Grain storage flowability monitoring and dredging device based on ai

Abstract

The application discloses a grain warehouse storage liquidity monitoring and dredging device based on AI, which comprises a multimodal liquidity monitoring module, an AI intelligent decision module, a composite non-contact dredging execution module and a safety closed-loop protection module; the multimodal liquidity monitoring module collects and fuses storage multi-dimensional data, the AI intelligent decision module identifies abnormalities, calculates risk indexes, generates dredging strategies, the composite non-contact dredging execution module dredges according to the strategies, and the safety closed-loop protection module monitors safety states and triggers protection, so that each module cooperates to realize storage monitoring and dredging. The application realizes full-dimensional real-time monitoring of storage, accurately identifies multiple types of abnormalities, dynamically generates adaptive strategies by AI, improves dredging efficiency, avoids grain loss by non-contact dredging, prevents silo collapse and personnel misentry by safety protection, and guarantees the safety of silo operation and the quality of storage.

Description

Technical Field

[0001] This invention belongs to the field of grain storage monitoring technology, specifically relating to an AI-based grain storage flow monitoring and unblocking device. Background Technology

[0002] Grain storage is a crucial link in ensuring food security and stabilizing the grain supply chain. The fluidity of stored materials directly affects the efficiency of grain silo discharge, the quality of stored materials, and the structural safety of the silo. In actual storage operations, due to the characteristics of the grain type (particle size, density, moisture content), environmental conditions (temperature, humidity), and silo structure (silo type, height), stored materials are prone to abnormal fluidity issues such as arching, adhesion to walls, blockage, and freezing. If these issues are not monitored and addressed in a timely manner, they can lead to interrupted discharge, localized mold or insect infestation, and even cause overload on the silo walls due to abnormal accumulation, potentially triggering silo collapses and other safety accidents, resulting in significant economic losses and grain waste.

[0003] In summary, existing grain storage flow monitoring and unblocking technologies are insufficient in terms of monitoring completeness, decision adaptability, unblocking effectiveness, and comprehensive safety protection. They are unable to meet the needs of modern grain storage for full-space monitoring, precise decision-making, efficient unblocking, and proactive protection. There is an urgent need for a technical solution that can overcome the above limitations and achieve synergistic optimization of the entire process of monitoring, decision-making, unblocking, and protection. Summary of the Invention

[0004] To address the problems existing in the background technology, this invention proposes an AI-based grain storage flow monitoring and unblocking device, aiming to achieve coordinated optimization of the entire process of monitoring, decision-making, unblocking, and protection in grain storage flow monitoring.

[0005] The first aspect of this application provides an AI-based method for monitoring the liquidity of grain storage, including:

[0006] The multimodal flow monitoring module is used to collect surface three-dimensional geometric data, surface physical property data, internal physical field data and dynamic flow characteristic data of grain storage materials. After spatiotemporal alignment, compression and fusion processing of the collected data, the fused data is output.

[0007] The AI ​​intelligent decision-making module is connected to the multimodal flow monitoring module. It is used to receive the fused data, identify the types of storage anomalies through a preset anomaly identification model, calculate the risk index based on the identification results and a preset risk quantification formula, and generate corresponding unblocking strategies based on the risk index through a preset decision-making model.

[0008] A composite non-contact dredging execution module is connected to the AI ​​intelligent decision module to receive the dredging strategy and execute the material storage dredging operation;

[0009] The safety closed-loop protection module is connected to the multimodal flow monitoring module, the AI ​​intelligent decision-making module, and the composite non-contact dredging execution module, respectively, and is used to monitor the structural safety status of the grain warehouse, the entry status of personnel, and the quality status of stored materials. When an abnormality is detected, the corresponding safety protection measures are triggered.

[0010] Optionally, the multimodal mobility monitoring module includes:

[0011] The surface monitoring unit includes a lidar and a camera. The lidar is used to collect three-dimensional geometric data of the storage surface, and the camera is used to collect image data of the storage surface.

[0012] The internal physical field monitoring unit includes FMCW radars deployed vertically at intervals along the inner wall of the grain silo, used to collect data on the internal density distribution, porosity, and local compaction of the stored material.

