Manicure pen direct current brushless driver operation speed regulation and control method based on artificial intelligence

Abstract

The invention discloses a manicure pen direct current brushless driver operation speed regulation and control method based on artificial intelligence, which is used in the field of manicure instrument regulation and control, and comprises the following steps: obtaining operation data of a manicure pen direct current brushless driver; according to the characteristic data, constructing a performance prediction model of the manicure pen direct current brushless driver, and utilizing the performance prediction model to predict the operation state of the manicure pen direct current brushless driver; predicting an operation state in a future preset time according to the output of the performance prediction model, and calculating a pulse width modulation duty ratio and a bus voltage; and the actual operation state of the adjusted direct current brushless driver of the manicure pen is monitored in real time, new operation data is fed back to the performance prediction model, and parameters of the performance prediction model are optimized. According to the method, key feature data are automatically screened out through a soft threshold function in the residual shrinkage unit; weights of easy-to-classify samples are reduced, and a maintenance plan is optimized.

Description

Technical Field

[0001] The invention relates to the field of beauty instrument control, and in particular to an artificial intelligence-based method for controlling the operating speed of a direct current brushless driver of a nail art pen. Background Art

[0002] A nail pen is an electric tool used for nail art and is widely used in nail art processes such as grinding, engraving, and polishing. A nail pen requires a very precise and controllable speed to complete different nail art tasks. Compared to traditional brushed motors, brushless motors use electronic devices to control the commutation inside the motor instead of using physical brushes. Brushless motors usually provide a smoother speed and higher control accuracy, which is very suitable for nail pens that require precise operations. The driver of the brushless motor is responsible for receiving control signals and adjusting the power supply of the motor accordingly to control the speed, direction and torque of the motor.

[0003] Different nail operations (such as polishing, drawing or cutting) require different speeds; for example, fine line drawing requires slow and stable speed, while removing old paint layers requires high speed. Accurate speed control is directly related to the quality and effect of the operation. Appropriate speed control can improve energy efficiency and reduce unnecessary energy consumption, thereby extending battery life and reducing energy costs. The introduction of artificial intelligence technology can further improve the intelligence and automation level of speed control.

[0004] The lack of real-time data monitoring and immediate response mechanisms in traditional control methods leads to incomplete equipment performance monitoring and inability to timely detect and respond to problems in equipment operation, such as mechanical failures or electrical problems, which increases the risk of equipment failure and downtime. Existing technologies fail to effectively screen and process noise and irrelevant information in the data, resulting in poor data quality, which affects the accuracy and reliability of subsequent analysis and decision-making. Traditional methods lack efficient algorithms to process and predict complex data patterns, especially when facing unknown or changing operating conditions. The model's predictive and generalization capabilities are insufficient, which limits the optimization of equipment performance and reliability assurance. It also lacks the ability to accurately calculate and adjust key control parameters (such as pulse width modulation duty cycle and bus voltage), which leads to energy waste and substandard performance, making it impossible to ensure that the equipment operates in the best condition. In addition, it does not support rapid dynamic adjustment of control parameters to adapt to immediate changes. At the same time, there is a lack of effective preventive maintenance strategies, making it difficult to achieve early diagnosis and treatment of equipment failures, which increases maintenance costs and equipment downtime. Summary of the invention

[0005] In order to overcome the above problems, the present invention aims to propose an artificial intelligence-based nail pen DC brushless driver operating speed control method, with the aim of solving the problem that traditional methods lack efficient algorithms to process and predict complex data patterns, especially when facing unknown or changing operating conditions, the model's predictive ability and generalization ability are insufficient, which limits the optimization of equipment performance and reliability assurance, and lacks the ability to accurately calculate and adjust key control parameters (such as pulse width modulation duty cycle and bus voltage), which may lead to energy waste and substandard performance, thereby failing to ensure that the equipment operates in the best state. In addition, it does not support rapid dynamic adjustment of control parameters to adapt to instant changes. At the same time, there is a lack of effective preventive maintenance strategies, and it is difficult to achieve early diagnosis and processing of equipment failures, which increases maintenance costs and equipment downtime.

[0006] To this end, the specific technical solution adopted by the present invention is as follows:

[0007] A method for controlling the operating speed of a DC brushless driver for a nail art pen based on artificial intelligence, the method comprising the following steps:

[0008] S1. Obtaining operation data of a nail art pen DC brushless driver, preprocessing the acquired operation data, and extracting feature data from the preprocessed operation data;

[0009] S2. Using the deep residual shrinkage network, a performance prediction model of the nail pen DC brushless driver is constructed according to the feature data, and the operating state of the nail pen DC brushless driver is predicted using the performance prediction model;

[0010] S3. Predicting the operating state within a preset time in the future according to the output of the performance prediction model, calculating the pulse width modulation duty cycle and the bus voltage, and adjusting the pulse width modulation duty cycle and the bus voltage to meet the preset target operating state based on the calculation results;

[0011] S4, real-time monitoring of the actual operating status of the adjusted nail art pen DC brushless driver, if it is detected that the performance is lower than the preset threshold, an alarm is immediately triggered, and the new operating data is fed back to the performance prediction model to optimize the parameters of the performance prediction model;

[0012] Among them, using the deep residual shrinkage network, building a performance prediction model of the nail pen DC brushless driver according to the feature data, and using the performance prediction model to predict the operating state of the nail pen DC brushless driver includes the following steps:

[0013] S21, using the residual shrinkage unit in the deep residual shrinkage network to automatically filter and optimize feature data through a soft threshold function;

[0014] S22. Based on the optimized and selected key features, a performance prediction model of the DC brushless driver for nail art pens is constructed;

[0015] S23. Use the focus loss function to train a performance prediction model, and use the trained performance prediction model to predict the operating status of the nail pen DC brushless driver.

