Intelligent air pressure cleaning device and method for filling pipeline of underground mine

Abstract

The application discloses an intelligent air pressure cleaning device and method for filling pipeline of underground mine, and belongs to the technical field of underground treatment. The device comprises an intelligent control module, a frequency converter, an air path assembly, a water path assembly, an air-water switching valve and a filling pipeline sealing adaptive device. The intelligent control module is connected with the frequency converter, the air path assembly, the water path assembly, the air-water switching valve and the filling pipeline sealing adaptive device respectively. The frequency converter is connected with the air path assembly and the water path assembly respectively. The air path assembly and the water path assembly are connected with the air-water switching valve respectively. The air-water switching valve is connected with the filling pipeline sealing adaptive device. The application solves the problems of large water resource consumption, incomplete cleaning, fixed parameters and time length depending on artificial experience of traditional clean water pipe cleaning, and has the advantages of high efficiency, energy saving, intelligent adaptation, stable operation and high automation degree, and improves the safety and economy of the filling system.

Description

Technical Field

[0001] This invention relates to the field of underground treatment technology, and in particular to an apparatus and method for cleaning and treating underground filling pipelines in mines. Background Technology

[0002] In the current underground mining sector, cemented backfilling technology has become a core technology for ensuring the safety of goaf areas and improving resource recovery rates. As a crucial carrier for transporting backfill slurry, the smoothness of the backfilling pipeline directly determines the continuity and safety of backfilling operations. If the residual backfill slurry on the inner wall of the pipeline is not thoroughly cleaned in a timely manner after the backfilling station completes production, it can easily accumulate and harden, forming a solidified layer. This not only reduces the pipeline's flow cross-section and increases the resistance to slurry transport, but also easily leads to pipe blockage accidents, causing backfilling operations to be interrupted. This significantly increases the labor and equipment costs for pipeline clearing and can even affect the overall progress of underground mining operations.

[0003] Currently, the mainstream method for cleaning backfill pipelines in domestic underground mines is still single-flow gravity flushing with clean water. While this technology is simple to operate, it has revealed many unresolved industry pain points in practical applications and is no longer suitable for the large-scale, high-concentration backfill production needs of modern mines: First, it consumes a large amount of water. Due to the long transportation distance of the backfill pipeline (the horizontal transportation distance in some mines exceeds 1,000 meters) and the presence of many branches and veins, a single cleaning requires the injection of a large amount of clean water. This not only causes serious waste of water resources, but the cleaning water entering the mine can also seep into the backfill, reducing the setting strength and forming quality of the backfill, while also significantly increasing the workload and energy consumption of the underground drainage system. Second, the cleaning effect is poor and the risk of pipe blockage is high. The cleaning ability of clean water pipe washing is affected by the pipeline path (bends, diameter changes), water flow velocity, and slurry characteristics (concentration, ash content). Multiple factors, including the sand-to-sand ratio, limit the effective cleaning of stubbornly adhered slurry on the inner wall of the pipeline, leading to slurry accumulation and gradual hardening, which becomes a core cause of pipe blockage accidents. Third, the fixed cleaning parameters cannot adapt to diverse working conditions. The pipeline conveying distance varies significantly across different veins, the slurry ratio (tailings sand to adhesive ratio) differs at different filling stages, and the degree of residual slurry accumulation varies with the pipeline's service life. Traditional water-washing uses a uniform flushing method, lacking targeted parameter adjustments, further exacerbating the problem of incomplete cleaning. Fourth, the cleaning duration is not scientifically planned. Current cleaning operations mostly rely on manual experience to determine the flushing time, which easily leads to insufficient or excessive cleaning. Insufficient cleaning cannot eliminate the risk of pipe blockage, while excessive cleaning will increase resource consumption and operating costs. Summary of the Invention

[0004] This invention aims to provide an intelligent compressed air cleaning device and method for backfill pipelines in underground mining, to solve the following technical problems existing in current backfill pipeline cleaning technologies: high water consumption from single-use water cleaning, incomplete cleaning leading to pipe blockage, fixed cleaning parameters unable to adapt to diverse working conditions, lack of scientific planning for cleaning duration, and limited parameter adjustment capabilities and lack of intelligent control in existing air-water linkage cleaning technologies. The specific solution is as follows: An intelligent compressed air cleaning device for filling pipelines in underground mining includes an intelligent control module, a frequency converter, an air circuit assembly, a water circuit assembly, an air-water switching valve, and a filling pipeline sealing adapter. The intelligent control module is connected to the frequency converter, the air circuit assembly, the water circuit assembly, the air-water switching valve, and the filling pipeline sealing adapter. The frequency converter is connected to the air circuit assembly and the water circuit assembly. The air circuit assembly and the water circuit assembly are connected to the air-water switching valve, and the air-water switching valve is connected to the filling pipeline sealing adapter.

[0005] Furthermore, the air duct assembly includes a variable frequency air compressor, an air tank, a main air duct, an electric regulating valve, an air volume sensor, and an air pressure sensor. The variable frequency drive is connected to the variable frequency air compressor, which is connected to the air tank. The air tank is connected to the main air duct, which is connected to the electric regulating valve. The electric regulating valve is connected to the air-water switching valve. The main air duct is equipped with an air volume sensor, an air pressure sensor, and an electro-proportional valve.

[0006] Furthermore, the water circuit assembly includes a variable frequency high-pressure water pump, a main water pipe, an electric regulating valve II, a water volume sensor, and a water pressure sensor. The variable frequency drive is connected to the variable frequency high-pressure water pump, which is connected to the main water pipe. The main water pipe is connected to the electric regulating valve II, which is connected to the air-water switching valve. The main water pipe is equipped with a water volume sensor, a water pressure sensor, and an electric proportional valve II.

[0007] Furthermore, the intelligent control module includes a PLC host, a human-machine interaction unit, a vein selection unit, a filling stage selection unit, a clock timing unit, a parameter storage unit, and a sealing adapter control unit; The human-machine interaction unit, the vein selection unit, the filling stage selection unit, the clock timing unit, the parameter storage unit, and the sealing adapter control unit are all connected to the PLC host. The human-machine interaction unit is connected to the vein selection unit and the filling stage selection unit respectively.

[0008] Furthermore, the filling pipe sealing adapter includes a main water inlet pipe and a mechanical clamping sleeve screwed onto the outside of the main water inlet pipe. One end of the mechanical clamping sleeve is provided with a rotating handle, and an anti-detachment safety chain is provided at the rotating handle. The other end is a compression head, which is adjacent to an intelligent control rubber expansion seal. The intelligent control rubber expansion seal is located at one end of the main water inlet pipe, and a pressure monitoring sensor is provided on the intelligent control rubber expansion seal.

[0009] A smart compressed air cleaning method for filling pipelines in underground mining includes the following steps: Step A: Device and system initialization; Step B: The operator selects the vein number, filling stage, and pipe number to be cleaned through the human-machine interaction unit of the intelligent control module; Step C: The intelligent control module receives the selection signal, reads the current time through the clock timing unit, and simultaneously reads the cumulative number of days the selected pipeline has been used. Based on the current time, the intelligent control module determines whether to perform high-pressure water cleaning alone or combined air and water cleaning; based on the cumulative number of days the pipeline has been used, it determines the cleaning duration. 1) If the current time is 12:00: Perform a single high-pressure water cleaning; 2) If the current time is 24:00: Perform a combined high-pressure water + high-pressure air cleaning; 3) If the pipeline has been used for a cumulative period of 3 days: the corresponding cleaning time will automatically double; Step D: The PLC host of the intelligent control module, based on the signals from the pulse selection unit and the filling stage selection unit, calls the preset data in the parameter storage unit, automatically matches and outputs the target parameters. 1) Match the corresponding water volume and water pressure parameters according to the vein number and filling stage; 2) If it is a 24:00 joint cleaning mode, the corresponding air volume and air pressure parameters will be matched synchronously; 3) Send the final parameters to the frequency converter, electric regulating valve 1, electric regulating valve 2, electric proportional valve 1, and electric proportional valve 2 to complete the actuator parameter setting; This step is the main timer T=0min stage; Step E, System Preprocessing Stage, T=0-10s, lasting 10 seconds to ensure equipment stability. Preprocessing includes: 1) The variable frequency high-pressure water pump starts, and the water system is filled with water and vented. 2) In the combined cleaning mode, the variable frequency air compressor starts synchronously, and the air tank stabilizes and stores energy; 3) The electric regulating valve and the electric proportional valve are opened to their initial opening degree; 4) The sealing adaptation structure enters the intelligent adaptive sealing process; 5) Confirm that the seal is complete and there are no leaks or abnormalities; the pretreatment is now complete. Step F: If the current time period is 12:00, T = 10s - the set total duration, the system enters the high-pressure water separate cleaning mode, the air-water switching valve group keeps the water circuit open, this stage will run until the set time ends automatically, and then proceed to step H; Step G: The current time period is 24:00. T = 10s - set total duration. The system enters the high-pressure water + high-pressure air combined cleaning mode. The air-water switching valve group keeps the air-water linkage conduction. This stage will run until the set time ends automatically and proceed to step H. After step H, the rinsing operation is completed, the PLC host executes the shutdown and reset procedure; Step 1: Parallel detection throughout the entire system process; Step J: Cyclic standby.

[0010] Furthermore, in step E, the sealing adapter structure enters the intelligent adaptive sealing process, including: the PLC host sends an expansion command to the sealing drive mechanism to control the rubber expansion seal to actively expand radially; the pressure monitoring sensor collects the expansion pressure of the seal in real time; when the pressure reaches the preset sealing completion threshold, the pressurization is automatically stopped, and the sealing is determined to be successful; if the sealing pressure threshold is not reached within 10 seconds, the system alarms and jumps to emergency stop processing. Step C determines the cleaning duration based on the cumulative number of days the pipeline has been used, including: If the current time is 12:00, the basic rinsing time is 15 minutes; If the current time is 24:00, the basic rinsing time is 30 minutes; If the pipeline is used for a cumulative period of 3 days, the corresponding cleaning time will automatically double, i.e., 12:00 becomes 30 minutes and 24:00 becomes 60 minutes.