[0013] The dynamic flow monitoring unit includes a high-frequency vibration sensor installed in the outlet pipe, used to collect particle collision frequency data during material flow.

[0014] The environmental monitoring unit includes temperature and humidity sensors for collecting temperature and humidity data inside the grain silo.

[0015] The data fusion unit is connected to the surface monitoring unit, the internal physical field monitoring unit, the dynamic flow monitoring unit, and the environmental monitoring unit, respectively, and is used to perform spatiotemporal alignment, compression, and fusion processing on the collected data.

[0016] Optionally, the data fusion unit is further configured to determine the accuracy coefficients of each sensor under different environmental parameters according to a preset environment and sensor accuracy mapping model, and then perform fusion processing after assigning dynamic weights to the data collected by each sensor according to the accuracy coefficients.

[0017] Optionally, the anomaly recognition model of the AI ​​intelligent decision-making module is a CNN-LSTM hybrid neural network model. The input features of the anomaly recognition model include three-dimensional geometric features of the storage surface, surface physical property features, internal physical field features, dynamic flow features, grain color spectrum features, and volatile gas concentration features. The types of storage anomalies used for identification include arching, wall adhesion, blockage, freezing, mold and caking, insect infestation and adhesion, impurity blockage, and local compaction.

[0018] Optionally, the risk quantification formula of the AI ​​intelligent decision-making module is as follows: Wherein, RI is the risk index; K is the grain type coefficient, preset based on grain density and particle hardness; E is the environmental coefficient, preset based on temperature and humidity inside the grain silo; h is the arching height; s is the wall area; H is the storage humidity; p is the degree of blockage; and I is the ice thickness. , , , , These are the weighting coefficients for the corresponding parameters, which are dynamically iterated by the AI ​​model based on historical fault data.

[0019] Optionally, the decision model of the AI ​​intelligent decision-making module is a DQN model that combines transfer learning and digital twin pre-simulation; the transfer learning is used to use the parameters of a pre-trained grain warehouse model of the same type as the initial value, and complete the model adaptation by supplementing local data; the digital twin pre-simulation is used to construct a grain warehouse digital twin model based on multimodal monitoring data, pre-simulate the candidate dredging strategies generated by the DQN model, and select the strategy with the highest dredging efficiency and the lowest risk of secondary anomalies for output.

[0020] Optionally, the composite non-contact unblocking execution module includes:

[0021] The pulse airflow unit is used to impact the stored material with pulse airflow;

[0022] A laser cleaning unit is used to clean stored materials using lasers.

[0023] A dual-end actuator unit is used to integrate the nozzle of the pulse airflow unit and the laser head of the laser cleaning unit, and is used to adjust the spatial position of the pulse airflow unit and the laser cleaning unit.

[0024] The collaborative control unit is connected to the pulse airflow unit, the laser cleaning unit and the dual-end execution unit respectively, and is used to control the start-up sequence, working parameters and spatial position coordination of each unit according to the dredging strategy.

[0025] Optionally, the composite non-contact dredging execution module further includes a real-time feedback adjustment unit, which includes an infrared temperature feedback device installed in the laser cleaning unit and a pressure feedback device installed in the pulse airflow unit. The real-time feedback adjustment unit is used to collect the surface temperature data of the stored material and the reaction force data of the stored material after the airflow impact, and transmit the data to the AI ​​intelligent decision module to dynamically adjust the laser power and airflow pressure.

[0026] Optionally, the safety closed-loop protection module includes:

[0027] The anti-collapse silo sub-module includes a strain sensor array and an LSTM deformation prediction model. The strain sensor array is used to collect stress data on the silo wall, and the LSTM deformation prediction model is used to predict the deformation trend of the silo wall based on the stress data.

[0028] The submodule for preventing accidental entry includes a UWB positioning tag dispenser, a UWB positioning base station, an audio-visual guidance system, and a central control room terminal. The UWB positioning tag dispenser is used to issue positioning tags, the UWB positioning base station is used to locate personnel positions in real time, the audio-visual guidance system is used to provide evacuation guidance when personnel accidentally enter the area, and the central control room terminal is used to receive and display personnel position information.

[0029] The grain protection submodule includes a high-speed vision sorting camera for identifying the level of grain damage.