[0016] Optionally, obtaining the operating data of the nail art pen DC brushless driver, preprocessing the obtained operating data, and extracting feature data from the preprocessed operating data includes the following steps:

[0017] S11, using a pre-configured sensor to collect operation data of a nail art pen DC brushless driver in real time;

[0018] S12, using a digital filter to remove high-frequency noise in the operating data, and performing standardization processing on the operating data;

[0019] S13, converting the normalized operating data into frequency domain data, and obtaining frequency components and corresponding amplitudes by fast Fourier transform;

[0020] S14. Based on the frequency domain data, analyze the harmonics in the spectrum, calculate the power spectrum density, and evaluate the power distribution of the signal at each frequency; identify the frequency components and peaks in the signal;

[0021] S15. Analyze each frequency component according to the obtained frequency domain data, and extract characteristic data of the nail art pen DC brushless driver performance.

[0022] Optionally, based on the frequency domain data, analyzing harmonics in the spectrum, calculating the power spectrum density, and evaluating the power distribution of the signal at each frequency; identifying the frequency components and peaks in the signal comprises the following steps:

[0023] S141, using the result of fast Fourier transform, identifying the main harmonics and subharmonics of the signal spectrum from the frequency domain data;

[0024] S142, based on identifying the main harmonics and subharmonics of the signal spectrum from the frequency domain data, calculating the power spectrum density of the entire frequency domain data, and evaluating the power distribution of the signal at all frequency points;

[0025] S143, based on power spectrum density analysis, identifying frequency points where power is concentrated;

[0026] S144. Evaluate the power distribution of the signal at each frequency point, and identify the frequency components and peaks in the signal.

[0027] Optionally, using the residual shrinkage unit in the deep residual shrinkage network to automatically screen and optimize feature data through a soft threshold function includes the following steps:

[0028] S211, sending the extracted feature data to the residual shrinkage unit in the deep residual shrinkage network;

[0029] S212, automatically identifying feature data that does not meet preset requirements using a soft threshold function inside the residual shrinkage unit;

[0030] S213. Set the feature data that does not meet the preset requirements to zero, retain the feature data that meets the preset requirements, and construct a feature data set.

[0031] Optionally, using a soft threshold function inside the residual shrinkage unit to automatically identify feature data that does not meet preset requirements includes the following steps:

[0032] S2121, embedding a residual shrinkage unit in the neural network, and selecting a threshold type according to a preset application requirement;

[0033] S2122, processing the feature data of each channel using a soft threshold function in the residual shrinkage unit, and calculating the absolute value of each feature;

[0034] S2123, using global average pooling and fully connected layers to calculate independent thresholds for each channel;

[0035] S2124, performing feature selection according to the calculated independent threshold and the absolute value of each feature;

[0036] Among them, the expression for processing the feature data of each channel using the soft threshold function in the residual shrinkage unit is:

[0037]

[0038] Where E represents the soft threshold function;

[0039] x represents feature data;

[0040] τ represents the threshold parameter.

[0041] Optionally, calculating the independent threshold for each channel using global average pooling and a fully connected layer includes the following steps:

[0042] S21231. For each channel's feature map, calculate the average value of all elements, simplify the multi-dimensional data of each channel into a single value, representing the global characteristics of the channel;

[0043] S21232, input the value obtained by global average pooling into the fully connected layer, and assign an independent threshold to each channel through the learning mechanism of the neural network;

[0044] S21233, using an activation function to normalize the output of the fully connected layer, and adjusting the independent threshold to a preset range;

[0045] S21234. Apply independent thresholds to the feature data of each channel, filter the feature data according to the threshold value, retain the feature data that meets the threshold, and eliminate the feature data that does not meet the threshold.

[0046] Optionally, training a performance prediction model using a focus loss function, and predicting the operating state of a nail art pen DC brushless driver using the trained performance prediction model includes the following steps:

[0047] S231, selecting a focal loss function as a loss function;

[0048] S232, training the performance prediction model using the focus loss function and the optimization algorithm, and setting the learning rate and the number of iterations;

[0049] S233. During training, regularly evaluate the performance of the performance prediction model using the validation dataset and monitor the generalization ability of the model on unseen data;

[0050] S234. After the training is completed, the performance prediction model is applied to predict the operating status of the nail pen DC brushless driver, and the prediction result is output by inputting the real-time operating data into the performance prediction model.

[0051] Optionally, predicting the operating state within a preset future time according to the output of the performance prediction model, calculating the pulse width modulation duty cycle and the bus voltage, and adjusting the pulse width modulation duty cycle and the bus voltage to meet the preset target operating state based on the calculation results includes the following steps:

[0052] S31, based on the model predictive control strategy, using the operating state prediction obtained by the performance prediction model to calculate the pulse width modulation duty cycle and bus voltage required to achieve the expected performance;

[0053] S32, sending a control signal through a control interface using the calculated pulse width modulation duty cycle and bus voltage parameters, and dynamically adjusting the operating parameters of the driver;

[0054] S33. After the working parameter adjustment is implemented, continuously monitor the actual operating state of the driver and compare the difference between the actual operating state and the predicted state;

[0055] S34. Record the actual operating data after adjustment, and verify the performance of the driver after each adjustment of the working parameters through data analysis.

[0056] Optionally, based on the model predictive control strategy, using the operating state prediction obtained by the performance prediction model, calculating the pulse width modulation duty cycle and bus voltage required to achieve the expected performance includes the following steps:

[0057] S311. Setting a specific objective function, determining decision variables, and setting constraints based on the future operating state output by the performance prediction model;

[0058] S312, calculating the pulse width modulation duty cycle and bus voltage that achieve the optimal objective function value under given constraints through a model predictive control strategy;

[0059] S313, organizing the calculated optimal pulse width modulation duty cycle and bus voltage parameters into a control signal, sending it through the control interface of the driver, and adjusting the operating state of the driver to the predicted state.

