A method for analyzing lithium carbonate production quality data

Abstract

The present invention relates to the technical field of data processing, and more specifically, to a method for analyzing lithium carbonate production quality data. The method comprises: collecting various quality data, obtaining the splitting degree of each data according to the differential sequence of each data; obtaining the splitting fitness of each data in each data sequence based on the splitting fitness; obtaining multiple splitting points in each data sequence according to the splitting fitness, obtaining the preference degree of each splitting point in each data sequence, and obtaining the final number of splitting points of each data sequence based on the preference degree; obtaining the number of clustering clusters according to the final number of splitting points of each data sequence, clustering various quality data according to the number of clustering clusters to obtain clustering results, and evaluating the quality of lithium carbonate according to the clustering results. The present invention obtains an accurate number of clusters, and a more accurate production quality evaluation result can be obtained according to the accurate clustering result.

Description

Technical Field

[0001] The present invention relates to the technical field of data processing, and more specifically, to a method for analyzing lithium carbonate production quality data. Background Art

[0002] As a key raw material for lithium batteries, the production quality of lithium carbonate directly affects the performance, safety and life of lithium batteries. Therefore, ensuring high-quality lithium carbonate products is crucial to the sustainable development of the industry. To this end, many lithium production companies have begun to implement systematic quality data analysis to optimize production processes and improve product consistency. With the advancement of Industry 4.0 and digital transformation, the large amount of data generated in the production process provides a basis for quality analysis. Therefore, by collecting various quality data in the lithium carbonate production process and performing cluster analysis on each quality data, the production quality of lithium carbonate can be graded and evaluated.

[0003] The patent application document with the current publication number CN109871412A discloses a lane flow analysis method based on K-Means clustering, including: 1. Obtaining traffic flow statistics at a road intersection; 2. Using the sliding average method to smooth the data of a certain week to obtain the flow curves of each week; 3. Setting the initial maximum number of clusters, and traversing each cluster value using the K-Means clustering method; 4. Processing the clustering results, and obtaining a cluster value using the elbow rule; 5. Using this cluster value as the cluster value, clustering again using the K-Means method to obtain the clustering effect of the last week.

[0004] In the process of analyzing the quality of lithium carbonate production by classifying the various quality data in the lithium carbonate production process through the conventional K-means algorithm, it is difficult to determine the number of classifications when classifying the various quality data in the lithium carbonate production process because the analysis of the various quality data in the lithium carbonate production process requires a huge amount of calculation. If the number of classifications is too large, the classification between clusters will be meaningless, resulting in over-segmentation, while too few classifications will lead to under-segmentation. Therefore, both too many and too few classifications will cause inaccurate classification, thereby affecting the accuracy of the analysis of the quality of lithium carbonate production. Summary of the invention

[0005] In order to solve the problem that too many classification numbers in the K-means algorithm will make the classification between clusters meaningless and lead to over-segmentation, and too few classification numbers will lead to under-segmentation, thereby affecting the accuracy of lithium carbonate production quality analysis, the present invention proposes a lithium carbonate production quality data analysis method, which includes the following steps:

[0006] Collect each quality data, each quality data includes several items of data; obtain each data sequence; obtain the split fitness of each data in each data sequence , Represents the split fitness of the i-th data in the j-th data sequence; Represents the degree of splitting of the j-th data; Represents the number of data in the j-th data sequence that is greater than or equal to its i-th data; Represents the mean of all data in the j-th data sequence that is greater than or equal to its i-th data; Represents the number of data in the j-th data sequence that is smaller than the i-th data; represents the mean of all data in the j-th data sequence that is smaller than the i-th data; obtaining each splitting point in each data sequence, wherein the splitting fitness of the splitting point is greater than a preset splitting fitness threshold;

[0007] Obtaining the preference degree of each splitting point in each data sequence; obtaining the final number of splitting points of each data sequence, wherein the preference degree of the final splitting point is greater than a preset preference degree threshold; clustering each quality data according to the final number of splitting points of each data sequence, obtaining a clustering result, and evaluating the quality of lithium carbonate.

[0008] The innovation of the present invention lies in that by calculating the degree of splitting of each data item, the splitting fitness of each data in each data sequence is further obtained, and based on the distribution fitness, the splitting points of each data sequence are obtained, and the preferred degree of each splitting point in each data sequence is combined to determine the number of final splitting points for each data item, and finally a relatively accurate classification number is obtained to cluster the quality data, so that a more accurate production quality assessment result can be obtained based on the accurate clustering result.

[0009] Preferably, the obtaining of each data sequence includes:

[0010] Arrange the j-th item of all quality data in ascending order to obtain the j-th data sequence.