[0011] Furthermore, the high-pressure water cleaning mode in step F includes: 1) The PLC host controls the variable frequency high-pressure water pump and the electric proportional valve, and outputs matching water volume and water pressure; 2) Water pressure and water volume sensors monitor in real time, forming a closed-loop precise adjustment; 3) Perform the full high-pressure water cleaning process according to the time set in step C; Step G, the combined high-pressure water and high-pressure air cleaning mode, includes: 1) The PLC host controls the variable frequency high-pressure water pump and the electric proportional valve, and outputs matching water volume and water pressure; 2) Synchronous control of the variable frequency air compressor and electric regulating valve to match the output air volume and air pressure; 3) Water pressure sensor, water volume sensor, wind pressure sensor, and air volume sensor collect data in real time and adjust to the set value in a closed loop; 4) Perform the full-process joint cleaning according to the time set in step C.

[0012] Further, step H includes: 1) Stop the variable frequency air compressor and variable frequency high-pressure water pump in sequence; 2) Close all electric regulating valves, electric proportional valves, and air-water switching valve assemblies; 3) Slowly release pressure in the air and water systems; 4) When the sealing adapter structure releases the seal, the rubber expansion seal automatically retracts to its initial state; 5) The pipeline usage days counter automatically increments by 1 day and saves the current operation parameters and duration; 6) After the reset is complete, proceed to step J; Step J includes: all cleaning processes are completed, all actuators are reset, data recording is completed, the system automatically returns to the initial state of Step A, and waits for the next vein puncture, filling stage, pipeline number selection and start signal to enter the next cleaning cycle.

[0013] Further, step I includes: As part of the parallel detection process throughout the entire system, the PLC host continuously monitors emergency stop signals, pressure over-limit signals, abnormal seal pressure signals, and equipment fault signals in real time during any stage of operation from step B to step H. If no abnormal signal is detected, the system continues to execute the current step logic; If an emergency stop / fault signal is detected, immediately interrupt all actuators, perform emergency pressure relief, close all valves, lock the fault code, and directly jump to step A to perform system reset.

[0014] Compared with the prior art, the present invention has at least one of the following technical effects: 1. This invention employs a combined compressed air and high-pressure water cleaning technology to replace the traditional single-flow water flushing mode. Through an intelligent control module, water volume and pressure are precisely matched according to working conditions such as pipe penetration distance and filling stage. During 24:00, a combined air-water operation is used, utilizing the powerful impact and stripping effect of compressed air to reduce dependence on water flow. Compared to single-flow water pipe washing, this reduces the water consumption per cleaning cycle. Simultaneously, due to the reduced water consumption, less cleaning water enters the well, preventing a large amount of cleaning water from seeping into the filling material and reducing its setting strength. This also reduces the workload and energy consumption of the well drainage system, achieving water conservation, energy saving, and cost reduction from the source.

[0015] 2. Traditional gravity-flow water methods lack effective removal capabilities for stubbornly adhered slurry on the inner walls of pipes, easily leading to slurry residue buildup and caking. This invention integrates the synergistic effect of compressed air and high-pressure water: high-pressure water provides flushing force, while compressed air provides impact pulsation and cavitation effects; together, they powerfully remove residual slurry from the inner walls of pipes. Simultaneously, the system uses real-time closed-loop adjustment via air pressure / water pressure sensors and airflow / water flow sensors to ensure that output parameters are always optimal. Regardless of whether the pipe path contains bends, diameter changes, or variations in slurry concentration and ash-sand ratio, thorough cleaning can be achieved, eliminating pipe blockage caused by slurry deposition and caking at its source.

[0016] 3. This invention establishes an intelligent adaptation mechanism of "condition perception - parameter decision-making - precise execution". Operators select the pipeline number (corresponding to different conveying distances) and filling stage (bottom / middle / top, corresponding to different slurry ratios) through the human-machine interface. The PLC host calls multiple sets of pre-stored airflow, air pressure, waterflow, and water pressure parameters from the parameter storage unit and automatically matches the output. Furthermore, the clock timing unit automatically doubles the flushing duration based on the pipeline's cumulative usage days (e.g., a 3-day cycle), achieving on-demand cleaning. This allows pipelines with different pipelines, different filling stages, and different aging levels to obtain customized cleaning parameters, completely changing the traditional "one-size-fits-all" fixed parameter mode.

[0017] 4. This invention achieves standardized and periodic control of cleaning duration through a clock timing unit and a PLC main control program. The system is set to perform 15 minutes of high-pressure water cleaning at 12:00 noon and 30 minutes of combined air and water cleaning at 24:00 every day. Simultaneously, it monitors the cumulative number of days the pipeline has been used, and automatically doubles the corresponding duration when a 3-day cycle is reached (12:00 becomes 30 minutes, 24:00 becomes 60 minutes). All durations are automatically assigned and executed by the system according to preset logic, eliminating the need for manual estimation based on experience. This avoids the risk of pipe blockage caused by insufficient cleaning and prevents resource waste caused by over-cleaning, making cleaning operations data-driven, precise, and controllable.

[0018] 5. This invention continuously optimizes the combination of air pressure, water volume, and cleaning time under various operating conditions through reinforcement learning algorithms, which can effectively reduce energy consumption and media consumption, and save electricity and water treatment costs. At the same time, the precise on-demand cleaning strategy avoids local scaling and accelerated wear caused by insufficient cleaning and pipe wall stress fatigue caused by excessive cleaning, which helps to extend the service life of pipelines. In addition, parameter configuration, effect evaluation, and strategy optimization do not require manual intervention, which not only reduces the burden of on-site operation, but also realizes the intelligent improvement from fixed parameter cleaning to data-driven self-optimizing cleaning. Attached Figure Description

[0019] To more clearly illustrate the technical solutions in the embodiments of this application, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0020] Figure 1 This is a connection diagram of an intelligent compressed air cleaning device for filling pipelines in underground mining, according to the present invention. Figure 2 This is a schematic diagram of the structure of the filling pipe sealing adapter of the present invention.

[0021] Figure label: 1-Intelligent control module; 2-Frequency inverter; 3-Filling pipe sealing adapter; 4-Variable frequency air compressor; 5-Air tank; 6-Main air pipe; 7-Electric regulating valve one; 8-Air volume sensor; 9-Air pressure sensor; 10-Air-water switching valve; 11-Electro-proportional valve one; 12-Main water pipe; 13-Electro-regulating valve two; 14-Water volume sensor; 15-Water pressure sensor; 16-Electro-proportional valve two; 17-PLC host; 18-Human machine interaction unit; 19-Pulse selection unit; 20-Filling stage selection unit; 21-Clock timing unit; 22-Parameter storage unit; 23-Sealing adapter mechanism control unit; 24-Mechanical clamping sleeve; 25-Rotating handle; 26-Extrusion column head; 27-Intelligent control type rubber expansion seal; 28-Anti-detachment safety chain; 29-Pressure monitoring sensor; 30-Variable frequency high-pressure water pump; 31-Main water inlet pipe. Detailed Implementation

[0022] In the following description, specific details such as particular system architectures and techniques are set forth for illustrative purposes and not for limitation, in order to provide a thorough understanding of the embodiments of this application. However, those skilled in the art will understand that this application may also be implemented in other embodiments without these specific details. In other instances, detailed descriptions of well-known systems, apparatuses, circuits, and methods have been omitted so as not to obscure the description of this application with unnecessary detail.

[0023] It should be understood that, when used in this application specification and the appended claims, the term "comprising" indicates the presence of the described features, integrals, steps, operations, elements and / or components, but does not exclude the presence or addition of one or more other features, integrals, steps, operations, elements, components and / or a collection thereof.

[0024] It should also be understood that the term “and / or” as used in this application specification and the appended claims means any combination of one or more of the associated listed items and all possible combinations, and includes such combinations.

[0025] As used in this application specification and the appended claims, the term "if" may be interpreted, depending on the context, as "when," "once," "in response to determination," or "in response to detection." Similarly, the phrase "if determined" or "if detected [the described condition or event]" may be interpreted, depending on the context, as meaning "once determined," "in response to determination," "once detected [the described condition or event]," or "in response to detection [the described condition or event]."

[0026] Furthermore, in the description of this application and the appended claims, the terms "first," "second," "third," etc., are used only to distinguish descriptions and should not be construed as indicating or implying relative importance.

[0027] References to "one embodiment" or "some embodiments" as described in this specification mean that one or more embodiments of this application include a specific feature, structure, or characteristic described in connection with that embodiment. Therefore, the phrases "in one embodiment," "in some embodiments," "in other embodiments," "in still other embodiments," etc., appearing in different parts of this specification do not necessarily refer to the same embodiment, but rather mean "one or more, but not all, embodiments," unless otherwise specifically emphasized. The terms "comprising," "including," "having," and variations thereof mean "including but not limited to," unless otherwise specifically emphasized.

[0028] See appendix Figure 1 -Appendix Figure 2 This invention discloses an intelligent compressed air cleaning device for filling pipelines in underground mining, comprising an intelligent control module 1, a frequency converter 2, an air circuit assembly, a water circuit assembly, an air-water switching valve 10, and a filling pipeline sealing adapter 3. The intelligent control module 1 is connected to the frequency converter 2, the air circuit assembly, the water circuit assembly, the air-water switching valve 10, and the filling pipeline sealing adapter 3. The frequency converter 2 is connected to the air circuit assembly and the water circuit assembly. The air circuit assembly and the water circuit assembly are connected to the air-water switching valve 10, and the air-water switching valve 10 is connected to the filling pipeline sealing adapter 3.

[0029] The intelligent control module 1 is a system component of the device. After the operator selects the vein number, filling stage, and pipe number to be cleaned at its input terminal and presses the start button, the built-in control unit of the intelligent control module 1 receives the above-mentioned vein number, filling stage, and pipe number to be cleaned signals, starts reading the current time, reads the cumulative number of days used for the selected pipe, executes the cleaning mode and duration assignment logic, calls the preset data in the parameter storage unit 22, automatically matches and outputs the target parameters to: 1) Variable frequency drive 2, which controls the air generation device of the air circuit component and the water generation device of the water circuit component to generate air volume, air pressure, water volume, water pressure and basic flushing time respectively, which are adapted to the above-mentioned through-pulse number, filling stage and the number of the pipe to be cleaned. 2) The airflow assembly is equipped with an air generating device and air pressure and air volume regulating devices; the waterflow assembly is equipped with a water generating device and water pressure and water volume regulating devices. The air pressure and air volume regulating devices enable fine-tuning of air pressure and air volume at two or more levels; the water pressure and water volume regulating devices enable fine-tuning of water pressure and water volume at two or more levels. 3) The air-water switching valve 10, depending on the current time, the control unit of the intelligent control module 1 decides whether to adopt the high-pressure water cleaning mode alone or the high-pressure water + high-pressure air combined cleaning mode, and controls the air-water switching valve 10 accordingly to complete the switching. 4) Filling pipe sealing adapter 3: Before cleaning, the control unit of the intelligent control module 1 controls the filling pipe sealing adapter 3 to adaptively seal the inlet end of the filling pipe, effectively preventing leakage of flushing medium.

