A spraying control method based on job area dust data prediction

Abstract

This application relates to a water spraying control method based on dust data prediction of the work area. The method includes a water spraying control system that predicts dust levels based on work area data and weather variables, and then controls water trucks for precise water spraying control based on the dust levels. It predicts dust levels within a specified time period and calculates the corresponding water spraying demand based on the dust levels; determines the number of water trucks, spraying time periods, and spraying areas based on the water spraying demand; and controls the water trucks to spray water on the specified areas during the specified spraying time periods. This application improves the ability to fit dust generation patterns in complex work environments, thus providing a reliable decision-making basis for subsequent water spraying control. This accurate predictive capability effectively avoids the blindness of traditional fixed-frequency water spraying methods, ensuring that water spraying operations can respond in advance to potential dust peaks.

Description

Technical Field

[0001] This application relates to the technical field of mining operations, and in particular to a water spraying control method based on dust data prediction of the operating area. Background Technology

[0002] Water spraying in mining areas is a common dust control measure, widely used in open-pit mines, quarries, coal mines, and various other mining operations. The large amounts of dust generated during ore crushing, transportation, loading and unloading, and open-pit mining not only pollute the environment but also threaten the health of workers. Therefore, water spraying for dust suppression has become an important environmental protection method. The principle is to use high-pressure spraying equipment to evenly spray water mist onto the work area, causing dust particles to combine with the water mist, increase their weight, and settle, thereby reducing the concentration of suspended particulate matter in the air. Water spraying dust suppression systems are usually equipped with automatic control devices that can intelligently adjust the water volume and spraying frequency based on factors such as ambient humidity, wind speed, and dust concentration to achieve the best dust suppression effect. Furthermore, this technology has advantages such as simple operation, low cost, and wide applicability, and has been incorporated into the mining environmental protection standards of many countries and regions. However, water spraying for dust control also has problems such as high water consumption and easy freezing in cold environments. Therefore, in practical applications, it is necessary to combine it with other dust control technologies, such as dustproof netting, green isolation belts, and dust collection devices, in order to achieve comprehensive and efficient environmental management goals.

[0003] However, existing technologies cannot dynamically adjust watering strategies based on actual dust changes in the work area, resulting in water waste and unsatisfactory dust control effects. Summary of the Invention

[0004] To address the problem of water waste and unsatisfactory dust control results caused by dynamically adjusting watering strategies based on actual dust changes, this application provides a watering control method based on dust data prediction of the work area.

[0005] This application provides a water spraying control method based on dust data prediction of the work area, which adopts the following technical solution: A water spraying control method based on dust data prediction of the work area includes the following steps:

[0006] S1: Collect construction data, weather data, and regional characteristic data within the mining operation area to provide basic data for subsequent data processing and model training;

[0007] S2: Clean and standardize the collected data, remove outliers and missing values, unify the data format, and transmit the processed data to the model training module.

[0008] S3: Based on the preprocessed data, a hybrid neural network model is built using a long short-term memory network architecture combined with a convolutional neural network, and the model is trained. During the training process, hyperparameters are adjusted to optimize the model performance.

[0009] S4: Using the trained hybrid neural network model and real-time collected data, predict the amount of dust in a specified time period sequence, and send the prediction results to the water demand calculation module.

[0010] S5: Calculate the water demand based on the predicted dust volume, dust particle settling efficiency, water density, and water flow rate of the sprinkler equipment, providing data support for sprinkler resource scheduling.

[0011] S6: Determine the required number of sprinkler trucks by combining parameters such as water demand, sprinkler truck water tank capacity, working time, and actual watering efficiency. At the same time, determine the watering time period and watering area based on the spatiotemporal distribution of dust, meteorological conditions, and work activity arrangements.

[0012] S7: Based on the determined number of sprinkler trucks, sprinkler time period and sprinkler area, combined with real-time dust monitoring data and weather changes, dynamically adjust the sprinkler strategy, such as increasing the sprinkler flow and frequency when the dust volume is high and the wind speed is high, and reducing the sprinkler flow and frequency when the dust volume is low and the humidity is high.

