Intelligent dust collection and spraying robot system for factories based on multi-sensor collaboration

Abstract

This invention belongs to the field of dust collection robot technology, specifically relating to a factory intelligent dust collection spray robot system based on multi-sensor collaboration. The system includes a vehicle body, a sweeper mounted on the vehicle body, an electronic control system within the vehicle body, a vision recognition module mounted on the vehicle body, a dust collection module mounted on the vehicle body, and a dry fog module mounted on the vehicle body. The electronic control system controls the operation of the vision recognition module, dust collection module, and dry fog module. The vision recognition module is used to construct a map and optimize the dynamic path in real time. The dust collection module absorbs dust through graded filtration. The dry fog module dynamically adjusts the atomized particle size and spray flow rate of the dry fog nozzles according to the dust concentration to prevent dust diffusion. This invention features improved dust collection efficiency, reduced development and maintenance costs, and dust suppression effects.

Description

Technical Field

[0001] This invention belongs to the field of dust collection robot technology, specifically relating to a factory intelligent dust collection spray robot system based on multi-sensor collaboration. Background Technology

[0002] Factories, as environments where large machinery operates, often generate large amounts of dust, which needs to be cleaned up in a timely manner, otherwise safety hazards may easily arise.

[0003] However, existing factory dust collection methods have the following limitations:

[0004] Manual cleaning:

[0005] Relying on manual labor using tools such as brooms and dustpans is not only extremely inefficient (the efficiency of a single person is only 50-80 cubic meters per minute), but also... 2 The dust concentration is high (per hour), and it cannot completely remove dust adhering to hidden areas such as equipment crevices and corners. Long-term exposure to high concentrations of dust can easily lead to occupational health problems such as pneumoconiosis. At the same time, the cleaning process can easily generate secondary dust, further deteriorating the working environment. In addition, manual cleaning is difficult to monitor and handle around the clock, and dust accumulation may cause equipment failure or fire hazards.

[0006] Traditional vacuum cleaners:

[0007] Most of these are small, mobile devices with limited suction power and insufficient battery life, making them unsuitable for handling large areas and high-concentration dust environments. Their filtration systems typically only have a single filter, which is prone to clogging, leading to decreased suction power and requiring frequent manual cleaning or filter replacement. Furthermore, traditional vacuum cleaners lack intelligent sensing capabilities and cannot automatically detect changes in dust concentration, requiring manual start-up and shutdown, which fails to meet the automation needs of continuous factory production.

[0008] Fixed dust collection device:

[0009] While central dust collection systems or fixed dust collection ducts in workshops can cover a large area, their fixed layout makes them difficult to adapt to changes in factory equipment or work processes, resulting in poor flexibility. Installation costs are high, the installation period is long, and maintenance is complex; a malfunction could lead to widespread dust collection disruptions. Furthermore, fixed systems are not specifically designed for high-dust-generating areas (such as material loading / unloading points or around crushing equipment), easily creating dust collection blind spots.

[0010] Visual interference and operational risks:

[0011] In high-dust work areas such as welding, grinding, and ore crushing, instantaneous dust concentrations can reach 50-200 mg / m³. 3Dust can severely obstruct operators' vision, leading to decreased equipment operating accuracy and increased risk of misoperation. For example, forklift drivers may find it difficult to accurately judge the position of goods in a dusty environment, increasing the risk of collisions; equipment maintenance personnel may be unable to clearly observe instrument data, potentially delaying troubleshooting, or even causing abnormal equipment conditions (such as overheating or oil leaks) to be obscured by dust accumulation, resulting in missed opportunities for optimal maintenance.

[0012] Dust explosion and fire hazards:

[0013] Combustible dust (such as metal dust, flour dust, and coal dust) in a high-concentration suspended state is highly susceptible to explosion upon contact with open flames, static electricity, or other ignition sources. Dust concentrations in high-dust-affected areas fluctuate greatly and are difficult to monitor in real time; once the dust reaches its explosion limits (e.g., the lower explosive limit for aluminum powder is 37-50 g / m³), it can easily ignite. 3 This could lead to major safety accidents. Furthermore, dust accumulation on the surface of electrical equipment can affect heat dissipation and increase the risk of short circuits and fires.

[0014] The contradiction between efficiency and cost:

[0015] Traditional dust collection methods cannot adapt to the large-scale, continuous production rhythm of factories. For example, manual cleaning requires production to stop, which seriously affects production efficiency; although fixed dust collection devices can operate online, they have high energy consumption (1.2-1.8 kW•h / 1000m³). 3 Furthermore, the inability to dynamically adjust power based on actual dust concentration results in persistently high operating costs. This is particularly problematic in large workshops (areas exceeding 10,000 m²). 2 In traditional systems, multiple devices are needed to cover the area, significantly increasing management and maintenance costs.

[0016] Insufficient adaptability to complex environments:

[0017] Factory environments are complex and varied, with equipment layouts at varying heights, narrow passageways, and dynamic obstacles (such as moving vehicles and workers). Existing equipment lacks environmental awareness and autonomous decision-making capabilities, and cannot automatically avoid obstacles or adjust dust collection strategies according to different dust characteristics (such as particle size and viscosity). For example, traditional vacuum cleaners have difficulty reaching dust that has settled on slopes or steps, while manual cleaning poses a risk of slipping and falling.

[0018] Lack of data-driven management:

[0019] Traditional dust collection equipment lacks data acquisition and analysis capabilities, making it impossible to provide real-time feedback on key information such as dust collection effectiveness and equipment status. Factory managers struggle to grasp dust distribution patterns and equipment operating efficiency, hindering their ability to optimize dust collection solutions through data-driven approaches and failing to meet environmental protection departments' requirements for real-time monitoring and source tracing of dust emissions.

