Model determination device, prediction device, and model determination method

Abstract

A model determination device according to the present invention includes: a first creation unit that creates first data indicating a set consisting of multiple pairs of a first variable and a second variable used when constructing a prediction model that outputs a predicted value of the discharge capacity of a manufactured secondary battery; a second creation unit that creates second data indicating a set consisting of multiple pairs of a first variable and a second variable measured when inspecting the secondary battery that is the object of the predicted value output; and a determination unit that determines whether the second data is similar to the first data on the basis of the results of statistical processing performed on the first data and the second data.

Description

Model determination device, prediction device, and model determination method

[0001] The present disclosure relates to a model determination device, a prediction device, a model determination method, and the like.

[0002] Patent Document 1 discloses a discharge capacity prediction device that predicts the discharge capacity of a chargeable and dischargeable battery. The discharge capacity prediction device in Patent Document 1 accepts, for each battery, inputs of measured values ​​for each material manufactured in the manufacturing process, measured values ​​for each item after each assembly operation in the assembly process, and measured values ​​for each item measured in the charge / discharge inspection process. The measured values ​​accepted for each battery are then input into a prediction model, and a predicted value of the discharge capacity is output.

[0003] Japanese Patent Application Publication No. 2014-071103

[0004] A model determination device according to one aspect of the present disclosure includes a first creation unit that creates first data indicating a set consisting of multiple pairs of a first variable and a second variable different from the first variable, which is used when constructing a prediction model that outputs a predicted value of the discharge capacity of a manufactured secondary battery; a second creation unit that creates second data indicating a set consisting of multiple pairs of the first variable and the second variable, which are measured when inspecting the secondary battery that is the target for outputting the predicted value; and a determination unit that determines whether the second data is similar to the first data based on results of statistical processing on the first data and the second data.

[0005] Furthermore, a model determination method according to one aspect of the present disclosure includes a first creation step of creating first data indicating a set consisting of multiple pairs of a first variable and a second variable different from the first variable, which is used when constructing a prediction model that outputs a predicted value of the discharge capacity of a manufactured secondary battery; a second creation step of creating second data indicating a set consisting of multiple pairs of the first variable and the second variable, which are measured when inspecting the secondary battery that is the target for outputting the predicted value; and a determination step of determining whether the second data is similar to the first data based on the results of statistical processing on the first data and the second data.

[0006] FIG. 1 is a block diagram showing an example of an inspection system of an embodiment; FIG. 2 is a flowchart showing an example of processing in the inspection system of an embodiment; FIG. 3 is a data diagram showing an example of first data and second data; FIG. 4 is a flowchart showing an example of a processing flow in a control unit of a model determination device of an embodiment; FIG. 5 is a table showing an example of various data related to the first data; and FIG. 6 is a table showing an example of various data related to the second data.

[0007] [Embodiment 1] <Configuration of Inspection System> Fig. 1 is a block diagram showing an example of an inspection system 1 according to this embodiment. The inspection system 1 is a system that inspects manufactured secondary batteries. The inspection system 1 may include, for example, a manufacturing device 2, an inspection device 3, a prediction device 4, a model construction device 5, and a model determination device 6. In Fig. 1, solid arrows indicate the flow of data, and dashed arrows indicate the direction of supply of secondary batteries.

[0008] The manufacturing apparatus 2 may be an apparatus for manufacturing chargeable and dischargeable secondary batteries. The secondary batteries manufactured by the manufacturing apparatus 2 are supplied to the inspection apparatus 3. Secondary batteries that have been inspected by the inspection apparatus 3 and determined to be non-defective may be shipped as products.

[0009] The secondary battery manufactured by the manufacturing apparatus 2 may include at least one negative electrode, at least one positive electrode, and a separator located between the negative electrode and the positive electrode.

[0010] The negative electrode may have an electrode conductor and a negative electrode active material layer. The negative electrode active material layer may be a layer of a negative electrode material that is a mixture of a negative electrode active material, a conductive additive, and an electrolyte.

[0011] The positive electrode may have an electrode conductor and a positive electrode active material layer. The positive electrode active material layer may be a layer of a positive electrode material that is a mixture of a positive electrode active material, a conductive additive, and an electrolyte.

[0012] The electrolyte may be filled in the housing that houses the negative electrode and the positive electrode, or may be held by impregnating the separator, for example.

[0013] The inspection device 3 may be a device that inspects the secondary battery manufactured by the manufacturing device 2. The inspection device 3 may perform a first inspection step of inspecting whether the secondary battery is a non-defective product, and a second inspection step of inspecting the capacity of the secondary battery. The first inspection step may include a formation step and an aging step.

[0014] The inspection device 3 may charge the secondary battery in the formation step. For example, the inspection device 3 may charge the secondary battery until it is fully charged (first charging step), and then discharge the secondary battery until it reaches the end of discharge (first discharging step). Furthermore, the inspection device 3 may perform a process of repeatedly charging and discharging the secondary battery in the formation step.

[0015] The inspection device 3 may not charge the secondary battery until it is fully charged in the first charging step. Also, the inspection device 3 may not discharge the secondary battery until it reaches the end of discharge in the first discharging step. The inspection device 3 may charge or discharge the secondary battery until it reaches a voltage that can dissolve and deposit metal impurities.

[0016] When repeatedly charging and discharging, the inspection device 3 may cause the secondary battery to wait inside the inspection device 3 for a predetermined time between the end of a charging step and the start of the next discharging step, or between the end of a discharging step and the start of the next charging step. In other words, the inspection device 3 does not need to charge or discharge the secondary battery for the predetermined time. The predetermined time may be, for example, several hours to about a day.