[0013] S8: Based on the adjusted watering strategy, control the number of water trucks matching the number of water trucks to carry out watering operations on the designated area within the specified time period, and monitor the operation of the water trucks in real time.

[0014] S9: After the watering operation is completed, collect the actual dust volume data of the operation area, compare and analyze it with the predicted dust volume, optimize the parameters of the hybrid neural network model based on the analysis results, and improve the accuracy of subsequent dust volume prediction and the precision of watering control.

[0015] In a preferred embodiment, in step S1, laser scattering dust sensors, wind speed sensors, humidity sensors, and temperature sensors are installed in the mining operation area. Simultaneously, the CAN bus data interface built into the operating equipment is integrated. Every 5 minutes, excavator operating status parameters, including digging depth and frequency, truck transport frequency parameters, including hourly transport trips and average load, as well as wind speed, humidity, temperature, and air pressure data, are collected. Real-time data is transmitted to edge computing nodes via industrial Ethernet. Simultaneously, high-definition cameras capture images of the operation area, and the YOLOv5 algorithm is used to identify the operating status of the excavators and trucks. Daily 24-hour weather forecast data for the operation area, including wind direction, wind speed, humidity, and precipitation probability, is obtained from the meteorological department. Topographical features such as slope and vegetation cover are acquired through drone aerial photography combined with a GIS system. Dust-generating sub-regions are divided according to the mining area, transport road area, and stockpile area, and soil particle composition data, including particle size distribution and density characteristics, are collected for each sub-region.

[0016] In a preferred embodiment, in step S2, outliers in the construction data are detected and removed using the IQR method, missing values ​​are filled using linear interpolation, meteorological data such as wind speed, humidity, and temperature are standardized using Z-score, excavator working status parameters and truck transportation frequency data are converted into dimensionless indicators, the data format is unified as JSON, the time synchronization of data in each sub-region is ensured through a data verification mechanism, the processed data is stored in a distributed database, and transmitted to the model training module in real time through a message queue, data quality assessment indicators including data integrity, accuracy, and timeliness are established, and a data quality report is generated daily.

[0017] In a preferred embodiment, in step S3, a hybrid model of a long short-term memory network and a convolutional neural network is constructed based on preprocessed historical construction data, weather data, and regional feature data. The convolutional neural network part uses three convolutional layers to extract spatial features of the data, while the long short-term memory network part has two hidden layers, each containing 128 neurons. The Adam optimizer is used with an initial learning rate of 0.001, and an early stopping method is used to prevent overfitting. The learning rate is adjusted every 50 epochs during training. The terrain slope and soil particle composition in the regional feature data are used as weight parameters of the model. The influence weight of each sub-region in the model is dynamically adjusted through an attention mechanism. The root mean square error is used as the loss function of the model, and the ratio of the training set, validation set, and test set is set to 7:2:1. Training stops when the loss of the model on the validation set no longer decreases for 10 consecutive epochs.

[0018] In a preferred embodiment, in step S4, the collected historical construction data, weather data, and regional feature data are cleaned and standardized to remove outliers and missing values ​​and unify the data format. Using the historical construction data, weather data, and regional feature data as input and the actual monitored dust volume as output, the hybrid neural network model is trained. During training, the model's performance is optimized by adjusting its hyperparameters (such as learning rate, number of hidden layer neurons, etc.), improving the accuracy and stability of dust volume prediction. Furthermore, weight parameters for regional features are set in the hybrid neural network model, and the model's dust volume prediction strategy for each sub-region is adjusted based on factors such as geographical features and soil particle composition, achieving regionally differentiated dust volume prediction.

[0019] In a preferred embodiment, in step S5, the formula for calculating the water demand based on the predicted dust volume is as follows:

[0020] ;

[0021] Where Q is the water demand, D is the predicted dust volume, A is the area of ​​the work area, η is the settling efficiency of dust particles, which is related to the particle size distribution and settling characteristics of particles in the area, ρ is the density of water, and v is the water flow rate of the sprinkler equipment.