[0020] Therefore, it is very important to design a factory intelligent dust collection spray robot system based on multi-sensor collaboration that can improve dust collection efficiency, save development and maintenance costs, and has dust suppression effects. Summary of the Invention

[0021] The present invention aims to overcome the problems of low dust collection efficiency, insufficient applicability to complex environments, and high maintenance costs of existing factory dust collection methods. It provides a factory intelligent dust collection spray robot system based on multi-sensor collaboration that can improve dust collection efficiency, save development and maintenance costs, and has a dust suppression effect.

[0022] To achieve the above-mentioned objectives, the present invention adopts the following technical solution:

[0023] The intelligent dust collection and spraying robot system for factories based on multi-sensor collaboration includes a vehicle body, a sweeper installed on the vehicle body, an electronic control system installed inside the vehicle body, a vision recognition module installed on the vehicle body, a dust collection module installed on the vehicle body, and a dry fog module installed on the vehicle body.

[0024] The electronic control system is used to control the operation of the vision recognition module, the dust collection module, and the dry fog module;

[0025] The visual recognition module is used to construct maps and optimize dynamic paths in real time.

[0026] The dust collection module is used to absorb dust through a graded filtration process;

[0027] The dry fog module is used to dynamically adjust the atomization particle size and spray flow rate of the dry fog nozzle according to the dust concentration to resist dust diffusion.

[0028] Preferably, the visual recognition module includes four cameras located on the front, rear, left and right sides of the vehicle body, a collision avoidance radar located at the front of the vehicle body, and an infrared sensor and a dust concentration sensor located on both sides of the front camera: the infrared sensor, the dust concentration sensor and each camera are electrically connected to the electronic control system.

[0029] Preferably, the dust collection module includes a dust collection pipe inlet, a primary filter chamber, a secondary filter chamber, a tertiary filter chamber, and an exhaust port; the dust collection pipe inlet is connected to the cleaner; the dust collection pipe inlet, the primary filter chamber, the secondary filter chamber, the tertiary filter chamber, and the exhaust port are connected in sequence; and a supporting column is provided below the primary filter chamber, the secondary filter chamber, and the tertiary filter chamber.

[0030] Preferably, the primary filtration chamber is a gravity settling chamber used to separate large dust particles; the secondary filtration chamber is a filter bag filtration zone used to filter submicron particles using PTFE membrane filter media; and the tertiary filtration chamber is an activated carbon filtration zone used to adsorb oily and odorous dust.

[0031] The filter bag is also equipped with a differential pressure sensor in the filtration zone to monitor the filter bag clogging status in real time.

[0032] Preferably, the dry fog module includes a water tank, a spray host, a water pump, and a nozzle; the water tank is connected to the spray host; the water pump is located inside the spray host; the water pump is connected to the water outlet on the spray host; a nozzle is provided at the water outlet; the water tank is provided with a water inlet, and a float and a water level probe are also provided inside the water tank.

[0033] Preferably, the electronic control system includes a vehicle control unit (VCU), a battery management system (BMS), an on-board power supply, a motor controller (MCU), a drive motor, and a reducer assembly. The BMS and MCU are both electrically connected to the VCU. The on-board power supply is electrically connected to the BMS. The drive motor is electrically connected to the MCU. The reducer assembly is electrically connected to the drive motor. The VCU receives input signals and controls the operation of the vision recognition module, dust collection module, and dry fog module via CAN communication.

[0034] Preferably, the vehicle body includes a suspension system; the suspension system includes an angle sensor (collecting the steering angle of the steering gear to determine the steering amplitude and direction), a speed sensor (collecting the movement speed of the suspension system to provide a basis for control system adjustment), a steering gear, a tie rod, a shock absorber, an upper wishbone, a lower wishbone, a wheel hub, and a swing bearing; the tie rod is connected to the steering gear; the upper and lower wishbones are connected via a swing bearing; the wheel hub is mounted on the swing bearing; the lower end of the shock absorber is connected to the lower wishbone, and the upper end of the shock absorber is connected to the upper wishbone.

[0035] Preferably, it also includes an automatic replenishment bin for completing water and electricity connection and dust transfer; the automatic replenishment bin has a built-in weighing sensor and a liquid level sensor; the automatic replenishment bin is equipped with a millimeter-wave radar; the automatic replenishment bin is also equipped with a water inlet, a solenoid valve water replenishment switch, a charging base and a display screen.

[0036] Compared with the prior art, the beneficial effects of this invention are: (1) Compared with the traditional solution, the path planning efficiency of this invention is improved by 40%, which can flexibly avoid dynamic obstacles, significantly reduce invalid paths and repeated dust collection operations, and increase the dust collection coverage rate to more than 98%; the visual recognition module is designed with a standardized interface, which supports plug-and-play of multiple types of sensors such as lidar and industrial cameras; the algorithm framework has reserved interfaces, which can quickly integrate advanced detection models such as YOLO series and Transformer, which is convenient for algorithm iteration upgrades for different dust environments (such as metal dust and fiber dust) and reduces secondary development costs; (2) The three-stage filtration system of this invention forms a progressive purification link of "coarse separation-fine filtration-deep purification", and the final emission dust concentration is less than 5mg / m³, which far exceeds the national environmental protection standard; (3) This invention can dynamically adjust the atomization particle size (5-20μm) and spray flow rate of the dry fog nozzle according to the dust concentration, and the dust suppression efficiency is increased to 90%, and the water consumption is reduced by 60% compared with the traditional spray system; combined with meteorological data (wind speed, wind direction) and working conditions, the system automatically adjusts the dry fog coverage area. When strong winds are detected, the amount of dry fog spray is increased on the windward side to form an "air curtain barrier" to effectively resist dust diffusion and ensure the stability of dust suppression effect; (4) Based on visual recognition and laser navigation, the automatic supply bin in this invention can automatically align with the equipment interface to complete water and electricity connection and dust transfer. It supports parallel supply scheduling of multiple devices, coordinates the operation sequence through the Internet of Things platform, improves supply efficiency by 50%, and reduces the risk of manual intervention; (5) After the angle sensor and speed sensor data in the suspension system of this invention are fused by Kalman filtering, the attitude and driving status of the suspension system are fed back in real time; combined with the PID control algorithm, the steering gear can respond to terrain changes within 0.5 seconds, automatically adjust the suspension height (±10cm) and damping coefficient, ensure the stability of the equipment when driving on slopes and potholes, and improve the passability by 70%. Attached Figure Description