[0011] Preferably, the acquisition of the degree of splitting includes:

[0012] Get the difference sequence of each data;

[0013] ;

[0014] In the formula, Represents the degree of splitting of the j-th data; Represents the value of the kth data in the jth data difference sequence; Represents the number of data in the j-th data difference sequence; Represents the mean of the neighborhood data of the kth data in the jth data difference sequence; Represents the mean of all data in the j-th data difference sequence; || represents the absolute value symbol.

[0015] This facilitates the subsequent acquisition of the split suitability of each data in each data sequence based on the degree of splitting.

[0016] Preferably, the step of obtaining each data differential sequence includes:

[0017] The absolute values ​​of the differences between all adjacent data in the j-th data sequence are obtained in sequence to obtain the j-th data difference sequence.

[0018] Preferably, obtaining each splitting point in each data sequence includes:

[0019] A split fitness threshold is preset, and the data in the j-th data sequence whose split fitness is greater than the split fitness threshold is recorded as a suspected split point; if there are multiple suspected split points appearing consecutively in the j-th data sequence, and the number of the consecutive suspected split points is an even number, any one of the suspected split points is recorded as a split point; if there are multiple suspected split points appearing consecutively, and the number of the consecutive suspected split points is an odd number, the suspected split point located at the middle position is recorded as the split point; if the suspected split point exists alone, it is recorded as a split point, and multiple split points in the j-th data sequence are obtained.

[0020] Delete the suspected split points that appear repeatedly to avoid over-segmentation in the future.

[0021] Preferably, obtaining the preference level of each splitting point in each data sequence includes:

[0022] ;

[0023] In the formula, Represents the preference of the vth split point in the jth data sequence; Represents the value of the vth split point in the jth data sequence; Represents the value of the v-1th split point in the jth data sequence; Represents the value of the v+1th split point in the jth data sequence; Represents the mean of all split segments in the j-th data sequence.

[0024] It is convenient to remove similar split points according to the degree of optimization.

[0025] Preferably, the acquisition of all split segments in the j-th data sequence includes:

[0026] The data segments formed by all adjacent splitting points in the j-th data sequence are recorded as all splitting segments in the j-th data sequence.

[0027] Preferably, clustering each quality data according to the final number of splitting points of each data sequence to obtain a clustering result, and evaluating the quality of lithium carbonate, including:

[0028] Obtain the mean of the final number of split points of each data sequence, record it as the number of clusters, and classify all quality data using the k-means clustering algorithm according to the number of clusters to obtain each classification of the clustering results;

[0029] Each classification of the clustering results is input into the existing lithium carbonate production quality assessment model. If the quality assessment result of any classification of the clustering results is qualified, the production raw material (lithium carbonate) corresponding to the quality data in this classification of the clustering results is qualified; otherwise, the production raw material corresponding to the quality data in this classification of the clustering results is unqualified.

[0030] The obtained clustering results are more accurate and can obtain more accurate production quality assessment results.

[0031] The present invention has the following beneficial effects: the purpose of the present invention is to calculate the degree of splitting of each data item, further obtain the splitting fitness of each data in each data sequence, obtain the splitting points of each data sequence based on the distribution fitness, and determine the number of final splitting points of each data item in combination with the preferred degree of each splitting point in each data sequence, and finally obtain a more accurate classification number to cluster the quality data, so that a more accurate production quality assessment result can be obtained based on the accurate clustering result. BRIEF DESCRIPTION OF THE DRAWINGS

[0032] By reading the following detailed description with reference to the accompanying drawings, the above and other objects, features and advantages of the exemplary embodiments of the present invention will become readily understood. In the accompanying drawings, several embodiments of the present invention are shown in an exemplary and non-limiting manner, and the same or corresponding reference numerals represent the same or corresponding parts, wherein:

[0033] Figure 1 It is a step flow chart of a lithium carbonate production quality data analysis method according to an embodiment of the present invention. DETAILED DESCRIPTION

[0034] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative work are within the scope of protection of the present invention.

[0035] The specific embodiments of the present invention are described in detail below with reference to the accompanying drawings.

[0036] See also Figure 1 , which shows a flow chart of the steps of a lithium carbonate production quality data analysis method provided by one embodiment of the present invention, the method comprising the following steps:

[0037] S001. Collect various quality data during the lithium carbonate production process.

[0038] It should be noted that by collecting various quality data in the lithium carbonate production process and performing cluster analysis on the various quality data, the production quality of lithium carbonate can be graded and evaluated.

