Method and system for automatically adjusting cleaning device based on data analysis

Through the YOLOV5 neural network positioning and grabbing, PID algorithm to control temperature, sensor monitoring liquid level and convolutional neural network to identify drying status, the automatic adjustment of the cleaning device is realized, solving the damage of traditional cleaning methods and the inadequate drying of drying in time and liquid level control issues improve cleaning efficiency and accuracy.

CN118915440APending Publication Date: 2024-11-08JIANGMEN SAITE PRECISION CERAMICS CO LTD
View PDF 1 Cites 0 Cited by

Patent Information

Application Number
CN202410953326.X
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-07-16
Publication Date
2024-11-08

AI Technical Summary

Technical Problem

The traditional cleaning methods have problems such as easily causing damage to clean items, untimely drying and lack of cleaning liquid, and cannot achieve full automation control, and the cleaning device cannot be controlled in real time based on the cleaning status data.

Method used

Using the automatic adjustment method of cleaning device based on data analysis, the cleaning materials are located and captured through the YOLOV5 neural network, the cleaning temperature is controlled using the PID algorithm and the tuna group optimization algorithm, the sensor is set to monitor the cleaning liquid level, and the drying is identified through the convolutional neural network. Dry state, realizes automatic grasping, liquid level replenishment and temperature regulation.

Benefits of technology

It improves the accuracy and speed of the robotic arm grabbing, realizes real-time supplementation of cleaning fluid and precise temperature control, improves the accuracy of drying status recognition, and solves the automation and optimization problems of traditional cleaning methods.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118915440A_ABST
    Figure CN118915440A_ABST
Patent Text Reader

Abstract

The invention relates to the technical field of general control or regulation systems, and discloses a method and a system for automatically regulating a cleaning device based on data analysis. The method comprises the following steps: firstly, checking the weight and quantity of a cleaning material, performing coordinate labeling and feature extraction on a to-be-detected cleaning material picture, inputting the to-be-detected cleaning material picture into a YOLOV5 neural network for training, and taking a central point of an output coordinate as a mechanical arm grabbing point; secondly, the cleaning fluid is automatically discharged or supplemented, and an abnormal state is prompted and alarmed; then a PID controller is used for adjusting the temperature, a tuna swarm optimization algorithm is used for optimizing a proportionality coefficient, an integral time coefficient and a differential time coefficient, and the algorithm is high in convergence speed and robustness; and finally, based on the convolutional neural network, through training the convolutional neural network, the drying state of the dried cleaning material image is identified, and the identification accuracy is high.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the technical field of general control or regulation systems, and in particular to an automatic regulation method and system for a cleaning device based on data analysis. Background Art

[0002] Chinese patent CN112232039B discloses a control method and system for an ultrasonic cleaning device, which specifically includes: relying on the historical data of the ultrasonic cleaning device, constructing a cleaning mode with cleaning materials as parameters, setting the index threshold in the cleaning mode, and establishing a perfect working mode; then integrating and applying the cleaning mode and index threshold after setting with the controller in the ultrasonic cleaning device, converting the control signal into control data based on the controller, and completing the automatic control of the ultrasonic cleaning device; then putting the cleaning material into the ultrasonic cleaning device, and selecting the corresponding cleaning mode according to the judgment of the controller; finally, monitoring the cleaning mode and preparation threshold in real time according to the actual situation, and optimizing and adjusting in real time to realize the control of the ultrasonic cleaning device. However, the present invention does not realize cleaning automation, nor optimize the cleaning process, and there are problems such as slow cleaning process.

[0003] With the continuous improvement of cleaning level, traditional cleaning methods have problems such as easy damage to the items being cleaned, untimely drying of the items being cleaned, and lack of cleaning fluid. At the same time, traditional cleaning methods require real-time control, which can easily lead to dangers during the cleaning process. It is time-consuming and laborious to identify whether the items being cleaned are dried after cleaning. It cannot achieve fully automatic cleaning, nor can it adjust the cleaning device in real time based on the cleaning status data. Summary of the invention

[0004] In view of the problems in the related art, the present invention provides an automatic adjustment method and system for a cleaning device based on data analysis to overcome the above-mentioned technical problems existing in the existing related art.

[0005] In order to solve the above technical problems, the present invention is achieved through the following technical solutions:

[0006] The present invention is a method for automatically adjusting a cleaning device based on data analysis, comprising the following steps:

[0007] S1. The cleaning materials are sent to the inside of the cleaning device by the conveyor belt, and the cleaning materials are weighed and counted, and the weight and quantity of the cleaning materials are checked to obtain the checked cleaning materials. The robotic arm locates the checked cleaning materials through the YOLOV5 neural network and grabs the cleaning materials to obtain the cleaning materials to be cleaned;

[0008] S2, cleaning the cleaning material to be cleaned, monitoring the amount of cleaning liquid, discharging the cleaning liquid and replenishing the cleaning liquid in real time when the cleaning liquid level does not meet the requirements, and alarming when the cleaning liquid level is abnormal, and obtaining the cleaning material being cleaned;

[0009] S3, using a PID algorithm to control the cleaning temperature in real time, adjusting parameters so that the cleaning material being cleaned is cleaned within a rated temperature threshold, and using a tuna school optimization algorithm to adjust the parameters in the PID algorithm to obtain the cleaned cleaning material;

[0010] S4, training the convolutional neural network to obtain a final convolutional neural network, drying the cleaned cleaning material to obtain an image of the dried cleaning material, and inputting the image of the dried cleaning material into the final convolutional neural network to identify a drying state.