[0022] In a preferred embodiment, in step S6, the formula for calculating the number of sprinkler trucks N based on the water demand Q and the sprinkler truck parameters is as follows:

[0023] ;

[0024] Where N is the number of sprinkler trucks, C is the water tank capacity of the sprinkler trucks, T is the working time of the sprinkler trucks, ε is the actual sprinkling efficiency of the sprinkler trucks, and ⌈⋅⌉ indicates rounding up to ensure that the number of sprinkler trucks is sufficient to meet the demand;

[0025] Furthermore, the watering schedule is determined based on the spatiotemporal distribution of dust, meteorological conditions, and the arrangement of work activities.

[0026] Furthermore, the spatiotemporal distribution of dust, regional characteristics, water spraying radius and driving speed of the sprinkler truck are used to determine the water spraying area.

[0027] In a preferred embodiment, in step S7, based on the determined number of sprinkler trucks, sprinkler time period, and sprinkler area, the system collects real-time data on current dust concentration, wind speed, humidity, etc., of each sub-area within the work area and compares them with the predicted dust volume and preset thresholds. When the real-time dust concentration of a sub-area exceeds 15% of the predicted value and the current wind speed is higher than 3 m / s, the sprinkler flow rate of that area is automatically increased to 1.2 times the base value, and the sprinkler interval is shortened to 80% of the original plan. If the real-time humidity has reached 65% or higher and the dust concentration is lower than the predicted value, the sprinkler flow rate is reduced to 0.7 times the base value, and the sprinkler interval is extended. For areas with assigned sprinkler tasks, the system dynamically adjusts the driving path based on the real-time location data of the sprinkler trucks. When adjacent areas simultaneously trigger sprinkler demands, the system prioritizes dispatching the nearest available sprinkler truck and updates the task list synchronously through the vehicle network system. If the meteorological department issues a short-term gale warning (wind speed ≥8m / s) during the watering operation, the system will immediately suspend the watering operation in the open area and recalculate the amount of water to be sprayed based on the dust rebound model after the warning is lifted.

[0028] The dynamic sprinkler flow correction formula is:

[0029] ;

[0030] In the formula:

[0031] v t This represents the real-time sprinkler flow rate at time t;

[0032] v0 represents the basic sprinkler flow rate;

[0033] D t This represents the measured dust concentration at time t;

[0034] D p Indicates the predicted dust concentration;

[0035] H t Indicates the current ambient humidity;

[0036] H0 represents the reference humidity;

[0037] w t Indicates the current wind speed;

[0038] w0 represents the critical wind speed;

[0039] γ represents the dust response coefficient;

[0040] κ represents the humidity influence coefficient;

[0041] λ represents the wind speed correction factor.

[0042] In a preferred embodiment, in step S8, operation instructions are sent to each sprinkler truck according to the adjusted sprinkling strategy. The instructions include the coordinates of the sprinkling area, the sprinkling flow rate, the sprinkling duration, and the driving route. The on-board terminal receives the operation data of the sprinkler trucks in real time, including the actual sprinkling volume, sprinkling time, and operation completion status. The location and operation status of each sprinkler truck are displayed in real time on the monitoring platform. When a sprinkler truck malfunctions or the sprinkling volume is insufficient, a backup sprinkler truck is automatically dispatched. After the operation is completed, the sprinkler truck uploads an electronic work order including the actual sprinkling area and the sprinkling volume, and the system automatically generates a sprinkling operation report.

[0043] In a preferred embodiment, in step S9, the actual dust volume data of each sub-area is collected 1 hour after the watering operation is completed. The data is compared with the predicted dust volume to calculate the error rate. When the error rate exceeds 10%, the model optimization process is initiated. The training set is updated using the latest actual dust volume data and construction data. The hybrid neural network model is retrained, and the learning rate and the number of hidden layer neurons are adjusted. The regional feature weight parameters are optimized. The model is comprehensively evaluated once a month. The model architecture or training strategy is adjusted according to the evaluation results. The optimized model is deployed to the production environment and synchronized to each edge computing node. A model optimization log is established to record the parameter adjustments and effect evaluation results of each optimization.

