR helmet, and supports the operation and query of voice commands. In this way, it is convenient to retrieve the emergency plan through simple operations. At the same time, a VR / AR-like interaction interface can be preset on the Microsoft Holo Lens device or the VR helmet. Specifically, the VR / AR interaction interface is developed based on the Unity 3D engine, and in combination with the Mixed Reality Toolkit (MRTK), it realizes the spatial perception, gesture, and voice interaction support for the HoloLens, and supports the real-time rendering and dynamic superimposed display of three-dimensional data. In this way, the need for holographic display of device status is met, and the convenient operation of users can also be realized.

[0044] Combined with a preferred embodiment of the present invention, a device-space-event association network can be constructed using the Neo4j graph database to facilitate composite queries based on geofences. It can include querying all associated sensors within a radius of 5 meters of a specified device. In this way, the real-time perception and capture of various types of data can be satisfied.

[0045] Furthermore, the dynamic texture mapping of the model and the rendering of the level of detail (LOD) are implemented through the Unity 3D engine. In this way, the dynamic rendering and the setting of the interaction logic can be achieved, the model accuracy can be automatically switched according to the user's viewing distance, and the real-time state parameters of the device can be reflected through the dynamic texture mapping. As a result, a smooth picture can be achieved, and the phenomenon of frame drops can be avoided. At the same time, the mapping accuracy in the current user's viewing distance can be improved, giving a better picture display. Moreover, in order to improve the simulation effect, the Unity 3D engine can be used to add physical properties to key devices such as storage tanks and reactors, and attach parameters such as material density and pressure resistance values. Furthermore, the sensor data stream can be dynamically associated with the 3D model nodes. For example, the data of the temperature sensor of the storage tank is mapped to the color gradient of the model surface in real time.

[0046] At the same time, for equipment fault prediction, Fourier transform is used to analyze the vibration spectrum, and the binding relationship between the unique UUID of the equipment and the geographical location coordinates is established. In this way, in the coordinate system that meets the WGS84 standard, the automatic dispatch of maintenance work orders can be realized when abnormal data is triggered. Furthermore, the pollutant diffusion simulation combines meteorological data and fluid dynamics models, and the prediction results are marked in the 3D scene and linked to close the sewage outlet valve, and pollution prevention and control suggestions are generated and pushed to the environmental protection department.

[0047] The working principle of the present invention is as follows:

[0048] A platform system is established, which includes a 3D modeling engine module, an Internet of Things perception layer module, a data fusion middleware module, an intelligent analysis engine module, a visualization interaction interface module, and a simulation module. Specifically:

[0049] The 3D modeling engine module constructs a refined 3D model of the park buildings, pipelines, and equipment based on drone aerial photography, lidar scanning, and BIM (Building Information Modeling) data. It supports dynamic texture mapping and LOD (level of detail) rendering. In this way, the model accuracy can be automatically switched according to the user's perspective, reducing the computational load. For the dynamic update of the model, the AI algorithm can be used to compare historical and real-time image data, automatically detect scene changes (such as new buildings and equipment displacement), and update the model.

[0050] The Internet of Things perception layer module integrates combustible gas sensors (methane, ethane), toxic gas sensors (hydrogen sulfide, carbon monoxide), temperature and humidity sensors, pressure sensors, and sensors of similar types. It can be used to deploy edge computing nodes. In this way, data preprocessing such as noise filtering and outlier removal can be completed at the device end. Finally, it is uploaded to the platform system through the MQTT protocol.

[0051] The data fusion middleware module adopts a distributed message queue such as Rocket MQ. It realizes the timestamp alignment and spatial association of multi-source data. It can support real-time data stream processing and define a unified data format such as JSON Schema. In this way, heterogeneous device data can be converted into structured data that can be parsed by the platform.

[0052] The intelligent analysis engine module can implement real-time video monitoring based on the YOLOv7 algorithm, and identify various abnormal behaviors including personnel intrusion, equipment oil leakage, flame and smoke. At the same time, combining historical data with the LSTM (Long Short-Term Memory) model, it generates a predicted risk heat map and generates warning signals. Then, it can use time series data analysis methods such as Fourier transform to detect abnormal equipment vibrations and predict the probability of failure.