[0011] The invention firstly weighs and counts the cleaning materials to check the weight and quantity of the cleaning materials, and after the coordinates of the images of the cleaning materials to be detected are marked and features are extracted, they are input into the YOLOV5 neural network for training to obtain the YOLOV5 detection model, and then the center point of the coordinates output by the YOLOV5 detection model is used as the grasping point of the robot arm, thereby improving the grasping accuracy and speed of the robot arm; secondly, a sensor is set in the cleaning liquid tank body to discharge or replenish the cleaning liquid according to the change of the cleaning liquid level, and to prompt an alarm for abnormal conditions; then a PID controller is used to adjust the temperature, and the proportional coefficient and the integral time coefficient are adjusted. And differential time coefficient make it a complete single closed-loop controller, and use tuna school optimization algorithm to optimize the proportional coefficient, integral time coefficient and differential time coefficient, and determine the parameters by iterating the position of tuna when preying to determine the best fitness function. The algorithm has fast convergence speed, good optimization performance and strong robustness. Finally, the drying image is filtered, labeled and data enhanced, and then input into the convolutional neural network for training to obtain the final convolutional neural network. The image of the washed material after drying is input into the final convolutional neural network to identify the drying state. Compared with other neural networks, it has stronger adaptability and higher recognition accuracy.

[0012] Preferably, the S1 comprises the following steps:

[0013] S11, the cleaning material is sent to the inside of the cleaning device through a conveyor belt, and the cleaning material passes through a weight detection device and a counting detection device. The amount of the cleaning material is set to a, and the weight of the cleaning material is set to b. The total weight of the cleaning material is calculated as follows:

[0014]

[0015] Wherein, A represents the total weight of the cleaning material, i1=1,2,3,...,a;

[0016] The actual total weight of the cleaning materials is calculated from the actual quantity of the cleaning materials and the actual weight of the cleaning materials, and the weight threshold is set to ω. When the absolute value of the difference between the actual total weight of the cleaning materials and the total weight of the cleaning materials is less than ω, the weight and quantity of the cleaning materials are accurate, and the verified cleaning materials are obtained; otherwise, the weight and quantity of the cleaning materials are inaccurate, and the weight and quantity of the cleaning materials are re-verified;

[0017] S12, set the rectangular positioning frame to A′, take the lower left corner of the rectangular positioning frame as the coordinate origin, obtain the coordinates of the upper left corner, lower left corner, upper right corner and lower right corner of the rectangular positioning frame as (a1, b1), (a2, b2), (a3, b3) and (a4, b4), respectively, and complete the coordinate labeling; set the activation function of the YOLOV5 neural network to the Sigmoid function, the loss function to f, select the cleaning material picture to be detected, and label the coordinates of the cleaning material picture to be detected to obtain the labeled cleaning material picture, and then divide the labeled cleaning material picture into a cleaning material training set and a cleaning material test set;

[0018] The cleaning material training set is input into the convolution layer, and a sliding window is set. The sliding window slides in the pixels of the labeled cleaning material pictures in the cleaning material training set with a step size of 1, and the features of the labeled cleaning material pictures in the cleaning material training set are extracted to obtain the cleaning material training set after feature extraction; the cleaning material training set after feature extraction is then input into the pooling layer, and a dimension reduction operation is performed to obtain a reduced dimension cleaning material training set, and iterates continuously, setting the current number of iterations to b′ and the maximum number of iterations to B. When the current number of iterations is equal to the maximum number of iterations, the iteration is stopped to obtain a trained YOLOV5 neural network; the cleaning material test set is input into the trained YOLOV5 neural network, and the accuracy error is set to B′ and the error threshold to ω. When the accuracy error is less than ω, the trained YOLOV5 neural network meets the accuracy requirements, and a YOLOV5 detection model is obtained; otherwise, the weight is adjusted until the accuracy error is less than ω;

[0019] S13, obtaining a verified cleaning material image from the verified cleaning material, inputting the verified cleaning material image into the YOLOV5 detection model after coordinate annotation, outputting the coordinates of the upper left corner, lower left corner, upper right corner and lower right corner of the cleaning material, selecting the coordinate center points of the upper left corner, lower left corner, upper right corner and lower right corner of the cleaning material as the robot arm grasping points, completing the cleaning material grasping, and obtaining the cleaning material to be cleaned.

[0020] This invention checks the weight and quantity of the cleaning materials, annotates the coordinates and extracts the features of the images of the cleaning materials to be detected, and then inputs them into the YOLOV5 neural network for training to obtain the YOLOV5 detection model. The center point of the coordinates output by the YOLOV5 detection model is then used as the robotic arm grasping point to achieve robotic arm grasping, thereby improving the robotic arm grasping accuracy and speed.

