A processing method and system for optimizing paper ejection parameters of a paper sheet separating apparatus

Abstract

The embodiment of the application relates to a kind of processing method and system for optimizing paper outlet parameter of paper distribution equipment, the method comprises: constructing paper outlet parameter optimization model M, constructing model data set for each equipment model, training paper outlet parameter optimization model M;And all model of type is deployed to remote server;Intelligent paper distribution equipment regularly sends recent cached temperature and humidity sequence to remote server;Remote server inputs current model of type coding vector and temperature and humidity sequence and obtains paper outlet parameter to intelligent paper distribution equipment and sends back;Intelligent paper distribution equipment is based on the paper outlet parameter sent by remote server and updates the paper outlet parameter stored in equipment side.The application can improve the environmental adaptation flexibility of equipment, reduce equipment failure rate.

Description

Technical Field

[0001] This invention relates to the fields of artificial intelligence and the Internet of Things, and in particular to a method and system for optimizing the paper output parameters of a paper separating device. Background Technology

[0002] In public restrooms, automatic paper dispensing devices (such as automatic toilet paper dispensers) incorporating infrared sensing technology are already quite common. Traditional automatic paper dispensing devices only have the ability to automatically dispense paper; they lack network connectivity and data collection capabilities (such as ambient temperature and humidity). With the development and application of IoT technology, some intelligent paper dispensing devices with network connectivity and data collection capabilities have emerged in recent years. These intelligent paper dispensing devices simultaneously possess capabilities such as automatic paper dispensing, remote network connectivity, and local data collection (such as ambient temperature and humidity).

[0003] However, research revealed that both traditional and new intelligent paper dispensing equipment use a set of factory-fixed parameters related to automatic paper dispensing (such as the shortest interval between two dispensing operations Δt, the motor speed v for a single dispensing operation, and the dispensing time T). These parameters cannot be corrected or dynamically optimized throughout the equipment's lifespan. Even newer equipment that can collect ambient temperature and humidity data only uses it for local display or remote monitoring, without applying this data to real-time adjustment of dispensing parameters. In actual use, the temperature and humidity of the toilet stall environment vary significantly with seasons, weather, and pedestrian traffic: in high-temperature and high-humidity environments, paper easily becomes damp and softens, reducing its tensile strength, and the fixed parameter mode easily leads to a high rate of paper jams or breaks; in low-temperature and dry environments, paper becomes brittle, causing similar problems. Therefore, there is an urgent need for a technical solution that can dynamically optimize the dispensing parameters (Δt, v, T) based on the temperature and humidity sequence of the toilet stall environment to improve the equipment's environmental adaptability and reduce its overall failure rate. Summary of the Invention

[0004] The purpose of this invention is to address the shortcomings of existing technologies by providing a method and system for optimizing the paper output parameters of paper-splitting equipment. This invention first constructs a paper output parameter optimization model M for intelligent paper-splitting equipment based on artificial intelligence technology. Then, through temperature and humidity zone control experiments, it constructs corresponding model datasets for various equipment models of intelligent paper-splitting equipment. Based on the model datasets of each model, it trains the paper output parameter optimization model M to obtain the corresponding model model, and deploys all model models to a remote server. Next, each intelligent paper-splitting device continuously samples and caches the ambient temperature and humidity of its respective toilet stall at the front end, and periodically sends its model code vector and the most recently cached temperature and humidity sequence to the remote server. The remote server, in the background, predicts the paper output parameters based on the corresponding model model for each device and promptly feeds back the prediction results to the front-end devices. This invention can improve the environmental adaptability of the equipment and reduce the equipment failure rate.

[0005] To achieve the above objectives, a first aspect of the present invention provides a method for optimizing the paper output parameters of a paper separating device, the method comprising: A paper output parameter optimization model M is constructed for the intelligent paper separating equipment; and corresponding model datasets are constructed for various equipment models of the intelligent paper separating equipment through temperature and humidity zoning control experiments; the paper output parameter optimization model M is trained based on the model datasets of each model to obtain the corresponding model model; and all model models are deployed to a remote server; wherein, the paper output parameter optimization model is used to optimize the paper output parameters according to the model input model encoding vector X and temperature and humidity sequence Z and output the corresponding paper output parameters Y; the model encoding vector X is the unique thermal encoding vector of the equipment model of the intelligent paper separating equipment, which is composed of N X Model code x j Composition, 1≤indexj≤N X N X X represents the total number of device models of the intelligent paper-splitting equipment, and there is one and only one x. j The value is 1 for all others and 0 for the rest; the temperature and humidity sequence Z is the environmental temperature and humidity sampling sequence of the toilet stall where the intelligent paper dispensing device is located, consisting of N Z Each sampled data z i Composition, the sampled data z i Including temperature a i Humidity b i 1 ≤ index i ≤ N Z N Z The preset total number of sampling points, z = every two sampling data points i z i+1 The time intervals between them are equal, which is the preset sampling interval duration; the paper output parameter Y is the optimized paper output parameter of the intelligent paper separation device, including the shortest interval duration Δt between two paper output operations, the motor speed v of a single paper output, and the paper output duration T; each of the model datasets consists of multiple first data records; the first data record includes the model encoding vector X, the temperature and humidity sequence Z, and the label parameter Y. tag The label parameter Y tag Including the tag interval duration △t tag Tag rotation speed v tag Label output time T tag The remote server is connected to each of the intelligent paper-splitting devices, and each intelligent paper-splitting device corresponds to a type of device model and is installed in a designated toilet stall. Each of the aforementioned intelligent paper-splitting devices continuously samples and caches the ambient temperature and humidity of the toilet stall where the device is located according to the sampling interval; and periodically reports the most recently cached N data according to a preset data reporting frequency. ZThe temperature and humidity sampling data are extracted to form the corresponding temperature and humidity sequence Z, and the corresponding model encoding vector X is set based on the current device model. The current model encoding vector X and the temperature and humidity sequence Z are then sent to the remote server. When the remote server receives the model code vector X and the temperature and humidity sequence Z sent by any of the intelligent paper-splitting devices, it first selects the corresponding model as the current model based on the current model code vector X; then it inputs the current model code vector X and the temperature and humidity sequence Z into the current model for processing to obtain the corresponding paper output parameter Y; and then it sends the current paper output parameter Y back to the current intelligent paper-splitting device. When each of the intelligent paper-splitting devices receives the paper output parameter Y sent by the remote server, it updates the paper output parameter Y stored on the device side based on the current paper output parameter Y. loc Wherein, the paper output parameter Y loc Including the shortest interval duration △t loc Motor speed v loc Paper output time T loc .

[0006] Preferably, the intelligent paper-splitting device includes at least a mobile communication module, an ambient temperature and humidity sensor, an infrared proximity sensor, an automatic paper dispensing module, a storage module, and a main control module; the main control module is connected to the mobile communication module, the ambient temperature and humidity sensor, the infrared proximity sensor, the automatic paper dispensing module, and the storage module, respectively. The mobile communication module includes 4G and 5G communication units; the mobile communication module is used to handle the communication interaction process between the remote server and the main control module. The ambient temperature and humidity sensor is used to periodically collect data on the ambient temperature and humidity of the toilet stall where the current device is located, according to the sensor's collection frequency, and generate corresponding temperature and humidity sampling data to send to the main control module. The infrared proximity sensor is used to send an object proximity signal to the main control module when it detects an object approaching the current sensor. The automatic paper output module is used to, upon receiving a paper output command from the main control module, operate at the motor speed v specified in the current command. loc and the paper output time T loc The system determines the motor speed and rotation duration for each automatic paper feeding operation, driving the paper feeding motor to rotate at a constant speed to complete the operation. Feedback on this operation is then sent to the main control module. This feedback includes both success and failure states. If a paper jam or paper breakage occurs during paper feeding, the automatic paper feeding module will generate a failure feedback. The single paper feeding length of the automatic paper feeding module is calculated as K × v.loc T loc K is the preset conversion coefficient; The storage module is used to store the previous paper output time and the paper output parameter Y. loc A data buffer queue; the data buffer queue is used to buffer the temperature and humidity sampling data of the environmental temperature and humidity sensor according to the first-in-first-out (FIFO) buffering principle, and the queue length of the data buffer queue is greater than or equal to N. Z ; The main control module is used to store the temperature and humidity sampling data sent by the ambient temperature and humidity sensor into the data cache queue; The main control module is also used to periodically update the data cache queue with the most recently cached N data according to the data reporting frequency. Z The temperature and humidity sampling data are extracted to form the corresponding temperature and humidity sequence Z, and the corresponding model code vector X is set based on the current device model. The current model code vector X and the temperature and humidity sequence Z are sent to the remote server through the mobile communication module. The main control module is also used to read the corresponding previous paper output time and the shortest interval duration Δt from the storage module when it receives the object approach signal sent by the infrared proximity sensor. loc The motor speed v loc The paper output time T loc And whether the interval between the current time and the previous paper output time is greater than or equal to the shortest interval length Δt. loc Perform identification; if so, then carry the motor speed v loc and the paper output time T loc The paper output command is sent to the automatic paper output module; if not, a delay is performed for a preset waiting time, and at the end of this wait, the interval between the latest current time and the previous paper output time is checked to see if it is greater than or equal to the shortest interval Δt. loc Perform a second identification; if the result of this identification is yes, then the motor speed v will be carried. loc and the paper output time T loc The paper output command is sent to the automatic paper output module. If the recognition result is negative, the system will wait again for the specified waiting time until the interval between the latest current time and the previous paper output time is greater than or equal to the shortest interval Δt. loc until; The main control module is also used to identify whether the current paper output operation feedback is successful when it receives the paper output operation feedback sent back by the automatic paper output module; if so, it resets the previous paper output time stored in the storage module based on the current time. The main control module is further configured to, upon receiving the paper output parameter Y sent by the remote server via the mobile communication module, adjust the paper output parameter Y stored in the storage module based on the current paper output parameter Y. loc Reset.