[0053] The visual interaction interface module can support the switching of perspectives of drones, inspection robots, and personnel positioning. In this way, it can provide an integrated air-ground-space display effect. During implementation, the holographic information of the equipment can be superimposed through Microsoft HoloLens or VR helmets. Thus, it can meet the requirements of displaying data such as pressure curves and maintenance records on the same screen. And it can mark the accident points in the 3D scene, automatically generate a resource scheduling plan and push it to relevant personnel.

[0054] The simulation module uses the CFD (Computational Fluid Dynamics) algorithm to simulate the gas leakage path. Then, it dynamically adjusts the diffusion model in combination with meteorological data. Thus, it can dynamically plan the escape route and simulate the rescue resource scheduling. And it can support multi-person collaborative drills and result evaluation.

[0055] During implementation, by combining the binary data writing mechanism with real-time stream processing technology, the data reading and writing efficiency is significantly optimized. At the same time, relying on the intelligent inspection path planning function, the intensity of manual participation can be effectively reduced and the inspection efficiency can be improved. Through 3D heat maps and visualization simulation means, the timeliness and accuracy of accident response are improved. Using AI algorithms and combining video analysis methods to participate in the processing reduces false alarm interference and enhances the stability of risk identification. The dynamic emergency plan mechanism cooperates with the edge computing architecture to effectively optimize the resource allocation strategy and at the same time reduce the data transmission burden on the cloud.

[0056] Example 1

[0057] High-precision 3D modeling is associated with the equipment. The storage tank area of a chemical plant is scanned by lidar to generate a model with millimeter-level accuracy. During implementation, through the mutual cooperation of lidar and drones, model data acquisition with an accuracy of ±1 mm can be satisfied, and a digital twin scenario can be constructed by the method disclosed in this application.

[0058] After that, bind the temperature sensor data to the nodes on the surface of the storage tank. In this way, the temperature gradient can be displayed in real time. When the temperature in a certain area of the storage tank exceeds the threshold, the corresponding position on the model turns red and an alarm is triggered. Then, through the Rocket MQ stream processing platform, align the timestamps of the video surveillance data and the sensor data. Thus, an association relationship of "storage tank - temperature sensor - alarm record" can be established in Neo4j to support complex queries, such as "storage tanks that triggered alarms in the past week and their associated sensors".

[0059] Example 2

[0060] Combine intelligent inspection with AR assistance, and deploy inspection robots to check the pipeline welds along the preset path. Then, use AI algorithms to identify cracks and mark their positions. The corresponding data is uploaded to the platform in real time to generate an inspection report. This inspection report includes the crack position, size, and recommended repair plan. Then, use the Microsoft HoloLens display device to show holographic information, such as maintenance records, pressure curves, etc. Support users to call the emergency plan using voice commands. For example, "Show the escape route of storage tank A" to facilitate rapid handling.

[0061] Example 3

[0062] Regarding emergency simulation and command and dispatch: Users input parameters such as the fire starting point and wind speed, and through the CFD algorithm, use software such as ANSYS Fluent to perform CFD-related simulation processing to simulate the smoke diffusion path. Specifically, through the CFD algorithm, it is known that: smoke diffusion, gas leakage, temperature field distribution, ventilation simulation, etc. Through the ANSYS Fluent software, it supports turbulence models (such as k-ε, LES), multi-component gas simulation, and flexible setting of boundary conditions, which is suitable for fine simulation of fire smoke diffusion. Combining with the open-source OpenFOAM can meet the needs of local deployment. If multi-physics field coupled simulation needs to be realized, COMSOL Multiphysics can be used for processing. Finally, through the cooperation of PyroSim and FDS (Fire Dynamics Simulator), it can be dedicated to fire smoke diffusion simulation and is suitable for emergency evacuation simulation in building scenarios.