[0030] In specific implementation, the intelligent control module 1 includes a PLC host 17, a human-machine interaction unit 18, a vein selection unit 19, a filling stage selection unit 20, a clock timing unit 21, a parameter storage unit 22, and a sealing adapter control unit 23. The human-machine interaction unit 18, the vein selection unit 19, the filling stage selection unit 20, the clock timing unit 21, the parameter storage unit 22, and the sealing adapter control unit 23 are all connected to the PLC host 17. The human-machine interaction unit 18 is connected to the vein selection unit 19 and the filling stage selection unit 20 respectively.

[0031] The PLC host 17 receives signals for pulse penetration, filling stage, and time cycle, calls up and calculates pre-stored parameters, and outputs control commands for air volume, air pressure, water volume, water pressure, and flushing duration.

[0032] The human-machine interaction unit 18 is used to select the cleaning and filling stages and display the operating status.

[0033] The vein selection unit 19 corresponds to different delivery distances of 10 veins, providing a basis for parameter matching.

[0034] The filling stage selection unit 20 distinguishes between bottom / top and middle filling conditions, such as the slurry mix ratio used.

[0035] The human-machine interface 18 serves as the operation input interface, which is used to receive the vein number and filling stage selected by the operator, and transmit these selection results to the vein selection unit 19 and the filling stage selection unit 20 respectively. These two units then provide the corresponding conveying distance parameters and slurry ratio parameters to the PLC, thereby realizing working condition identification and parameter matching.

[0036] The clock timing unit 21 automatically determines the rinsing duration according to the 0 o'clock and 12 o'clock nodes and the three-day cycle rule.

[0037] The clock timing unit 21 provides the current standard time signal to the PLC host 17 in real time. The PLC host 17 determines whether the preset cleaning period (such as 12:00 or 24:00) has been reached based on the signal and executes the corresponding cleaning mode and duration assignment logic.

[0038] The parameter storage unit 22 pre-stores multiple sets of air volume, air pressure, water volume, and water pressure parameters, enabling quick retrieval.

[0039] The PLC host 17 sends an expansion command to the sealing adapter control unit 23 and receives the sealing pressure signal fed back by the pressure monitoring sensor 29. When the pressure reaches the set threshold, the sealing is automatically determined to be complete.

[0040] The PLC host 17 retrieves the corresponding preset air volume, air pressure, water volume, and water pressure parameters from the parameter storage unit 22 based on the signals from the pulse selection unit 19 and the filling stage selection unit 20, for subsequent control of the actuators.

[0041] This module, through multi-unit collaboration, completes the entire process of intelligent control from "condition perception to parameter decision-making to command output," providing a reliable command basis for subsequent precise execution.

[0042] In specific implementation, the air duct assembly includes a variable frequency air compressor 4, an air tank 5, a main air duct 6, an electric regulating valve 7, an air volume sensor 8, and an air pressure sensor 9. The variable frequency drive 2 is connected to the variable frequency air compressor 4, which is connected to the air tank 5. The air tank 5 is connected to the main air duct 6, which is connected to the electric regulating valve 7. The electric regulating valve 7 is connected to the air-water switching valve 10. The main air duct 6 is connected to the air volume sensor 8, the air pressure sensor 9, and an electro-proportional valve 11.

[0043] Inverter 2 is used to adjust the air volume of variable frequency air compressor 4.

[0044] The PLC main unit 17 regulates the speed of the variable frequency air compressor 4 via the frequency converter 2, thereby controlling the output air volume and air pressure. In other words, the variable frequency air compressor 4 receives the adjustment signals from the PLC main unit 17 of the intelligent control module 1, and continuously adjusts the exhaust volume and exhaust pressure through frequency conversion speed regulation, providing the system with a compressed air source with adjustable pressure and flow rate. The air storage tank 5 provides the system with pressure-stabilized compressed air. The PLC host 17 indirectly monitors the outlet pressure of the air storage tank 5 through the downstream air pressure sensor 9. That is, the air storage tank 5 stabilizes the pressure of the compressed air, stores energy, and performs preliminary dehydration, buffers the pressure fluctuations caused by the start and stop of the air compressor, stabilizes the air path pressure, and ensures the accuracy of subsequent adjustment and the stability of the medium output.

[0045] The main air duct 6 serves as the compressed air delivery channel. The PLC host 17 obtains pressure and flow feedback signals through the air pressure sensor 9 and air volume sensor 8 installed on it. The main air duct 6 is used to deliver compressed air, ensuring smooth media delivery and low pressure loss.

[0046] The PLC host 17 sends an opening control command to the electric regulating valve 7 based on the target air pressure / air volume value, performing secondary fine adjustment of the air volume / air pressure. The electric regulating valve 7 further improves output stability and accuracy.

[0047] The air volume sensor 8 collects the air volume signal in the main air duct 6 in real time and uploads it to the PLC host 17 for closed-loop regulation and control.

[0048] The wind pressure sensor 9 collects the wind pressure signal in the main air duct 6 in real time and uploads it to the PLC host 17 for closed-loop regulation and control.

[0049] Based on the deviation between the feedback signal from the wind pressure sensor 9 and the set value, the PLC host 17 sends a control command to the electro-proportional valve 11, which continuously and finely adjusts the wind pressure so that the actual output value quickly approaches the set value.

[0050] In specific implementation, the water circuit assembly includes a variable frequency high-pressure water pump 30, a main water pipe 12, an electric regulating valve 2 13, a water volume sensor 14, and a water pressure sensor 15. The variable frequency drive 2 is connected to the variable frequency high-pressure water pump 30, which is connected to the main water pipe 12. The main water pipe 12 is connected to the electric regulating valve 2 13, which is connected to the air-water switching valve 10. The main water pipe 12 is connected to the water volume sensor 14, the water pressure sensor 15, and the electric proportional valve 2 16.

[0051] Inverter 2 is used to adjust the speed of the variable frequency water pump to achieve coarse adjustment of the water volume.

[0052] The PLC host 17 adjusts the speed of the variable frequency high-pressure water pump 30 through the frequency converter 2, thereby controlling the output water volume and pressure. That is, the variable frequency high-pressure water pump 30 adjusts its output power in real time according to the control command, providing a variable high-pressure water flow, and achieving precise adjustment of water volume and pressure.

[0053] The main water pipe 12 serves as a high-pressure water delivery channel. The PLC host 17 obtains pressure and flow feedback signals through the water pressure sensor 15 and water flow sensor 14 installed on it. The main water pipe 12 is used to transport high-pressure water, ensuring smooth media delivery and low pressure loss.

[0054] The PLC host 17 sends an opening control command to the electric regulating valve 13 based on the target water pressure / volume value, performing secondary fine adjustment of the water volume / pressure. The electric regulating valve 13 further improves output stability and accuracy.

[0055] The water sensor 14 collects the water volume signal in the main water pipe 12 in real time and uploads it to the PLC host 17 to form real-time feedback, which is used to adjust the water volume to the set value in a closed loop.

[0056] The water pressure sensor 15 collects the water pressure signal in the main water pipe 12 in real time and uploads it to the PLC host 17 to form real-time feedback, which is used to adjust the water pressure to the set value in a closed loop.

[0057] The PLC host 17 controls the operation of the air-water switching valve 10 according to a preset timing sequence (such as high-pressure water cleaning alone or air-water combined cleaning), thereby realizing the switching or linkage output of air and water media. That is, the air-water switching valve group completes the switching and linkage output of air and water media according to the preset timing sequence, meeting the needs of various cleaning modes such as water first and then air, and air-water synergy.

[0058] Based on the deviation between the feedback signal from the water pressure sensor 15 and the set value, the PLC host 17 sends a control command to the electro-proportional valve 16, which continuously and finely adjusts the water pressure so that the actual output value quickly approaches the set value.

[0059] This invention utilizes a variable frequency power source and regulating valves to work in tandem, automatically outputting differentiated air pressure, air volume, water pressure, and water volume according to the needs of different veining and filling stages, providing stable and adjustable media power for multi-condition adaptive cleaning. The wind pressure sensor 9, air volume sensor 8, water pressure sensor 15, water volume sensor 14, frequency converter 2, and electric proportional valve together constitute the parameter detection and adjustment component. Through the "detection-feedback-adjustment" closed-loop mechanism, it ensures that the wind pressure, air volume, water pressure, and water volume strictly follow the output of the decision parameters, ensuring accurate and reliable cleaning operations.

[0060] The filling pipe sealing adapter 3 includes a main water inlet pipe 31 and a mechanical compression sleeve 24 screwed onto the outside of the main water inlet pipe 31. One end of the mechanical compression sleeve 24 is provided with a rotating handle 25, and the other end is a compression head 26. The compression head 26 is adjacent to an intelligent control rubber expansion seal 27. The intelligent control rubber expansion seal 27 is located at one end of the main water inlet pipe 31, and a pressure monitoring sensor 29 is provided on the intelligent control rubber expansion seal 27.

[0061] The mechanical clamping mechanism enables manual compression expansion: by rotating the handle 25, the mechanical clamping sleeve 24 moves along the thread toward the intelligent control rubber expansion seal 27, and the compression head 26 compresses the intelligent control rubber expansion seal 27, controlling the tight fit with the inner wall of different pipes according to the amount of compression expansion.

[0062] In addition, the PLC host 17 can also send an expansion command to the sealing adapter control unit 23, which wirelessly controls the expansion of the intelligent control rubber expansion seal 27 and receives the sealing pressure signal fed back by the pressure monitoring sensor 29. When the pressure reaches the set threshold, the sealing is automatically determined to be completed.

[0063] The cleaning water / air from the air-water switching valve enters the pipe to be cleaned through the main water inlet pipe.