[0007] Preferably, the first model input terminal of the paper output parameter optimization model M is used to receive the model code vector X, the second model input terminal is used to receive the temperature and humidity sequence Z, and the model output terminal is used to output the paper output parameter Y; The paper output parameter optimization model M includes a feature mapping layer, a preprocessing layer, a feature encoder, a feature pooling layer, a feature fusion layer, a fused feature prediction layer, and a prediction output layer. The input of the feature mapping layer is connected to the input of the first model, and its output is connected to the first input of the feature fusion layer; the input of the preprocessing layer is connected to the input of the second model, and its output is connected to the input of the feature encoder; the output of the feature encoder is connected to the input of the feature pooling layer; the output of the feature pooling layer is connected to the second input of the feature fusion layer; the output of the feature fusion layer is connected to the input of the fused feature prediction layer; the output of the fused feature prediction layer is connected to the second input of the prediction output layer; the first input of the prediction output layer is connected to the input of the first model, and the output of the prediction output layer is connected to the model output. The feature mapping layer is used to perform linear mapping processing on the model encoding vector X to obtain the corresponding feature vector H1, which is then sent to the feature fusion layer. The linear mapping method of the feature vector H1 is as follows: ; W X The weight matrix of the feature mapping layer has a shape of D1×N. X D1 is the preset first feature dimension; the shape of the model encoding vector X is 1×N. X The shape of the feature vector H1 is 1×D1. The pretreatment layer is used to determine the global minimum temperature A. min Global maximum temperature A max Global minimum humidity B min Global maximum humidity B max The temperature and humidity sequence Z is normalized to obtain the corresponding normalized sequence Z. * The normalized sequence Z is encoded according to the embedding encoding method of the Transformer model encoder. * Embedding encoding is performed to obtain the corresponding embedding vector E. Z Send to the feature encoder; Wherein, the normalized sequence Z * Including N Z Normalized data The normalized data Including normalized temperature Normalized humidity The normalized data With the sampled data z i One-to-one correspondence, the normalized temperature With the temperature a i One-to-one correspondence, the normalized humidity With the humidity b i One-to-one correspondence; , ; The embedding vector E Z The shape is N Z ×D2, where D2 is the preset second feature dimension; The feature encoder is implemented based on the encoder structure of the Transformer model; the feature encoder is used to process the embedding vector E. Z Feature encoding is performed to obtain the corresponding feature vector H. Z Send to the feature pooling layer; Wherein, the feature vector H Z The shape is N Z ×D2; The feature pooling layer is used to process the feature vector H. Z Each feature channel is subjected to average pooling to obtain the corresponding feature vector H2, which is then sent to the feature fusion layer. The shape of the feature vector H2 is 1×D2; The feature fusion layer is used to perform feature concatenation on feature vectors H1 and H2 to obtain the corresponding feature vector H3, which is then sent to the fused feature prediction layer. The shape of the feature vector H3 is 1×(D1+D2); The fusion feature prediction layer is implemented based on the MLP model; the fusion feature prediction layer is used to perform normalized vector prediction processing on the feature vector H3 to obtain the corresponding prediction vector H4 and send it to the prediction output layer; The prediction method for the prediction vector H4 is as follows: ; W1 and W2 are the first and second weight matrices of the fused feature prediction layer, and b1 and b2 are the first and second bias vectors of the fused feature prediction layer; the shape of W1 is D3×(D1+D2), where D3 is a preset third feature dimension, and the shape of W2 is 3×D3; the shape of b1 is 1×D3, and the shape of b2 is 1×3; ReLU() is the ReLU activation function, and sigmoid() is the Sigmoid activation function; activation vectors The shape of the prediction vector H4 is 1×D3; the shape of the prediction vector H4 is 1×3, consisting of three corresponding vector data. , , composition; The prediction output layer is used to convert the model code x that is 1 in the model code vector X. j Let index j be the index of the current model. * ; and based on the prediction vector H4 and the current model index j * Corresponding global minimum interval duration Global maximum interval duration Global minimum speed Global maximum speed Minimum global output time Maximum global paper output time The three types of paper output parameters are inversely normalized to obtain the corresponding paper output parameter Y and output it. The inverse normalization method for the three types of paper output parameters Δt, v, and T of the paper output parameter Y is as follows: , , .

[0008] Preferably, the step of constructing corresponding model datasets for various equipment models of the intelligent paper-splitting device through temperature and humidity zone control experiments specifically includes: Step 41, set the global temperature range [A] min A max Divide into N equal parts A Individual temperature ranges [a] min,u ,a max,u ], and set the global humidity range [B min B max Divide into N equal parts B Individual humidity ranges [b] min,s ,b max,s ], and for N A Individual temperature ranges and N B The corresponding N is obtained by combining the individual humidity ranges with the temperature and humidity ranges.A ×N B Temperature and humidity combination range P u,s [(a min,u ,a max,u ),(b min,s ,b max,s The duration of the temperature and humidity sequence Z is taken as the corresponding experimental duration T. test =N Z × Sampling interval duration; and according to the preset N R Temperature and humidity control modes, for each of the aforementioned temperature and humidity combination ranges P u,s Set the corresponding N R Group temperature and humidity adjustment curve R u,s,r ; Where 1 ≤ index u ≤ N A 1 ≤ index s ≤ N B 1 ≤ index r ≤ N R N A The total number of preset temperature ranges, N B The total number of preset humidity ranges, N R This represents the total number of preset adjustment modes; a min,u a max,u Let be the minimum and maximum temperatures of the u-th sub-temperature range; b min,s b max,s Let be the minimum and maximum humidity of the s-th sub-humidity interval; The N R The temperature and humidity regulation modes include at least the following modes: constant humidity and linear temperature increase mode, constant humidity and linear temperature decrease mode, constant humidity and step temperature increase mode, constant humidity and step temperature decrease mode, constant humidity and temperature sinusoidal fluctuation mode, constant temperature and humidity and linear temperature increase mode, constant temperature and humidity and linear temperature decrease mode, constant temperature and humidity and step temperature increase mode, constant temperature and humidity and step temperature decrease mode, constant temperature and humidity and sinusoidal fluctuation mode, temperature and humidity increase linearly with the same slope mode, temperature and humidity increase linearly with different slope modes, temperature and humidity decrease linearly with the same slope mode, temperature and humidity decrease linearly with different slope modes, temperature increases linearly but humidity decreases linearly, temperature decreases linearly but humidity increases linearly, temperature and humidity fluctuate based on the same phase sinusoidal fluctuation mode, temperature and humidity increase in the same direction step mode, and temperature and humidity decrease in the same direction step mode. The temperature and humidity control curves R for each group u,s,r Each curve consists of a set of corresponding temperature and humidity control curves, and the duration of both curves is equal to the experimental duration T. test The trends of the two curves match the corresponding temperature and humidity regulation modes. Step 42: Select each model of the intelligent paper-splitting device as the current model; and select one test sample for the current model as the current sample. Step 43, for the current prototype in each of the temperature and humidity combination ranges P u,s Target paper output length Configure settings; Step 44, set the global interval duration range [△t] of the current prototype. min ,△t max Divide into N equal parts △t Sample Δt q and the global speed range [v] of the current prototype min ,v max Divide into N equal parts v Sample v g and the global paper output time interval [T] of the current prototype. min ,T max Divide into N equal parts T Sample T o ; and for the N corresponding to the current prototype △t The sampling Δt q N v The sample v g N T The sample T o By combining them, we get N. △t ×N v ×N T Group sampling parameters Y q,g,o (△t q ,v g ,T o ); Where 1 ≤ index q ≤ N △t 1 ≤ index g ≤ N v 1 ≤ index o ≤ N T N △t The preset interval duration, the total number of sampling points, N v The preset total number of speed sampling points, N T The total number of sampling points for the preset paper output time; Step 45, convert the temperature and humidity adjustment curves R of each group into... u,s,r As the current environment curve; and each of the sampling parameters Y is used as the current environment curve; q,g,o Using the current paper output parameters as the current parameters, and in a temperature and humidity adjustable experimental environment, with the current environmental curve and the current paper output parameters as the current experimental conditions, the current prototype is subjected to a test run for a duration of T. test A set of corresponding temperature and humidity sequences Z was obtained from the continuous paper output experiment. u,s,r,q,g,o Experimental score C u,s,r,q,g,o; and the N corresponding to the current environment curve △t ×N v ×N T The experimental score C mentioned above u,s,r,q,g,o The sampling parameter Y corresponding to the maximum score in the data. q,g,o As the label parameter corresponding to the current environment curve ; Step 46: Use the index j corresponding to the current device model in the model encoding vector X as the current model index j. * And set a length of N X The all-zero vector is taken as the current vector, and the j-th element in the current vector is... * Each vector data is reset to 1, and the current vector after reset is used as the model code vector X corresponding to the current device model; and the temperature and humidity sequences Z of the current prototype are used. u,s,r,q,g,o and the corresponding tag parameters The model code vector X forms a corresponding first data record; and the obtained N A ×N B ×N R ×N △t ×N v ×N T The first data record is deduplicated; and the model dataset corresponding to the current device model is composed of all the remaining first data records after deduplication.

[0009] Furthermore, in an experimental environment with adjustable temperature and humidity, using the current environmental curve and the current paper output parameters as the current experimental conditions, the current prototype undergoes a test run for a duration of T. test A set of corresponding temperature and humidity sequences Z was obtained from the continuous paper output experiment. u,s,r,q,g,o Experimental score C u,s,r,q,g,o Specifically, it includes: Step 51: Before starting this round of continuous paper output experiment, adjust the paper output parameter Y on the current prototype based on the current paper output parameters. loc Perform a reset; and based on the sampled v of the current paper output parameters. g The sampling T o And the corresponding experimental paper output length L is calculated based on the conversion coefficient K of the current prototype. u,s,r,q,g,o ; and install a continuous paper output program run by the main control module on the current prototype; Among them, L u,s,r,q,g,o =K×v g T o ; The continuous paper output program is used to, after the program runs, first store an initialized first failure time parameter in the storage module of the current prototype; then, use the shortest interval duration Δt stored in the storage module. loc The system sends continuous paper output commands to the automatic paper output module of the current prototype at continuous command intervals; it also identifies whether each paper output operation feedback returned by the automatic paper output module is a failure, and resets the first failure time parameter based on the current time when the first specific failure paper output operation feedback is received. Step 52: At the start of this round of continuous paper output experiment, drive the main control module of the current prototype to run the continuous paper output program; Step 53: During the continuous paper output experiment, the temperature and humidity of the current experimental environment are continuously modulated based on the current environmental curve; and the temperature and humidity of the current experimental environment are continuously sampled using the sampling interval as the continuous sampling interval, and the sampled temperature and humidity sequence is used as the corresponding temperature and humidity sequence Z. u,s,r,q,g,o ; Step 54: At the end of this round of continuous paper output experiment, drive the main control module to shut down the continuous paper output program; read the corresponding first failure time parameter from the storage module; and record the end time of this round of experiment. Step 55: Identify whether the first fault time parameter is empty; if so, set the corresponding fault duration T. err If the value is 0, then the interval between the first failure time parameter and the experiment end time is calculated, and the calculation result is used as the corresponding failure duration T. err Based on the fault duration T err The experimental duration T test Calculate the corresponding failure duration ratio f u,s,r,q,g,o f u,s,r,q,g,o =T err / T test Based on the fault duration ratio f u,s,r,q,g,o The experimental paper output length L u,s,r,q,g,o And the target paper output length corresponding to the current prototype and the current environment curve. Calculate the corresponding experimental score C u,s,r,q,g,o ; Wherein, the fault duration ratio f u,s,r,q,g,o The value of is between 0 and 1; The experimental score C u,s,r,q,g,o The calculation method is as follows: ; α and β are preset weighting coefficients, where α > β > 0.