[0063] After that, dynamically mark the safe evacuation route in the 3D scene and push it to the corresponding APP on the mobile phones of relevant personnel. Then, automatically match the nearest fire hydrants and rescue teams to generate an optimal dispatching plan, which may include route navigation and resource list. During the implementation period, the corresponding APP can be customized according to needs, or directly import existing APPs related to on-site and virtual inspections, and coordinate and dock through data interfaces.

[0064] Example 4

[0065] Take environmental supervision and emission control as an example. In the digital twin scene, the sewage outlets of enterprises are marked, and monitoring data such as COD (chemical oxygen demand) and pH value are displayed in real time at the location of the area. When the monitoring value exceeds the standard, the corresponding sewage outlet valve can be automatically closed and the environmental protection department can be notified. At the same time, based on meteorological data and fluid mechanics models, the diffusion path of pollutants in the park is predicted to generate pollution prevention and control recommendations.

[0066] It can be seen from the above textual description and the accompanying drawings that the present invention has the following advantages:

[0067] 1. Possessing 3D space holographic perception and dynamic update capabilities. Through LiDAR, drone aerial photography and BIM data fusion, a millimeter-level precision 3D space model (such as storage tanks, pipeline interlayers) can be constructed, and dynamic updates can be achieved based on AI algorithms (such as CNN image comparison models), solving the problem that traditional 2D systems cannot reflect 3D structures. For example, when a pipeline leaks, the model can accurately locate the height of the leak point and the diffusion path, avoiding blind spots in manual inspections.

[0068] 2. It can realize real-time linkage and intelligent early warning of multi-source data. It uses distributed message queues (RocketMQ) for timestamp alignment and spatial association, and combines edge computing nodes (such as noise filtering and 3σ outlier removal) to realize real-time linkage of sensor data (methane concentration, temperature) and video monitoring (YOLOv7 flame recognition). For example, when the methane concentration exceeds the threshold, the video screen is automatically triggered to locate the leak source, solving the problem of response delay caused by the isolation of subsystem data.

[0069] 3. Dynamic emergency plan generation and simulation can be performed. Based on CFD algorithms (such as ANSYS Fluent), the gas leakage diffusion path is simulated, the evacuation route is dynamically adjusted in combination with meteorological data, and the historical case library is automatically matched to generate rescue resource scheduling plans (such as fire hydrant locations and personnel evacuation navigation). For example, in a fire scene, the escape route can be updated in real time to avoid the lag of traditional static plans.

[0070] 4. It is easy to carry out integrated management. The Neo4j graph database is used to build an associated network and support compound queries based on geographic fences.

[0071] 5. Satisfy the compatibility of high-precision models and low-load rendering. Use the Unity 3D engine to achieve dynamic texture mapping and LOD (Level of Detail) rendering, automatically switch model accuracy according to the user's viewing distance (such as the display of the temperature gradient on the surface of the tank), and reduce the computing load through GPU acceleration to avoid the jamming problem caused by the large amount of data in traditional 3D systems.

[0072] 6. It can improve environmental supervision and pollution prevention and control. It can predict the diffusion path of pollutants based on fluid mechanics models and link the valve control of sewage outlets.

[0073] In addition, the indicated orientations or positional relationships described in the present invention are all based on the orientations or positional relationships shown in the accompanying drawings and are only for the convenience of describing the present invention and simplifying the description. They do not indicate or imply that the device or structure referred to must have a specific orientation or be operated with a specific orientation structure. Therefore, they cannot be understood as limitations on the present invention.

[0074] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. It should be pointed out that a person skilled in the art can make several improvements and modifications without departing from the technical principles of the present invention, and these improvements and modifications should also be regarded as within the scope of protection of the present invention.

