Equipment and method for circularly cleaning sludge in irrigation canal

Abstract

The invention provides equipment and a method for circularly cleaning sludge in an irrigation canal, and belongs to the technical field of sludge cleaning equipment.The equipment for circularly cleaning the sludge in the irrigation canal comprises a water storage pond; an opening is formed in the top of the silt storage pool, the silt storage pool is fixedly installed in the water storage pool, and a water storage tank is reserved between the silt storage pool and the inner wall of the water storage pool; the two overflow ports are formed in the two side ends of the silt storage pool respectively; the silt inlet pipeline is arranged on the upper side of the silt storage tank; through the design of the water storage tank and the silt storage tank, after silt and water enter the silt storage tank, the silt is deposited at the bottom, and the water is accumulated above. When water is accumulated to submerge the edge of the uppermost side of the inclined baffle, the water overflows from the upper surface of the inclined baffle into the water storage tank between the water storage pond and the silt storage pond and is finally discharged into the water channel through the water outlet pipeline, water circulation is achieved, waste of water resources is avoided, and meanwhile silt can be effectively cleaned.

Description

Technical Field

[0001] The invention belongs to the technical field of sludge cleaning equipment, and in particular relates to irrigation canal sludge circulation cleaning equipment and method. Background Art

[0002] Farmland irrigation often uses water from rivers, which is introduced into farmland through "ditches" dug on the ground. Canals are artificially dug waterways, which are divided into main canals and branch canals. Main canals and branch canals are generally built with stone or cement. However, silt will accumulate at the bottom of the canal after long-term use, affecting its normal use. When dredging the canal, a sludge pump is often used to enter the bottom of the canal for cleaning.

[0003] In the prior art, when a sludge pump is used to desilt the bottom of a canal, the sludge and water are generally pumped out together, and then the pumped water is also processed and discharged together with the sludge. This method is likely to cause a decrease in the water level or an increase in the water level in the sludge treatment area when the sludge is circulated, and the water cannot be circulated and discharged into the canal again. Therefore, we provide an irrigation canal sludge circulation cleaning device and method to solve the above problems. Summary of the invention

[0004] The purpose of the present invention is to provide an irrigation canal sludge recycling cleaning device and method, aiming to solve the problem in the prior art that when a sludge pump is placed at the bottom of a canal for dredging, the sludge and water are generally pumped out together, and then the pumped water is also processed and discharged together with the sludge. This method is likely to cause a decrease in the water level or an increase in the water level at the sludge processing site when the sludge is recycled, and the water cannot be recycled and discharged into the canal again.

[0005] To achieve the above object, the present invention provides the following technical solutions:

[0006] An irrigation canal sludge circulation cleaning device, comprising:

[0007] water storage tanks;

[0008] A silt storage tank, the top of which is open, is fixedly installed in the water storage tank and has a water storage tank between it and the inner wall of the water storage tank;

[0009] Two overflow ports are provided, which are respectively opened at the two side ends of the silt storage tank;

[0010] A silt inlet pipeline is arranged on the upper side of the silt storage tank;

[0011] A transmission pipeline, one end of which is fixed to the top of the silt inlet pipeline;

[0012] A crushing box, fixed to the other end of the transmission pipeline and connected to the transmission pipeline;

[0013] A silt suction pipe, one end of which is fixed to and connected with the crushing box, and the other end of which is located in the water channel;

[0014] A sludge pump installed at the end of the sludge suction pipe in the canal;

[0015] A crushing mechanism, disposed in the crushing box, for crushing the sludge entering the silt storage tank;

[0016] A stirring mechanism is provided in the silt storage tank to prevent sludge from agglomerating and blocking the pipeline; and

[0017] The driving mechanism is connected with both the crushing mechanism and the stirring mechanism to realize the operation of both.

[0018] As a preferred solution of the present invention, a circular support frame is fixed on the inner wall of the water storage tank, a filter plate is arranged on the circular support frame, and a plurality of evenly distributed filter holes are opened on the top of the filter plate.

[0019] As a preferred solution of the present invention, a water outlet pipe is provided at one side end of the water storage tank, a silt outlet pipe is fixed at the bottom of the silt storage tank, and a valve is fixedly installed on the silt outlet pipe.

[0020] As a preferred solution of the present invention, inclined baffles are obliquely fixed on the inner walls of both sides of the silt storage tank, and the silt inlet pipeline is located between the two inclined baffles.

[0021] As a preferred solution of the present invention, a fixing plate is fixed on the top of the silt storage tank, and the bottom of the silt inlet pipe passes through and extends to the bottom of the fixing plate.

[0022] As a preferred solution of the present invention, the crushing mechanism includes a roller and a crushing blade, the roller is rotatably connected between the inner walls of both sides of the crushing box, and a plurality of crushing blades are provided, and the plurality of crushing blades are evenly fixed on the circumferential surface of the roller.

[0023] As a preferred solution of the present invention, the stirring mechanism includes a stirring shaft and a breaking blade, one end of the stirring shaft is rotatably connected to an inner wall of one side of the silt storage tank, the other end of the stirring shaft movably penetrates through the outside of the water storage tank and extends, and a plurality of the breaking blades are provided, and the plurality of breaking blades are evenly fixed on the circumferential surface of the stirring shaft.

[0024] As a preferred solution of the present invention, the driving mechanism includes:

[0025] A motor is fixed to one side of the crushing box, and an output shaft of the motor movably passes through the other side of the crushing box and extends outward;

[0026] The roller is fixed on the output shaft of the motor;

[0027] A small I-shaped wheel is fixed on the circumferential surface of the output shaft of the motor;

[0028] A large I-shaped wheel fixed on the circumferential surface of the extended end of the stirring shaft; and

[0029] A belt is connected between the circumferential surfaces of the small I-shaped wheel and the large I-shaped wheel.

[0030] As a preferred solution of the present invention, a box cover is movably provided on the top of the crushing box, and fixing blocks are fixed to both side ends of the crushing box and the box cover, and two adjacent fixing blocks are fixed by bolts.

[0031] As a preferred solution of the present invention, a method for circulating and cleaning sludge in an irrigation canal is applied to an irrigation canal sludge circulating and cleaning device as described in claim 9, and is characterized in that it comprises the following steps:

[0032] S1. First, place the sludge pump in the canal, and then start the sludge pump. The sludge pump can suck the sludge and water in the canal into the crushing box through the sludge suction pipe, and then discharge it into the sludge inlet pipe through the transmission pipe, and finally enter the bottom of the sludge storage tank. Since the sludge will be deposited at the bottom of the sludge storage tank, the water entering the sludge storage tank will also accumulate in the sludge storage tank;

[0033] S2. When the water in the silt storage tank accumulates and submerges the uppermost edge of the inclined baffle, it will overflow from the upper surface of the inclined baffle and flow from the two overflow ports into the water tank formed between the water storage tank and the silt storage tank, and finally be discharged into the water channel through the water outlet pipe. Since a valve is installed on the silt outlet pipe, the silt in the silt storage tank can be discharged regularly through the silt outlet pipe through the control of the valve;

[0034] S3. When the sludge passes through the crushing box, the motor is started, and the output end of the motor drives the roller and the small I-shaped wheel to rotate. The roller can drive multiple crushing blades to rotate, so that the sludge can be crushed. At the same time, through the transmission of the belt, the small I-shaped wheel rotates and drives the large I-shaped wheel to rotate, so that the breaking blades can drive the stirring shaft to rotate, and the sludge in the silt storage tank can be slowly stirred to prevent the sludge from agglomerating.

