Motor operation control method and system of a nanogrinder and storage medium

Abstract

This invention relates to the field of motor control technology, specifically to a motor operation control method, system, and storage medium for a nano-grinding machine. The method includes: acquiring an initial dataset, which includes a particle size data sequence showing the change in grinding particle size with grinding time under multiple control parameters; extracting a first sample set and a second sample set from the initial dataset based on the particle size data sequence; constructing a fused feature vector based on the first sample set and the second sample set; training a pre-constructed neural network model using the fused feature vector to obtain a trained model; acquiring the target particle size for grinding lithium battery cathode materials; inputting the target particle size into the trained model to obtain corresponding control parameters; and controlling the motor operation according to the control parameters corresponding to the particle size range. This invention can improve grinding efficiency and processing accuracy.

Description

Technical Field

[0001] This invention relates to the field of motor control technology, specifically to a motor operation control method, system, and storage medium for a nano-grinding machine. Background Technology

[0002] Cathode materials are the main source of lithium ions in lithium-ion batteries. Based on the particle size, lithium battery cathode material precursors can be broadly classified into small-particle precursors, medium-particle precursors, and large-particle precursors. To prepare lithium battery cathode materials with different particle sizes, it is necessary to use a motor with a variable frequency speed control system or multiple motors with different rated power, combined with stirrers with different stirring intensities, for grinding.

[0003] However, most existing motor control methods rely on manual parameter setting and adjustment based on experience, which leads to cumbersome operation and makes it difficult to ensure that the produced lithium battery cathode material can stably meet the particle size requirements. Summary of the Invention

[0004] The purpose of this invention is to provide a motor operation control method, system, and storage medium for a nano-grinding machine, which can automatically generate corresponding control parameters according to the particle size of the grinding particles, thereby improving grinding efficiency and processing accuracy.

[0005] To achieve the above objectives, the present invention provides the following technical solution:

[0006] In a first aspect, embodiments of the present invention provide a method for controlling the operation of a motor in a nano-grinding machine, the method comprising the following steps:

[0007] An initial dataset is obtained, which includes a particle size data sequence showing the change in grinding particle size with grinding time under multiple control parameters; wherein, the control parameters include the opening degree of the electric ball valve, the grinding power of the motor, and the grinding speed; the particle size data sequence includes multiple particle size distribution data sorted by time series.

[0008] Based on the particle size data sequence, a first sample set and a second sample set are extracted from the initial dataset; wherein, the grinding particle size changes faster with grinding time in the first sample set than in the second sample set.

[0009] A fusion feature vector is constructed based on the first sample set and the second sample set. The pre-constructed neural network model is then trained using the fusion feature vector to obtain a trained model.

[0010] The target particle size for grinding the lithium battery cathode material is obtained, and the target particle size is input into the trained model to obtain the corresponding control parameters. The motor is then controlled according to the control parameters corresponding to the particle size range.

[0011] Optionally, the step of extracting the first sample set and the second sample set from the initial dataset based on the granular data sequence includes:

[0012] The frequency distribution and standard deviation of granularity are determined based on the granularity distribution data in the initial dataset.

[0013] The initial dataset is divided into a positive sample dataset with a standard deviation below a standard deviation threshold and a negative sample dataset with a standard deviation above a standard deviation threshold;

[0014] For each particle size data sequence in the positive sample dataset, the change in the average particle size with the grinding time is used as the gradient. The particle size data sequences are arranged in ascending order of the gradient. The first few particle size data sequences and their corresponding control parameters are selected as the first sample set, and the last few particle size data sequences and their corresponding control parameters are selected as the second sample set. The average value is the interval with the largest frequency distribution.

[0015] Optionally, the step of constructing a fused feature vector based on the first sample set and the second sample set, and training a pre-built neural network model using the fused feature vector to obtain a trained model, includes:

[0016] The first sample set is input into the time feature extraction channel to extract time features, and the second sample set is input into the spatial feature extraction channel to extract spatial features.

[0017] The temporal features extracted by the temporal feature extraction channel and the spatial features extracted by the spatial feature extraction channel are concatenated in a weighted manner to obtain a fused feature vector;

[0018] The fused feature vectors are input into a pre-built neural network model for training. The loss value is calculated based on the pre-built loss function and backpropagated to the neural network model until the loss value of the neural network model reaches the set loss threshold, thus obtaining the trained model.