[0060] Optionally, recording the actual operating data after adjustment and verifying the performance of the driver after each adjustment of the working parameters through data analysis includes the following steps:

[0061] S341, obtaining the operating data of the driver after parameter adjustment;

[0062] S342, using the same measurement parameters and conditions, recording the operating data of the driver before adjustment;

[0063] S343, analyzing the difference between the operating data of the driver after the adjustment and the operating data of the driver before the adjustment, and evaluating the effect of the adjustment of the working parameters;

[0064] S344. Calculate the percentage of performance improvement and analyze the performance indicators, and regularly evaluate and adjust control parameters based on new performance data.

[0065] Compared with the prior art, this application has the following beneficial effects:

[0066] 1. The present invention improves overall performance and operational accuracy by continuously monitoring key performance parameters and processing data in real time. Real-time data analysis and frequency domain analysis help to promptly detect potential mechanical failures and electrical problems, such as motor imbalance and bearing damage, thereby achieving preventive maintenance and reducing the risk of sudden failures. By analyzing the collected data and identifying the changing trends of equipment performance, maintenance and replacement plans can be arranged more intelligently, reducing unnecessary maintenance costs and equipment downtime.

[0067] 2. The present invention uses a soft threshold function in the residual shrinkage unit, so that the model can automatically screen out key feature data and suppress unimportant noise and irrelevant information. The refined feature processing not only improves the quality of the data, but also enhances the prediction accuracy and robustness of the subsequent model. The focus loss function helps the model achieve better classification results on class-imbalanced data sets by paying more attention to samples that are difficult to classify and reducing the weight of samples that are easy to classify. This is particularly important in practical applications because it can ensure that the model remains highly sensitive to all potential operating anomalies, thereby more effectively preventing failures. By optimizing the algorithm and configuring training parameters (such as learning rate and number of iterations), the model can not only effectively learn complex data patterns during the training process, but also show good generalization ability for unseen data in practical applications, and can accurately predict the drive status under various operating conditions, thereby improving the reliability and safety of the equipment. The trained model can predict the operating status of the drive in real time, detect possible failures and performance degradation in a timely manner, help to achieve preventive maintenance, optimize maintenance plans, and reduce unnecessary maintenance costs and equipment downtime.

[0068] 3. The present invention uses a model predictive control strategy to accurately calculate the required pulse width modulation duty cycle and bus voltage according to the predicted future operating state, ensuring that the driver operates under the optimal operating conditions, thereby improving performance and efficiency. Correct adjustment of the pulse width modulation duty cycle and bus voltage helps to optimize energy consumption and mechanical efficiency. By maintaining the driver in the optimal operating state, energy consumption can be significantly reduced and overall performance can be improved. Model predictive control allows rapid response to external changes and internal state changes, and dynamically adjusts control parameters to meet immediate operational needs. By continuously monitoring the actual operating state after adjustment and comparing it with the predicted state, any deviation or deficiency can be discovered and corrected in a timely manner to ensure that the equipment continues to operate in the optimal state. BRIEF DESCRIPTION OF THE DRAWINGS

[0069] The above characteristics, features and advantages of the present invention and their implementation methods and methods will become more clearly understood in conjunction with the following description of the embodiments, which are described in detail in conjunction with the accompanying drawings. Herein, a schematic diagram is shown:

[0070] Figure 1 The present invention is a flowchart of a method for controlling the operating speed of a DC brushless driver of a nail art pen based on artificial intelligence according to an embodiment of the present invention. DETAILED DESCRIPTION

[0071] In order to enable those skilled in the art to better understand the solution of the present application, the technical solution in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0072] According to an embodiment of the present invention, a method for controlling the operating speed of a nail art pen DC brushless driver based on artificial intelligence is provided.

[0073] The present invention is further described with reference to the accompanying drawings and specific embodiments. Figure 1 As shown, according to the artificial intelligence-based nail art pen DC brushless driver operating speed control method of the embodiment of the present invention, the nail art pen DC brushless driver operating speed control method comprises the following steps:

[0074] S1. Obtaining the operating data of the nail art pen DC brushless driver, preprocessing the acquired operating data, and extracting feature data from the preprocessed operating data.

[0075] Preferably, obtaining the operating data of the nail art pen DC brushless driver, preprocessing the obtained operating data, and extracting characteristic data from the preprocessed operating data includes the following steps:

[0076] S11. Use pre-configured sensors to collect operation data of the nail art pen DC brushless driver in real time.

[0077] S12. Use a digital filter to remove high-frequency noise in the operating data and perform standardization on the operating data.

[0078] S13, converting the normalized operating data into frequency domain data, and using fast Fourier transform to obtain frequency components and corresponding amplitudes.

[0079] It should be explained that pre-configured sensors, such as current sensors, voltage sensors, and temperature sensors, are used to monitor and record the operating status of the nail pen DC brushless driver in real time. These sensors are connected to a data acquisition terminal (such as a DAQ system) that can continuously record data at a high sampling rate, and the selection of sensors depends on the type of parameters that need to be monitored and the accuracy requirements. For example, current and voltage sensors are used to capture the electrical performance of the motor, while temperature sensors help monitor the overheating of the device.

[0080] Use digital filters, such as low-pass filters, to remove high-frequency noise from the collected data, thereby reducing interference and improving data quality. After that, the data is standardized and scaled to a uniform level. Common methods include maximum and minimum normalization or Z-score normalization. Digital filters set a cutoff frequency to limit the noise part of the signal that is higher than this frequency. Low-pass filters are particularly suitable for removing electrical noise. Standardization is a common step in data preprocessing, which helps subsequent algorithms process data more effectively.

[0081] The pre-processed time series data is converted into frequency domain data, usually using Fast Fourier Transform (FFT). FFT is an algorithm that can decompose time series data into components of different frequencies and provide amplitude and phase information of each component. Frequency domain analysis is an important part of signal processing, especially when analyzing periodic fluctuations or vibration data. Through FFT, the main frequency components in the data can be identified, which is of great value for diagnosing mechanical failures, electrical problems or performance evaluation.

[0082] S14. Based on the frequency domain data, analyze the harmonics in the spectrum, calculate the power spectrum density, evaluate the power distribution of the signal at each frequency, and identify the frequency components and peaks in the signal.

