A brush cleaning machine nozzle running track control system

Abstract

The present application relates to the technical field of brush cleaning, and specifically discloses a brush cleaning machine nozzle running track control system, which comprises the following steps: obtaining brush data and cleaning data thereof in each cleaning task, performing feature extraction on the brush data and cleaning data thereof to obtain a cleaning reference value; synchronously obtaining a cleaning duration; statistically obtaining the cleaning reference value and cleaning duration of a preset number of cleaning tasks; training a neural network model of the cleaning reference value to the cleaning duration; and generating feedback data based on the neural network model. The present application obtains brush data and cleaning data thereof in multiple cleaning tasks, extracts a characteristic value, synchronously obtains a cleaning duration, constructs a neural network model of the characteristic value to the cleaning duration, and in actual use, the constructed neural network model can provide feedback on the user's trial adjustment, has extremely high flexibility, and can assist the user in making an optimal cleaning scheme in combination with actual conditions.

Description

Technical Field

[0001] This invention relates to the field of calligraphy brush cleaning technology, specifically a control system for the running trajectory of a calligraphy brush cleaning machine nozzle. Background Technology

[0002] Chinese calligraphy, an art form with a long history spanning approximately 4,000 years, is one of the most elegant arts and an important expression of traditional Chinese culture. Currently, calligraphy is increasingly popular, with over 50 million calligraphy enthusiasts worldwide, according to incomplete statistics. The aftermath of calligraphy practice and creation, including cleaning the brushes and ink dishes and disposing of wastewater, is referred to here as "calligraphy aftercare." Typically, calligraphy enthusiasts must clean their brushes and ink dishes after each session; otherwise, the gum in the ink can cause irreversible damage to the brush bristles. Alas, the remaining ink in the ink dish cannot be used after it dries. Even a small amount of ink that dries and clumps can damage the brush bristles if used for writing. Therefore, post-calligraphy cleaning is an unavoidable step. However, in reality, post-calligraphy cleaning often brings negative emotions and effects: it takes 3-5 minutes to clean a brush, and improper cleaning methods can damage the brush. Due to the nature of ink, it will get into dirty sinks, basins, and other containers during cleaning, requiring secondary cleaning. Ink on hands requires cleaning, and if ink is accidentally splashed on clothing or other items, it is even more difficult to clean.

[0003] Patent application CN111596610B discloses an industrial computer control system with trajectory measurement and control functions. It uses different levels of trajectory analysis (highly abnormal trajectory, excellent trajectory, in-depth trajectory anomaly bending, and in-depth trajectory constant variation) to edit and send abnormal trajectory curvature text to a display screen. The system then categorizes the trajectory into different regions and intervals. Based on the trajectory length, boundary area, and projected area, it makes targeted divisions of the measurement and control status of each segment of the trajectory. Furthermore, it performs in-depth analysis of the trajectory curvature by combining peak and trough amplitude, peak and trough number, and feature point spacing. Based on this analysis, it provides precise trajectory feedback and recording operations, making targeted judgments and in-depth processing of the overall trajectory to achieve a progressively detailed trajectory measurement and control effect.

[0004] Existing cleaning equipment lacks predictive capabilities, meaning users cannot easily anticipate the cleaning effects of different cleaning methods. This implies that the cleaning process relies entirely on a pre-set, fixed scheme, making it difficult for users to make adjustments. While a fixed scheme can provide a superior cleaning effect, the fact that different brushes have varying ink content (e.g., different usage times) reduces the likelihood of achieving even better cleaning results. Therefore, providing a feedback framework to offer users adjustment references is the technical problem this invention aims to solve. Summary of the Invention

[0005] The purpose of this invention is to provide a control system for the running trajectory of the nozzle of a calligraphy brush cleaning machine, so as to solve the problems mentioned in the background art.

[0006] To achieve the above objectives, the present invention provides the following technical solution:

[0007] A control system for the running trajectory of a brush cleaning machine nozzle, the system comprising:

[0008] The feature extraction module is used to acquire brush data and its cleaning data in each cleaning task, and to extract features from the brush data and its cleaning data to obtain cleaning reference values; wherein, the brush data includes weight data, length data and size data, and the cleaning data includes intensity data, frequency data and angle data;

[0009] The cleaning time calculation module is used to obtain the cleaning process for each cleaning task and calculate the cleaning time based on the cleaning process.

[0010] The model training module is used to statistically analyze the cleaning reference value and its cleaning duration for a preset number of cleaning tasks, and to train a neural network model from the cleaning reference value to the cleaning duration.

