Electronic device and method for analyzing battery performance influencing factor thereof

Abstract

Provided is a method for analyzing a battery performance influencing factor of an electronic device, the method comprising the steps of: identifying a causal graph indicating a relationship between a plurality of factors related to a battery manufacturing process and process data including, as each data instance, data on the plurality of factors identified for each execution of the battery manufacturing process; and identifying influence information on the characteristics of a final product for each of the plurality of factors on the basis of the causal graph, the process data, and a model trained to predict the characteristics of the final product of the battery manufacturing process by receiving the process data.

Description

Method for analyzing factors affecting the performance of electronic devices and their batteries

[0001] The present disclosure relates to an electronic device and a method for analyzing factors affecting battery performance thereof.

[0002] The battery manufacturing process involves a complex, nonlinear interplay of various sensors and process factors, making it difficult to identify the variables that significantly impact the process. Research has been conducted to identify factors that significantly impact battery cell performance using machine learning and explainable artificial intelligence (XAI) algorithms. However, these prior techniques are problematic in that they assume independence between unit processes within the overall process and conduct analyses. Specifically, this approach, which assumes independence between unit processes, fails to reflect the characteristics of the battery manufacturing process, in which unit processes responsible for multiple mechanisms are sequentially connected, and provides analyses that deviate from actual system behavior.

[0003] If based on such flawed analytical methods, key factors among the numerous factors that comprise the battery process will be incorrectly identified, making efficient quality control and product design difficult. Therefore, it is necessary to explore data-driven interpretation methods that can reflect the causal relationships between these process factors and measured values, thereby improving the accuracy of determining key factor contributions.

[0004] The disclosed embodiments provide a method for analyzing factors affecting the performance of electronic devices and their batteries. Specifically, the purpose is to provide a data-based interpretation method capable of reflecting causal relationships between process factors and measured values, thereby improving the accuracy of determining key factor contributions.

[0005] The technical tasks to be achieved by this embodiment are not limited to the technical tasks described above, and other technical tasks can be inferred from the following embodiments.

[0006] One aspect of the present disclosure provides a method for analyzing battery performance influencing factors of an electronic device, comprising: a step of identifying process data including, as data instances, a causal graph representing relationships between a plurality of factors related to a battery manufacturing process and data for the plurality of factors identified for each execution of the battery manufacturing process; and a step of identifying information on influences on the characteristics of the final product for each of the plurality of factors based on the causal graph, the process data, and a model learned to predict the characteristics of the final product of the battery manufacturing process by inputting the process data, wherein the causal graph represents a graph including: a property node unidirectionally connected to correspond to the order of each sub-process included in the battery manufacturing process, each property node corresponding to a characteristic of an intermediate product and the final product produced for each of the sub-processes; and a plurality of process variable nodes unidirectionally connected toward each of the property nodes, each process variable corresponding to a process variable controlled in each of the sub-processes corresponding to each of the property nodes.

[0007] In one embodiment of the present disclosure, a method for analyzing battery performance influencing factors may be included, including: a step of identifying a plurality of paths connecting the plurality of process variable nodes to a property node corresponding to a characteristic of the final product based on the causal graph; and a step of identifying a virtual data instance by adjusting at least a portion of the data of the data instance included in the process data based on the identified plurality of paths; a step of inputting the virtual data instance into the model; and a step of identifying the influence information by analyzing an output of the model for the virtual data instance.

[0008] In addition, in one embodiment of the present disclosure, the step of identifying the virtual data instance may include a step of identifying at least one first path including an edge between a first node corresponding to a first factor and a second node corresponding to a child node of the first node on the causal graph based on the plurality of paths; and a step of identifying at least one first virtual data instance based on the at least one first path, and the step of identifying the influence information may include a step of identifying first value information for the first path based on the first virtual data instance and the model; and a step of identifying an influence value of the first factor for a second factor corresponding to the second node based on the first value information.

[0009] In addition, in one embodiment of the present disclosure, the step of identifying at least one first virtual data instance includes: identifying at least one path located in a previous order than the at least one first path among a plurality of permutations in which the plurality of paths are randomly listed; identifying, for each of the plurality of permutations, a first subset including the at least one path located in a previous order than the at least one first path and a second subset in which the first path is added to the first subset; and identifying a 1-1 virtual data instance by randomly sampling and adjusting data of a node that does not correspond to the first subset among the data instances, and identifying a 1-2 virtual data instance by randomly sampling and adjusting data of a node that does not correspond to the second subset among the data instances, wherein the step of identifying first value information for the first path includes: identifying the 1-1 and 1-2 value information based on a result of inputting the 1-1 and 1-2 virtual data instances identified for each of the plurality of permutations into the model; And it may include a method for analyzing battery performance influencing factors, including a step of confirming the first value information based on difference information between the first-1 and first-2 value information.

[0010] In addition, in one embodiment of the present disclosure, the step of confirming the first value information may include a method for analyzing a battery performance influencing factor, including a step of confirming the first value information for the first path based on average information of the difference information between the first-1 value information and the first-2 value information confirmed for each of the plurality of permutations.

[0011] In addition, in one embodiment of the present disclosure, the step of confirming the first value information may include a method for analyzing battery performance influencing factors, including a step of confirming the first-first output information by inputting the first-first virtual data instance into the model; and a step of confirming the first-first value information based on difference information between the average information of the confirmed output information and the first-first output information by inputting the unadjusted data instance into the model.

[0012] In addition, in one embodiment of the present disclosure, the step of checking the influence information of the first factor may include a battery performance influence factor analysis method including a step of checking the first influence value of the first factor on the second factor based on the sum information of the first value information for the first path including the edge between the first and second nodes.

[0013] In addition, in one embodiment of the present disclosure, the step of checking the influence information may include a method for analyzing battery performance influence factors, including a step of checking a matrix including, as each component, an influence value for a first factor corresponding to a first coordinate of a first axis and a second factor corresponding to a second coordinate of a second axis; and a step of checking the influence information including direct influence information and indirect influence information on the characteristics of the final product of the plurality of factors based on the matrix.

[0014] In addition, in one embodiment of the present disclosure, the step of checking the influence information may include a method for analyzing battery performance influence factors, including a step of checking the influence value of at least one factor corresponding to at least one node directly connected to a node corresponding to a characteristic of the final product on the matrix; and a step of checking the direct influence information based on the influence value of the at least one factor.

[0015] In addition, in one embodiment of the present disclosure, the step of checking the influence value of the at least one factor may include a method for analyzing battery performance influencing factors, including: checking a coordinate corresponding to a characteristic of the final product on the second axis; checking a vector on the matrix parallel to the first axis, the vector having a coordinate corresponding to the characteristic of the final product with respect to the second axis; and checking the influence value of the at least one factor based on the vector.