[0083] Preferably, based on the frequency domain data, the harmonics in the spectrum are analyzed, the power spectrum density is calculated, and the power distribution of the signal at each frequency is evaluated; identifying the frequency components and peaks in the signal comprises the following steps:

[0084] S141, using the result of fast Fourier transform, identifying the main harmonics and subharmonics of the signal spectrum from the frequency domain data;

[0085] S142, based on identifying the main harmonics and subharmonics of the signal spectrum from the frequency domain data, calculating the power spectrum density of the entire frequency domain data, and evaluating the power distribution of the signal at all frequency points;

[0086] S143, based on power spectrum density analysis, identifying frequency points where power is concentrated;

[0087] S144. Evaluate the power distribution of the signal at each frequency point, and identify the frequency components and peaks in the signal.

[0088] S15. Analyze each frequency component according to the obtained frequency domain data, and extract characteristic data of the nail art pen DC brushless driver performance.

[0089] It should be explained that the results of the fast Fourier transform (FFT) are used to analyze the significant peaks in the spectrum diagram. These peaks usually represent the main and secondary harmonics in the operation of the motor. Identifying these harmonics helps to understand the dynamics of the motor operation and potential imbalance or faults. Harmonic analysis is an important tool for evaluating the health of the motor. The main harmonics usually correspond to the fundamental operating frequency of the motor, while the secondary harmonics may indicate mechanical imbalance, electrical faults or other problems. Based on the FFT results, the formula PSD(f)=|X(f)| 2 Calculate the power spectral density of each frequency point, where X(f) represents the amplitude of the frequency domain data. This step provides the contribution of different frequency components to the total power of the signal. The power spectral density is the power distribution of the signal in the frequency domain. It can reveal which frequency components are the most energy-concentrated. It is particularly useful for discovering abnormal vibrations or noise. From the power spectral density analysis, frequency points with concentrated power are identified. These points may be caused by specific mechanical activities or electrical events. The identification of these frequency points helps detect specific failure modes or performance issues, such as the increase in energy of specific frequencies caused by bearing damage or electrical failure. Analyze the power distribution of the entire spectrum and identify the There are significant frequency components and peaks, which help to fully understand the operating status and performance of the motor. The evaluation of the entire power distribution can reveal the overall health and performance efficiency of the motor and provide signals about the need for maintenance or adjustment. According to the above analysis results, combined with the design and operation requirements of the motor, the frequency components directly related to the motor performance are selected as feature data. The feature data includes the amplitude of specific harmonics, the peak in the power spectral density, etc. Which features are selected for tracking and analysis depends on the performance indicators of concern, such as efficiency, durability or early fault diagnosis. It is a key link to convert raw data into information that has practical significance for the health and performance of the drive.

[0090] S2. Use the deep residual shrinkage network to build a performance prediction model of the nail pen DC brushless driver based on the feature data, and use the performance prediction model to predict the operating status of the nail pen DC brushless driver.

[0091] Using the deep residual shrinkage network, a performance prediction model of the nail pen DC brushless driver is constructed according to the feature data, and the operating state of the nail pen DC brushless driver is predicted by using the performance prediction model, including the following steps:

[0092] S21. Utilize the residual shrinkage unit in the deep residual shrinkage network to automatically filter and optimize feature data through a soft threshold function.

[0093] Preferably, using the residual shrinkage unit in the deep residual shrinkage network to automatically screen and optimize feature data through a soft threshold function includes the following steps:

[0094] S211, sending the extracted feature data to the residual shrinkage unit in the deep residual shrinkage network;

[0095] S212, automatically identifying feature data that does not meet preset requirements using a soft threshold function inside the residual shrinkage unit;

[0096] S213. Set the feature data that does not meet the preset requirements to zero, retain the feature data that meets the preset requirements, and construct a feature data set.

[0097] It should be explained that the feature data obtained from the steps of S1 (such as FFT analysis and power spectral density calculation) are used as input and sent to the deep residual shrinkage network. The feature data includes the amplitude of the frequency component, the harmonic intensity, etc. The deep residual shrinkage network has multiple layers of residual shrinkage units, each of which contains some standard deep learning layers (such as convolutional layers) and unique shrinkage units for processing and optimizing input features. This type of network solves the degradation problem in deep networks by introducing a residual learning framework, so that the network can improve its performance as the depth increases. A shrinkage layer is added to the standard residual structure, which is usually implemented by a soft threshold function. To enhance the sparsity and denoising capabilities of the network, in the residual shrinkage unit, a soft threshold function is applied to the feature data. This function evaluates each element in the data. If the absolute value of an element is less than the preset threshold, the element is set to zero, which helps to reduce unnecessary information and noise. This step automatically removes features that do not meet the preset conditions, such as excessive noise or low-contribution frequency components. The soft threshold function is a sparse representation technique commonly used in signal and image processing, and is particularly suitable for removing noise while retaining important signal features. After processing by the residual shrinkage unit, all non-zero feature data are retained and integrated into a new feature data set. The data is now more streamlined and targeted, containing only the features that contribute most to the drive performance analysis. This screened and optimized feature data set can be used for further machine learning model training or performance evaluation tasks.

[0098] Preferably, using the soft threshold function inside the residual shrinkage unit to automatically identify feature data that does not meet the preset requirements includes the following steps:

[0099] S2121, embedding a residual shrinkage unit in the neural network, and selecting a threshold type according to a preset application requirement;

[0100] S2122, processing the feature data of each channel using a soft threshold function in the residual shrinkage unit, and calculating the absolute value of each feature;

[0101] S2123, using global average pooling and fully connected layers to calculate independent thresholds for each channel;

[0102] S2124, performing feature selection according to the calculated independent threshold and the absolute value of each feature;

[0103] Among them, the expression for processing the feature data of each channel using the soft threshold function in the residual shrinkage unit is:

[0104]

[0105] Where E represents the soft threshold function;

[0106] x represents feature data;

[0107] τ represents the threshold parameter.