[0030] Optionally, it also includes a scene adaptation module, the scene adaptation module comprising:

[0031] The grain warehouse structure adaptation unit is used to add side wall liftable cameras along the length of the warehouse for tall, flat warehouses, and to deploy 360° rotating 16-line lidar and top ring camera for shallow, round warehouses.

[0032] The extreme environment adaptive unit includes a graphene heating element and an anti-condensation coating for the sensor. The graphene heating element is used to maintain the sensor's operating temperature in low-temperature environments, and the anti-condensation coating is used to prevent condensation on the sensor in high-humidity environments.

[0033] Compared with the prior art, the present invention has the following beneficial effects:

[0034] This invention discloses an AI-based grain storage flowability monitoring and unblocking device, comprising: a multimodal flowability monitoring module, an AI intelligent decision-making module, a composite non-contact unblocking execution module, and a safety closed-loop protection module; the multimodal flowability monitoring module is used to collect surface three-dimensional geometric data, surface physical property data, internal physical field data, and dynamic flow characteristic data of the grain storage material, and outputs the collected data after spatiotemporal alignment, compression, and fusion processing; the AI ​​intelligent decision-making module is connected to the multimodal flowability monitoring module and is used to receive the fused data output by the multimodal flowability monitoring module, and to identify anomalies in the stored material through a preset anomaly identification model. The system identifies common types of abnormalities, calculates a risk index based on the identification results and a preset risk quantification formula, and generates corresponding dredging strategies according to the risk index through a preset decision model. A composite non-contact dredging execution module, connected to the AI ​​intelligent decision module, receives the dredging strategies output by the AI ​​intelligent decision module and executes the storage dredging operation accordingly. A safety closed-loop protection module, connected to the multimodal flow monitoring module, the AI ​​intelligent decision module, and the composite non-contact dredging execution module, monitors the structural safety status of the grain silo, personnel entry status, and storage quality status, triggering corresponding safety protection measures when an anomaly is detected. This invention achieves real-time monitoring of storage materials across all dimensions, accurately identifies multiple types of anomalies, dynamically generates adaptive strategies using AI to improve dredging efficiency, avoids grain damage through non-contact dredging, and provides safety protection against silo collapse and accidental personnel entry, ensuring the safety of grain silo operations and the quality of stored materials. Attached Figure Description

[0035] Figure 1 This is a schematic diagram of an AI-based grain storage flow monitoring and unblocking device in one embodiment of the present invention. Detailed Implementation

[0036] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0037] In one embodiment, such as Figure 1 As shown, an AI-powered grain storage flow monitoring and unblocking device is provided. This device includes: a multimodal flow monitoring module, an AI intelligent decision-making module, a composite non-contact unblocking execution module, and a safety closed-loop protection module. Detailed descriptions of each functional module are as follows:

[0038] A multimodal flowability monitoring module is used to collect surface three-dimensional geometric data, surface physical property data, internal physical field data, and dynamic flow characteristic data of grain storage materials. The collected data is then spatiotemporally aligned, compressed, and fused before being output. An AI intelligent decision-making module, connected to the multimodal flowability monitoring module, receives the fused data output by the multimodal flowability monitoring module, identifies the type of storage material anomaly using a preset anomaly identification model, calculates a risk index based on the identification results and a preset risk quantification formula, and generates a corresponding unblocking strategy based on the risk index using a preset decision-making model. A composite non-contact unblocking execution module, connected to the AI ​​intelligent decision-making module, receives the unblocking strategy output by the AI ​​intelligent decision-making module and executes the storage material unblocking operation according to the strategy. A safety closed-loop protection module, connected to the multimodal flowability monitoring module, the AI ​​intelligent decision-making module, and the composite non-contact unblocking execution module, monitors the structural safety status of the grain silo, the personnel entry status, and the storage material quality status, triggering corresponding safety protection measures when an anomaly is detected.

