Prediction method and apparatus for crushing particle size of ternary positive electrode material

Abstract

Disclosed in the present invention are a prediction method and apparatus for a crushing particle size of a ternary positive electrode material. The method comprises: for each crushing device, collecting crushing particle size measurement results of the crushing device under different process parameters to form a sample data set, the sample data set comprising the crushing particle size measurement results corresponding to the crushing device under different process parameters; performing preprocessing and standard-compliant data screening on the sample data set; on the basis of different products, selecting corresponding sample data sub-sets from the sample data set for training to obtain crushing particle size prediction models corresponding to different products; and inputting process parameters, obtained on site, of the crushing device into a corresponding crushing particle size prediction model so as to obtain a crushing particle size index prediction result. By using the present invention, crushing particle size prediction models are used to obtain predicted measurement particle sizes corresponding to process parameters of current devices, so that a worker can adjust the process parameters of the devices on the basis of the deviation between the predicted measurement particle sizes and target particle size data.

Description

A method and device for predicting the crushing particle size of ternary positive electrode materials Technical Field

[0001] The present invention relates to the field of cathode material production, and in particular to a method and device for predicting the crushing particle size of a ternary cathode material. Background Art

[0002] In the production process of ternary positive electrode materials, the precursor and lithium source are mixed and sintered at high temperature to generate ternary positive electrode materials, and then the block materials out of the furnace are converted into powdered materials through air flow mill equipment. During the crushing process, according to the particle size of the sampling test, the classification frequency, induced draft frequency, feeding frequency and crushing air pressure of the air flow mill need to be adjusted to control the particle size within the internal control range.

[0003] Currently, particle size sampling and testing are performed, and the sampling and testing cycle is long. After the data is produced, relevant personnel will decide whether to adjust the parameters based on the particle size data indicators. The parameters to adjust require long-term experience accumulation and understanding of the equipment. Different technicians cannot debug the same crushing equipment in a completely consistent manner. There is no scientific and unified parameter adjustment method, and multiple adjustments are often required to achieve qualified indicators. At the same time, particle size debugging is quite difficult for new technicians. Finally, after the adjustments are completed, samples are taken for testing until the indicators are qualified.

[0004] In summary, the entire equipment process parameter adjustment process requires extremely high human resources and the adjustment efficiency is low.

[0005] Summary of the Invention

[0006] An embodiment of the present invention provides a method and device for predicting the crushing particle size of a ternary positive electrode material, which uses a crushing particle size prediction model to obtain the predicted detection particle size corresponding to the current equipment process parameters, making it convenient for staff to adjust the equipment process parameters according to the deviation between the predicted detection particle size and the target particle size data.

[0007] To achieve the above-mentioned objectives, a first aspect of an embodiment of the present application provides a method for predicting the crushing particle size of a ternary positive electrode material, comprising:

[0008] For each pulverizing equipment, the pulverizing particle size test results of the pulverizing equipment under different process parameters are collected to form a sample data set;

[0009] Preprocessing the sample data set and screening qualified standard data;

[0010] Selecting corresponding sample data subsets from the sample data set according to different products for training, and obtaining crushing particle size prediction models corresponding to different products;

[0011] The process parameters of the corresponding pulverizing equipment obtained on site are input into the corresponding pulverizing particle size prediction model to obtain the pulverizing particle size index prediction results.

[0012] In a possible implementation of the first aspect, preprocessing the sample data set and screening qualified standard data specifically includes:

[0013] Performing non-data format data removal, column data type conversion, data filtering and normalization processing on the sample data set;

[0014] According to the preset upper and lower limits of process control, outliers in the sample data set are removed and qualified standard data judgment and marking are performed on the sample data set.

[0015] In a possible implementation manner of the first aspect, removing outliers from the sample data set specifically includes:

[0016] The upper limit of process control and the lower limit of process control are respectively used as the upper quartile and the lower quartile to calculate the interquartile range, and the points where the sample data are concentrated in the interval (upper quartile + 1.5 × interquartile range, lower quartile - 1.5 × interquartile range) are removed as outliers in the sample data set.

[0017] In a possible implementation of the first aspect, the identifying and marking qualified standard data on the sample dataset specifically includes:

[0018] According to the upper limit and lower limit of process control, each data in the sample data set is judged on the crushing particle size parameter and the broadening coefficient. When a data in the sample data set passes the judgment, a new column is added to the sample data set to store qualified information.