[0108] Preferably, calculating the independent threshold of each channel using global average pooling and a fully connected layer comprises the following steps:

[0109] S21231. For each channel's feature map, calculate the average value of all elements, simplify the multi-dimensional data of each channel into a single value, representing the global characteristics of the channel;

[0110] S21232, input the value obtained by global average pooling into the fully connected layer, and assign an independent threshold to each channel through the learning mechanism of the neural network;

[0111] S21233, using an activation function to normalize the output of the fully connected layer, and adjusting the independent threshold to a preset range;

[0112] S21234. Apply independent thresholds to the feature data of each channel, filter the feature data according to the threshold value, retain the feature data that meets the threshold, and eliminate the feature data that does not meet the threshold.

[0113] It needs to be explained that the residual shrinkage unit is integrated into the deep neural network. These units can be embedded in specific layers of the network. According to the preset application requirements, the appropriate threshold type is selected, such as fixed threshold, dynamically adjusted threshold, etc. Residual shrinkage unit: The residual shrinkage unit increases the network's sparse representation ability of features through soft threshold processing, which helps to suppress unimportant features and enhance key information. The choice of threshold directly affects the feature screening effect. Different types of thresholds are applied to different scenarios and data characteristics. In the residual shrinkage unit, the soft threshold function is applied to process the feature data of each channel, and the absolute value of the processed data is calculated to facilitate the next step of threshold comparison. By setting the threshold, the value less than this threshold is reduced. The influence of the feature of the value is usually set to zero to achieve feature sparsification. The feature map of each channel is globally averaged pooled and simplified into a single value. This value represents the average feature strength of the channel. The result of average pooling is sent to the fully connected layer. Through network learning, an independent threshold is calculated for each channel. Global average pooling is a dimensionality reduction technique. The global information of the channel is obtained by calculating the average value of all elements in the feature map. The fully connected layer is usually used in the decision-making stage in deep learning. Here it is used to adjust the threshold based on global information; the corresponding independent threshold is applied to the feature data of each channel, and the absolute value of each feature is compared with its threshold. Features greater than the threshold are retained, and features less than the threshold are eliminated.

[0114] S22. Based on the optimized and selected key features, a performance prediction model for the DC brushless driver of a nail art pen is constructed.

[0115] S23. Use the focus loss function to train a performance prediction model, and use the trained performance prediction model to predict the operating status of the nail pen DC brushless driver.

[0116] Preferably, using the focus loss function to train the performance prediction model, and using the trained performance prediction model to predict the operating state of the nail art pen DC brushless driver includes the following steps:

[0117] S231, selecting a focal loss function as a loss function;

[0118] S232, training the performance prediction model using the focus loss function and the optimization algorithm, and setting the learning rate and the number of iterations;

[0119] S233. During training, regularly evaluate the performance of the performance prediction model using the validation dataset and monitor the generalization ability of the model on unseen data;

[0120] S234. After the training is completed, the performance prediction model is applied to predict the operating status of the nail pen DC brushless driver, and the prediction result is output by inputting the real-time operating data into the performance prediction model.

[0121] It should be explained that the key feature data screened from the residual shrinkage network is used to build a model for predicting the performance of the DC brushless drive, including the frequency response, power spectrum density characteristics, and other key indicators of the motor. The appropriate machine learning algorithm (such as support vector machine, random forest or deep learning model) is selected to build the model according to the complexity of the problem and the characteristics of the data. The focal loss function (Focal Loss) is particularly suitable for processing datasets with imbalanced categories. It improves the recognition ability of the model by increasing the focus on samples that are difficult to classify and reducing the focus on samples that are easy to classify. The focus loss function is an improved cross entropy loss. By adjusting the weight coefficient in the loss function, the model is more focused on samples that are difficult to classify. Configure training parameters such as learning rate and number of iterations, and use optimization algorithms such as Adam or SGD for model training. During the training process, the focus loss function continuously adjusts the loss weight to help the model effectively learn complex patterns during the training process. The learning rate is a key parameter that affects the speed and effect of model training, and the number of iterations determines the depth and breadth of model training. During training, the validation dataset is regularly used to test the performance of the model, which helps monitor the generalization ability of the model on unseen data and evaluates whether the model is overfitting or underfitting. After training is completed, the real-time operation data is input into the performance prediction model, and the model will output the operating status of the drive. The model can help predict the status before a failure occurs, thereby achieving preventive maintenance.

[0122] In addition, the calculation step of global average pooling of feature maps of each channel is a simple but critical process. The main purpose is to simplify the multi-dimensional data of each channel into a single value to represent the global characteristics of the channel. For each channel, the average value of all pixel values ​​is calculated. If the dimension of a channel is m×n, the global average value Avg(C) of the channel is given by the following formula:

[0123]

[0124] Where Avg(C) represents the global average value of channel C;

[0125] C ij Represents the pixel value in row i and column j;

[0126] m represents the number of rows of channel C;

[0127] n represents the column number of channel C.

[0128] S3. Predict the operating state within a preset future time according to the output of the performance prediction model, calculate the pulse width modulation duty cycle and the bus voltage, and adjust the pulse width modulation duty cycle and the bus voltage to meet the preset target operating state based on the calculation results.

[0129] Preferably, predicting the operating state within a preset future time according to the output of the performance prediction model, calculating the pulse width modulation duty cycle and the bus voltage, and adjusting the pulse width modulation duty cycle and the bus voltage to meet the preset target operating state based on the calculation results includes the following steps:

[0130] S31. Based on the model predictive control strategy, the pulse width modulation duty cycle and bus voltage required to achieve the expected performance are calculated using the operating state prediction obtained by the performance prediction model.

[0131] Preferably, based on the model predictive control strategy, using the operating state prediction obtained by the performance prediction model, calculating the pulse width modulation duty cycle and bus voltage required to achieve the expected performance includes the following steps:

[0132] S311. Setting a specific objective function, determining decision variables, and setting constraints based on the future operating state output by the performance prediction model;

[0133] S312, calculating the pulse width modulation duty cycle and bus voltage that achieve the optimal objective function value under given constraints through a model predictive control strategy;

[0134] S313, organizing the calculated optimal pulse width modulation duty cycle and bus voltage parameters into a control signal, sending it through the control interface of the driver, and adjusting the operating state of the driver to the predicted state.