[0039] The surface physical property data, collected by the multimodal flowability monitoring module, reflects the physical state of the grain storage surface (i.e., the outer surface and shallow area of ​​the storage exposed to direct detection by the monitoring equipment). Specifically, it includes temperature distribution data, humidity data, and flow pressure data at the discharge port. This data is collected by temperature and humidity sensors in the environmental monitoring unit and micro-pressure sensors in the dynamic flow monitoring unit, providing the AI ​​intelligent decision-making module with a basis for identifying surface physical state anomalies (such as signs of freezing or arching). The internal physical field data, collected by the internal physical field monitoring unit within the multimodal flowability monitoring module, reflects the internal physical state of the grain storage. Specifically, it includes density distribution data, porosity data, and local compaction data. This data is collected by 77GHz FMCW radars deployed vertically along the inner wall of the grain storage silo, supplementing blind spots in surface monitoring and providing the AI ​​intelligent decision-making module with a basis for identifying internal physical state anomalies (such as decreased flowability due to local compaction).

[0040] Dynamic flow characteristic data refers to parameters collected by the dynamic flow monitoring unit in the multimodal flowability monitoring module, reflecting changes in the physical state of grain storage materials during flow. Specifically, this includes particle collision frequency data as the storage materials flow through the outlet pipe. These data are collected by a high-frequency vibration sensor installed in the outlet pipe to characterize the activity level of the storage material flow, providing a dynamic flow status basis for the AI ​​intelligent decision-making module to determine whether the storage material flowability is normal and to predict blockage risks.

[0041] In this application, the multimodal flow monitoring module includes: a surface monitoring unit, an internal physical field monitoring unit, a dynamic flow monitoring unit, an environmental monitoring unit, and a data fusion unit. The surface monitoring unit includes a lidar and a camera; the lidar is used to collect three-dimensional geometric data of the storage surface, and the camera is used to collect image data of the storage surface. The internal physical field monitoring unit includes 77GHz FMCW radars deployed vertically at intervals along the inner wall of the grain silo, used to collect data on the internal density distribution, porosity, and local compaction of the storage material. The dynamic flow monitoring unit includes a high-frequency vibration sensor installed in the outlet pipe, used to collect particle collision frequency data during storage material flow. The environmental monitoring unit includes temperature and humidity sensors, used to collect temperature and humidity data within the grain silo. The data fusion unit is connected to the surface monitoring unit, internal physical field monitoring unit, dynamic flow monitoring unit, and environmental monitoring unit respectively, and is used to perform spatiotemporal alignment, compression, and fusion processing on the data collected by each unit. The data fusion unit is also used to determine the accuracy coefficient of each sensor under different environmental parameters according to the preset environment and sensor accuracy mapping model, and to perform fusion processing after assigning dynamic weights to the data collected by each sensor according to the accuracy coefficients.

[0042] In this application, the anomaly recognition model of the AI ​​intelligent decision-making module is a CNN-LSTM hybrid neural network model. The input features of the anomaly recognition model include three-dimensional geometric features of the storage surface, surface physical property features, internal physical field features, dynamic flow features, grain color spectrum features, and volatile gas concentration features. The types of storage anomalies used for identification include arching, wall adhesion, blockage, freezing, mold caking, insect infestation, impurity blockage, and local compaction. The volatile gas concentration features are collected by a specific unit in the multimodal flowability monitoring module. Specifically, this module is equipped with a miniature gas sensor array unit, which consists of multiple different types of gas sensors, each with high sensitivity for different volatile gases.

[0043] The risk quantification formula for the AI ​​intelligent decision-making module is as follows: Wherein, RI is the risk index; K is the grain type coefficient, preset based on grain density and particle hardness; E is the environmental coefficient, preset based on temperature and humidity inside the grain silo; h is the arching height; s is the wall area; H is the storage humidity; p is the degree of blockage; and I is the ice thickness. , , , , These are the weighting coefficients for the corresponding parameters, which can be dynamically iterated using an AI model based on historical fault data.

[0044] The decision-making model of the AI ​​intelligent decision-making module is a DQN model that combines transfer learning and digital twin pre-simulation. The transfer learning is used to take the parameters of a pre-trained grain warehouse model of the same type as the initial value and complete the model adaptation by supplementing local data. The digital twin pre-simulation is used to construct a grain warehouse digital twin model based on multimodal monitoring data, pre-simulate the candidate dredging strategies generated by the DQN model, and select the strategy with the highest dredging efficiency and the lowest risk of secondary anomalies for output.