[0016] In addition, in one embodiment of the present disclosure, the step of checking the influence information may include a method for analyzing battery performance influence factors, including a step of checking the sum information of influence values ​​for each factor corresponding to each of the process variable nodes; and a step of checking the indirect influence information for each factor corresponding to each of the process variable nodes based on the sum information.

[0017] In addition, in one embodiment of the present disclosure, the step of checking the sum information may include a battery performance influence factor analysis method including the steps of: checking each coordinate corresponding to each process variable on the first axis; checking each vector on the matrix parallel to the second axis with each coordinate corresponding to each process variable for the first axis; and checking each sum information of influence values ​​included in each vector for each factor corresponding to each process variable node.

[0018] In addition, in one embodiment of the present disclosure, a battery performance influence factor analysis method may be included, further comprising a step of checking contribution ratio information indicating a ratio of values ​​of the direct influence information and the indirect influence information for each of the plurality of factors based on the influence information.

[0019] In addition, in one embodiment of the present disclosure, a method for analyzing battery performance influencing factors may be included, further including a step of supporting control of the characteristics of the intermediate product based on the direct influence information.

[0020] In addition, in one embodiment of the present disclosure, a method for analyzing battery performance influencing factors may be included, further including a step of supporting control of the process variables for each detailed process based on the indirect influence information.

[0021] In addition, in one embodiment of the present disclosure, the step of checking the causal graph and the process data may include a method for analyzing battery performance influencing factors, including: checking the entire process data checked for each execution of the battery manufacturing process; calculating a correlation coefficient between all factors of the battery manufacturing process based on the entire process data; checking the plurality of factors by filtering at least some of the factors whose correlation coefficient is greater than or equal to a threshold value among the entire factors; and checking at least some of the data corresponding to the plurality of factors among the entire process data as the process data.

[0022] In addition, in one embodiment of the present disclosure, the step of identifying at least some of the data corresponding to the plurality of factors among the entire process data as the process data may include a method for analyzing battery performance influencing factors, including the step of identifying, among the data instances included in the entire process data, a data instance in which the measurement temperature of the characteristic of the final product corresponds to an effective temperature range as the process data.

[0023] In addition, in one embodiment of the present disclosure, a method for analyzing battery performance influencing factors may be included, including a step of inputting the process data into the model for each data instance; and a step of training the model based on a result of comparing the output of the model with actual data corresponding to the characteristics of the final product identified for each data instance.

[0024] Another aspect of the present disclosure provides an electronic device for analyzing battery performance influencing factors, comprising: a processor; and a memory storing one or more instructions, wherein the processor, by executing the one or more instructions, identifies a causal graph representing relationships between a plurality of factors related to a battery manufacturing process and process data including data for the plurality of factors identified for each run of the battery manufacturing process as each data instance, and identifies influence information on the characteristics of the final product for each of the plurality of factors based on a model learned by receiving the causal graph, the process data, and the process data to predict characteristics of the final product of the battery manufacturing process, wherein the causal graph represents a graph including a property node unidirectionally corresponding to the order of each sub-process included in the battery manufacturing process and corresponding to characteristics of an intermediate product and the final product generated for each of the sub-processes; and a plurality of process variable nodes unidirectionally connected toward each of the property nodes and corresponding to process variables controlled in each of the sub-processes corresponding to each of the property nodes.

[0025] Another aspect of the present disclosure may provide a non-transitory computer-readable recording medium having recorded thereon a program for executing the above-described battery performance influencing factor analysis method on a computer.

[0026] Specific details of other embodiments are included in the detailed description and drawings.

[0027] According to the proposed embodiment, one or more of the following effects can be expected.

[0028] According to the embodiments of this specification, the accuracy of determining the contribution of major factors can be improved through a data-based interpretation method that can reflect the causal relationship between process factors and measured values.

[0029] In addition, according to the embodiments of this specification, the influence of process factors can be interpreted by dividing them into direct and indirect influences depending on the purpose.

[0030] The effects of the present disclosure are not limited to the effects mentioned above, and other effects not mentioned will be clearly understood by those skilled in the art from the description of the claims.

[0031] FIG. 1 is a diagram showing an example of how an electronic device for analyzing battery performance influencing factors operates according to one embodiment.

[0032] Figure 2 is a flowchart illustrating a method for analyzing battery performance influencing factors according to one embodiment.

[0033] Figure 3 is an example diagram of a causal graph according to one embodiment.

[0034] FIG. 4 is an exemplary diagram of a path connecting a process variable node on a causal graph to a property node corresponding to a characteristic of a final product according to one embodiment.

[0035] FIG. 5 is a diagram illustrating an example of a matrix including influence values ​​according to one embodiment.

[0036] FIG. 6 is a diagram illustrating an example of confirming direct influence information using a matrix according to one embodiment.

[0037] FIG. 7 is a diagram illustrating an example of checking indirect influence information using a matrix according to one embodiment.

[0038] Figures 8a and 8b are graphs showing direct and indirect influence information according to one embodiment.

[0039] Figure 9 is an example diagram of contribution ratio information according to one embodiment.

[0040] Fig. 10 illustrates a block diagram of an electronic device according to one embodiment.

[0041] The terms used in the examples have been selected from widely used, current terms, taking into account the functions of the present disclosure. However, these terms may vary depending on the intentions of those skilled in the art, precedents, the emergence of new technologies, etc. Furthermore, in certain cases, terms may be arbitrarily selected by the applicant, in which case their meanings will be described in detail in the relevant description. Therefore, the terms used in this disclosure should not be defined simply as names, but rather based on the meanings of the terms and the overall content of the present disclosure.

[0042] When a part of a specification is said to "include" a component, this does not mean that it excludes other components, but rather that it may include other components, unless otherwise stated.

[0043] The expression "at least one of a, b, and c" described throughout the specification may encompass 'a alone', 'b alone', 'c alone', 'a and b', 'a and c', 'b and c', or 'all of a, b, and c'.

[0044] The "terminal" mentioned below may be implemented as a computer or portable terminal capable of connecting to a server or other terminal via a network. Here, the computer includes, for example, a notebook, desktop, or laptop equipped with a web browser, and the portable terminal may include, for example, a wireless communication device that guarantees portability and mobility, such as a communication-based terminal such as IMT (International Mobile Telecommunication), CDMA (Code Division Multiple Access), W-CDMA (W-Code Division Multiple Access), LTE (Long Term Evolution), and all types of handheld-based wireless communication devices such as smartphones and tablet PCs.