[0011] The feedback data generation module is used to open the information acquisition port when a cleaning request is received, acquire the current brush data and its cleaning data, extract the cleaning reference value of the current brush data and its cleaning data, and generate feedback data based on a neural network model.

[0012] As a further aspect of the present invention: the feature extraction module includes:

[0013] The weight data acquisition unit is used to monitor the device's operating signals in real time. When the device's operating signals are detected, the weight of the brush is collected by the weight sensing module built into the device and used as weight data.

[0014] The size data acquisition unit is used to acquire the length value and penholder circumference value of the brush based on the vision sensor built into the device, and obtain size data and length data.

[0015] The equipment parameter acquisition unit is used to acquire spray intensity, spray frequency and spray angle values ​​based on the sensors connected to the nozzle, and obtain intensity data, frequency data and angle data.

[0016] The data statistics unit is used to use weight data, size data, and length data as brush data, and intensity data, frequency data, and angle data as cleaning data.

[0017] The execution unit is used to extract features from the brush data and its cleaning data to obtain cleaning reference values.

[0018] As a further aspect of the present invention: the feature extraction of the brush data and its cleaning data to obtain the cleaning reference value includes:

[0019] By substituting into the calculation formula: Obtain the pen tip reference value, where Ab is the weight data, Ac is the size data, Ad is the length data, and A1 is the length coefficient;

[0020] By substituting into the calculation formula: The cleaning reference values ​​are obtained, where Bb is the intensity data, Bc is the frequency data, and Bd is the angle data.

[0021] As a further aspect of the present invention: the cleaning time calculation module includes:

[0022] The image acquisition unit is used to acquire images of the cleaning water flow in real time based on the vision sensor built into the device;

[0023] A color value recognition unit is used to perform color value recognition on the image and record the cleaning time based on the color value recognition result.

[0024] As a further aspect of the present invention: the content of color value recognition of the image includes:

[0025] Perform grayscale conversion on the image, read the grayscale value of each pixel in the converted image, and calculate the average grayscale value of the image;

[0026] Calculate the rate of change of the average grayscale value based on the timestamp of the image;

[0027] When the rate of change reaches a preset threshold, the corresponding time point is marked;

[0028] Read the image at the specified time point and determine the start and end times based on the average grayscale value;

[0029] Specifically, when the average grayscale value is less than a preset first threshold, the earliest time point is marked as the start time, and when the average grayscale value is greater than a preset second threshold, the latest time point is marked as the end time.

[0030] As a further aspect of the present invention: the model training module includes:

[0031] The sample set construction unit is used to statistically analyze the cleaning reference values ​​and cleaning duration of a preset number of cleaning tasks and construct a sample set.

[0032] The sample set splitting unit is used to divide the sample set into a training set containing 70% of the data, a test set containing 15% of the data, and a validation set containing 15% of the data.

[0033] The first center selection unit is used to obtain historical feature vectors in the training set. It pre-defines inefficient data clusters L1, normal data clusters L2, and efficient data clusters L3. In the training set, three data points are manually selected as the first cluster centers, representing inefficient data clusters L1, normal data clusters L2, and efficient data clusters L3, respectively.

[0034] By substituting into the calculation formula The distances between data items are obtained. The distances between each data item in the training set and the inefficient data cluster L1, the normal data cluster L2, and the efficient data cluster L3 are calculated. Each data item is then assigned to the nearest data cluster, resulting in three new data clusters that serve as the second cluster centers. Here, x and y are the coordinates of the data points, where x... i and y i Let be the values ​​of the two data points on the i sub-data items, and n be the number of sub-data items;

[0035] Calculate the mean of the three new data clusters in the second cluster center, and use them as the new first cluster center for recalculation;

[0036] The process is repeated until the preset number of iterations is reached, resulting in a neural network model from the cleaning reference value to the cleaning duration.

[0037] As a further aspect of the present invention: the feedback data generation module includes:

[0038] The port opening unit is used to open the information acquisition port when a cleaning request is received.

[0039] The port application unit is used to acquire the current brush data and its cleaning data based on the information acquisition port.

[0040] The extraction execution unit is used to extract features from the acquired current brush data and its cleaning data to obtain the current cleaning reference value;

[0041] The duration output unit is used to input the current cleaning reference value into the trained neural network model and output the cleaning duration as feedback data.

[0042] As a further aspect of the present invention: the system includes:

[0043] The image acquisition unit is used to acquire images of the cleaning water flow in real time based on the vision sensor built into the device;

[0044] The cleaning position acquisition unit is used to acquire the cleaning position of the brush at various times based on the pressure sensor built into the fixture;

[0045] The duration segmentation unit is used to cluster images at each time point based on the cleaning location and calculate the cleaning duration at different cleaning locations.

