A maintenance system for a tobacco drum moisture removal filter anti-sticking device

Abstract

The present invention provides a maintenance system for a tobacco drum dehumidification filter anti-sticking device, comprising a remote server that records the update and maintenance time points of the tobacco drum dehumidification filter anti-sticking device, and at least one testing device identical in structure to the tobacco drum dehumidification filter anti-sticking device. The tobacco drum dehumidification filter anti-sticking device and the testing device operate synchronously to record a graph of the overall performance degradation of at least one set of stainless steel circulating filters, a cleaning assembly, a drive assembly, and a collection assembly. The remote server uses the time points and a learning model constructed based on the time-dependent degradation function graph to predict the overall performance degradation of the current device in real time. After replacement of the system of the present invention, the dehumidification system exhausts smoothly, with no noticeable accumulation of tobacco shreds on or within the screen, and dehumidification is uniform and smooth. Furthermore, the testing device uses artificial intelligence learning to predict the contribution of performance degradation to each component, accurately determining the update and maintenance time points.

Description

Technical Field

[0001] The invention relates to the technical field of tobacco production equipment, in particular to a maintenance system for a tobacco drum moisture removal filter anti-sticking device. Background Art

[0002] The anti-sticking, anti-clogging, and moisture-removing device is a key feature of the tobacco yarn drum dehumidification process. Its purpose is to remove lint and hemp from the drum discharge chamber filter screen, preventing sticking and blocking. Tobacco processing typically requires moisture conditioning, flavoring, and adding flavoring to tobacco leaves under specific conditions. During the production process, tobacco foam and floating debris are attracted to the dehumidification filter screen. If not cleaned regularly, debris can easily clog the filter screen, paralyzing dehumidification and affecting tobacco material quality. Therefore, the design of an anti-sticking device is urgently needed. However, maintenance is a challenge. Without considering ease of assembly and disassembly, determining the coordination and updating of various components has become a major challenge for anti-sticking devices. Summary of the Invention

[0003] In order to solve the above problems, the purpose of the present invention is to provide a maintenance system for the tobacco drum moisture removal filter anti-sticking device.

[0004] The present invention provides a maintenance system for a tobacco roller dehumidification filter anti-sticking device, comprising a remote server for recording the update and maintenance time points (different from the time points described below) of the tobacco roller dehumidification filter anti-sticking device having a stainless steel circulation filter, a drive component, a cleaning component and a collection component, and at least one test device having the same structure as the tobacco roller dehumidification filter anti-sticking device. The tobacco roller dehumidification filter anti-sticking device operates synchronously with the test device, and records a time-dependent performance degradation function graph of at least one group of stainless steel circulation filters, cleaning components, drive components and collection components according to the degree of decline in the quality level of tobacco products produced by the at least one test device during operation. The remote server judges the overall performance degradation prediction results of at least one group of stainless steel circulation filters, cleaning components, drive components and collection components of the current device in real time according to the time-dependent time points and a learning model constructed according to the time-dependent degradation function graph. The remote server can display the prediction results in real time and view them by downloading them to the mobile smart device of the staff, wherein,

[0005] The stainless steel circulating filter screen includes an active roller, a driven roller and a mesh chain plate arranged inside the shell, and the mesh chain plate is arranged between the active roller and the driven roller, wherein a first high-definition industrial camera is provided on the shell of the at least one test device, which is sealed and fixed through a first perforation at a preset position on the shell, and is used to regularly take images of the mesh chain plate and the scraper assembly; the driving component includes a casing, a motor, a reducer, a driving shaft and a transmission shaft, the motor is connected to the driving shaft through the reducer, and the driving shaft is connected to the transmission shaft, wherein a second high-definition industrial camera is provided at a preset position on the shell of the at least one test device, which is sealed and fixed through a second perforation at a preset position on the shell, and is used to regularly take images of the cleaning brush assembly, and the cleaning assembly includes a cleaning brush assembly and a scraper assembly; the collection component includes a flap sealing component arranged at the bottom of the device, wherein the images of the mesh chain plate and the scraper assembly and the image of the cleaning brush assembly are uploaded to the remote server, and a comprehensive evaluation of the performance degradation of the current each net and component is given through the pre-trained artificial intelligence model and the degradation prediction result.

[0006] It is understandable that while the test device participates in normal production operations, degradation function data is collected in real time, so that the performance degradation prediction results of each network and component are obtained through the data training model. This not only expands production but also solves the problem of collecting and obtaining training data. There is no need to maintain the test device because it is used to collect data and needs to truly reflect the actual degradation laws of each network and component in a maintenance-free state.

[0007] Optionally, the learning model is constructed by:

[0008] S1 obtains multiple groups of overall performance degradation degrees corresponding to different time points on a time-dependent performance degradation function graph of at least one set of stainless steel circulating filter, cleaning component, driving component, and collection component, and divides them into training sets and validation sets;

[0009] S2 constructs the corresponding long-short time memory network for the stainless steel circulating filter, cleaning component, driving component, and collection component. It uses different time points as different node units and constructs input vectors based on the degree of performance degradation. It converts the training set and validation set into training vectors and validation vectors.