[0045] Below, embodiments of the present disclosure are described in detail with reference to the attached drawings so that those skilled in the art can easily implement the present disclosure. However, the present disclosure may be implemented in various different forms and is not limited to the embodiments described herein.

[0046] Hereinafter, embodiments of the present disclosure will be described in detail with reference to the drawings.

[0047] FIG. 1 is a diagram showing an example of how an electronic device for analyzing battery performance influencing factors operates according to one embodiment.

[0048] Referring to FIG. 1, the electronic device (100) can operate by receiving causal graphs and process data, as will be described later. In addition, the electronic device (100) may also include a model (110).

[0049] Here, the model (110) can be implemented using various algorithms such as machine learning algorithms that can reflect nonlinear correlations between process factors, such as XGBoost, Random Forest, or Deep Neural Network (DNN), but is not limited to those listed above. Such a model (110) may be a model trained to receive data corresponding to factors corresponding to process variables (Process Parameters, PP) controlled in each detailed process of the battery manufacturing process and data corresponding to factors corresponding to the characteristics of intermediate products (Intermediate Product Features, IPF) generated for each detailed process, and output the characteristics (Final Product Properties, FPP) of the final product generated by the entire battery manufacturing process, and the learning method may be determined in relation to a model construction method according to the various algorithms described above or other algorithms.

[0050] Meanwhile, only components related to the present embodiment are illustrated in FIG. 1. Therefore, those skilled in the art will understand that other general components may be included in addition to the components illustrated in FIG. 1.

[0051] Figure 2 is a flowchart illustrating a method for analyzing battery performance influencing factors according to one embodiment.

[0052] In step S210, the electronic device (100) can identify process data including a causal graph representing relationships between multiple factors related to the battery manufacturing process and data for the multiple factors identified for each execution of the battery manufacturing process as each data instance. In step S220, the electronic device (100) can identify information on the influence of each of the multiple factors on the characteristics of the final product based on a model (110) trained to predict the characteristics of the final product of the battery manufacturing process by receiving the causal graph, the process data, and the process data. Hereinafter, a method for analyzing battery performance influencing factors according to an embodiment of the present disclosure will be described in detail.

[0053] First, the electronic device (100) can obtain a causal graph. According to one embodiment, the causal graph may be input into the electronic device (100) by a user with domain knowledge related to the battery manufacturing process, transmitted from another server and input into the electronic device (100), or obtained by the electronic device (100) analyzing process factors related to the battery manufacturing process.

[0054] Such a causal graph may include a unidirectionally connected property node corresponding to the order of each sub-process included in the battery manufacturing process, which corresponds to the characteristics of intermediate products and final products produced in each sub-process, and a plurality of process variable nodes corresponding to process variables controlled in each sub-process corresponding to each property node, which are unidirectionally connected to each property node. As can be inferred from the above explanation, the causal graph may be a type of directed acyclic graph (DAG).

[0055] Here, examples of each detailed process included in the battery manufacturing process include a mixing process, a coating process, a roll pressing process, a slitting and notching process, an assembly process, an activation process, an aging process, and a degassing process, but are not limited to the processes listed above, and do not have to include all of the processes listed above. In addition, examples of process variables controlled in each detailed process include slurry temperature, slurry flow rate, pump, coating gap length, drying temperature and air volume, coating speed, coating thickness, and coating loading for the coating process; rolling roll gap length, rolling pressure, rolling speed, and rolling thickness for the rolling process; lamination plate temperature, lamination force, lamination roller temperature, lamination speed, and dimension for the assembly process; and J / F temperature and pressure, aging time, inter-process waiting time, electrolyte amount, performance measurement temperature, and final cell thickness for the activation process, post-process, and overall process. This is also not limited to what is listed above, and does not have to include everything listed above. In addition, examples of the characteristics of the intermediate product include rolling ratio, loading after rolling, negative electrode thickness after activation, loading amount of one monocell, and void amount of the positive and negative electrodes after coating / rolling / activation, and examples of the characteristics of the final product include discharge capacity, discharge capacity fitted to the measurement temperature, and charge resistance, but is not limited to what is listed above, and does not have to include everything listed above.

[0056] According to one embodiment, the battery performance influencing factor analysis method of the present disclosure may be performed by selecting some of the overall factors related to the battery manufacturing process. For example, the electronic device (100) may check the overall process data confirmed for each run of the battery manufacturing process. Thereafter, the electronic device (100) may calculate a correlation coefficient between the overall factors of the battery manufacturing process based on the overall process data. The electronic device (100) may check a plurality of factors by filtering at least some of the factors whose correlation coefficients are greater than or equal to a threshold among the overall factors. In other words, among the factors whose correlation coefficients are greater than or equal to the threshold, only representative factors may be retained and the rest may be filtered. For example, if the combination of factors A, B, C, and D whose correlation coefficients are greater than or equal to the threshold is one of these factors, for example, factor A, only one of these factors may be retained and the rest may be filtered. Here, according to one embodiment, the factor that is not filtered may be the one with the highest permutation importance within the group of factors whose correlation coefficients are greater than or equal to the threshold. According to one embodiment, the correlation coefficient may be calculated using various correlation coefficient calculation methods such as the Pearson correlation coefficient, Spearman rank correlation coefficient, or Kendall tau coefficient, but is not limited to the above-listed methods, and the threshold may be set as a boundary value that can be determined to have a significant correlation depending on the characteristics of each correlation coefficient calculation method. The aforementioned causal graph may be generated and input based on a selected portion of all factors related to the battery manufacturing process.

[0057] To examine an example of the causal graph described above, refer to Figure 3.

[0058] Figure 3 is an example diagram of a causal graph according to one embodiment.

[0059] Referring to Fig. 3, an example of a causal graph can be examined. In Fig. 3, one of the property nodes (310) (311) is It refers to a node corresponding to the characteristics of an intermediate product according to the process, (321) and (322) is It can mean a node corresponding to a process variable that can be controlled in relation to the process. For example, If the process is a rolling process, (321) is a node corresponding to the rolling pressure, (322) may denote a node corresponding to the rolling speed. In this way, a causal graph can be confirmed in the form of a unidirectional edge connecting from a process variable node (320) corresponding to a process variable that can be controlled in the respective detailed process to a material property node (310) corresponding to the characteristics of the intermediate product for each detailed process. In addition, as can be confirmed in Fig. 3, a unidirectional edge connected from noise to a node related to the characteristics of each intermediate product may exist to reflect measurement errors or other various errors. Meanwhile, Node (315) may be a node corresponding to the characteristics of the final product, Nodes (311, 312, 313, and 314) indicated as may be nodes corresponding to the characteristics of the intermediate product. The causal graph of Fig. 3 described above is merely an example and may be significantly simplified for the convenience of explanation. For example, There may be multiple processes, each process variable linked to multiple intermediate products, or a structure much more complex than that in Figure 3. The relationships between fields may be connected. In Fig. 3, (321) The edge (330) connected to (311) is marked for later explanation.