[0021] Preferably, S2 comprises the following steps:

[0022] S21, setting the cleaning liquid tank body to have a first liquid level sensor, a second liquid level sensor, a third liquid level sensor and a fourth liquid level sensor from top to bottom, the cleaning liquid level is initially at the position of the second liquid level sensor, the cleaning material to be cleaned is cleaned, when the cleaning liquid level drops to the position of the third liquid level sensor, the cleaning liquid is replenished until it reaches the position of the second liquid level sensor; when the cleaning liquid level rises to the position of the first liquid level sensor, the cleaning liquid is discharged until it reaches the position of the second liquid level sensor; when the cleaning liquid level drops below the position of the fourth liquid level sensor, the cleaning liquid level is in an abnormal state, and the cleaning liquid tank body prompts an alarm;

[0023] S22. Record the time interval for replenishing the cleaning liquid and the time interval for discharging the cleaning liquid, set the expected time interval for replenishing the cleaning liquid and the expected time interval for discharging the cleaning liquid, the replenishment error threshold is ξ1, and the discharge error threshold is ξ2. When the absolute value of the difference between the time interval for replenishing the cleaning liquid and the expected time interval for replenishing the cleaning liquid is greater than ξ1, or the absolute value of the difference between the time interval for discharging the cleaning liquid and the expected time interval for discharging the cleaning liquid is greater than ξ2, the cleaning liquid level is in an abnormal state, and the cleaning liquid tank prompts an alarm; otherwise, the cleaning liquid level is in a normal state.

[0024] Preferably, S3 comprises the following steps:

[0025] S31, set the actual cleaning temperature of the cleaning device at time t to c(t), set the rated cleaning temperature of the cleaning device at time t to c′(t), then the cleaning temperature deviation of the cleaning device at time t c″(t)=c(t)-c′(t), set the proportional coefficient to C1, the integral time coefficient to C2, and the differential time coefficient to C3, perform proportional, integral and differential calculations on the cleaning temperature deviation of the cleaning device, and obtain the output cleaning temperature c″′ at time t through the PID algorithm formula. The calculation formula is as follows:

[0026]

[0027] S32, repeat S31, set the first threshold value to ω1, when the error between the output cleaning temperature at time t and the rated cleaning temperature of the cleaning device at time t is less than ω1, the output cleaning temperature at time t is the temperature adjusted by the PID (proportional, integral and differential) controller; otherwise, use the tuna school optimization algorithm to adjust the proportional coefficient, integral time coefficient and differential time coefficient in the PID algorithm, so that the error between the output cleaning temperature at time t and the rated cleaning temperature of the cleaning device at time t is less than ω1, and the specific process is as follows:

[0028] S321, taking the error integral of the output cleaning temperature at time t as the fitness function Assume that the tuna population set is D = {d1, d2, d3, ..., d m}, where d m represents the mth tuna individual, α1 is a random number and α1∈[0,1], the upper bound of the tuna population in the search space is e1, and the lower bound of the tuna population in the search space is e2. The calculation formula for the initial position x1 of the tuna individual is as follows:

[0029] x1=e2+(e1-e2)·α1

[0030] Tuna in the tuna population forages in a spiral according to the optimal tuna individual, and iterates continuously to find the minimum fitness function value; set the current number of iterations to e, the maximum number of iterations to E, the optimal tuna individual position to x, and the tuna individual position at the e-1th iteration to x e-1 , the position of the tuna in the eth iteration is x e , the position of the tuna in the e+1th iteration is x e+1 , the optimal tuna individual movement trend coefficient is β1, the tuna individual movement trend coefficient is β2, the distance parameter between the tuna individual and the optimal tuna individual is χ, α2 is a random number and α2∈[0,1], then the tuna individual position x in the e+1th iteration e+1 The calculation formula is as follows,

[0031] x e+1 =(x+χ·|xx e )·β1+x e-1 β2

[0032]

[0033] S322, the tunas in the tuna population forage according to the parabola of the optimal tuna individual, and the optimal tuna individual updates its position. The optimal tuna individual position is the optimal proportional coefficient, the optimal integral time coefficient, and the optimal differential time coefficient. The adaptive parameters of the tuna population are set to δ, α3 is a random number and α3∈[0,1], α4 represents a random number of -1 or 1, then the tuna individual updates its position x in the e+1th iteratione ' +1 The calculation formula is as follows,

[0034] x e ' +1 =x+(xx e )·α3+(xx e )·α4·δ,

[0035] The individual tuna continuously updates its position to find the minimum fitness function value. When the current number of iterations is equal to the maximum number of iterations, the iteration is stopped to obtain the optimal individual tuna position. The three-dimensional coordinate values ​​corresponding to the optimal individual tuna position correspond to the optimal proportional coefficient, the optimal integral time coefficient and the optimal differential time coefficient respectively. The PID controller adjusts the temperature according to the cleaning temperature output at time t to obtain the cleaned cleaning material.

[0036] The invention uses a PID controller to adjust the temperature, adjusts the proportional coefficient, integral time coefficient and differential time coefficient to make the PID control a complete single closed-loop controller to achieve temperature control; and uses a tuna school optimization algorithm to optimize the proportional coefficient, integral time coefficient and differential time coefficient, and determines the best fitness function and thus the parameters by iterating the position of the tuna when hunting. The algorithm has fast convergence speed, good optimization performance and strong robustness.