[0135] It should be explained that, based on the operating status prediction output by the performance prediction model, a specific objective function is set, such as minimizing energy consumption, maximizing efficiency or other performance indicators, and the decision variables are the control parameters that need to be optimized, such as pulse width modulation (PWM) duty cycle and bus voltage. Reasonable constraints are set, including the voltage and current limits for safe operation of the equipment, the range of the duty cycle, etc.

[0136] Model Predictive Control (MPC): MPC is an advanced process control method that uses models to predict future behavior and optimizes future control signals to achieve the best performance.

[0137] The objective function and constraints are the core components of the optimization problem, which determine the search direction of the optimization algorithm and the feasibility of the solution.

[0138] Using the MPC strategy, combined with the set objective function and constraints, the pulse width modulation duty cycle and bus voltage that make the objective function optimal are calculated through numerical optimization methods (such as linear programming, nonlinear programming, etc.), and the pulse width modulation duty cycle and bus voltage obtained by the optimized calculation are converted into actual electronic control signals, which are sent through the control interface of the drive to adjust the operating state of the drive to the predicted state. The control signal usually needs to be sent through a specific hardware interface, such as through a digital-to-analog converter (DAC) or through a specific industrial communication protocol.

[0139] S32, sending the calculated pulse width modulation duty cycle and bus voltage parameters as a control signal through a control interface, and dynamically adjusting the operating parameters of the driver.

[0140] S33. After the working parameter adjustment is implemented, the actual operating status of the driver is continuously monitored, and the difference between the actual operating status and the predicted status is compared.

[0141] S34. Record the actual operating data after adjustment, and verify the performance of the driver after each adjustment of the working parameters through data analysis.

[0142] Preferably, recording the actual operating data after adjustment and verifying the performance of the driver after each adjustment of the working parameters through data analysis includes the following steps:

[0143] S341, obtaining the operating data of the driver after parameter adjustment;

[0144] S342, using the same measurement parameters and conditions, recording the operating data of the driver before adjustment;

[0145] S343, analyzing the difference between the operating data of the driver after the adjustment and the operating data of the driver before the adjustment, and evaluating the effect of the adjustment of the working parameters;

[0146] S344. Calculate the percentage of performance improvement and analyze the performance indicators, and regularly evaluate and adjust control parameters based on new performance data.

[0147] It should be explained that the pulse width modulation duty cycle and bus voltage parameters calculated by the model predictive control are converted into control signals, and these signals are sent through the control interface of the electric drive. According to the received control signals, the control of the drive dynamically adjusts the working parameters to achieve the preset performance goals. After the parameters are adjusted, the operating status of the drive is monitored in real time, including key indicators such as power output, speed, and temperature. The actual operating status monitored is compared with the status predicted by the model to identify any deviations or unmet performance indicators. A data acquisition terminal (such as a SCADA system) is used to collect operating data from the adjusted drive. Before adjusting the parameters, the operating data of the drive is also recorded using data acquisition as a baseline for subsequent comparisons. Statistical and data analysis methods (such as mean comparison, variance analysis, etc.) are used to analyze the differences in data before and after the adjustment. Based on the data analysis results, the specific impact of the parameter adjustment on the drive performance is evaluated, and the percentage of performance improvement is calculated, such as power increase, efficiency improvement, etc. Based on the performance improvement results, the control parameters are regularly evaluated and adjusted to continuously optimize the performance of the drive.

[0148] S4. Real-time monitoring of the actual operating status of the adjusted nail art pen DC brushless driver. If the performance is detected to be lower than the preset threshold, an alarm is triggered immediately, and the new operating data is fed back to the performance prediction model to optimize the parameters of the performance prediction model.

[0149] It should be explained that integrating sensors and data acquisition devices in the drive control to collect data about speed, power output, temperature, etc. in real time, using real-time data transmission technology (such as wireless networks or wired connections) to ensure that data is transmitted to the monitoring center or directly to the processing unit in real time, and continuously monitoring the operating status of the equipment is the key to ensuring that the equipment operates in the best condition. Through real-time monitoring, any signs of performance degradation or failure can be responded to quickly. According to the output of historical performance data and prediction models, specific performance thresholds are set. Once the monitored data is lower than these preset thresholds, alarms are automatically triggered to notify the maintenance team or automatically execute preset response measures. In industrial automation, alarms are an important part of preventive maintenance, which can reduce downtime and maintenance costs. The operating data collected from real-time monitoring is fed back to the performance prediction model regularly or as needed, and the new data is used to optimize or retrain the performance prediction model, and the model parameters are updated to more accurately reflect the current operating status and performance of the equipment.

[0150] Specific examples are as follows:

[0151] Suppose a company is designing high-precision nail art equipment that needs to precisely control the rotation speed of the drill to accommodate the processing needs of different materials. The DC brushless drive of the nail art pen is responsible for driving the rotation mechanism and needs to be precisely adjusted to ensure optimal performance and user safety.

[0152] The implementation steps and specific applications are as follows:

[0153] 1) Data acquisition and preprocessing

[0154] Scenario: During a typical workday, use pre-configured sensors to monitor the drive's current, voltage, and temperature in real time to obtain operational data;

[0155] Operation: The data are first passed through a digital filter to remove noise and normalize for subsequent analysis;

[0156] 2) Performance prediction model construction

[0157] Scenario: Using the collected feature data, such as the frequency components of current and voltage, a performance prediction model based on a deep residual shrinkage network is established;

[0158] Operation: The model predicts the performance of the drive under different operating conditions by automatically screening and optimizing input features;

[0159] 3) Model predictive control

[0160] Scenario: In a specific design task, such as fine grinding for a complex nail design, the required pulse width modulation duty cycle and bus voltage are calculated based on the output of the performance prediction model;

[0161] Operation: The calculated parameters are sent to the driver through the control interface, and the rotation speed is adjusted in real time to meet the needs of fine grinding;

[0162] 4) Real-time monitoring and feedback optimization

[0163] Scenario: After adjusting the parameters, the operating status of the equipment is monitored in real time. If the performance is detected to be lower than the preset threshold, an alarm will be triggered;

[0164] Operation: Abnormal data is fed back to the performance prediction model to optimize model parameters to improve future prediction accuracy and control effects.