[0017] In the charging step, the inspection device 3 may charge the secondary battery in a constant current-constant voltage mode or only in the constant current mode. In the constant current-constant voltage mode, the inspection device 3 may charge the secondary battery at a constant current, and then, after the voltage at a pre-designated connection terminal reaches a predetermined voltage, charge the secondary battery at a constant voltage. The connection terminal is a terminal that is electrically connected to an external terminal.

[0018] The constant current during charging and discharging may be, for example, 0.01 C or more and 3 C or less. The C rate is an index that represents the magnitude of the current when charging and discharging a secondary battery. 1 C is defined as the current value when a battery goes from a fully charged state to an almost or completely discharged state in one hour, or the current value when a battery goes from an almost or completely discharged state to a fully charged state. Furthermore, the temperature inside the inspection device 3 when the formation process is performed may be, for example, 20°C or more and 100°C or less.

[0019] Furthermore, the secondary battery may be charged to a predetermined voltage before the aging process is performed, and then placed outside the inspection device 3. The secondary battery may be left outside the inspection device 3 to wait until the aging process is performed.

[0020] The inspection device 3 may perform an aging process after performing the formation process. In the aging process, the secondary battery is left in an open circuit state. At this time, the secondary battery is left at a voltage or charge level determined when the formation process is performed. The inspection device 3 measures the voltage difference between the open circuit voltage at the start of the aging process and the open circuit voltage at the end of the aging process. The inspection device 3 determines whether the secondary battery is good or bad based on the value of this voltage difference. This allows defective products to be removed. The temperature inside the inspection device 3 when the aging process is performed may be, for example, 15°C or higher and 80°C or lower. The secondary battery may be left for, for example, 12 hours or higher and 30 days or shorter.

[0021] In the second inspection process, the inspection device 3 charges the secondary battery that was determined to be a good product in the aging process. In the second inspection process, the inspection device 3 may charge the secondary battery until it is fully charged after being almost or completely discharged. The inspection device 3 may determine whether the secondary battery is good or bad based on the amount of charge in the second inspection process. The inspection device 3 may determine whether the amount of charge in the second inspection process is equal to or greater than a predetermined charge capacity, thereby making it possible to further eliminate defective products.

[0022] The inspection device 3 may measure various values ​​in the first inspection process and the second inspection process. In the second inspection process, the inspection device 3 may determine whether the secondary battery is good or bad based on a predicted value of the discharge capacity of the secondary battery output based on at least two of these values. In the second inspection process, the inspection device 3 may determine whether the predicted value of the discharge capacity is equal to or greater than a predetermined discharge capacity, thereby making it possible to further remove defective products.

[0023] Examples of the types of variables that the inspection device 3 measures include the following. However, variables other than those listed below may also be measured. (1) Examples of variables measured in the formation process Here, it is assumed that the charging process and the discharging process are repeated multiple times. - Charge amount in at least any of the charging processes. - Discharge amount in at least any of the discharging processes. - Voltage value at the start of at least any of the charging processes, or voltage value at the start of at least any of the discharging processes. - Voltage value at the end of at least any of the charging processes, or voltage value at the end of at least any of the discharging processes. - Resistance value of the secondary battery at a predetermined timing. (2) Examples of variables measured in the aging process - Voltage value at the start or end of the aging process. - Resistance value at the start or end of the aging process. - Temperature near the secondary battery at the start or end of the aging process. This temperature may be the ambient temperature when the aging process is being performed. The ambient temperature may be the internal temperature of the inspection device 3 when the aging process is being performed. Average temperature in the vicinity of the secondary battery during the aging process. This average temperature may be the average temperature of the ambient temperature. Rate of voltage decrease. (3) Examples of variables measured in the second inspection process: Amount of charge. Charging time. Voltage value at the start or end of charging.

[0024] In addition, the variables may include a waiting time between each process. The waiting time is the time from the end of one process to the start of the next process. The waiting time may include the time between the formation process and the aging process, and the time between the aging process and the second inspection process. The waiting time may include the time from the manufacture of the secondary battery until the start of the formation process.

[0025] The prediction device 4 may be a device that predicts the discharge capacity of a manufactured secondary battery using a prediction model 421 that outputs a predicted value of the discharge capacity of the manufactured secondary battery. The prediction model 421 is a target for judgment by the model judgment device 6.

[0026] The prediction device 4 may include a prediction unit 41 and a model storage unit 42. The prediction unit 41 may be a function of a control unit that comprehensively controls each unit of the prediction device 4. The model storage unit 42 may store a prediction model 421 that is constructed by the model construction device 5 and used by the prediction unit 41.

[0027] The prediction unit 41 may output a predicted value of the discharge capacity by inputting values ​​of variables measured when the inspection device 3 inspects the secondary battery in the first inspection process and the second inspection process into the prediction model 421. The prediction unit 41 may output the predicted value of the discharge capacity to the inspection device 3. The measured values ​​may be values ​​measured for each of the above-mentioned types of variables.

[0028] By providing the prediction device 4, the inspection device 3 does not need to calculate the discharge capacity in the second inspection process. This reduces the time required to inspect the secondary battery. Furthermore, when predicting the discharge capacity, the measured values ​​in the first and second inspection processes are used, but the measured values ​​in the previous processes are not used. This reduces the equipment costs for measuring the measured values.

[0029] If the charge amount measured when inspecting the secondary battery in the second inspection process is equal to or less than a predetermined reference value, the prediction unit 41 does not need to input the measurement value measured by the inspection device 3 into the prediction model 421. The prediction unit 41 does not predict the discharge capacity of a secondary battery whose charge amount is measured to be equal to or less than the reference value. This allows for efficient inspection of secondary batteries. The reference value may be set, for example, through experiments, to a charge amount that exceeds the acceptable range for a product. In other words, if the charge amount is equal to or less than the reference value, the secondary battery whose charge amount is measured may be identified as a defective product.