[0035] According to S3, vibration and sound sensors are installed in the crushing box 19 to monitor the working status of the crushing blade 24 and the roller 23. An intelligent algorithm is added to perform intelligent stirring control, analyze the operating data of the stirring shaft 16 and the breaking blade 17, and automatically adjust the stirring speed and strength to ensure that the sludge is uniform and does not clump.

[0036] Compared with the prior art, the present invention has the following beneficial effects:

[0037] 1. The present invention uses the design of the water storage tank and the silt storage tank so that after the silt and water enter the silt storage tank, the silt is deposited at the bottom and the water accumulates on the top. When the water accumulates to submerge the uppermost edge of the inclined baffle, it overflows from the upper surface of the inclined baffle into the water storage tank between the water storage tank and the silt storage tank, and is finally discharged into the water channel through the outlet pipe, thus realizing the circulation of water, avoiding the waste of water resources, and helping to maintain the water level in the water channel stable.

[0038] 2. The present invention provides a crushing mechanism in the crushing box. When the sludge passes through the crushing box, the motor drives the roller and the crushing blade to rotate to crush the sludge, preventing the sludge from clumping too large, making it easier for the sludge to settle and store in the silt storage tank, reducing the possibility of sludge agglomeration blocking the pipeline and ensuring the smooth progress of subsequent sludge treatment.

[0039] 3. The present invention provides a stirring mechanism in the silt storage tank. Driven by a motor, the breaking blades drive the stirring shaft to slowly stir the silt in the silt storage tank, so that the silt does not agglomerate, thereby ensuring the smooth discharge of the silt, effectively avoiding the problem of pipeline blockage, and improving the operating stability and reliability of the equipment.

[0040] 4. The present invention installs a valve on the silt discharge pipe at the bottom of the silt storage tank. Through the control of the valve, the silt in the silt storage tank can be discharged regularly according to actual needs, ensuring the timely cleaning of the silt, ensuring the normal operation of the equipment, and improving the working efficiency and service life of the equipment.

[0041] 5. The present invention can filter weeds and other debris in the water through the filter plate arranged in the water storage tank. When water overflows from the silt storage tank into the water storage tank, the weeds and other debris in the water are intercepted by the filtration of the filter plate, thereby ensuring the cleanliness of the water discharged into the canal, reducing pollution to the canal, and being beneficial to maintaining the ecological environment of the canal and the irrigation water quality.

[0042] 6. The intelligent algorithm of this patent significantly improves the efficiency and reliability of the sludge treatment system by integrating advanced deep learning technology with real-time data monitoring. First, the system deploys a variety of sensors in key components such as the crushing box and the stirring shaft to collect multi-dimensional data such as vibration, sound, torque and speed in real time, and ensures the high quality and consistency of the input data through efficient data preprocessing and feature extraction. The Transformer-based deep learning model is adopted to make full use of its advantages in processing long sequence data and capturing complex dependencies to achieve accurate prediction and intelligent control of the equipment operation status. Through transfer learning and automated hyperparameter optimization, the model training process is more efficient, and the prediction accuracy is significantly improved, ensuring the system's adaptability under different working conditions and sludge characteristics. In addition, the real-time deployment and reasoning of the model enable the system to generate control instructions instantly, automatically adjust the speed and strength of the crushing blade and the stirring shaft, optimize the equipment operation parameters, and prevent sludge agglomeration and equipment overload. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] The accompanying drawings are used to provide a further understanding of the present invention and constitute a part of the specification. Together with the embodiments of the present invention, they are used to explain the present invention and do not constitute a limitation of the present invention. In the accompanying drawings:

[0044] Figure 1 It is a first-view stereogram in the present invention;

[0045] Figure 2 It is a main cross-sectional stereoscopic view of the present invention;

[0046] Figure 3 It is a side cross-sectional perspective view of the present invention;

[0047] Figure 4 It is a partial stereoscopic view of the water storage tank in the present invention;

[0048] Figure 5 It is a partial stereoscopic view of the filter plate in the present invention;

[0049] Figure 6 It is a partial stereoscopic view of the silt storage tank in the present invention;

[0050] Figure 7 It is a second viewing angle stereogram in the present invention;

[0051] Figure 8 This is a flow chart of the intelligent control algorithm in the present invention.

[0052] In the figure: 1. water storage tank; 2. silt storage tank; 3. silt inlet pipe; 4. transmission pipe; 5. silt suction pipe; 6. sludge pump; 7. inclined baffle; 8. filter plate; 9. filter hole; 10. handle; 11. circular support frame; 12. overflow port; 13. water outlet pipe; 14. silt outlet pipe; 15. valve; 16. stirring shaft; 17. scattering blade; 18. fixing plate; 19. crushing box; 20. box cover; 21. fixing block; 22. bolt; 23. roller; 24. crushing blade; 25. small I-shaped pulley; 26. belt; 27. large I-shaped pulley; 28. motor. DETAILED DESCRIPTION

[0053] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0054] Example 1

[0055] See also Figure 1-Figure 7 , the present invention provides the following technical solutions:

[0056] An irrigation canal sludge circulation cleaning device, characterized in that it comprises:

[0057] Water storage tank 1;

[0058] The silt storage tank 2 has an opening at the top, and the silt storage tank 2 is fixedly installed in the water storage tank 1 and a water storage tank is left between the silt storage tank 2 and the inner wall of the water storage tank 1;

[0059] There are two overflow ports 12, which are respectively opened at two side ends of the silt storage tank 2;

[0060] The silt inlet pipeline 3 is arranged on the upper side of the silt storage tank 2;

[0061] A transmission pipeline 4, one end of which is fixed to the top of the silt inlet pipeline 3;

[0062] A crushing box 19 is fixed to the other end of the transmission pipeline 4 and communicated with the transmission pipeline 4;

[0063] A sludge suction pipe 5, one end of which is fixed to and connected with the crushing box 19, and the other end of which is located in the water channel;

[0064] A sludge pump 6 is installed at the end of a sludge suction pipe 5 in the canal;

[0065] A crushing mechanism is arranged in the crushing box 19 and is used to crush the sludge entering the silt storage tank 2;

[0066] A stirring mechanism is provided in the silt storage tank 2 to prevent sludge from agglomerating and blocking the pipeline; and

[0067] The driving mechanism is connected with both the crushing mechanism and the stirring mechanism to realize the operation of both.