[0019] Optionally, the step of extracting time features from the first sample set via the time feature extraction channel includes:

[0020] The first sample set is used to form a first sample matrix with a first granularity. The first sample matrix includes d-dimensional control parameters of different types collected at n sampling times. Specifically, for each element in the first sample matrix... The superscript j represents the type of control parameter, and the subscript i represents the sampling time, i=1,2,...,n; j=1,2,...,d;

[0021] The first sample matrix is ​​input into the first bidirectional LSTM network to obtain the first expected feature vector.

[0022] Optionally, the step of extracting spatial features from the second sample set input spatial feature extraction channel includes:

[0023] The second sample set is used to form a second sample matrix and a corresponding second granularity. The second sample matrix includes d-dimensional control parameters of different types collected at m sampling times. Specifically, for each element in the second sample matrix... The superscript j represents the type of control parameter, and the subscript i represents the sampling time, i=1,2,...,n; j=1,2,...,d,m<n;

[0024] A set of one-dimensional convolutional kernels is used to slide in variable order and traverse all time periods in the granular data sequence, and a ReLU activation function is connected to perform initial feature extraction on the granular data sequence;

[0025] The extracted initial features are processed using a multi-scale convolution module to obtain multiple feature vectors. The multiple feature vectors are then fused using the ReLU activation function to obtain a second desired feature vector.

[0026] Optionally, the loss function is:

[0027] ;

[0028] Among them, L FU For the fusion loss function, L MSE Here, k is the weight of the fusion loss function, where 0 < k < 1;

[0029] ;

[0030] L FU =meanKL(softmax(X(i,j)),softmax(Y(i,j)));

[0031] Where X(i,j) represents the fused feature vector, Y(i,j) represents the feature vector predicted by the neural network model, KL() represents the KL divergence loss, mean() represents the average value, and softmax() represents the Softmax function.

[0032] Secondly, embodiments of the present invention provide a motor operation control system for a nano-grinding machine, the system comprising:

[0033] At least one processor;

[0034] At least one memory for storing at least one program;

[0035] When the at least one program is executed by the at least one processor, the at least one processor implements the motor operation control method for the nano-grinding machine as described in any of the preceding claims.

[0036] Thirdly, embodiments of the present invention provide a computer-readable storage medium storing a processor-executable program, characterized in that the processor-executable program, when executed by a processor, is used to perform the motor operation control method for a nano-grinding machine as described in any of the preceding claims.

[0037] The beneficial effects of this invention are as follows: This invention discloses a motor operation control method, system, and storage medium for a nano-grinding machine. By extracting a first sample set and a second sample set from an initial dataset, this invention can not only better distinguish the effect of control parameters on grinding particles, but also predict the optimal combination of control parameters based on the fused feature vector constructed from the first and second sample sets through processing efficiency. This invention can automatically generate corresponding control parameters according to the particle size of the grinding particles using a trained model, thereby improving grinding efficiency and processing accuracy. Attached Figure Description

[0038] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0039] Figure 1 This is a flowchart illustrating the motor operation control method of the nano-grinding machine in an embodiment of the present invention;

[0040] Figure 2 This is a schematic diagram of the motor operation control system of the nano-grinding machine in an embodiment of the present invention. Detailed Implementation

[0041] The following will provide a clear and complete description of the concept, specific structure, and technical effects of the present invention in conjunction with embodiments and accompanying drawings, so as to fully understand the purpose, solution, and effects of the present invention. It should be noted that, unless otherwise specified, the embodiments and features described in the embodiments of the present invention can be combined with each other.