[0046] The ink collection volume acquisition unit is used to determine the ink collection volume at different cleaning positions based on the cleaning time and the preset cleaning speed; the cleaning speed is the amount of ink collected per unit time.

[0047] The 3D modeling unit is used to construct a 3D model of the brush based on the amount of ink collected at different cleaning locations, and then feed this model back to the user.

[0048] Compared with the prior art, the beneficial effects of the present invention are: the present invention obtains brush data and cleaning data from multiple cleaning tasks, extracts feature values, synchronously obtains cleaning time, and constructs a neural network model from feature values ​​to cleaning time. In actual use, the constructed neural network model can provide feedback on the user's trial adjustments, which is highly flexible and can help the user make the optimal cleaning plan based on the actual situation. Attached Figure Description

[0049] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention.

[0050] Figure 1 This is a block diagram showing the composition of the nozzle trajectory control system for a calligraphy brush cleaning machine. Detailed Implementation

[0051] To make the technical problems to be solved, the technical solutions, and the beneficial effects of the present invention clearer, the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the present invention and are not intended to limit the present invention.

[0052] Figure 1 This is a block diagram illustrating the structural composition of a brush cleaning machine nozzle trajectory control system. In this embodiment of the invention, a brush cleaning machine nozzle trajectory control system 10 includes:

[0053] The feature extraction module 11 is used to acquire brush data and its cleaning data in each cleaning task, and to extract features from the brush data and its cleaning data to obtain cleaning reference values; wherein, the brush data includes weight data, length data and size data, and the cleaning data includes intensity data, frequency data and angle data;

[0054] The cleaning time calculation module 12 is used to obtain the cleaning process of each cleaning task and calculate the cleaning time according to the cleaning process.

[0055] Model training module 13 is used to statistically analyze the cleaning reference value and its cleaning duration for a preset number of cleaning tasks, and to train a neural network model from the cleaning reference value to the cleaning duration.

[0056] The feedback data generation module 14 is used to open the information acquisition port when a cleaning request is received, acquire the current brush data and its cleaning data, extract the cleaning reference value of the current brush data and its cleaning data, and generate feedback data based on the neural network model.

[0057] The cleaning process is continuous. Each time a cleaning request is received, the cleaning equipment will execute a cleaning task. In each cleaning task, the brush data and its cleaning data are acquired. Feature extraction is performed on the brush data and its cleaning data to obtain cleaning reference values. The brush data is the physical information of the brush itself, including weight data, length data and size data. The cleaning data is the parameters of the equipment during the cleaning process, including intensity data, frequency data and angle data.

[0058] Furthermore, the cleaning process for each cleaning task is obtained, and the cleaning time is calculated based on the cleaning process. The cleaning process is the cleaning procedure, which includes rinsing the brush with the nozzle along a preset trajectory. The cleaning time is the time consumed by one cleaning task. The cleaning reference value and its cleaning time for multiple cleaning tasks are statistically analyzed, and a neural network model from the cleaning reference value to the cleaning time is trained, where the cleaning reference value is used as input and the cleaning time is used as output.

[0059] In practical applications, the client inputs a cleaning request and opens an information acquisition port. Based on this port, the current brush data and its cleaning data are obtained. During this process, the brush data is fixed, while the cleaning data is input by the client, indicating how to clean. After extracting the cleaning reference values ​​of the current brush data and its cleaning data, the data is input into a trained neural network model, which outputs the cleaning duration. The cleaning duration itself reflects the cleaning efficiency. After feedback to the client, the client can adjust the cleaning data independently, resulting in different cleaning durations. After multiple attempts, the client can obtain a better cleaning duration, and the corresponding cleaning data serves as the final reference for the equipment, which is then converted into control commands.

[0060] It is worth mentioning that the working process of the feedback data generation module 14 is actually equivalent to a cleaning task. It will be used as a new sample to recursively update the training process of the neural network model.

[0061] In a preferred embodiment of the technical solution of the present invention, the feature extraction module 11 includes:

[0062] The weight data acquisition unit is used to monitor the device's operating signals in real time. When the device's operating signals are detected, the weight of the brush is collected by the weight sensing module built into the device and used as weight data.

[0063] The size data acquisition unit is used to acquire the length and circumference of the brush based on the vision sensor built into the device, and obtain size data and length data.

[0064] The above information is used to obtain data about the calligraphy brush, including weight, size, and length.

[0065] The equipment parameter acquisition unit is used to acquire spray intensity, spray frequency and spray angle values ​​based on the sensors connected to the nozzle, and obtain intensity data, frequency data and angle data; wherein, the signals acquired by the sensors connected to the nozzle all contain timestamps.