[0030] For example, the inspection device 3 may determine whether the charge amount measured in the second inspection process is equal to or less than a reference value. If the inspection device 3 determines that the measured charge amount is equal to or less than the reference value, the inspection device 3 may determine that the secondary battery is defective. In this case, the inspection device 3 does not need to transmit various measurement values ​​measured on the secondary battery in each process to the prediction device 4. As a result, the prediction unit 41 cannot input the measurement values ​​measured by the inspection device 3 into the prediction model 421, and therefore does not predict the discharge capacity of the secondary battery.

[0031] However, the above determination may be made by the prediction unit 41. When the prediction unit 41 determines that the measured charge amount is equal to or less than the reference value, the prediction unit 41 may output information indicating that the secondary battery is defective to the inspection device 3 without outputting the predicted value of the discharge capacity.

[0032] The model construction device 5 may be a device that constructs a prediction model 421. The model construction device 5 outputs the constructed prediction model 421 to the prediction device 4. The model construction device 5 outputs the values ​​of variables used in constructing the prediction model 421 to the model determination device 6. Furthermore, upon receiving a notification from the model determination device 6, the model construction device 5 may reconstruct the prediction model 421 by excluding variables of the types indicated in the notification.

[0033] Examples of variables used to build the prediction model 421 include variables measured in the formation process, variables measured in the aging process, and variables measured in the second inspection process. The model construction device 5 may build the prediction model 421 using, for example, values ​​measured for each variable from multiple secondary batteries manufactured in the past. The values ​​of each variable used to build the prediction model 421 are not limited to values ​​measured in the past. For example, for each variable, a value measured by simulating the manufacturing of a secondary battery, a value obtained by appropriately adjusting the measured value, or the like may be used.

[0034] The prediction model 421 may be constructed by regression analysis using at least one of the above-described various variables as an explanatory variable and the predicted value of the discharge capacity as a response variable. The regression analysis may be, for example, linear regression analysis. In this case, for example, Lasso regression may be used. Furthermore, for example, Ridge regression may be used instead of Lasso regression. Lasso regression and Ridge regression are regression analyses in which a regularization term is added to the objective function, and are regression analyses that can reduce overlearning.

[0035] Nonlinear regression analysis, rather than linear regression analysis, may be used to build the prediction model 421. For example, a neural network model may be used to build the prediction model 421.

[0036] A plurality of data sets may be prepared, each including a value of each variable and a value of discharge capacity at that value of the variable, in order to construct the prediction model 421. Some of the data sets may be used as a training data set, and the remaining data sets may be used as a test data set.

[0037] When the prediction model 421 is constructed by linear regression analysis, it may be expressed by, for example, the following equation: Predicted value of discharge capacity = A0X0 + A1X1 + ... + An-1Xn-1 + AnXn, where A0 to An are regression coefficients of each variable adjusted by regression analysis. X0 to Xn are values ​​of each variable measured by the inspection device 3 that are input to the prediction unit 41 when predicting the discharge capacity.

[0038] The model determination device 6 is a device that determines the quality of the prediction model 421 constructed by the model construction device 5. An example of a specific configuration and processing of the model determination device 6 will be described later.

[0039] <Processing Flow of Inspection System> FIG. 2 is a flowchart showing an example of processing in the inspection system 1.

[0040] 2, first, the model construction device 5 may construct a prediction model 421 (S1). The prediction device 4 may store the prediction model 421 constructed by the model construction device 5 in the model storage unit 42 (S2).

[0041] Thereafter, the manufacturing apparatus 2 may manufacture a secondary battery (S3). The inspection apparatus 3 may inspect the secondary battery manufactured by the manufacturing apparatus 2 (S4). The inspection apparatus 3 may perform a first inspection process to determine whether the secondary battery is a good product. The inspection apparatus 3 may also output the values ​​of each variable measured in the first inspection process and the second inspection process to the prediction apparatus 4. The prediction apparatus 4 may input these values ​​into the prediction model 421 and output a predicted value of discharge capacity obtained as a result to the inspection apparatus 3. The inspection apparatus 3 may further determine whether the secondary battery is a good product based on this predicted value of discharge capacity.

[0042] Thereafter, the model determination device 6 determines whether the secondary battery has been inspected for a predetermined period using the prediction model 421 stored in the model storage unit 42 in S2 (S5). The predetermined period may be set, for example, to a period during which sufficient measurement values ​​can be acquired for the model determination device 6 to determine the quality of the prediction model 421. The predetermined period may be, for example, one month.

[0043] If the model determination device 6 determines that the secondary battery has been inspected for a predetermined period of time (YES in S5), it may determine whether the prediction model 421 is good or bad (S6). If the model determination device 6 determines that the prediction model 421 is usable as a result of the process in S6 (YES in S7), this process ends. If the model determination device 6 determines that the prediction model 421 is unusable (NO in S7), it may notify the model construction device 5 of the types of variables identified in the process described below. When the model construction device 5 receives a notification indicating the types of identified variables, it may reconstruct the prediction model 421 by excluding the identified variables (S8). The prediction device 4 may store the prediction model 421 reconstructed by the model construction device 5 in the model storage unit 42 (S9). This storage updates the prediction model 421.

[0044] In S5, the model determination device 6 may determine whether the number of secondary battery inspections has reached a predetermined number, rather than whether the secondary battery inspections have been performed for a predetermined period of time. If it is determined that the number of secondary battery inspections has reached the predetermined number, the process may proceed to S6, and if the number of secondary battery inspections has not reached the predetermined number, the process may end.

[0045] Generally, secondary battery inspections include a formation process, an aging process, and a capacity inspection process. The capacity inspection process includes, for example, a charging process in which secondary batteries from which defective products have been removed in the aging process are fully charged, and a discharging process in which the secondary batteries after the charging process are discharged to the end of discharge. The discharge capacity is calculated using the discharge current and discharge time measured in this discharging process. Then, the quality of the secondary battery is determined based on the calculated discharge capacity.