[0010] Preferably, the step of training the paper output parameter optimization model M based on the model dataset of each model to obtain the corresponding model model specifically includes: Step 61: Take each device model of the intelligent paper-splitting device as the current device model; take the model dataset corresponding to the current device model as the current model dataset; and initialize the paper output parameter optimization model M based on a set of preset initial model parameters to obtain the corresponding current base model; Step 62: Based on a preset first segmentation ratio, randomly divide the current model dataset into two subsets, denoted as the first training set and the first evaluation set; and calculate the total number of records N in the first training set. tr The total number of records N in the first evaluation set av Statistical analysis was conducted separately; Both the first training set and the first evaluation set consist of multiple first data records; the ratio N of the total number of records in the first training set to the total number of records in the first evaluation set is... tr :N av The first segmentation ratio is satisfied; Step 63: Take each of the first data records in the first training set as the current record; input the model code vector X and the temperature and humidity sequence Z of the current record into the current base model for processing, and record the paper output parameter Y of the current model as the corresponding prediction parameter. ; and the currently recorded tag parameter Y tag Record as the corresponding tag parameter ; and by the prediction parameters and the label parameters Form a corresponding first prediction-label pair; and obtain N tr The first prediction-label pair is substituted into the model loss function L. M The corresponding first loss value is obtained through calculation; Where 1 ≤ index k ≤ N tr ; The prediction parameters Including the shortest interval duration Motor speed Paper output time ; The tag parameters Including tag interval duration Tag rotation speed Label output time ; The model loss function L M for: ; Step 64: Identify whether the first loss value meets the preset first loss value range; if it does, proceed to step 65; if not, based on the preset model optimizer, move towards making the model loss function L... M The model parameters of the current base model are optimized in one round in the direction of reaching the minimum value, and the process returns to step 63 after this round of optimization is completed; The model optimizer includes the Adam optimizer and the SGD optimizer; Step 65: Take each of the first data records in the first evaluation set as the current record; input the model code vector X and the temperature and humidity sequence Z of the current record into the current base model for processing, and record the paper output parameter Y of the current model as the corresponding prediction parameter. ; and the currently recorded tag parameter Y tag Record as the corresponding tag parameter ; and by the prediction parameters and the label parameters Form a corresponding second prediction-label pair; and obtain N av The second prediction-label pair is substituted into the model evaluation function L. A The corresponding first evaluation value is obtained through calculation; Where 1 ≤ index l ≤ N av ; The prediction parameters Including the shortest interval duration Motor speed Paper output time ; The tag parameters Including tag interval duration Tag rotation speed Label output time ; The model evaluation function L A for: ; Step 66: Identify whether the first evaluation value meets the preset first evaluation value range; if not, return to step 62; if yes, stop training and use the current base model as the model corresponding to the current device model.

[0011] A second aspect of the present invention provides a system for implementing the processing method for optimizing the paper output parameters of a paper-splitting device provided in the first aspect above, the system comprising: a model preparation system, a remote server, and multiple intelligent paper-splitting devices; The remote server is connected to the model preparation system and each of the intelligent paper-splitting devices; each intelligent paper-splitting device corresponds to a type of device model and is installed in a designated toilet stall. The model preparation system is used to construct a paper output parameter optimization model M for the intelligent paper separating device; and to construct corresponding model datasets for various models of the intelligent paper separating device through temperature and humidity zoning control experiments; and to train the paper output parameter optimization model M based on the model datasets of each model to obtain the corresponding model model; and to deploy all the model models to the remote server; wherein, the paper output parameter optimization model is used to optimize the paper output parameters according to the model input model encoding vector X and temperature and humidity sequence Z and output the corresponding paper output parameters Y; the model encoding vector X is the unique thermal encoding vector of the intelligent paper separating device model, which is composed of N X Model code x j Composition, 1≤indexj≤N X N X X represents the total number of device models of the intelligent paper-splitting equipment, and there is one and only one x. j The value is 1 for all others and 0 for the rest; the temperature and humidity sequence Z is the environmental temperature and humidity sampling sequence of the toilet stall where the intelligent paper dispensing device is located, consisting of N Z Each sampled data z i Composition, the sampled data z i Including temperature a i Humidity b i 1 ≤ index i ≤ N Z N Z The preset total number of sampling points, z = every two sampling data points i z i+1 The time intervals between them are equal, which is the preset sampling interval duration; the paper output parameter Y is the optimized paper output parameter of the intelligent paper separation device, including the shortest interval duration Δt between two paper output operations, the motor speed v of a single paper output, and the paper output duration T; each of the model datasets consists of multiple first data records; the first data record includes the model encoding vector X, the temperature and humidity sequence Z, and the label parameter Y. tag The label parameter Y tag Including the tag interval duration △t tag Tag rotation speed v tag Label output time T tag ; Each of the aforementioned intelligent paper-splitting devices is used to continuously sample and cache the ambient temperature and humidity of the toilet stall where the device is located according to the sampling interval; and periodically upload the most recently cached N data according to a preset data reporting frequency. ZThe temperature and humidity sampling data are extracted to form the corresponding temperature and humidity sequence Z, and the corresponding model encoding vector X is set based on the current device model. The current model encoding vector X and the temperature and humidity sequence Z are then sent to the remote server. Each of the aforementioned intelligent paper-splitting devices is further configured to, upon receiving the paper output parameter Y sent by the remote server, update the paper output parameter Y stored on the device side based on the current paper output parameter Y. loc Wherein, the paper output parameter Y loc Including the shortest interval duration △t loc Motor speed v loc Paper output time T loc ; When the remote server receives the model code vector X and the temperature and humidity sequence Z sent by any of the intelligent paper-splitting devices, it first selects the corresponding model as the current model based on the current model code vector X; then it inputs the current model code vector X and the temperature and humidity sequence Z into the current model for processing to obtain the corresponding paper output parameter Y; and then it sends the current paper output parameter Y back to the current intelligent paper-splitting device.

[0012] This invention provides a method and system for optimizing paper output parameters in a paper-splitting device. First, an optimization model M for paper output parameters is constructed for the intelligent paper-splitting device based on artificial intelligence technology. Then, through temperature and humidity zone control experiments, corresponding model datasets are built for various models of the intelligent paper-splitting device. Based on these model datasets, the optimization model M is trained to obtain the corresponding model, and all model models are deployed to a remote server. Next, each intelligent paper-splitting device continuously samples and caches the ambient temperature and humidity of its respective toilet stall at the front end, and periodically sends its model code vector and the most recently cached temperature and humidity sequence to the remote server. The remote server then predicts the paper output parameters based on the corresponding model model for each device in the background and promptly feeds back the prediction results to the front-end device. This invention improves the environmental adaptability of the equipment and reduces the equipment failure rate. Attached Figure Description

[0013] Figure 1 This is a schematic diagram of a method for optimizing the paper output parameters of a paper separating device according to Embodiment 1 of the present invention; Figure 2 This is a schematic diagram of the paper output parameter optimization model provided in Embodiment 1 of the present invention; Figure 3 This is a schematic diagram of a processing system for optimizing the paper output parameters of a paper separating device, provided in Embodiment 2 of the present invention. Detailed Implementation

[0014] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings. Obviously, the described embodiments are merely some embodiments of this invention, and not all embodiments. Based on the embodiments of this invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this invention.

[0015] The method for optimizing the paper output parameters of a paper separating device provided in Embodiment 1 of the present invention, such as... Figure 1 The schematic diagram shows a method for optimizing the paper output parameters of a paper separating device according to Embodiment 1 of the present invention. The method includes the following steps: Step 1: Construct a paper output parameter optimization model M for the intelligent paper separating equipment; construct corresponding model datasets for various equipment models of the intelligent paper separating equipment through temperature and humidity zoning control experiments; train the paper output parameter optimization model M based on the model datasets of each model to obtain the corresponding model model; and deploy all model models to a remote server.

[0016] Here, in this embodiment of the invention, the remote server is connected to each intelligent paper-splitting device, and each intelligent paper-splitting device corresponds to a type of device model and is installed in a designated toilet stall.

[0017] The paper output parameter optimization model of this invention is used to optimize the paper output parameters based on the model input model code vector X and temperature and humidity sequence Z, and output the corresponding paper output parameters Y.

[0018] Wherein, the model coding vector X is the unique thermal coding vector of the intelligent paper-splitting equipment model, which is composed of N X Model code x j Composition, 1≤indexj≤N X N X X represents the total number of intelligent paper-splitting equipment models, and X contains exactly one x. j The value is 1, and the rest are 0.

[0019] The temperature and humidity sequence Z is the environmental temperature and humidity sampling sequence of the toilet stall where the intelligent paper dispensing device is located, consisting of N. Z Each sampled data z i Composition, sampled data z i Including temperature a i Humidity b i 1 ≤ index i ≤ N Z N Z The preset total number of sampling points, z = every two sampling data points i z i+1 The time intervals between them are equal and are the preset sampling interval duration.

[0020] The paper output parameter Y is the optimized paper output parameter of the intelligent paper separation device, including the shortest interval Δt between two paper output operations, the motor speed v of a single paper output, and the paper output time T.

[0021] Each model dataset in this embodiment of the invention consists of multiple first data records. Each first data record includes a model encoding vector X, a temperature and humidity sequence Z, and label parameters Y. tag ; Label parameter Y tag Including the tag interval duration △t tag Tag rotation speed v tag Label output time T tag .

[0022] like Figure 2 As shown in the schematic diagram of the paper output parameter optimization model provided in Embodiment 1 of the present invention, the model components of the paper output parameter optimization model M include: a feature mapping layer, a preprocessing layer, a feature encoder, a feature pooling layer, a feature fusion layer, a fused feature prediction layer, and a prediction output layer. The first model input terminal of the paper output parameter optimization model M is used to receive the model encoding vector X, the second model input terminal is used to receive the temperature and humidity sequence Z, and the model output terminal is used to output the paper output parameters Y.