[0066] The above information is used to obtain cleaning data, including intensity data, frequency data, and angle data. It should be explained that spray intensity is the amount of water sprayed by the nozzle per unit time, for example, 0.2 liters of water sprayed per minute; spray frequency is the number of times the nozzle sprays water per unit time, for example, 60 times of water sprayed per minute; and spray angle is the angle between the direction of water spraying by the nozzle and the axis of the nozzle, for example, a spray angle of 30 degrees.

[0067] The data statistics unit is used to use weight data, size data, and length data as brush data, and intensity data, frequency data, and angle data as cleaning data.

[0068] The execution unit is used to extract features from the brush data and its cleaning data to obtain cleaning reference values.

[0069] Finally, the brush data and cleaning data were statistically obtained, and features were extracted from the brush data and cleaning data to obtain cleaning reference values.

[0070] Specifically, the process of extracting features from the brush data and its cleaning data to obtain cleaning reference values ​​includes:

[0071] By substituting into the calculation formula: Obtain the pen tip reference value, where Ab is the weight data, Ac is the size data, Ad is the length data, and A1 is the length coefficient;

[0072] By substituting into the calculation formula: The cleaning reference values ​​are obtained, where Bb is the intensity data, Bc is the frequency data, and Bd is the angle data.

[0073] In a preferred embodiment of the technical solution of the present invention, the cleaning time calculation module 12 includes:

[0074] The image acquisition unit is used to acquire images of the cleaning water flow in real time based on the vision sensor built into the device;

[0075] A color value recognition unit is used to perform color value recognition on the image and record the cleaning time based on the color value recognition result.

[0076] In one embodiment of the technical solution of the present invention, an image of the cleaning water flow is acquired in real time based on a vision sensor built into the device, the image is color value identified, and the cleaning time is recorded according to the color value identification result. Since ink has a clear color value, such as black (a small part is other colors), the identification process is not complicated. Based on the identification process to identify the ink color, the precise start time and end time of cleaning can be determined, and thus the accurate cleaning time can be determined.

[0077] Furthermore, the content of color value recognition of the image includes:

[0078] Perform grayscale conversion on the image, read the grayscale value of each pixel in the converted image, and calculate the average grayscale value of the image;

[0079] Calculate the rate of change of the average grayscale value based on the timestamp of the image;

[0080] When the rate of change reaches a preset threshold, the corresponding time point is marked;

[0081] Read the image at the specified time point and determine the start and end times based on the average grayscale value;

[0082] Specifically, when the average grayscale value is less than a preset first threshold, the earliest time point is marked as the start time, and when the average grayscale value is greater than a preset second threshold, the latest time point is marked as the end time.

[0083] In one embodiment of the technical solution of this invention, a specific color value recognition process is provided. The image is converted to grayscale, the grayscale values ​​of each pixel in the converted image are read, and the average grayscale value of the image is calculated. Since each image has a timestamp, the obtained average grayscale value is also the average grayscale value at each time point. The difference between the average grayscale values ​​of adjacent time points is calculated, called the rate of change. Since this application only needs to locate the start and end times, the rate of change at these times will be very high, such as transparent turning black and black turning transparent. Therefore, it is only necessary to specifically analyze the images corresponding to the times with larger rates of change to filter out the start and end times. Specifically, when the average grayscale value is less than a preset first threshold, it indicates that the image is sufficiently black (a grayscale value of 0 is black), and the earliest time point is marked as the start time. When the average grayscale value is greater than a preset second threshold, it indicates that the image is sufficiently white (a grayscale value of 255 is white), and the latest time point is marked as the end time.

[0084] In a preferred embodiment of the technical solution of the present invention, the model training module 13 includes:

[0085] The sample set construction unit is used to statistically analyze the cleaning reference values ​​and cleaning duration of a preset number of cleaning tasks and construct a sample set.

[0086] The sample set splitting unit is used to divide the sample set into a training set containing 70% of the data, a test set containing 15% of the data, and a validation set containing 15% of the data.

[0087] The first center selection unit is used to obtain historical feature vectors in the training set. It pre-defines inefficient data clusters L1, normal data clusters L2, and efficient data clusters L3. In the training set, three data points are manually selected as the first cluster centers, representing inefficient data clusters L1, normal data clusters L2, and efficient data clusters L3, respectively.