[0044] In summary, this application includes at least one of the following beneficial technical effects:

[0045] 1. By constructing a hybrid model combining Long Short-Term Memory (LSTM) networks and convolutional neural networks, and introducing regional feature weight parameters, dust emission prediction can simultaneously capture temporal variation patterns and spatial distribution differences. During model training, dynamic adjustment of hyperparameters and attention mechanisms further enhances the fitting ability to dust generation patterns in complex operating environments, thus providing a reliable decision-making basis for subsequent sprinkler control. This accurate predictive capability effectively avoids the blindness of traditional fixed-frequency sprinkler methods, ensuring that sprinkler operations can respond in advance to potential dust peaks.

[0046] 2. When calculating water demand based on prediction results, the system comprehensively considers the dust particle settling characteristics of the work area and equipment parameters, achieving quantitative allocation of water resources. By dynamically adjusting the watering strategy in conjunction with real-time monitoring data and weather changes, the system can promptly increase watering intensity when dust levels exceed expectations and automatically reduce watering when environmental conditions are favorable for dust suppression. This dynamic adjustment mechanism ensures effective dust control while avoiding ineffective water consumption, significantly improving the overall economic efficiency of watering operations.

[0047] 3. By continuously collecting and comparing actual dust data with predicted values, the model parameters and regional feature weights are constantly optimized, forming a closed-loop adaptive improvement mechanism. During watering operations, intelligent scheduling of water trucks is achieved through vehicle networking and GPS positioning, which can adjust driving routes and operation sequences according to real-time needs, improving equipment utilization. Simultaneously, the emergency response mechanism for sudden weather changes further ensures the safety and effectiveness of watering operations, making the entire system robust enough to handle complex working conditions. Attached Figure Description

[0048] Figure 1 This is a schematic diagram illustrating the process principle of this application. Detailed Implementation

[0049] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0050] The following is in conjunction with the appendix Figure 1 This application will be described in further detail.

[0051] Example:

[0052] See Figure 1 A water spraying control method based on dust data prediction of the work area, the method includes the following steps:

[0053] S1: Collect construction data, weather data, and regional characteristic data within the mining operation area to provide basic data for subsequent data processing and model training;

[0054] S2: Clean and standardize the collected data, remove outliers and missing values, unify the data format, and transmit the processed data to the model training module.

[0055] S3: Based on the preprocessed data, a hybrid neural network model is built using a long short-term memory network architecture combined with a convolutional neural network, and the model is trained. During the training process, hyperparameters are adjusted to optimize the model performance.

[0056] S4: Using the trained hybrid neural network model and real-time collected data, predict the amount of dust in a specified time period sequence, and send the prediction results to the water demand calculation module.

[0057] S5: Calculate the water demand based on the predicted dust volume, dust particle settling efficiency, water density, and water flow rate of the sprinkler equipment, providing data support for sprinkler resource scheduling.

[0058] S6: Determine the required number of sprinkler trucks by combining parameters such as water demand, sprinkler truck water tank capacity, working time, and actual watering efficiency. At the same time, determine the watering time period and watering area based on the spatiotemporal distribution of dust, meteorological conditions, and work activity arrangements.

[0059] S7: Based on the determined number of sprinkler trucks, sprinkler time period and sprinkler area, combined with real-time dust monitoring data and weather changes, dynamically adjust the sprinkler strategy, such as increasing the sprinkler flow and frequency when the dust volume is high and the wind speed is high, and reducing the sprinkler flow and frequency when the dust volume is low and the humidity is high.

[0060] S8: Based on the adjusted watering strategy, control the number of water trucks matching the number of water trucks to carry out watering operations on the designated area within the specified time period, and monitor the operation of the water trucks in real time.

[0061] S9: After the watering operation is completed, collect the actual dust volume data of the operation area, compare and analyze it with the predicted dust volume, optimize the parameters of the hybrid neural network model based on the analysis results, and improve the accuracy of subsequent dust volume prediction and the precision of watering control.

[0062] In step S1, laser scattering dust sensors, wind speed sensors, humidity sensors, and temperature sensors are installed in the mining operation area. Simultaneously, the CAN bus data interface built into the operating equipment is integrated. Every 5 minutes, excavator operating status parameters, including digging depth and frequency, truck transport frequency parameters, including hourly transport trips and average load, as well as wind speed, humidity, temperature, and air pressure data, are collected. Real-time data is transmitted to the edge computing node via industrial Ethernet. High-definition cameras capture images of the operation area, and the YOLOv5 algorithm is used to identify the operating status of the excavators and trucks. Daily 24-hour weather forecast data, including wind direction, wind speed, humidity, and precipitation probability, is obtained from the meteorological department. Topographical features such as slope and vegetation cover are acquired through drone aerial photography combined with a GIS system. Dust-generating sub-regions are divided according to the mining area, transport road area, and stockpile area, and soil particle composition data, including particle size distribution and density characteristics, are collected for each sub-region.