[0042] See Figure 1 , Figure 1 This invention provides a motor operation control method for a nano-grinding machine, the method comprising the following steps:

[0043] S100, Obtain an initial dataset, which includes a particle size data sequence showing the change in grinding particle size with grinding time under multiple control parameters; wherein, the control parameters include the opening degree of the electric ball valve, the grinding power of the motor, and the grinding speed; the particle size data sequence includes multiple particle size distribution data sorted by time series;

[0044] Specifically, the control parameters and grinding particle size of the nano-grinding mill in multiple past grinding tasks are obtained. Grinding particles are collected at set time intervals, and the particle size of the collected grinding particles is measured using particle size testing methods such as dynamic light scattering to obtain particle size distribution data. It should be noted that in the nano-grinding mill, the feed inlet is the entry point for material input. The material to be ground is fed into the nano-grinding mill through the feed inlet for subsequent grinding and dispersion processing. An electric ball valve is used to control the feeding and discharging processes. The opening degree of the electric ball valve controls the grinding speed, adjusting the feeding and discharging speeds of the nano-grinding mill to meet the needs of different material grinding processes. Controlling the feeding speed controls the grinding volume, thus affecting the particle size. The higher the feeding speed, the smaller the crushing force acting on each particle, resulting in a smaller particle size. The motor is the power source of the nano-grinding mill, driving its rotation to achieve material grinding and dispersion. Precise adjustment and control of the motor can provide stable speed and power output. Higher grinding speeds can accelerate the grinding process but may also affect particle size uniformity. Higher grinding power results in smaller particle size, but may also affect particle size uniformity. Extending grinding time can further refine the particles, but excessive grinding time can lead to over-grinding and uneven particle size sampling. By changing different control parameters and recording the data sequence of particle size variation with grinding time for each grinding task, particle size data sequences corresponding to each control parameter are obtained, reflecting the impact of different control parameters on particle size. In practical applications, experiments and optimizations are needed based on specific circumstances to determine the optimal grinding parameters to obtain the ideal particle size.

[0045] S200, based on the particle size data sequence, a first sample set and a second sample set are extracted from the initial dataset; wherein, the grinding particle size changes faster with grinding time in the first sample set than in the second sample set.

[0046] It should be noted that the slower the change in grinding particle size with grinding time, the better the effect of control parameters on grinding particles can be identified, which is beneficial for the subsequent training of neural network models; conversely, the higher the processing efficiency, the better the feedback learning of the subsequent neural network model, and the better the combination of control parameters can be formed.

[0047] S300, construct a fusion feature vector based on the first sample set and the second sample set, and train the pre-constructed neural network model using the fusion feature vector to obtain the trained model;

[0048] S400: Obtain the target particle size for grinding the lithium battery cathode material, input the target particle size into the trained model to obtain the corresponding control parameters, and control the motor operation according to the control parameters corresponding to the particle size range.

[0049] It should be noted that the nano-grinding machine can process particles ranging from 1µm to 50nm, and features a variable frequency speed control device with an adjustable speed range of 0-2800 rpm. In some embodiments, a target particle size range is set according to the required particle size of the lithium battery cathode material. The target particle size is input into a trained model to obtain corresponding control parameters, which are then sent to the controller to control the motor output.

[0050] This invention trains a model by collecting control parameters and grinding particle size. The model outputs control parameters corresponding to the target particle size, thus meeting different particle size requirements. This invention uses a trained model for control parameter prediction, which improves grinding efficiency and processing accuracy.

[0051] In some embodiments, S200, the step of extracting a first sample set and a second sample set from the initial dataset based on the granular data sequence includes:

[0052] S210, determine the frequency distribution and standard deviation of granularity based on the granularity distribution data in the initial dataset;

[0053] It should be noted that frequency distribution, also called interval distribution or differential distribution, refers to the percentage of particles between two particle sizes. Standard deviation represents the degree of deviation of grinding particles, reflecting the particle size uniformity of the grinding process.

[0054] S220, divide the initial dataset into a positive sample dataset with a standard deviation below the standard deviation threshold and a negative sample dataset with a standard deviation above the standard deviation threshold;

[0055] Specifically, based on the uniformity requirements of the grinding process, a standard deviation threshold is preset, and the particle size distribution data is divided according to the standard deviation threshold. The initial data corresponding to the particle size distribution data is then assigned to either the positive sample dataset or the negative sample dataset.

[0056] S230, for each particle size data sequence in the positive sample dataset, the change in the average particle size with the grinding time is used as the gradient. The particle size data sequences are arranged in ascending order of the gradient. The first few particle size data sequences and their corresponding control parameters are selected as the first sample set, and the last few particle size data sequences and their corresponding control parameters are selected as the second sample set; wherein, the average value is the interval with the largest frequency distribution.