[0045] In this application, the composite non-contact dredging execution module includes: a pulse airflow unit, a laser cleaning unit, a dual-end execution unit, and a collaborative control unit. The pulse airflow unit is used to impact the stored material with pulse airflow; the laser cleaning unit is used to clean the stored material with a laser; the dual-end execution unit integrates the nozzle of the pulse airflow unit and the laser head of the laser cleaning unit, and is used to adjust the spatial positions of the pulse airflow unit and the laser cleaning unit; the collaborative control unit is connected to the pulse airflow unit, the laser cleaning unit, and the dual-end execution unit respectively, and is used to control the start-up sequence, working parameters, and spatial position coordination of each unit according to the dredging strategy.

[0046] The laser cleaning unit is a component of the composite non-contact unblocking module. It emits a laser beam of a specific wavelength to act on the surface of the stored material and the areas where it adheres to the walls. The energy generated by the laser beam raises the surface temperature of the stored material, reducing the adhesion between the particles. Simultaneously, the laser's directional impact on the adhered material causes it to detach from the silo wall or arched structure, achieving non-contact cleaning of abnormal conditions such as material adhesion and localized arching. This unit works in conjunction with other actuators, such as the pulse airflow unit, to complete the unblocking operation. The laser power, irradiation angle, and action time of the laser cleaning unit can be dynamically adjusted according to the unblocking strategy output by the AI ​​intelligent decision-making module to adapt to different types of stored material anomalies and grain characteristics. The dual-end actuator is the core actuator of the composite non-contact unblocking module. Its main body is a multi-degree-of-freedom mechanical structure driven by a servo motor. A key feature is the integration of two functional ends—the nozzle of the pulse airflow unit and the laser head of the laser cleaning unit. The dual-end execution unit is a robotic arm that can adjust the spatial position and working angle of the two ends through multi-degree-of-freedom motion according to the unblocking strategy output by the AI ​​intelligent decision module. This allows the pulse airflow nozzle to accurately aim at the arched area of ​​the stored material and spray airflow, while the laser head can be directed to the wall or icy area. At the same time, it can coordinate and control the working sequence of the two ends to avoid mutual interference between airflow and laser, ensuring efficient connection of the non-contact unblocking process and adapting to the operational needs of different abnormal storage scenarios.

[0047] The composite non-contact dredging execution module also includes a real-time feedback adjustment unit. The real-time feedback adjustment unit includes an infrared temperature feedback device installed in the laser cleaning unit and a pressure feedback device installed in the pulse airflow unit. It is used to collect the surface temperature data of the stored material and the reaction force data of the stored material after the airflow impact, and transmit the data to the AI ​​intelligent decision module to dynamically adjust the laser power and airflow pressure.

[0048] In this application, the safety closed-loop protection module includes: an anti-collapse silo submodule, an anti-personnel intrusion submodule, and a grain protection submodule. The anti-collapse silo submodule includes a strain sensor array and an LSTM deformation prediction model. The strain sensor array is used to collect stress data on the silo wall, and the LSTM deformation prediction model is used to predict the deformation trend of the silo wall based on the stress data. The anti-personnel intrusion submodule includes a UWB positioning tag dispenser, a UWB positioning base station, an audio-visual guidance system, and a central control room terminal. The UWB positioning tag dispenser is used to issue positioning tags, the UWB positioning base station is used to locate personnel positions in real time, the audio-visual guidance system is used to provide evacuation guidance in case of accidental intrusion, and the central control room terminal is used to receive and display personnel position information. The grain protection submodule includes a high-speed visual sorting camera for identifying the grain damage level, which includes mild, moderate, and severe damage. Mild damage is defined as cracks less than 2mm, moderate damage as cracks 2-5mm, and severe damage as breakage greater than 5mm.

[0049] Optionally, the AI-based grain storage flow monitoring and unblocking device further includes a scene adaptation module, which includes:

[0050] The grain warehouse structure adaptation unit is used to add side wall liftable cameras along the length of the warehouse for tall, flat warehouses, and to deploy 360° rotating 16-line lidar and top ring camera for shallow, round warehouses.