[0046] On the other hand, in the inspection system 1 of this embodiment, the formation process, the aging process, and the second inspection process may be performed. That is, the second inspection process is performed instead of the capacity inspection process, and the above-mentioned discharge process and calculation of the discharge capacity are not performed. Instead, the predicted result of the discharge capacity by the prediction device 4 is used to determine the quality of the secondary battery. Therefore, the time required for inspecting the secondary battery can be shortened.

[0047] 1 , the model determination device 6 may include a control unit 61 and a storage unit 62. The control unit 61 may have a function of comprehensively controlling each unit of the model determination device 6. The storage unit 62 may store data and programs used by the control unit 61.

[0048] The control unit 61 may include a first creation unit 611 , a second creation unit 612 , a first calculation unit 613 , a second calculation unit 614 , a third calculation unit 615 , a setting unit 616 , a determination unit 617 , and a notification unit 618 .

[0049] The first creation unit 611 may create first data indicating a set of multiple pairs of a first variable and a second variable different from the first variable. The first variable and the second variable used to create the first data may be values ​​of variables used when the model construction device 5 builds the prediction model 421.

[0050] The second creation unit 612 may create second data indicating a set consisting of multiple pairs of first variables and second variables. The first variables and second variables used in creating the second data may be values ​​of variables measured when inspecting a secondary battery for which a predicted value of discharge capacity is to be output. The secondary battery for which a predicted value of discharge capacity is to be output may be a secondary battery manufactured by the manufacturing apparatus 2 in preparation for shipment and a secondary battery to be inspected by the inspection apparatus 3. In other words, the second data may be created using measured values ​​measured by the inspection apparatus 3.

[0051] The first variable may be one of a plurality of variables to be measured when inspecting the secondary battery in the first inspection process and the second inspection process. The second variable may be one or more of the plurality of variables. The variables used when constructing the prediction model 421 may be measurement values ​​measured in the first inspection process and the second inspection process, or values ​​corresponding to the measurement values.

[0052] In creating the first data and the second data, variables to be measured in the first inspection process and the second inspection process are used as the first variable and the second variable, and variables to be measured in the previous process are not used as the first variable and the second variable. This reduces the cost of equipment for measuring the measured values. Also, it reduces the number of types of variables used to create the first data and the second data.

[0053] Furthermore, the type of first variable may be the charge amount of the secondary battery. The charge amount has a stronger correlation with the discharge capacity than other variables. Therefore, by using the charge amount as the first variable, it is possible to accurately determine whether the predicted value of the discharge capacity and the actual measured value are consistent. The charge amount used as the first variable may be the charge amount to be measured in the second inspection process.

[0054] 3 is a data diagram showing an example of the first data and the second data. Reference numeral 101 in Fig. 3 indicates an example of the first data created when constructing the prediction model 421. Reference numeral 102 in Fig. 3 indicates an example of the second data created when inspecting the secondary battery.

[0055] In the example of Figure 3, the vertical axis represents the first variable and the horizontal axis represents the second variable. In the example of Figure 3, the first variable is the charge amount of the secondary battery in the second inspection process. The second variable is one of the variables measured in the formation process, the variables measured in the aging process, and the variables measured in the second inspection process other than the charge amount. In the example of Figure 3, the second variable is the charge amount in the first charging process in the formation process.

[0056] To construct the prediction model 421, the first creation unit 611 may create, for example, first data indicating a set of pairs of first variables and second variables measured in the past when inspecting each of a plurality of secondary batteries in the first inspection process and the second inspection process. The second creation unit 612 may create, for example, second data indicating a set of pairs of first variables and second variables measured when inspecting each of a plurality of secondary batteries in the first inspection process and the second inspection process during the manufacture of secondary batteries to be shipped. The second creation unit 612 may create, for example, second data indicating a set of pairs of first variables and second variables measured during the above-mentioned predetermined period. The example of FIG. 3 shows a state in which multiple pairs of first variables and second variables are plotted.

[0057] The first creation unit 611 may create first data for each type of second variable. The first creation unit 611 may create first data by using two or more types of variables different from the type of the first variable as second variables among the above-mentioned variables. Furthermore, the second creation unit 612 may create second data corresponding to the first data. In other words, the second creation unit 612 may create second data using first variables and second variables of the same type as the first variables and second variables constituting the first data, for each of the first data. However, the first creation unit 611 and the second creation unit 612 may each create one piece of first data and one piece of second data.

[0058] 5 , which will be described later, in each of the multiple first data, the type of the first variable is “V1,” and the types of the second variables are different from the type of the first variable, such as “V2,” “V3,” “V4,” etc. In other words, first data may be created for each type of second variable, such as first data in which the first variable is “V1” and the second variable is “V2,” first data in which the first variable is “V1” and the second variable is “V3,” first data in which the first variable is “V1” and the second variable is “V4,” etc.

[0059] 5 is created, in each of the plurality of second data, the type of the first variable is "V1" and the types of the second variables are different from the type of the first variable, such as "V2," "V3," "V4," etc. In other words, the same number of second data as the number of first data may be created, such as second data in which the first variable is "V1" and the second variable is "V2," second data in which the first variable is "V1" and the second variable is "V3," second data in which the first variable is "V1" and the second variable is "V4," etc.

[0060] The first calculation unit 613 may calculate the first Mahalanobis distance for each of a plurality of pairs included in the first data. The first calculation unit 613 may calculate the first Mahalanobis distance for each pair using, as a population, the total number of pairs of first variables and second variables acquired when constructing the prediction model 421. When there are a plurality of pieces of first data, the first calculation unit 613 may calculate the first Mahalanobis distance for each piece of first data.