[0064] The intelligent controllable rubber expansion seal 27 adopts a dual-mode expansion drive design. It can achieve active radial expansion under control system commands, with the expansion pressure collected in real time by the pressure monitoring sensor 29. When the pressure reaches the preset sealing threshold, pressurization automatically stops, completing an adaptive seal. Alternatively, it can achieve manual compression expansion via a mechanical clamping mechanism to meet emergency sealing needs. After expansion, the seal fits tightly against the inner wall of the pipe, forming a high-pressure seal that effectively prevents media leakage. The dual-mode drive mechanism combines automated intelligent adaptation with manual emergency operation, adaptively matching filling pipes of different diameters. The anti-detachment safety chain 28 resists the axial impact force generated by wind and water linkage, preventing the device from falling off or shifting, thus improving the safety of downhole operations. This structure can intelligently adapt to filling pipes of different diameters, combining automated and manual modes in the sealing process. It is easy to install, provides reliable high-pressure sealing, and offers structural protection for the stable operation of the entire cleaning device.

[0065] To achieve identification of working conditions (different pipelines, different filling stages), automatic adjustment of air and water parameters, time-based cleaning, and cycle duration control, this invention employs a PLC control system. The complete working sequence, logical judgments, and cycle rules of the intelligent compressed air cleaning device for filling pipelines in underground mining operations, from start to stop, are as follows: Step A (Initial State): After the system is powered on and connected to the network, it is in the initial state by default. The variable frequency air compressor 4 and the variable frequency high-pressure water pump 30 stop running, and the electric regulating valve, the electric proportional valve, and the air-water switching valve group are all closed and reset. The main timer T, pipeline usage days counter, flushing duration register, and time-division flag bit inside the PLC are all reset to zero. The clock timing unit 21 automatically synchronizes with the standard time. The sealed adapter structure is in an unexpanded state, all actuators are inactive, and the system continuously waits for the parameter selection and start trigger signal from the human-machine interaction unit 18 / host computer.

[0066] Step B (Operating Condition Selection and Startup Signal Judgment): The system monitors the input signals from the human-machine interface unit 18 / host computer in real time, and the operator makes the following selections: 1) Select the pipeline number (1~10, number 1 is 1.6km, and the pipeline length increases by 103m for each additional number). 2) Select the filling stage (bottom filling, middle filling, top filling, corresponding to different ratios of adhesive and tailings); 3) Select the pipe number to be cleaned. If the selection is not completed or the start button is not pressed, the system maintains the initial state of step A; if the selection is completed and a start signal is detected, the system locks the parameters and immediately proceeds to step C, while the main timer T starts counting.

[0067] Step C (Determining the pipeline usage cycle and cleaning period): The PLC host 17 reads the current time through the clock timing unit 21, and at the same time reads the cumulative number of days the selected pipeline has been used, and executes the cleaning mode and duration assignment logic: 1) If the current time is 12:00: Perform a single high-pressure water cleaning, with a basic rinsing time of 15 minutes; 2) If the current time is 24:00: Perform high-pressure water + high-pressure air combined cleaning, with a basic rinsing time of 30 minutes; 3) If the pipeline has been used for a cumulative period of 3 days: the corresponding cleaning time will be automatically doubled (12:00 becomes 30 minutes, 24:00 becomes 60 minutes).

[0068] After the duration and mode are determined, the system proceeds directly to step D.

[0069] Step D (Automatic matching of operating conditions and Feng Shui parameters): Based on the signals from the pulse selection unit 19 and the filling stage selection unit 20, the PLC host 17 calls the preset data in the parameter storage unit 22, automatically matches and outputs the target parameters: 1) Match the corresponding water volume and water pressure parameters according to the vein number and filling stage; 2) If it is a 24:00 joint cleaning mode, the corresponding air volume and air pressure parameters will be matched synchronously; 3) Send the final parameters to inverter 2, electric regulating valve, and electric proportional valve to complete the actuator parameter setting.

[0070] This step is the main timer T=0min stage. After the parameters are set, proceed to step E.

[0071] Step E (System preprocessing stage, T=0-10s): The system enters the preprocessing preparation phase, which lasts for 10 seconds to ensure equipment stability. 1) The variable frequency high-pressure water pump starts at 30, and the water system is filled with water and vented. 2) In the combined cleaning mode, the variable frequency air compressor 4 starts synchronously, and the air tank 5 stabilizes the pressure and stores energy; 3) The electric regulating valve and the electric proportional valve are opened to their initial opening degree; 4) The sealing adaptation structure enters the intelligent adaptive sealing process; The PLC sends an expansion command to the sealing drive mechanism to control the active radial expansion of the rubber expansion seal. The expansion pressure of the sealing body is collected in real time by pressure monitoring sensor 29; When the pressure reaches the preset sealing completion threshold, the pressurization will automatically stop, and the sealing will be determined to be successful. If the sealing pressure threshold is not reached within 10 seconds, the system will alarm and switch to emergency stop mode. 5) Once the seal is confirmed to be complete and there are no leaks or abnormalities, the pretreatment is finished, and proceed to step F. Step F (12:00 High-pressure water cleaning stage, T=10s - set total time): If the current time period is 12:00, the system enters the high-pressure water separate cleaning mode, and the air-water switching valve group keeps the water circuit open: 1) PLC controls variable frequency high-pressure water pump 30 and electric proportional valve, outputting matching water volume and water pressure; 2) Water pressure sensor 15 and water volume sensor 14 monitor in real time to form a closed-loop precise adjustment; 3) Perform the full high-pressure water cleaning process according to the time set in step C.

[0072] This stage will automatically end after the set time, and then proceed to step H.

[0073] Step G (24:00 Feng Shui joint cleaning stage, T=10s - set total duration) If the current time period is 24:00, the system enters the high-pressure water + high-pressure air combined cleaning mode, and the air-water switching valve group maintains the air-water linkage conduction: 1) PLC controls variable frequency high-pressure water pump 30 and electric proportional valve, outputting matching water volume and water pressure; 2) Synchronous control of the variable frequency air compressor 4) Electric regulating valve output matching air volume and air pressure; 3) Water pressure / water volume and air pressure / air volume sensors 8 collect data in real time and adjust to the set value in a closed loop; 4) Perform the full-process joint cleaning according to the time set in step C.

[0074] This stage will automatically end after the set time, and then proceed to step H.

[0075] Step H (System Shutdown and Reset) After the rinsing operation is completed, the PLC executes the shutdown and reset procedure: 1) Stop the variable frequency air compressor 4 and the variable frequency high-pressure water pump 30 in sequence; 2) Close all electric regulating valves, electric proportional valves, and air-water switching valve assemblies; 3) Slowly release pressure in the air and water systems; 4) When the sealing adapter structure releases the seal, the rubber expansion seal automatically retracts to its initial state; 5) The pipeline usage days counter automatically increments by 1 day and saves the current operation parameters and duration.

[0076] After the reset is complete, proceed to step I.

[0077] Step I: As part of the system's parallel detection process, the PLC continuously monitors emergency stop signals, pressure over-limit signals, abnormal seal pressure signals, and equipment fault signals during any stage of operation from step B to step H. If no abnormal signal is detected, the system continues to execute the current step logic; If an emergency stop / fault signal is detected, immediately interrupt all actuators, perform emergency pressure relief, close all valves, lock the fault code, and directly jump to step A to perform system reset.

[0078] Step J (Standby Cycle): Once all cleaning processes are completed, all actuators are reset, data recording is finished, and the system automatically returns to the initial state of step A, awaiting the next vein puncture, filling stage, pipe number selection, and start signal before entering the next cleaning cycle.

[0079] The present invention has the following beneficial effects: 1. Cleaning efficiency and cleanliness are greatly improved. By employing a combined compressed air and high-pressure water cleaning technology, along with intelligent multi-condition matching parameters, it can powerfully remove residual slurry from the inner wall of the pipeline, completely solving the problems of incomplete cleaning with traditional water and easy slurry accumulation and caking. This reduces the incidence of pipe blockage accidents from the source and ensures long-term unobstructed pipeline flow.

[0080] 2. Significant energy conservation, emission reduction, and environmental benefits. It replaces the single water rinsing mode, and precisely adjusts the water volume, air volume and rinsing time according to the working conditions, which greatly reduces the consumption of cleaning water and the underground drainage load, avoids the cleaning water from affecting the strength of the filling body, reduces energy consumption and production costs, and meets the requirements of green production in mines.

[0081] 3. Intelligent self-adaptation, highly adaptable The device intelligently outputs corresponding air volume, air pressure, water volume, water pressure and flushing time by identifying the location of the vein, the filling stage and the duration of pipeline use. No manual adjustment is required. It is adaptable to pipeline cleaning needs with different conveying distances, different slurry ratios and different usage cycles, and has strong versatility.

[0082] 4. Reliable structure, stable and safe operation. The system integrates an intelligent control module 1, a wind-water power output unit, parameter adjustment components, and a sealing adapter structure. The overall modular design ensures reliable sealing, convenient installation, and compatibility with filling pipes of different diameters. It operates stably without leakage.

[0083] 5. High degree of automation, reducing labor costs. The entire cleaning process is automated, with cyclical control of rinsing time, reducing manual operation and on-site monitoring, lowering labor intensity and human error, and improving the overall intelligence level of the filling system.

[0084] Furthermore, current technology relies on pre-stored parameter tables (pulse number, fixed air volume and water volume corresponding to the filling stage), which cannot be dynamically adjusted according to pipeline aging, seasonal changes in slurry characteristics, etc., potentially leading to insufficient or excessive cleaning. Specific problems are as follows: First, parameter response is lagging. The values ​​in the pre-stored parameter table remain fixed after deployment, making it impossible to detect the increase in inner wall roughness caused by long-term use of the pipeline or the fluctuation of slurry viscosity due to seasonal changes. As the pipeline ages, the original parameter cleaning power gradually becomes insufficient; when the slurry fluidity improves, the original parameter may over-clean, resulting in a waste of air, water, and electricity resources.

[0085] Second, the calculation rules for cleaning duration are crude. The rule of doubling the cleaning time every three days lacks process awareness. If a pipeline has accumulated a lot of deposits in the first two days, doubling the cleaning time on the third day may be barely sufficient; however, if the pipeline is used infrequently and has little deposits, doubling the cleaning time means over-cleaning. Conversely, if the pipeline has undergone unusually frequent filling operations on a particular day, resulting in a deposit volume far exceeding the usual amount, doubling the cleaning time may still be insufficient to remove all residues. The actual degree of siltation cannot be fully described by the "number of days" alone; a more refined state awareness and dynamic adjustment mechanism is needed.