[0019] In a possible implementation of the first aspect, selecting corresponding sample data subsets from the sample data set for training according to different products specifically includes:

[0020] Determining a sample data subset required for this training round from the sample data set according to the type of product;

[0021] Dividing the sample data subset into a training set and a test set according to a preset ratio;

[0022] The importance of different features of the training set is calculated using a random forest algorithm;

[0023] Sort the importance of different features, and select a preset number of features as the feature set in descending order of importance;

[0024] Selecting multiple samples from the training set to train a decision tree root node;

[0025] Selecting multiple features from the feature set as split features to construct decision tree child nodes;

[0026] According to the preset information gain strategy, a split feature is selected as the split attribute of the child node, and each decision tree child node is split until the decision tree can no longer split;

[0027] The voting results of all nodes in the decision tree are counted and the category with the largest number of voting results is used as the crushing particle size prediction model.

[0028] In a possible implementation of the first aspect, after selecting corresponding sample data subsets from the sample data set according to different products for training to obtain crushing particle size prediction models corresponding to different products, the method further includes:

[0029] The crushing particle size prediction model is subjected to model verification and model evaluation, and the crushing particle size prediction model that fails the model verification or the model evaluation is retrained until the crushing particle size prediction model passes the model verification and the model evaluation.

[0030] In a possible implementation of the first aspect, the model verification specifically includes:

[0031] The test set in the sample data subset is input into the crushing particle size prediction model as sample data, and the difference between the prediction result output by the crushing particle size prediction model and the actual test result is compared. If the difference is within a preset range, the crushing particle size prediction model passes the model verification.

[0032] In a possible implementation of the first aspect, the model evaluation specifically includes:

[0033] Different evaluation index values ​​of the crushing particle size prediction model in various parameter and algorithm combinations are obtained. If all evaluation index values ​​are within a preset range, the crushing particle size prediction model passes the model evaluation.

[0034] In a possible implementation of the first aspect, before inputting the process parameters of the corresponding pulverizing equipment obtained on-site into the corresponding pulverizing particle size prediction model to obtain the pulverizing particle size index prediction result, the method further includes:

[0035] The crushing particle size detection results of the corresponding crushing equipment under historical process parameters in the factory data are used as the training set, and an online learning mechanism is introduced to iteratively optimize the crushing particle size prediction model.

[0036] A second aspect of the embodiments of the present application provides a device for predicting the crushing particle size of a ternary positive electrode material, comprising:

[0037] A collection module is used to collect the pulverization particle size test results of each pulverization device under different process parameters to form a sample data set;

[0038] A screening module, used for preprocessing the sample data set and screening qualified standard data;

[0039] A training module is used to select corresponding sample data subsets from the sample data set according to different products for training, so as to obtain crushing particle size prediction models corresponding to different products;

[0040] The prediction module is used to input the process parameters of the corresponding crushing equipment obtained on site into the corresponding crushing particle size prediction model to obtain the crushing particle size index prediction result.

[0041] Compared with the prior art, the embodiment of the present invention provides a method and device for predicting the crushing particle size of a ternary positive electrode material. By obtaining the crushing data of each crushing device for different products to form a sample data subset to train a crushing particle size prediction model, a crushing particle size prediction model for each crushing device for different products is obtained. The crushing particle size prediction model is then iterated until the crushing particle size prediction model passes the model verification and model evaluation, thereby ensuring the validity and reliability of the crushing particle size prediction results of the crushing particle size prediction model. After obtaining an accurate and reliable predicted detection particle size, the staff can flexibly adjust the equipment process parameters according to the deviation between the predicted detection particle size and the target particle size data, so that the predicted detection particle size is as close as possible to the target particle size data. Since the predicted detection particle size does not need to be obtained after the equipment is tested, the detection process is omitted, making the equipment process parameter adjustment efficient. Since it does not rely on the workers' work experience, it also saves labor costs. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] FIG1 is a schematic flow chart of a method for predicting the crushing particle size of a ternary cathode material according to an embodiment of the present invention;

[0043] FIG2 is a flow chart of a ternary cathode material particle size prediction and correction process according to an embodiment of the present invention. DETAILED DESCRIPTION

[0044] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. All other embodiments obtained by ordinary technicians in this field based on the embodiments of the present invention without making any creative efforts shall fall within the scope of protection of the present invention.