[0051] The extreme environment adaptive unit includes a graphene heating element and an anti-condensation coating for the sensor. The graphene heating element is used to maintain the sensor's operating temperature in low-temperature environments, and the anti-condensation coating is used to prevent condensation on the sensor in high-humidity environments.

[0052] The multimodal flowability monitoring module, as the core of the system's information acquisition and preprocessing, first collects three-dimensional geometric data and surface image data of the storage surface through the lidar and camera of the surface monitoring unit. Then, it collects internal density distribution data, porosity data, and local compaction data of the storage material through the 77GHz FMCW radar of the internal physical field monitoring unit. Finally, it collects particle collision frequency data during storage material flow through the high-frequency vibration sensor of the dynamic flow monitoring unit, and collects temperature and humidity data inside the grain silo through the temperature and humidity sensors of the environmental monitoring unit. All of the above multi-source raw data are transmitted to the data fusion unit. The data fusion unit first determines the accuracy coefficient of each sensor under the current environmental parameters based on a preset environment-sensor accuracy mapping model. It then assigns dynamic weights to the data collected by each sensor according to the accuracy coefficients, and performs spatiotemporal alignment and compression processing on all weighted data to form standardized fused data. Finally, the fused data is transmitted to the AI ​​intelligent decision-making module.

[0053] After receiving the fused data from the multimodal flowability monitoring module, the AI ​​intelligent decision-making module first calls the anomaly recognition model composed of a CNN-LSTM hybrid neural network. Using the three-dimensional geometric features of the storage surface, surface physical properties, internal physical field features, dynamic flow characteristics, grain color spectrum characteristics, and volatile gas concentration characteristics extracted from the fused data as input, it identifies eight types of anomalies: arching, wall adhesion, blockage, freezing, mold caking, insect infestation, impurity blockage, and localized compaction. Subsequently, the risk quantification unit, based on the identified anomaly types, substitutes them into a preset risk quantification formula. The risk index is calculated as follows: RI is the risk index; K is the grain type coefficient, which is preset based on grain density and particle hardness; E is the environmental coefficient, which is preset based on temperature and humidity inside the grain silo; h is the arching height; s is the wall area; H is the storage humidity; p is the degree of blockage; and I is the ice thickness. , , , , These are the weight coefficients for the corresponding parameters. Next, combining transfer learning and digital twin pre-simulation, the DQN decision model, with anomaly type and risk index as input, first uses transfer learning to call the parameters of a trained model of the same type of grain warehouse as initial values, supplements local data to complete model adaptation, and then constructs a grain warehouse digital twin model based on fused data to pre-simulate the generated candidate dredging strategies, selecting the final dredging strategy with the highest dredging efficiency and the lowest risk of secondary anomalies; the AI ​​intelligent decision module simultaneously transmits the final dredging strategy to the composite non-contact dredging execution module and transmits safety monitoring instructions matching the current anomaly type and dredging strategy to the safety closed-loop protection module.

[0054] After receiving the dredging strategy from the AI ​​intelligent decision-making module, the composite non-contact dredging execution module analyzes the start-up timing, working parameters, and spatial position requirements in the strategy through the collaborative control unit. Simultaneously, it sends control signals to the pulse airflow unit, laser cleaning unit, and dual-end execution unit: controlling the dual-end execution unit to adjust the spatial position of the integrated pulse airflow nozzle and laser head; controlling the pulse airflow unit to spray pulse airflow at a set pressure to impact the stored material; and controlling the laser cleaning unit to emit laser light at a set power to clean the stored material. Simultaneously, the real-time feedback adjustment unit of the composite non-contact dredging execution module collects surface temperature data of the stored material through an infrared temperature feedback device and reaction force data of the stored material after airflow impact through a pressure feedback device. This real-time feedback data is continuously transmitted to the AI ​​intelligent decision-making module. The AI ​​intelligent decision-making module dynamically adjusts the laser power and airflow pressure parameters in the dredging strategy based on the feedback data and sends the adjusted strategy back to the composite non-contact dredging execution module, achieving dynamic optimization of the dredging process.