[0088] By substituting into the calculation formula The distances between data items are obtained. The distances between each data item in the training set and the inefficient data cluster L1, the normal data cluster L2, and the efficient data cluster L3 are calculated. Each data item is then assigned to the nearest data cluster, resulting in three new data clusters that serve as the second cluster centers. Here, x and y are the coordinates of the data points, where x... i and y i Let be the values ​​of the two data points on the i sub-data items, and n be the number of sub-data items;

[0089] Calculate the mean of the three new data clusters in the second cluster center, and use them as the new first cluster center for recalculation;

[0090] The process is repeated until the preset number of iterations is reached, resulting in a neural network model from the cleaning reference value to the cleaning duration.

[0091] In a preferred embodiment of the technical solution of the present invention, the feedback data generation module 14 includes:

[0092] The port opening unit is used to open the information acquisition port when a cleaning request is received.

[0093] The port application unit is used to acquire the current brush data and its cleaning data based on the information acquisition port.

[0094] The extraction execution unit is used to extract features from the acquired current brush data and its cleaning data to obtain the current cleaning reference value;

[0095] The duration output unit is used to input the current cleaning reference value into the trained neural network model and output the cleaning duration as feedback data.

[0096] In one example of the technical solution of the present invention, when a cleaning request is received, the information acquisition port is opened, and the current brush data and its cleaning data are acquired based on the information acquisition port. Feature extraction is performed on the acquired current brush data and its cleaning data to obtain the current cleaning reference value. The feature extraction scheme adopts the feature extraction scheme mentioned above. The current cleaning reference value is input into the trained neural network model, and the cleaning time is output as feedback data.

[0097] As a preferred embodiment of the technical solution of the present invention, the system includes:

[0098] The image acquisition unit is used to acquire images of the cleaning water flow in real time based on the vision sensor built into the device;

[0099] The cleaning position acquisition unit is used to acquire the cleaning position of the brush at various times based on the pressure sensor built into the fixture;

[0100] The duration segmentation unit is used to cluster images at each time point based on the cleaning location and calculate the cleaning duration at different cleaning locations.

[0101] The ink collection volume acquisition unit is used to determine the ink collection volume at different cleaning positions based on the cleaning time and the preset cleaning speed; the cleaning speed is the amount of ink collected per unit time.

[0102] The 3D modeling unit is used to construct a 3D model of the brush based on the amount of ink collected at different cleaning locations, and then feed this model back to the user.

[0103] In one embodiment of the technical solution of this invention, an additional solution is also provided. Based on the visual sensor introduced in this application, during the actual cleaning process, the cleaning position of the brush at each moment can be obtained based on the pressure sensor built into the fixture. On this basis, the image of the cleaning water flow is obtained in real time based on the visual sensor built into the device. At this time, the obtained data includes the cleaning position and the image at each time point. Based on time, the cleaning position and the image are registered, and then all images of the same cleaning position are clustered into one category to obtain the image of the same cleaning position at different times. Finally, the cleaning time of different cleaning positions is obtained. This process is equivalent to a detailed refinement of the original cleaning time calculation process, and the cleaning time of each cleaning position is obtained.

[0104] For a printhead, the cleaning speed corresponding to its cleaning data is known and obtainable, that is, the amount of ink washed per unit time. By calculating the product of the cleaning time and the cleaning speed, the total amount of ink washed can be obtained, which is used as the amount of ink gathered at the cleaning position. Finally, a three-dimensional model of the brush is constructed based on the amount of ink gathered at different cleaning positions and fed back to the user.

[0105] The practical significance of the above solution is that the person who cleans the brush is usually the owner of the brush. When using the brush, they have their own habits. When they know how much ink is gathered in different places when using the brush, they can adjust their usage habits during use, thereby increasing the lifespan of the brush and making the brush use process more even.

[0106] As a preferred embodiment of the technical solution of the present invention, the cleaning time reflects the efficiency issue. The cleaning time can also have other applications, as follows:

[0107] A preset cleaning threshold range (W1, W2) is defined. When the cleaning time is less than W1, an efficient signal is generated. When the cleaning time is greater than W1 and less than W2, a normal signal is generated. When the cleaning time is greater than W2, an inefficient signal is generated.

[0108] The high-efficiency signal contains a set of fields representing high brush cleaning efficiency, the normal signal contains a set of fields representing average brush cleaning efficiency, and the low-efficiency signal contains a set of fields representing low cleaning efficiency.

[0109] High-efficiency signals, normal signals, and low-efficiency signals are packaged to obtain the cleaning and classification results.

[0110] Based on this, the nozzle cleaning trajectory can be adjusted according to the cleaning classification results. The methods for adjusting the nozzle cleaning trajectory based on the cleaning classification results include:

[0111] Based on the fast positioning command, a stop command is generated when the cleaning and classification result is a high-efficiency signal, a low-speed command is generated when the cleaning and classification result is a normal signal, and a high-speed command is generated when the cleaning and classification result is an inefficient signal.