[0045] Referring to FIG1 , an embodiment of the present invention provides a method for predicting the crushing particle size of a ternary cathode material, comprising:

[0046] S10. For each pulverizing device, collect the pulverizing particle size test results of the pulverizing device under different process parameters to form a sample data set; the sample data set includes the pulverizing particle size test results corresponding to the pulverizing device under different process parameters.

[0047] S11. Preprocessing the sample data set and screening qualified standard data.

[0048] S12. Select corresponding sample data subsets from the sample data set according to different products for training, and obtain crushing particle size prediction models corresponding to different products.

[0049] S13. Input the process parameters of the corresponding pulverizing equipment obtained on site into the corresponding pulverizing particle size prediction model to obtain the pulverizing particle size index prediction result.

[0050] In this embodiment, S12 obtains a crushing particle size prediction model corresponding to different products, so that the staff can give the process parameters required for the corresponding equipment according to the particle size range required by different products in actual application; when the particle size test results do not meet the standards, the actual process parameters of the equipment can also be adjusted by combining the actual process parameters of the current equipment with the test results to confirm whether the adjustment direction of the process parameters is incorrect and make timely corrections. Since the sample data set obtained from S10 is a sample data set formed based on the crushing particle size test results of each crushing equipment under different process parameters, the crushing particle size prediction model trained on this basis can be used to perform targeted corrections on the process parameters of the corresponding model of equipment, achieving the effect of remotely writing to the on-site equipment to adjust the crushing parameters.

[0051] The model training process in S12 essentially uses big data to analyze the dynamic relationship between pulverization equipment process parameters and particle size results, thereby achieving the goal of predicting particle size test results. After S13, staff can dynamically adjust parameters based on the predicted results, reducing the number of offline manual adjustments to particle size based on actual test results, improving the response speed of particle size adjustment, ensuring the stability of particle size indicators, and reducing the difficulty of particle size adjustment. This is applied to the adjustment of parameters in the airflow mill pulverization process in various workshops and the analysis of particle size indicator deviations.

[0052] Compared with the prior art, the embodiment of the present invention provides a method for predicting the crushing particle size of a ternary positive electrode material. By obtaining the crushing data of each crushing device for different products to form a sample data subset to train a crushing particle size prediction model, a crushing particle size prediction model for each crushing device for different products is obtained. The crushing particle size prediction model is then iterated until the crushing particle size prediction model passes the model verification and model evaluation, thereby ensuring the validity and reliability of the crushing particle size prediction results of the crushing particle size prediction model. After obtaining an accurate and reliable predicted detection particle size, the staff can flexibly adjust the equipment process parameters according to the deviation between the predicted detection particle size and the target particle size data, so that the predicted detection particle size is as close as possible to the target particle size data. Since the predicted detection particle size does not need to be obtained after the equipment is tested, the detection process is omitted, making the equipment process parameter adjustment efficient. Since it does not rely on the workers' work experience, it also saves labor costs.

[0053] Exemplarily, S11 specifically includes:

[0054] S110 , performing non-data format data removal, column data type conversion, data filtering, and normalization processing on the sample data set.

[0055] S111 , according to the preset upper and lower limits of process control, removing outliers in the sample data set and performing qualified standard data identification and marking on the sample data set.

[0056] The process parameters of the pulverizing equipment in S110 may include pulverizing air pressure, grading frequency, feeding frequency, induced draft frequency, and bottom grinding material. Taking the equipment LL5503 as an example, see Table 1; the sample data set includes detection items of parameters such as Dv10, Dv50, Dv90, Dv99, Dn10 and widening coefficient. Taking the equipment LL5503 as an example, see Table 2.

[0057] Table 1 Crushing equipment process parameters

[0058] Table 2 Crushing particle size test items

[0059] In this embodiment, the sample data set includes a large number of process parameters collected by each crushing equipment and the results of multiple corresponding material particle size detection items. The data is preliminarily processed by preprocessing methods such as removing outliers, filling missing values, feature scaling and standardization, and the sample data set is preprocessed and cleaned.