[0023] like Figure 2 As shown, the connection relationships of the model components in the paper output parameter optimization model M are as follows: the input of the feature mapping layer is connected to the input of the first model, and its output is connected to the first input of the feature fusion layer; the input of the preprocessing layer is connected to the input of the second model, and its output is connected to the input of the feature encoder; the output of the feature encoder is connected to the input of the feature pooling layer; the output of the feature pooling layer is connected to the second input of the feature fusion layer; the output of the feature fusion layer is connected to the input of the fused feature prediction layer; the output of the fused feature prediction layer is connected to the second input of the prediction output layer; the first input of the prediction output layer is connected to the input of the first model, and the output of the prediction output layer is connected to the model output.

[0024] The model component functions of the paper output parameter optimization model M are shown below.

[0025] 1) Feature mapping layer: In this embodiment of the invention, the feature mapping layer is used to perform linear mapping processing on the model encoding vector X to obtain the corresponding feature vector H1, which is then sent to the feature fusion layer.

[0026] Here, the linear mapping method of feature vector H1 in this embodiment of the invention is as follows: ; Among them, W X The weight matrix of the feature mapping layer has a shape of D1×N. X D1 is the preset first feature dimension; the shape of the model code vector X is 1×N.X The shape of the eigenvector H1 is 1×D1.

[0027] 2) Preprocessing layer: The preprocessing layer in this embodiment of the invention is used to determine the global minimum temperature A. min Global maximum temperature A max Global minimum humidity B min Global maximum humidity B max Normalizing the temperature and humidity sequence Z yields the corresponding normalized sequence Z. * ; and the normalized sequence Z is encoded according to the embedding encoding method of the Transformer model encoder. * Embedding encoding is performed to obtain the corresponding embedding vector E. Z Send to the feature encoder.

[0028] Here, the normalized sequence Z in this embodiment of the invention * Including N Z Normalized data Normalized data Including normalized temperature Normalized humidity Normalized data With sampled data z i One-to-one correspondence, normalized temperature With temperature a i One-to-one correspondence, normalized humidity With humidity b i One-to-one correspondence.

[0029] Normalized temperature ,humidity The calculation method is as follows: , .

[0030] Embedded vector E Z The shape is N Z ×D2, where D2 is the preset second feature dimension.

[0031] 3) Feature encoder: The feature encoder in this embodiment of the invention is implemented based on the encoder structure of the Transformer model.

[0032] The feature encoder of this embodiment is used to process the embedding vector E Z Feature encoding is performed to obtain the corresponding feature vector H. Z Send to the feature pooling layer.

[0033] Here, the feature vector H in the embodiment of the present invention Z The shape is NZ ×D2.

[0034] 4) Feature pooling layer: The feature pooling layer in this embodiment of the invention is used to process the feature vector H Z The feature vectors H2 are obtained by average pooling each feature channel and then sent to the feature fusion layer.

[0035] Here, the shape of the feature vector H2 in this embodiment of the invention is 1×D2.

[0036] 5) Feature fusion layer: In this embodiment of the invention, the feature fusion layer is used to perform feature concatenation processing on feature vectors H1 and H2 to obtain the corresponding feature vector H3, which is then sent to the fusion feature prediction layer.

[0037] Here, the shape of the feature vector H3 in this embodiment of the invention is 1×(D1+D2).

[0038] 6) Feature fusion prediction layer: The fusion feature prediction layer in this embodiment of the invention is implemented based on an MLP model.

[0039] In this embodiment of the invention, the fusion feature prediction layer is used to perform normalized vector prediction processing based on feature vector H3 to obtain the corresponding prediction vector H4, which is then sent to the prediction output layer.

[0040] Here, the prediction method for the prediction vector H4 in this embodiment of the invention is as follows: .

[0041] W1 and W2 are the first and second weight matrices of the fused feature prediction layer, and b1 and b2 are the first and second bias vectors of the fused feature prediction layer. W1 has a shape of D3×(D1+D2), where D3 is the preset third feature dimension. W2 has a shape of 3×D3. b1 has a shape of 1×D3, and b2 has a shape of 1×3. ReLU() is the ReLU activation function, and sigmoid() is the Sigmoid activation function. Activation vectors The shape of the prediction vector H4 is 1×D3. The shape of the prediction vector H4 is 1×3, consisting of three corresponding vector data. , , composition.

[0042] 7) Predicting the output layer: The prediction output layer is used to convert the model codes x that are 1 in the model code vector X into model codes x. j Let index j be the index of the current model. * ; and based on the prediction vector H4 and the current model index j * Corresponding global minimum interval duration Global maximum interval duration Global minimum speed Global maximum speed Minimum global output time Maximum global paper output time The three types of paper output parameters are inversely normalized to obtain the corresponding paper output parameters Y and then output.

[0043] Here, the inverse normalization method for the three types of paper output parameters Δt, v, and T of the paper output parameter Y in this embodiment of the invention is as follows: , , .

[0044] Step 1 specifically includes: Step 11: Construct an optimized paper output parameter model M for the intelligent paper separating equipment.

[0045] Here, the paper output parameter optimization model M is as shown above.

[0046] Step 12: Construct corresponding model datasets for various models of intelligent paper-splitting equipment through temperature and humidity zone control experiments.

[0047] Specifically, it includes: Step 121, set the global temperature range [A] min A max Divide into N equal parts A Individual temperature ranges [a] min,u ,a max,u ], and set the global humidity range [B min B max Divide into N equal parts B Individual humidity ranges [b] min,s ,b max,s ], and for N A Individual temperature ranges and N B The corresponding N is obtained by combining the individual humidity ranges with the temperature and humidity ranges. A ×N B Temperature and humidity combination range P u,s [(a min,u ,a max,u ),(b min,s ,b max,s The duration of the temperature and humidity sequence Z is taken as the corresponding experimental duration T. test =N Z × Sampling interval duration; and according to the preset N R Temperature and humidity control modes, for each temperature and humidity combination range P u,s Set the corresponding NR Group temperature and humidity adjustment curve R u,s,r .

[0048] Where 1 ≤ index u ≤ N A 1 ≤ index s ≤ N B 1 ≤ index r ≤ N R N A The total number of preset temperature ranges, N B The total number of preset humidity ranges, N R This represents the total number of preset adjustment modes.

[0049] a min,u a max,u Let be the minimum and maximum temperatures of the u-th sub-temperature range.

[0050] b min,s b max,s Let be the minimum and maximum humidity of the s-th sub-humidity range.

[0051] N R Temperature and humidity regulation modes include at least the following: constant humidity and linear temperature increase mode, constant humidity and linear temperature decrease mode, constant humidity and step temperature increase mode, constant humidity and step temperature decrease mode, constant humidity and temperature sinusoidal fluctuation mode, constant temperature and linear humidity increase mode, constant temperature and linear humidity decrease mode, constant temperature and step humidity increase mode, constant temperature and step humidity decrease mode, constant temperature and linear humidity fluctuation mode, temperature and humidity increase linearly with the same slope mode, temperature and humidity increase linearly with different slope modes, temperature and humidity decrease linearly with the same slope mode, temperature and humidity decrease linearly with different slope modes, temperature increases linearly but humidity decreases linearly, temperature decreases linearly but humidity increases linearly, temperature and humidity fluctuate based on in-phase sinusoidal fluctuation mode, temperature and humidity increase in the same direction step mode, and temperature and humidity decrease in the same direction step mode.

[0052] Each set of temperature and humidity control curves R u,s,r Each curve consists of a set of corresponding temperature and humidity control curves, and the duration of both curves is equal to the experimental duration T. test The trends of the two curves match the corresponding temperature and humidity control modes.

[0053] Step 122: Select each model of the intelligent paper-splitting equipment as the current model; and select one test sample for the current model as the current sample.

[0054] Step 123, for the current prototype in various temperature and humidity combination ranges P u,s Target paper output length Configure the settings.

[0055] Here, in this embodiment of the invention, various temperature and humidity combination ranges P are set.u,s Target paper output length First, a baseline temperature and humidity range P is set based on historical application experience. base and for P base Set the corresponding reference length L base and for L base Set the corresponding sliding range; then, based on historical statistical data and / or historical application experience, adjust each P... u,s Compared to P base The analysis focuses on the types of faults that increase the paper output failure rate (such as paper jams and paper breaks), and is based on the analysis results and L. base The sliding range is used to formulate corresponding paper output length adjustment strategies (e.g., increasing by 5%, increasing by 10%, shortening by 5%, shortening by 10%, etc.), and then based on the formulated paper output length adjustment strategy and L... base To set each P u,s corresponding .

[0056] Step 124, set the global interval duration range [△t] of the current prototype. min ,△t max Divide into N equal parts △t Sample Δt q and the global speed range of the current prototype [v min ,v max Divide into N equal parts v Sample v g and set the global paper output time range [T] of the current prototype. min ,T max Divide into N equal parts T Sample T o ; and for the N corresponding to the current prototype △t Sample Δt q N v Sample v g N T Sample T o By combining them, we get N. △t ×N v ×N T Group sampling parameters Y q,g,o (△t q ,v g ,T o ).

[0057] Where 1 ≤ index q ≤ N △t 1 ≤ index g ≤ N v 1 ≤ index o ≤ N T N △t The preset interval duration, the total number of sampling points, N v The preset total number of speed sampling points, NT This represents the total number of sampling points for the preset paper output time.

[0058] Step 125, convert the temperature and humidity adjustment curves R of each group to... u,s,r This is used as the current environment curve; and each sampling parameter Y is used as... q,g,o Using the current paper output parameters as the current parameters, and in a temperature and humidity adjustable experimental environment, with the current environmental curve and current paper output parameters as the current experimental conditions, the current prototype is subjected to a test run for a duration of T. test A set of corresponding temperature and humidity sequences Z was obtained from the continuous paper output experiment. u,s,r,q,g,o Experimental score C u,s,r,q,g,o ; and the N corresponding to the current environmental curve △t ×N v ×N T Each experiment score C u,s,r,q,g,o The sampling parameter Y corresponding to the maximum score in q,g,o As the label parameter corresponding to the current environment curve .

[0059] In the experimental environment with adjustable temperature and humidity, using the current environmental curve and current paper output parameters as the current experimental conditions, the current prototype was subjected to a test run for a duration of T. test A set of corresponding temperature and humidity sequences Z was obtained from the continuous paper output experiment. u,s,r,q,g,o Experimental score C u,s,r,q,g,o Specifically, it includes: Step A1: Before starting this round of continuous paper output experiment, adjust the paper output parameter Y on the current prototype based on the current paper output parameters. loc Perform a reset; and sample v based on the current paper output parameters. g Sampling T o And the corresponding experimental paper output length L is calculated based on the conversion factor K of the current prototype. u,s,r,q,g,o And install a continuous paper output program run by the main control module on the current prototype.