[0039] In an embodiment of the present invention, for any batch of production raw materials, data such as lithium content, reaction temperature, reactor pressure, reaction duration, product lithium purity, etc. are obtained as a quality data. Similarly, based on several batches of production raw materials, several quality data are obtained, and each quality data contains several items of data. Each item of all quality data is standardized by a linear normalization algorithm.

[0040] S002. Obtain each data differential sequence according to each data sequence, and obtain the degree of splitting of each data according to each data differential sequence; obtain the splitting fitness of each data in each data sequence based on the splitting fitness; obtain multiple splitting points in each data sequence according to the splitting fitness.

[0041] It should be noted that in the process of analyzing the quality of lithium carbonate production by classifying the various quality data in the lithium carbonate production process through the conventional K-means algorithm, since the analysis of the various quality data in the lithium carbonate production process requires a huge amount of calculation, it is difficult to determine the number of classifications when classifying the various quality data in the lithium carbonate production process. If the number of classifications is too large, the classification between clusters will be meaningless, resulting in over-segmentation, and too few classifications will lead to under-segmentation. Therefore, too many and too few classifications will cause inaccurate classification, thereby affecting the accuracy of the quality analysis of lithium carbonate production. Therefore, the present invention needs to analyze the distribution of each data in all quality data, obtain the number of splitting points of each data, and then determine the number of classifications when classifying each quality data in the lithium carbonate production process.

[0042] It should be further explained that, for any item of all quality data, the more uneven its distribution is, the more this item of data in all quality data can be split. Therefore, the degree of splitting of each data is obtained according to the distribution of this item of data in all quality data. For any item of data, the more uneven the interval between adjacent data is, the more data aggregation exists in this item of data. Then, the overall distribution of this item of data is uneven, and its degree of splitting is greater; the greater the difference between each data in this item of data and the data as a whole, the more uneven the overall distribution of this data is, and its degree of splitting is greater.

[0043] In an embodiment of the present invention, the j-th data of all quality data are arranged in ascending order to obtain the j-th data sequence, and the absolute values ​​of the differences between all adjacent data in the j-th data sequence are obtained in turn to obtain the j-th data differential sequence.

[0044] The preset number of data N=5. In other embodiments, the implementer may preset the value of N according to the specific implementation situation, and use the first N data and the last N data of each data in the j-th data differential sequence as the neighborhood data of each data in the j-th data differential sequence. It should be noted that if N data is not met, it can be discarded.

[0045] Get the splitting degree of each data:

[0046] ;

[0047] In the formula, Represents the degree of splitting of the j-th data; Represents the value of the kth data in the jth data difference sequence; Represents the number of data in the j-th data difference sequence; Represents the mean of the neighborhood data of the kth data in the jth data difference sequence; Represents the mean of all data in the j-th data difference sequence; Represents the mean difference between the kth data and its neighboring data in the jth data difference sequence. The smaller the value, the smaller the average difference between the kth data and its five neighboring data, indicating that there is aggregation in the local area of ​​the kth data. Then the overall distribution of the data in the jth data sequence is uneven, indicating that the jth data is more divisible; || represents the absolute value symbol; Represents the difference between the kth data in the jth data difference sequence and the mean of all data in the jth data difference sequence. The larger the value, the more uneven the data distribution in the jth data sequence is, and the more the data should be split;

[0048] It should be noted that the degree to which each data item can be split only calculates the degree to which each data item should be split, and the specific data on which the data item can be split is determined based on the performance of each data item in each data sequence.

[0049] In the embodiment of the present invention, the split fitness of the i-th data in the j-th data sequence is obtained:

[0050] ;

[0051] In the formula, Represents the split fitness of the i-th data in the j-th data sequence; Represents the degree of splitting of the j-th data; Represents the number of data in the j-th data sequence that is greater than or equal to its i-th data; Represents the mean of all data in the j-th data sequence that is greater than or equal to its i-th data; Represents the number of data in the j-th data sequence that is smaller than the i-th data; Represents the mean of all data in the j-th data sequence that are less than the i-th data; Represents the degree of centering of the i-th data in the j-th data sequence. The larger its value is, the closer the i-th data in the j-th data sequence is to the middle data, and the greater its split fitness is; Represents the difference between the average values ​​of the two parts after the j-th data sequence is divided into two parts using the i-th data. The larger the difference, the better the segmentation effect. The larger the value is, the more splittable the j-th data is, and the larger the value is when calculating its split fitness.