[0060] In parallel with the process of verifying the causal graph described above, the electronic device (100) can verify information about process data. The process data may include data on multiple factors identified for each run of the battery manufacturing process as individual data instances. Here, each data instance may refer to data on factors related to the process of manufacturing each cell. For example, a specific data instance related to a specific cell may include information on process variables for each detailed process set in the manufacturing process of the specific cell and the characteristics of intermediate products. The process data may include such data instances for each run of the battery manufacturing process.

[0061] Additionally, the process data may include data corresponding to a selected subset of factors, if the aforementioned causal graph is generated based on a selected subset of factors related to the battery manufacturing process. For example, if some of the overall factors of the battery manufacturing process are filtered based on correlation coefficients, the process data may include data for the unfiltered factors, per data instance. Accordingly, the electronic device (100) may identify at least a subset of the data corresponding to a plurality of selected factors among the overall process data as process data.

[0062] According to one embodiment, after filtering factors based on the correlation coefficient as described above, some data instances can be filtered based on the measurement temperature of the characteristics of the final product. That is, the electronic device (100) can identify, as process data, data instances in which the measurement temperature of the characteristics of the final product corresponds to an effective temperature range among the data instances included in the entire process data. Since temperature has a significant influence when measuring the characteristics of the final product, such an effective temperature range can be set in order to analyze performance-influencing factors based on the characteristics of the final product measured at similar temperatures within the effective temperature range. The effective temperature range can be set to a range in which the characteristics of the final product are not measured differently due to temperature, and can be set to a constant temperature range, such as 28.5 to 29.5 degrees Celsius, for example. Here, each data instance included in the process data may not include data on the characteristics of the final product according to the execution of the corresponding manufacturing process, and the model (110) may be trained by taking the data instance as input and using data on the characteristics of the actual measured final product corresponding to the data instance as the correct answer.

[0063] In conjunction with the causal graph and process data described above, an embodiment of using a model (110) to determine information on the impact of an electronic device (100) on battery performance by multiple factors is described below.

[0064] According to one embodiment, the electronic device (100) can identify multiple paths connecting a plurality of process variable nodes to a material node corresponding to the characteristics of the final product based on a causal graph. For example, the electronic device (100) can identify all possible paths for the causal graph. According to one embodiment, the electronic device (100) can identify multiple paths connecting a plurality of process variable nodes to a material node corresponding to the characteristics of the final product through a depth-first search (DFS). An example of identifying paths in this way for the example of FIG. 3 described above will be described with reference to FIG. 4.

[0065] Referring to Fig. 4, one of the process variable nodes (320) described in Fig. 3 An example (410) of the first path connecting from node (321) to FPP node (315) corresponding to the characteristics of the final product can be confirmed. Although not shown in Fig. 4 for convenience of explanation, multiple paths are similar to the first path (410) and each process variable node (320), i.e., Fig. 3. (322 to 329) may include a path leading to the FPP node (315). For convenience of explanation, in the following paragraphs, FIG. 3 Each path from (322 to 329) to the FPP node (315) is described as the second to ninth paths.

[0066] Thereafter, the electronic device (100) can verify a virtual data instance by adjusting at least some of the data of the data instance included in the process data based on the plurality of paths verified in this way. The virtual data instance may be data that is verified by randomly sampling within a batch (process data in the present disclosure) while using data corresponding to some factors of the data instance as is, and values ​​corresponding to the remaining factors. The electronic device (100) can input the virtual data instance verified in this way into the model (110), and analyze the output of the model (110) for the virtual data instance accordingly to verify the influence information. This process will be described in more detail below.

[0067] According to one embodiment, the electronic device (100) can check the influence value for each edge on the causal graph. Since the edge is unidirectional, the influence value of the corresponding edge can mean a numerical value of the influence that a factor corresponding to a parent node has on a factor corresponding to a child node. For example, the electronic device (100) calculates the influence value of the edge (330) of FIG. 3 described above, (321) Setting of process variables corresponding to nodes (311) It is possible to calculate the numerical value of the influence on the characteristics of the intermediate product corresponding to the node.

[0068] In order to calculate the influence value of an edge in this way, the electronic device (100) can identify at least one first path including an edge between a first node corresponding to a first factor and a second node corresponding to a child node of the first node on the causal graph. For example, the edge (330) of FIG. 3 described above may be included in the first path (410) of FIG. 4. In this example, there is only one path including the edge, but in other cases, for example, There may be multiple paths that contain edges of the liver.

[0069] The electronic device (100) can identify at least one first virtual data instance based on the first path. Hereinafter, a process for identifying the first virtual data instance according to one embodiment will be described.

[0070] According to one embodiment, the electronic device (100) can check multiple permutations in which multiple paths are randomly listed. For example, in the example of FIG. 4 described above, the electronic device (100) can check multiple permutations in which the first to ninth paths are randomly listed. The permutations of the dog can be confirmed. All such permutations can be confirmed, but an embodiment in which only some of them are sampled and confirmed is also possible. The electronic device (100) can confirm, for each permutation, at least one path located in a previous order than the first path described above. For example, if a certain permutation is composed of [second path, third path, first path, fourth path, ...], the electronic device (100) can confirm the second path and the third path as at least one path located in a previous order than the first path for the permutation. The electronic device (100) can confirm, for each of a plurality of permutations, a first subset including at least one path located in a previous order than the first path described above, and a second subset in which the first path is added to the first subset. For example, regarding the example described above, the first subset may be composed of [second path, third path], and the second subset may be composed of [second path, third path, first path].