[0037] Preferably, S4 comprises the following steps:

[0038] S41, selecting images of cleaning materials before drying and images of cleaning materials after drying as a set of cleaning material images to be processed, filtering, labeling and data enhancement processing the images of cleaning materials before drying and images of cleaning materials after drying in the set of cleaning material images to be processed respectively to obtain a set of processed cleaning material images, dividing the set of processed cleaning material images into a training set for cleaning material drying recognition and a test set for cleaning material drying recognition, setting the learning rate of the convolutional neural network to ε, the activation function to the ReLU function, inputting the training set for cleaning material drying recognition into the convolutional layer of the convolutional neural network, and outputting the result using the softmax function;

[0039] S42, setting the current number of iterations to g, the maximum number of iterations to G, the training error to g′, and the second threshold to ω2. When the current number of iterations in the convolutional neural network is equal to the maximum number of iterations or the training error is less than ω2, stop the iteration to obtain a trained convolutional neural network; otherwise, adjust the weight until the current number of iterations in the convolutional neural network is equal to the maximum number of iterations or the training error is less than ω2; then input the cleaning material drying recognition test set into the trained convolutional neural network, set the recognition accuracy error to g″, and the third threshold to ω3. When the recognition accuracy error is less than ω3, the trained convolutional neural network meets the requirements and obtains a final convolutional neural network; otherwise, repeat S42 until the recognition accuracy error is less than ω3; dry the cleaned cleaning material to obtain a dried cleaning material image, input the dried cleaning material image into the final convolutional neural network, identify the drying state, and obtain the cleaned cleaning material.

[0040] The invention obtains the final convolutional neural network by filtering, labeling and data enhancement processing on the drying image and then inputting it into the convolutional neural network for training. The image of the cleaned material after drying is input into the final convolutional neural network to identify the drying status. Compared with other neural networks, the convolutional neural network has stronger adaptability and higher recognition accuracy.

[0041] This embodiment also discloses an automatic adjustment system for a cleaning device based on data analysis, which specifically includes: a cleaning material preparation module, a cleaning liquid adjustment alarm module, a cleaning temperature control module, and a cleaning material drying identification module;

[0042] The cleaning material preparation module is used to check the weight and quantity of the cleaning materials and use a neural network to locate and grab the cleaning materials;

[0043] The cleaning fluid adjustment alarm module is used to discharge and replenish cleaning fluid in real time and to give an alarm when the cleaning fluid level is abnormal;

[0044] The cleaning temperature control module is used to use the PID algorithm to control the cleaning temperature in real time and use the tuna school optimization algorithm to adjust the parameters in the PID algorithm;

[0045] The cleaning material drying recognition module is used to use a convolutional neural network to recognize the image of the cleaning material after drying to obtain a drying state.

[0046] The present invention has the following beneficial effects:

[0047] 1. The invention first weighs and counts the cleaning materials to check the weight and quantity of the cleaning materials. After the coordinates of the cleaning material images to be detected are marked and features are extracted, they are input into the YOLOV5 neural network for training to obtain the YOLOV5 detection model. The center point of the coordinates output by the YOLOV5 detection model is used as the robotic arm grasping point, thereby improving the robotic arm grasping accuracy and speed.

[0048] 2. The invention sets a sensor in the cleaning liquid tank, discharges or replenishes the cleaning liquid according to the change of the cleaning liquid level, and gives an alarm for abnormal conditions; then uses a PID controller to adjust the temperature, and uses a tuna school optimization algorithm to optimize the proportional coefficient, integral time coefficient and differential time coefficient. The optimal fitness function is determined by iterating the position of the tuna when hunting, so as to determine the parameters. The algorithm has fast convergence speed, good optimization performance and strong robustness.

[0049] 3. The invention obtains the final convolutional neural network by filtering, labeling and data enhancement processing on the drying image and then inputting it into the convolutional neural network for training. The image of the cleaned material after drying is input into the final convolutional neural network to identify the drying state. Compared with other neural networks, it has stronger adaptability and higher recognition accuracy.

[0050] Of course, any product implementing the present invention does not necessarily need to achieve all of the advantages described above at the same time. BRIEF DESCRIPTION OF THE DRAWINGS

[0051] In order to more clearly illustrate the technical solutions of the embodiments of the invention, the drawings required for describing the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the invention. For ordinary technicians in this field, they can also obtain drawings based on these drawings without paying creative work.

[0052] Figure 1 The present invention provides a flow chart of automatic adjustment of a cleaning device by an automatic adjustment system for a cleaning device based on data analysis. DETAILED DESCRIPTION

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

[0054] In the description of the present invention, it is necessary to understand that the terms "opening", "upper", "lower", "top", "middle", "inside" and the like indicating orientation or positional relationship are only for the convenience of describing the invention and simplifying the description, and do not indicate or imply that the components or elements referred to must have a specific orientation, be constructed and operate in a specific orientation, and therefore cannot be understood as limiting the invention.