[0052] In the embodiment of the present invention, the split fitness threshold T is preset to 0.7. In other embodiments, the implementer may preset the value of the split fitness threshold T according to the specific implementation situation, and record the data with a split fitness greater than the split fitness threshold T in the j-th data sequence as a suspected split point;

[0053] If there are multiple suspected splitting points appearing consecutively in the j-th data sequence, and the number of the suspected splitting points appearing consecutively is an even number, any one of the suspected splitting points is recorded as the splitting point; if there are multiple suspected splitting points appearing consecutively, and the number of the suspected splitting points appearing consecutively is an odd number, the suspected splitting point located at the middle position is recorded as the splitting point; if the suspected splitting point exists alone, it is recorded as the splitting point, and multiple splitting points in the j-th data sequence are obtained.

[0054] S003. Obtain the preference level of each splitting point in each data sequence, and based on the preference level, obtain the final number of splitting points of each data sequence; and obtain the number of clustering clusters according to the final number of splitting points of each data sequence.

[0055] It should be noted that, for multiple splitting points in the j-th data sequence, there may be a situation where the distance between two splitting points is too close. If the distance between the splitting points is too close, there may be an over-segmentation situation, which should be eliminated. Therefore, according to the distance distribution between each splitting point in the j-th data sequence and its adjacent splitting points, the preference degree of each splitting point in the j-th data sequence is obtained, and then the splitting points are eliminated according to the preference degree of each splitting point in the j-th data sequence.

[0056] In the embodiment of the present invention, the data segments formed by all adjacent splitting points in the j-th data sequence are recorded as all splitting segments in the j-th data sequence;

[0057] Get the preference of the vth split point in the jth data sequence:

[0058] ;

[0059] In the formula, Represents the preference of the vth split point in the jth data sequence; Represents the value of the vth split point in the jth data sequence; Represents the value of the v-1th split point in the jth data sequence; Represents the value of the v+1th split point in the jth data sequence; Represents the mean of all split segments in the j-th data sequence;

[0060] Represents the ratio of the difference between the vth split point and its previous split point in the jth data sequence to the difference between the split point after the vth split point and its previous split point. The larger the gap is, the larger the gap is between the data segment of the vth split point and its previous split point and the data segment of the next split point, which means that the vth split point is more likely to be removed; Represents the uniformity between the data segment of the vth split point and the data segment of the previous split point and the data segment of the next split point. The larger the value, the higher the preference of the split point.

[0061] It represents the ratio of the sum of the data segments on both sides of the v-th split point to twice the mean of all the split segments in the j-th data sequence. The smaller the ratio is, the smaller the length of the data segments on both sides of the v-th split point is, and the more they can be merged into the same data segment, that is, the v-th split point should be eliminated, and the less preferred the v-th split point is.

[0062] The preset preference threshold T1=0.8. The implementer can preset the value of the preference threshold according to the specific implementation situation, and record the number of splitting points in the j-th data sequence whose preference is greater than the preference threshold T1 as the final number of splitting points in the j-th data sequence.

[0063] The mean of the number of final split points of each data sequence is obtained and recorded as the number of clusters. The number of clusters is the number of classifications mentioned above.

[0064] S004. Clustering each quality data according to the number of clusters to obtain a clustering result, and evaluating the quality of lithium carbonate according to the clustering result.

[0065] It should be noted that, since similar quality data of lithium carbonate are classified into one category by clustering the quality data of lithium carbonate, the quality data in each category of the clustering results can be evaluated, and then the production quality of lithium carbonate can be graded and evaluated based on the quality evaluation results of each category of the clustering results.

[0066] In the embodiment of the present invention, all quality data are classified using the k-means clustering algorithm according to the number of clusters, and each classification of the clustering results can better adapt to the splitting requirements of each data in each quality data.

[0067] Each classification in the clustering results is input into the existing lithium carbonate production quality assessment model to obtain the quality assessment result of each classification in the clustering results. If the quality assessment result of any classification of the clustering results is qualified, the quality data in this classification of the clustering results is qualified, and the production raw materials (lithium carbonate) corresponding to the quality data in this classification of the clustering results are qualified; otherwise, the production raw materials corresponding to the quality data in this classification of the clustering results are unqualified.

[0068] The specific training process of the lithium carbonate production quality assessment model is as follows:

[0069] Each category of the clustering result is input into the trained neural network to obtain the quality evaluation result of each category of the clustering result; the neural network used in this embodiment is Resnet50, and the method for obtaining the data set for training the neural network is:

[0070] A large amount of clustering results of quality data are collected, and the quality evaluation results of each category of the clustering results are manually marked, and the marked result is recorded as the label of the clustering result of the quality data; a large amount of quality data and the labels of the corresponding clustering results are collected to form a data set; the neural network is trained using the data set, and the loss function used in the training process is a cross loss function; the specific training process is a well-known content of the neural network, and this embodiment will not repeat the specific training process.