[0112] The stop command contains a set of fields representing that the nozzle position does not change; the low speed command contains a set of fields representing that the nozzle moves at a slow speed; the high speed command contains a set of fields representing that the nozzle moves at a higher speed.

[0113] The packaged stop command, low-speed command, and high-speed command are used to obtain the cleaning and adjustment results.

[0114] Furthermore, the cleaning and sorting results can be analyzed. The specific analysis process is as follows:

[0115] Methods for analyzing cleaning and sorting results include:

[0116] When the cleaning and sorting result is a high-efficiency signal, a reminder SMS is sent to the user through the communication unit; when the cleaning and sorting result is a normal signal, an SMS message indicating adjustment is needed is sent to the user through the communication unit; when the cleaning and sorting result is an inefficient signal, an SMS message indicating urgent adjustment is sent to the user through the communication unit.

[0117] The reminder SMS message includes information that the current equipment is cleaning efficiently, and reminds users to maintain the equipment regularly, change the cleaning solution regularly, and observe the cleaning effect.

[0118] The text message to be adjusted includes an explanation that the current equipment cleaning efficiency is average, reminding the user to adjust the cleaning dataset according to their own needs;

[0119] The urgent adjustment message includes an explanation that the current equipment cleaning efficiency is low, requiring the user to adjust the cleaning dataset, observe the nozzle operation status, and choose whether to replace the cleaning nozzle based on the operation status.

[0120] The system packages reminder SMS messages, SMS messages awaiting adjustment, and SMS messages requiring urgent adjustment to obtain the cleaning decision results.

[0121] The mechanical architecture of the technical solution of this invention generally includes four modules: a brush cleaning module, an ink disc cleaning module, a water circulation module, and an electromechanical module, which are described in detail below:

[0122] I. Brush Cleaning Module:

[0123] 1. Main components: pen holder, servo motor, gears, rack, push rod, waterproof silicone sleeve, nozzle, water valve, cleaning chamber body and other supporting components.

[0124] 2. Working Principle: This application employs a mechanical structure that simulates the reciprocating motion of a finger squeezing the brush bristles. During the cleaning process, the most stubborn part is the base of the bristles, as this is the most compact and the ink that has seeped in is the most difficult to remove efficiently and quickly. In hand washing, pressing the brush belly is the most practical method. Therefore, in the experiment, we immersed the brush bristles below the water surface and, simulating finger squeezing, repeatedly squeezed the bristles 100 times in 125mL of clean water, then changed the water. This was one cleaning sub-process; the sub-process was repeated 4 times, using a total of 500mL of water. The cleanliness of the brush cleaned using this method is close to that of hand washing. Simultaneously, the waterproof silicone sleeve can fully transmit the squeezing force while protecting the bristles and brush handle from damage caused by the mechanical squeezing force. This design not only upgrades the cleaning effect but also saves cleaning water.

[0125] 3. Technical Solution:

[0126] A. The brush is fixed by the brush holder, and after fixing, the brush bristles are in the middle of the two waterproof silicone sleeves.

[0127] B. Pour 125ml of water into the cleaning chamber through the nozzle, ensuring the water level covers the brush bristles.

[0128] C. The servo motor drives the gears to rotate, causing two racks to reciprocate in opposite directions. The push rods on the racks create a cyclical squeezing effect on the brush bristles within the cleaning chamber. After a certain number of repetitions, the ink dissolved in the water within the cleaning chamber reaches its maximum concentration. The motor then stops operating.

[0129] D. The water valve is opened, and the sewage flows out of the cleaning chamber. The water valve is then closed.

[0130] The above BCD steps constitute a cleaning sub-process, controlled by the circuit system. The cleaning sub-process is repeated a certain number of times according to the characteristics of the brush and the user's requirements, and the brush cleaning work is considered to be completed.

[0131] II. Disc Cleaning Module:

[0132] 1. Main components: ink disc cleaning chamber, servo motor, gears, elastic telescopic rod, brush head, bottom suction cup, water inlet, water valve, etc.

[0133] 2. Working Principle: This invention is primarily compatible with near-circular ink discs with a diameter less than 150mm. The upper end of the elastic telescopic rod connects to the drive mechanism, and the lower end connects to the brush head. When the ink disc is placed in the cleaning chamber, it is fixed to the suction cup at the bottom by the pushing force of the brush head. The motor drives the brush head to rotate, and the brush head rubs against the ink disc, thus cleaning it.

[0134] 3. Technical Solution:

[0135] A. After the ink disc is placed in the cleaning chamber, it sinks to the bottom.

[0136] B. Close the cleaning chamber lid. The spring in the elastic telescopic rod is compressed, pushing the ink disc to adhere to the suction cup at the bottom.