[0055] This embodiment discloses a method for automatic adjustment of a cleaning device based on data analysis, which specifically includes the following contents:

[0056] S1. The cleaning materials are sent to the inside of the cleaning device by the conveyor belt, and the cleaning materials are weighed and counted, and the weight and quantity of the cleaning materials are checked to obtain the checked cleaning materials. The robotic arm locates the checked cleaning materials through the YOLOV5 neural network and grabs the cleaning materials to obtain the cleaning materials to be cleaned;

[0057] The S1 comprises the following steps:

[0058] S11, the cleaning material is sent to the inside of the cleaning device through a conveyor belt, and the cleaning material passes through a weight detection device and a counting detection device. The amount of the cleaning material is set to a, and the weight of the cleaning material is set to b. The total weight of the cleaning material is calculated as follows:

[0059]

[0060] Wherein, A represents the total weight of the cleaning material, i1=1,2,3,...,a;

[0061] The actual total weight of the cleaning materials is calculated from the actual quantity of the cleaning materials and the actual weight of the cleaning materials, and the weight threshold is set to ω. When the absolute value of the difference between the actual total weight of the cleaning materials and the total weight of the cleaning materials is less than ω, the weight and quantity of the cleaning materials are accurate, and the verified cleaning materials are obtained; otherwise, the weight and quantity of the cleaning materials are inaccurate, and the weight and quantity of the cleaning materials are re-verified;

[0062] S12, set the rectangular positioning frame to A′, take the lower left corner of the rectangular positioning frame as the coordinate origin, obtain the coordinates of the upper left corner, lower left corner, upper right corner and lower right corner of the rectangular positioning frame as (a1, b1), (a2, b2), (a3, b3) and (a4, b4), respectively, and complete the coordinate labeling; set the activation function of the YOLOV5 neural network to the Sigmoid function, the loss function to f, select the cleaning material picture to be detected, and label the coordinates of the cleaning material picture to be detected to obtain the labeled cleaning material picture, and then divide the labeled cleaning material picture into a cleaning material training set and a cleaning material test set;

[0063] The cleaning material training set is input into the convolution layer, and a sliding window is set. The sliding window slides in the pixels of the labeled cleaning material pictures in the cleaning material training set with a step size of 1, and the features of the labeled cleaning material pictures in the cleaning material training set are extracted to obtain the cleaning material training set after feature extraction; the cleaning material training set after feature extraction is then input into the pooling layer, and a dimension reduction operation is performed to obtain a reduced dimension cleaning material training set, and iterates continuously, setting the current number of iterations to b′ and the maximum number of iterations to B. When the current number of iterations is equal to the maximum number of iterations, the iteration is stopped to obtain a trained YOLOV5 neural network; the cleaning material test set is input into the trained YOLOV5 neural network, and the accuracy error is set to B′ and the error threshold to ω. When the accuracy error is less than ω, the trained YOLOV5 neural network meets the accuracy requirements, and a YOLOV5 detection model is obtained; otherwise, the weight is adjusted until the accuracy error is less than ω;

[0064] S13, obtaining a verified cleaning material picture from the verified cleaning material, inputting the verified cleaning material picture into the YOLOV5 detection model after coordinate annotation, outputting the coordinates of the upper left corner, lower left corner, upper right corner and lower right corner of the cleaning material, selecting the coordinate center points of the upper left corner, lower left corner, upper right corner and lower right corner of the cleaning material as the grasping points of the robot arm, completing the grasping of the cleaning material, and obtaining the cleaning material to be cleaned;

[0065] S2, cleaning the cleaning material to be cleaned, monitoring the amount of cleaning liquid, discharging the cleaning liquid and replenishing the cleaning liquid in real time when the cleaning liquid level does not meet the requirements, and alarming when the cleaning liquid level is abnormal, and obtaining the cleaning material being cleaned;

[0066] The S2 comprises the following steps:

[0067] S21, setting the cleaning liquid tank body to have a first liquid level sensor, a second liquid level sensor, a third liquid level sensor and a fourth liquid level sensor from top to bottom, the cleaning liquid level is initially at the position of the second liquid level sensor, the cleaning material to be cleaned is cleaned, when the cleaning liquid level drops to the position of the third liquid level sensor, the cleaning liquid is replenished until it reaches the position of the second liquid level sensor; when the cleaning liquid level rises to the position of the first liquid level sensor, the cleaning liquid is discharged until it reaches the position of the second liquid level sensor; when the cleaning liquid level drops below the position of the fourth liquid level sensor, the cleaning liquid level is in an abnormal state, and the cleaning liquid tank body prompts an alarm;

[0068] S22, recording the time interval for replenishing the cleaning liquid and the time interval for discharging the cleaning liquid, setting the expected time interval for replenishing the cleaning liquid and the expected time interval for discharging the cleaning liquid, the replenishment error threshold is ξ1, and the discharge error threshold is ξ2. When the absolute value of the difference between the time interval for replenishing the cleaning liquid and the expected time interval for replenishing the cleaning liquid is greater than ξ1, or the absolute value of the difference between the time interval for discharging the cleaning liquid and the expected time interval for discharging the cleaning liquid is greater than ξ2, the cleaning liquid level is in an abnormal state, and the cleaning liquid tank prompts an alarm; otherwise, the cleaning liquid level is in a normal state;

[0069] S3, using a PID algorithm to control the cleaning temperature in real time, adjusting parameters so that the cleaning material being cleaned is cleaned within a rated temperature threshold, and using a tuna school optimization algorithm to adjust the parameters in the PID algorithm to obtain the cleaned cleaning material;

[0070] The S3 comprises the following steps:

[0071] S31, set the actual cleaning temperature of the cleaning device at time t to c(t), set the rated cleaning temperature of the cleaning device at time t to c′(t), then the cleaning temperature deviation of the cleaning device at time t c″(t)=c(t)-c′(t), set the proportional coefficient to C1, the integral time coefficient to C2, and the differential time coefficient to C3, perform proportional, integral and differential calculations on the cleaning temperature deviation of the cleaning device, and obtain the output cleaning temperature c″′ at time t through the PID algorithm formula. The calculation formula is as follows:

[0072]

[0073] S32, repeat S31, set the first threshold value to ω1, when the error between the output cleaning temperature at time t and the rated cleaning temperature of the cleaning device at time t is less than ω1, the output cleaning temperature at time t is the PID controller adjustment temperature; otherwise, use the tuna school optimization algorithm to adjust the proportional coefficient, integral time coefficient and differential time coefficient in the PID algorithm, so that the error between the output cleaning temperature at time t and the rated cleaning temperature of the cleaning device at time t is less than ω1, the specific process is as follows:

[0074] S321, taking the error integral of the output cleaning temperature at time t as the fitness function Assume that the tuna population set is D = {d1, d2, d3, ..., d m}, where d m represents the mth tuna individual, α1 is a random number and α1∈[0,1], the upper bound of the tuna population in the search space is e1, and the lower bound of the tuna population in the search space is e2. The calculation formula for the initial position x1 of the tuna individual is as follows:

[0075] x1=e2+(e1-e2)·α1

[0076] Tuna in the tuna population forages in a spiral according to the optimal tuna individual, and iterates continuously to find the minimum fitness function value; set the current number of iterations to e, the maximum number of iterations to E, the optimal tuna individual position to x, and the tuna individual position at the e-1th iteration to x e-1 , the position of the tuna in the eth iteration is x e , the position of the tuna in the e+1th iteration is x e+1 , the optimal tuna individual movement trend coefficient is β1, the tuna individual movement trend coefficient is β2, the distance parameter between the tuna individual and the optimal tuna individual is χ, α2 is a random number and α2∈[0,1], then the tuna individual position x in the e+1th iteration e+1 The calculation formula is as follows,

[0077] x e+1 =(x+χ·|xx e )·β1+x e-1 β2

[0078]

[0079] S322, the tunas in the tuna population forage according to the parabola of the optimal tuna individual, and the optimal tuna individual updates its position. The optimal tuna individual position is the optimal proportional coefficient, the optimal integral time coefficient, and the optimal differential time coefficient. The adaptive parameters of the tuna population are set to δ, α3 is a random number and α3∈[0,1], α4 represents a random number of -1 or 1, then the tuna individual updates its position x in the e+1th iteration e ' +1 The calculation formula is as follows,

[0080] x e ' +1 =x+(xx e )·α3+(xx e )·α4·δ,

[0081] The individual tuna continuously updates its position to find the minimum fitness function value. When the current number of iterations is equal to the maximum number of iterations, the iteration is stopped to obtain the optimal individual tuna position. The three-dimensional coordinate values ​​corresponding to the optimal individual tuna position correspond to the optimal proportional coefficient, the optimal integral time coefficient and the optimal differential time coefficient respectively. The PID controller adjusts the temperature according to the cleaning temperature output at time t to obtain the cleaned cleaning material.

[0082] S4, training the convolutional neural network to obtain a final convolutional neural network, drying the cleaned cleaning material to obtain an image of the dried cleaning material, and inputting the image of the dried cleaning material into the final convolutional neural network to identify the drying state;

[0083] The S4 comprises the following steps:

[0084] S41, selecting images of cleaning materials before drying and images of cleaning materials after drying as a set of cleaning material images to be processed, filtering, labeling and data enhancement processing the images of cleaning materials before drying and images of cleaning materials after drying in the set of cleaning material images to be processed respectively to obtain a set of processed cleaning material images, dividing the set of processed cleaning material images into a training set for cleaning material drying recognition and a test set for cleaning material drying recognition, setting the learning rate of the convolutional neural network to ε, the activation function to the ReLU function, inputting the training set for cleaning material drying recognition into the convolutional layer of the convolutional neural network, and outputting the result using the softmax function;

[0085] S42, setting the current number of iterations to g, the maximum number of iterations to G, the training error to g′, and the second threshold to ω2. When the current number of iterations in the convolutional neural network is equal to the maximum number of iterations or the training error is less than ω2, stop the iteration to obtain a trained convolutional neural network; otherwise, adjust the weight until the current number of iterations in the convolutional neural network is equal to the maximum number of iterations or the training error is less than ω2; then input the cleaning material drying recognition test set into the trained convolutional neural network, set the recognition accuracy error to g″, and the third threshold to ω3. When the recognition accuracy error is less than ω3, the trained convolutional neural network meets the requirements and obtains a final convolutional neural network; otherwise, repeat S42 until the recognition accuracy error is less than ω3; dry the cleaned cleaning material to obtain a dried cleaning material image, input the dried cleaning material image into the final convolutional neural network, identify the drying state, and obtain the cleaned cleaning material.

[0086] This embodiment also discloses an automatic adjustment system for a cleaning device based on data analysis, which specifically includes: a cleaning material preparation module, a cleaning liquid adjustment alarm module, a cleaning temperature control module, and a cleaning material drying identification module;

[0087] The cleaning material preparation module is used to check the weight and quantity of the cleaning materials and use a neural network to locate and grab the cleaning materials;

[0088] The cleaning fluid adjustment alarm module is used to discharge and replenish cleaning fluid in real time and to give an alarm when the cleaning fluid level is abnormal;

[0089] The cleaning temperature control module is used to use the PID algorithm to control the cleaning temperature in real time and use the tuna school optimization algorithm to adjust the parameters in the PID algorithm;

[0090] The cleaning material drying recognition module is used to use a convolutional neural network to recognize the image of the cleaning material after drying to obtain a drying state.