[0086] This invention further develops a parameter self-tuning algorithm based on reinforcement learning: the system records subsequent pipeline operation data (such as resistance changes and blockage probability) after each cleaning, and automatically adjusts the optimal air pressure, water pressure, and flushing duration under different operating conditions, allowing the parameters to continuously evolve with environmental changes. This reduces energy consumption and media consumption, while extending pipeline lifespan, saving operating costs for large mines, and solving the aforementioned technical problems. Detailed technical solutions are as follows: The core software module of this solution is the "Reinforcement Learning Parameter Self-Learning Engine," which is deployed as an independent program in the PLC host 17 of the intelligent control module 1 or in an industrial control computer that communicates with the PLC host 17. It exchanges data with the existing PLC host 17 in real time via standard industrial protocols (such as OPCUA or ModbusTCP). The complete data closed-loop operation process is as follows: Step 1: Data Acquisition and Consolidation. During the daily operation of the device, the software module continuously receives real-time data uploaded by the PLC host 17, including the cumulative number of pipeline fillings or slurry throughput before the start of this cleaning process, the actual output air pressure / air volume / water pressure / water volume during the cleaning process, and subsequent pipeline operation data after the cleaning is completed. This data is stored systematically, forming a continuously enriched historical training sample set over time.

[0087] Step 2: After each cleaning cycle, the software module analyzes indicators such as pressure fluctuations and resistance changes during subsequent pipeline transport to automatically determine whether the cleaning has achieved the expected cleanliness target. Specifically: if the resistance along the pipeline decreases significantly and remains stable during subsequent transport, the cleaning effect is good; if the resistance decreases only slightly or rises again in a short time, the cleaning is incomplete; if there is no change in resistance after cleaning or the initial resistance is already very low, it may indicate over-cleaning.

[0088] Step 3: Reinforcement Learning Model Update. After accumulating a certain number of cleaning samples, the software module initiates the offline training process, using reinforcement learning algorithms to iteratively optimize the current control strategy. The model's goal is: when the system identifies the pipe state corresponding to a certain vein number and filling stage combination, it can output an optimal combination of air pressure, air volume, water pressure, water volume, and total duration, ensuring that the cleanliness of the pipe meets the standards while minimizing total energy consumption (air compressor power consumption + water pump power consumption + water consumption).

[0089] Step 4: Decision Parameter Distribution. After model training is complete, the software module packages the updated optimal strategy parameters (i.e., the mapping rules from state to action) into a control instruction format recognizable by the PLC host 17 and sends it back to the PLC host 17. The original fixed parameter table will be gradually replaced by dynamically generated strategy parameters. The above training can be performed periodically or manually triggered by the operator as needed, ensuring that the parameters continuously evolve with the accumulation of operating data.

[0090] Core algorithm design: I. Algorithm for Estimating Cleaning Effect In mine-filled pipeline environments, it is difficult to directly install visual or optical sensors to detect the actual cleanliness of the pipeline's inner wall. Therefore, this solution uses an indirect estimation method to infer the cleaning effect. The estimation process consists of three stages: baseline measurement before cleaning, immediate measurement after cleaning, and subsequent operational monitoring.

[0091] Phase 1: Pre-cleaning baseline measurement. Before the official start of the cleaning operation, the PLC host 17 controls the frequency converter 2 to pump cleaning water into the pipeline at a standard flow rate for a fixed duration (e.g., 30 seconds). During this period, the water pressure sensor 15 and the water flow sensor 14 record the instantaneous flow rate and inlet water pressure values ​​of the pipeline. According to the Darcy-Weisbach formula in pipeline hydraulics, the magnitude of friction resistance directly reflects the additional flow resistance increased by the adhesion of residual slurry inside the pipeline. The higher the water pressure measured under the same flow rate conditions, the greater the friction resistance of the pipeline in its current state, and the more severe the adhesion on the pipe wall; the lower the measured water pressure, the more unobstructed the pipeline is.

[0092] Phase Two: Immediate Measurement After Cleaning. After the cleaning operation is completed according to the preset parameters, the baseline flow rate test is performed again in the exact same manner, and the inlet water pressure value after cleaning is recorded. The cleaning effect is judged based on the pressure drop: the greater the drop, the more deposits are removed, and the more thorough the cleaning.

[0093] Phase 3: Subsequent Operation Monitoring. After all cleaning operations are completed, the system continuously records changes in pipeline resistance during subsequent filling operations. If the resistance value remains at the low level after cleaning throughout the subsequent operation, it indicates that the cleaning effect is stable; if the resistance value rises rapidly in a short period of time, it indicates that deposits have reformed, and the cleaning depth was insufficient; if the resistance value itself hardly decreases after cleaning, it indicates that the cleaning parameters are insufficient to effectively remove the deposits.

[0094] Using the data from the three stages mentioned above (benchmark resistance measurement data before cleaning, instantaneous resistance reduction data after cleaning, and resistance change trend data in several subsequent filling operations), the software module quantifies the "effect score" of a cleaning operation into a comprehensive index, which serves as the basis for reward signals in subsequent model training.

[0095] The specific content and acquisition methods of the three-stage data are as follows: The baseline resistance measurement data before cleaning includes the inlet water pressure value P under a fixed standard flow rate. before The frequency / flow rate setpoint of the water pump is controlled by the PLC to operate the variable frequency high-pressure water pump 30 at the same standard frequency (corresponding to a fixed flow rate, such as 30m³ / h). 3 After the water pressure sensor 15 and flow sensor have been in operation for a stable time (e.g., 30 seconds), the data is collected; the instantaneous resistance reduction data after cleaning includes the inlet water pressure value P measured at the exact same standard flow rate after cleaning. after Similarly, the water pressure is collected by water pressure sensor 15 and compared with P. before Calculate the percentage decrease ΔP = (P before -P after ) / P before×100%; The resistance change trend data in subsequent filling operations includes the inlet pressure value and the corresponding filling flow rate value during the stable delivery phase (2-5 minutes after the start) of each normal filling operation. If the filling flow rate differs from the standard flow rate, the Darcy-Weisbach formula h is applied. f =λ(L / D)(v 2 The formula ) / 2g (resistance is proportional to the square of the flow velocity) is used to convert the measured pressure to the equivalent pressure under the standard flow rate, thus obtaining a series of comparable pressure values ​​P1, P2, ..., P n (n is usually taken as 10 fillings), and the pressure recovery rate and magnitude are calculated accordingly. All data are acquired in real time by existing sensors and automatically recorded by the PLC, without the need for manual intervention.

[0096] In the Darcy-Weisbach formula above, h f Here, λ represents the head loss along the pipeline (unit: m), and λ is the friction factor (dimensionless). Under standard flow test conditions (flow velocity approximately 1.5-2.5 m / s, pipe inner wall in a state containing residual slurry after filling), λ is typically between 0.02 and 0.05, and can be calculated by solving this equation from the measured water pressure and flow rate. L is the total conveying distance of the filling pipeline (unit: m), determined by the actual length of the cross-sections in the mine, and can be obtained from the parameter storage unit by reading the selected cross-section number through the PLC. D is the inner diameter of the pipeline (unit: m), determined by the model and specifications of the filling pipeline, and is a known constant. v is the average flow velocity of the fluid in the pipeline (unit: m / s), calculated by dividing the volumetric flow rate by the cross-sectional area of ​​the pipeline. g is the acceleration due to gravity (unit: m / s²), taken as 9.8. In practical applications, since the standard flow test fixes the pumping flow rate (e.g., 30 m³ / s), the actual flow rate is determined by the measured flow rate. 3 / h), the flow velocity v is a constant value, therefore the measured water pressure value P and the friction resistance h are related. f The relationship is linearly positively correlated. By comparing the P-values ​​at different time points, the trend of changes in the roughness of the pipe inner wall can be directly determined without explicit calculation of h. f The specific values ​​are as follows. Among the parameters mentioned above, L and D are known quantities of the system, v is measured in real time by the sensor, and λ is obtained through inverse calculation. Those skilled in the art can implement the technical solution of the present invention based on the above description.

[0097] The performance score, as a positive incentive term in the reward function, is a comprehensive quantitative indicator obtained by weighted fusion calculation of the three-stage data. The specific steps are as follows: First, calculate the resistance reduction rate ΔP = (P...) based on the baseline pressure measurements before and after cleaning. before -P after ) / P before A piecewise linear function f1(ΔP) is used to map it to the interval [0,1] (for example, 1 when ΔP≥30%, 0 when ΔP≤0%, and linear interpolation in between); where Pbefore P represents the inlet water pressure (unit: kPa) under standard flow rate testing before cleaning. after The inlet water pressure (unit: kPa) is represented by the standard flow rate test after cleaning. ΔP represents the resistance reduction (dimensionless percentage, formula: ΔP=(P before -P after ) / P before ×100%).

[0098] Secondly, based on the pressure sequence P1, P2, ..., P at standard flow rates during the subsequent n filling operations (usually 10),... n Calculate the resistance recovery rate k (i.e., the pressure growth rate per unit time) and the maximum recovery amplitude δ = (P max -P after ) / P after The scores f2(k) and f3(δ) in the interval [0,1] are converted by mapping functions f2 and f3 respectively (1 for a recovery rate of less than 0.5% / day and 0 for a recovery rate of more than 2% / day; 1 for a recovery magnitude of less than 10% and 0 for a recovery magnitude of more than 50%).

[0099] Finally, the three scores are weighted and summed according to weight coefficients A, B, and C, resulting in the performance score = A·f1(ΔP) + B·f2(k) + C·f3(δ), where A + B + C = 1. Recommended values ​​are A = 0.5, B = 0.25, and C = 0.25. This score is calculated entirely from sensor data automatically collected by the PLC and serves as an immediate reward signal for the reinforcement learning model after each cleaning, driving the strategy towards more efficient and thorough cleaning.

[0100] f1(ΔP) is the score mapped based on the resistance reduction (0≤f1≤1), f2(k) is the score mapped based on the recovery rate (0≤f2≤1), and f3(δ) is the score mapped based on the maximum recovery amplitude (0≤f3≤1).