[0057] It should be noted that the selected positive sample dataset excludes control parameters that result in poor grinding effects. A smaller gradient indicates a slower change in particle size, making it easier to discern the effect of control parameters on the grinding particles, which is beneficial for subsequent neural network model training. Provided the sample data in the positive sample dataset satisfies particle size uniformity, a larger gradient indicates higher processing efficiency, which is beneficial for the feedback learning of the subsequent neural network model.

[0058] In some embodiments, in S300, the step of constructing a fused feature vector based on the first sample set and the second sample set, and training a pre-constructed neural network model using the fused feature vector to obtain a trained model, includes:

[0059] S310, input the first sample set into the time feature extraction channel to extract time features, and input the second sample set into the spatial feature extraction channel to extract spatial features;

[0060] S320, the temporal features extracted by the temporal feature extraction channel and the spatial features extracted by the spatial feature extraction channel are concatenated in a weighted manner to obtain a fused feature vector;

[0061] S330, the fused feature vector is input into the pre-built neural network model for training. The loss value is calculated based on the pre-built loss function and backpropagated to the neural network model until the loss value of the neural network model reaches the set loss threshold, thus obtaining the trained model.

[0062] It's important to note that parameter calculation for a neural network model is an iterative process. By continuously adjusting weights and biases, the model can learn patterns and rules in the data. Specifically, the neural network model is constructed, determining the number of layers, the number of neurons in each layer, and the connection methods. Initial weights and biases are assigned to each neuron. The fused feature vector is passed through each layer of the neural network, and the output of each neuron is calculated. Based on the difference between the expected and actual output granularity, the loss function is calculated to measure the model's prediction error. The gradient of the loss function is passed to each layer of the neural network model using the backpropagation algorithm, updating the weights and biases to reduce the loss. The forward and backward propagation process is repeated until the loss of the neural network model on the fused feature vector reaches a set loss threshold.

[0063] In some embodiments, S310, the step of extracting time features from the input time feature extraction channel of the first sample set includes:

[0064] S311, the first sample set is formed into a first granularity corresponding to a first sample matrix, wherein the first sample matrix includes d-dimensional control parameters of different types collected at n sampling times; wherein, for each element in the first sample matrix The superscript j represents the type of control parameter, and the subscript i represents the sampling time, i=1,2,...,n; j=1,2,...,d;

[0065] S312, input the first sample matrix into the first bidirectional LSTM network to obtain the first expected feature vector.

[0066] A first bidirectional LSTM network is used to analyze the sample matrix. This can comprehensively capture the dynamic changes in the granular data sequence, improve the accuracy and effectiveness of sample matrix extraction, and obtain a comprehensive expected feature that reflects the granularity.

[0067] In some embodiments, S310, the step of extracting spatial features from the second sample set input spatial feature extraction channel includes:

[0068] S313, the second sample set is used to form a second sample matrix and a corresponding second granularity. The second sample matrix includes d-dimensional control parameters of different types collected at m sampling times; wherein, for each element in the second sample matrix... The superscript j represents the type of control parameter, and the subscript i represents the sampling time, i=1,2,...,n; j=1,2,...,d,m<n;

[0069] It should be noted that there is a correlation between different types of control parameters. Due to the different grinding times, the correlation between control parameters changes accordingly. The spatial features of control parameters have dynamic characteristics. Therefore, this invention designs a multi-scale convolution module to extract the spatial features of control parameters from multiple aspects as much as possible.

[0070] S314, a set of one-dimensional convolution kernels are used to slide in the order of variables and traverse all time periods in the granular data sequence, and a ReLU activation function is connected to perform initial feature extraction on the granular data sequence;

[0071] S315, the extracted initial features are processed by a multi-scale convolution module to obtain multiple feature vectors, and the multiple feature vectors are fused by the ReLU activation function to obtain a second desired feature vector.

[0072] Specifically, one-dimensional convolutional kernels of different lengths in multiple convolutional branches are traversed from left to right along the data variable axis with a step count of 1 to generate preliminary feature vectors and complete convolution operations for all time periods, outputting multiple feature vectors along the variable axis to represent different spatial features; the extracted feature vectors are then concatenated and fused along the time axis to obtain feature vectors corresponding to the multiple convolutional branches, each representing the feature vectors extracted at each scale of the one-dimensional convolutional kernel; the multiple feature vectors are then fused using the ReLU activation function to obtain the second desired feature vector.