[0063] In step S2, outliers in the construction data are detected and removed using the IQR method, missing values ​​are filled using linear interpolation, meteorological data such as wind speed, humidity, and temperature are standardized using Z-score, excavator working status parameters and truck transportation frequency data are converted into dimensionless indicators, the data format is unified to JSON, the time synchronization of data in each sub-region is ensured through a data verification mechanism, the processed data is stored in a distributed database, and transmitted to the model training module in real time through a message queue. Data quality assessment indicators are established, including data integrity, accuracy, and timeliness, and a data quality report is generated daily.

[0064] In step S3, based on the preprocessed historical construction data, weather data, and regional feature data, a hybrid model of Long Short-Term Memory (LSTM) network and Convolutional Neural Network (CNN) is constructed. The CNN part uses three convolutional layers to extract spatial features of the data, while the LSTM part has two hidden layers, each containing 128 neurons. The Adam optimizer is used with an initial learning rate of 0.001. Early stopping is used to prevent overfitting. The learning rate is adjusted every 50 epochs during training. The terrain slope and soil particle composition in the regional feature data are used as the model's weight parameters. The influence weight of each sub-region in the model is dynamically adjusted through an attention mechanism. The root mean square error is used as the model's loss function. The ratio of the model's loss on the training set, validation set, and test set is set to 7:2:1. Training stops when the model's loss on the validation set no longer decreases for 10 consecutive epochs.

[0065] In step S4, the collected historical construction data, weather data, and regional feature data are cleaned and standardized to remove outliers and missing values ​​and unify the data format. Using these data as input and the actual monitored dust levels as output, the hybrid neural network model is trained. During training, the model's hyperparameters (such as learning rate and the number of neurons in the hidden layer) are adjusted to optimize performance and improve the accuracy and stability of dust level prediction. Furthermore, weight parameters for regional features are set in the hybrid neural network model, and the model's dust level prediction strategy for each sub-region is adjusted based on factors such as geographical features and soil particle composition, achieving differentiated dust level prediction across regions.

[0066] In step S5, the formula for calculating the water demand based on the predicted dust volume is as follows:

[0067] ;

[0068] Where Q is the water demand, D is the predicted dust volume, A is the area of ​​the work area, η is the settling efficiency of dust particles, which is related to the particle size distribution and settling characteristics of particles in the area, ρ is the density of water, and v is the water flow rate of the sprinkler equipment.

[0069] In step S6, the formula for calculating the number of sprinkler trucks N based on the water demand Q and the sprinkler truck parameters is as follows:

[0070] ;

[0071] Where N is the number of water trucks, C is the water tank capacity of the water truck, T is the working time of the water truck, ϵ is the actual water spraying efficiency of the water truck, and ⌈⋅⌉ indicates rounding up to ensure that the number of water trucks is sufficient to meet the demand.

[0072] Furthermore, the timing of watering is determined based on the spatial and temporal distribution of dust, meteorological conditions, and work schedules. For example, priority is given to times with higher dust levels and times with lower wind speeds and higher humidity, while watering is avoided during peak excavator or truck operation periods.

[0073] Furthermore, the spatiotemporal distribution of dust, regional characteristics, water spraying radius and driving speed of the sprinkler truck are used to determine the water spraying area.