[0037] Figure 1 This is a schematic diagram of the overall structure of the intelligent dust collection and spraying robot system for factories based on multi-sensor collaboration according to the present invention; Figure 2 This is a schematic diagram of a chassis battery pack in the system of the present invention; Figure 3 This is a schematic diagram of the structure of the visual recognition module in the system of the present invention; Figure 4 This is a schematic diagram of the structure of the dust collection module in the system of the present invention; Figure 5 This is a schematic diagram of the dry fog module in the system of the present invention; Figure 6 This is a schematic diagram of the structure of the electronic control system in the system of the present invention; Figure 7This is a schematic diagram of a suspension system in the system of the present invention; Figure 8 This is a schematic diagram of the structure of the automatic supply bin in the system of the present invention; Figure 9 This is a schematic diagram of a vehicle body frame in the system of the present invention.

[0038] In the diagram: 1. Vehicle body; 2. Sweeper; 3. Charging port; 4. Frame; 5. Battery pack; 6. Terminal; 7. Front camera; 8. Rear camera; 9. Left camera; 10. Right camera; 11. Collision avoidance radar; 12. Infrared sensor; 13. Dust concentration sensor; 14. Electronic control system; 15. Dust collection pipe inlet; 16. Primary filter chamber; 17. Secondary filter chamber; 18. Tertiary filter chamber; 19. Exhaust port; 20. Support column; 21. Water tank; 22. Spray host; 23. Water pump; 24. Nozzle; 25. Water outlet; 26. Water inlet; 27. Float; 28. Water level probe; 29. ​​Steering gear; 30. Lateral tie rod; 31. Shock absorber; 32. Upper fork arm; 33. Lower fork arm; 34. Wheel hub; 35. Swing bearing; 36. Water inlet; 37. Solenoid valve water inlet switch; 38. Charging base; 39. Display screen. Detailed Implementation

[0039] To more clearly illustrate the embodiments of the present invention, specific implementation methods will be described below with reference to the accompanying drawings. Obviously, the drawings described below are merely some embodiments of the present invention. For those skilled in the art, other drawings and other implementation methods can be obtained based on these drawings without any creative effort.

[0040] like Figures 1 to 6 As shown, the present invention provides a factory intelligent dust collection spray robot system based on multi-sensor collaboration, including a vehicle body 1, a sweeper 2 installed on the vehicle body, an electronic control system 14 installed inside the vehicle body, a vision recognition module installed on the vehicle body, a dust collection module installed on the vehicle body, and a dry fog module installed on the vehicle body; the vehicle body is also provided with a charging interface 3.

[0041] The electronic control system is used to control the operation of the vision recognition module, the dust collection module, and the dry fog module;

[0042] The visual recognition module is used to construct maps and optimize dynamic paths in real time.

[0043] The dust collection module is used to absorb dust through a graded filtration process;

[0044] The dry fog module is used to dynamically adjust the atomization particle size and spray flow rate of the dry fog nozzle according to the dust concentration to resist dust diffusion.

[0045] Specifically, such as Figure 9As shown, the chassis 4 of the vehicle body is made of steel.

[0046] Specifically, such as Figure 2 and Figure 6 As shown, the electronic control system includes a vehicle control unit (VCU), a battery management system (BMS), an on-board power supply, a motor controller (MCU), a drive motor, and a reducer assembly. The BMS and MCU are both electrically connected to the VCU. The on-board power supply is electrically connected to the BMS. The drive motor is electrically connected to the MCU. The reducer assembly is electrically connected to the drive motor. The VCU receives input signals (switch inputs, including power-on signals, charging switch signals, braking signals, power-off signals, and spray signals) and controls the operation of the vision recognition module, dust collection module, and dry fog module via CAN communication. Additionally, a battery pack 5 is installed inside the vehicle body (the battery pack mainly supplies power to the on-board power supply and the drive motor), and the battery pack has two terminals 6.

[0047] Specifically, such as Figure 3 As shown, the visual recognition module includes a front camera 7, a rear camera 8, a left-side camera 9, and a right-side camera 10 mounted on the vehicle body; a collision avoidance radar 11 mounted on the front of the vehicle body; and infrared sensors 12 and dust concentration sensors 13 mounted on either side of the front cameras. The infrared sensors, dust concentration sensors, and each camera are electrically connected to the electronic control system. The visual recognition module deeply integrates SLAM mapping with the Dijkstra path planning algorithm, achieving centimeter-level high-precision map construction and real-time dynamic path optimization through multimodal perception of LiDAR and sensor data. Compared to traditional solutions, path planning efficiency is improved by 40%, allowing for flexible avoidance of dynamic obstacles, significantly reducing invalid paths and repetitive dust collection operations, and increasing dust collection coverage to over 98%.

[0048] The real-time mapping process of SLAM is dynamically combined with Dijkstra's path planning, enabling robots / agents to optimize paths in real time while exploring unknown environments, rather than waiting for a complete map to be built before planning. This fusion significantly improves navigation efficiency, real-time performance, and adaptability in dynamic environments.