[0073] In some embodiments, the window size of the one-dimensional convolutional kernel is 1 to 3, the number of convolutional branches is set to 2 to 6, and the number of convolutional kernels is 16 or 32; the first expected feature vector reflects the temporal features; the second expected feature vector reflects the spatial features. The temporal features extracted by the temporal feature extraction channel and the spatial features extracted by the spatial feature extraction channel are concatenated in a weighted manner to obtain the fused features.

[0074] After obtaining the first expected feature vector and the second expected feature vector, the first expected feature vector and the second expected feature vector are assigned corresponding weights and then summed to obtain the fused feature vector.

[0075] In some embodiments, the loss function is:

[0076] ;

[0077] Among them, L FU For the fusion loss function, L MSE Here, k is the weight of the fusion loss function, where 0 < k < 1;

[0078] ;

[0079] L FU =meanKL(softmax(X(i,j)),softmax(Y(i,j)));

[0080] Where X(i,j) represents the fused feature vector, Y(i,j) represents the feature vector predicted by the neural network model, KL() represents the KL divergence loss, mean() represents the average value, and softmax() represents the Softmax function.

[0081] It should be noted that both the first and second sample sets are used in the loss calculation. After the granularity of the input is properly matched, the feature vectors are fused as the actual data, and the feature vectors predicted by the neural network model are used as the prediction data.

[0082] and Figure 1 The corresponding method is referenced. Figure 2 This invention provides a motor operation control system for a nano-grinding machine, comprising:

[0083] At least one processor;

[0084] At least one memory for storing at least one program;

[0085] When the at least one program is executed by the at least one processor, the at least one processor performs the method described above.

[0086] It is evident that the content of the above method embodiments is applicable to this system embodiment. The specific functions implemented in this system embodiment are the same as those in the above method embodiments, and the beneficial effects achieved are also the same as those achieved in the above method embodiments.

[0087] Furthermore, embodiments of the present invention also disclose a computer program product or computer program stored in a computer-readable storage medium. A processor of a computer device can read the computer program from the computer-readable storage medium, and the processor executes the computer program, causing the computer device to perform the described method. Similarly, the content of the above method embodiments is applicable to this storage medium embodiment. The specific functions implemented in this storage medium embodiment are the same as those in the above method embodiments, and the beneficial effects achieved are also the same as those achieved in the above method embodiments.

[0088] It will be understood by those skilled in the art that all or some of the methods and systems disclosed above can be implemented as software, firmware, hardware, and suitable combinations thereof. Some or all of the physical components can be implemented as software executed by a processor, such as a central processing unit, digital signal processor, or microprocessor, or as hardware, or as an integrated circuit, such as an application-specific integrated circuit. Such software can be distributed on a computer-readable medium, which can include computer storage media (or non-transitory media) and communication media (or transient media). As is known to those skilled in the art, the term computer storage media includes volatile and non-volatile, removable and non-removable media implemented in any method or technology for storing information (such as computer-readable instructions, data structures, program modules, or other data). Computer storage media includes, but is not limited to, RAM, ROM, EEPROM, flash memory or other memory technologies, CD-ROM, digital versatile disc (DVD) or other optical disc storage, magnetic cartridges, magnetic tape, disk storage or other magnetic storage devices, or any other medium that can be used to store desired information and is accessible to a computer. Furthermore, as is known to those skilled in the art, communication media typically contain computer-readable instructions, data structures, program modules, or other data in modulated data signals such as carrier waves or other transmission mechanisms, and may include any information delivery medium.

[0089] The above is a detailed description of the preferred embodiments of this disclosure. However, this disclosure is not limited to the above embodiments. Those skilled in the art can make various equivalent modifications or substitutions without departing from the spirit of this disclosure. All such equivalent modifications or substitutions are included within the scope defined by the claims of this disclosure.