[0101] The rate of increase in resistance, k (i.e., the pressure increase rate per unit time), is calculated as follows: First, record the specific time points of each filling operation (based on the completion time of cleaning, in days), obtaining the time series t1, t2, ..., t n (n is generally taken as 10), and the equivalent pressure values ​​P1, P2, ..., P corresponding to the standard flow rate at each filling time. n Then, for the data points (t) i ,P i Linear fitting is performed using the least squares method, and the fitted line equation is P = k·t + b, where the slope k is the pressure growth rate per unit time (unit: pressure / day). In simplified engineering implementation, the growth rate k can also be calculated directly using the first and last two points: k = (Pn -P1) / (t n -t1), but the linear fitting method can better eliminate the interference of single outlier data. A positive value of the calculated k indicates that the pressure is rising, while a negative value indicates that the pressure continues to fall or remains stable. Its value directly reflects the speed of the resistance recovery, and is then used to calculate the score of the f2 mapping function.

[0102] Among them, the pressure sequence P1, P2, ..., P n This represents the equivalent pressure value (unit: kPa) converted to standard flow rate in the subsequent n (usually 10) filling operations; time series t1, t2, ..., t n Indicates the corresponding filling time (based on the cleaning completion time, unit: days); P max δ represents the maximum pressure value among these n pressure values ​​(unit: kPa), δ represents the maximum recovery amplitude (dimensionless), and k represents the pressure growth rate per unit time (unit: kPa / day), given by the slope in the linear fitting equation P=k·t+b.

[0103] The resistance recovery rate thresholds (0.5% / day and 2% / day) were obtained through statistical analysis of three consecutive months of operational data from a large iron ore filling pipeline. Under conditions of good cleaning effectiveness, the average daily increase in equivalent pressure (converted to standard flow rate) over the subsequent 10 filling operations did not exceed 0.5% / day; under conditions of insufficient cleaning effectiveness, the daily increase could reach over 2% / day. Therefore, when the recovery rate is below 0.5% / day, it is considered "stable cleaning effectiveness," and the mapping function f2 scores 1 point; when the recovery rate is above 2% / day, it is considered "unstable cleaning effectiveness," and the mapping function f2 scores 0 points; when it falls between these two values, the score is calculated using linear interpolation, i.e., f2(k) = (k... max -k) / (k max -k min )×1, where k is the measured recovery rate (unit: pressure / day), k min =0.5% / day, k max =2% / day. The above threshold can be adjusted via the human-machine interface unit according to the actual working conditions of the mine. For example, for cross-cuts with long conveying distances (such as cross-cuts 6-10), due to the inherently large frictional resistance, the relative sensitivity of pressure recovery differs, and k can be adjusted accordingly. min and k max Appropriately increase the threshold; for filling stages with higher slurry viscosity (such as bottom filling), the threshold can be appropriately reduced to increase the strictness of judgment due to the stronger adhesion of residues. Operators can gradually optimize the threshold setting through the accumulation of field experience.

[0104] The formula for calculating the effectiveness score adopts a weighted linear combination form, designed based on the following considerations: the comprehensive evaluation of the cleaning effect should simultaneously consider three dimensions: instantaneous effect (resistance reduction), short-term stability (resistance recovery rate), and long-term stability (maximum recovery amplitude). Analysis of 156 sets of cleaning data collected on-site using principal component analysis revealed that the cumulative variance contribution rate of the first three principal components reached 92.6%, indicating that these three dimensions are sufficient to characterize the main features of the cleaning effect, and there is no need to introduce higher-dimensional parameters. The recommended values ​​for the weight coefficients A, B, and C are A=0.5, B=0.25, and C=0.25, determined using the analytic hierarchy process: five experts in the field of mine backfilling were invited to conduct pairwise comparisons of the importance of the three evaluation indicators, constructing a judgment matrix and calculating the weight vector of each indicator. The consistency test showed CR=0.037<0.1, indicating satisfactory consistency. Among these factors, the resistance reduction rate (instantaneous effect) is considered the most important (weight 0.5) because it directly reflects the immediate effect of the cleaning operation; the resistance recovery rate (short-term stability) is the second most important (weight 0.25); and the maximum recovery amplitude (long-term stability) is the third most important (weight 0.25). Operators can adjust the weighting coefficients through the human-machine interface unit according to the specific needs of the actual mine. For example, when pipe blockage accidents occur frequently, the weight of the resistance recovery amplitude can be appropriately increased to more rigorously evaluate the cleaning quality.

[0105] Reinforcement learning state space and action space: This scheme uses the Markov decision process framework to describe the cleaning parameter decision problem.

[0106] The state space is defined as a combination of the following multidimensional vectors: Pipeline number: a discrete variable, with values ​​from 1 to 10, corresponding to different pipeline transport distances, and determining the basic hydraulic characteristics.

[0107] Filling stage: a discrete variable, with values ​​for bottom filling, middle filling, or top filling, corresponding to different slurry ash-sand ratios and concentrations. A higher ash-sand ratio indicates more cementitious components and a stronger tendency for the slurry to adhere to the inner wall of the pipe; a higher slurry concentration results in a faster settling rate.

[0108] Cumulative slurry throughput since the last cleaning: a continuous variable, representing the total volume of filling slurry transported in the pipeline since the last cleaning. This value can be obtained from the filling production management system and is the most direct indicator of the pipeline's current "contamination status." A larger throughput indicates more severe pipe wall deposition.

[0109] Current pipeline reference resistance coefficient: a continuous variable, estimated by substituting water pressure and flow rate data obtained during the baseline measurement phase before cleaning into the Darcy-Weisbach formula. A higher resistance coefficient indicates higher pipe wall roughness or the presence of deposits.

[0110] The action space is defined as a set of the following multidimensional actions, each corresponding to a set of executable cleaning parameters: Air pressure: Continuous value, adjusted by variable frequency air compressor 4.

[0111] Air volume: Continuous value, controlled by variable frequency air compressor 4 and electric regulating valve.

[0112] Water pressure: Continuous value, controlled by a variable frequency high-pressure water pump 30 and an electric regulating valve.

[0113] Water volume: Continuous value, controlled by variable frequency high-pressure water pump 30.

[0114] Total cleaning time: continuous value, in seconds.

[0115] In engineering practice, the actual range of values ​​for the above continuous variables is limited by the physical properties of the frequency converter 2 and the electric regulating valve. The software module will impose constraints based on these boundary conditions during model training.

[0116] Reward function design: The reward function is the core of guiding policy optimization in reinforcement learning models. This scheme defines the total reward obtained by the system after each complete cleansing as a weighted sum of three components, allowing the model to pursue cleansing effectiveness while considering operational costs.

[0117] Component 1 (Cleaning Effectiveness Score): Calculated based on the comprehensive measurement data from the three stages mentioned above. If the baseline flow test after cleaning shows that the resistance reduction reaches or exceeds the preset target threshold (e.g., 30%), and the resistance does not rebound significantly in subsequent fillings, this component takes a positive value, with a higher score for a larger reduction. If the resistance reduction fails to reach the target threshold, this component takes a zero or negative value. If the resistance value increases instead of decreasing after cleaning (indicating an operational anomaly), this component takes a larger negative value as a penalty.

[0118] Component Two (Energy Consumption Score): The total electrical energy consumed during the cleaning process, including the operating power consumption of the variable frequency air compressor 4 and the variable frequency high-pressure water pump 30. This component has a negative value in the total reward, meaning the more electrical energy consumed, the more points are deducted.

[0119] Component 3 (Water Consumption Score): The total amount of water consumed during the cleaning process, in liters. This component has a negative value in the total reward, meaning the more water consumed, the more points are deducted.

[0120] The formula for calculating the total reward is: Total reward = w1 × effect score + w2 × (-total energy consumption) + w3 × (-total water consumption).

[0121] w1, w2, and w3 are weighting coefficients that can be adjusted according to the specific operating cost structure of the mine. If a mine has high electricity costs and relatively cheap water resources, the weight of w2 can be appropriately increased; conversely, if the mine has scarce water resources and high drainage costs, the weight of w3 can be appropriately increased. These weighting parameters can be manually configured by the operator through the human-machine interface of the host computer.

[0122] The performance score is calculated using the method described above (value 0-1). The total energy consumption is the electrical energy consumed during the cleaning process (unit: kWh), and the total water consumption is the amount of water consumed during the cleaning process (unit: L).

[0123] The "-" in the formula represents a negative sign, meaning that the total energy consumption and total water consumption are used as penalty items. The larger the value, the more the total reward is deducted, thus guiding the model to minimize resource consumption while ensuring the cleaning effect.

[0124] The aforementioned 30% resistance reduction target threshold was determined based on field test data from mine backfilling pipelines. Tests showed that when the inlet water pressure reduction at standard flow rate reached over 30%, more than 85% of the residual slurry adhering to the pipeline inner wall was stripped away. The friction loss during subsequent backfilling operations could be stably maintained at a low level for over 72 hours, reducing the pipe blockage rate by over 90%. When the reduction was less than 15%, a large amount of residue remained in the pipeline, causing the resistance to rise rapidly during subsequent backfilling operations, significantly increasing the risk of blockage. Therefore, 30% was set as the engineering threshold for determining "cleaning qualification." It should be noted that this threshold is not fixed—for different pipelines in different mines, operators can adjust the threshold within the range of 15%-45% based on actual operating experience through the human-machine interface. In the effect score calculation, this threshold is used for linear interpolation mapping of the f1 function: 1 is obtained when ΔP ≥ threshold, 0 is obtained when ΔP ≤ 0%, and intermediate linear interpolation is used. In the reward function, the resistance reduction target threshold is also used to determine whether the cleaning effect score meets the qualified standard (the effect score takes a higher positive value when ΔP reaches the threshold). The two thresholds are logically consistent.

[0125] Reinforcement learning training strategies: This scheme selects the Q-learning algorithm from value-based reinforcement learning methods as the core training framework. Q-learning is a model-independent reinforcement learning algorithm applicable to discrete state and discrete action spaces. Its basic idea is to maintain a "state-action" value table Q(s,a), recording the maximum cumulative reward expected to be obtained by performing action a in state s. Through continuous interaction with the environment, the Q-value gradually converges, and the action that maximizes the Q-value can be directly selected from the table as the optimal policy.

[0126] The Q-value update rule uses the standard Bellman equation: Q(s,a)←Q(s,a)+α×[R(s,a)+γ×max a′ Q(s′,a′)-Q(s,a)].

[0127] The meanings of each symbol are as follows: Q(s,a): The current estimated value corresponding to the action a performed in the current state s.