[0071] Thereafter, the electronic device (100) can confirm the 1-1 virtual data instance by randomly sampling and adjusting data of nodes that do not correspond to the first subset of data instances, and can confirm the 1-2 virtual data instance by randomly sampling and adjusting data of nodes that do not correspond to the second subset of data instances. For example, if the above-described example is explained in combination with the path example of FIG. 4, for the first subset consisting of [the second path, the third path], the electronic device (100) leaves the values ​​for the nodes included in the path of the first subset as they are for each data instance, and adds the values ​​for the nodes that are not included in the path of the first subset. The values ​​of (321, 324, 325, 326, 327, 328, 329 and 312) can be randomly sampled from other data instances included in the process data to identify the 1-1 virtual data instance. For example, the 1-1 virtual data instance can be identified by performing this random sampling operation on all data instances, or the 1-1 virtual data instance can be identified by performing this operation on only some of the data instances. Similarly, for the 2nd subset, for each data instance, the values ​​corresponding to the nodes included in the path of the 2nd subset are left as they are, and the values ​​corresponding to the nodes not included in the path of the 2nd subset are The 1-2 virtual data instances can be identified by randomly sampling values ​​from other data instances for (324, 325, 326, 327, 328, 329, and 312). For the 1-2 virtual data instances, the same random sampling operation can be performed on all data instances to identify the 1-2 virtual data instances, or the same operation can be performed on some of the data instances to identify the 1-2 virtual data instances.

[0072] According to one embodiment, the electronic device (100) can verify the first-first and first-second value information based on the results of inputting the first-first and first-second virtual data instances into the model (110). Specifically, the electronic device (100) can input the first-first virtual data instance into the model (110) to verify the first-first output information. Thereafter, the electronic device (100) can input the unadjusted original data instances included in the process data into the model (110) to verify the difference information between the average information of the verified output information and the first-first output information, and verify this as the first-first value information. The average information of the verified output information by inputting the unadjusted original data instances into the model (110) may be pre-calculated. In addition, similarly, the electronic device (100) can input the first-second virtual data instance into the model (110) to verify the first-second output information. Thereafter, the electronic device (100) inputs the data instances of the unadjusted original data included in the process data into the model (110) to check the difference information between the average information of the confirmed output information and the first-second output information, and can check this as the first-second value information.

[0073] According to one embodiment, the electronic device (100) can calculate difference information between the first-1 and first-2 value information. Since the first subset and second subset, the first-1 and first-2 virtual data instances described above are identified for each permutation, the first-1 and first-2 value information can also be identified for each permutation. The electronic device (100) can calculate difference information between the first-1 and first-2 value information identified for each permutation. Thereafter, the electronic device (100) can calculate average information of the difference information between the first-1 and first-2 value information and identify it as the first value information for the first path. That is, to explain again with respect to the above-described example, the electronic device (100) can calculate the difference information between the 1-1 value information for the first subset consisting of [the second path, the third path] and the 1-2 value information for the second subset consisting of [the second path, the third path, the first path]. Accordingly, the difference information between the 1-1 and 1-2 value information for the corresponding permutation can be calculated, and in a similar manner, when the difference information between the 1-1 and 1-2 value information for each permutation is calculated, the average of such difference information can be identified as the 1-1 value information for the first path. The method described above, which will be described in more detail later with mathematical formulas, may be a suitable method for checking how much the values ​​included in the first path contributed when the model (110) receives a data instance and outputs the characteristics of the final product.

[0074] According to one embodiment, the electronic device (100) can check the first influence value of the first factor on the second factor based on the sum information of the first value information on the first path including the edge between the first and second nodes. Since the edge (330) according to the example of FIG. 3 described above is included in the first path (410) of FIG. 4, i.e., only one path, the first value information on the first path (410) can be checked as the first influence value of the first factor on the second factor. However, unlike this, in FIG. 4 and Since the edge connecting is included in all the paths 1 to 9, the sum of the value information of the paths 1 to 9 is the value of that edge, i.e. of intermediate products can be calculated as an influence value. For another example, in Fig. 4 and Since the edge connecting is included in the 5th and 6th paths, the sum of the value information of the 5th and 6th paths is the value of that edge, i.e. of intermediate products It can be calculated as an influence value on the intermediate product.

[0075] The above-described series of influence value calculation processes are explained below using mathematical equations 1 to 3. First, mathematical equation 1 related to the function for calculating value information is explained.

[0076]

[0077] In mathematical expression 1, means a set containing multiple permutations as described above, Is It can mean any one permutation included in . Also, is the first route, is a permutation Path within can mean the order of. In light of this definition, is a set that includes as elements a path that is in the same order as or faster than the first path within the permutation, and means the second subset mentioned above. It can be understood that refers to the first subset mentioned above as a set that includes a path that is in a faster order than the first path within the permutation as an element. In addition, as a function for calculating value information, as described above, it may be a function that receives a subset including at least one path as an input, leaves the values ​​corresponding to the nodes included in the path in the data instance as they are, and outputs information on the difference between the average of the output information when the values ​​of the factors corresponding to the nodes not included in the path are input to the model (110) by randomly sampling from other data instances of the process data and the average of the output information when the entire data instance of the process data is input to the model (110). is the mathematical formula 2 It is a function with the same purpose, but the format of the input value is slightly different, as explained below.

[0078]

[0079] In mathematical expression 2, represents a set (or coalition) that contains at least one node, Is Each value corresponding to the entire argument of the dog can mean, is a set About the possible combinations of variables not included in , which means that the function corresponding to the model (110) will be integrated, and the set It can be understood as being equivalent to the average of the results of inputting data instances that are randomly sampled from variables not included in the model (110). can mean the average of the model (110) output information for the entire process data set X. As can be seen from the interpretation of the above mathematical expression 2, v receives a set containing nodes as input, The meaning is slightly different in that it takes as input a set containing paths, but the usual technician takes as input a set containing paths from Equation 2. I think you will be able to understand about it.

[0080]

[0081] In mathematical expression 3, is a causal graph may mean any of the aforementioned multiple permutations, is the path Ga edge It can be a function that has a value of 1 only if it contains , and 0 otherwise. is the path It can mean value information about. According to mathematical formula 3, is the edge All paths containing The output value is the sum of the value information for the edge described above. Same definition as described for the influence value.

[0082] The calculation of influence values ​​according to the mathematical formula described above is adapted from the Shapley Value, a value derived from game theory and used to assess the fair contribution of each participant. In other words, the influence value of each edge can be understood as a numerical value that evaluates the contribution of each edge, similar to the Shapley Value.

[0083] According to the embodiment described above, when the influence value for the factor corresponding to the child node of the factor corresponding to each edge, that is, each parent node connected through the edge, is calculated, the electronic device (100) can check a matrix including the same. For example, the electronic device (100) can check a matrix including, as each component, the influence value of the first factor corresponding to the first coordinate of the first axis and the influence value of the second factor corresponding to the second coordinate of the second axis. Since the influence value of the first factor corresponding to the first coordinate of the first axis and the influence value of the second factor corresponding to the second coordinate of the second axis are included as components, the matrix has the form of a triangular matrix, and the diagonal components and the components indicating the influence values ​​between nodes that are not directly connected to each other can be automatically input as 0. Refer to FIG. 5 to explain such an example.