[0010] The S3 zero vector is input to the transmission end of the first node unit, and the training vectors are input to the corresponding different node units respectively. The result of the node unit output end is compared with the training vector of the input end of the next node unit, and the training vector of the final node unit input end is compared with the predicted result of its own output end to obtain the loss function value, thereby continuously optimizing the network parameters, and verifying the prediction accuracy of each node unit through the verification vector. When the prediction accuracy of all node units stabilizes, a learning model is obtained, wherein the method for obtaining multiple groups of performance degradation degrees corresponding to different time points is to randomly select one day every 1-15 days to obtain one group of overall performance degradation degrees every hour, which are regarded as multiple groups of data on the performance degradation degrees at one time point, wherein the performance degradation degree is evaluated by randomly sampling and measuring the moisture content of the tobacco material multiple times every hour on a randomly selected day.

[0011] Therefore, for the function point on the overall performance degradation function graph recorded at any time point, the degree of performance degradation at the next time point can be predicted, and the next time point can be continuously predicted in the time direction to obtain a series of prediction results.

[0012] Optionally, multiple random samplings are performed at an optional statistical decrease in moisture content of 0.5-2% as a grade decrease.

[0013] It's easy to understand that the smaller the number of days selected from 1-15, the more statistically accurate the learning model's predictions. Therefore, within a year, a single test device can acquire between 584 and 8,760 sets of data at a single point in time. Without considering prediction accuracy, a single test device can collect enough training data (over 10,000 sets) in a year. Therefore, taking the median of 8 days and randomly selecting a single day for acquisition, there are 1,095 sets of data. Therefore, building 10 test devices within a year will yield sufficient training data. Therefore, manufacturers can arbitrarily choose the number of test devices they need based on their capital investment and accuracy requirements.

[0014] The artificial intelligence model pre-training method is:

[0015] Q1: Within every 1-15 days, within the randomly selected day, acquire one image of the mesh chain plate and scraper assembly and one image of the cleaning brush assembly every hour, segment the images of the mesh chain plate and scraper assembly, and divide the two regions belonging to the mesh chain plate and scraper assembly into two regions, and establish corresponding training atlases and validation atlases together with the images of the cleaning brush assembly;

[0016] Q2 establishes a convolutional neural network (Res-CNN) with a residual mechanism, inputs the training atlas into Res-CNN, and classifies the degree of performance degradation at the output through the fully connected LC and softmax functions. A loss function is constructed to obtain the loss function value for each training. Backpropagation is used to optimize the Res-CNN parameters, and the accuracy is verified by the validation set. When the loss function value and accuracy stabilize, training is stopped to obtain a pre-trained artificial intelligence model.

[0017] Through targeted image recognition of mesh chain plates and cleaning components that are prone to performance degradation, the main part of the overall performance degradation is separated, so that the driving component can be calculated through weighted coefficients.

[0018] Collect components that contribute to this non-prone performance degradation.

[0019] Optionally, one end of the mesh chain plate passes around the active roller and the driven roller and is connected to the other end thereof to form a circulating mesh chain plate.

[0020] Optionally, the motor is connected to the first end of the driving shaft through a reducer, the second end of the driving shaft is transmission-connected to the first end of the transmission shaft and the first end of the active roller, and the second end of the transmission shaft is transmission-connected to the brush roller.

[0021] Optionally, the transmission connection may be a pulley transmission or a sprocket transmission.

[0022] Preferably, a first window embedded with a heating wire is provided on the first perforation, and the first high-definition industrial camera is viewed through the first window and a first lighting system for irradiating fill light into the first window; a second window embedded with a heating wire is provided on the second perforation, and the second high-definition industrial camera is viewed through the second window and a first lighting system for irradiating fill light into the second window, and the inner surfaces of the shells of the first window and the second window are cleaned regularly by opening the third inspection door and the first inspection door respectively.

[0023] Optionally, the second viewing window, the second lighting system, and the second high-definition industrial camera are two sets, which respectively regularly capture images of the cleaning components of the mesh chain plate and scraper assembly, and the cleaning brush assembly.

[0024] Optionally, the cleaning brush assembly includes a brush roller, which includes a rotating shaft and a cleaning rod. The cleaning rod is arranged at intervals on the side wall of the rotating shaft to form a mace-shaped brush roller.

[0025] Preferably, at least one flexible branch rod is provided on one of the cleaning rods to further increase the cleaning area.

[0026] Optionally, the rotating shaft is a square rotating shaft.

[0027] Optionally, the flap sealing assembly includes a rotating shaft that is laterally rotatable and arranged in the sewage pipe, a flap fixedly arranged on the rotating shaft, a silicone pad wrapped around the edges of the flap, and a control handle that is arranged outside the sewage pipe and connected to the rotating shaft.

[0028] After the tobacco drum dehumidification filter anti-sticking device of the present invention is replaced, the dehumidification system exhausts smoothly, there is no obvious accumulation of tobacco fragments on the surface and inside of the screen, and the dehumidification is uniform and smooth. Since the present invention only improves the dehumidification system at the outlet of the original drum and does not change the main structure of the original equipment, it will not have a negative impact on the discharge process of the main machine. Since the present invention adopts a guide rail drawer installation, it is convenient for maintenance. Since the screen cleaning system equipment operates stably, it is convenient to repair by setting up an inspection port, which can ensure that there is no residue on the main body of the screen after daily cleaning and there are no dead corners. The entire device uses stainless steel as the basic skeleton to prevent the skeleton from being deformed by heat; the mesh chain and connecting and transmission parts are all made of 304 stainless steel to avoid corrosion, deformation or odor.