[0165] Therefore, through this integrated approach, the DC brushless drive of the nail pen can achieve more precise speed control and improve the operating accuracy and safety of the equipment. At the same time, continuous data monitoring and model optimization can significantly improve the reliability and maintenance efficiency of the equipment and reduce maintenance costs.

[0166] In summary, with the help of the above technical solutions of the present invention, the present invention improves the overall performance and operation accuracy by continuously monitoring key performance parameters and processing data in real time. Real-time data analysis and frequency domain analysis help to timely discover potential mechanical failures and electrical problems, such as motor imbalance and bearing damage, so as to achieve preventive maintenance and reduce the risk of sudden failures. By analyzing the collected data and identifying the changing trend of equipment performance, maintenance and replacement plans can be arranged more intelligently to reduce unnecessary maintenance costs and equipment downtime. The present invention uses the soft threshold function in the residual shrinkage unit, and the model can automatically filter out key feature data and suppress unimportant noise and irrelevant information. The fine feature processing not only improves the quality of the data, but also enhances the prediction accuracy and robustness of the subsequent model. The focus loss function helps the model achieve better classification results on unbalanced data sets by paying more attention to samples that are difficult to classify and reducing the weight of samples that are easy to classify. It is particularly important in practical applications because it can ensure that the model is highly sensitive to all potential operating anomalies, thereby more effectively preventing failures. By optimizing the algorithm and the configured training parameters (such as learning rate and number of iterations), the model The model can not only effectively learn complex data patterns during the training process, but also show good generalization ability for unseen data in practical applications, and can accurately predict the driver status under various operating conditions, thereby improving the reliability and safety of the equipment; the model after training can predict the operating status of the driver in real time, timely discover possible faults and performance degradation, help to achieve preventive maintenance, optimize maintenance plans, and reduce unnecessary maintenance costs and equipment downtime; the present invention can accurately calculate the required pulse width modulation duty cycle and bus voltage according to the predicted future operating status through the model predictive control strategy, ensure that the driver works according to the optimal operating conditions, thereby improving performance and efficiency, and the correct adjustment of the pulse width modulation duty cycle and bus voltage helps to optimize energy consumption and mechanical efficiency. By maintaining the driver in the optimal operating state, energy consumption can be significantly reduced and overall performance can be improved. Model predictive control allows rapid response to external changes and internal state changes, and dynamically adjusts control parameters to meet immediate operational needs. By continuously monitoring the adjusted actual operating state and comparing it with the predicted state, any deviation or deficiency can be discovered and corrected in time to ensure that the equipment continues to operate in the optimal state.

[0167] Although the present invention has been disclosed as above with preferred embodiments, the embodiments are merely examples for the convenience of description and are not intended to limit the present invention. Those skilled in the art may make several changes and modifications without departing from the spirit and scope of the present invention. The scope of protection claimed by the present invention shall be based on the claims.

Claims

1. A method for controlling the operating speed of a DC brushless driver for a nail art pen based on artificial intelligence, characterized in that: The following steps are involved: S1. Obtaining operation data of a nail art pen DC brushless driver, preprocessing the acquired operation data, and extracting feature data from the preprocessed operation data; S2. Using the deep residual shrinkage network, a performance prediction model of the nail pen DC brushless driver is constructed according to the feature data, and the operating state of the nail pen DC brushless driver is predicted using the performance prediction model; S3. Predicting the operating state within a preset time in the future according to the output of the performance prediction model, calculating the pulse width modulation duty cycle and the bus voltage, and adjusting the pulse width modulation duty cycle and the bus voltage to meet the preset target operating state based on the calculation results; S4, real-time monitoring of the actual operating status of the adjusted nail art pen DC brushless driver, if it is detected that the performance is lower than the preset threshold, an alarm is triggered, and the new operating data is fed back to the performance prediction model to optimize the parameters of the performance prediction model; The method of using a deep residual shrinkage network to construct a performance prediction model of a nail pen DC brushless driver according to feature data, and using the performance prediction model to predict the operating state of the nail pen DC brushless driver includes the following steps: S21, using the residual shrinkage unit in the deep residual shrinkage network to automatically filter and optimize feature data through a soft threshold function; S22. Based on the optimized and selected key features, a performance prediction model of the DC brushless driver for nail art pens is constructed; S23. Use the focus loss function to train a performance prediction model, and use the trained performance prediction model to predict the operating status of the nail pen DC brushless driver.

2. The method for controlling the operating speed of a DC brushless driver for a nail art pen based on artificial intelligence according to claim 1 is characterized in that: The method of obtaining the operating data of the nail art pen DC brushless driver, preprocessing the obtained operating data, and extracting feature data from the preprocessed operating data includes the following steps: S11, using a pre-configured sensor to collect operation data of a nail art pen DC brushless driver in real time; S12, using a digital filter to remove high-frequency noise in the operating data, and performing standardization processing on the operating data; S13, converting the normalized operating data into frequency domain data, and obtaining frequency components and corresponding amplitudes by fast Fourier transform; S14. Based on the frequency domain data, analyze the harmonics in the spectrum, calculate the power spectrum density, evaluate the power distribution of the signal at each frequency, and identify the frequency components and peaks in the signal; S15. Analyze each frequency component according to the obtained frequency domain data, and extract characteristic data of the nail art pen DC brushless driver performance.