Claims

1. A method for building a digital twin planning and display platform for chemical fine materials, characterized in that The following steps are involved: Step 1: Generate millimeter-level precision 3D models of the park buildings, equipment and pipelines through LiDAR and drone aerial photography, build spatial structures based on BIM model data, and dynamically update the model by comparing real-time image data through AI algorithms; Step 2: deploy several sensors to collect real-time data, perform timestamp alignment and spatial association processing through distributed message queues, and use the InfluxDB database to store structured data; Step 3: Use the YOLOv7 algorithm to identify abnormal behaviors in video surveillance, and use the LSTM model to generate a risk heat map. Use time series data analysis to predict equipment failures. Step 4: Use computational fluid dynamics algorithms to simulate the diffusion path of gas leakage, adjust the diffusion model based on meteorological data, and dynamically plan the evacuation routes and rescue resource scheduling; Step five: Build the platform backend to support a visual interface for VR / AR interaction, realize holographic display of equipment status, dynamic push of emergency plans, and multi-source data fusion analysis.

2. The method for building a digital twin planning and display platform for chemical fine materials according to claim 1, characterized in that: In the step 1, after the laser radar scanning and drone aerial photography generate point cloud data, the three-dimensional data of the complex indoor area is supplemented by ground mobile scanning equipment, and refined modeling is performed in Blender software; the dynamic update can match the physical environment changes of the park, including the addition of new buildings and the displacement of equipment; The AI ​​algorithm includes an open source framework consisting of PyTorch, TensorFlow, MMDetection, and OpenMMLab, performs point cloud processing through Open3D and PCL, and uses public data sets for pre-training.

3. The method for building a digital twin planning and display platform for chemical fine materials according to claim 1, characterized in that: In the step 2, the MQTT protocol is used to transmit sensor data, and the edge computing node is used to perform noise filtering and outlier removal preprocessing. When storing data, lossless compression is used for low-frequency change data and lossy compression is used for high-frequency data; The sensors include combustible gas sensors, toxic gas sensors, temperature and humidity sensors, and pressure sensors.

4. The method for building a digital twin planning and display platform for chemical fine materials according to claim 1, characterized in that: In step three, the YOLOv7 algorithm is adapted to edge device deployment through quantitative perception training, and targets identified include flames, smoke, and equipment oil leakage, and is linked with sensor data to trigger multi-level warnings; the risk heat map is generated by combining historical data and real-time data.

5. The method for building a digital twin planning and display platform for chemical fine materials according to claim 1, characterized in that: In step 4, computational fluid dynamics is used to simulate the wind speed influencing factor and the smoke diffusion coefficient variable, and GPU accelerated calculations are performed through the NVIDIA PhysX engine to obtain a gas leakage diffusion path simulation or a smoke leakage diffusion path simulation. The diffusion model is adjusted in combination with meteorological data, which can simulate the evacuation of personnel and the dispatch of rescue resources in emergency scenarios, and generate a dynamic plan including escape route navigation and a list of firefighting resources through simulation results.

6. The method for building a digital twin planning and display platform for chemical fine materials according to claim 1, characterized in that: In step five, the visualization interface displays the pressure curve of the corresponding device, maintains and records three-dimensional overlay information through a Microsoft Holo Lens device or a VR helmet, and supports operation and query of voice commands.

7. The method for building a digital twin planning and display platform for chemical fine materials according to claim 1, characterized in that: The Neo4j graph database is used to build a device-space-event association network, which supports compound queries based on geo-fences, including querying all associated sensors within a 5-meter radius of a specified device.

8. The method for building a digital twin planning and display platform for chemical fine materials according to claim 1, characterized in that: The Unity 3D engine is used to realize dynamic texture mapping and detail level rendering of the model, automatically switch the model accuracy according to the user's viewing distance, and reflect the real-time status parameters of the device through dynamic texture mapping.

9. The method for building a digital twin planning and display platform for chemical fine materials according to claim 1, characterized in that: Equipment failure prediction uses Fourier transform to analyze the vibration spectrum and establish a binding relationship between the device's unique UUID and geographic location coordinates.

10. The method for building a digital twin planning and display platform for chemical fine materials according to claim 1, characterized in that: The pollutant diffusion simulation combines meteorological data with fluid mechanics models. The prediction results are annotated on the three-dimensional scene and the sewage outlet valves are closed in conjunction to generate pollution prevention and control recommendations that are pushed to the environmental protection department.

Images