[0029] The present invention employs at least one test device with the same structure as the tobacco drum dehumidification filter anti-sticking device, along with a remote server. By selecting time points, plotting degradation functions, and building a learning model based on these functions, the system can predict future degradation based on the current time point. Simultaneously, individual images of vulnerable parts are identified to determine the contribution of each network and component to performance degradation. The degradation prediction results are then displayed remotely on the remote server.

[0030] By arranging a part of the conveyor belt in the dehumidification duct and a part outside the dehumidification duct, a cleaning roller is arranged outside the duct to clean the conveyor belt. In the dehumidification duct, tobacco and floating debris move upward and adhere to the conveyor belt and are transported to the outside. At the same time, the debris falls into the hopper, and the clean conveyor belt returns to the dehumidification duct. By adopting the above method to transport the adhered debris outward, the impurities are not blown into the drum and mixed with good quality tobacco. The conveyor belt can be kept clean at any time under the cleaning of the cleaning roller.

[0031] In the present invention, a special solenoid valve cleaning structure is used to prevent the screen from sticking to debris without dead corners. As a cleaning component, a high-temperature steam pipeline, a heated water pipeline, and a Venturi water pipeline are used in conjunction with the cleaning pipeline. By setting up a steam pipeline, stubborn sticking debris can be removed forcefully. By setting up a heated water pipeline, an atomized environment can be created to further prevent the adhesion of fine tobacco particles to the filter. By setting up a Venturi water pipeline, mold and other fine bacteria can be further prevented from taking root on the filter for a long time and contaminating the filter. A cleaning solenoid valve is set between these pipelines to achieve pipeline control. When the pressure in the cleaning component chamber is too high, part of the air pressure needs to be released through the cleaning solenoid valve. This special design was first created by the applicant team. It does not change the main structure of the original equipment, so it will not have a negative impact on the discharge process of the main machine. On the contrary, it adopts a guide rail drawer installation, so it is convenient for maintenance. By setting up such a cleaning component, the filter can be further effectively prevented from sticking.

[0032] In the present invention, by designing the flap sealing assembly, sundries can be collected in a centralized manner, so as to separate the sundries from the finished product materials, thereby having the function of removing lint and avoiding affecting the product quality. BRIEF DESCRIPTION OF THE DRAWINGS

[0033] Figure 1 This is a schematic diagram of the configuration of a maintenance system for a tobacco drum moisture removal filter anti-sticking device according to the present invention;

[0034] Figure 2 This is a schematic diagram of the overall structure of the tobacco drum moisture removal filter anti-sticking device in the maintenance system of an embodiment of the present invention;

[0035] Figure 3a yes Figure 2 Schematic diagram of the structure of the stainless steel circulating filter connected to the drive assembly of the tobacco drum moisture removal filter anti-sticking device;

[0036] Figure 3b It is a schematic diagram of the structure of the stainless steel circulating filter connection drive assembly in each of the two test devices in the maintenance system;

[0037] Figure 4 2. It is a schematic structural diagram of a stainless steel circulating filter screen of a tobacco drum moisture removal filter screen anti-sticking device according to an embodiment of the present invention;

[0038] Figure 5 This is a schematic diagram of the structure of one side of the drive assembly of the tobacco drum moisture removal filter anti-sticking device according to an embodiment of the present invention;

[0039] Figure 6 This is a schematic structural diagram of the other side of the driving assembly of the tobacco drum moisture removal filter anti-sticking device according to an embodiment of the present invention;

[0040] Figure 7This is an enlarged structural diagram of the cleaning brush assembly of the tobacco drum moisture removal filter anti-sticking device according to an embodiment of the present invention;

[0041] Figure 8 This is an enlarged structural diagram of the scraper assembly of the tobacco drum moisture removal filter anti-sticking device according to an embodiment of the present invention;

[0042] Figure 9 2. It is a structural schematic diagram of the brush roller of the tobacco drum moisture removal filter anti-sticking device according to an embodiment of the present invention;

[0043] Figure 10 It is a structural schematic diagram of the flap sealing assembly of the tobacco drum moisture removal filter anti-sticking device according to an embodiment of the present invention.

[0044] Figure 11 This is a schematic diagram of the construction method of the learning model with a median of 8 days as a time point.

[0045] Figure 12 This is the overall time-dependent performance degradation function diagram with a median of 8 days as a time point.

[0046] Figure 13 It is a training flowchart of the artificial intelligence model pre-training method.

[0047] Figure markings: 1-driving body; 2-first inspection door; 3-mesh chain body; 4-stainless steel mesh chain; 5-driving motor (front view); 6-inspection door seal; 7-main drive bearing (cross-sectional view); 8-cleaning pipeline; 9-cleaning solenoid valve; 10-bearing seat; 12-first discharge chamber; 13-second discharge chamber (second inspection door); 14-negative pressure isolation chamber; 17-sewage pipe; 21-active roller (mesh chain driving mechanism); 22-driven roller (tensioning device); 31-driving motor (side view); 33-active shaft (lateral view of main drive bearing); 34-transmission shaft (brush bearing); 41-rotating shaft (brush shaft); 42-cleaning rod (cleaning brush); 51-rotating shaft (flip shaft); 52-sealing flap; 53-sealing silicone pad; 54-switch control handle. DETAILED DESCRIPTION

[0048] The embodiments of the present invention are described below with reference to the accompanying drawings. Those skilled in the art should understand that these embodiments are only used to illustrate the present invention and are not intended to limit the present invention.