[0055] After receiving the safety monitoring command from the AI ​​intelligent decision-making module, the safety closed-loop protection module activates the monitoring functions of the corresponding sub-modules: the anti-collapse sub-module collects the stress data of the silo wall through a strain sensor array, inputs the stress data into the LSTM deformation prediction model to predict the deformation trend of the silo wall, and if the real-time stress value reaches the preset threshold or the predicted deformation trend will trigger the risk of silo collapse, it immediately transmits a silo collapse anomaly signal to the AI ​​intelligent decision-making module; the anti-personnel trespassing sub-module receives the signal of the UWB positioning tag carried by personnel through the UWB positioning base station, obtains the personnel location data in real time and transmits it to the central control room terminal, and if personnel are detected entering the preset restricted area, it immediately transmits a personnel trespassing anomaly signal to the AI ​​intelligent decision-making module. Simultaneously, the sound and light guidance system is triggered to activate LED evacuation path indicators and voice guidance; the grain protection submodule collects images of grains at the discharge port through a high-speed visual sorting camera, identifies the grain damage level (mild is cracks less than 2mm, moderate is cracks 2-5mm, severe is broken grains greater than 5mm), and if the damage level reaches the preset protection threshold, it transmits an abnormal grain damage signal to the AI ​​intelligent decision module; after receiving any abnormal signal, the AI ​​intelligent decision module immediately sends a command to the composite non-contact dredging execution module to pause or adjust the dredging operation, and triggers the corresponding safety protection linkage according to the abnormality type, until the safety closed-loop protection module feeds back an abnormality cancellation signal, and then restarts or resumes the dredging operation.

[0056] The various modules in the AI-based grain storage flow monitoring and unblocking device can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in the processor of a computer device in hardware form or independent of it, or stored in the memory of the computer device in software form, so that the processor can call and execute the corresponding operations of each module.

[0057] Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing embodiments or make equivalent substitutions for some of the technical features. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. An AI-based grain storage flow monitoring and unblocking device, characterized in that, include: The multimodal flow monitoring module is used to collect surface three-dimensional geometric data, surface physical property data, internal physical field data and dynamic flow characteristic data of grain storage materials. After spatiotemporal alignment, compression and fusion processing of the collected data, the fused data is output. The AI ​​intelligent decision-making module is connected to the multimodal flow monitoring module. It is used to receive the fused data, identify the types of storage anomalies through a preset anomaly identification model, calculate the risk index based on the identification results and a preset risk quantification formula, and generate corresponding unblocking strategies based on the risk index through a preset decision-making model. A composite non-contact dredging execution module is connected to the AI ​​intelligent decision module to receive the dredging strategy and execute the material storage dredging operation; The safety closed-loop protection module is connected to the multimodal flow monitoring module, the AI ​​intelligent decision-making module and the composite non-contact dredging execution module respectively. It is used to monitor the structural safety status of the grain warehouse, the entry status of personnel and the quality status of stored materials, and trigger corresponding safety protection measures when an abnormality is detected. The risk quantification formula for the AI ​​intelligent decision-making module is as follows: Wherein, RI is the risk index; K is the grain type coefficient, preset based on grain density and particle hardness; E is the environmental coefficient, preset based on temperature and humidity inside the grain silo; h is the arching height; s is the wall area; H is the storage humidity; p is the degree of blockage; and I is the ice thickness. , , , , These are the weighting coefficients for the corresponding parameters, which are dynamically iterated by the AI ​​model based on historical fault data.

2. The AI-based grain storage flow monitoring and unblocking device according to claim 1, characterized in that, The multimodal liquidity monitoring module includes: The surface monitoring unit includes a lidar and a camera. The lidar is used to collect three-dimensional geometric data of the storage surface, and the camera is used to collect image data of the storage surface. The internal physical field monitoring unit includes FMCW radars deployed vertically at intervals along the inner wall of the grain silo, used to collect data on the internal density distribution, porosity, and local compaction of the stored material. The dynamic flow monitoring unit includes a high-frequency vibration sensor installed in the outlet pipe, used to collect particle collision frequency data during material flow. The environmental monitoring unit includes temperature and humidity sensors for collecting temperature and humidity data inside the grain silo. The data fusion unit is connected to the surface monitoring unit, the internal physical field monitoring unit, the dynamic flow monitoring unit, and the environmental monitoring unit, respectively, and is used to perform spatiotemporal alignment, compression, and fusion processing on the collected data.