[0049] (1) Incremental SLAM and dynamic map update

[0050] Traditional SLAM (such as LiDAR SLAM and visual SLAM) typically builds a global map first and then uses it for path planning.

[0051] Improvement of this invention: Incremental SLAM (such as LOAM, Cartographer) is adopted to update the local map in real time during robot movement and dynamically expand the global map.

[0052] Key point: Map data (such as occupancy grid maps) is fed into the Dijkstra algorithm in a streaming manner, rather than waiting for the complete map to be built.

[0053] (2) Dynamic Dijkstra: Progressive path optimization based on partially known maps

[0054] Traditional Dijkstra's algorithm requires a complete map, but it cannot be directly applied in unknown environments.

[0055] Improvements in this invention:

[0056] Local Dijkstra: Run Dijkstra within the currently known area constructed by SLAM to calculate the shortest optimal path (e.g., the optimal route within 10 meters in front of the robot).

[0057] Global Dijkstra (progressive): As SLAM continuously expands the map, the Dijkstra algorithm dynamically adjusts the global path (e.g., from point A to point B, the path is re-optimized as new map information is added).

[0058] Hybrid strategy: Combining front-end local planning (such as D Lite) and back-end global Dijkstra's algorithm to achieve a balance between fast response and global optimization.

[0059] (3) SLAM-Dijkstra collaborative decision-making mechanism

[0060] The trade-off between exploration and navigation:

[0061] If the robot has no idea of ​​the target direction (e.g., a completely unknown environment), SLAM prioritizes map building, while Dijkstra does not plan (or only based on local information).

[0062] If the robot knows the target location (e.g., SLAM has partially mapped and identified the target point), Dijkstra calculates the optimal path in real time based on the currently known map and adjusts it dynamically (e.g., replanning when encountering new obstacles).

[0063] Dynamic obstacle handling:

[0064] SLAM detects new obstacles in real time and updates the map, and Dijkstra immediately recalculates the path (such as avoiding newly appearing obstacles).

[0065] The visual recognition module adopts a standardized interface design, supporting plug-and-play functionality for various types of sensors such as LiDAR and industrial cameras. The algorithm framework reserves interfaces for rapid integration of advanced detection models such as the YOLO series and Transformer, facilitating algorithm iteration and upgrades for different dust environments (such as metal dust and fiber dust) and reducing secondary development costs.

[0066] Furthermore, such as Figure 1 and Figure 4 As shown, the dust collection module includes a dust collection pipe inlet 15, a primary filter chamber 16, a secondary filter chamber 17, a tertiary filter chamber 18, and an exhaust port 19; the dust collection pipe inlet is connected to the cleaner; the dust collection pipe inlet, the primary filter chamber, the secondary filter chamber, the tertiary filter chamber, and the exhaust port are connected in sequence; a support column 20 is provided below the primary filter chamber, the secondary filter chamber, and the tertiary filter chamber.

[0067] The primary filtration chamber is a gravity settling chamber used to separate large dust particles; the secondary filtration chamber is a filter bag filtration area used to filter submicron particles using PTFE membrane filter media; the tertiary filtration chamber is an activated carbon filtration area used to adsorb oily and odorous dust; a differential pressure sensor is also installed in the filter bag filtration area.

[0068] The three-stage filtration system forms a progressive purification chain of "coarse separation - fine filtration - deep purification". The gravity settling chamber uses the inertia of dust gravity to separate large dust particles, the filter bag filtration zone uses PTFE membrane filter material to achieve submicron particle interception, and the activated carbon filtration zone adsorbs oily and odorous dust, ultimately resulting in an emission dust concentration of less than 5mg / m³. 3 It far exceeds national environmental protection standards.

[0069] A differential pressure sensor is installed in the filter bag filtration area to monitor the filter bag clogging status in real time; the lifespan of the filter element is predicted by AI algorithm, and maintenance reminders are automatically pushed when the degree of clogging or the usage time reaches the threshold, avoiding secondary dust caused by filter element failure and reducing maintenance costs by 30%.

[0070] The process of predicting filter lifespan using AI algorithms is as follows:

[0071] 1. Hardware layer: Deployment of differential pressure sensors and data acquisition.

[0072] (1) Differential pressure sensor installation location

[0073] Upstream pressure point (P1): Installed on the air inlet side (unfiltered airflow side) of the filter bag filtration zone.

[0074] Downstream pressure point (P2): Installed on the outlet side (filtered airflow side) of the filter bag filtration zone.

[0075] Pressure difference calculation: ΔP = P1-P2 (unit: Pa or mmH2O).

[0076] (2) Real-time monitoring data

[0077] Sampling frequency: 1Hz-10Hz (adjusted according to working conditions, such as increasing the sampling rate in high dust environments).

[0078] Data storage: Edge computing devices (such as industrial gateways) store differential pressure data in real time and upload it to the cloud / AI analysis platform.

[0079] (3) Setting of abnormal threshold

[0080] Normal operating differential pressure range: ΔP_normal (e.g., 500-1500Pa, depending on the filter bag type).

[0081] Clogging warning threshold: ΔP_warning (e.g., ΔP > 1800Pa indicates that the filter bag is about to become clogged).

[0082] Emergency shutdown threshold: ΔP_critical (e.g., if ΔP > 2500Pa, the system will automatically shut down or switch to a backup filter bag).

[0083] 2. AI Algorithm Layer: Filter Cartridge Life Prediction Model.

[0084] (1) Key input features (used for AI training), as shown in the table below: Table 1 Key Input Feature Data for AI Model Training

[0085] (2) AI model selection

[0086] LSTM (Long Short-Term Memory Network): Pressure difference time series prediction (long-term dependency) for trend analysis.