[0060] In particular, in S110, removing non-data format data includes marking all non-numeric format data of crushing equipment process parameters and crushing particle size test results as NaN, deleting rows including NaN and empty values, and ensuring that each data is of numerical type; column data type conversion includes converting the data type of the above data to int or float as required, and unifying the numerical type of each column of data; data filtering includes defining the data range for each column of data, cleaning up the row data whose values ​​are not within the normal range (i.e. unreasonable), and ensuring that there are no abnormal values ​​(normal range: [specification lower limit × 50%, specification upper limit × 150%]); normalization processing: due to the problem of inconsistent dimensions between the original features, it will directly affect the understanding of the feature rules of some modeling algorithms. When the values ​​are relatively concentrated, the maximum and minimum normalization method is used to normalize the feature vectors.

[0061] Exemplarily, removing outliers from the sample data set specifically includes:

[0062] The upper limit of process control and the lower limit of process control are respectively used as the upper quartile and the lower quartile to calculate the interquartile range, and the points of the sample data set outside the interval (upper quartile + 1.5 × interquartile range, lower quartile - 1.5 × interquartile range) are removed as outliers.

[0063] This example uses the interquartile range clustering algorithm to remove outliers from the sample set. The upper and lower control limits are used as the upper and lower quartiles, and the difference is calculated to determine whether an outlier is present. Using 1.5 times the IQR (interquartile range) as the standard, points exceeding (upper quartile + 1.5 times the IQR distance, or lower quartile - 1.5 times the IQR distance) are considered outliers.

[0064] Exemplarily, the performing qualified standard data identification and marking on the sample data set specifically includes:

[0065] According to the upper limit and lower limit of process control, each data in the sample data set is judged on the crushing particle size parameter and the broadening coefficient. When a data in the sample data set passes the judgment, a new column is added to the sample data set to store qualified information.

[0066] That is, the qualification is judged according to the upper limit and lower limit of process control (the crushing particle size parameters include Dv10, Dv50, Dv90, Dv99, Dn10, and the broadening coefficient = (Dv90-Dv10) / Dv50), and a new column of Label is added to store the qualification information, where qualified is 1 and unqualified is 0.

[0067] Exemplarily, the selecting corresponding sample data subsets from the sample data set for training according to different products specifically includes:

[0068] Determine the subset of sample data required for this training round from the sample dataset according to the type of the product;

[0069] Divide the subset of sample data into a training set and a test set according to a preset ratio;

[0070] Calculate the importance of different features of the training set using the random forest algorithm;

[0071] Sort the importance of different features, and select a preset number of features as the feature set according to the sorting from large to small importance;

[0072] Select multiple samples from the training set to train the root node of the decision tree.

[0073] Select multiple features from the feature set as splitting features to construct the child nodes of the decision tree.

[0074] Select a splitting feature as the splitting attribute of this child node according to a preset information gain strategy, and split each child node of the decision tree until the decision tree cannot be split anymore.

[0075] Statistically count the voting results of all nodes of the decision tree and take the category with the largest number of voting results as the comminution granularity prediction model.

[0076] The purpose of this embodiment is to train multiple decision trees based on the comminution granularity and equipment process parameter data, and screen to obtain the comminution granularity prediction model. The general process can be referred to as follows:

[0077] ① Determine the sample data set required for this training round;

[0078] ② Divide the sample data set into a training set and a test set according to a ratio;

[0079] ③ Set the overall training set as T, and there are N samples in T. Each time, randomly select N samples with replacement to train the first decision tree, which is used as the sample at the root node of the decision tree;

[0080] ④ Set the number of features of the training set as d (selected from 3), and each time only select k (k < d) to construct the child nodes of the decision tree. When each node of the decision tree needs to be split, randomly select k features from d, satisfying the condition k < x < d. Then select 1 feature from these k features as the splitting attribute of this node using the information gain strategy;

[0081] ⑤ Each node in the process of forming the decision tree should be split according to step ④ until it cannot be split anymore;

[0082] ⑥According to steps ③ to ⑤, a large number of decision tree nodes, namely random forests, are established. Each decision tree node will have a voting result. The category with the most voting results will be selected as the crushing particle size prediction model.

[0083] ⑦ Use the crushing particle size optimization model selected in ⑥ as the decision tree root node for the next round of training, and perform the next training and testing according to the process of ① to ⑥ until the final round is reached.

[0084] ⑧ Different products have different acceptance criteria for material particle size test results. It is necessary to determine the sample data subset for the corresponding product and perform model training for each product according to the process from ① to ⑦ to obtain multiple prediction models.