[0061] The second calculation unit 614 may calculate the second Mahalanobis distance for each of the plurality of pairs included in the second data. The second calculation unit 614 may calculate the second Mahalanobis distance for each pair using the total number of pairs of first variables and second variables obtained in inspection of secondary batteries manufactured for shipment as a population. When there are a plurality of first data, there is second data corresponding to each of the first data. In this case, the second calculation unit 614 may calculate the second Mahalanobis distance for each of the second data.

[0062] Reference numeral 101 in Fig. 3 indicates the Mahalanobis distance calculated from the first data indicated by reference numeral 101 using a contour line. Reference numeral 102 in Fig. 3 also indicates the Mahalanobis distance calculated from the second data indicated by reference numeral 102 using a contour line. The numerical values ​​on the contour lines indicate the Mahalanobis distances on the lines.

[0063] The third calculation unit 615 may calculate a standard deviation for the plurality of first Mahalanobis distances calculated by the first calculation unit 613. The third calculation unit 615 may calculate a standard deviation using the plurality of first Mahalanobis distances calculated by the first calculation unit 613 as a population. When there are a plurality of first data, the third calculation unit 615 may calculate a standard deviation for each of the first data.

[0064] The setting unit 616 may set a first threshold value for the second Mahalanobis distance based on the standard deviation calculated by the third calculation unit 615. For example, the setting unit 616 may set the first threshold value to a value obtained by multiplying the standard deviation calculated by the third calculation unit 615 by a predetermined factor. For example, the setting unit 616 may set the first threshold value to a value obtained by multiplying the standard deviation calculated by the third calculation unit 615 by 2 to 10 factors. The first threshold value may be set, for example, through experiments, so that the prediction model 421 can be changed in accordance with changes in the trend of the predicted value of the discharge capacity. When there are multiple pieces of first data, the setting unit 616 may set a first threshold value corresponding to each piece of first data. When there are multiple pieces of first data, the control unit 61 may manage each piece of first data by associating it with the second data, the standard deviation, and the first threshold value.

[0065] The determination unit 617 may determine whether the second data is similar to the first data based on the results of statistical processing on the first data and the second data. The determination unit 617 determines whether second data using first variables and second variables of the same type as the first variables and second variables constituting the first data is similar to the first data. When there are multiple pairs of first data and second data, the determination unit 617 determines whether the second data is similar to the first data for each of the pairs.

[0066] The above-described determination determines whether the predicted value of discharge capacity output by the prediction model 421 deviates from the actual measured value of discharge capacity, thereby determining the quality of the prediction model 421. Therefore, if a change in the trend of the predicted value of discharge capacity occurs due to, for example, a change in material, causing the predicted value of discharge capacity to deviate from the actual measured value, the prediction model 421 can be reconstructed by the model construction device 5. This reduces the possibility that secondary batteries whose predicted value of discharge capacity deviates from the actual measured value will be released onto the market.

[0067] This also reduces the possibility of an increase in the amount of secondary batteries discarded due to secondary batteries with a discrepancy between the predicted and measured discharge capacities being released onto the market. Such effects also contribute to the achievement of Goal 12 of the United Nations' Sustainable Development Goals (SDGs), "Responsible Consumption and Production."

[0068] In this embodiment, the determination unit 617 may determine whether the second data is similar to the first data by using the second Mahalanobis distance and the first threshold. As described above, the first threshold is set based on the standard deviation of the multiple first Mahalanobis distances calculated from the first data. Therefore, by using the second Mahalanobis distance and the first threshold, it is possible to determine the similarity of the second data to the first data based on the variance of the calculated first Mahalanobis distances.

[0069] The determination unit 617 may determine that the second data is not similar to the first data in any of the following cases, for example: (A) when it is determined that the ratio of the number of pairs of first variables and second variables for which second Mahalanobis distances equal to or greater than the first threshold to the number of elements that are at least a portion of the multiple pairs of first variables and second variables included in the second data is equal to or greater than a first preset value; or (B) when it is determined that the number of pairs of first variables and second variables for which second Mahalanobis distances equal to or greater than the first threshold are calculated is equal to or greater than a second preset value.

[0070] The number of elements in (A) above may be, for example, the total number of pairs of first variables and second variables included in the second data. The number of elements may be, for example, the number of pairs of first variables and second variables measured in the most recent predetermined number of tests. In this case, the number of elements corresponds to the most recent predetermined number of tests. The most recent predetermined number of tests may be set, for example, through experiments, so that the prediction model 421 can be changed in accordance with changes in the trend of the predicted value of discharge capacity.

[0071] The determination unit 617 may compare each second Mahalanobis distance with a first threshold and identify second Mahalanobis distances equal to or greater than the first threshold. The determination unit 617 may count the second Mahalanobis distances equal to or greater than the first threshold and identify the counting result as the number of pairs of first variables and second variables for which second Mahalanobis distances equal to or greater than the first threshold are calculated. The first and second set values ​​may be set, for example, through experiments, so that the prediction model 421 can be changed in accordance with changes in the trend of the predicted value of the discharge capacity. The first set value may be set to a value at least greater than 0, and the second set value may be set to a value at least greater than 1.

[0072] Furthermore, when the determination unit 617 determines that any of the second Mahalanobis distances calculated for the second data is equal to or greater than a predetermined second threshold, the determination unit 617 may determine that the second data is dissimilar to the first data. The second threshold is a threshold for the second Mahalanobis distances and is equal to or greater than the first threshold.

[0073] If there is a prominent outlier in the second data, the determination unit 617 can determine that the second data is not similar to the first data regardless of the result of the similarity determination using the first threshold. For example, even if the ratio in (A) is less than the first set value or the number of pairs in (B) is less than the second set value, the determination unit 617 can determine that the second data is not similar to the first data if any of the second Mahalanobis distances is equal to or greater than the second threshold.