[0137] C. A certain volume of water is injected into the tank through the water inlet.

[0138] D. The motor drives the brush head to rotate at a speed of M revolutions per minute. The brush head rubs against the ink disk to wash away the ink.

[0139] E. The water valve is opened, and the sewage flows out of the cleaning chamber. The water valve is then closed.

[0140] The CDE steps above constitute a cleaning sub-process B, controlled by the circuit system. The cleaning sub-process is repeated a certain number of times, at which point the disc cleaning operation is considered complete.

[0141] III. Water Circulation Module:

[0142] 1. Main components: water pump, clean water tank, wastewater tank, PVDF ultrafiltration cartridge, and several water pipes.

[0143] 2. Working Principle: Ink consists of two main components: carbon particles and colloids. The working principle of ink is that the carbon particles adhere to the colloids, which then stick to paper or other objects. In this application, to solve the series of problems caused by ink, it is temporarily impossible to start by decomposing the carbon particles and colloids. On the one hand, carbon particles are extremely stable and almost impossible to decompose or transform; on the other hand, in the water circulation design of this application, any chemical reagents will eventually come into contact with the brush bristles with the water, damaging them and thus negating the meaning and value of the entire solution. Therefore, we chose a physical method: separating the carbon particles, colloids, and water. The size of carbon particles is approximately 50-300nm, and the size of colloid particles is approximately 10-200nm. Utilizing this physical property, a 10-inch PVDF ultrafiltration cartridge with a pore size of 10nm was selected as the filtration tool in the experiment, successfully achieving "black and white separation" with stable results, completing the most important link in the water circulation module of this application. The separated water can be reused to clean the brush and ink tray.

[0144] 3. Technical Solution: A water pump draws purified water from the purified water tank, which then flows through water pipes into the brush cleaning module and ink disc cleaning module. After the cleaning process (regardless of the specific work sequence), the wastewater generated flows through water pipes into the wastewater tank. The water pump then draws the wastewater into the PVDF ultrafiltration cartridge. After filtration by the ultrafiltration cartridge, carbon particles and most colloids are separated within the PVDF ultrafiltration cartridge, while water and a small amount of colloids flow through water pipes into the purified water tank, awaiting the next cleaning process or replacement. After filtering a certain amount of ink, the PVDF ultrafiltration cartridge will reach its filtration limit and will need to be replaced before filtration can resume.

[0145] IV. Electromechanical Modules:

[0146] 1. Main components: circuit boards, motors, water pumps, valves, wires, etc.

[0147] 2. Working principle: The brush cleaning module, ink disc cleaning module and water circulation module work together under the control of the circuit program to complete the post-calligraphy cleaning work.

[0148] It is important to note that calligraphy brushes come in different materials and sizes, and calligraphy enthusiasts have individualized usage preferences. Therefore, there is no universal cleaning process to complete the cleaning of calligraphy brushes. It is necessary to provide options for combining cleaning time T or allow users to directly set the time T to flexibly adjust the cleaning process and time in the scenario. The combination scheme of preset time for different specifications needs to be written into the embedded program in advance.

[0149] Assuming we preset the cleaning time: the brush hair material determines t1, with a cleaning time of 5 minutes for sheep hair brushes, 4 minutes for mixed hair brushes, and 3 minutes for wolf hair brushes; the brush hair size determines t2, with a cleaning time of 3 minutes for brush hairs longer than 5cm, 2 minutes for 3-5cm, and 1 minute for brush hairs shorter than 3cm (this data is for illustrative purposes only). In one use, the user selects a sheep hair brush with a bristle length of 3-5cm; the system presets the cleaning time t1 for sheep hair brushes to be 5 minutes and the cleaning time t2 for brushes with a bristle length of 3-5cm to be 2 minutes, then T = t1 + t2 = 7 minutes.