[0085] Exemplarily, after selecting corresponding sample data subsets from the sample data set for training according to different products to obtain the crushing particle size prediction models corresponding to different products, the method further includes:

[0086] The crushing particle size prediction model is subjected to model verification and model evaluation, and the crushing particle size prediction model that fails the model verification or the model evaluation is retrained until the crushing particle size prediction model passes the model verification and the model evaluation.

[0087] Exemplarily, the model verification specifically includes:

[0088] The test set in the sample data subset is input into the crushing particle size prediction model as sample data, and the difference between the prediction result output by the crushing particle size prediction model and the actual test result is compared. If the difference is within a preset range, the crushing particle size prediction model passes the model verification.

[0089] In this example, the test data in the dataset is used as sample data to verify the training effect of the previously obtained crushing particle size optimization model. The equipment process parameters and actual test results are input, and the model is used to analyze whether the current particle size is qualified, thereby verifying the effectiveness of the model.

[0090] Exemplarily, the model evaluation specifically includes:

[0091] Different evaluation index values ​​of the crushing particle size prediction model in various parameter and algorithm combinations are obtained. If all evaluation index values ​​are within a preset range, the crushing particle size prediction model passes the model evaluation.

[0092] In this embodiment, the analytical performance of the model is compared between a set of parameters, different parameter combinations or multiple algorithms to test the reliability of the crushing particle size tuning model; finally, based on some evaluation indicators (such as mean square error, root mean square error, mean absolute error, mean relative error, etc.) or graphical display, a prediction model with reliable quality is obtained.

[0093] For example, before inputting the process parameters of the corresponding pulverizing equipment obtained on-site into the corresponding pulverizing particle size prediction model to obtain the pulverizing particle size index prediction result, the method further includes:

[0094] The crushing particle size detection results of the corresponding crushing equipment under historical process parameters in the factory data are used as the training set, and an online learning mechanism is introduced to iteratively optimize the crushing particle size prediction model.

[0095] By inputting the process parameters of the corresponding pulverizing equipment into the particle size optimization model, which has been trained, learned, and evaluated by the algorithm, the model's principles can be quickly and efficiently applied to new data, thereby generating prediction results for the process parameters of the corresponding pulverizing equipment. This method can be used to access the particle size test results of the corresponding pulverizing equipment under historical process parameters in factory data, and introduce an online learning mechanism to iteratively optimize the model's generalization capabilities. This allows for application to large amounts of actual production data, and adjust the model to better fit the actual situation.

[0096] Furthermore, as shown in Figure 2, in practical applications, the above embodiments can be used to establish different dynamic analysis models for different products. This allows for the impact of different equipment on material indices to be considered, with multiple models established or equipment incorporated into the influencing factors. The logic behind the impact of process parameters on indices can be determined by using data curves to determine possible fitting functions between parameters and indices. Finally, a fitting mathematical model for all relevant parameters can be developed for each particle size index.

[0097] Referring to Figure 2, another application of this embodiment in actual application is to input the actual process parameters of the equipment into the crushing particle size optimization prediction model to obtain the crushing particle size index prediction result. If the prediction result is outside the qualified index range, an abnormal crushing particle size index warning is issued.

[0098] Compared with the prior art, the embodiment of the present invention provides a method for predicting the crushing particle size of a ternary positive electrode material. By obtaining the crushing data of each crushing device for different products to form a sample data subset to train a crushing particle size prediction model, a crushing particle size prediction model for each crushing device for different products is obtained. The crushing particle size prediction model is then iterated until the crushing particle size prediction model passes the model verification and model evaluation, thereby ensuring the validity and reliability of the crushing particle size prediction results of the crushing particle size prediction model. After obtaining an accurate and reliable predicted detection particle size, the staff can flexibly adjust the equipment process parameters according to the deviation between the predicted detection particle size and the target particle size data, so that the predicted detection particle size is as close as possible to the target particle size data. Since the predicted detection particle size does not need to be obtained after the equipment is tested, the detection process is omitted, making the equipment process parameter adjustment efficient. Since it does not rely on the workers' work experience, it also saves labor costs.

[0099] An embodiment of the present application provides a device for predicting the crushing particle size of a ternary cathode material, including: a collection module, a screening module, a training module, and a prediction module.

[0100] The collection module is used to collect the pulverization particle size test results of each pulverization device under different process parameters to form a sample data set. The sample data set includes the pulverization particle size test results corresponding to the pulverization device under different process parameters.