[0049] Example 1

[0050] This embodiment provides an overall description of the overall structure of the maintenance system and the maintenance method.

[0051] Figure 1A remote server is provided for recording updates and maintenance times for a tobacco drum dehumidification filter anti-sticking device (i.e., the structure of the stainless steel circulating filter connected to the drive assembly in the device labeled "Device" in the diagram). Two test devices, A and B, identical in structure to the tobacco drum dehumidification filter anti-sticking device (the structure of the stainless steel circulating filter connected to the drive assembly in red in the diagram), communicate with the remote server to obtain these times. The tobacco drum dehumidification filter anti-sticking device in the diagram operates synchronously with the test device, so the time-containing times obtained for the test devices are equivalent to those for the tobacco drum dehumidification filter anti-sticking device.

[0052] According to the degree of decline in the quality level of the tobacco product produced by the at least one test device during operation, a time-dependent degradation function graph of the performance of at least one group of stainless steel circulation filters, cleaning components, drive components, and collection components as a whole (that is, it can be identified as the test device as a whole or the device as a whole, because the two are isomorphic) is recorded. The remote server judges the overall performance degradation prediction results of at least one group of stainless steel circulation filters, cleaning components, drive components, and collection components of the current device in real time based on the time point and the learning model constructed according to the time-dependent degradation function graph. The remote server can display the prediction results in real time and download them to the mobile smart device of the staff for viewing.

[0053] Figure 2 yes Figure 1 A schematic diagram of the overall structure of the tobacco drum dehumidification filter anti-sticking device in the maintenance system. The filter anti-sticking device is detachably connected to the tobacco drum's dehumidification duct. As shown in the figure, the tobacco drum dehumidification filter anti-sticking device generally consists of four parts: a conveying system including a stainless steel circulating filter, a drive system, a cleaning brush system, and a collection system. Figure 3a This is a structural diagram of the connection between the cleaning box and the stainless steel circulating filter of the tobacco drum moisture removal filter anti-sticking device according to an embodiment of the present invention. Figure 4 This is a schematic diagram of the structure of the stainless steel circulating filter screen of the tobacco drum moisture removal filter screen anti-sticking device according to an embodiment of the present invention. Figure 2-4 As shown, the stainless steel circulating filter system of the tobacco drum moisture removal filter anti-sticking device includes a shell, and an active roller 21, a driven roller 22 and a mesh chain plate arranged inside the shell. The active roller 21 is a mesh chain driving mechanism for driving the mesh chain body 3 and the stainless steel mesh chain 4 to move along the conveyor belt. The driven roller 22 is a tensioning device for keeping the conveyor belt in a tensioned state during movement. The mesh chain plate is arranged between the active roller 21 and the driven roller 22. The driving mechanism is connected to the conveyor belt of the circulating mesh chain plate and is connected to the brush roller. The driving component includes a motor ( Figure 2 5 in the figure, Figure 5Reference numeral 31 in the figure), reducer (located at Figure 5 The middle part), the driving shaft 33 and the transmission shaft 34, the motor 5 (ie 31) is connected to the driving shaft 33 through the reducer, and the driving shaft 33 is connected to the transmission shaft 34; the driving shaft 33 is the main driving bearing (ie the following Figure 6 The main drive bearing 7) in the transmission shaft 34 is a brush bearing, which is Figure 8 The rotating shaft 41 is also called the brush shaft 41. In a specific embodiment, the transmission shaft 34 can be directly used as the brush shaft 41. The cleaning assembly includes a cleaning brush assembly and a scraper assembly. The cleaning brush assembly mainly includes a brush shaft 41 (i.e., the rotating shaft 41) and a cleaning brush 42 (i.e., the cleaning rod 42). The scraper assembly mainly includes a scraper, a partition, a wind shear plate, etc., which are located below the cleaning box near the network chain body 3; the collection assembly includes a flap shaft, a sealing flap, a sealing silicone pad, and other flap sealing components arranged at the bottom of the device.

[0054] In addition, if Figure 2 and Figure 3a The figure also shows that the tobacco drum moisture removal filter anti-sticking device according to an embodiment of the present invention also includes a main discharge portion including a drive body 1, a first inspection door 2, a first discharge chamber 12, a second discharge chamber 13 (with a second inspection door), a negative pressure isolation chamber 14, and a collection portion including a sewage pipe 17. The second inspection door 2 allows for real-time viewing of the operating status of the circulating mesh chain conveyor mechanism and facilitates replacement of the brush roller.

[0055] The lower end of the sewage pipe 17 is connected to the debris collection area. A flap sealing assembly is provided in the sewage pipe 17 . The flap sealing assembly can adjust the communication cross-section size of the upper and lower parts of the sewage pipe 17 .

[0056] Figure 3b This is a structural schematic diagram of the stainless steel circulating filter connection drive assembly commonly illustrated for test device a and test device b. A first high-definition industrial camera is provided on the outer shell of at least one of the test devices and is sealed and fixed through a first perforation at a preset position on the outer shell as shown in the figure, for regularly capturing images of the mesh chain plate.

[0057] A second high-definition industrial camera is provided at a preset position on the casing of the at least one test device, and is sealed and fixed by second perforations provided on two symmetrical sides of the lower part of the casing as shown in the figure, for regularly taking images of the cleaning components.