[0068] In a specific embodiment of the present invention, the water tank 1 is used to store water overflowing from the silt tank 2, and discharge the water into the canal through the outlet pipe 13 to achieve water circulation. The water tank 1 is mainly used to store water overflowing from the silt tank 2. During the entire cleaning process, water and silt will first enter the silt tank 2. Since the silt will be deposited at the bottom of the tank, and the water will accumulate on the top, when the water accumulates to a certain extent, it will flow into the water tank between the water tank 1 and the silt tank 2 through the overflow port 12. The outlet pipe 13 set at one side of the water tank 1 will discharge the water back to the canal, thereby achieving water circulation and avoiding the waste of water resources. At the same time, it also helps to maintain the water level in the canal stable. The silt tank 2 stores the silt sucked from the canal, and the top opening is convenient for the silt to enter. After the pump 6 sucks the silt and water in the canal, the silt will be deposited at the bottom of the silt storage tank 2. As the silt accumulates, water will accumulate in the silt storage tank 2. When the water accumulates to submerge the uppermost edge of the inclined baffle 7, it overflows from the upper surface of the inclined baffle 7 to the water storage tank, thereby achieving a preliminary separation of the silt and water. The overflow port 12 allows the water accumulated in the silt storage tank 2 to flow into the water tank formed between the water storage tank 1 and the silt storage tank 2. When the water in the silt storage tank 2 accumulates to a certain extent, the overflow port 12 is set so that the water can smoothly flow out of the silt storage tank 2 and enter the water storage tank, thereby ensuring the orderly flow and recycling of water. The silt inlet pipe 3 guides the silt transported by the transmission pipe 4 into the silt storage tank 2. The silt inlet pipe 3 is located on the upper side of the silt storage tank 2 and between the two inclined baffles 7. , ensuring that the silt can accurately enter the silt storage tank 2 for sedimentation and storage. The transmission pipeline 4 connects the silt inlet pipeline 3 and the crushing box 19 to transport the treated silt in the crushing box 19 to the silt inlet pipeline 3. The transmission pipeline 4 plays the role of a bridge, so that the silt can be orderly transmitted inside the equipment, from the crushing box 19 through the transmission pipeline 4 into the silt inlet pipeline 3, and finally into the silt storage tank 2. The crushing box 19 crushes the silt entering the silt storage tank 2. When the silt passes through the crushing box 19, the roller 23 in the crushing mechanism drives the crushing blade 24 to rotate to crush the silt to prevent the silt from agglomerating too large and affecting subsequent processing and discharge. One end of the silt suction pipeline 5 is fixed and connected to the crushing box 19, and the other end is located in the canal, which is used to transfer water The silt and water in the canal are sucked into the equipment. The silt pump 6 is installed at the end of the silt suction pipe 5 located in the canal. Through the action of the silt pump 6, the silt suction pipe 5 can suck the silt and water in the canal into the crushing box 19 for treatment. The silt pump 6 provides suction to suck the silt and water in the canal into the crushing box 19 through the silt suction pipe 5. The silt pump 6 is the power source for the whole equipment to absorb silt. It can effectively suck the silt and water at the bottom of the canal into the equipment. After starting the silt pump 6, the silt and water pass through the silt suction pipe 5, the crushing box 19, the transmission pipe 4, and the silt inlet pipe 3 in turn, and finally enter the silt storage tank 2. The crushing mechanism crushes the silt entering the silt storage tank 2 to prevent the silt from agglomerating. The stirring mechanism is arranged in the silt storage tank 2 to prevent the silt from agglomerating and blocking the pipe.The driving mechanism is connected with both the crushing mechanism and the stirring mechanism to realize the operation of both.

[0069] Specifically, a circular support frame 11 is fixed on the inner wall of the water storage tank 1 , a filter plate 8 is arranged on the circular support frame 11 , and a plurality of evenly distributed filter holes 9 are opened on the top of the filter plate 8 .

[0070] In this embodiment: the filter plate 8 is arranged on the inner wall of the water storage tank 1 through the circular support frame 11, and is used to filter weeds and other debris in the water. A plurality of evenly distributed filter holes 9 are provided on the top of the filter plate 8. When water overflows from the silt storage tank 2 into the water storage tank 1, weeds and other debris in the water are intercepted by the filtration of the filter plate 8, thereby ensuring the cleanliness of the water discharged into the canal and reducing the pollution to the canal. Two handles 10 are fixed on the top of the filter plate 8 for easy removal of the green weed plate 8.

[0071] Specifically, a water outlet pipe 13 is provided at one side end of the water storage tank 1 , a silt outlet pipe 14 is fixed at the bottom of the silt storage tank 2 , and a valve 15 is fixedly installed on the silt outlet pipe 14 .

[0072] In this embodiment: the water outlet pipe 13 discharges the water in the water storage tank 1 into the canal to realize water circulation. The water in the water storage tank 1 returns to the canal through the water outlet pipe 13, completing the water recycling and avoiding the waste of water resources. It also helps to maintain the water level stability in the canal; the silt discharge pipe 14 is used to discharge the silt in the silt storage tank 2, and the valve 15 controls the silt discharge time. Through the control of the valve 15, the silt in the silt storage tank 2 can be discharged regularly according to actual needs, ensuring the timely cleaning of the silt and ensuring the normal operation of the equipment.

[0073] Specifically, inclined baffles 7 are obliquely fixed on the inner walls of both sides of the silt storage tank 2 , and the silt inlet pipeline 3 is located between the two inclined baffles 7 .

[0074] In this embodiment: the inclined baffle 7 is used to guide the water accumulated in the silt storage tank 2 to overflow from the overflow port 12, and at the same time prevent the silt from directly entering the water storage tank through the overflow port 12. The inclined baffles 7 are fixed on the inner walls of both sides of the silt storage tank 2, so that when the water accumulates to a certain extent, it can smoothly overflow to the overflow port 12 along the upper surface of the inclined baffle 7, while the silt is deposited at the bottom of the silt storage tank 2 and will not overflow with the water, thereby achieving further separation of water and silt.

[0075] Specifically, a fixing plate 18 is fixed to the top of the silt storage tank 2, and the bottom of the silt inlet pipe 3 passes through and extends to the bottom of the fixing plate 18.

[0076] In this embodiment: the fixed plate 18 is used to fix the silt inlet pipe 3 to ensure the stability of the silt inlet pipe 3. The fixed plate 18 fixed on the top of the silt storage tank 2 allows the bottom of the silt inlet pipe 3 to pass through and extend to the bottom of the fixed plate 18, ensuring that the silt inlet pipe 3 will not shake or shift during operation, thereby improving the stability of the equipment.

[0077] Specifically, the crushing mechanism includes a roller 23 and a crushing blade 24 . The roller 23 is rotatably connected between the inner walls of both sides of the crushing box 19 . A plurality of crushing blades 24 are provided, and the plurality of crushing blades 24 are evenly fixed on the circumferential surface of the roller 23 .

[0078] In this embodiment: the crushing mechanism includes a roller 23 and a crushing blade 24, which crushes the sludge entering the silt storage tank 2 to prevent the sludge from clumping. When the sludge passes through the crushing box 19, the output end of the motor 28 drives the roller 23 and the crushing blade 24 to rotate to crush the sludge, making it easier for the sludge to settle and store in the silt storage tank 2, while also reducing the possibility of sludge clumping and clogging the pipeline.

[0079] Specifically, the stirring mechanism includes a stirring shaft 16 and a scattering blade 17. One end of the stirring shaft 16 is rotatably connected to an inner wall of one side of the silt storage tank 2, and the other end of the stirring shaft 16 is movable through and extends outside the water storage tank 1. A plurality of scattering blades 17 are provided, and the plurality of scattering blades 17 are evenly fixed on the circumferential surface of the stirring shaft 16.

[0080] In this embodiment: the stirring mechanism is arranged in the silt storage tank 2 to prevent the sludge from agglomerating and blocking the pipeline. Driven by the motor 28, the breaking blades 17 drive the stirring shaft 16 to rotate, and slowly stir the sludge in the silt storage tank 2 to prevent the sludge from agglomerating, thereby ensuring the smooth discharge of the sludge and avoiding the problem of pipeline blockage.

[0081] Specifically, the driving mechanism includes:

[0082] The motor 28 is fixed to one side of the crushing box 19, and the output shaft of the motor 28 movably passes through the other side of the crushing box 19 and extends outward;

[0083] The roller 23 is fixed on the output shaft of the motor 28;

[0084] A small I-shaped wheel 25 is fixed on the circumferential surface of the output shaft of the motor 28;

[0085] A large I-shaped wheel 27 is fixed to the circumferential surface of the extended end of the stirring shaft 16; and

[0086] The belt 26 is drivingly connected between the circumferential surfaces of the small I-shaped wheel 25 and the large I-shaped wheel 27 .