3. The method for controlling the operating speed of a DC brushless driver for a nail art pen based on artificial intelligence according to claim 2 is characterized in that: The method of analyzing harmonics in the spectrum, calculating power spectrum density, evaluating the power distribution of the signal at each frequency, and identifying frequency components and peaks in the signal based on frequency domain data includes the following steps: S141, using the result of fast Fourier transform, identifying the main harmonics and subharmonics of the signal spectrum from the frequency domain data; S142, based on identifying the main harmonics and subharmonics of the signal spectrum from the frequency domain data, calculating the power spectrum density of the entire frequency domain data, and evaluating the power distribution of the signal at all frequency points; S143, based on power spectrum density analysis, identifying frequency points where power is concentrated; S144. Evaluate the power distribution of the signal at each frequency point, and identify the frequency components and peaks in the signal.

4. The method for controlling the operating speed of a DC brushless driver for a nail art pen based on artificial intelligence according to claim 1 is characterized in that: The method of using the residual shrinkage unit in the deep residual shrinkage network to automatically screen and optimize feature data through a soft threshold function includes the following steps: S211, sending the extracted feature data to the residual shrinkage unit in the deep residual shrinkage network; S212, automatically identifying feature data that does not meet preset requirements using a soft threshold function inside the residual shrinkage unit; S213. Set the feature data that does not meet the preset requirements to zero, retain the feature data that meets the preset requirements, and construct a feature data set.

5. The method for controlling the operating speed of a DC brushless driver for a nail art pen based on artificial intelligence according to claim 4 is characterized in that: The method of automatically identifying feature data that does not meet preset requirements by using the soft threshold function inside the residual shrinkage unit comprises the following steps: S2121, embedding a residual shrinkage unit in the neural network, and selecting a threshold type according to a preset application requirement; S2122, processing the feature data of each channel using a soft threshold function in the residual shrinkage unit, and calculating the absolute value of each feature; S2123, using global average pooling and fully connected layers to calculate independent thresholds for each channel; S2124, performing feature selection according to the calculated independent threshold and the absolute value of each feature; The expression for processing the feature data of each channel using the soft threshold function in the residual shrinkage unit is: Where E represents the soft threshold function; χ represents characteristic data; τ represents the threshold parameter.

6. The method for controlling the operating speed of a DC brushless driver for a nail art pen based on artificial intelligence according to claim 5 is characterized in that: The method of calculating the independent threshold of each channel using global average pooling and a fully connected layer includes the following steps: S21231. For each channel's feature map, calculate the average value of all elements, simplify the multi-dimensional data of each channel into a single value, representing the global characteristics of the channel; S21232, input the value obtained by global average pooling into the fully connected layer, and assign an independent threshold to each channel through the learning mechanism of the neural network; S21233, using an activation function to normalize the output of the fully connected layer, and adjusting the independent threshold to a preset range; S21234. Apply independent thresholds to the feature data of each channel, filter the feature data according to the threshold value, retain the feature data that meets the threshold, and eliminate the feature data that does not meet the threshold.

7. The method for controlling the operating speed of a DC brushless driver for a nail art pen based on artificial intelligence according to claim 6 is characterized in that: The method of training the performance prediction model by using the focus loss function and predicting the operating state of the nail art pen DC brushless driver by using the trained performance prediction model comprises the following steps: S231, selecting a focal loss function as a loss function; S232, training the performance prediction model using the focus loss function and the optimization algorithm, and setting the learning rate and the number of iterations; S233. During training, regularly evaluate the performance of the performance prediction model using the validation dataset and monitor the generalization ability of the model on unseen data; S234. After the training is completed, the performance prediction model is applied to predict the operating status of the nail pen DC brushless driver, and the prediction result is output by inputting the real-time operating data into the performance prediction model.

8. The method for controlling the operating speed of a DC brushless driver for a nail art pen based on artificial intelligence according to claim 1, characterized in that: The method of predicting the operating state within a preset time in the future according to the output of the performance prediction model, calculating the pulse width modulation duty cycle and the bus voltage, and adjusting the pulse width modulation duty cycle and the bus voltage to meet the preset target operating state based on the calculation results includes the following steps: S31, based on the model predictive control strategy, using the operating state prediction obtained by the performance prediction model to calculate the pulse width modulation duty cycle and bus voltage required to achieve the expected performance; S32, sending a control signal through a control interface using the calculated pulse width modulation duty cycle and bus voltage parameters, and dynamically adjusting the operating parameters of the driver; S33. After the working parameter adjustment is implemented, continuously monitor the actual operating state of the driver and compare the difference between the actual operating state and the predicted state; S34. Record the actual operating data after adjustment, and verify the performance of the driver after each adjustment of the working parameters through data analysis.

9. The method for controlling the operating speed of a DC brushless driver for a nail art pen based on artificial intelligence according to claim 8 is characterized in that: The method of calculating the pulse width modulation duty cycle and bus voltage required to achieve the expected performance based on the model predictive control strategy and using the operating state prediction obtained by the performance prediction model includes the following steps: S311. Setting a specific objective function, determining decision variables, and setting constraints based on the future operating state output by the performance prediction model; S312, calculating the pulse width modulation duty cycle and bus voltage that achieve the optimal objective function value under given constraints through a model predictive control strategy; S313, organizing the calculated optimal pulse width modulation duty cycle and bus voltage parameters into a control signal, sending it through the control interface of the driver, and adjusting the operating state of the driver to the predicted state.

10. The method for controlling the operating speed of a DC brushless driver for a nail art pen based on artificial intelligence according to claim 9, characterized in that: The recording of the adjusted actual operating data and verifying the performance of the driver after each adjustment of the working parameters through data analysis include the following steps: S341, obtaining the operating data of the driver after parameter adjustment; S342, using the same measurement parameters and conditions, recording the operating data of the driver before adjustment; S343, analyzing the difference between the operating data of the driver after the adjustment and the operating data of the driver before the adjustment, and evaluating the effect of the adjustment of the working parameters; S344. Calculate the percentage of performance improvement and analyze the performance indicators, and regularly evaluate and adjust control parameters based on new performance data.

Images