3. The AI-based grain storage flow monitoring and unblocking device according to claim 2, characterized in that, The data fusion unit is also used to determine the accuracy coefficient of each sensor under different environmental parameters according to the preset environment and sensor accuracy mapping model, and to perform fusion processing after assigning dynamic weights to the data collected by each sensor according to the accuracy coefficients.

4. The AI-based grain storage flow monitoring and unblocking device according to claim 1, characterized in that, The anomaly recognition model of the AI ​​intelligent decision-making module is a CNN-LSTM hybrid neural network model. The input features of the anomaly recognition model include three-dimensional geometric features of the storage surface, surface physical property features, internal physical field features, dynamic flow features, grain color spectrum features, and volatile gas concentration features. The types of storage anomalies used for identification include arching, wall adhesion, blockage, freezing, mold and caking, insect infestation and adhesion, impurity blockage, and local compaction.

5. The AI-based grain storage flow monitoring and unblocking device according to claim 1, characterized in that, The decision-making model of the AI ​​intelligent decision-making module is a DQN model that combines transfer learning and digital twin pre-simulation. The transfer learning is used to take the parameters of a pre-trained grain warehouse model of the same type as the initial value and complete the model adaptation by supplementing local data. The digital twin pre-simulation is used to construct a grain warehouse digital twin model based on multimodal monitoring data, pre-simulate the candidate dredging strategies generated by the DQN model, and select the strategy with the highest dredging efficiency and the lowest risk of secondary anomalies for output.

6. The AI-based grain storage flow monitoring and unblocking device according to claim 1, characterized in that, The composite non-contact unblocking execution module includes: The pulse airflow unit is used to impact the stored material with pulse airflow; A laser cleaning unit is used to clean stored materials using lasers. A dual-end actuator unit is used to integrate the nozzle of the pulse airflow unit and the laser head of the laser cleaning unit, and is used to adjust the spatial position of the pulse airflow unit and the laser cleaning unit. The collaborative control unit is connected to the pulse airflow unit, the laser cleaning unit and the dual-end execution unit respectively, and is used to control the start-up sequence, working parameters and spatial position coordination of each unit according to the dredging strategy.

7. The AI-based grain storage flow monitoring and unblocking device according to claim 6, characterized in that, The composite non-contact dredging execution module also includes a real-time feedback adjustment unit, which includes an infrared temperature feedback device installed in the laser cleaning unit and a pressure feedback device installed in the pulse airflow unit. It is used to collect the surface temperature data of the stored material and the reaction force data of the stored material after the airflow impact, and transmit the data to the AI ​​intelligent decision module to dynamically adjust the laser power and airflow pressure.

8. The AI-based grain storage flow monitoring and unblocking device according to claim 1, characterized in that, The safety closed-loop protection module includes: The anti-collapse silo sub-module includes a strain sensor array and an LSTM deformation prediction model. The strain sensor array is used to collect stress data on the silo wall, and the LSTM deformation prediction model is used to predict the deformation trend of the silo wall based on the stress data. The submodule for preventing accidental entry includes a UWB positioning tag dispenser, a UWB positioning base station, an audio-visual guidance system, and a central control room terminal. The UWB positioning tag dispenser is used to issue positioning tags, the UWB positioning base station is used to locate personnel positions in real time, the audio-visual guidance system is used to provide evacuation guidance when personnel accidentally enter the area, and the central control room terminal is used to receive and display personnel position information. The grain protection submodule includes a high-speed vision sorting camera for identifying the level of grain damage.

9. The AI-based grain storage flow monitoring and unblocking device according to claim 1, characterized in that, It also includes a scene adaptation module, which includes: The grain warehouse structure adaptation unit is used to add side wall liftable cameras along the length of the warehouse for tall, flat warehouses, and to deploy 360° rotating 16-line lidar and top ring camera for shallow, round warehouses. The extreme environment adaptive unit includes a graphene heating element and an anti-condensation coating for the sensor. The graphene heating element is used to maintain the sensor's operating temperature in low-temperature environments, and the anti-condensation coating is used to prevent condensation on the sensor in high-humidity environments.

Images