[0128] α: Learning rate, a decimal between 0 and 1. The learning rate determines the extent to which new information corrects the original estimate in a single update: the closer the learning rate is to 1, the faster the model responds to new samples; the closer the learning rate is to 0, the stronger the model's dependence on historical experience. In engineering practice, the learning rate is usually taken as a value between 0.1 and 0.2.

[0129] R(s,a): The immediate reward obtained after performing action a from state s to transition to the new state, calculated by the total reward function described in the previous section.

[0130] γ: Discount factor, also a decimal between 0 and 1. The discount factor determines the degree to which future rewards influence current decisions: the closer the discount factor is to 1, the more the model values ​​long-term returns; the closer the discount factor is to 0, the more the model focuses on immediate rewards. In the scenario of filling pipeline cleaning, the long-term impact of cleaning effectiveness on subsequent production far outweighs the direct cost savings of a single cleaning, therefore, a higher value for the discount factor between 0.9 and 0.95 is recommended.

[0131] s′: The new state the system enters after executing action a.

[0132] a′ represents one of the possible actions that the system can perform after entering a new state s′, and is used to iterate through and find the maximum Q value.

[0133] max a′ Q(s′,a′): This represents the maximum value of Q(s′,a′) for all possible actions a′ in the new state s′ (where a′ belongs to A(s′), and As′ is the set of all possible actions in the new state s′). In engineering implementation, since both the state space and action space are finite after discretization, this maximum value is obtained by traversing all possible actions. This term in the Bellman equation represents the "discounted value of the maximum possible future reward," reflecting the core idea of ​​the Q-learning algorithm to comprehensively evaluate an action from the perspectives of both "current reward" and "future optimal reward."

[0134] In this technical improvement, Q(s,a) represents the expected cumulative reward (i.e., the state-action value function) that can be obtained by performing action a in state s, and is used to evaluate the long-term benefits of a certain set of cleaning parameters under the current pipeline conditions. State s is a multi-dimensional discretized vector composed of the vein number, filling stage, cumulative slurry throughput after the last cleaning (discretized into low / medium / high grade) and the current pipeline reference resistance coefficient (discretized into five levels), which fully describes the current operating conditions and contamination level of the pipeline to be cleaned. Action a is a set of values ​​in a multi-dimensional continuous action space, specifically including the air pressure, air volume, water pressure, water volume and total cleaning time during cleaning. These five parameters together determine the execution strategy of this cleaning.

[0135] s′ represents the new state the system enters after performing action a (i.e., using a specific set of air pressure, air volume, water pressure, water volume, and cleaning duration) in state s and completing cleaning and several subsequent filling operations. This new state is composed of updated parameters such as the cumulative slurry flow through the pipeline and the resistance coefficient measured after cleaning, which are re-discreteized. a′ represents one of all possible actions (i.e., all possible combinations of cleaning parameters) that the system can perform when in the new state s′, and is used to iterate through and calculate the maximum future reward value max in the update formula. a′ Q(s′,a′); R(s,a) represents the immediate reward obtained after performing action a in state s. In this scheme, it is specifically reflected as the total reward value calculated by the cleaning effect score, total energy consumption and total water consumption. Total reward = w1×effect score + w2×(-total energy consumption) + w3×(-total water consumption). This value is directly used as the feedback signal after an interaction to update Q(s,a).

[0136] The intuitive meaning of the above update rule is that in each training iteration, the model calculates a target Q-value using "the current immediate reward plus the maximum possible future reward," and then adjusts the current estimate a small step towards that target value (the adjustment magnitude is controlled by the learning rate α). Through repeated iterations with a large number of samples, the Q-value table gradually approaches the true policy value.

[0137] In engineering implementation, the continuous variables in the state space (slurry throughput and resistance coefficient) are first discretized into several intervals, each interval corresponding to a discrete state number, to ensure that the size of the Q-value table is within a manageable range. For example, the slurry throughput is divided into three levels, "low," "medium," and "high," according to the distribution of historical data, and the resistance coefficient is divided into five levels according to its numerical value. Combined with the discrete vein numbers and filling stages, this forms an enumeration space of no more than 150 finite discrete states, which is perfectly suitable for offline training of Q-learning.

[0138] The balance between exploration and utilization: Reinforcement learning models need to maintain a balance between "utilizing existing experience" (selecting actions currently considered optimal) and "exploring unknown possibilities" (trying actions not yet fully evaluated). Otherwise, the policy may prematurely converge to a local optimum and miss the true optimal solution. This scheme adopts the classic ε-greedy strategy: during the training phase, when selecting an action, an action is randomly selected with probability ε for exploration, and the action with the largest Q-value is selected according to the current Q-value table with probability 1-ε. The initial value of ε is set to 0.3, and it gradually decays to a small final value (e.g., 0.05) as the number of training iterations increases. This decay mechanism allows the model to explore various parameter combinations more in the early stages of training, and gradually converges to a stable policy as experience accumulates.

[0139] Strategy Deployment and Continuous Learning Cycle: The entire training process described above operates offline, without interfering with the online operation of the device. After training, the software module extracts the optimal policy function π(s) from the Q-value table, i.e., for each discrete state s, the optimal action a = argmax_{a∈A}Q(s,a). These mapping rules are then encoded into a control parameter table recognizable by the PLC host 17. As the device continues to operate and new cleaned data accumulates, the software module periodically initiates a new round of offline training, updating the Q-value table with new samples and regenerating the optimal policy parameters, which are then sent back to the PLC host 17. This cycle of "offline training—online deployment—data feedback—retraining" makes the parameter self-optimization process a continuously evolving closed loop.

[0140] This solution does not alter any hardware components of the original invention, including the variable frequency air compressor 4, air tank 5, variable frequency high-pressure water pump 30, electric regulating valve, electric proportional valve, air volume sensor 8, air pressure sensor 9, water volume sensor 14, water pressure sensor 15, air-water switching valve, filling pipe sealing adapter 3, etc. All new functions are implemented through software modules.

[0141] The reinforcement learning module in this solution obtains status information by reading existing sensor data from the PLC host 17, and executes the cleaning strategy by issuing action commands to the PLC host 17. The communication between the module and the PLC host 17 adopts existing industrial communication protocols without introducing new hardware interfaces or external dependencies.

[0142] This software solution only takes over the setting of cleaning parameters in "automatic parameter decision" mode. When the system is in manual mode or a sensor failure occurs, the original fixed parameter table can still be used as a backup solution to ensure that the device can maintain basic cleaning capabilities under any circumstances and will not shut down due to software module failure.

[0143] Before its official deployment and launch, this software solution needs to be verified offline through historical data playback simulation.

[0144] Verification Method: Real cleaning records from the past three or six months will be extracted from the device's operation logs. These records include the pipeline status before each cleaning (slurry throughput, baseline resistance coefficient), the action parameters executed during the cleaning process (air pressure, air volume, water pressure, water volume, duration), and the post-cleaning effects and energy consumption data. This historical data will be used to train the reinforcement learning model. The model's recommended action parameters will be compared with the actual historical action parameters. If the model's recommended air pressure, water volume, or duration is lower than the historical values, but the resistance reduction after cleaning still reaches or exceeds the historical level, it proves that the model has found a more energy-efficient parameter combination. If the model's recommended parameters are higher than the historical values, but the resistance reduction after cleaning is greater and subsequent operation is more stable, it proves that the historical cleaning intensity was insufficient.

[0145] The evaluation metrics mainly include: the percentage reduction in average energy consumption per cleaning cycle, the percentage reduction in average water consumption per cleaning cycle, the rate of pipe resistance compliance after cleaning (i.e., the percentage of cleaning cycles in which the resistance reduction reaches the target threshold), and the number of training samples required for strategy convergence. Based on simulation experience, it is expected that after three to five rounds of iterative training, the strategy recommended by the model can reduce average energy consumption and water consumption by 10% to 15%, while reducing the probability of cleaning failure by approximately 50%.

[0146] Actual testing was conducted on a backfill pipeline (vessel number 5, corresponding to a transport distance of 2.0 km, pipeline inner diameter 150 mm) in an iron ore mine. The standard flow rate before cleaning (30 m³ / h) was... 3 Inlet pressure P at / h) before =850kPa; The device of this invention is used to perform air-water combined cleaning (PLC automatically matches parameters according to working conditions: air pressure 0.60MPa, air volume 8.0m³ / h). 3 / min, water pressure 1.20MPa, water volume 30m 3 / h, cleaning time 45min). After cleaning, the inlet pressure P was obtained by testing again with the same standard flow rate. after =560kPa, the calculated resistance reduction ΔP=(850-560) / 850×100%=34.1%, exceeding the preset qualified threshold of 30%, and the effect score component f1 is taken as 1.0 by linear interpolation. In the subsequent 10 filling operations, the equivalent pressure sequence P1~P at the standard flow rate at the initial stage of each filling (2-5 minutes after the start) is calculated. 10The pressure increases (in kPa) are: 570, 575, 580, 578, 585, 590, 588, 595, 600, 605; the corresponding time points (with the end of cleaning as time 0, in days) are: 0.5, 1.0, 1.5, 2.0, 2.5, 3.0, 3.5, 4.0, 4.5, 5.0. Using the least squares linear fitting method, the pressure increase rate k = 7.2 kPa / day is obtained. after The relative growth rate, calculated based on the baseline, is 1.2% / day, falling between the lower limit of 0.5% / day and the upper limit of 2% / day. Therefore, f2 = (2% - 1.2%) / (2% - 0.5%) = 0.53. The maximum pressure rebound is δ = (605 - 560) / 560 ≈ 8.0%, which is below the 10% threshold. Therefore, f3 = 1.0. Taking weights A = 0.5, B = 0.25, and C = 0.25, the calculated score is 0.5 × 1.0 + 0.25 × 0.53 + 0.25 × 1.0 = 0.8825. The total energy consumption for this cleaning (air compressor + water pump) is 200 kWh, and the water consumption is 22.5 m³. 3 If the reward weights are set as w1=0.6, w2=0.3, w3=0.1, then the total reward = 0.6×0.8825 + 0.3×(-200) + 0.1×(-22.5) = -61.72. After 5 rounds of offline reinforcement learning training, the model recommends the following new parameters: wind pressure 0.55MPa, air volume 7.5m³ / h. 3 / min, water pressure 1.15MPa, water volume 28m 3 / h, duration 40min; P was obtained after execution. before =860kPa, P after =580kPa, ΔP=32.6%, performance score 0.92, energy consumption reduced to 185kWh, water consumption reduced to 18.7m 3 The total reward increased to -55.31. The above data proves that the self-learning algorithm of this invention can effectively reduce energy consumption by about 7.5% and water consumption by about 16.9% while maintaining the cleaning effect (resistance reduction of more than 30%), demonstrating significant economic benefits.