[0074] The second threshold may be set so that the second Mahalanobis distance can be identified as a prominent outlier in the second data. Therefore, the second threshold may be set to a value equal to or greater than the first threshold set by the setting unit 616. However, the second threshold does not have to be set after the first threshold is set by the setting unit 616, and may be set in advance to a value equal to or greater than the first threshold that can be predicted by, for example, experimentation or the like.

[0075] The notification unit 618 may notify the model construction device 5 of the type of second variables constituting the second data determined to be dissimilar to the first data. This makes it possible to prevent second variables constituting the second data that have low similarity to the first data from being used in reconstructing the prediction model 421. Therefore, the model construction device 5 can reconstruct the prediction model 421 in accordance with changes in the trend of the predicted value of the discharge capacity.

[0076] For example, if the determination unit 617 determines in (A) above that the ratio is equal to or greater than a first set value, the notification unit 618 may notify the model construction device 5 of the type of second variables constituting the second data that was the subject of this determination. If the determination unit 617 determines in (B) above that the number of pairs is equal to or greater than a second set value, the notification unit 618 may notify the model construction device 5 of the type of second variables constituting the second data that was the subject of this determination. If the determination unit 617 determines that any of the second Mahalanobis distances calculated for the second data is equal to or greater than a second threshold, the notification unit 618 may notify the model construction device 5 of the type of second variables constituting the second data that was the subject of this determination.

[0077] The model construction device 5, which has received the notification from the notification unit 618, may exclude the second variables of the type indicated in the notification and reconstruct the prediction model 421. In other words, the model construction device 5 does not use the second variables constituting the second data determined by the determination unit 617 to be not similar to the first data as variables for predicting the discharge capacity.

[0078] In this embodiment, the determination unit 617 may perform a determination using the first set value or a determination using the second set value, and may also perform a determination using the second threshold value, provided that the determination unit 617 only needs to perform at least one of the determination using the first set value, the determination using the second set value, and the determination using the second threshold value.

[0079] Furthermore, when the control unit 61 determines that any of the second Mahalanobis distances calculated in the second data is equal to or greater than a preset third threshold, the control unit 61 may output identification information of the secondary battery for which the second Mahalanobis distance was calculated to the inspection device 3. The third threshold may be set so that a second Mahalanobis distance that is an outstanding outlier in the second data can be identified.

[0080] The inspection device 3 may not adopt the predicted value of discharge capacity output from the prediction device 4 for a secondary battery for which a second Mahalanobis distance equal to or greater than the third threshold value has been calculated. The inspection device 3 may perform the discharging step of the capacity inspection process described above instead of the second inspection step for this secondary battery. That is, the inspection device 3 may calculate the discharge capacity for this secondary battery using the discharge current and discharge time measured in this discharging step. Furthermore, the inspection device 3 may identify this secondary battery as a defective product.

[0081] <Processing Flow of Model Determination Device> Fig. 4 is a flowchart showing an example of the processing flow in the control unit 61 of the model determination device 6. Fig. 5 is a table showing an example of various data related to the first data. Fig. 6 is a table showing an example of various data related to the second data. The processing flow shown in Fig. 4 may be an example of a model determination method for determining the quality of the prediction model 421. The control unit 61 may start this processing, for example, when it is determined that inspection of a secondary battery using the prediction model 421 stored in the prediction device 4 has been performed for a predetermined period of time.

[0082] For the first data and second data used in the subsequent processing, the first variables and second variables constituting these data are of the same type. When there are multiple first data with different types of second variables, there are multiple pairs of first data and second data in which the first variables are of the same type and the second variables are of the same type. In this case, the control unit 61 may perform the subsequent processing for each pair of first data and second data.

[0083] For example, in the example of Fig. 5, there are a plurality of first data of different types, in which the type of the first variable is "V1" and the types of the second variables are "V2", "V3", "V4", .... Also, in the example of Fig. 6, there are a plurality of second data of different types but the same type as the first data, in which the type of the first variable is "V1" and the types of the second variables are "V2", "V3", "V4", ....

[0084] First, the first creation unit 611 may create first data indicating a set of multiple pairs of first variables and second variables to be used when constructing a prediction model (S11; first creation step). For example, in Fig. 5, the first data, in which the type of first variable is "V1" and the type of second variable is "V2", includes multiple pairs such as a pair of the first variable "Aa" and the second variable "Ba", a pair of the first variable "Ab" and the second variable "Bb", a pair of the first variable "Ac" and the second variable "Bc", and so on.

[0085] Furthermore, the second creation unit 612 may create second data indicating a set of multiple pairs of first variables and second variables measured from the secondary battery for which the predicted value of discharge capacity is to be output (S12; second creation step). For example, in Fig. 6, the second data in which the type of first variable is "V1" and the type of second variable is "V2" includes multiple pairs of the first variable "Ax" and the second variable "Bx", the first variable "Ay" and the second variable "By", the first variable "Az" and the second variable "Bz", and so on.

[0086] Next, the first calculation unit 613 may calculate a first Mahalanobis distance for each of the multiple pairs included in the first data created by the first creation unit 611 (S13). For example, as shown in Fig. 5, in first data in which the first variable type is "V1" and the second variable type is "V2," the first Mahalanobis distance is calculated for each pair of the first variable and the second variable. For other types of first data, such as first data in which the first variable type is "V1" and the second variable type is "V3," the first Mahalanobis distance is also calculated for each pair of the first variable and the second variable.

[0087] Furthermore, the second calculation unit 614 may calculate a second Mahalanobis distance for each of a plurality of pairs included in the second data created by the second creation unit 612 (S14). For example, as shown in Fig. 6, in second data in which the first variable type is "V1" and the second variable type is "V2," the second Mahalanobis distance is calculated for each pair of the first variable and the second variable. For other types of second data, such as second data in which the first variable type is "V1" and the second variable type is "V3," the first Mahalanobis distance is also calculated for each pair of the first variable and the second variable.