[0087] In this embodiment, the driving mechanism is connected to both the crushing mechanism and the stirring mechanism to realize the operation of both. The motor 28 is fixed to one side of the crushing box 19, and its output shaft movably penetrates the other side of the crushing box 19 and extends outward. The roller 23 is fixed to the output shaft of the motor 28, and the small I-shaped wheel 25 is also fixed to the circumferential surface of the output shaft of the motor 28. The large I-shaped wheel 27 is fixed to the circumferential surface of the extended end of the stirring shaft 16. The belt 26 is connected between the circumferential surfaces of the small I-shaped wheel 25 and the large I-shaped wheel 27. After starting the motor 28, the output end of the motor 28 drives the roller 23 and the small I-shaped wheel 25 to rotate. Thereby, the crushing blade 24 crushes the sludge; at the same time, through the transmission of the belt 26, the small I-shaped wheel 25 drives the large I-shaped wheel 27 to rotate when it rotates. The I-shaped wheels of different diameters have different transmission ratios during transmission, so that when the small I-shaped wheel 25 rotates, the large I-shaped wheel 27 rotates slowly, so that the sludge in the silt storage tank 2 will not splash everywhere, but only prevent the sludge from condensing into clumps, so that the breaking blades 17 can drive the stirring shaft 16 to rotate, so as to stir the sludge in the silt storage tank 2. The entire driving mechanism ensures the coordinated work of the crushing mechanism and the stirring mechanism, thereby improving the operating efficiency of the equipment.

[0088] Specifically, a box cover 20 is movably provided on the top of the crushing box 19 , and fixing blocks 21 are fixed to both side ends of the crushing box 19 and the box cover 20 , and two adjacent fixing blocks 21 are fixed by bolts 22 .

[0089] In this embodiment, the box cover 20 is used to protect the components inside the crushing box 19, and the fixing block 21 and the bolt 22 are used to fix the box cover 20. The box cover 20 movably arranged on the top of the crushing box 19 can be closed when the equipment is running to prevent foreign debris from entering the crushing box 19. The fixing blocks 21 fixed on both sides of the crushing box 19 and the box cover 20 are fixed by bolts 22, which ensures the firm installation of the box cover 20, protects the components inside the equipment, and also facilitates the maintenance and repair of the inside of the crushing box 19.

[0090] The working principle and use process of the present invention:

[0091] A method for cleaning irrigation canal sludge circulation, comprising the following steps:

[0092] S1. First, place the sludge pump 6 in the canal, and then start the sludge pump 6. The sludge pump 6 can suck the sludge and water in the canal into the crushing box 19 through the sludge suction pipe 5, and then discharge it into the sludge inlet pipe 3 through the transmission pipe 4, and finally enter the bottom of the sludge storage tank 2. Since the sludge will be deposited at the bottom of the sludge storage tank 2, the water entering the sludge storage tank 2 will also accumulate in the sludge storage tank 2;

[0093] S2. When the water in the silt storage tank 2 accumulates and submerges the uppermost edge of the inclined baffle 7, it will overflow from the upper surface of the inclined baffle 7 and flow from the two overflow ports 12 to the water tank formed between the water storage tank 1 and the silt storage tank 2, and finally be discharged into the water channel through the outlet pipe 13. Since the silt discharge pipe 14 is equipped with a valve 15, the silt in the silt storage tank 2 can be discharged regularly through the silt discharge pipe 14 through the control of the valve 15;

[0094] S3. When the sludge passes through the crushing box 19, the motor 28 is started, and the output end of the motor 28 drives the roller 23 and the small I-shaped wheel 25 to rotate. The roller 23 can drive the multiple crushing blades 24 to rotate, so as to crush the sludge. At the same time, through the transmission of the belt 26, the small I-shaped wheel 25 drives the large I-shaped wheel 27 to rotate when it rotates, so that the breaking blades 17 can drive the stirring shaft 16 to rotate, and the sludge in the silt storage tank 2 can be slowly stirred, so that the sludge does not clump and is not easy to cause pipeline blockage.

[0095] According to S3, vibration and sound sensors are installed in the crushing box 19 to monitor the working status of the crushing blade 24 and the roller 23. Intelligent algorithms are added to perform intelligent stirring control, analyze the operating data of the stirring shaft 16 and the breaking blade 17, and automatically adjust the stirring speed and strength to ensure that the sludge is uniform and does not clump. The specific steps are as follows:

[0096] Step A: Sensor Installation and Selection

[0097] Vibration sensor: installed on the crushing blade 24 and the roller 23 inside the crushing box 19. A three-axis piezoelectric accelerometer is selected to measure the vibration acceleration in the three directions of X, Y, and Z.

[0098] Sound sensor: Installed inside the crushing box 19, close to the crushing blade 24 and the roller 23. Choose a high-sensitivity condenser microphone with a wide frequency response range, suitable for industrial environments.

[0099] Torque sensor: installed on the stirring shaft 16, close to the output end of the motor 28; a non-contact torque sensor is selected to measure the torque of the stirring shaft in real time;

[0100] Speed ​​sensor: Installed on the stirring shaft 16; select the photoelectric encoder to provide high-resolution speed measurement;

[0101] Among them, vibration sensors and sound sensors are used to monitor the crushing process, and torque sensors and speed sensors are used to monitor the mixing process;

[0102] Step B: Data collection and preprocessing

[0103] Sensor data collection: vibration acceleration a(t) measured by the vibration sensor in the crushing box, sound pressure level L measured by the sound sensorp (t), the torque τ(t) measured by the torque sensor on the stirring shaft, and the speed ω(t) measured by the speed sensor of the stirring shaft.

[0104] Sound pressure level calculation:

[0105]

[0106] Among them, L p (t) is the sound pressure level in decibels (dB), p(t) is the sound pressure measured at time t in Pascals (Pa), p0 = 20 × 10 -6Pa is the reference sound pressure.

[0107] Torque change rate: It is used to judge the resistance change during the stirring process and detect the viscosity change and agglomeration of the sludge. The torque τ(t) is the torque value measured at time t, in Newton meter (Nm).

[0108] Data synchronization: Ensure that the data collected from vibration, sound, torque, and speed sensors are aligned in the time dimension. i , data in (t ij ,d ij ) in the form of ij is the timestamp, d ij is at time t ij The data value of, where i is the i-th sensor, j represents the sensor S i The jth data point index of . Different sensors have different sampling rates, so the data needs to be resampled to a common time base. Then use the linear interpolation method to increase the low sampling rate data to a high sampling rate.

[0109] Linear interpolation formula:

[0110] For the sensor data point (t i ,d i ) and (t i+1 ,d i+1 ), the interpolation value d(t) at time t is:

[0111]

[0112] Where d(t) is the interpolated data value at time t, t is the time point that needs to be interpolated, and t i+1 and t i are the timestamp of the next known data point and the timestamp of the previous known data point, respectively, i+1 With d i Represent the sensor data values ​​at the next moment and the previous moment respectively.

[0113] Data cleaning: Use a low-pass filter to remove high-frequency noise, and the filtered signal is recorded as

[0114]

[0115] Among them, d(t) is the original signal, is the signal after low-pass filtering, and are Fourier transform and inverse Fourier transform, and H(f) is the frequency response function of the filter.