[0084] FIG. 5 is a diagram illustrating an example of a matrix including influence values ​​according to one embodiment.

[0085] Referring to FIG. 5, the matrix may include a vertical axis corresponding to a parent node and a horizontal axis corresponding to a child node, and each component may mean an influence value for a factor corresponding to the coordinate of the horizontal axis of the factor corresponding to the coordinate of the vertical axis of the corresponding component. For example, an example component (501) may include: The argument It can mean an influence value for the factor. In the example of Fig. 5, the first axis can correspond to the vertical axis, and the second axis can correspond to the horizontal axis.

[0086] According to one embodiment, the electronic device (100) can determine influence information including direct influence information and indirect influence information on the characteristics of the final product of multiple factors based on the aforementioned matrix.

[0087] First, regarding the direct influence information, the electronic device (100) can check the influence value of at least one factor corresponding to at least one node directly connected to a node corresponding to the characteristics of the final product on the matrix. For example, according to the example of FIG. 3 described above, the electronic device (100) and of The influence value can be confirmed. The electronic device (100) can directly confirm influence information based on at least one influence value confirmed in this way. The influence value of at least one factor corresponding to at least one directly connected node described above can be easily confirmed through a matrix. That is, the electronic device (100) can confirm a coordinate corresponding to a characteristic of the final product on a second axis corresponding to a child node of the matrix, confirm a vector on the matrix that has a coordinate corresponding to a characteristic of the final product with respect to the second axis and is parallel to the first axis, and confirm the influence value of at least one factor corresponding to at least one directly connected node described above based on the vector. More specifically, the components included in the vector on the confirmed matrix can be confirmed as an influence value of at least one factor corresponding to at least one directly connected node described above. Refer to FIG. 6 to examine an example of such a vector.

[0088] FIG. 6 is a diagram illustrating an example of confirming direct influence information using a matrix according to one embodiment.

[0089] Referring to FIG. 6, the electronic device (100) can confirm that the coordinate on the second axis corresponding to the FPP is the coordinate at the right end, and can confirm a vector (600) having the corresponding coordinate and parallel to the first axis. The electronic device (100) can confirm the values ​​included in the vector (600) as influence values ​​of at least one factor corresponding to at least one directly connected node.

[0090] Next, regarding indirect influence information, the electronic device (100) can check the sum information of influence values ​​for each factor corresponding to each process variable node in the matrix. And, based on the sum information, indirect influence information for each factor corresponding to each process variable node can be checked. For example, according to the example of FIG. 3 described above, the electronic device (100) can: The sum information of the influence values ​​can be confirmed for each factor. The electronic device (100) can confirm the indirect influence information based on the sum information of at least one influence value confirmed for each factor in this way. The influence value for each factor corresponding to each process variable node described above can be easily confirmed through a matrix. For example, the electronic device (100) can confirm each coordinate corresponding to each process variable on the first axis corresponding to the parent node in the matrix. The electronic device (100) can confirm each vector on the matrix parallel to the second axis with each coordinate corresponding to each process variable for the first axis. The electronic device (100) can confirm the sum information of each influence value included in each vector for each factor corresponding to each process variable node, and confirm the indirect influence information based on the sum information. Refer to FIG. 7 to examine an example of such a vector.

[0091] FIG. 7 is a diagram illustrating an example of checking indirect influence information using a matrix according to one embodiment.

[0092] Referring to FIG. 7, the electronic device (100) can check the coordinates on the first axis corresponding to each process variable node, i.e., PP, and check vectors (700) having the corresponding coordinates and parallel to the second axis. The electronic device (100) can check the indirect influence information for each process variable by adding the values ​​included in the vectors (700) for each vector. In FIG. 7, the vectors (700) parallel to the second axis are displayed as if they correspond to only a part, not the entire row having the corresponding coordinates. This is because the PPs, i.e., the process variable nodes, do not influence each other and are therefore set to 0, and the result is the same whether or not the part related to the influence value between the PPs is included.

[0093] According to one embodiment, the electronic device (100) can identify direct and indirect influence information and graph it according to the embodiment described above. To examine an example of such graphed direct and indirect influence information, refer to FIGS. 8a and 8b .

[0094] Figures 8a and 8b are graphs showing direct and indirect influence information according to one embodiment.

[0095] Figure 8a is a graph showing direct influence information, where the vertical axis represents each factor, and the bar graph parallel to the horizontal axis allows you to check the direct influence value of each factor on the characteristics of the final product. Factors with an X prefix in each factor on the vertical axis, for example, The back refers to the process variables, and the factors with the prefix Y, for example, The back can refer to the properties of intermediate products.

[0096] Figure 8b is a graph showing indirect influence information. The vertical axis represents each factor, and the bar graph parallel to the horizontal axis shows the indirect influence value of each process variable. Because this is indirect influence information, unlike Figure 8a, it can be confirmed that the characteristics of intermediate products prefixed with "Y" do not exist.

[0097] Comparing Figures 8a and 8b, the direct influence information does not exist, but indirect impact information is available It can be confirmed that it exists in a high ranking of 3rd place. This is Although it has a relatively small direct influence because it is not a process variable that primarily affects the process of creating a final product from an intermediate product just before the final product, it can mean that it has a large influence on the intermediate products before it. Similarly, it is a factor that is not included in the direct influence information but accounts for a significant proportion of the indirect influence information. and There is room for different uses of direct and indirect influence information, as they provide information about slightly different factors.

[0098] According to one embodiment, the electronic device (100) can support controlling the characteristics of intermediate products based on direct impact information. Specifically, the direct impact information can be beneficially utilized by product designers who wish to analyze and adjust intermediate products to efficiently design cells. Accordingly, the electronic device (100) can support product designers in performing cell design by adjusting intermediate products by providing direct impact information to the product designer terminal.

[0099] According to one embodiment, the electronic device (100) can support the control of process variables for each sub-process based on indirect impact information. Specifically, the indirect impact information can be beneficially utilized by process operators who wish to control quality by efficiently setting process variables associated with each sub-process. Accordingly, the electronic device (100) can support process operators in efficiently controlling process variables and ensuring good cell quality by providing the indirect impact information to the process operator terminal.

[0100] According to one embodiment, the electronic device (100) may also check additional analysis information. Specifically, the electronic device (100) may check contribution ratio information indicating the ratio of the values ​​of direct influence information and indirect influence information for each of a plurality of factors based on the influence information. That is, the electronic device (100) may check contribution ratio information by adding the values ​​indicated by the direct influence information and the values ​​indicated by the indirect influence information for each factor and then dividing each value by the sum to calculate the ratio of direct influence and indirect influence among the influences that each factor has on the overall battery manufacturing process. Refer to FIG. 9 to examine an example of such contribution ratio information.