[0091] In the description of this specification, the description with reference to the terms "one embodiment", "example", "specific example", etc. means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the invention. In this specification, the schematic representation of the above terms does not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described can be combined in any one or more embodiments or examples in a suitable manner.

[0092] The preferred embodiments of the invention disclosed above are only used to help explain the invention. The preferred embodiments do not describe all the details in detail, nor do they limit the invention to the specific implementation methods described. Obviously, many modifications and changes can be made according to the content of this specification. This specification selects and specifically describes these embodiments in order to better explain the principles and practical applications of the invention, so that those skilled in the art can understand and use the invention well.

Claims

1. A method for automatic adjustment of a cleaning device based on data analysis, characterized in that: The steps include: S1. The cleaning materials are sent to the inside of the cleaning device by the conveyor belt, and the cleaning materials are weighed and counted, and the weight and quantity of the cleaning materials are checked to obtain the checked cleaning materials. The robotic arm locates the checked cleaning materials through the YOLOV5 neural network and grabs the cleaning materials to obtain the cleaning materials to be cleaned; S2, cleaning the cleaning material to be cleaned, monitoring the amount of cleaning liquid, discharging the cleaning liquid and replenishing the cleaning liquid in real time when the cleaning liquid level does not meet the requirements, and alarming when the cleaning liquid level is abnormal, and obtaining the cleaning material being cleaned; S3, using a PID algorithm to control the cleaning temperature in real time, adjusting parameters so that the cleaning material being cleaned is cleaned within a rated temperature threshold, and using a tuna school optimization algorithm to adjust the parameters in the PID algorithm to obtain the cleaned cleaning material; S4, training the convolutional neural network to obtain a final convolutional neural network, drying the cleaned cleaning material to obtain an image of the dried cleaning material, and inputting the image of the dried cleaning material into the final convolutional neural network to identify a drying state.

2. The method for automatic adjustment of a cleaning device based on data analysis according to claim 1, characterized in that: The S1 comprises the following steps: S11, sending the cleaning material to the inside of the cleaning device through a conveyor belt, and the cleaning material passes through a weight detection device and a counting detection device to check the weight and quantity of the cleaning material to obtain the checked cleaning material; S12, setting the activation function of the YOLOV5 neural network to the Sigmoid function and the loss function to f, selecting a picture of a cleaning material to be detected, marking the coordinates of the picture of the cleaning material to be detected, obtaining a marked picture of the cleaning material, and then dividing the marked picture of the cleaning material into a cleaning material training set and a cleaning material test set; The cleaning material training set is input into the convolutional layer, and it is continuously iterated. The current number of iterations is set to b′ and the maximum number of iterations is set to B. When the current number of iterations is equal to the maximum number of iterations, the iteration is stopped to obtain a trained YOLOV5 neural network. The cleaning material test set is input into the trained YOLOV5 neural network, and the precision error is set to B′ and the error threshold is ω. When the precision error is less than ω, the trained YOLOV5 neural network meets the precision requirements and a YOLOV5 detection model is obtained. Otherwise, the weight is adjusted until the precision error is less than ω.

3. The method for automatic adjustment of a cleaning device based on data analysis according to claim 1, characterized in that: The checked cleaning material obtains a checked cleaning material picture, and the checked cleaning material picture is input into the YOLOV5 detection model after coordinate annotation, and the coordinates of the upper left corner, lower left corner, upper right corner and lower right corner of the cleaning material are output, and the coordinate center points of the upper left corner, lower left corner, upper right corner and lower right corner of the cleaning material are selected as the robot arm grasping points, and the cleaning material grasping is completed to obtain the cleaning material to be cleaned.

4. The method for automatic adjustment of a cleaning device based on data analysis according to claim 2, characterized in that: The S2 comprises the following steps: S21. A sensor is set in the cleaning liquid tank to monitor the amount of cleaning liquid, discharge the cleaning liquid and replenish the cleaning liquid in real time; when the cleaning liquid level is in an abnormal state, the cleaning liquid tank prompts an alarm.

5. The method for automatic adjustment of a cleaning device based on data analysis according to claim 2, characterized in that: Record the time interval for replenishing the cleaning liquid and the time interval for discharging the cleaning liquid, set the expected time interval for replenishing the cleaning liquid and the expected time interval for discharging the cleaning liquid, the replenishment error threshold is ξ1, and the discharge error threshold is ξ2. When the absolute value of the difference between the time interval for replenishing the cleaning liquid and the expected time interval for replenishing the cleaning liquid is greater than ξ1, or the absolute value of the difference between the time interval for discharging the cleaning liquid and the expected time interval for discharging the cleaning liquid is greater than ξ2, the cleaning fluid level is in an abnormal state, and the cleaning liquid tank prompts an alarm; otherwise, the cleaning fluid level is in a normal state.