[0116] Feature extraction: Perform frequency domain analysis on vibration and sound signals, convert time domain signals into frequency domain, and obtain the spectrum information of the signal; then calculate the power distribution of the signal at different frequencies and extract the main frequency f max And power spectral density P(f).

[0117]

[0118] P(f)=|D(f)| 2

[0119]

[0120] Where D(f) is the frequency domain signal spectrum, h(t) is the time domain signal, and f max is the main frequency and P(f) is the power spectral density.

[0121] The statistical characteristics of torque and speed include mean, variance, standard deviation, peak value and kurtosis.

[0122] Step C: Model selection and establishment

[0123] The Transformer-based deep learning model requires real-time monitoring and analysis of the operating data of the crushing box (19) and the agitator shaft (16) and other equipment; these data include vibration signals, sound signals, torque and rotation speed, which are all time series data; the Transformer-based model can make full use of these multi-dimensional time series data to achieve accurate prediction and control of the equipment operation status.

[0124] 1. Model input: sensor data collection preprocessing and feature extraction data sequence

[0125] Vibration signal characteristics: acceleration a(t), spectrum amplitude A(f)

[0126] Sound signal characteristics: sound pressure level L p (t), spectrum amplitude S(f)

[0127] Torque signal characteristics: Torque mean μ τ, standard deviation σ τ

[0128] Speed ​​signal characteristics: speed mean μ ω , standard deviation σ ω

[0129] Organize all features in time series to form the input feature matrix X:

[0130]

[0131] Where T is the time step, the total data sequence length, d is the feature dimension, which is equal to the number of all sensor features, and X is the input feature matrix; x t,i is the value of the i-th feature at time t.

[0132] 2. Model construction: The structure of the Transformer model mainly includes position encoding, encoder layer and decoder layer.

[0133] Positional encoding

[0134] For even dimension 2i:

[0135]

[0136] For odd dimension 2i+1:

[0137]

[0138] Add positional encoding to the input X′:

[0139] X′=X+PE

[0140] Where t is the time step index, t = 1, 2, ... T; i is the feature dimension index, i = 1, 2, ..., d model ;d model is the hidden layer dimension of the model, PE (t,i) is the position information of time step t in dimension i, X′ is the input after adding position encoding, and X is the input feature matrix.

[0141] Encoder layer

[0142] Linear mapping: The input X is transformed linearly to obtain the query, key, and value matrices.

[0143] Q=X′W Q ,K=X′W K ,V=X′W V

[0144] Among them, Q, K, V represent query, key and value matrices respectively, and W Q , WK , W V is the learnable parameter matrix corresponding to Q, K, and V, and X′ is the input matrix after adding position encoding;

[0145] Calculate the attention mechanism:

[0146]

[0147] Among them, Attention is the attention mechanism, d k is the dimension of the key and query vector; the Softmax function normalizes the attention weights to form a probability distribution;

[0148] Multi-head mechanism: Use h independent attention heads and connect the results:

[0149] MultiHead(Q,K,V)=Concat(head1,…,head h )W O

[0150] Among them, MultiHead represents the number of attention heads, W O is the output transformation matrix, Concat means concatenating the outputs of h attention heads according to the dimension, head i Representing the head of the attention mechanism i =Attention(Q i ,K i ,V i )

[0151] Feedforward neural network: two linear transformation layers with an activation function in between;

[0152] FFN(x)=ReLU(xW1+b1)w2+b2

[0153] Among them, W1 and W2 are the weight matrices of the first and second layers respectively, b1 and b2 are the bias vectors of the first and second layers respectively, ReLU is the activation function, and FFN is the output of the feedforward neural network;

[0154] Residual connections and layer normalization:

[0155] For the multi-head self-attention sublayer:

[0156] x SA =LayerNorm(x+MultiHead(x))

[0157] For a feed-forward neural network sublayer:

[0158] x FFN =LayerNorm(x SA+FFN(x SA ))

[0159] Among them, x is the input of the encoder layer, x SA is the output after self-attention and residual connection, x FFN is the output after the feedforward network and residual connection, and LayerNorm represents residual connection and layer normalization.

[0160] Decoder layer: Based on the output of the encoder layer, the decoder layer generates the parameters of the stirring shaft speed and the crushing blade speed that need to be adjusted, that is, generates control instructions. This layer uses a masked multi-head self-attention mechanism to prevent the decoder from seeing future output information and ensure the causality of the sequence.

[0161] Linear Mapping:

[0162]

[0163] Calculating attention:

[0164] Attention(Q dec ,K enc ,V enc )

[0165] Among them, x dec represents the input of the decoder layer, x enc is the output of the encoder layer, Query the weight matrix for the decoder, is the encoder key-value weight matrix, is the encoder numerical weight matrix, Q dec ,K enc ,V enc They represent the encoder query vector, encoder key vector, and encoder value vector respectively.

[0166] The decoder layer feedforward neural network has the same structure as the encoder feedforward network, and the residual connections and layer normalization are also handled in the same way as in the encoder.

[0167] 3. Model output: Based on the output of the decoder, generate the equipment parameters that need to be adjusted, the new speed r of the stirring shaft new , Crushing blade speed ω new .

[0168] Linear transformation:

[0169] y=x dec W out +b out

[0170] Among them, y is the adjustment parameter vector output by the model, x dec represents the input of the decoder layer, Wout and b out are the weights and biases of the output layer.

[0171] New stirring shaft speed:

[0172] r new =r current +Δr

[0173] Among them, r new is the new speed of the stirring shaft, Δr is the speed increment of the stirring shaft predicted by the model, and r current is the current stirring shaft speed.

[0174] New speed of crushing blade:

[0175] ω new =ω current +Δω

[0176] Among them, ω new is the new speed of the crushing blade, ω current is the current crushing blade speed, and Δω is the crushing blade speed increment predicted by the model.

[0177] Step D: Model training and optimization

[0178] The loss function of the training model is L pre , the loss function after fine-tuning is L fine ,but:

[0179]

[0180] Where N is the number of samples, y i is the actual control parameter, are the control parameters predicted by the model.

[0181] Automated Hyperparameter Optimization Methods

[0182] In the present invention, Bayesian optimization and AutoML automatic machine learning methods are used to perform hyperparameter tuning, aiming to find the optimal model configuration and improve model performance.

[0183] Bayesian Optimization

[0184] Objective function: defined as the model performance indicator on the validation set, mean square error MSE;

[0185] f(θ)=MSE validation (θ)

[0186] Among them, f(θ) is the objective function of Bayesian optimization; θ is the set of hyperparameters, MSE validation is the mean square error on the validation set.

[0187] Gaussian process regression: used to model the target function

[0188]

[0189] Among them, f(θ) is the objective function of Bayesian optimization; represents a Gaussian process; m(θ) is the mean function, and k(θ,θ′) is the kernel function, which measures the similarity between the hyperparameter combinations θ and θ′.

[0190] Acquisition function: used to balance exploration and utilization and select the next hyperparameter combination for evaluation.

[0191] EI(θ)=max(f min -f(θ),0)

[0192] Among them, f min is the currently known minimum objective function value, and EI(θ) is the expected improvement.

[0193] Automatic Machine Learning AutoML

[0194] Define the search space S:

[0195] S={θ1,θ2,…,θ n}

[0196] Among them, S is the defined search space, θ i is the i-th hyperparameter combination.

[0197] Optimization goal:

[0198]

[0199] in, It means that the goal is to minimize the MSE on the validation set, θ is the hyperparameter set, and MSE validation is the mean square error on the validation set.