[0101] Figure 9 is an example diagram of contribution ratio information according to one embodiment.

[0102] Referring to Fig. 9, In the case of process variables, the proportion of direct impact information is overwhelmingly high. and In the case of process variables, it can be confirmed that the proportion of indirect influence information is overwhelmingly high. By analyzing this contribution ratio information, managers can There is no intermediate product characteristic that well reflects the influence of and Conversely, it can be seen that the characteristics of the intermediate products connected to it can be greatly influenced. In addition, as an additional example, managers can, on the causal graph, and If the node corresponding to is connected to a node corresponding to the same intermediate product characteristic, and It will be possible to analyze whether the characteristics of the intermediate product can be controlled by adjusting the .

[0103] The model (110) used to perform the method of analyzing battery performance influencing factors described above may be one learned by the electronic device (100), or may be a model whose learning has already been completed and input into the electronic device (100). An example of a model (110) being learned by the electronic device (100) will be described below.

[0104] According to one embodiment, the electronic device (100) can input the aforementioned process data into the model (110) for each data instance. The electronic device (100) can check the results of comparing the output of the model (110) with actual data corresponding to the characteristics of the final product identified for each data instance. Thereafter, the electronic device (100) can train the model (110) based on the results of such comparison. For example, if the model (110) is constructed using a DNN, the electronic device (100) can calculate a loss function based on the results of the aforementioned comparison and train the model (110) by backpropagating the loss.

[0105] Fig. 10 illustrates a block diagram of an electronic device according to one embodiment.

[0106] The electronic device (100) may include, according to one embodiment, a memory (101) and a processor (102). The electronic device (100) illustrated in FIG. 10 only illustrates components related to the present embodiment. Therefore, it will be understood by those skilled in the art related to the present embodiment that other general components may be included in addition to the components illustrated in FIG. 10. In one embodiment, the processor (102) may be included in a controller.

[0107] The processor (102) can control the overall operation of the electronic device (100) and process data and signals. The processor (102) can be composed of at least one hardware unit. In addition, the processor (102) can operate by one or more software modules generated by executing program codes stored in the memory (101). The processor (102) can include a memory, and the processor (102) can control the overall operation of the electronic device (100) and process data and signals by executing program codes stored in the memory.

[0108] The processor (102) may be configured to perform one or more instructions to identify process data including, as data instances, a causal graph representing relationships between a plurality of factors related to a battery manufacturing process and data for a plurality of factors identified for each run of the battery manufacturing process, and to identify information on the influence of each of the plurality of factors on the characteristics of the final product based on a model learned to receive the causal graph, the process data, and the process data and to predict the characteristics of the final product of the battery manufacturing process.

[0109] Depending on the embodiment, the electronic device (100) may additionally include a transceiver for performing wired / wireless communication. The electronic device (100) may communicate with an external electronic device using the transceiver. The external electronic device may be a terminal or a server. In addition, the communication technologies used by the transceiver may include GSM (Global System for Mobile communication), CDMA (Code Division Multi Access), LTE (Long Term Evolution), 5G, WLAN (Wireless LAN), Wi-Fi (Wireless-Fidelity), Bluetooth (Bluetooth), RFID (Radio Frequency Identification), Infrared Data Association (IrDA), ZigBee, NFC (Near Field Communication), etc.

[0110] The electronic device according to the above-described embodiments may include a processor, a memory for storing and executing program data, permanent storage such as a disk drive, a communication port for communicating with an external device, a user interface device such as a touch panel, a key, a button, etc. The methods implemented as software modules or algorithms may be stored on a computer-readable recording medium as computer-readable codes or program instructions executable on the processor. Here, the computer-readable recording medium includes a magnetic storage medium (e.g., read-only memory (ROM), random-access memory (RAM), floppy disk, hard disk, etc.) and an optical reading medium (e.g., CD-ROM, DVD: Digital Versatile Disc)). The computer-readable recording medium may be distributed to computer systems connected to a network, so that the computer-readable code may be stored and executed in a distributed manner. The medium may be readable by a computer, stored in a memory, and executed by a processor.

[0111] The present embodiment may be represented by functional block configurations and various processing steps. These functional blocks may be implemented by various hardware and / or software configurations that perform specific functions. For example, the embodiment may employ direct circuit configurations such as memory, processing, logic, look-up tables, etc., which may perform various functions under the control of one or more microprocessors or other control devices. Similarly, the present embodiment may be implemented in a programming or scripting language such as C, C++, Java, assembler, etc., including various algorithms implemented as a combination of data structures, processes, routines, or other programming configurations. Functional aspects may be implemented as algorithms that execute on one or more processors. Furthermore, the present embodiment may employ conventional techniques for electronic configuration, signal processing, and / or data processing. Terms such as "mechanism," "element," "means," and "composition" can be used broadly and are not limited to mechanical or physical structures. These terms can also encompass a series of software routines, such as those associated with a processor.

[0112] The above-described embodiments are merely examples, and other embodiments may be implemented within the scope of the claims set forth below.

Claims

1. In a method for analyzing factors affecting battery performance of an electronic device, A step of identifying process data including a causal graph representing relationships between multiple factors related to a battery manufacturing process and data for the multiple factors identified for each execution of the battery manufacturing process as each data instance; and A step of checking information on the influence on the characteristics of the final product by each of the plurality of factors based on the causal graph, the process data, and a model learned to predict the characteristics of the final product of the battery manufacturing process by inputting the process data, The above causal graph is, A property node corresponding to the characteristics of the intermediate product and the final product produced for each detailed process, each of which is unidirectionally connected to the order of each detailed process included in the battery manufacturing process; and A method for analyzing battery performance influencing factors, which represents a graph including a plurality of process variable nodes corresponding to process variables controlled in each of the detailed processes corresponding to each of the property nodes, each of which is unidirectionally connected toward each of the above property nodes.

2. In paragraph 1, Based on the above causal graph, a step of identifying multiple paths connecting the multiple process variable nodes to the property nodes corresponding to the characteristics of the final product; and A step of verifying a virtual data instance by adjusting at least some of the data of the data instance included in the process data based on the plurality of paths verified above; a step of inputting the virtual data instance into the model; and A method for analyzing battery performance influence factors, comprising a step of analyzing the output of the model for the virtual data instance and confirming the influence information.