[0060] Here, the conversion relationship between the paper output length L, motor speed v, and paper output time T of the intelligent paper separating device in this embodiment of the invention is: L = K × vT; where the conversion coefficient K is related to the diameter of the paper roller of the paper output module of the device, and under normal circumstances K = πD, where D is the diameter of the paper roller. Because there are multiple models of intelligent paper separating devices, the parameters of the components (such as the paper roller diameter D) of each model may have individual differences, so each model of device corresponds to a specified conversion coefficient K.

[0061] The experimental paper output length L in this embodiment of the invention u,s,r,q,g,o The calculation method is as follows: L u,s,r,q,g,o =K×v g T o .

[0062] The continuous paper output program of this embodiment of the invention is used to, after the program runs, first store an initialized first failure time parameter in the storage module of the current prototype; then, based on the shortest interval duration Δt stored in the storage module... loc The system sends continuous paper output commands to the automatic paper output module of the current prototype at continuous command intervals; it also identifies whether each paper output operation feedback returned by the automatic paper output module is a failure, and resets the first failure time parameter based on the current time when the first specific failure paper output operation feedback is received.

[0063] Step A2: At the start of this round of continuous paper output experiment, drive the main control module of the current prototype to run the continuous paper output program.

[0064] Step A3: During this round of continuous paper output experiment, the temperature and humidity of the current experimental environment are continuously modulated based on the current environmental curve; and the temperature and humidity of the current experimental environment are continuously sampled at the sampling interval, with the sampling interval duration as the continuous sampling interval, and the sampled temperature and humidity sequence is used as the corresponding temperature and humidity sequence Z. u,s,r,q,g,o .

[0065] Step A4: At the end of this round of continuous paper output experiment, drive the main control module to shut down the continuous paper output program; read the corresponding first failure time parameter from the storage module; and record the end time of this round of experiment.

[0066] Step A5: Check if the first fault time parameter is empty; if so, set the corresponding fault duration T. err If the value is 0, then the interval between the first failure time parameter and the experiment end time is calculated, and the calculation result is used as the corresponding failure duration T. err And based on the fault duration T err Experiment duration T test Calculate the corresponding failure duration ratio f u,s,r,q,g,o f u,s,r,q,g,o =T err / T test And based on the proportion of fault duration f u,s,r,q,g,o Experimental paper output length L u,s,r,q,g,o And the target paper output length corresponding to the current prototype and the current environmental curve. Calculate the corresponding experimental score C u,s,r,q,g,o .

[0067] Here, the fault duration ratio f in the embodiment of the present invention u,s,r,q,g,o The value of is between 0 and 1.

[0068] Experimental score C of this invention embodiment u,s,r,q,g,o The calculation method is as follows: ; Where α and β are preset weighting coefficients, α > β > 0.

[0069] Step 126: Use the index j corresponding to the current device model in the model coding vector X as the current model index j. * And set a length of N X The zero vector is taken as the current vector, and the j-th element in the current vector is... * Each vector data point is reset to 1, and the current vector after reset is used as the model code vector X corresponding to the current device model; and the current temperature and humidity sequences Z of the current prototype are used. u,s,r,q,g,o and its corresponding tag parameters The model code vector X forms a corresponding first data record; and the obtained N A ×N B ×N R ×N △t ×N v ×N T The first data record is deduplicated; and the model dataset corresponding to the current device model is composed of all the remaining first data records after deduplication.

[0070] Here, under theoretical conditions, each temperature and humidity regulation curve R u,s,r The corresponding N △t ×N v ×N T Z u,s,r,q,g,o They are the same, that is, each temperature and humidity control curve R u,s,r Only one test sample (i.e., the first data record) can be generated; however, in actual experiments, due to a series of practical factors such as experimental operation and environmental sensors, each temperature and humidity adjustment curve R... u,s,r The corresponding N △t ×N v ×N T Z u,s,r,q,g,o Most, if not all, are not exactly the same. This is similar to adding a large number of perturbation sequences in an experiment, thereby making each temperature and humidity regulation curve R... u,s,r A maximum of N can be obtained. △t ×N v ×N T The first test sample (i.e., the first data record).

[0071] Step 13: Train the paper parameter optimization model M based on the model datasets for each model to obtain the corresponding model model. Specifically, this includes: Step 131: Take each device model of the intelligent paper-splitting device as the current device model; take the model dataset corresponding to the current device model as the current model dataset; and initialize the paper output parameter optimization model M based on a set of preset initial model parameters to obtain the corresponding current base model model.

[0072] Step 132: Based on a preset first segmentation ratio, randomly divide the current model dataset into two subsets, denoted as the first training set and the first evaluation set; and assign each subset N to a number of records in the first training set. tr The total number of records N in the first evaluation set av Statistical analysis was conducted separately.

[0073] Here, the first segmentation ratio in this embodiment of the invention is a preset ratio parameter, such as 8:2; both the first training set and the first evaluation set consist of multiple first data records; the ratio N of the total number of records in the first training set to the total number of records in the first evaluation set is... tr :N av The first segmentation ratio is satisfied.

[0074] Step 133: Take each of the first data records in the first training set as the current record; input the model code vector X and the temperature and humidity sequence Z of the current record into the current base model for processing, and record the paper output parameter Y of the current model as the corresponding prediction parameter. ; and the currently recorded label parameter Y tag Record as the corresponding tag parameter ; and by prediction parameters and label parameters Form a corresponding first prediction-label pair; and obtain N tr The first prediction-label pair is substituted into the model loss function L. M The corresponding first loss value is obtained through calculation.

[0075] Where 1 ≤ index k ≤ N tr .

[0076] Prediction parameters Including the shortest interval duration Motor speed Paper output time .

[0077] Tag parameters Including tag interval duration Tag rotation speed Label output time .

[0078] The model loss function L in this embodiment of the invention M for: .

[0079] Step 134: Identify whether the first loss value meets the preset first loss value range; if it does, proceed to step 135; if not, based on the preset model optimizer, move towards making the model loss function L... M The direction that reaches the minimum value is used to perform a round of optimization on the model parameters of the current base model, and after this round of optimization is completed, return to step 133.

[0080] Here, the first loss value range in this embodiment of the invention is a pre-set numerical range. The model optimizer in this embodiment of the invention includes the Adam optimizer and the SGD optimizer.

[0081] Step 135: Take each of the first data records in the first evaluation set as the current record; input the model code vector X and temperature and humidity sequence Z of the current record into the current base model for processing, and record the paper output parameter Y of the current model as the corresponding prediction parameter. ; and the currently recorded label parameter Y tag Record as the corresponding tag parameter ; and by prediction parameters and label parameters Form a corresponding second prediction-label pair; and obtain N av The second prediction-label pair is substituted into the model evaluation function L. A The corresponding first evaluation value is obtained through calculation.

[0082] Where 1 ≤ index l ≤ N av .

[0083] Prediction parameters Including the shortest interval duration Motor speed Paper output time .

[0084] Tag parameters Including tag interval duration Tag rotation speed Label output time .

[0085] The model evaluation function L in this embodiment of the invention A for: .

[0086] Step 136: Identify whether the first evaluation value meets the preset first evaluation value range; if not, return to step 132; if it does, stop training and use the current base model as the model corresponding to the current device model.

[0087] Here, the first evaluation value range of this embodiment of the invention is a pre-set numerical range.

[0088] Step 14, and deploy all model types to the remote server.

[0089] Step 2: Each intelligent paper-dispensing device continuously samples and caches the ambient temperature and humidity of the toilet stall where the device is located at sampling intervals; and periodically reports the most recently cached N data according to a preset data reporting frequency. Z The temperature and humidity sampling data are extracted to form the corresponding temperature and humidity sequence Z, and the corresponding model code vector X is set based on the current device model. The current model code vector X and the temperature and humidity sequence Z are then sent to the remote server.

[0090] The intelligent paper-splitting device of this invention includes at least a mobile communication module, an ambient temperature and humidity sensor, an infrared proximity sensor, an automatic paper dispensing module, a storage module, and a main control module. The main control module is connected to the mobile communication module, the ambient temperature and humidity sensor, the infrared proximity sensor, the automatic paper dispensing module, and the storage module, respectively.

[0091] The functions of each component of the intelligent paper-splitting device are as follows.

[0092] 1) Mobile communication module: The mobile communication module of this invention includes 4G and 5G communication units.

[0093] The mobile communication module in this embodiment of the invention is used to handle the communication interaction process between the remote server and the main control module.

[0094] 2) Ambient temperature and humidity sensor: The environmental temperature and humidity sensor in this embodiment of the invention is used to periodically collect data on the ambient temperature and humidity of the toilet stall where the current device is located, according to the sensor's acquisition frequency, and generate corresponding temperature and humidity sampling data to send to the main control module.

[0095] 3) Infrared proximity sensor: The infrared proximity sensor in this embodiment of the invention is used to send an object proximity signal to the main control module when it detects an object approaching the current sensor.

[0096] 4) Automatic paper output module: In this embodiment of the invention, the automatic paper output module has a paper inlet at the top and a paper outlet at the bottom. The paper outlet of the automatic paper output module also serves as the paper outlet of the intelligent paper sorting device. The automatic paper output module incorporates a motor, a guide roller assembly, and a cutter assembly. When the motor rotates at a constant speed, it drives the guide roller to rotate. As the guide roller rotates, the friction between the roller surface and the paper strip drives the paper strip fed from the inlet to the paper outlet. When the motor stops rotating, the cutter assembly cuts the paper strip exiting the paper outlet. The single paper output length of the automatic paper output module = K × v loc T loc K is the preset conversion coefficient. Under normal circumstances, K=πD, where D is the diameter of the guide roller.

[0097] The automatic paper output module of this embodiment of the invention is used to, upon receiving a paper output command from the main control module, operate at the motor speed v specified in the current command. loc and paper output time T loc The system determines the motor speed and rotation duration for each operation, drives the paper output motor to rotate at a constant speed to complete the automatic paper output operation, and sends the feedback of this paper output operation to the main control module.

[0098] The paper output command includes the motor speed v. loc and paper output time T loc The paper output operation feedback includes two states: success and failure. If a paper jam or paper breakage occurs during paper output, the automatic paper output module will generate a specific failure paper output operation feedback.

[0099] 5) Storage module: The storage module in this embodiment of the invention is used to store the previous paper output time and paper output parameter Y. loc Data caching queue; The data caching queue is used to cache the temperature and humidity sampling data of the environmental temperature and humidity sensor according to the first-in-first-out caching principle.

[0100] Here, in this embodiment of the invention, the queue length of the data cache queue is greater than or equal to N. Z .