[0058] In this embodiment, the first window containing the heating wire is arranged on the first through hole (by Figure 3bThe enlarged part in the middle circle is shown), and the second window containing the heating wire is arranged on the second perforation to achieve sealing and fixation (the same setting method as the first window, not enlarged in the figure). The first lighting system and the second lighting system are respectively arranged on the edge sides of the first window and the second window, and the lighting light is respectively incident on the outer shell and the casing to illuminate the mesh chain plate and the scraper assembly, as well as the shooting part of the cleaning assembly. A third inspection door is provided next to the first window. After opening it, the surface of the first window facing the inside of the casing can be cleaned; for cleaning the surface of the second window facing the inside of the casing, it is opened by Figure 3b The first access door 2 is shown.

[0059] The mesh chain plate image and the cleaning component image are uploaded to the remote server, and a comprehensive evaluation of the performance degradation of each network and component is given through the pre-trained artificial intelligence model and the degradation prediction result.

[0060] The comprehensive evaluation specifically includes: assuming that the number of overall performance degradation levels is deg1, the numbers of performance degradation levels obtained from the degradation prediction results of the mesh chain plate and the cleaning component are deg2 and deg3 respectively, then the contribution of the driving component and the collection component to the overall performance degradation level number is deg4, then deg1 = αgdeg2 + βgdeg3 + γgdeg4, α+β+γ = 1, and α>β>γ, the α, β, γ weighting coefficients can be statistically calculated based on the corresponding mesh and component life and time points in the specific test device a and the test device b, for example, the ratio of the number of days and life corresponding to the randomly selected time point within the median of 8 days can be normalized.

[0061] like Figure 4 As shown, the stainless steel circulating filter includes a housing (see Figure 3a ), rotate the active roller 21 set in the device, rotate the driven roller 22 set inside the other end of the roller discharge chamber, and wrap around the two traction chains between the active roller and the driven roller and spaced apart from each other; one end of the mesh chain plate bypasses the active roller 21 and the driven roller 22 and is connected to its other end to form a circulating mesh chain plate, and the two ends of the mesh chain plate are connected to the traction chains on the corresponding sides, and the active roller 21 is connected to the drive assembly.

[0062] In a preferred embodiment, side plates are respectively provided on both sides of the circulating mesh chain plate conveying mechanism and along the conveying direction, and the side plates are connected by a plurality of reinforcing ribs passing through the upper mesh chain plate and the lower mesh chain plate.

[0063] Figure 5 This is a schematic structural diagram of one side of a drive assembly of a tobacco drum moisture removal filter anti-sticking device according to an embodiment of the present invention. Figure 6 This is a schematic structural diagram of the other side of the driving assembly of the tobacco drum moisture removal filter anti-sticking device according to an embodiment of the present invention. Figure 5As shown, the drive assembly includes a housing (see Figure 3a )、motor 31、reducer (located at Figure 5 The apparatus further comprises a driving shaft 33 and a transmission shaft 34, both ends of which are disposed on the side walls of the apparatus. The motor 31 is connected to the first end of the driving shaft 33 via a speed reducer. The second end of the driving shaft 33 is in transmission connection with the first end of the transmission shaft 34 and the first end of the driving roller 21. The second end of the transmission shaft 34 is in transmission connection with the brush roller. The transmission connection method described above may be a pulley drive, a sprocket drive, or the like, preferably a sprocket drive. In a preferred embodiment, the transmission shaft 34 can directly serve as the brush roller.

[0064] like Figure 6 As shown, reference numeral 6 denotes the inspection door seal, used to seal the cleaning box. Reference numeral 7 denotes the aforementioned drive shaft 33, which serves as the main drive bearing. One view shows the main drive bearing 7 from a cross-sectional perspective, while the other shows the main drive bearing 33 from a transverse perspective. Reference numeral 8 denotes the cleaning line, reference numeral 9 denotes the cleaning solenoid valve, and reference numeral 10 denotes the bearing seat. The cleaning line 8, as part of the stainless steel circulating filter cleaning assembly, is used to spray a high-pressure water jet. In one preferred embodiment, this is achieved using a nozzle, in another preferred embodiment, using a spray gun, and in yet another preferred embodiment, using a hydraulic cylinder. As a cleaning component, the cleaning pipeline 8 is also coordinated with a high-temperature steam pipeline, a heated water pipeline, a Venturi water pipeline, etc. By setting up a steam pipeline, stubborn sticking debris can be forcefully removed. By using a heated water pipeline, an atomized environment can be created to further avoid the adhesion of fine tobacco particles to the filter. By setting up a Venturi water pipeline, mold and other fine bacteria can be further prevented from taking root on the filter for a long time and contaminating the filter. A cleaning solenoid valve is set between these pipelines to achieve pipeline control. When the pressure in the cleaning component chamber is too high, part of the air pressure needs to be released through the cleaning solenoid valve.