[0200] Evaluation and selection: Select the model configuration based on the performance metric on the validation set.

[0201] Step E: Model deployment and application

[0202] After completing model training and optimization, the trained Transformer-based deep learning model is deployed to the actual system to achieve real-time intelligent control.

[0203] Real-time inference: Deploy the trained deep learning model on the central processing unit (CPU) to quickly process real-time sensor data and generate control instructions.

[0204] Automatic control: According to the control instructions output by the model, the rotation speed and strength of the crushing blade 24 and the stirring shaft 16 are automatically adjusted to ensure the high efficiency and stability of the sludge treatment process.

[0205] Control strategy: PID control further refines the speed adjustment (PID adjustment is a conventional adjustment, and the parameters can be adjusted manually) to ensure smooth system response and reduce overshoot and oscillation.

[0206]

[0207] Among them, u(t) is the control signal, e(t) is the error, that is, the difference between the target speed and the current speed, K p , K i , K d are the proportional, integral and differential coefficients respectively.

[0208] Human-computer interaction: Provides operators with a friendly visual interface that displays sensor data, model output, and equipment status in real time to facilitate monitoring and decision-making.

[0209] Real-time data monitoring module: displays the real-time data of each sensor, vibration acceleration a(t), sound pressure level L p (t), torque τ(t), speed ω(t).

[0210] Model output display module: displays the predictive control instructions of the Transformer model.

[0211] Equipment status module: displays the current stirring shaft speed, crushing blade speed and adjusted speed.

[0212] Alarm and prompt module: When sensor data is abnormal or equipment status deviates from the set range, alarm information and maintenance suggestions are displayed.

[0213] Numerical example:

[0214] Data collection

[0215] Sensor data collection: vibration acceleration a(t) measured by the vibration sensor in the crushing box, sound pressure level L measured by the sound sensor p (t), the torque τ(t) measured by the torque sensor on the stirring shaft, and the speed ω(t) measured by the speed sensor of the stirring shaft are sampled once per second for 10 seconds.

[0216] Data synchronization: The data of all sensors have been time-stamped and aligned through the Network Time Protocol (NTP) to ensure the consistency of the data in the time dimension.

[0217] Data cleaning

[0218] Filtering: Use a low-pass filter to remove high-frequency noise and set the cutoff frequency f c =1kHz

[0219] Outlier elimination: Use the Z-score method to detect and eliminate outliers. Calculate the mean value of vibration acceleration μ a =14.0m / s 2 , standard deviation σ a =2.0m / s. Set the threshold value Z threshold =3,|Z i |>3 is considered an outlier.

[0220] Feature extraction

[0221] Frequency domain analysis of vibration and sound signals:

[0222] Perform fast Fourier transform FFT on the data per second to extract the main frequency f max And power spectrum density P(f). At the 10th second, the main frequency f of the vibration signal max =50Hz, power spectrum density P(50Hz)=200W / Hz

[0223] Statistical characteristics of torque and speed:

[0224] Calculate the mean, variance and peak value of torque and speed. Data characteristics at the 10th second:

[0225] Torque mean: μ τ =160Nm

[0226] Torque variance:

[0227] Average speed: μ ω =124RPM

[0228] Speed ​​variance:

[0229] Model building and training

[0230] Model selection

[0231] A Transformer-based deep learning model is adopted, and its self-attention mechanism is used to process multi-dimensional time series data to capture the complex patterns of equipment operation status.

[0232] Model Input

[0233] Organize the preprocessed and feature extracted data into the input matrix X:

[0234]

[0235] Model Training

[0236] Pre-trained model transfer learning: Load the pre-trained Transformer model, freeze the first 12 layers, and only fine-tune the last 4 layers and the newly added decoder layer.

[0237] Hyperparameter tuning: Find the best hyperparameter combination through Bayesian optimization, learning rate Batch size B = 64, number of attention heads h = 8, feedforward network dimension d ff =1024.

[0238] Model Output

[0239] After the model training is completed, it can predict the control instructions based on the input data

[0240] Δr: agitator shaft speed adjustment, in RPM.

[0241] Δω: Crushing blade speed adjustment, in RPM.

[0242] Model deployment and application

[0243] Real-time inference

[0244] Deploy the trained Transformer model on the central processing unit GPU and receive sensor data in real time real , perform forward propagation and generate control instructions

[0245] Real-time input data at the 10th second:

[0246] X real =[17.5m / s 2 ,96dB,185Nm,128RPM]

[0247] Model prediction output:

[0248]

[0249] Explanation of numerical meaning:

[0250] Δr: The model recommends increasing the agitator shaft speed by 5 RPM from the current 120 RPM.

[0251] Δω: The model recommends reducing the crushing blade speed by 3RPM from the current 200RPM.

[0252] Automatic Control

[0253] Automatically adjust device parameters based on model output:

[0254] Stirring shaft speed adjustment:

[0255] rnew =r current +Δr=120+5=125RPM

[0256] Crushing blade speed adjustment:

[0257] ω new =ω current +Δω=200-3=197RPM

[0258] Human-computer interaction

[0259] Through the visual interface, operators can monitor the following in real time:

[0260] Real-time data: Display current vibration acceleration a(t)=17.5m / s, sound pressure level L p (t) = 96dE, torque τ(t) = 185Nm, speed ω(t) = 128RPM

[0261] Model output: Display control instructions Δr = +5RPM, Δω = -3RPM

[0262] Equipment status: Displays the adjusted stirring shaft speed r new 125RPM and crushing blade speed ω new =197RPM.

[0263] Alarm and prompt: If the sensor data is abnormal, the system will automatically issue an alarm prompting "abnormal vibration, it is recommended to check the wear of the crushing blade."

[0264] Finally, it should be noted that the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art can still modify the technical solutions described in the aforementioned embodiments or replace some of the technical features therein by equivalents. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. An irrigation canal sludge circulation cleaning device, characterized in that: include: Water storage tank (1); A silt storage tank (2) having an opening at the top, the silt storage tank (2) being fixedly installed in the water storage tank (1) and having a water storage tank between the silt storage tank (2) and the inner wall of the water storage tank (1); Two overflow ports (12) are provided, and are respectively opened at two side ends of the silt storage tank (2); A silt inlet pipeline (3) is arranged on the upper side of the silt storage tank (2); A transmission pipeline (4), one end of which is fixed to the top of the silt inlet pipeline (3); A crushing box (19) is fixed to the other end of the transmission pipeline (4) and is in communication with the transmission pipeline (4); A silt suction pipe (5), one end of which is fixed to and connected to the crushing box (19), and the other end of which is located in the water channel; A sludge pump (6) is installed at the end of the sludge suction pipe (5) located in the canal; A crushing mechanism, arranged in the crushing box (19), for crushing the sludge entering the sludge storage tank (2); A stirring mechanism is arranged in the silt storage tank (2) to prevent sludge from agglomerating and blocking the pipeline; and The driving mechanism is connected with both the crushing mechanism and the stirring mechanism to realize the operation of both.

2. The irrigation canal sludge circulation cleaning equipment according to claim 1 is characterized in that: A circular support frame (11) is fixed on the inner wall of the water storage tank (1), a filter plate (8) is arranged on the circular support frame (11), and a plurality of evenly distributed filter holes (9) are provided on the top of the filter plate (8).