[0101] 6) Main control module: The main control module in this embodiment of the invention is used to store the temperature and humidity sampling data sent by the ambient temperature and humidity sensor into a data cache queue.

[0102] The main control module is also used to periodically update the data cache queue with the most recently cached N data according to the data reporting frequency. Z The temperature and humidity sampling data are extracted to form the corresponding temperature and humidity sequence Z, and the corresponding model code vector X is set based on the current device model. The current model code vector X and the temperature and humidity sequence Z are sent to the remote server through the mobile communication module.

[0103] The main control module is also used to read the corresponding previous paper output time and the shortest interval duration Δt from the storage module when it receives an object approach signal sent by the infrared proximity sensor. loc Motor speed v loc Paper output time T loc And check whether the interval between the current time and the previous paper output time is greater than or equal to the shortest interval length Δt. loc Perform identification; if so, then carry the motor speed v. loc and paper output time T loc The paper output command is sent to the automatic paper output module; otherwise, a delay is performed for a preset waiting time, and at the end of this wait, the interval between the latest current time and the previous paper output time is checked to see if it is greater than or equal to the shortest interval Δt. loc Perform a second identification; if the result of this identification is yes, then the motor speed v will be carried. loc and paper output time T loc The paper output command is sent to the automatic paper output module. If the recognition result is negative, a delay is performed again until the interval between the latest current time and the previous paper output time is greater than or equal to the shortest interval Δt. loc until.

[0104] Here, the waiting time in this embodiment of the invention is a pre-set time length parameter.

[0105] The main control module is also used to identify whether the current paper output operation feedback is successful when it receives the paper output operation feedback sent back by the automatic paper output module; if so, it resets the previous paper output time stored in the storage module based on the current time.

[0106] The main control module is also used to, upon receiving the paper output parameter Y sent by the remote server via the mobile communication module, modify the paper output parameter Y stored in the storage module based on the current paper output parameter Y. loc Reset.

[0107] Step 3: When the remote server receives the model code vector X and temperature and humidity sequence Z sent by any intelligent paper sorting device, it first selects the corresponding model as the current model based on the current model code vector X; then it inputs the current model code vector X and temperature and humidity sequence Z into the current model for processing to obtain the corresponding paper output parameters Y; and then it sends the current paper output parameters Y back to the current intelligent paper sorting device.

[0108] Step 4: When each intelligent paper-splitting device receives the paper output parameter Y sent by the remote server, it updates the paper output parameter Y stored on the device side based on the current paper output parameter Y. loc .

[0109] Among them, the paper output parameter Y locIncluding the shortest interval duration △t loc Motor speed v loc Paper output time T loc .

[0110] The system provided in Embodiment 2 of the present invention for implementing the method described in Embodiment 1 above, such as Figure 3 The schematic diagram of a processing system for optimizing paper output parameters of a paper-splitting device provided in Embodiment 2 of the present invention is shown. The system includes: a model preparation system 201, a remote server 202, and multiple intelligent paper-splitting devices 203.

[0111] The remote server 202 is connected to the model preparation system 201 and each intelligent paper-splitting device 203 respectively; each intelligent paper-splitting device 203 corresponds to a type of equipment model and is installed in a designated toilet stall.

[0112] The model preparation system 201 is used to construct a paper output parameter optimization model M for the intelligent paper separating device 203; and to construct corresponding model datasets for various models of the intelligent paper separating device 203 through temperature and humidity zoning control experiments; and to train the paper output parameter optimization model M based on the model datasets of each model to obtain the corresponding model model; and to deploy all model models to the remote server 202; wherein, the paper output parameter optimization model is used to optimize the paper output parameters according to the model input model encoding vector X and temperature and humidity sequence Z and output the corresponding paper output parameters Y; the model encoding vector X is the unique thermal encoding vector of the intelligent paper separating device 203, which is composed of N X Model code x j Composition, 1≤indexj≤N X N X For the total number of device models of intelligent paper-splitting equipment 203, X contains exactly one x. j The value is 1 for all others and 0 for the rest; the temperature and humidity sequence Z is the environmental temperature and humidity sampling sequence of the toilet stall where the intelligent paper-splitting device 203 is located, which is composed of N Z Each sampled data z i Composition, sampled data z i Including temperature a i Humidity b i 1 ≤ index i ≤ N Z N Z The preset total number of sampling points, z = every two sampling data points i z i+1 The time intervals between them are equal, which is the preset sampling interval duration; the paper output parameter Y is the optimized paper output parameter of the intelligent paper separation device 203, including the shortest interval duration Δt between two paper output operations, the motor speed v of a single paper output, and the paper output duration T; each model dataset consists of multiple first data records; the first data record includes the model code vector X, the temperature and humidity sequence Z, and the label parameter Y. tag; Label parameter Y tag Including the tag interval duration △t tag Tag rotation speed v tag Label output time T tag .

[0113] Each intelligent paper-splitting device 203 is used to continuously sample and cache the ambient temperature and humidity of the toilet stall where the device is located according to the sampling interval; and periodically uploads the most recently cached N data according to the preset data reporting frequency. Z The temperature and humidity sampling data are extracted to form the corresponding temperature and humidity sequence Z, and the corresponding model code vector X is set based on the current device model. The current model code vector X and the temperature and humidity sequence Z are then sent to the remote server 202.

[0114] Each intelligent paper-splitting device 203 is also used to update the paper output parameters Y stored on the device side based on the current paper output parameters Y when it receives the paper output parameters Y sent by the remote server 202. loc Among them, the paper output parameter Y loc Including the shortest interval duration △t loc Motor speed v loc Paper output time T loc .

[0115] When the remote server 202 receives the model code vector X and temperature and humidity sequence Z sent by any intelligent paper sorting device 203, it first selects the corresponding model as the current model based on the current model code vector X; then it inputs the current model code vector X and temperature and humidity sequence Z into the current model for processing to obtain the corresponding paper output parameters Y; and then it sends the current paper output parameters Y back to the current intelligent paper sorting device 203.

[0116] The processing system for optimizing the paper output parameters of a paper separating device provided in Embodiment 2 of the present invention can execute the method steps in the above method embodiments. Its implementation principle and technical effect are similar, and will not be described again here.

[0117] In summary, the technical solution of the method and system for optimizing paper output parameters of a paper-splitting device provided by this invention first constructs a paper output parameter optimization model M for the intelligent paper-splitting device based on artificial intelligence technology. Then, through temperature and humidity zoning control experiments, corresponding model datasets are constructed for various models of the intelligent paper-splitting device. Based on the model datasets of each model, the paper output parameter optimization model M is trained to obtain the corresponding model, and all model models are deployed to a remote server. Next, each intelligent paper-splitting device continuously samples and caches the ambient temperature and humidity of its respective toilet stall at the front end, and periodically sends the model code vector and the most recently cached temperature and humidity sequence to the remote server. The remote server, in the background, predicts the paper output parameters based on the corresponding model model of each device and promptly feeds back the prediction results to the front-end device. This invention improves the environmental adaptability of the equipment and reduces the equipment failure rate.

[0118] The processing steps described in the methods or system embodiments disclosed herein may be implemented based on hardware modules, software modules executed by a processor, or a combination of both. The software modules may be located in random access memory (RAM), main memory, read-only memory (ROM), electrically programmable ROM, electrically erasable programmable ROM, registers, hard disks, removable disks, CD-ROMs, or any other form of storage medium known in the art.

[0119] The specific embodiments described above further illustrate the purpose, technical solution, and beneficial effects of the present invention. It should be understood that the above description is only a specific embodiment of the present invention and is not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A method for optimizing paper output parameters of a paper separating device, characterized in that, The method includes: A paper output parameter optimization model M is constructed for the intelligent paper separating equipment; and corresponding model datasets are constructed for various equipment models of the intelligent paper separating equipment through temperature and humidity zoning control experiments; the paper output parameter optimization model M is trained based on the model datasets of each model to obtain the corresponding model model; and all model models are deployed to a remote server; wherein, the paper output parameter optimization model is used to optimize the paper output parameters according to the model input model encoding vector X and temperature and humidity sequence Z and output the corresponding paper output parameters Y; the model encoding vector X is the unique thermal encoding vector of the equipment model of the intelligent paper separating equipment, which is composed of N X Model code x j Composition, 1≤indexj≤N X N X X represents the total number of device models of the intelligent paper-splitting equipment, and there is one and only one x. j The value is 1 for all others and 0 for the rest; the temperature and humidity sequence Z is the environmental temperature and humidity sampling sequence of the toilet stall where the intelligent paper dispensing device is located, consisting of N Z Each sampled data z i Composition, the sampled data z i Including temperature a i Humidity b i 1 ≤ index i ≤ N Z N Z The preset total number of sampling points, z = every two sampling data points i z i+1 The time intervals between them are equal, which is the preset sampling interval duration; the paper output parameter Y is the optimized paper output parameter of the intelligent paper separation device, including the shortest interval duration Δt between two paper output operations, the motor speed v of a single paper output, and the paper output duration T; each of the model datasets consists of multiple first data records; the first data record includes the model encoding vector X, the temperature and humidity sequence Z, and the label parameter Y. tag The label parameter Y tag Including the tag interval duration △t tag Tag rotation speed v tag Label output time T tag The remote server is connected to each of the intelligent paper-splitting devices, and each intelligent paper-splitting device corresponds to a type of device model and is installed in a designated toilet stall. Each of the aforementioned intelligent paper-splitting devices continuously samples and caches the ambient temperature and humidity of the toilet stall where the device is located according to the sampling interval; and periodically reports the most recently cached N data according to a preset data reporting frequency. Z The temperature and humidity sampling data are extracted to form the corresponding temperature and humidity sequence Z, and the corresponding model encoding vector X is set based on the current device model. The current model encoding vector X and the temperature and humidity sequence Z are then sent to the remote server. When the remote server receives the model code vector X and the temperature and humidity sequence Z sent by any of the intelligent paper-splitting devices, it first selects the corresponding model as the current model based on the current model code vector X; then it inputs the current model code vector X and the temperature and humidity sequence Z into the current model for processing to obtain the corresponding paper output parameter Y; and then it sends the current paper output parameter Y back to the current intelligent paper-splitting device. When each of the intelligent paper-splitting devices receives the paper output parameter Y sent by the remote server, it updates the paper output parameter Y stored on the device side based on the current paper output parameter Y. loc Wherein, the paper output parameter Y loc Including the shortest interval duration △t loc Motor speed v loc Paper output time T loc .