[0147] A further solution of this invention continuously optimizes the combination of air pressure, water volume, and cleaning time under various operating conditions through reinforcement learning algorithms, which can effectively reduce energy consumption and media consumption, and save electricity and water treatment costs. At the same time, the precise on-demand cleaning strategy avoids local scaling and accelerated wear caused by insufficient cleaning and pipe wall stress fatigue caused by excessive cleaning, which helps to extend the service life of pipelines. In addition, parameter configuration, effect evaluation, and strategy optimization do not require manual intervention, which not only reduces the burden of on-site operation, but also realizes the intelligent improvement from fixed parameter cleaning to data-driven self-optimizing cleaning.

Claims

1. An intelligent compressed air cleaning device for filling pipelines in underground mining, characterized in that, The device includes an intelligent control module (1), a frequency converter (2), an air circuit assembly, a water circuit assembly, an air-water switching valve (10), and a filling pipe sealing adapter (3). The intelligent control module (1) is connected to the frequency converter (2), the air circuit assembly, the water circuit assembly, the air-water switching valve (10), and the filling pipe sealing adapter (3). The frequency converter (2) is connected to the air circuit assembly and the water circuit assembly. The air circuit assembly and the water circuit assembly are connected to the air-water switching valve (10). The air-water switching valve (10) is connected to the filling pipe sealing adapter (3).

2. The intelligent compressed air cleaning device for filling pipelines in underground mining as described in claim 1, characterized in that, The air circuit assembly includes a variable frequency air compressor (4), an air tank (5), a main air circuit (6), an electric regulating valve (7), an air volume sensor (8), and an air pressure sensor (9). The variable frequency drive (2) is connected to the variable frequency air compressor (4), which is connected to the air tank (5). The air tank (5) is connected to the main air circuit (6), which is connected to the electric regulating valve (7). The electric regulating valve (7) is connected to the air-water switching valve (10). The main air circuit (6) is connected to the air volume sensor (8), the air pressure sensor (9), and the electric proportional valve (11).

3. The intelligent compressed air cleaning device for filling pipelines in underground mining as described in claim 1, characterized in that, The water circuit assembly includes a variable frequency high-pressure water pump (30), a water main pipe (12), an electric regulating valve II (13), a water volume sensor (14), and a water pressure sensor (15). The variable frequency drive (2) is connected to the variable frequency high-pressure water pump (30), which is connected to the water main pipe (12). The water main pipe (12) is connected to the electric regulating valve II (13), which is connected to the air-water switching valve (10). The water main pipe (12) is connected to the water volume sensor (14), the water pressure sensor (15), and the electric proportional valve II (16).

4. The intelligent compressed air cleaning device for filling pipelines in underground mining as described in claim 1, characterized in that, The intelligent control module (1) includes a PLC host (17), a human-machine interaction unit (18), a pulse selection unit (19), a filling stage selection unit (20), a clock timing unit (21), a parameter storage unit (22), and a sealing adapter control unit (23). The human-machine interaction unit (18), the pulse selection unit (19), the filling stage selection unit (20), the clock timing unit (21), the parameter storage unit (22), and the sealing adapter control unit (23) are all connected to the PLC host (17). The human-machine interaction unit (18) is connected to the pulse selection unit (19) and the filling stage selection unit (20) respectively.

5. The intelligent compressed air cleaning device for filling pipelines in underground mining as described in claim 1, characterized in that, The filling pipe sealing adapter (3) includes a main water inlet pipe (31) and a mechanical clamping sleeve (24) screwed onto the outside of the main water inlet pipe (31). One end of the mechanical clamping sleeve (24) is provided with a rotating handle (25) and an anti-detachment safety chain (28) is provided at the rotating handle (25). The other end is a compression head (26). The compression head (26) is adjacent to the intelligent control rubber expansion seal (27). The intelligent control rubber expansion seal (27) is located at one end of the main water inlet pipe (31). A pressure monitoring sensor (29) is provided on the intelligent control rubber expansion seal (27).

6. A smart compressed air cleaning method for filling pipelines in underground mining, characterized in that, Includes the following steps: Step A: Device and system initialization; Step B: The operator selects the vein number, filling stage and pipe number to be cleaned through the human-machine interaction unit (18) of the intelligent control module (1); Step C: The intelligent control module (1) receives the above selection signal, reads the current time through the clock timing unit (21), and reads the cumulative number of days the selected pipeline has been used; the intelligent control module (1) determines whether to use high-pressure water for cleaning alone or a combination of air and water cleaning based on the current time; and determines the cleaning duration based on the cumulative number of days the pipeline has been used. 1) If the current time is 12: 00: Perform a single high-pressure water cleaning; 2) If the current time is 24:00: Perform a combined high-pressure water + high-pressure air cleaning; 3) If the pipeline has been used for a cumulative period of 3 days: the corresponding cleaning time will automatically double; Step D: The PLC host (17) of the intelligent control module (1) calls the preset data in the parameter storage unit (22) according to the signals from the pulse selection unit (19) and the filling stage selection unit (20), and automatically matches and outputs the target parameters: 1) Match the corresponding water volume and water pressure parameters according to the vein number and filling stage; 2) If it is a 24:00 joint cleaning mode, the corresponding air volume and air pressure parameters will be matched synchronously; 3) Send the final parameters to the frequency converter (2), electric regulating valve one, electric regulating valve two, electric proportional valve one, and electric proportional valve two to complete the parameter setting of the actuator; This step is the main timer T=0min stage; Step E, System Preprocessing Stage, T=0-10s, lasting 10 seconds to ensure equipment stability. Preprocessing includes: 1) The variable frequency high-pressure water pump (30) starts, and the water system is filled with water and vented; 2) If it is a combined cleaning mode, the variable frequency air compressor (4) starts synchronously, and the air tank (5) stabilizes the pressure and stores energy; 3) The electric regulating valve and the electric proportional valve are opened to their initial opening degree; 4) The sealing adaptation structure enters the intelligent adaptive sealing process; 5) Confirm that the seal is complete and there are no leaks or abnormalities; the pretreatment is now complete. Step F: If the current time period is 12:00, T = 10s - the set total duration, the system enters the high-pressure water separate cleaning mode, the air-water switching valve group keeps the water circuit open, this stage will run until the set time ends automatically, and then proceed to step H; Step G: The current time period is 24:

00. T = 10s - set total duration. The system enters the high-pressure water + high-pressure air combined cleaning mode. The air-water switching valve group keeps the air-water linkage conduction. This stage will run until the set time ends automatically and proceed to step H. After step H, the rinsing operation is completed, the PLC host (17) executes the shutdown and reset process; Step 1: Parallel detection throughout the entire system process; Step J: Cyclic standby.

7. The intelligent compressed air cleaning method for filling pipelines in underground mining as described in claim 6, characterized in that, Step C determines the cleaning duration based on the cumulative number of days the pipeline has been used, including: If the current time is 12:00, the basic rinsing time is 15 minutes; If the current time is 24:00, the basic rinsing time is 30 minutes; If the pipeline is used for a cumulative period of 3 days, the corresponding cleaning time will automatically double, i.e., 12:00 becomes 30 minutes and 24:00 becomes 60 minutes.

8. The intelligent compressed air cleaning method for filling pipelines in underground mining as described in claim 6, characterized in that, In step E, the sealing adapter structure enters the intelligent adaptive sealing process, including: the PLC host (17) sends an expansion command to the sealing drive mechanism to control the rubber expansion seal to actively expand radially; the pressure monitoring sensor (29) collects the expansion pressure of the seal in real time; when the pressure reaches the preset sealing completion threshold, the pressurization is automatically stopped and the sealing is determined to be successful; if the sealing pressure threshold is not reached within 10 seconds, the system alarms and jumps to emergency stop processing. Step F includes the following high-pressure water cleaning modes: 1) The PLC host (17) controls the variable frequency high-pressure water pump (30) and the electric proportional valve, and outputs matching water volume and water pressure; 2) Water pressure sensor (15) and water volume sensor (14) monitor in real time to form a closed-loop precise adjustment; 3) Perform the full high-pressure water cleaning process according to the time set in step C; Step G, the combined high-pressure water and high-pressure air cleaning mode, includes: 1) The PLC host (17) controls the variable frequency high-pressure water pump (30) and the electric proportional valve, and outputs matching water volume and water pressure; 2) Synchronous control of variable frequency air compressor (4) Electric regulating valve output matching air volume and air pressure; 3) The water pressure sensor (15), water volume sensor (14), wind pressure sensor (9), and air volume sensor (8) collect data in real time and adjust them to the set value in a closed loop. 4) Perform the full-process joint cleaning according to the time set in step C.

9. The intelligent compressed air cleaning method for backfilling pipelines in underground mining as described in claim 6, characterized in that, Step H includes: 1) Stop the variable frequency air compressor (4) and the variable frequency high-pressure water pump (30) in sequence; 2) Close all electric regulating valves, electric proportional valves, and air-water switching valve assemblies; 3) Slowly release pressure in the air and water systems; 4) When the sealing adapter structure releases the seal, the rubber expansion seal automatically retracts to its initial state; 5) The pipeline usage days counter automatically increments by 1 day and saves the current operation parameters and duration; 6) After the reset is complete, proceed to step J; Step J includes: all cleaning processes are completed, all actuators are reset, data recording is completed, the system automatically returns to the initial state of Step A, and waits for the next vein puncture, filling stage, pipeline number selection and start signal to enter the next cleaning cycle.

10. The intelligent compressed air cleaning method for backfilling pipelines in underground mining as described in claim 6, characterized in that, Step I includes: As part of the parallel detection process throughout the entire system, during any stage of operation from step B to step H, the PLC host (17) continuously detects emergency stop signals, pressure over-limit signals, abnormal seal pressure signals, and equipment fault signals in real time. If no abnormal signal is detected, the system continues to execute the current step logic; If an emergency stop / fault signal is detected, immediately interrupt all actuators, perform emergency pressure relief, close all valves, lock the fault code, and directly jump to step A to perform system reset.

Images