3. In paragraph 2, The steps to verify the above virtual data instance are: A step of identifying at least one first path including an edge between a first node corresponding to a first factor and a second node corresponding to a child node of the first node on the causal graph based on the plurality of paths; and A step of identifying at least one first virtual data instance based on at least one first path, The steps to check the above impact information are: A step of confirming first value information for the first path based on the first virtual data instance and the model; and A method for analyzing battery performance influencing factors, comprising a step of confirming the influence value of the first factor on the second factor corresponding to the second node based on the first value information.

4. In paragraph 3, The step of verifying at least one first virtual data instance is: A step of identifying at least one path that is positioned in a preceding order than at least one first path among a plurality of permutations in which the plurality of paths are randomly listed; A step of checking a first subset including at least one path located in a previous order than the at least one first path, and a second subset in which the first path is added to the first subset, for each of the plurality of permutations; and A step of confirming a 1-1 virtual data instance by randomly sampling and adjusting data of nodes that do not correspond to the first subset among data of the data instance, and a step of confirming a 1-2 virtual data instance by randomly sampling and adjusting data of nodes that do not correspond to the second subset among data of the data instance, The step of confirming the first value information for the first path above is: A step of confirming the 1-1 and 1-2 value information based on the results of inputting the 1-1 and 1-2 virtual data instances confirmed by the plurality of permutations into the model; and A method for analyzing battery performance influencing factors, comprising a step of confirming the first value information based on difference information between the first-1 and first-2 value information.

5. In paragraph 4, The step of confirming the above first value information is: A method for analyzing battery performance influencing factors, comprising a step of confirming first value information for the first path based on average information of the difference information between the first-1 and first-2 value information confirmed for each of the plurality of permutations.

6. In paragraph 4, The step of confirming the above first value information is: A step of inputting the above 1-1 virtual data instance into the above model and confirming the 1-1 output information; and A method for analyzing battery performance influencing factors, comprising a step of inputting the unadjusted data instance into the model and confirming the first-first value information based on the difference information between the average information of the confirmed output information and the first-first output information.

7. In paragraph 3, The step of checking the influence information of the first factor above is: A method for analyzing battery performance influencing factors, comprising a step of confirming a first influence value of the first factor on the second factor based on the sum information of the first value information for the first path including the edge between the first and second nodes.

8. In paragraph 1, The steps to check the above impact information are: A step of checking a matrix including as each component an influence value of a first factor corresponding to a first coordinate of a first axis and a second factor corresponding to a second coordinate of a second axis; A method for analyzing battery performance influencing factors, comprising a step of identifying the influence information including direct influence information and indirect influence information on the characteristics of the final product of the plurality of factors based on the matrix.

9. In paragraph 8, The steps to check the above impact information are: A step of checking the influence value of at least one factor corresponding to at least one node directly connected to a node corresponding to the characteristic of the final product on the above matrix; and A method for analyzing battery performance influencing factors, comprising a step of confirming the direct influence information based on the influence value of at least one factor.

10. In paragraph 9, The step of checking the influence value of at least one of the above factors is: A step of confirming coordinates corresponding to the characteristics of the final product on the second axis; A step of identifying a vector on the matrix parallel to the first axis with coordinates corresponding to the characteristics of the final product with respect to the second axis; and A method for analyzing battery performance influencing factors, comprising a step of checking the influence value of at least one factor based on the above vector.

11. In paragraph 8, The steps to check the above impact information are: A step of checking the sum information of the influence values ​​for each factor corresponding to each of the above process variable nodes; and A method for analyzing battery performance influencing factors, comprising a step of confirming the indirect influence information for each factor corresponding to each process variable node based on the above total information.

12. In paragraph 11, The steps to check the above total information are: A step of confirming each coordinate corresponding to each process variable on the first axis; A step of checking each vector on the matrix parallel to the second axis with each coordinate corresponding to each process variable for the first axis; and A method for analyzing battery performance influencing factors, comprising a step of checking each sum information of the influence values ​​included in each of the above vectors for each factor corresponding to each of the above process variable nodes.

13. In paragraph 8, A battery performance influence factor analysis method further comprising a step of checking contribution ratio information indicating the ratio of the values ​​of the direct influence information and the indirect influence information for each of the plurality of factors based on the influence information.

14. In paragraph 8, A method for analyzing battery performance influencing factors, further comprising a step of supporting control of the characteristics of the intermediate product based on the above direct influence information.

15. In paragraph 8, A method for analyzing battery performance influencing factors, further comprising a step of supporting control of the process variables for each detailed process based on the indirect influence information.

16. In paragraph 1, The step of checking the above causal graph and the above process data is: A step of confirming the entire process data confirmed for each execution of the above battery manufacturing process; A step of calculating a correlation coefficient between overall factors of the battery manufacturing process based on the above overall process data; A step of confirming the plurality of factors by filtering at least some of the factors among the entire factors whose correlation coefficient is greater than or equal to a threshold value; and A method for analyzing battery performance influencing factors, comprising a step of identifying at least some of the data corresponding to the plurality of factors among the entire process data as the process data.

17. In paragraph 16, Among the above entire process data, the step of confirming at least some of the data corresponding to the plurality of factors as the process data is: A method for analyzing battery performance influencing factors, comprising a step of identifying, as the process data, a data instance in which the measurement temperature of the characteristics of the final product corresponds to an effective temperature range among the data instances included in the entire process data.

18. In paragraph 1, A step of inputting the above process data into the model for each data instance; and A method for analyzing battery performance influencing factors, comprising a step of learning the model based on the results of comparing the output of the model with actual data corresponding to the characteristics of the final product confirmed for each data instance.

19. A non-transitory computer-readable recording medium having recorded thereon a program for executing the method of any one of clauses 1 to 18 on a computer.

20. An electronic device that analyzes factors affecting battery performance. processor; and Contains memory that stores one or more instructions, The processor, by performing one or more of the instructions, A process data including a causal graph representing the relationship between multiple factors related to a battery manufacturing process and data for the multiple factors identified for each execution of the battery manufacturing process as each data instance is identified, and based on the causal graph, the process data, and a model learned to predict the characteristics of the final product of the battery manufacturing process by inputting the process data, information on the influence on the characteristics of the final product is identified for each of the multiple factors. The above causal graph is, A property node corresponding to the characteristics of the intermediate product and the final product produced for each detailed process, each of which is unidirectionally connected to the order of each detailed process included in the battery manufacturing process; and An electronic device representing a graph including a plurality of process variable nodes corresponding to process variables controlled in each of the detailed processes corresponding to each of the material nodes, each of which is unidirectionally connected toward each of the material nodes.

Images