[0071] The specific method of the artificial marking is:

[0072] For any item of quality data in any category in the clustering results, the mean of the item of quality data in the category is recorded as the first mean. If the first mean is within the standard range of the item, the item is recorded as qualified item data; if all items of quality data in the category are qualified production data, the quality assessment result of the category is recorded as qualified, otherwise, the quality assessment result of the category is recorded as unqualified. The standard range of any item of quality data is a priori knowledge and can be obtained through the lithium carbonate quality description document.

[0073] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the principles of the present invention should be included in the protection scope of the present invention.

Claims

1. A lithium carbonate production quality data analysis method, characterized in that: include: Collect each quality data, each quality data includes several items of data; obtain each data sequence; obtain the split fitness of each data in each data sequence , Represents the split fitness of the i-th data in the j-th data sequence; Represents the degree of splitting of the j-th data; Represents the number of data in the j-th data sequence that is greater than or equal to its i-th data; Represents the mean of all data in the j-th data sequence that is greater than or equal to its i-th data; Represents the number of data in the j-th data sequence that is smaller than the i-th data; represents the mean of all data in the j-th data sequence that is smaller than the i-th data; obtaining each splitting point in each data sequence, wherein the splitting fitness of the splitting point is greater than a preset splitting fitness threshold; Obtaining the preference degree of each splitting point in each data sequence; obtaining the final number of splitting points of each data sequence, wherein the preference degree of the final splitting point is greater than a preset preference degree threshold; clustering each quality data according to the final number of splitting points of each data sequence, obtaining a clustering result, and evaluating the quality of lithium carbonate.

2. A lithium carbonate production quality data analysis method according to claim 1, characterized in that, The step of obtaining each data sequence includes: Arrange the j-th item of all quality data in ascending order to obtain the j-th data sequence.

3. A lithium carbonate production quality data analysis method according to claim 1, characterized in that, The acquisition of the degree of splitting includes: Get the difference sequence of each data; ； In the formula, Represents the degree of splitting of the j-th data; Represents the value of the kth data in the jth data difference sequence; Represents the number of data in the j-th data difference sequence; Represents the mean of the neighborhood data of the kth data in the jth data difference sequence; Represents the mean of all data in the j-th data difference sequence; || represents the absolute value symbol.

4. A lithium carbonate production quality data analysis method according to claim 3, characterized in that, The step of obtaining each data difference sequence includes: The absolute values ​​of the differences between all adjacent data in the j-th data sequence are obtained in sequence to obtain the j-th data difference sequence.

5. A lithium carbonate production quality data analysis method according to claim 1, characterized in that, The step of obtaining each splitting point in each data sequence includes: A split fitness threshold is preset, and the data in the j-th data sequence whose split fitness is greater than the split fitness threshold is recorded as a suspected split point; if there are multiple suspected split points appearing consecutively in the j-th data sequence, and the number of the consecutive suspected split points is an even number, any one of the suspected split points is recorded as a split point; if there are multiple suspected split points appearing consecutively, and the number of the consecutive suspected split points is an odd number, the suspected split point located at the middle position is recorded as the split point; if the suspected split point exists alone, it is recorded as a split point, and multiple split points in the j-th data sequence are obtained.

6. A lithium carbonate production quality data analysis method according to claim 1, characterized in that, The obtaining of the preferred degree of each splitting point in each data sequence includes: ； In the formula, Represents the preference of the vth split point in the jth data sequence; Represents the value of the vth split point in the jth data sequence; Represents the value of the v-1th split point in the jth data sequence; Represents the value of the v+1th split point in the jth data sequence; Represents the mean of all split segments in the j-th data sequence.

7. A lithium carbonate production quality data analysis method according to claim 6, characterized in that, The acquisition of all split segments in the j-th data sequence includes: The data segments formed by all adjacent splitting points in the j-th data sequence are recorded as all splitting segments in the j-th data sequence.

8. A lithium carbonate production quality data analysis method according to claim 1, characterized in that, The method clusters the quality data according to the final number of split points of each data sequence to obtain a clustering result, and evaluates the quality of lithium carbonate, including: Obtain the mean of the final number of split points of each data sequence, record it as the number of clusters, and classify all quality data using the k-means clustering algorithm according to the number of clusters to obtain each classification of the clustering results; Each classification of the clustering results is input into the existing lithium carbonate production quality assessment model. If the quality assessment result of any classification of the clustering results is qualified, the production raw material lithium carbonate corresponding to the quality data in this classification of the clustering results is qualified; otherwise, the production raw material lithium carbonate corresponding to the quality data in this classification of the clustering results is unqualified.

Images