[0087] (3) Sources of training data

[0088] Historical data backtracking: Automatically extracts differential pressure changes and replacement records of filter bags.

[0089] Experimental data: The filter bag clogging process under different dust concentrations and wind speeds was simulated in the laboratory, and data on the change of ΔP over time were collected.

[0090] Online learning: AI models are continuously optimized, and prediction accuracy is constantly adjusted as new data is input.

[0091] (4) Predicted output

[0092] Remaining Useful Life (RUL): e.g., "The current filter bag can still be used for 720 hours".

[0093] Blockage trend prediction: The pressure difference change curve for the next 24 / 72 hours will help determine whether early maintenance is needed.

[0094] Maintenance recommendations:

[0095] If the system is in normal condition (ΔP is stable, RUL > 1000h), it will continue to operate.

[0096] If a warning status is in effect (ΔP rises rapidly, RUL < 300h), a replacement plan is proposed.

[0097] In an emergency (ΔP exceeds the threshold, RUL < 50h), the machine must be shut down and replaced immediately.

[0098] Furthermore, such as Figure 5 As shown, the dry fog module includes a water tank 21, a spray host 22, a water pump 23, and nozzles 24. The water tank is connected to the spray host; the water pump is located inside the spray host; the water pump is connected to the water outlet 25 on the spray host; a nozzle is provided at the water outlet; the water tank has an inlet 26, and a float 27 and a water level probe 28 are also provided inside the water tank. The dry fog module constructs a dust concentration-humidity dual closed-loop feedback control system. The humidity probe monitors the ambient humidity in real time, and the dust concentration probe collects PM2.5 / PM10 data at a frequency of seconds. Based on a fuzzy control algorithm, the system can dynamically adjust the atomization particle size (5-20μm) and spray flow rate of the dry fog nozzles according to the dust concentration, increasing the dust suppression efficiency to 90% and reducing water consumption by 60% compared to traditional spray systems. The specific process is as follows:

[0099] 1. Multi-dimensional environmental perception layer.

[0100] Dust concentration monitoring: A laser scattering dust sensor (accuracy ±5%) is used to collect dust concentration data in the range of 0.1-1000mg / m³ in real time.

[0101] Particle size distribution analysis: The median diameter of dust (such as the PM2.5 / PM10 ratio) is obtained by using a particle counter.

[0102] Environmental parameter compensation: Synchronously collect key parameters that affect droplet condensation, such as temperature and humidity.

[0103] 2. Fuzzy control decision engine.

[0104] Input variable fuzzification:

[0105] Dust concentration (low / medium / high / extremely high, corresponding to 0-30 / 30-100 / 100-300 / 300+ mg / m³);

[0106] Dust particle size (fine / medium / coarse, corresponding to <2.5 / 2.5-10 / >10μm percentage);

[0107] Ambient humidity (dry / moderate / humid, corresponding to <40% / 40-70% / >70%RH);

[0108] Output variable adjustment:

[0109] Atomized particle size (5-20μm continuously adjustable, prioritizing matching dust particle size);

[0110] Injection flow rate (0-100% graded adjustment, minimum adjustment step 1%).

[0111] 3. Precise control by the implementing agency.

[0112] Atomized particle size adjustment: The nozzle air pressure (0.2-0.6MPa) is adjusted in real time by a piezoelectric ceramic micro-valve (response time <10ms).

[0113] Flow control: A proportional valve driven by a servo motor (accuracy ±1%) is used to achieve continuous and adjustable flow.

[0114] Adaptive spray angle: dynamically adjusts the spray coverage area in conjunction with the gimbal mechanism.

[0115] In addition, by combining meteorological data (wind speed, wind direction) with operational conditions, the dry fog module can automatically adjust the dry fog coverage area. When strong winds are detected, the dry fog spray volume is increased on the windward side first, forming an "air curtain barrier" to effectively resist dust diffusion and ensure the stability of dust suppression effect.

[0116] Furthermore, such as Figure 7 As shown, the vehicle body includes a suspension system; the suspension system includes an angle sensor, a speed sensor, a steering gear 29, a tie rod 30, a shock absorber 31, an upper wishbone 32, a lower wishbone 33, a wheel hub 34, and a swing bearing 35; the tie rod is connected to the steering gear; the upper and lower wishbones are connected via the swing bearing; the wheel hub is mounted on the swing bearing; the lower end of the shock absorber is connected to the lower wishbone, and the upper end is connected to the upper wishbone. The angle sensor collects the steering angle of the steering gear to determine the steering amplitude and direction; the speed sensor collects the movement speed of the suspension system to provide a basis for control system adjustments.

[0117] After the angle and speed sensor data are fused using Kalman filtering (by establishing a state-space model, calculating dynamic weights, and performing adaptive noise adjustment), the suspension system's attitude and driving status are fed back in real time. Combined with a PID control algorithm, the steering gear can respond to terrain changes within 0.5 seconds, automatically adjusting the suspension height (±10cm) and damping coefficient to ensure the stability of the equipment when driving on slopes and uneven roads, improving passability by 70%.

[0118] The precise control process of the PID control algorithm is as follows:

[0119] 1. Control objective:

[0120] When the terrain undulation is detected to exceed the preset threshold (e.g., ΔH > 5cm), the suspension height is adjusted to the target value (error < ±2cm) within 0.5 seconds, and the damping coefficient is optimized simultaneously to suppress vehicle body sway.

[0121] 2. Parameter Design:

[0122] Proportional (P): Rapid response to terrain deviations, such as a 5cm suspension adjustment for every 1cm height error, ensuring instantaneous response speed.