Claims

1. A method for controlling the operation of a motor in a nano-grinding machine, characterized in that, The method includes the following steps: An initial dataset is obtained, which includes a particle size data sequence showing the change in grinding particle size with grinding time under multiple control parameters; wherein, the control parameters include the opening degree of the electric ball valve, the grinding power of the motor, and the grinding speed; the particle size data sequence includes multiple particle size distribution data sorted by time series. Based on the granularity data sequence, a first sample set and a second sample set are extracted from the initial dataset; wherein, the grinding particle size changes faster with grinding time in the first sample set than in the second sample set. A fusion feature vector is constructed based on the first sample set and the second sample set. The pre-constructed neural network model is then trained using the fusion feature vector to obtain a trained model. The target particle size for grinding lithium battery cathode material is obtained, and the target particle size is input into the trained model to obtain the corresponding control parameters. The motor is then controlled to run according to the control parameters corresponding to the particle size range. The extraction of the first sample set and the second sample set from the initial dataset based on the granular data sequence includes: The frequency distribution and standard deviation of granularity are determined based on the granularity distribution data in the initial dataset. The initial dataset is divided into a positive sample dataset with a standard deviation below a standard deviation threshold and a negative sample dataset with a standard deviation above a standard deviation threshold; For each particle size data sequence in the positive sample dataset, the change in the average particle size with the grinding time is used as the gradient. The particle size data sequences are arranged in ascending order of the gradient. The first few particle size data sequences and their corresponding control parameters are selected as the first sample set, and the last few particle size data sequences and their corresponding control parameters are selected as the second sample set. The average value is the interval with the largest frequency distribution. The step of constructing a fused feature vector based on the first sample set and the second sample set, and training a pre-constructed neural network model using the fused feature vector to obtain a trained model includes: The first sample set is input into the time feature extraction channel to extract time features, and the second sample set is input into the spatial feature extraction channel to extract spatial features. The temporal features extracted by the temporal feature extraction channel and the spatial features extracted by the spatial feature extraction channel are concatenated in a weighted manner to obtain a fused feature vector; The fused feature vectors are input into a pre-built neural network model for training. The loss value is calculated based on the pre-built loss function and backpropagated to the neural network model until the loss value of the neural network model reaches a set loss threshold, thus obtaining a trained model. The step of extracting time features from the first sample set via the time feature extraction channel includes: The first sample set is used to form a first sample matrix with a first granularity. The first sample matrix includes d-dimensional control parameters of different types collected at n sampling times. Specifically, for each element in the first sample matrix... The superscript j represents the type of control parameter, and the subscript i represents the sampling time, i=1,2,...,n; j=1,2,...,d; The first sample matrix is ​​input into the first bidirectional LSTM network to obtain the first expected feature vector; The step of extracting spatial features from the second sample set input spatial feature extraction channel includes: The second sample set is used to form a second sample matrix and a corresponding second granularity. The second sample matrix includes d-dimensional control parameters of different types collected at m sampling times. Specifically, for each element in the second sample matrix... The superscript j represents the type of control parameter, and the subscript i represents the sampling time, i=1,2,...,n; j=1,2,...,d,m<n; A set of one-dimensional convolutional kernels is used to slide in variable order and traverse all time periods in the granular data sequence, and a ReLU activation function is connected to perform initial feature extraction on the granular data sequence; The extracted initial features are processed using a multi-scale convolution module to obtain multiple feature vectors, and the multiple feature vectors are fused using the ReLU activation function to obtain a second desired feature vector; The loss function is: ； Among them, L FU For the fusion loss function, L MSE Here, k is the weight of the fusion loss function, where 0 < k < 1; ； L FU =meanKL(softmax(X(i,j)),softmax(Y(i,j)))? Where X(i,j) represents the fused feature vector, Y(i,j) represents the feature vector predicted by the neural network model, KL() represents the KL divergence loss, mean() represents the average value, and softmax() represents the Softmax function.

2. A motor operation control system for a nano-grinding machine, characterized in that, The system includes: At least one processor; At least one memory for storing at least one program; When the at least one program is executed by the at least one processor, the at least one processor implements the motor operation control method for the nano-grinding machine as described in claim 1.

3. A computer-readable storage medium storing a processor-executable program, characterized in that, The processor-executable program is used to perform the method as described in claim 1 when executed by the processor.

Images