[0074] In step S7, based on the determined number of sprinkler trucks, sprinkler time period, and sprinkler area, the system collects real-time data on current dust concentration, wind speed, and humidity in each sub-area within the work area and compares it with the predicted dust volume and preset thresholds. When the real-time dust concentration in a sub-area exceeds 15% of the predicted value and the current wind speed is higher than 3 m / s, the system automatically increases the sprinkler flow rate in that area to 1.2 times the base value and shortens the sprinkler interval to 80% of the original plan. If the real-time humidity has reached 65% or higher and the dust concentration is lower than the predicted value, the system reduces the sprinkler flow rate to 0.7 times the base value and extends the sprinkler interval. For areas with assigned sprinkler tasks, the system dynamically adjusts the driving path based on the real-time location data of the sprinkler trucks. When adjacent areas simultaneously trigger sprinkler demand, the system prioritizes dispatching the nearest available sprinkler truck and updates the task list synchronously through the vehicle network system. During sprinkler operations, if the meteorological department issues a short-term gale warning (wind speed ≥ 8 m / s), the system immediately suspends sprinkler operations in open areas. After the warning is lifted, the system recalculates the replenishment amount based on the dust bounce model.

[0075] The dynamic sprinkler flow correction formula is:

[0076] ;

[0077] In the formula:

[0078] v t This represents the real-time sprinkler flow rate (L / min) at time t.

[0079] v0 represents the basic sprinkler flow rate (L / min);

[0080] D t This represents the measured dust concentration at time t (mg / m³).

[0081] D p This indicates the predicted dust concentration (mg / m³).

[0082] H t Indicates the current ambient humidity (%);

[0083] H0 represents the baseline humidity (50%).

[0084] w tIndicates the current wind speed (m / s);

[0085] w0 represents the critical wind speed (3m / s);

[0086] γ represents the dust response coefficient (value 1.2);

[0087] κ represents the humidity influence coefficient (value 0.02);

[0088] λ represents the wind speed correction factor (value 0.15).

[0089] In step S8, operation instructions are sent to each sprinkler truck according to the adjusted sprinkling strategy. The instructions include the coordinates of the sprinkling area, the sprinkling flow rate, the sprinkling duration, and the driving route. The on-board terminal receives the sprinkling truck's operation data in real time, including the actual sprinkling volume, sprinkling time, and operation completion status. The location and operation status of each sprinkler truck are displayed in real time on the monitoring platform. When a sprinkler truck malfunctions or the sprinkling volume is insufficient, a backup sprinkler truck is automatically dispatched. After the operation is completed, the sprinkler truck uploads an electronic work order, including the actual sprinkling area and the sprinkling volume. The system automatically generates a sprinkling operation report.

[0090] In step S9, one hour after the watering operation is completed, the actual dust volume data of each sub-area is collected and compared with the predicted dust volume to calculate the error rate. When the error rate exceeds 10%, the model optimization process is initiated. The training set is updated using the latest actual dust volume data and construction data, the hybrid neural network model is retrained, the learning rate and the number of hidden layer neurons are adjusted, the regional feature weight parameters are optimized, a comprehensive evaluation of the model is conducted monthly, the model architecture or training strategy is adjusted according to the evaluation results, the optimized model is deployed to the production environment and synchronized to each edge computing node, and a model optimization log is established to record the parameter adjustments and effect evaluation results of each optimization.

[0091] From the above, we can conclude that:

[0092] By constructing a hybrid model combining Long Short-Term Memory (LSTM) networks and convolutional neural networks, and introducing regional feature weight parameters, dust emission prediction can simultaneously capture temporal variation patterns and spatial distribution differences. During model training, dynamic adjustment of hyperparameters and an attention mechanism further enhances the model's ability to fit dust generation patterns in complex operating environments, thus providing a reliable decision-making basis for subsequent sprinkler control. This accurate predictive capability effectively avoids the blindness of traditional fixed-frequency sprinkler methods, ensuring that sprinkler operations can respond in advance to potential dust peaks.

[0093] When calculating water demand based on forecast results, the system comprehensively considers the dust particle settling characteristics of the work area and equipment parameters, achieving quantitative allocation of water resources. By dynamically adjusting the watering strategy in conjunction with real-time monitoring data and weather changes, the system can promptly increase watering intensity when dust levels exceed expectations and automatically reduce watering when environmental conditions are favorable for dust suppression. This dynamic adjustment mechanism ensures effective dust control while avoiding ineffective water consumption, significantly improving the overall economic efficiency of watering operations.