6. The method for automatic adjustment of a cleaning device based on data analysis according to claim 3, characterized in that: The S3 comprises the following steps: S31. Obtain the output cleaning temperature at time t through the PID algorithm formula, set the first threshold value to ω1, and when the error between the output cleaning temperature at time t and the rated cleaning temperature of the cleaning device at time t is less than ω1, the output cleaning temperature at time t is the temperature adjusted by the PID controller; otherwise, use the tuna school optimization algorithm to adjust the proportional coefficient, integral time coefficient and differential time coefficient in the PID algorithm, so that the error between the output cleaning temperature at time t and the rated cleaning temperature of the cleaning device at time t is less than ω1; the PID controller adjusts the temperature according to the output cleaning temperature at time t to obtain the cleaning material after cleaning.

7. The method for automatic adjustment of a cleaning device based on data analysis according to claim 3, characterized in that: The S31 comprises the following steps: S311, taking the error integral of the output cleaning temperature at time t as the fitness function Assume that the tuna population set is D = {d1, d2, d3, ..., d m }, where d m represents the mth tuna individual, α1 is a random number and α1∈[0,1], the upper bound of the tuna population in the search space is e1, and the lower bound of the tuna population in the search space is e2. The calculation formula for the initial position x1 of the tuna individual is as follows: x1=e2+(e1-e2)·α1 Tuna in the tuna population forages in a spiral according to the optimal tuna individual, and iterates continuously to find the minimum fitness function value; set the current number of iterations to e, the maximum number of iterations to E, the optimal tuna individual position to x, and the tuna individual position at the e-1th iteration to x e-1 , the position of the tuna in the eth iteration is x e , the position of the tuna in the e+1th iteration is x e+1 , the optimal tuna individual movement trend coefficient is β1, the tuna individual movement trend coefficient is β2, the distance parameter between the tuna individual and the optimal tuna individual is χ, α2 is a random number and α2∈[0,1], then the tuna individual position x in the e+1th iteration e+1 The calculation formula is as follows, x e+1 =(x+χ·|xx e )·β1+x e-1 ·β2 S312, tuna in the tuna population forages according to the parabola of the optimal tuna individual, and the optimal tuna individual updates its position. The optimal tuna individual position is the optimal proportional coefficient, the optimal integral time coefficient, and the optimal differential time coefficient. The adaptive parameter of the tuna population is set to δ, α3 is a random number and α3∈[0,1], α4 represents a random number of -1 or 1, then the tuna individual updates its position x in the e+1th iteration e ' +1 The calculation formula is as follows, The individual tuna continuously updates its position to find the minimum fitness function value. When the current number of iterations is equal to the maximum number of iterations, the iteration is stopped to obtain the optimal individual tuna position. The three-dimensional coordinate values ​​corresponding to the optimal individual tuna position correspond to the optimal proportional coefficient, the optimal integral time coefficient and the optimal differential time coefficient respectively.

8. The method for automatic adjustment of a cleaning device based on data analysis according to claim 4, characterized in that: The S4 comprises the following steps: S41, selecting images of cleaning materials before drying and images of cleaning materials after drying as a set of cleaning material images to be processed, filtering, labeling and data enhancement processing the images of cleaning materials before drying and images of cleaning materials after drying in the set of cleaning material images to be processed respectively to obtain a set of processed cleaning material images, dividing the set of processed cleaning material images into a training set for cleaning material drying recognition and a test set for cleaning material drying recognition, setting the learning rate of the convolutional neural network to ε, the activation function to the ReLU function, inputting the training set for cleaning material drying recognition into the convolutional layer of the convolutional neural network, and outputting the result using the softmax function; S42, setting the current number of iterations to g, the maximum number of iterations to G, the training error to g′, and the second threshold to ω2. When the current number of iterations in the convolutional neural network is equal to the maximum number of iterations or the training error is less than ω2, stop the iteration to obtain a trained convolutional neural network; otherwise, adjust the weight until the current number of iterations in the convolutional neural network is equal to the maximum number of iterations or the training error is less than ω2; then input the cleaning material drying recognition test set into the trained convolutional neural network, set the recognition accuracy error to g″, and the third threshold to ω3. When the recognition accuracy error is less than ω3, the trained convolutional neural network meets the requirements and obtains a final convolutional neural network; otherwise, repeat S42 until the recognition accuracy error is less than ω3; dry the cleaned cleaning material to obtain a dried cleaning material image, input the dried cleaning material image into the final convolutional neural network, identify the drying state, and obtain the cleaned cleaning material.

9. An automatic adjustment system for a cleaning device based on data analysis according to any one of claims 1 to 8, characterized in that: Specifically include: Cleaning material preparation module, cleaning liquid adjustment alarm module, cleaning temperature control module and cleaning material drying identification module; The cleaning material preparation module is used to check the weight and quantity of the cleaning materials and use a neural network to locate and grab the cleaning materials; The cleaning fluid adjustment alarm module is used to discharge and replenish cleaning fluid in real time and to give an alarm when the cleaning fluid level is abnormal; The cleaning temperature control module is used to use the PID algorithm to control the cleaning temperature in real time and use the tuna school optimization algorithm to adjust the parameters in the PID algorithm; The cleaning material drying recognition module is used to use a convolutional neural network to recognize the image of the cleaning material after drying to obtain a drying state.

Citation Information

Patent Citations

  • Segment editing method, device, equipment and storage medium combining RPA and AI

    CN112232039B