[0123] Integral term (I): Eliminates long-term accumulated errors (such as suspension offset caused by continuous bumps), and achieves accurate positioning by integrating the accumulated errors.

[0124] Differential term (D): Predicts the trend of terrain changes (e.g., adjusts damping in advance based on dH / dt) and suppresses overshoot (e.g., avoids excessive uplift leading to secondary shocks).

[0125] In addition, a health model for the suspension system can be established. Based on sensor data and a historical failure case library, machine learning algorithms can be used to predict potential failures such as bearing wear and hydraulic leakage, and issue early warnings 3-5 days in advance to reduce production losses caused by sudden failures and increase equipment availability to over 95%.

[0126] Furthermore, such as Figure 8 As shown, the system of the present invention also includes an automatic replenishment bin for completing water and electricity connection and dust transfer; the automatic replenishment bin has a built-in weighing sensor and a liquid level sensor; the automatic replenishment bin is equipped with a millimeter-wave radar; the automatic replenishment bin is also equipped with a water inlet 36, a solenoid valve water replenishment switch 37, a charging base 38, and a display screen 39. The display screen has a touch-screen interactive function, allowing commands to be sent to the vehicle controller through the operating interface, and providing real-time feedback on the operation status of the multi-sensor collaborative intelligent dust collection spray robot system in the factory.

[0127] The automatic replenishment bin integrates millimeter-wave radar, weighing sensors, and liquid level sensors to achieve non-contact and accurate monitoring of power, water, and dust collection bin capacity. Employing edge computing technology, it analyzes remaining resources and operational needs in real time, and uses path planning algorithms to trigger replenishment commands 20% in advance, preventing downtime due to resource depletion.

[0128] The real-time advantages of edge computing, dynamic resource demand prediction, and forward-looking replenishment triggering mechanism (20% margin in advance) are deeply integrated to form a closed-loop optimization system of "perception-analysis-decision-execution", breaking through the limitations of passive response in traditional resource management.

[0129] (1) Edge computing-driven real-time resource-job collaborative analysis

[0130] Low-latency data processing: Edge nodes are deployed close to the factory production line to collect environmental data (such as dust concentration levels in various work areas) and operational requirements (task priorities) in real time. A lightweight model dynamically calculates the remaining time that resources can support, reducing analysis latency from seconds in the cloud center to milliseconds. For irregular operations in the factory (such as scenarios where dust concentration exceeds the standard), reinforcement learning is used to dynamically adjust resource thresholds.

[0131] (2) Dynamic margin feedforward control strategy (triggering 20% ​​in advance)

[0132] Nonlinear safety margin design: Based on historical downtime data analysis (such as low power or low water), Monte Carlo simulation is used to determine "20% margin" as the optimal balance point (which avoids over-replenishment and waste, and covers more than 95% of sudden demand scenarios).

[0133] Linked path planning algorithms: When it is predicted that the remaining resources are below the threshold, the edge computing model calls the localized algorithm to calculate the optimal supply path in real time (considering the path congestion in the work area and the time characteristics of peak and off-peak operations in the workshop), and generates a supply instruction with time window constraints (e.g., "Water needs to be replenished before 12:30").

[0134] (3) Closed-loop feedback and continuous optimization

[0135] Digital twin calibration: Edge nodes synchronize actual resource consumption data to the cloud digital twin, and reversely optimize the parameters of the prediction model (such as adjusting the time step of LSTM or the reward function of reinforcement learning), forming a hierarchical intelligent architecture of "real-time edge control + long-term cloud learning".

[0136] Based on visual recognition and laser navigation, the automated replenishment bin can automatically align with equipment interfaces to complete water and electricity connections and dust transfer. It supports parallel replenishment scheduling of multiple devices, coordinates the operation sequence through an IoT platform, improves replenishment efficiency by 50%, and reduces the risk of manual intervention.

[0137] Based on the technical solution of this invention, and through the following case scenarios, the implementation process of this invention in practical applications is illustrated. The specific application implementation scheme is as follows:

[0138] 1. The basic setup and installation of the automatic supply bins are completed, enabling power and water supply.

[0139] Power supply system:

[0140] Circuit planning: Based on the equipment layout and power requirements of the automated replenishment bins, use professional circuit design software (such as AutoCAD Electrical) to draw circuit diagrams, clearly defining the main power lines and branch lines to ensure stable power supply to each device and compliance with safety regulations. Reserve 10%-20% of redundant power to meet future equipment expansion needs.

[0141] Equipment installation: Install distribution boxes, circuit breakers, residual current devices and other equipment. Select cables that meet industrial-grade standards and lay the lines using cable trays or conduits to ensure good insulation performance. Ensure electrical safety through grounding devices. After completion, perform insulation resistance tests and power-on tests.

[0142] Intelligent monitoring: Deploy a power monitoring system, install smart meters and current / voltage sensors to monitor the power consumption of each device in real time, and realize remote monitoring and abnormal alarm through the Internet of Things platform to facilitate timely detection and handling of power supply faults.

[0143] Water supply system:

[0144] Pipeline installation: Based on the distribution of water supply points in the supply warehouse, PPR or stainless steel pipes will be used for water supply pipeline installation, with a reasonable water flow path designed to ensure stable water pressure. Valves and water meters will be installed at key points to facilitate water consumption control and measurement.

[0145] Water purification and storage equipment: Pre-filters and reverse osmosis water purifiers are installed to ensure water quality meets usage requirements. A water storage tank with a water level sensor is provided, and automatic water replenishment is achieved through a PLC control system to maintain the water level within a reasonable range.

[0146] Drainage treatment: Plan drainage pipes, install drainage facilities such as floor drains and drainage ditches, and introduce sewage into the factory's sewage treatment system. For special water (such as cleaning water containing chemicals), pretreatment is required before discharge.