[0101] The screening module is used to pre-process the sample data set and screen qualified standard data.

[0102] The training module is used to select corresponding sample data subsets from the sample data set according to different products for training, so as to obtain crushing particle size prediction models corresponding to different products.

[0103] The prediction module is used to input the process parameters of the corresponding crushing equipment obtained on site into the corresponding crushing particle size prediction model to obtain the crushing particle size index prediction result.

[0104] Compared with the prior art, the embodiment of the present invention provides a device for predicting the crushing particle size of a ternary positive electrode material. By obtaining the crushing data of each crushing device for different products to form a sample data subset to train a crushing particle size prediction model, a crushing particle size prediction model for each crushing device for different products is obtained. The crushing particle size prediction model is then iterated until the crushing particle size prediction model passes the model verification and model evaluation, thereby ensuring the validity and reliability of the crushing particle size prediction result of the crushing particle size prediction model. After obtaining an accurate and reliable predicted detection particle size, the staff can flexibly adjust the equipment process parameters according to the deviation between the predicted detection particle size and the target particle size data, so that the predicted detection particle size is as close as possible to the target particle size data. Since the predicted detection particle size does not need to be obtained after the equipment is tested, the detection process is omitted, making the equipment process parameter adjustment efficient. Since it does not rely on the workers' work experience, it also saves labor costs.

[0105] Those skilled in the art will clearly understand that, for the convenience and brevity of description, the specific working process of the system described above can refer to the corresponding process in the aforementioned method embodiment and will not be described again here.

[0106] An embodiment of the present application provides a computer-readable storage medium storing a computer program, which, when executed by a processor, implements the above-mentioned method for predicting the crushing particle size of a ternary positive electrode material.

[0107] The computer device may be a computing device such as a smartphone, tablet computer, desktop computer, or cloud server. The computer device may include, but is not limited to, a processor and memory. Those skilled in the art will appreciate that the diagrams are merely examples of computer devices and do not limit the computer device. The computer device may include more or fewer components than shown, or a combination of certain components, or different components. For example, the computer device may also include input / output devices, network access devices, etc.

[0108] The processor may be a central processing unit (CPU), other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA), other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor may be a microprocessor or any conventional processor.

[0109] In some embodiments, the memory may be an internal storage unit of the computer device, such as a hard disk or memory of the computer device. In other embodiments, the memory may also be an external storage device of the computer device, such as a plug-in hard disk equipped on the computer device, a smart memory card (Smart Media Card, SMC), a secure digital (Secure Digital, SD) card, a flash card, etc. Furthermore, the memory may also include both an internal storage unit of the computer device and an external storage device. The memory is used to store an operating system, an application program, a boot loader (BootLoader), data, and other programs, such as the program code of the computer program. The memory may also be used to temporarily store data that has been output or is to be output.

[0110] An embodiment of the present application provides a computer program product. When the computer program product is run on a computer device, the computer device implements the steps in the above-mentioned various method embodiments when executing the computer program product.

[0111] In several embodiments provided in the present application, it is understood that each box in the flow chart or block diagram can represent a part of a module, program segment or code, and the part of the module, program segment or code contains one or more executable instructions for realizing the specified logical function. It should also be noted that in some alternative implementations, the functions marked in the box can also occur in an order different from that marked in the accompanying drawings. For example, two consecutive boxes can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, which depends on the functions involved.

[0112] If the functions are implemented in the form of software function modules and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application, or the part that contributes to the prior art, or the part of the technical solution, can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes several instructions for enabling a computer device to execute all or part of the steps of the method described in each embodiment of the present application. The aforementioned storage medium includes various media that can store program codes, such as a USB flash drive, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk.

[0113] The above is a preferred embodiment of the present invention. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principles of the present invention. These improvements and modifications are also considered to be within the scope of protection of the present invention.

Claims

1. A method for predicting the crushing particle size of a ternary cathode material, characterized in that Including: For each crushing device, collect the crushing particle size detection results of the crushing device under different process parameters to form a sample data set; The sample data set includes the crushing particle size detection results corresponding to the crushing device under different process parameters; Preprocess the sample data set and screen the qualified standard data; Select corresponding sample data subsets from the sample data set according to different products for training to obtain crushing particle size prediction models corresponding to different products; Input the process parameters of the corresponding crushing device obtained on-site into the corresponding crushing particle size prediction model to obtain the prediction result of the crushing particle size index.