[0088] Next, the third calculation unit 615 may calculate a standard deviation for the plurality of first Mahalanobis distances calculated by the first calculation unit 613. For example, as shown in FIG. 5 , the standard deviation for the first Mahalanobis distances in the first data in which the first variable type is "V1" and the second variable type is "V2" is calculated as "L." The standard deviation for the first Mahalanobis distances in the first data in which the first variable type is "V1" and the second variable type is "V3" is calculated as "M." The standard deviation for the first Mahalanobis distances in the first data in which the first variable type is "V1" and the second variable type is "V4" is calculated as "N."

[0089] The setting unit 616 may set a first threshold value for the second Mahalanobis distance based on the standard deviation calculated by the third calculation unit 615 (S15). For example, as shown in FIG. 5 , the first threshold value for the first data in which the first variable type is "V1" and the second variable type is "V2" is set to "3L," which is three times the standard deviation. The first threshold value for the first data in which the first variable type is "V1" and the second variable type is "V3" is set to "3M," which is three times the standard deviation. The first threshold value for the first data in which the first variable type is "V1" and the second variable type is "V4" is set to "4M," which is four times the standard deviation. In this way, the setting unit 616 may set the first threshold value by changing the multiple of the standard deviation depending on the type of the second variable.

[0090] Next, the determination unit 617 may determine whether any of the second Mahalanobis distances calculated by the second calculation unit 614 is equal to or greater than a second threshold (S16). The processing of S16 is an example of a determination step of determining whether the second data is similar to the first data based on the results of statistical processing on the first data and the second data.

[0091] If the determination unit 617 determines that any of the second Mahalanobis distances is equal to or greater than the second threshold (YES in S16), it may determine that the second data is dissimilar to the first data. In this case, the notification unit 618 may notify the model construction device 5 of the type of second variables constituting the second data that is dissimilar to the first data (S17). This makes it possible to prevent the second variables from being used in the reconstruction of the prediction model 421.

[0092] On the other hand, if the determination unit 617 determines that all of the second Mahalanobis distances are less than the second threshold value (NO in S16), it may determine whether the second data is similar to the first data using the second Mahalanobis distance and the first threshold value (S18). The process of S18 is an example of the above-mentioned determination step. An example of the process of S18 is the determination process using the above-mentioned first set value or second set value.

[0093] As a result of the process of S18, if the determination unit 617 determines that the second data is similar to the first data (YES in S19), the control unit 61 may end this process. On the other hand, if the determination unit 617 determines that the second data is not similar to the first data (NO in S19), the notification unit 618 may execute the process of S17. In this case, too, it is possible to prevent the second variables constituting the second data that are not similar to the first data from being used in reconstructing the prediction model 421.

[0094] The above-described process flow is merely an example. The first data creation process (S11), the first Mahalanobis distance calculation process (S13), and the first threshold value setting process (S15) may be performed before the second data creation process (S12). The processes of S11, S13, and S15 may be performed, for example, when the prediction model 421 is constructed. The control unit 61 may store the first threshold value in the storage unit 62 and use it in the process of determining the similarity between the first data and the second data using the first threshold value (S18). The control unit 61 may also store the first data and the first Mahalanobis distance in the storage unit 62 before the process of S12. Furthermore, the processes of S11, S13, and S15 may be omitted unless the first data is changed. For example, the processes of S11, S13, and S15 may be omitted unless the prediction model 421 is reconstructed.

[0095] Furthermore, the judgment using the second Mahalanobis distance and the second threshold value (S16) and the judgment using the second Mahalanobis distance and the first threshold value (S18 and S19) may be performed in parallel or in the reverse order.

[0096] [Embodiment 2] In the inspection system 1 of embodiment 1, when inspecting a secondary battery manufactured for shipment, the discharge capacity is predicted using the prediction model 421 constructed by the model construction device 5. In the inspection system 1 of embodiment 1, when inspecting a secondary battery manufactured for shipment, the appropriateness of the output result of the established prediction model 421 is determined, thereby determining whether the prediction model 421 is suitable for the current specifications of the secondary battery.

[0097] However, the application of the inspection system 1 of the first embodiment is not limited to determining the appropriateness of the output result of the established prediction model 421. The inspection system 1 can also be applied at the stage when the model construction device 5 constructs the prediction model 421. In this embodiment, a prediction model that the model construction device 5 constructs before establishing the prediction model 421 to be used in inspecting secondary batteries manufactured for shipment is referred to as a provisional prediction model 421.

[0098] In this case, in the model determination device 6, the first creation unit 611 may create first data indicating a set consisting of multiple pairs of first variables and second variables used when constructing the tentative prediction model 421. The second creation unit 612 may create second data indicating a set consisting of multiple pairs of first variables and second variables measured from the secondary battery that is the target for outputting a predicted value.

[0099] In this embodiment, the secondary battery to which the predicted value is output may be a secondary battery manufactured by the manufacturing apparatus 2 to determine the suitability of the output result of the tentative prediction model 421 and to be inspected by the inspection apparatus 3. Furthermore, the number of measurements for this determination, i.e., the number of pairs of the first variable and the second variable that are configured as the second data for this determination, may be set in advance.

[0100] As described in the first embodiment, the determination unit 617 may determine whether the second data is similar to the first data based on the results of statistical processing on the first data and the second data. If the determination unit 617 determines that the second data is not similar to the first data, the notification unit 618 may notify the model construction device 5 of the type of second variables constituting the second data. This allows the model construction device 5 to reconstruct the tentative prediction model 421.

[0101] For example, if the model determination device 6 performs a similarity determination of the second data and does not receive a notification from the notification unit 618, the model construction device 5 may establish the constructed prediction model 421 as the prediction model 421 to be used for inspecting secondary batteries manufactured for shipment.