[0094] By continuously collecting and comparing actual dust data with predicted values, and optimizing model parameters and regional feature weights, a closed-loop adaptive improvement mechanism is formed. During watering operations, intelligent scheduling of water trucks is achieved through vehicle networking and GPS positioning, enabling adjustments to driving routes and operation sequences based on real-time needs, thus improving equipment utilization. Simultaneously, an emergency response mechanism for sudden weather changes further ensures the safety and effectiveness of watering operations, making the entire system robust enough to handle complex working conditions.

[0095] The above are all preferred embodiments of this application, and are not intended to limit the scope of protection of this application. Therefore, all equivalent changes made in accordance with the structure, shape and principle of this application should be covered within the scope of protection of this application.

Claims

1. A sprinkler control method based on dust data prediction of the work area, characterized in that: The method includes the following steps: S1: Collect construction data, weather data, and regional characteristic data within the mining operation area to provide basic data for subsequent data processing and model training; S2: Clean and standardize the collected data, remove outliers and missing values, unify the data format, and transmit the processed data to the model training module. S3: Based on the preprocessed data, a hybrid neural network model is built using a long short-term memory network architecture combined with a convolutional neural network, and the model is trained. During the training process, hyperparameters are adjusted to optimize the model performance. S4: Using the trained hybrid neural network model and real-time collected data, predict the dust volume within a specified time period sequence, and send the prediction results to the water demand calculation module. S5: Calculate the water demand based on the predicted dust volume, the dust particle settling efficiency of the work area, the density of water, and the water flow parameters of the sprinkler equipment, to provide data support for water resource scheduling. S6: Determine the required number of sprinkler trucks by combining the water demand, sprinkler truck water tank capacity, working time and actual watering efficiency parameters. At the same time, determine the watering time period and watering area based on the spatiotemporal distribution of dust, meteorological conditions and work activities. S7: Based on the determined number of sprinkler trucks, sprinkler time period and sprinkler area, combined with real-time dust monitoring data and weather changes, dynamically adjust the sprinkler strategy, increase the sprinkler flow and frequency when the dust volume is high and the wind speed is high, and reduce the sprinkler flow and frequency when the dust volume is low and the humidity is high. S8: Based on the adjusted watering strategy, control the number of water trucks matching the number of water trucks to carry out watering operations on the designated area within the specified time period, and monitor the operation of the water trucks in real time. S9: After the watering operation is completed, collect the actual dust volume data of the operation area, compare and analyze it with the predicted dust volume, optimize the parameters of the hybrid neural network model based on the analysis results, and improve the accuracy of subsequent dust volume prediction and the precision of watering control.

2. The water spraying control method based on dust data prediction of the work area according to claim 1, characterized in that: In step S1, a laser scattering dust sensor, a wind speed sensor, a humidity sensor, and a temperature sensor are installed in the mining operation area. At the same time, the CAN bus data interface built into the operating equipment is integrated to collect the excavator's working status parameters, including digging depth and digging frequency, truck transportation frequency parameters, including the number of transport vehicles per hour and the average load, as well as wind speed, humidity, temperature, and air pressure data every 5 minutes. The real-time data is transmitted to the edge computing node via industrial Ethernet, and images of the operation area are collected by a high-definition camera.

3. The water spraying control method based on dust data prediction of the work area according to claim 1, characterized in that: In step S2, the IQR method is used to detect and remove outliers in the construction data, the linear interpolation method is used to fill in missing values, the wind speed, humidity and temperature meteorological data are Z-score standardized, the excavator working status parameters and truck transportation frequency data are converted into dimensionless indicators, and the data format is unified as JSON format.

4. The water spraying control method based on dust data prediction of the work area according to claim 1, characterized in that: In step S3, a hybrid model of long short-term memory network and convolutional neural network is constructed based on preprocessed historical construction data, weather data, and regional feature data. The convolutional neural network part uses three convolutional layers to extract spatial features of the data, while the long short-term memory network part has two hidden layers, each containing 128 neurons. The Adam optimizer is used with an initial learning rate of 0.