[0065] Figure 7 It is an enlarged structural schematic diagram of the cleaning brush assembly of the tobacco drum moisture removal filter anti-sticking device according to an embodiment of the present invention. Figure 8 It is an enlarged structural schematic diagram of the scraper assembly of the tobacco drum moisture removal filter anti-sticking device according to an embodiment of the present invention. Figure 9 This is a schematic diagram of the structure of the brush roller of the tobacco drum moisture removal filter anti-sticking device according to an embodiment of the present invention. The cleaning brush assembly is connected to the stainless steel circulating filter at one end and close to the scraper assembly at the other end. The cleaning brush assembly includes a driving roller, air holes and a cleaning roller. The outer side is connected to a connecting rod. The connecting rod is provided with a rack-type transmission structure. When it moves, it drives the cleaning brush assembly to move together, thereby cleaning the debris on the filter screen and scraping it off through the scraper assembly. Figure 9As shown, the brush roller includes a rotating shaft 41 and cleaning rods 42. The ends of the rotating shaft 41 are rotatably mounted on the side walls of the device. A plurality of cleaning rods 42 are spaced apart on the side walls of the rotating shaft 41 to form a mace-shaped brush roller. By making such improvements to the cleaning brush assembly, the present invention ensures that each roller can operate smoothly during the cleaning process, minimizing the possibility of undesirable jamming or even tooth chipping caused by the drive mechanism.

[0066] In a preferred embodiment, the rotating shaft 41 adopts a square rotating shaft, and a plurality of cleaning rods 42 are arranged on each side wall of the square rotating shaft 41 and parallel to and spaced apart along its length direction. The middle part of each cleaning rod 42 is fixed on the side wall of the square rotating shaft, and the two ends of each cleaning rod 42 extend out of the two sides of the square rotating shaft 41 as cleaning teeth. Each cleaning rod 42 forms an angle with the cross section of the square rotating shaft 41. Such a brush roller can greatly improve the cleaning effect and avoid the end of the cleaning rod 42 from being stuck in the mesh of the mesh chain plate. At the same time, it can also reduce the breakage of the cleaning rod 42.

[0067] Figure 10 This is a schematic diagram of the flap seal assembly of the tobacco drum moisture-removal filter anti-sticking device according to an embodiment of the present invention. As shown, the flap seal assembly comprises a shaft 51 that rotates transversely within the drainage pipe, a flap 52 fixedly mounted on the shaft 51, a silicone pad 53 wrapped around the edges of the flap 52, and a control handle 54 attached to the exterior of the drainage pipe and connected to the shaft 51. The shape of the flap 52 is adapted to the cross-sectional shape of the drainage pipe. The control handle 54 can be used to adjust the angle between the flap 52 and the drainage pipe's cross-section, thereby varying the cross-sectional area between the upper and lower sections of the drainage pipe. As the angle between the flap 52 and the drainage pipe's cross-section increases between greater than 0° and less than 90°, the cross-sectional area between the upper and lower sections of the drainage pipe gradually increases. When the flap 52 is parallel to the drainage pipe's cross-section, the silicone pad 53 around the flap 52's edges seal against the inner wall of the drainage pipe, completely separating the upper and lower sections. When the flap 52 is perpendicular to the drainage pipe's cross-section, the upper and lower sections of the drainage pipe are fully connected.

[0068] In a preferred embodiment, the net chain can also be provided with a separate tensioning device. In a preferred embodiment, the moisture exhaust air volume is adjustable.

[0069] In addition to the above structure, in another preferred embodiment, the tobacco drum dehumidification filter anti-sticking device can be further equipped with other drive devices, such as drive bearings, frequency converters, and drive reducers. Key components of the frequency converter and reducer can be made of Siemens, ABB, Schneider, and other products of equal or higher quality. The circulating anti-sticking chain and housing are all made of 304 stainless steel to match the main equipment. Other components are selected from high-quality brand-name products; supporting parts are also made of stainless steel.

[0070] Example 2

[0071] This embodiment illustrates the construction of the learning model and pre-trained artificial intelligence model involved in the maintenance system.

[0072] Among them, Figure 12 As shown in the figure, a graphic representation of the degree of overall performance degradation is given. A 1% increase in the moisture content of the tobacco product is considered a decrease in performance level, and the long arrow represents the degree of overall performance degradation. For the sake of simplicity, the part after 20 time points, starting from the 25th time point, is shown in a small figure. Figure 12 The horizontal axis of the sub-graph corresponds to the corresponding position on the moisture content axis. The dashed lines in the figure indicate the 10% and 15% positions in the sub-graph. After 35 time points, the initial value of the drainage is approached, forming the moisture content increase curve shown in the figure, also known as the degradation function graph.

[0073] Figure 13 Based on Figure 12 The degradation function graph and time points of the learning model are constructed, and the method flow specifically includes:

[0074] S1 obtains the overall performance degradation levels of at least one set of stainless steel circulating filter, cleaning component, driving component, and collecting component at different time points on the time-dependent performance degradation function graph, and divides the data into training set and validation set.

[0075] S2 constructs the corresponding long-short time memory network of the stainless steel circulating filter, cleaning component, driving component, and collection component. Different time points are used as different node units (indicated by different shades of grass green) and the input vector is constructed according to the degree of performance degradation, thereby converting the training set and verification set into training vectors and verification vectors.

[0076] The S3 zero vector is input to the transmission end of the first node unit, and the training vectors are input to the corresponding different node units respectively. The result of the node unit output end is compared with the training vector of the input end of the next node unit. Finally, the training vector of the node unit input end is compared with the predicted result of its own output end to obtain the loss function value, thereby continuously optimizing the network parameters and verifying the prediction accuracy of each node unit through the verification vector (not shown in the figure). When the prediction accuracy of all node units stabilizes, the learning model is obtained.

[0077] like Figure 13 As shown, the artificial intelligence model pre-training method is:

[0078] Q1: Within every 1-15 days, select the randomly selected day and obtain an image of the mesh chain plate and scraper assembly and an image of the cleaning brush assembly every hour. Segment the images of the mesh chain plate and scraper assembly to divide the two areas belonging to the mesh chain plate and scraper assembly (represented by black and gray areas in the figure), and establish corresponding training and verification atlases together with the cleaning brush assembly images.