[0102] [Example of Implementation by Software] The functions of the model determination device 6 (hereinafter referred to as the "device") can be realized by a program that causes a computer to function as the device, and a program that causes a computer to function as each control block of the device. Examples of each control block include, in particular, each unit included in the control unit 61.

[0103] In this case, the device includes a computer having at least one control device (e.g., a processor) and at least one storage device (e.g., a memory) as hardware for executing the program. The functions described in each of the above embodiments are realized by executing the program using the control device and storage device.

[0104] The program may be non-transitory and may be recorded on one or more computer-readable recording media. The recording media may or may not be included in the device. In the latter case, the program may be supplied to the device via any wired or wireless transmission medium.

[0105] In addition, some or all of the functions of each of the control blocks can be realized by logic circuits. For example, integrated circuits in which logic circuits that function as each of the control blocks are formed are also included in the scope of the present disclosure. In addition, the functions of each of the control blocks can also be realized by, for example, a quantum computer.

[0106] Furthermore, each process described in each of the above embodiments may be executed by AI (Artificial Intelligence). In this case, the AI ​​may run on the control device or on another device (for example, an edge computer or a cloud server).

[0107] The device may also be, for example, a prediction device 4. In this case, the control block may in particular be a prediction unit 41.

[0108] [Additional Notes] The invention according to the present disclosure has been described above based on the drawings and examples. However, the invention according to the present disclosure is not limited to the above-described embodiments. In other words, the invention according to the present disclosure can be modified in various ways within the scope of the present disclosure, and embodiments obtained by appropriately combining the technical means disclosed in different embodiments are also included in the technical scope of the invention according to the present disclosure. In other words, it should be noted that a person skilled in the art could easily make various modifications or corrections based on the present disclosure. It should also be noted that these modifications or corrections are included in the scope of the present disclosure.

[0109] 4 Prediction device 6 Model determination device 41 Prediction unit 421 Prediction model 611 First creation unit 612 Second creation unit 615 Third calculation unit 616 Setting unit 617 Determination unit 618 Notification unit

Claims

1. A model determination device comprising: a first creation unit that creates first data indicating a set consisting of multiple pairs of a first variable and a second variable different from the first variable, which is used when constructing a predictive model that outputs a predicted value of the discharge capacity of a manufactured secondary battery; a second creation unit that creates second data indicating a set consisting of multiple pairs of the first variable and the second variable, which are measured when inspecting the secondary battery that is the target for outputting the predicted value; and a determination unit that determines whether the second data is similar to the first data based on the results of statistical processing of the first data and the second data.

2. A model determination device as described in claim 1, further comprising: a first calculation unit that calculates a first Mahalanobis distance for each of the plurality of pairs included in the first data; a second calculation unit that calculates a second Mahalanobis distance for each of the plurality of pairs included in the second data; a third calculation unit that calculates a standard deviation for the calculated plurality of first Mahalanobis distances; and a setting unit that sets a first threshold value for the second Mahalanobis distance based on the standard deviation, wherein the determination unit uses the second Mahalanobis distance and the first threshold value to determine whether the second data is similar to the first data.

3. The model determination device of claim 2, wherein the determination unit determines that the second data is not similar to the first data when it determines that the ratio of the number of pairs of the first variable and the second variable for which the second Mahalanobis distance equal to or greater than the first threshold value to the number of elements that are at least a portion of the pairs included in the second data is equal to or greater than a first set value set in advance.

4. A model determination device as described in claim 2, wherein the determination unit determines that the second data is not similar to the first data when it determines that the number of pairs of the first variable and the second variable for which the second Mahalanobis distance equal to or greater than the first threshold is equal to or greater than a predetermined second set value.

5. A model determination device as claimed in any one of claims 2 to 4, wherein a second threshold value for the second Mahalanobis distance is set in advance and is equal to or greater than the first threshold value, and the determination unit determines that the second data is not similar to the first data if it determines that any of the second Mahalanobis distances is equal to or greater than the second threshold value.

6. A model determination device as described in any one of claims 1 to 5, further comprising a notification unit that notifies a model construction device that constructs the predictive model of the type of the second variable constituting the second data that is determined to be not similar to the first data.

7. A model determination device as claimed in any one of claims 1 to 6, wherein the first variable is the charge amount of the secondary battery, which is the measurement target in the second inspection process, between a first inspection process for inspecting whether the secondary battery is a non-defective product and a second inspection process for inspecting the capacity of the secondary battery.

8. A model determination device described in any one of claims 1 to 7, wherein the second variable is one or more of a plurality of variables that are to be measured when inspecting the secondary battery in a first inspection process for inspecting whether the secondary battery is a non-defective product and in a second inspection process for inspecting the capacity of the secondary battery.

9. A prediction device that predicts the discharge capacity using the prediction model that is the subject of judgment by the model judgment device described in any one of claims 1 to 8, comprising a prediction unit that outputs a predicted value of the discharge capacity by inputting measurement values ​​measured when inspecting the secondary battery into the prediction model in a first inspection process for inspecting whether the secondary battery is a non-defective product and a second inspection process for inspecting the capacity of the secondary battery.

10. A prediction device as described in claim 9, wherein the prediction unit does not input the measured charge amount measured when inspecting the secondary battery in the second inspection process into the prediction model if the measured charge amount is less than or equal to a predetermined reference value.

11. A model judgment method comprising: a first creation step of creating first data indicating a set consisting of a plurality of pairs of a first variable and a second variable different from the first variable, which is used when constructing a prediction model that outputs a predicted value of the discharge capacity of a manufactured secondary battery; a second creation step of creating second data indicating a set consisting of a plurality of pairs of the first variable and the second variable, which are measured when inspecting the secondary battery that is the target for outputting the predicted value; and a judgment step of judging whether the second data is similar to the first data based on the results of statistical processing of the first data and the second data.

Images