3. The irrigation canal sludge circulation cleaning equipment according to claim 2, characterized in that: A water outlet pipe (13) is provided at one side end of the water storage tank (1), a silt outlet pipe (14) is fixed at the bottom of the silt storage tank (2), and a valve (15) is fixedly installed on the silt outlet pipe (14); Inclined baffles (7) are fixedly disposed on both inner walls of the silt storage tank (2) at an angle, and the silt inlet pipeline (3) is located between the two inclined baffles (7).

4. The irrigation canal sludge circulation cleaning equipment according to claim 3 is characterized in that: A fixing plate (18) is fixed to the top of the silt storage tank (2), and the bottom of the silt inlet pipe (3) penetrates to the bottom of the fixing plate (18) and extends.

5. The irrigation canal sludge circulation cleaning equipment according to claim 4, characterized in that: The crushing mechanism comprises a rotating roller (23) and a crushing blade (24); the rotating roller (23) is rotatably connected between the inner walls of both sides of the crushing box (19); a plurality of crushing blades (24) are provided, and the plurality of crushing blades (24) are evenly fixed on the circumferential surface of the rotating roller (23).

6. The irrigation canal sludge circulation cleaning equipment according to claim 5, characterized in that: The stirring mechanism comprises a stirring shaft (16) and a scattering blade (17), one end of the stirring shaft (16) is rotatably connected to an inner wall of one side of the silt storage tank (2), and the other end of the stirring shaft (16) movably penetrates and extends outside the water storage tank (1), and a plurality of scattering blades (17) are provided, and the plurality of scattering blades (17) are evenly fixed on the circumferential surface of the stirring shaft (16).

7. The irrigation canal sludge circulation cleaning equipment according to claim 6, characterized in that: The driving mechanism comprises: A motor (28) is fixed to one side end of the crushing box (19), and an output shaft of the motor (28) movably passes through the other side end of the crushing box (19) and extends outward; The rotating roller (23) is fixed on the output shaft of the motor (28); A small I-shaped wheel (25) is fixed on the circumferential surface of the output shaft of the motor (28); A large I-shaped wheel (27) fixed to the circumferential surface of the extended end of the stirring shaft (16); and A belt (26) is drivingly connected between the circumferential surfaces of the small I-shaped wheel (25) and the large I-shaped wheel (27).

8. The irrigation canal sludge circulation cleaning equipment according to claim 7, characterized in that: A box cover (20) is movably provided on the top of the crushing box (19), and fixing blocks (21) are fixed to both side ends of the crushing box (19) and the box cover (20), and two adjacent fixing blocks (21) are fixed by bolts (22).

9. A method for recycling and cleaning sludge in an irrigation canal, applied to the device for recycling and cleaning sludge in an irrigation canal as claimed in claim 9, characterized in that: The steps include: S1. First, place the sludge pump (6) in the water channel, and then start the sludge pump (6). The sludge pump (6) can suck the sludge and water in the water channel into the crushing box (19) through the sludge suction pipe (5), and then discharge them into the sludge inlet pipe (3) through the transmission pipe (4), and finally enter the bottom of the sludge storage tank (2). Since the sludge will be deposited at the bottom of the sludge storage tank (2), the water entering the sludge storage tank (2) will also accumulate in the sludge storage tank (2); S2. When the water in the silt storage tank (2) accumulates and submerges the uppermost edge of the inclined baffle (7), it will overflow from the upper surface of the inclined baffle (7) and flow from the two overflow ports (12) to the water tank formed between the water storage tank (1) and the silt storage tank (2), and finally be discharged into the water channel through the water outlet pipe (13). Since a valve (15) is installed on the silt discharge pipe (14), the silt in the silt storage tank (2) can be discharged regularly through the silt discharge pipe (14) by controlling the valve (15); S3. When the sludge passes through the crushing box (19), the motor (28) is started, and the output end of the motor (28) drives the roller (23) and the small I-shaped wheel (25) to rotate. The roller (23) can drive the plurality of crushing blades (24) to rotate, so as to crush the sludge. At the same time, through the transmission of the belt (26), the small I-shaped wheel (25) drives the large I-shaped wheel (27) to rotate, so that the breaking blades (17) can drive the stirring shaft (16) to rotate, so as to slowly stir the sludge in the sludge storage tank (2) so that the sludge does not agglomerate.