[0079] Q2 establishes a convolutional neural network (Res-CNN) with a residual mechanism, inputs the training atlas into Res-CNN, and classifies the degree of performance degradation at the output through the fully connected LC and softmax functions. A loss function is constructed to obtain the loss function value for each training. Backpropagation is used to optimize the Res-CNN parameters, and the accuracy is verified by the validation set. When the loss function value and accuracy stabilize, training is stopped to obtain a pre-trained artificial intelligence model.

[0080] Example 3

[0081] This embodiment is a technical description of the maintenance system.

[0082] In the present invention, the box installation and pipeline layout are beautiful and reasonable, easy to disassemble, and the welding parts are smooth without obvious welding points.

[0083] The technical parameters of the tobacco drum moisture removal filter anti-sticking device of the embodiment of the present invention and the test device a and the test device b in Example 1 are as follows:

[0084]

[0085]

[0086] When installing the equipment, first install the main body of the equipment, adjust it to be level, and tighten the anchor bolts; during installation, place the loose main body on the conveying bracket, connect the connecting bolts; and connect the power supply.

[0087] The preparations before the test run are as follows:

[0088] After installation is complete, inspect the installation quality. Develop a system commissioning plan and personnel arrangements. Organize relevant drawings and technical documentation, and familiarize yourself with the equipment's technical performance and the system's key technical parameters. Ensure that the power and other energy supplies required for the commissioning meet operational requirements. The commissioning site should be clean.

[0089] The test run of a single machine includes: checking the oil level of the reducer; checking whether the tension adjustment of the network chain meets the requirements; checking whether the blanking valve is closed; adjusting the air pressure and steam valves to ensure they are normal; and checking the direction of the chain operation.

[0090] The equipment needs to be maintained in the later stage, and the maintenance content includes:

[0091] 1. Routine maintenance operations: including inspections by machine operators or routine maintenance by personnel specifically responsible for inspections.

[0092] 2. Regular maintenance: This includes component replacement, lubrication, and adjustment, as well as maintenance of the test equipment's camera and lighting systems. This operation must be performed by qualified personnel. Regular maintenance of the machine is crucial to reducing both malfunctions and premature component wear. HDF assumes no responsibility for any component damage caused directly or indirectly by improper maintenance and will not accept any claims resulting therefrom. Always disconnect the main power supply during maintenance on mechanical and electrical components.

[0093] 1. Daily inspection: clean the waste collection bin; the motor runs without noise.

[0094] 2. Weekly inspection: Check that there is no leakage in steam, compressed air solenoid valves and pipelines; check whether the circulation network chain operates smoothly.

[0095] 3. Monthly inspection: Check the oil level of the reducer. If it is insufficient, add the specified lubricant as required; check the operation of the reducer to see if there is any abnormality; check whether the moving parts are loose or damaged, and eliminate any problems found.

[0096] Proper lubrication ensures machine efficiency and longevity, so regular lubrication of components is essential. The purpose of lubricating moving parts is to prevent direct contact between them. For optimal lubrication, a minimum amount of lubricant should be used. The minimum amount of lubricant used should be determined by the lubricant's additional functions, such as heat absorption. Over time, as the lubricant level decreases, lubricated parts must be cleaned and the lubricant replaced. Lubricants must strictly comply with the relevant requirements in this manual.

[0097] Combine Figure 4 , equipment lubrication can include the following:

[0098]

[0099] The following table shows possible fault operating conditions that may occur under normal operating conditions.

[0100]

[0101]

[0102] Whenever performing electrical work, ensure that all equipment drivers are disconnected. Lock any electrical switches to prevent accidental activation of the equipment by touching the switch button. Always read the manufacturer's instructions before performing any related operations.

[0103] On the equipment, all electronic control devices, except for those connected to the corresponding controlled motors, should be insulated from other parts of the equipment.

[0104] Before operating the equipment, check whether the emergency stop button is working properly. Factories can also add some emergency buttons according to their needs to ensure operational safety.

[0105] Before turning on the power, check whether the alarm indicator light and sound are normal.

[0106] The above description is merely a preferred embodiment of the present invention. The scope of protection of the present invention is not limited to the above embodiment. All technical solutions based on the concept of the present invention are within the scope of protection of the present invention. It should be noted that for those skilled in the art, various improvements and modifications that do not depart from the principles of the present invention should also be considered within the scope of protection of the present invention.