001. Early stopping is used to prevent overfitting. The learning rate is adjusted every 50 epochs during training. The terrain slope and soil particle composition in the regional feature data are used as the weight parameters of the model. The influence weight of each sub-region in the model is dynamically adjusted through an attention mechanism. The root mean square error is used as the loss function of the model. The ratio of the model on the training set, validation set, and test set is set to 7:2:

1. Training stops when the loss of the model on the validation set no longer decreases for 10 consecutive epochs.

5. The water spraying control method based on dust data prediction of the work area according to claim 1, characterized in that: In step S4, the collected historical construction data, weather data, and regional feature data are cleaned and standardized to remove outliers and missing values ​​and unify the data format. The hybrid neural network model is trained using the historical construction data, weather data, and regional feature data as input and the actual monitored dust volume as output.

6. The water spraying control method based on dust data prediction of the work area according to claim 1, characterized in that: In step S5, the formula for calculating the water demand based on the predicted dust volume is as follows: ； Where Q is the water demand, D is the predicted dust volume, A is the area of ​​the work area, η is the settling efficiency of dust particles, which is related to the particle size distribution and settling characteristics of particles in the area, ρ is the density of water, and v is the water flow rate of the sprinkler equipment.

7. The water spraying control method based on dust data prediction of the work area according to claim 1, characterized in that: In step S6, the formula for calculating the number of sprinkler trucks N based on the water demand Q and the sprinkler truck parameters is as follows: ； Where N is the number of sprinkler trucks, C is the water tank capacity of the sprinkler trucks, T is the working time of the sprinkler trucks, ε is the actual sprinkling efficiency of the sprinkler trucks, and ⌈⋅⌉ indicates rounding up to ensure that the number of sprinkler trucks is sufficient to meet the demand; Furthermore, the watering schedule is determined based on the spatiotemporal distribution of dust, meteorological conditions, and the arrangement of work activities.

8. The water spraying control method based on dust data prediction of the work area according to claim 1, characterized in that: In step S7, based on the determined number of sprinkler trucks, sprinkler time period, and sprinkler area, the system collects the current dust concentration, wind speed, and humidity data of each sub-area within the operation area in real time, and compares them with the predicted dust amount and preset thresholds. For areas with assigned sprinkler tasks, the system dynamically adjusts the driving path according to the real-time location data of the sprinkler trucks. When adjacent areas trigger sprinkler demand simultaneously, the system prioritizes dispatching the nearest available sprinkler truck and updates the task list synchronously through the vehicle network system. During the sprinkler operation, if the meteorological department issues a short-term gale warning, the system immediately suspends the sprinkler operation in the open area and recalculates the replenishment amount according to the dust rebound model after the warning is lifted. The dynamic sprinkler flow correction formula is: ； In the formula: v t This represents the real-time sprinkler flow rate at time t; v0 represents the basic sprinkler flow rate; D t This represents the measured dust concentration at time t; D p Indicates the predicted dust concentration; H t Indicates the current ambient humidity; H0 represents the reference humidity; w t Indicates the current wind speed; w0 represents the critical wind speed; γ represents the dust response coefficient; κ represents the humidity influence coefficient; λ represents the wind speed correction factor.

9. The water spraying control method based on dust data prediction of the work area according to claim 1, characterized in that: In step S8, operation instructions are sent to each sprinkler truck according to the adjusted sprinkling strategy. The instructions include the coordinates of the sprinkling area, the sprinkling flow rate, the sprinkling duration, and the driving route. The on-board terminal receives the operation data of the sprinkler truck in real time, including the actual sprinkling volume, sprinkling time, and operation completion status. The location and operation status of each sprinkler truck are displayed in real time on the monitoring platform.

10. The water spraying control method based on dust data prediction of the work area according to claim 1, characterized in that: In step S9, one hour after the watering operation is completed, the actual dust volume data of each sub-area is collected, and the error rate is calculated by comparing it with the predicted dust volume. When the error rate exceeds 10%, the model optimization process is initiated. The training set is updated using the latest actual dust volume data and construction data, the hybrid neural network model is retrained, the learning rate and the number of hidden layer neurons are adjusted, the regional feature weight parameters are optimized, a comprehensive evaluation of the model is conducted once a month, the model architecture or training strategy is adjusted according to the evaluation results, the optimized model is deployed to the production environment and synchronized to each edge computing node, and a model optimization log is established to record the parameter adjustments and effect evaluation results of each optimization.

Images