[0147] 2. Data collection and labeling of factory equipment and facilities samples

[0148] Data collection:

[0149] Equipment data: Industrial sensors (such as vibration sensors, temperature sensors, pressure sensors, etc.) are used to collect equipment operating parameters, including speed, temperature, pressure, vibration frequency, etc. The data is transmitted to a local server via a PLC controller or data acquisition card. The sampling frequency is set according to the characteristics of the equipment (e.g., 100 times per second for high-frequency vibration equipment).

[0150] Image data: Use industrial cameras (resolution not less than 1920×1080) to capture images of the equipment's appearance, operating status, and facility details from multiple angles, covering different scenarios such as normal operating conditions and fault conditions, with no less than 1,000 images collected for each piece of equipment or facility.

[0151] Video data: Record video of the complete operating cycle of the device to capture the dynamic process of the device's operation. The video frame rate is set to 25fps or higher, and the storage format is common formats such as MP4.

[0152] Data annotation:

[0153] Tool selection: Use professional annotation tools (such as LabelImg, CVAT) to annotate image and video data. The annotation content includes the location of key equipment components, fault type, facility structural features, etc.

[0154] Labeling Standards: Establish unified labeling standards and category systems, such as categorizing equipment failures into "bearing wear" and "pipeline leakage," to ensure the accuracy and consistency of labeling. After labeling is completed, conduct cross-checking and review to reduce labeling errors.

[0155] 3. Set up YOLOv5 and complete the training and validation of the model.

[0156] Environment setup:

[0157] Install Python 3.8 or later, and configure the CUDA and cuDNN environments to support GPU acceleration (if using a GPU).

[0158] Install deep learning frameworks such as PyTorch and torchvision, as well as the dependency libraries required by YOLOv5 (such as numpy, opencv-python, etc.) via pip.

[0159] Clone the YOLOv5 code from the official GitHub repository and modify the configuration files (such as data paths and model structure parameters) according to your actual needs.

[0160] Data preparation:

[0161] The collected and labeled data is divided into training set (70%), validation set (20%) and test set (10%), and organized according to the YOLOv5 data format to generate the corresponding dataset configuration file (such as data.yaml).

[0162] Model training:

[0163] Choose a suitable YOLOv5 model version (such as YOLOv5s, YOLOv5m) and set the training hyperparameters, including the learning rate (initial value 0.001), batch size (adjust according to GPU memory, such as 16), and number of training epochs (usually 100-300 epochs).

[0164] Start the training script to monitor metrics such as loss function (classification loss, regression loss) and mAP (mean average accuracy) in real time during the training process, and save the optimal model weight file.

[0165] Model validation:

[0166] The trained model is evaluated using a validation set. Metrics such as mAP, precision, and recall are calculated to analyze the model's detection performance in different categories and scenarios. A confusion matrix is ​​plotted to identify false positives and false negatives in the model, providing a basis for subsequent optimization.

[0167] 4. Deploy the model on the control system

[0168] Hardware compatibility:

[0169] Evaluate the performance of the control system hardware (such as CPU and GPU computing power, memory capacity) and select an appropriate deployment method. If hardware resources are limited, lightweight models can be used or model quantization (such as INT8 quantization) can be performed to reduce computational requirements.

[0170] For edge computing devices (such as the NVIDIA Jetson series), install the corresponding version of the TensorRT inference acceleration engine to improve model inference speed.

[0171] Software deployment:

[0172] Convert the trained model weight file into a format suitable for deployment (such as the .engine format of ONNX and TensorRT).

[0173] Integrate inference frameworks (such as OpenVINO and TensorRT) into the control system, and write interface programs to implement model loading, data preprocessing, inference calculation and result postprocessing, so as to ensure seamless integration between the model and other modules of the control system (such as PLC and sensors).

[0174] Testing and optimization:

[0175] Perform functional tests on the deployed model in a real-world operating environment to check the accuracy and real-time performance of the model's inference results.

[0176] Based on test feedback, we optimized model inference parameters (such as confidence threshold and NMS threshold), adjusted data transmission and processing procedures, and further improved the overall system performance.

[0177] 5. SLAM Mapping

[0178] Sensor selection and installation:

[0179] Select a combination of lidar (such as 16-line or 32-line lidar) and inertial measurement unit (IMU) to ensure that the sensor installation position can fully cover the working area. The lidar must be installed horizontally, and the IMU must be aligned with the lidar coordinate system.

[0180] SLAM algorithm implementation:

[0181] Choose a suitable SLAM algorithm (such as Cartographer or LOAM) and configure and run it in the ROS (Robot Operating System) environment.

[0182] The system uses LiDAR to collect environmental point cloud data, IMU to provide attitude information, and SLAM algorithm to build an environmental map in real time, while calculating the device's own positioning information.

[0183] Map optimization and management:

[0184] The generated map is optimized by removing noise points, smoothing the map surface, and improving map accuracy.

[0185] Establish a map storage and update mechanism to update the map in a timely manner when the working environment changes (such as the addition of equipment or facilities) to ensure that the map is consistent with the actual environment.

[0186] 6. Startup and Model Parameter Tuning

[0187] System startup and monitoring: (1) Start the power supply and water supply system of the automatic replenishment bin.

[0188] (2) Press the power button to start the "Multi-sensor collaborative intelligent dust collection spray robot system".

[0189] (3) Send the “SLAM mapping” command to the “multi-sensor collaborative intelligent dust collection spray robot system” through the display screen (with touch interaction function) of the automatic replenishment bin to complete the path data collection.

[0190] (4) Send a “model training” instruction to the “multi-sensor collaborative factory intelligent dust collection spray robot system” through the display screen (with touch interaction function) of the automatic replenishment bin.