Claims

1. A maintenance system for a tobacco drum moisture removal filter anti-sticking device, characterized in that: The invention comprises a remote server for recording the update and maintenance time points of a tobacco drum dehumidification filter anti-sticking device having a stainless steel circulation filter, a drive component, a cleaning component and a collection component, and at least one test device with the same structure as the tobacco drum dehumidification filter anti-sticking device. The tobacco drum dehumidification filter anti-sticking device operates synchronously with the test device. According to the degree of decline in the quality level of tobacco products produced by the at least one test device during operation, a time-dependent performance degradation function diagram of the overall performance of at least one group of stainless steel circulation filters, cleaning components, driving components and collection components is recorded. The remote server judges the overall performance degradation prediction results of at least one group of stainless steel circulation filters, cleaning components, driving components and collection components of the current device in real time according to the time-dependent time points and the learning model constructed according to the time-dependent performance degradation function diagram. The remote server can display the prediction results in real time and download them to the mobile smart device of the staff for viewing. The stainless steel circulating filter screen includes an active roller, a driven roller and a mesh chain plate arranged inside the shell. The mesh chain plate is arranged between the active roller and the driven roller. One end of the mesh chain plate passes around the active roller and the driven roller and is connected to the other end of itself to form a circulating mesh chain plate; the driving assembly includes a housing, a motor, a reducer, a driving shaft and a transmission shaft. The motor is connected to the first end of the driving shaft through the reducer, and the second end of the driving shaft is transmission-connected to the first end of the transmission shaft and the first end of the active roller; the cleaning assembly includes a cleaning brush assembly and a scraper assembly. The cleaning brush assembly includes a brush roller, and the second end of the transmission shaft is transmission-connected to the brush roller; the collecting assembly includes a flap sealing assembly arranged at the bottom of the device; the second end of the transmission shaft is transmission-connected to the brush roller; A first high-definition industrial camera is provided on the casing of the at least one test device, and the first high-definition industrial camera is sealed and fixed by a first perforation at a preset position on the casing, and is used to regularly capture images of the mesh chain plate; a second high-definition industrial camera is provided at a preset position on the casing of the at least one test device, and the second high-definition industrial camera is sealed and fixed by a second perforation at a preset position on the casing, and is used to regularly capture images of the cleaning component; the mesh chain plate images and the cleaning component images are uploaded to the remote server, and a comprehensive evaluation of the performance degradation of each current circulation filter and component is given through the pre-trained artificial intelligence model and the degradation prediction results.

2. The maintenance system according to claim 1, characterized in that: The learning model is constructed by: S1 obtaining multiple groups of overall performance degradation degrees corresponding to different time points on a time-dependent performance degradation function graph of at least one set of stainless steel circulating filter, cleaning component, driving component, and collection component, and dividing the data into a training set and a validation set; S2 constructs the corresponding long-short time memory network for the stainless steel circulating filter, cleaning component, driving component, and collection component. It uses different time points as different node units and constructs input vectors based on the degree of performance degradation. It converts the training set and validation set into training vectors and validation vectors. The S3 zero vector is input into the transmission end of the first node unit, and the training vectors are input into the corresponding different node units respectively. The result of the node unit output end is compared with the training vector of the input end of the next node unit. Finally, the training vector of the node unit input end is compared with the predicted result of its own output end to obtain the loss function value, thereby continuously optimizing the network parameters, and verifying the prediction accuracy of each node unit through the verification vector. When the prediction accuracy of all node units stabilizes, the learning model is obtained, wherein the method for obtaining the degree of overall performance degradation of multiple groups corresponding to different time points is to obtain a group of overall performance degradation degrees every hour on a randomly selected day within every 1-15 days, and regard them as multiple groups of data on the degree of performance degradation at a time point, wherein the degree of performance degradation is evaluated by randomly sampling and measuring the moisture content of the tobacco material multiple times every hour on a randomly selected day; The artificial intelligence model pre-training method is: Q1: Within every 1-15 days, within the randomly selected day, acquire one image of the mesh chain plate and scraper assembly and one image of the cleaning brush assembly every hour, segment the images of the mesh chain plate and scraper assembly, and divide the two regions belonging to the mesh chain plate and scraper assembly into two regions, and establish corresponding training atlases and validation atlases together with the images of the cleaning brush assembly; Q2: Build a convolutional neural network (Res-CNN) with a residual mechanism. The training graph is fed into the Res-CNN. At the output, the fully connected LC and softmax functions are used to classify the degree of performance degradation. A loss function is constructed to obtain the loss value for each training session. Backpropagation is then used to optimize the Res-CNN parameters. The accuracy is verified using a validation set. Training is stopped when the loss value and accuracy stabilize, resulting in a pre-trained AI model. The comprehensive evaluation specifically includes: assuming the number of overall performance degradation levels is The performance degradation levels of mesh chain plate and cleaning component degradation prediction results are and , then the contribution of the driving component and the collection component to the number of overall performance degradation levels is , then and , The weighting factors can be statistically calculated based on the life and time points of the corresponding circulating filters and components in a specific individual test device.

3. The maintenance system according to claim 2, characterized in that: The transmission connection is a pulley drive or a sprocket drive.

4. The maintenance system according to claim 2 or 3, characterized in that: A first window embedded with a heating wire is provided on the first perforation, and the first high-definition industrial camera has a first window and a first lighting system for irradiating fill light into the first window; a second window embedded with a heating wire is provided on the second perforation, and the second high-definition industrial camera has a second window and a first lighting system for irradiating fill light into the second window, and the inner surfaces of the shells of the first window and the second window are cleaned regularly by opening the third inspection door and the first inspection door respectively.

5. The maintenance system according to claim 4, characterized in that: The brush roller includes a brush shaft and a cleaning rod. The cleaning rods are arranged at intervals on the side walls of the rotating shaft to form a mace-shaped brush roller. At least one flexible branch rod is arranged on one of the cleaning rods, and the brush shaft is a square rotating shaft; the flap sealing assembly includes a flap shaft that is arranged to rotate horizontally in the sewage pipe, a flap fixedly arranged on the flap shaft, a silicone pad wrapped around the edges of the flap, and a control handle that is arranged on the outside of the sewage pipe and connected to the rotating shaft.

Images