10. The irrigation canal sludge circulation cleaning method according to claim 9, characterized in that: According to S3, a sensor is installed in the crushing box (19) to monitor the working status of the crushing blade (24) and the roller (23); an intelligent algorithm is added to perform intelligent stirring control, analyze the operating data of the stirring shaft (16) and the breaking blade (17), and automatically adjust the stirring speed and strength; the specific steps are as follows: Step A: Sensor Installation and Selection Vibration sensor: installed on the crushing blade (24) and the roller (23) inside the crushing box (19); a three-axis piezoelectric accelerometer is selected to measure the vibration acceleration in the three directions of X, Y and Z; Sound sensor: installed inside the crushing box (19), close to the crushing blade (24) and the roller (23); a high-sensitivity condenser microphone is selected, with a wide frequency response range, suitable for industrial environments; Torque sensor: installed on the stirring shaft (16), close to the output end of the motor (28); a non-contact torque sensor is selected to measure the torque of the stirring shaft in real time; Speed ​​sensor: installed on the stirring shaft (16); select a photoelectric encoder to provide high-resolution speed measurement; Among them, vibration sensors and sound sensors are used to monitor the crushing process, and torque sensors and speed sensors are used to monitor the mixing process; Step B: Data collection and preprocessing Sensor data collection: vibration acceleration a(t) measured by the vibration sensor in the crushing box, sound pressure level L measured by the sound sensor p (t), the torque τ(t) measured by the torque sensor on the stirring shaft, and the speed ω(t) measured by the speed sensor of the stirring shaft; Sound pressure level calculation: Among them, L p (t) is the sound pressure level in decibels (dB), p(t) is the sound pressure measured at time t in Pascals (Pa), p0 = 20 × 10 -6Pa is the reference sound pressure; Torque change rate: It is used to judge the resistance change during the stirring process and detect the viscosity change and agglomeration of the sludge. The torque τ(t) is the torque value measured at time t, in Newton meter (Nm). Data synchronization: Ensure that the data collected from vibration, sound, torque, and speed sensors are aligned in the time dimension. i , data in (t ij ,d ij ) in the form of ij is the timestamp, d ij is at time t ij The data value of; where i is the i-th sensor, j represents the sensor S i The jth data point index of the sensor is different; different sensors have different sampling rates, so the data needs to be resampled to a common time base; then the linear interpolation method is used to increase the low sampling rate data to a high sampling rate; Linear interpolation formula: For the sensor data point (t i ,d i ) and (t i+1 ,d i+1 ), the interpolation value d(t) at time T is: Where d(t) is the interpolated data value at time t, t is the time point at which interpolation is required, and t i+1 and t i are the timestamp of the next known data point and the timestamp of the previous known data point, respectively, d i+1 With d i Respectively represent the sensor data values ​​at the next moment and the previous moment; Data cleaning: Use a low-pass filter to remove high-frequency noise, and the filtered signal is recorded as Among them, d(t) is the original signal, is the signal after low-pass filtering, and is the Fourier transform and inverse Fourier transform, H(f) is the frequency response function of the filter; Feature extraction: Perform frequency domain analysis on vibration and sound signals, convert time domain signals into frequency domain, and obtain the signal Spectrum information; then calculate the power distribution of the signal at different frequencies and extract the main frequency f max and power spectral density P(f); P(f)=|D(f)| 2 Where D(f) is the frequency domain signal spectrum, h(t) is the time domain signal, and f max is the main frequency, P(f) is the power spectrum density; The statistical characteristics of torque and speed include mean, variance, standard deviation, peak value, and kurtosis; Step C: Model selection and establishment By using a Transformer-based deep learning model, it is necessary to monitor and analyze the operating data of the crushing box (19) and the stirring shaft (16) in real time; these data include vibration signals, sound signals, torque and rotation speed, which are all time series data; by using a Transformer-based model, these multi-dimensional time series data can be fully utilized to achieve accurate prediction and control of the operating status of the equipment; 1. Model input: sensor data collection preprocessing and feature extraction data sequence Vibration signal characteristics: acceleration a(t), spectrum amplitude A(f); Sound signal characteristics: sound pressure level L p (t), spectrum amplitude S(f); Torque signal characteristics: Torque mean μ τ , standard deviation σ τ ; Speed ​​signal characteristics: speed mean μ ω , standard deviation σ ω ; Organize all features in time series to form the input feature matrix X: Where T is the time step, the total data sequence length, d is the feature dimension, which is equal to the number of all sensor features, and X is the input feature matrix; x t,i is the value of the i-th feature at time t; 2. Model construction: The structure of the Transformer model mainly includes position encoding, encoder layer and decoder layer; Positional encoding For even dimension 2i: For odd dimension 2i+1: Add positional encoding to the input X′: X′=X+PE Where t is the time step index, t = 1, 2, ... T; i is the feature dimension index, i = 1, 2, ..., d model ;d model is the hidden layer dimension of the model, PE (t,i) is the position information of time step t in dimension i, X′ is the input after adding position encoding, and X is the input feature matrix; Encoder layer Linear mapping: The input X is transformed linearly to obtain the query, key and value matrices; Q=X′W Q ,K=X′W K ,V=X′W V Among them, Q, K, V represent query, key and value matrices respectively, and W Q , W K , W V is the learnable parameter matrix corresponding to Q, K, and V, and X′ is the input matrix after adding position encoding; Calculate the attention mechanism: Among them, Attention is the attention mechanism, d k is the dimension of the key and query vector; the Softmax function normalizes the attention weights to form a probability distribution; Multi-head mechanism: Use h independent attention heads and connect the results: MultiHead(Q,K,V)=Concat(head1,…,head h )W O Among them, MultiHead represents the number of attention heads, W O is the output transformation matrix, Concat means concatenating the outputs of h attention heads according to the dimension, head i Representing the head of the attention mechanism i =Attention(Q i ,K i ,V i ); Feedforward neural network: two linear transformation layers with an activation function in between; FFN(x)=ReLU(xW1+b1)W2+b2 Among them, W1 and W2 are the weight matrices of the first and second layers respectively, b1 and b2 are the bias vectors of the first and second layers respectively, ReLU is the activation function, and FFN is the output of the feedforward neural network; Residual connections and layer normalization: For the multi-head self-attention sublayer: x SA =LayerNorm(x+MultiHead(x)) For a feed-forward neural network sublayer: x FFN =LayerNorm(x SA +FFN(x SA )) Among them, x is the input of the encoder layer, x SA is the output after self-attention and residual connection, x FFN is the output after the feedforward network and residual connection, LayerNorm represents residual connection and layer normalization; Decoder layer: Based on the output of the encoder layer, the decoder layer generates the parameters of the stirring shaft speed and the crushing blade speed that need to be adjusted, that is, generates control instructions; this layer uses the masked multi-head self-attention mechanism to prevent the decoder from seeing future output information and ensure the causality of the sequence; Encoder-Decoder Attention Mechanism: Linear Mapping: Calculating attention: Attention(Q dec ,K enc ,V enc ) Among them, x dec represents the input of the decoder layer, x enc is the output of the encoder layer, Query the weight matrix for the decoder, is the encoder key-value weight matrix, is the encoder numerical weight matrix, Q dec ,K enc ,V enc Represent the encoder query vector, encoder key vector, and encoder value vector respectively; The decoder layer feedforward neural network has the same structure as the encoder feedforward network, and the residual connection and layer normalization are also handled in the same way as the encoder; 3. Model output: Based on the output of the decoder, generate the equipment parameters that need to be adjusted, the new speed r of the stirring shaft new , Crushing blade speed ω new ; Linear transformation: y=x dec W out +b out Among them, y is the adjustment parameter vector output by the model, x dec represents the input of the decoder layer, W out and b out are the weights and biases of the output layer; New stirring shaft speed: r new =r current +Δr Among them, r new is the new speed of the stirring shaft, Δr is the speed increment of the stirring shaft predicted by the model, and r current is the current stirring shaft speed; New speed of crushing blade: oh new =ω current +See Among them, ω new is the new speed of the crushing blade, ω current is the current crushing blade speed, Δω is the crushing blade speed increment predicted by the model; Step D: Model training and optimization The loss function of the training model is L pre , the loss function after fine-tuning is L fine ,but: Where N is the number of samples, y i is the actual control parameter, Control parameters predicted by the model; Automated hyperparameter optimization methods In the present invention, Bayesian optimization and AutoML automatic machine learning methods are used to perform hyperparameter tuning, aiming to find the best model configuration and improve model performance; Bayesian Optimization Objective function: defined as the model performance indicator on the validation set, mean square error MSE; f(θ)=MSE validation (i) Among them, f(θ) is the objective function of Bayesian optimization, θ is the hyperparameter set, and MSE validation is the mean square error on the validation set; Gaussian process regression: used to model the target function; Among them, f(θ) is the objective function of Bayesian optimization; represents a Gaussian process; m(θ) is the mean function, and k(θ,θ′) is the kernel function, which measures the similarity between the hyperparameter combinations θ and θ′; Acquisition function: used to balance exploration and utilization and select the next hyperparameter combination for evaluation; EI(θ)=max(f min -f(θ),0) Among them, f min is the currently known minimum objective function value, and EI(θ) is the expected improvement; Automatic Machine Learning AutoML Define the search space S: S={θ1,θ2,…,θ n } Among them, S is the defined search space, θ i is the i-th hyperparameter combination; Optimization goal: in, It means that the goal is to minimize the MSE on the validation set, θ is the hyperparameter set, and MSE validation is the mean square error on the validation set; Evaluation and selection: Select the model configuration based on the performance indicators on the validation set; Step E: Model deployment and application After completing model training and optimization, the trained Transformer-based deep learning model is deployed to the actual system to achieve real-time intelligent control; Real-time reasoning: Deploy the trained deep learning model on the central processing unit (CPU) to quickly process real-time sensor data and generate control instructions; Automatic control: According to the control instructions output by the model, the speed and force of the crushing blade (24) and the stirring shaft (16) are automatically adjusted to ensure the efficiency and stability of the sludge treatment process; PID control further refines the speed adjustment to ensure smooth system response and reduce overshoot and oscillation: Among them, u(t) is the control signal, e(t) is the error, that is, the difference between the target speed and the current speed, K p , K i , K d are the proportional, integral and differential coefficients respectively; Human-computer interaction: Provides operators with a friendly visual interface that displays sensor data, model output, and equipment status in real time to facilitate monitoring and decision-making; Real-time data monitoring module: displays the real-time data of each sensor, vibration acceleration a(t), sound pressure level L p (t), torque τ(t), speed ω(t); Model output display module: displays the predictive control instructions of the Transformer model; Equipment status module: displays the current stirring shaft speed, crushing blade speed and adjusted speed; Alarm and prompt module: When sensor data is abnormal or equipment status deviates from the set range, alarm information and maintenance suggestions are displayed.

Images