[0191] (5) The operating status and key parameters (such as equipment temperature and model inference time) of each module can be viewed in real time through the monitoring interface of the automatic replenishment bin's display screen (with touch interaction function).

[0192] (6) After completing the model parameter tuning through the display screen (with touch interaction function) of the automatic supply bin, save the model training version and select to enable it.

[0193] (7) Send automatic operation instructions to the "Multi-sensor collaborative intelligent dust collection spraying robot system" through the display screen (with touch interaction function) of the automatic replenishment bin. The "Multi-sensor collaborative intelligent dust collection spraying robot system" automatically performs dust collection spraying, charging and water replenishment operations according to the path planning.

[0194] Model parameter tuning:

[0195] During system operation, actual detection data is collected, and the deviation between the model's detection results and the actual situation is analyzed.

[0196] Adjusting model hyperparameters (such as learning rate and anchor box size) or retraining the model (by adding data from specific scenarios) can gradually improve the model's detection accuracy and generalization ability.

[0197] System optimization and iteration:

[0198] Based on system operation feedback, optimize the collaborative workflow between modules and resolve compatibility issues and performance bottlenecks that occur during operation.

[0199] Regularly maintain and upgrade the system, introduce new technologies and algorithms, and keep the system advanced and reliable.

[0200] The above description is merely a detailed explanation of preferred embodiments and principles of the present invention. For those skilled in the art, there may be changes in specific implementation methods based on the ideas provided by the present invention, and these changes should also be considered within the scope of protection of the present invention.

Claims

1. A factory intelligent dust collection spray robot system based on multi-sensor collaboration, characterized in that, It includes the vehicle body, a sweeper installed on the vehicle body, an electronic control system installed inside the vehicle body, a vision recognition module installed on the vehicle body, a dust collection module installed on the vehicle body, and a dry fog module installed on the vehicle body. The electronic control system is used to control the operation of the vision recognition module, the dust collection module, and the dry fog module; The visual recognition module is used to construct maps and optimize dynamic paths in real time. The dust collection module is used to absorb dust through a graded filtration process; The dry fog module is used to dynamically adjust the atomization particle size and spray flow rate of the dry fog nozzle according to the dust concentration to resist dust diffusion.

2. The intelligent dust collection spray robot system for factories based on multi-sensor collaboration according to claim 1, characterized in that, The visual recognition module includes four cameras located on the front, rear, left, and right sides of the vehicle body, a collision avoidance radar located at the front of the vehicle body, and an infrared sensor and a dust concentration sensor located on both sides of the front camera. The infrared sensor, dust concentration sensor, and each camera are all electrically connected to the electronic control system.

3. The intelligent dust collection spray robot system for factories based on multi-sensor collaboration according to claim 2, characterized in that, The dust collection module includes a dust collection pipe inlet, a primary filter chamber, a secondary filter chamber, a tertiary filter chamber, and an exhaust port; the dust collection pipe inlet is connected to the cleaner; the dust collection pipe inlet, the primary filter chamber, the secondary filter chamber, the tertiary filter chamber, and the exhaust port are connected in sequence; support columns are provided below the primary filter chamber, the secondary filter chamber, and the tertiary filter chamber.

4. The intelligent dust collection spray robot system for factories based on multi-sensor collaboration according to claim 3, characterized in that, The primary filtration chamber is a gravity settling chamber used to separate large dust particles; the secondary filtration chamber is a filter bag filtration area used to filter submicron particles using PTFE membrane filter media; and the tertiary filtration chamber is an activated carbon filtration area used to adsorb oily and odorous dust. The filter bag is also equipped with a differential pressure sensor in the filtration zone to monitor the filter bag clogging status in real time.

5. The intelligent dust collection spray robot system for factories based on multi-sensor collaboration according to claim 4, characterized in that, The dry fog module includes a water tank, a spray host, a water pump, and nozzles; the water tank is connected to the spray host; the water pump is located inside the spray host; the water pump is connected to the water outlet on the spray host; a nozzle is provided at the water outlet; the water tank is provided with a water inlet, and a float and a water level probe are also provided inside the water tank.

6. The intelligent dust collection spray robot system for factories based on multi-sensor collaboration according to claim 1, characterized in that, The electronic control system includes a vehicle control unit (VCU), a battery management system (BMS), an on-board power supply, a motor controller (MCU), a drive motor, and a reducer assembly. The BMS and MCU are both electrically connected to the VCU. The on-board power supply is electrically connected to the BMS. The drive motor is electrically connected to the MCU. The reducer assembly is electrically connected to the drive motor. The VCU receives input signals and controls the operation of the vision recognition module, dust collection module, and dry fog module via CAN communication.

7. The intelligent dust collection spray robot system for factories based on multi-sensor collaboration according to claim 1, characterized in that, The vehicle body includes a suspension system; the suspension system includes an angle sensor, a speed sensor, a steering gear, a tie rod, a shock absorber, an upper wishbone, a lower wishbone, a wheel hub, and a swing bearing; the tie rod is connected to the steering gear; the upper and lower wishbones are connected via the swing bearing; the wheel hub is mounted on the swing bearing; the lower end of the shock absorber is connected to the lower wishbone, and the upper end of the shock absorber is connected to the upper wishbone.

8. The intelligent dust collection spray robot system for factories based on multi-sensor collaboration according to claim 1, characterized in that, It also includes an automatic replenishment bin for connecting water and electricity and transferring dust; the automatic replenishment bin has a built-in weighing sensor and a liquid level sensor; the automatic replenishment bin is equipped with a millimeter-wave radar; the automatic replenishment bin is also equipped with a water inlet, a solenoid valve water replenishment switch, a charging base and a display screen.

Images