2. The method for predicting the crushing particle size of the ternary cathode material according to claim 1, wherein The preprocessing of the sample data set and the screening of the qualified standard data specifically include: For the sample data set, perform non-data format data removal, column data type conversion, data filtering, and normalization processing; According to the preset upper and lower limits of process control, remove the outliers in the sample data set and perform qualified standard data discrimination and marking on the sample data set.

3. The method for predicting the crushing particle size of the ternary cathode material according to claim 2, wherein The removal of the outliers in the sample data set specifically includes: Take the upper limit of process control and the lower limit of process control as the upper quartile and the lower quartile respectively to calculate the interquartile range, and take the points outside the interval (upper quartile + 1.5 × interquartile range, lower quartile - 1.5 × interquartile range) in the sample data set as the outliers in the sample data set to be removed.

4. The ternary cathode material crushing particle size prediction method according to claim 2, wherein The qualified standard data discrimination and marking of the sample data set specifically include: Perform crushing particle size parameter and broadening coefficient discrimination on each data in the sample data set according to the upper limit of process control and the lower limit of process control. When a data in the sample data set passes the discrimination, add a column in the sample data set to store the qualified information.

5. The method for predicting the pulverization particle size of the ternary cathode material according to claim 1, wherein, The selection of the corresponding sample data subsets from the sample data set according to different products for training specifically includes: Determine the sample data subset required for this training round from the sample data set according to the type of product; Divide the sample data subset into a training set and a test set according to a preset ratio; Use the random forest algorithm to calculate the importance of different features in the training set; Sort the importance of different features, and select a preset number of features as the feature set according to the descending order of importance; Select multiple samples from the training set to train the root nodes of the decision tree; Select multiple features from the feature set as the splitting features to construct the decision tree sub-nodes; Select a splitting feature as the splitting attribute of the sub-node according to the preset information gain strategy, and split each decision tree sub-node until the decision tree cannot be split anymore; Count the voting results of all nodes of the decision tree and take the category with the most voting results as the crushing particle size prediction model.

6. The method for predicting the crushing particle size of the ternary cathode material according to claim 1, wherein, After the selection of the corresponding sample data subsets from the sample data set according to different products for training to obtain the crushing particle size prediction models corresponding to different products, it further includes: Perform model verification and model evaluation on the crushing particle size prediction model, and retrain the crushing particle size prediction model that fails to pass the model verification or model evaluation until the crushing particle size prediction model passes the model verification and model evaluation.

7. The method for predicting the crushing particle size of the ternary cathode material according to claim 6, wherein, The model verification specifically includes: Use the test set in the sample data subset as example data and input it into the comminution particle size prediction model. Compare the difference between the prediction result output by the comminution particle size prediction model and the actual detection result. If the difference is within the preset range, the comminution particle size prediction model passes the model verification.

8. The method for predicting the crushing particle size of the ternary cathode material according to claim 6, wherein, The model evaluation specifically includes: Obtain different evaluation index values of the comminution particle size prediction model in various parameter and algorithm combinations. If all the evaluation index values are within the preset range, the comminution particle size prediction model passes the model evaluation.

9. The method for predicting the crushing particle size of the ternary cathode material according to claim 1, wherein, Before inputting the process parameters of the corresponding comminution equipment obtained on-site into the corresponding comminution particle size prediction model to obtain the prediction result of the comminution particle size index, it also includes: Access the comminution particle size detection results corresponding to the corresponding comminution equipment in the factory data under historical process parameters as the training set, and introduce an online learning mechanism to iteratively optimize the comminution particle size prediction model.

10. A device for predicting the crushing particle size of a ternary cathode material, characterized in that, It includes: A collection module for collecting the comminution particle size detection results of each comminution equipment under different process parameters to form a sample data set for each comminution equipment; The sample data set includes the comminution particle size detection results corresponding to the comminution equipment under different process parameters; A screening module for preprocessing the sample data set and screening qualified standard data; A training module for selecting corresponding sample data subsets from the sample data set according to different products for training to obtain comminution particle size prediction models corresponding to different products; A prediction module for inputting the process parameters of the corresponding comminution equipment obtained on-site into the corresponding comminution particle size prediction model to obtain the prediction result of the comminution particle size index.

Images