2. The method for optimizing paper output parameters of a paper separating device according to claim 1, characterized in that, The intelligent paper-splitting device includes at least a mobile communication module, an ambient temperature and humidity sensor, an infrared proximity sensor, an automatic paper dispensing module, a storage module, and a main control module; the main control module is connected to the mobile communication module, the ambient temperature and humidity sensor, the infrared proximity sensor, the automatic paper dispensing module, and the storage module, respectively. The mobile communication module includes 4G and 5G communication units; the mobile communication module is used to handle the communication interaction process between the remote server and the main control module. The ambient temperature and humidity sensor is used to periodically collect data on the ambient temperature and humidity of the toilet stall where the current device is located, according to the sensor's collection frequency, and generate corresponding temperature and humidity sampling data to send to the main control module. The infrared proximity sensor is used to send an object proximity signal to the main control module when it detects an object approaching the current sensor. The automatic paper output module is used to, upon receiving a paper output command from the main control module, operate at the motor speed v specified in the current command. loc and the paper output time T loc The system determines the motor speed and rotation duration for each automatic paper feeding operation, driving the paper feeding motor to rotate at a constant speed to complete the operation. Feedback on this operation is then sent to the main control module. This feedback includes both success and failure states. If a paper jam or paper breakage occurs during paper feeding, the automatic paper feeding module will generate a failure feedback. The single paper feeding length of the automatic paper feeding module is calculated as K × v. loc T loc K is the preset conversion coefficient; The storage module is used to store the previous paper output time and the paper output parameter Y. loc A data buffer queue; the data buffer queue is used to buffer the temperature and humidity sampling data of the environmental temperature and humidity sensor according to the first-in-first-out (FIFO) buffering principle, and the queue length of the data buffer queue is greater than or equal to N. Z ; The main control module is used to store the temperature and humidity sampling data sent by the ambient temperature and humidity sensor into the data cache queue; The main control module is also used to periodically update the data cache queue with the most recently cached N data according to the data reporting frequency. Z The temperature and humidity sampling data are extracted to form the corresponding temperature and humidity sequence Z, and the corresponding model code vector X is set based on the current device model. The current model code vector X and the temperature and humidity sequence Z are sent to the remote server through the mobile communication module. The main control module is also used to read the corresponding previous paper output time and the shortest interval duration Δt from the storage module when it receives the object approach signal sent by the infrared proximity sensor. loc The motor speed v loc The paper output time T loc And whether the interval between the current time and the previous paper output time is greater than or equal to the shortest interval length Δt. loc Perform identification; if so, then carry the motor speed v loc and the paper output time T loc The paper output command is sent to the automatic paper output module; if not, a delay is performed for a preset waiting time, and at the end of this wait, the interval between the latest current time and the previous paper output time is checked to see if it is greater than or equal to the shortest interval Δt. loc Perform a second identification; if the result of this identification is yes, then the motor speed v will be carried. loc and the paper output time T loc The paper output command is sent to the automatic paper output module. If the recognition result is negative, the system will wait again for the specified waiting time until the interval between the latest current time and the previous paper output time is greater than or equal to the shortest interval Δt. loc until; The main control module is also used to identify whether the current paper output operation feedback is successful when it receives the paper output operation feedback sent back by the automatic paper output module; if so, it resets the previous paper output time stored in the storage module based on the current time. The main control module is further configured to, upon receiving the paper output parameter Y sent by the remote server via the mobile communication module, adjust the paper output parameter Y stored in the storage module based on the current paper output parameter Y. loc Reset.

3. The method for optimizing paper output parameters of a paper separating device according to claim 1, characterized in that, The first model input terminal of the paper output parameter optimization model M is used to receive the model code vector X, the second model input terminal is used to receive the temperature and humidity sequence Z, and the model output terminal is used to output the paper output parameter Y. The paper output parameter optimization model M includes a feature mapping layer, a preprocessing layer, a feature encoder, a feature pooling layer, a feature fusion layer, a fused feature prediction layer, and a prediction output layer. The input of the feature mapping layer is connected to the input of the first model, and its output is connected to the first input of the feature fusion layer; the input of the preprocessing layer is connected to the input of the second model, and its output is connected to the input of the feature encoder; the output of the feature encoder is connected to the input of the feature pooling layer; the output of the feature pooling layer is connected to the second input of the feature fusion layer; the output of the feature fusion layer is connected to the input of the fused feature prediction layer; the output of the fused feature prediction layer is connected to the second input of the prediction output layer; the first input of the prediction output layer is connected to the input of the first model, and the output of the prediction output layer is connected to the model output. The feature mapping layer is used to perform linear mapping processing on the model encoding vector X to obtain the corresponding feature vector H1, which is then sent to the feature fusion layer. The linear mapping method of the feature vector H1 is as follows: ; W X The weight matrix of the feature mapping layer has a shape of D1×N. X D1 is the preset first feature dimension; the shape of the model encoding vector X is 1×N. X The shape of the feature vector H1 is 1×D1. The pretreatment layer is used to determine the global minimum temperature A. min Global maximum temperature A max Global minimum humidity B min Global maximum humidity B max The temperature and humidity sequence Z is normalized to obtain the corresponding normalized sequence Z. * The normalized sequence Z is encoded according to the embedding encoding method of the Transformer model encoder. * Embedding encoding is performed to obtain the corresponding embedding vector E. Z Send to the feature encoder; Wherein, the normalized sequence Z * Including N Z Normalized data The normalized data Including normalized temperature Normalized humidity The normalized data With the sampled data z i One-to-one correspondence, the normalized temperature With the temperature a i One-to-one correspondence, the normalized humidity With the humidity b i One-to-one correspondence; ， ； The embedding vector E Z The shape is N Z ×D2, where D2 is the preset second feature dimension; The feature encoder is implemented based on the encoder structure of the Transformer model; the feature encoder is used to process the embedding vector E. Z Feature encoding is performed to obtain the corresponding feature vector H. Z Send to the feature pooling layer; Wherein, the feature vector H Z The shape is N Z ×D2; The feature pooling layer is used to process the feature vector H. Z Each feature channel is subjected to average pooling to obtain the corresponding feature vector H2, which is then sent to the feature fusion layer. The shape of the feature vector H2 is 1×D2; The feature fusion layer is used to perform feature concatenation on feature vectors H1 and H2 to obtain the corresponding feature vector H3, which is then sent to the fused feature prediction layer. The shape of the feature vector H3 is 1×(D1+D2); The fusion feature prediction layer is implemented based on the MLP model; the fusion feature prediction layer is used to perform normalized vector prediction processing on the feature vector H3 to obtain the corresponding prediction vector H4 and send it to the prediction output layer; The prediction method for the prediction vector H4 is as follows: ； W1 and W2 are the first and second weight matrices of the fused feature prediction layer, and b1 and b2 are the first and second bias vectors of the fused feature prediction layer; the shape of W1 is D3×(D1+D2), where D3 is a preset third feature dimension, and the shape of W2 is 3×D3; the shape of b1 is 1×D3, and the shape of b2 is 1×3; ReLU() is the ReLU activation function, and sigmoid() is the Sigmoid activation function; activation vectors The shape of the prediction vector H4 is 1×D3; the shape of the prediction vector H4 is 1×3, consisting of three corresponding vector data. , , composition; The prediction output layer is used to convert the model code x that is 1 in the model code vector X. j Let index j be the index of the current model. * ; and based on the prediction vector H4 and the current model index j * Corresponding global minimum interval duration Global maximum interval duration Global minimum speed Global maximum speed Minimum global output time Maximum global paper output time The three types of paper output parameters are inversely normalized to obtain the corresponding paper output parameter Y and output it. The inverse normalization method for the three types of paper output parameters Δt, v, and T of the paper output parameter Y is as follows: ， ， 。 4. The method for optimizing paper output parameters of a paper separating device according to claim 3, characterized in that, The process of constructing corresponding model datasets for various models of the intelligent paper-splitting equipment through temperature and humidity zone control experiments specifically includes: Step 41, set the global temperature range [A] min A max Divide into N equal parts A Individual temperature ranges [a] min,u ,a max,u ], and set the global humidity range [B min B max Divide into N equal parts B Individual humidity ranges [b] min,s ,b max,s ], and for N A Individual temperature ranges and N B The corresponding N is obtained by combining the individual humidity ranges with the temperature and humidity ranges. A ×N B Temperature and humidity combination range P u,s [(a min,u ,a max,u ),(b min,s ,b max,s The duration of the temperature and humidity sequence Z is taken as the corresponding experimental duration T. test =N Z × Sampling interval duration; and according to the preset N R Temperature and humidity control modes, for each of the aforementioned temperature and humidity combination ranges P u,s Set the corresponding N R Group temperature and humidity adjustment curve R u,s,r ; Where 1 ≤ index u ≤ N A 1 ≤ index s ≤ N B 1 ≤ index r ≤ N R N A The total number of preset temperature ranges, N B The total number of preset humidity ranges, N R This represents the total number of preset adjustment modes; a min,u a max,u Let be the minimum and maximum temperatures of the u-th sub-temperature range; b min,s b max,s Let be the minimum and maximum humidity of the s-th sub-humidity interval; The N R The temperature and humidity regulation modes include at least the following modes: constant humidity and linear temperature increase mode, constant humidity and linear temperature decrease mode, constant humidity and step temperature increase mode, constant humidity and step temperature decrease mode, constant humidity and temperature sinusoidal fluctuation mode, constant temperature and humidity and linear temperature increase mode, constant temperature and humidity and linear temperature decrease mode, constant temperature and humidity and step temperature increase mode, constant temperature and humidity and step temperature decrease mode, constant temperature and humidity and sinusoidal fluctuation mode, temperature and humidity increase linearly with the same slope mode, temperature and humidity increase linearly with different slope modes, temperature and humidity decrease linearly with the same slope mode, temperature and humidity decrease linearly with different slope modes, temperature increases linearly but humidity decreases linearly, temperature decreases linearly but humidity increases linearly, temperature and humidity fluctuate based on the same phase sinusoidal fluctuation mode, temperature and humidity increase in the same direction step mode, and temperature and humidity decrease in the same direction step mode. The temperature and humidity control curves R for each group u,s,r Each curve consists of a set of corresponding temperature and humidity control curves, and the duration of both curves is equal to the experimental duration T. test The trends of the two curves match the corresponding temperature and humidity regulation modes. Step 42: Select each model of the intelligent paper-splitting device as the current model; and select one test sample for the current model as the current sample. Step 43, for the current prototype in each of the temperature and humidity combination ranges P u,s Target paper output length Configure settings; Step 44, set the global interval duration range [△t] of the current prototype. min ,△t max Divide into N equal parts △t Sample Δt q and the global speed range [v] of the current prototype min ,v max Divide into N equal parts v Sample v g and the global paper output time interval [T] of the current prototype. min ,T max Divide into N equal parts T Sample T o ; and for the N corresponding to the current prototype △t The sampling Δt q N v The sample v g N T The sample T o By combining them, we get N. △t ×N v ×N T Group sampling parameters Y q,g,o (△t q ,v g ,T o ); Where 1 ≤ index q ≤ N △t 1 ≤ index g ≤ N v 1 ≤ index o ≤ N T N △t The preset interval duration, the total number of sampling points, N v The preset total number of speed sampling points, N T The total number of sampling points for the preset paper output time; Step 45, convert the temperature and humidity adjustment curves R of each group into... u,s,r As the current environment curve; and each of the sampling parameters Y is used as the current environment curve; q,g,o Using the current paper output parameters as the current parameters, and in a temperature and humidity adjustable experimental environment, with the current environmental curve and the current paper output parameters as the current experimental conditions, the current prototype is subjected to a test run for a duration of T. test A set of corresponding temperature and humidity sequences Z was obtained from the continuous paper output experiment. u,s,r,q,g,o Experimental score C u,s,r,q,g,o ; and the N corresponding to the current environment curve △t ×N v ×N T The experimental score C mentioned above u,s,r,q,g,o The sampling parameter Y corresponding to the maximum score in q,g,o As the label parameter corresponding to the current environment curve ; Step 46: Use the index j corresponding to the current device model in the model encoding vector X as the current model index j. * And set a length of N X The all-zero vector is taken as the current vector, and the j-th element in the current vector is... * Each vector data is reset to 1, and the current vector after reset is used as the model code vector X corresponding to the current device model; and the temperature and humidity sequences Z of the current prototype are used. u,s,r,q,g,o and the corresponding tag parameters The model code vector X forms a corresponding first data record; and the obtained N A ×N B ×N R ×N △t ×N v ×N T The first data record is deduplicated; and the model dataset corresponding to the current device model is composed of all the remaining first data records after deduplication.