[0150] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A control system for the running trajectory of a brush cleaning machine nozzle, characterized in that, The system includes: The feature extraction module is used to acquire brush data and its cleaning data in each cleaning task, and to extract features from the brush data and its cleaning data to obtain cleaning reference values; wherein, the brush data includes weight data, length data and size data, and the cleaning data includes intensity data, frequency data and angle data; The cleaning time calculation module is used to obtain the cleaning process for each cleaning task and calculate the cleaning time based on the cleaning process. The model training module is used to statistically analyze the cleaning reference value and its cleaning duration for a preset number of cleaning tasks, and to train a neural network model from the cleaning reference value to the cleaning duration. The feedback data generation module is used to open the information acquisition port when a cleaning request is received, acquire the current brush data and its cleaning data, extract the cleaning reference value of the current brush data and its cleaning data, and generate feedback data based on the neural network model. The feature extraction module includes: The weight data acquisition unit is used to monitor the device's operating signals in real time. When the device's operating signals are detected, the weight of the brush is collected by the weight sensing module built into the device and used as weight data. The size data acquisition unit is used to acquire the length value and penholder circumference value of the brush based on the vision sensor built into the device, and obtain size data and length data. The equipment parameter acquisition unit is used to acquire spray intensity, spray frequency and spray angle values ​​based on the sensors connected to the nozzle, and obtain intensity data, frequency data and angle data. The data statistics unit is used to use weight data, size data, and length data as brush data, and intensity data, frequency data, and angle data as cleaning data. The execution unit is used to extract features from the brush data and its cleaning data to obtain cleaning reference values. The process of extracting features from the brush data and its cleaning data to obtain the cleaning reference value includes: By substituting into the calculation formula: The pen tip reference value is obtained, where, For weight data, For size data, For length data, This is the length coefficient; By substituting into the calculation formula: The cleaning reference value is obtained, where, For intensity data, For frequency data, This is angle data.

2. The brush cleaning machine nozzle trajectory control system according to claim 1, characterized in that, The cleaning time calculation module includes: The image acquisition unit is used to acquire images of the cleaning water flow in real time based on the vision sensor built into the device; A color value recognition unit is used to perform color value recognition on the image and record the cleaning time based on the color value recognition result.

3. The brush cleaning machine nozzle trajectory control system according to claim 2, characterized in that, The content of color value recognition of the image includes: Perform grayscale conversion on the image, read the grayscale value of each pixel in the converted image, and calculate the average grayscale value of the image; Calculate the rate of change of the average grayscale value based on the timestamp of the image; When the rate of change reaches a preset threshold, the corresponding time point is marked; Read the image at the specified time point and determine the start and end times based on the average grayscale value; Specifically, when the average grayscale value is less than a preset first threshold, the earliest time point is marked as the start time, and when the average grayscale value is greater than a preset second threshold, the latest time point is marked as the end time.

4. The brush cleaning machine nozzle trajectory control system according to claim 1, characterized in that, The model training module includes: The sample set construction unit is used to statistically analyze the cleaning reference values ​​and cleaning duration of a preset number of cleaning tasks and construct a sample set. The sample set splitting unit is used to divide the sample set into a training set containing 70% of the data, a test set containing 15% of the data, and a validation set containing 15% of the data. The first center selection unit is used to obtain historical feature vectors in the training set. It pre-defines inefficient data clusters L1, normal data clusters L2, and efficient data clusters L3. In the training set, three data points are manually selected as the first cluster centers, representing inefficient data clusters L1, normal data clusters L2, and efficient data clusters L3, respectively. By substituting into the calculation formula The distances between data items are obtained. The distances between each data item in the training set and the inefficient data cluster L1, the normal data cluster L2, and the efficient data cluster L3 are calculated. Each data item is then assigned to the nearest data cluster, resulting in three new data clusters that serve as the second cluster centers. and These are the coordinates of the data points. and For the two data points respectively at The values ​​of each sub-data item The number of sub-data items; Calculate the mean of the three new data clusters in the second cluster center, and use them as the new first cluster center for recalculation; The process is repeated until the preset number of iterations is reached, resulting in a neural network model from the cleaning reference value to the cleaning duration.

5. The brush cleaning machine nozzle trajectory control system according to claim 1, characterized in that, The feedback data generation module includes: The port opening unit is used to open the information acquisition port when a cleaning request is received. The port application unit is used to acquire the current brush data and its cleaning data based on the information acquisition port. The extraction execution unit is used to extract features from the acquired current brush data and its cleaning data to obtain the current cleaning reference value; The duration output unit is used to input the current cleaning reference value into the trained neural network model and output the cleaning duration as feedback data.

6. The brush cleaning machine nozzle trajectory control system according to claim 2, characterized in that, The system includes: The image acquisition unit is used to acquire images of the cleaning water flow in real time based on the vision sensor built into the device; The cleaning position acquisition unit is used to acquire the cleaning position of the brush at various times based on the pressure sensor built into the fixture; The duration segmentation unit is used to cluster images at each time point based on the cleaning location and calculate the cleaning duration at different cleaning locations. The ink collection volume acquisition unit is used to determine the ink collection volume at different cleaning positions based on the cleaning time and the preset cleaning speed; the cleaning speed is the amount of ink collected per unit time. The 3D modeling unit is used to construct a 3D model of the brush based on the amount of ink collected at different cleaning locations, and then feed this model back to the user.

Images