5. The method for optimizing paper output parameters of a paper separating device according to claim 4, characterized in that, In a temperature and humidity adjustable experimental environment, using the current environmental curve and the current paper output parameters as the current experimental conditions, the current prototype undergoes a test run for a duration of T. test A set of corresponding temperature and humidity sequences Z was obtained from the continuous paper output experiment. u,s,r,q,g,o Experimental score C u,s,r,q,g,o Specifically, it includes: Step 51: Before starting this round of continuous paper output experiment, adjust the paper output parameter Y on the current prototype based on the current paper output parameters. loc Perform a reset; and based on the sampled v of the current paper output parameters. g The sampling T o And the corresponding experimental paper output length L is calculated based on the conversion coefficient K of the current prototype. u,s,r,q,g,o ; and install a continuous paper output program run by the main control module on the current prototype; Among them, L u,s,r,q,g,o =K×v g T o ; The continuous paper output program is used to, after the program runs, first store an initialized first failure time parameter in the storage module of the current prototype; then, use the shortest interval duration Δt stored in the storage module. loc The system sends continuous paper output commands to the automatic paper output module of the current prototype at continuous command intervals; it also identifies whether each paper output operation feedback returned by the automatic paper output module is a failure, and resets the first failure time parameter based on the current time when the first specific failure paper output operation feedback is received. Step 52: At the start of this round of continuous paper output experiment, drive the main control module of the current prototype to run the continuous paper output program; Step 53: During the continuous paper output experiment, the temperature and humidity of the current experimental environment are continuously modulated based on the current environmental curve; and the temperature and humidity of the current experimental environment are continuously sampled using the sampling interval as the continuous sampling interval, and the sampled temperature and humidity sequence is used as the corresponding temperature and humidity sequence Z. u,s,r,q,g,o ; Step 54: At the end of this round of continuous paper output experiment, drive the main control module to shut down the continuous paper output program; read the corresponding first failure time parameter from the storage module; and record the end time of this round of experiment. Step 55: Identify whether the first fault time parameter is empty; if so, set the corresponding fault duration T. err If the value is 0, then the interval between the first failure time parameter and the experiment end time is calculated, and the calculation result is used as the corresponding failure duration T. err Based on the fault duration T err The experimental duration T test Calculate the corresponding failure duration ratio f u,s,r,q,g,o f u,s,r,q,g,o =T err / T test Based on the fault duration ratio f u,s,r,q,g,o The experimental paper output length L u,s,r,q,g,o And the target paper output length corresponding to the current prototype and the current environment curve. Calculate the corresponding experimental score C u,s,r,q,g,o ; Wherein, the fault duration ratio f u,s,r,q,g,o The value of is between 0 and 1; The experimental score C u,s,r,q,g,o The calculation method is as follows: ； α and β are preset weighting coefficients, where α > β > 0.

6. The method for optimizing paper output parameters of a paper separating device according to claim 1, characterized in that, The process of training the paper output parameter optimization model M based on the model dataset of each model to obtain the corresponding model model specifically includes: Step 61: Take each device model of the intelligent paper-splitting device as the current device model; take the model dataset corresponding to the current device model as the current model dataset; and initialize the paper output parameter optimization model M based on a set of preset initial model parameters to obtain the corresponding current base model; Step 62: Based on a preset first segmentation ratio, randomly divide the current model dataset into two subsets, denoted as the first training set and the first evaluation set; and calculate the total number of records N in the first training set. tr The total number of records N in the first evaluation set av Statistical analysis was conducted separately; Both the first training set and the first evaluation set consist of multiple first data records; the ratio N of the total number of records in the first training set to the total number of records in the first evaluation set is... tr :N av The first segmentation ratio is satisfied; Step 63: Take each of the first data records in the first training set as the current record; input the model code vector X and the temperature and humidity sequence Z of the current record into the current base model for processing, and record the paper output parameter Y of the current model as the corresponding prediction parameter. ; and the currently recorded tag parameter Y tag Record as the corresponding tag parameter ; and by the prediction parameters and the label parameters Form a corresponding first prediction-label pair; and obtain N tr The first prediction-label pair is substituted into the model loss function L. M The corresponding first loss value is obtained through calculation; Where 1 ≤ index k ≤ N tr ; The prediction parameters Including the shortest interval duration Motor speed Paper output time ; The tag parameters Including tag interval duration Tag rotation speed Label output time ; The model loss function L M for: ； Step 64: Identify whether the first loss value meets the preset first loss value range; if it does, proceed to step 65; if not, based on the preset model optimizer, move towards making the model loss function L... M The model parameters of the current base model are optimized in one round in the direction of reaching the minimum value, and the process returns to step 63 after this round of optimization is completed; The model optimizer includes the Adam optimizer and the SGD optimizer; Step 65: Take each of the first data records in the first evaluation set as the current record; input the model code vector X and the temperature and humidity sequence Z of the current record into the current base model for processing, and record the paper output parameter Y of the current model as the corresponding prediction parameter. ; and the currently recorded tag parameter Y tag Record as the corresponding tag parameter ; and by the prediction parameters and the label parameters Form a corresponding second prediction-label pair; and obtain N av The second prediction-label pair is substituted into the model evaluation function L. A The corresponding first evaluation value is obtained through calculation; Where 1 ≤ index l ≤ N av ; The prediction parameters Including the shortest interval duration Motor speed Paper output time ; The tag parameters Including tag interval duration Tag rotation speed Label output time ; The model evaluation function L A for: ； Step 66: Identify whether the first evaluation value meets the preset first evaluation value range; if not, return to step 62; if yes, stop training and use the current base model as the model corresponding to the current device model.

7. A system for implementing the processing method for optimizing the paper output parameters of a paper separating device as described in any one of claims 1-6, characterized in that, The system includes: a model preparation system, a remote server, and multiple intelligent paper-splitting devices; The remote server is connected to the model preparation system and each of the intelligent paper-splitting devices; each intelligent paper-splitting device corresponds to a type of device model and is installed in a designated toilet stall. The model preparation system is used to construct a paper output parameter optimization model M for the intelligent paper separating device; and to construct corresponding model datasets for various models of the intelligent paper separating device through temperature and humidity zoning control experiments; and to train the paper output parameter optimization model M based on the model datasets of each model to obtain the corresponding model model; and to deploy all the model models to the remote server; wherein, the paper output parameter optimization model is used to optimize the paper output parameters according to the model input model encoding vector X and temperature and humidity sequence Z and output the corresponding paper output parameters Y; the model encoding vector X is the unique thermal encoding vector of the intelligent paper separating device model, which is composed of N X Model code x j Composition, 1≤indexj≤N X N X X represents the total number of device models of the intelligent paper-splitting equipment, and there is one and only one x. j The value is 1 for all others and 0 for the rest; the temperature and humidity sequence Z is the environmental temperature and humidity sampling sequence of the toilet stall where the intelligent paper dispensing device is located, consisting of N Z Each sampled data z i Composition, the sampled data z i Including temperature a i Humidity b i 1 ≤ index i ≤ N Z N Z The preset total number of sampling points, z = every two sampling data points i z i+1 The time intervals between them are equal, which is the preset sampling interval duration; the paper output parameter Y is the optimized paper output parameter of the intelligent paper separation device, including the shortest interval duration Δt between two paper output operations, the motor speed v of a single paper output, and the paper output duration T; each of the model datasets consists of multiple first data records; the first data record includes the model encoding vector X, the temperature and humidity sequence Z, and the label parameter Y. tag The label parameter Y tag Including the tag interval duration △t tag Tag rotation speed v tag Label output time T tag ; Each of the aforementioned intelligent paper-splitting devices is used to continuously sample and cache the ambient temperature and humidity of the toilet stall where the device is located according to the sampling interval; and periodically upload the most recently cached N data according to a preset data reporting frequency. Z The temperature and humidity sampling data are extracted to form the corresponding temperature and humidity sequence Z, and the corresponding model encoding vector X is set based on the current device model. The current model encoding vector X and the temperature and humidity sequence Z are then sent to the remote server. Each of the aforementioned intelligent paper-splitting devices is further configured to, upon receiving the paper output parameter Y sent by the remote server, update the paper output parameter Y stored on the device side based on the current paper output parameter Y. loc Wherein, the paper output parameter Y loc Including the shortest interval duration △t loc Motor speed v loc Paper output time T loc ; When the remote server receives the model code vector X and the temperature and humidity sequence Z sent by any of the intelligent paper-splitting devices, it first selects the corresponding model as the current model based on the current model code vector X; then it inputs the current model code vector X and the temperature and humidity sequence Z into the current model for processing to obtain the corresponding paper output parameter Y; and then it sends the current paper output parameter Y back to the current intelligent paper-splitting device.

Images