Defect Type Classifying System and Defect Type Classifying Method

Abstract

A defect type classifying system includes: a measurer connected to a battery from which a defect type is detected, and measuring at least one of electrical characteristics of the battery for a predetermined test period and generating measured data; a converter for generating input data by converting the measured data; and a defect type predicting module for machine-learning learning data, and determining a defect type of the battery based on the input data. The input data may be appropriate for an input node of the defect type predicting module, and the defect type may be classified according to a cause of a short-circuit defect of the battery.

Claims

1. A defect type classifying system for a battery comprising: a measurer connected to the battery, the measurer configured to detect a defect type, the measurer configured to measure at least one of electrical characteristics of the battery for a predetermined test period, the measurer configured to generate measured data; a converter configured to generate input data by converting the measured data; and a defect type predicting module configured to use machine-learning learning data, the defect type predicting module configured to determine a defect type of the battery based on the input data, wherein the input data are appropriate for an input node of the defect type predicting module, and the defect type is classified according to a cause of a short-circuit defect of the battery.

2. The defect type classifying system of claim 1, wherein the measurer is configured to measure a battery voltage that is a voltage at respective ends of the battery during a test period and is configured to store the measured data indicating the measured result.

3. The defect type classifying system of claim 2, wherein the converter is configured to analyze the measured data and is configured to extract at least one of a maximum value, a minimum value, a mean value, or a center value of the battery voltage and a number of peaks of the battery voltage to generate the input data.

4. The defect type classifying system of claim 1, wherein the measured data are based on a result obtained by measuring a battery voltage that is a voltage at respective ends of the battery for the predetermined test period, and the input data include at least one of a maximum value, a minimum value, a mean value, a center value of the battery voltage, or a number of peaks of the battery voltage.

5. The defect type classifying system of claim 1, further comprising a database configured to store the learning data, wherein the database is configured to store the input data based on a voltage at respective ends of the battery and a defect type when a short-circuit defect of the battery is determined.

6. The defect type classifying system of claim 1, wherein the defect type predicting module is configured to calculate deviations that are distances of differences between at least one of electrical characteristics measured for a target battery of which a defect type is classified and respective values of the learning data, the defect type predicting module is configured to deduce a predetermined number of reference data in least order of the calculated deviations, and the defect type predicting module is configured to classify the defect type of the target battery with the class that has the greatest result of counting the classes to which the deduced reference data respectively belong.

7. The defect type classifying system of claim 6, wherein the defect type predicting module is configured to receive at least one of electrical characteristics from the learning data, and the defect type predicting module is configured to learn to predict a defect type to be one of a plurality of classes that corresponds to the received one of the electrical characteristics.

8. A defect type classifying method comprising: learning, by a defect type predicting module, to classify a defect type according to a cause of a short-circuit defect of a battery based on learning data; measuring, by a measurer, at least one of electrical characteristics of the battery for a test period and generating measured data; converting, by a converter, the measured data and generating input data; and determining, by the defect type predicting module, a defect type of the battery based on the input data, wherein the input data are appropriate for an input node of the defect type predicting module.

9. The defect type classifying method of claim 8, wherein the generating of the measured data includes measuring a battery voltage that is a voltage at respective ends of the battery for the test period, and generating the measured data indicating a measured result.

10. The defect type classifying method of claim 9, wherein the generating of the input data includes analyzing the measured data, extracting at least one of a maximum value, a minimum value, a mean value, or a center value of the battery voltage and a number of peaks of the battery voltage, and generating the input data.

11. The defect type classifying method of claim 8, wherein the measured data are based on a result obtained by measuring a battery voltage that is a voltage at respective ends of the battery for the test period, and the input data include at least one of a maximum value, a minimum value, a mean value, or a center value of the battery voltage and a number of peaks of the battery voltage.

12. The defect type classifying method of claim 8, further comprising: storing, by a database, the input data based on a voltage at respective ends of the battery and the defect type when a short-circuit defect of the battery is generated as the learning data.

13. The defect type classifying method of claim 8, wherein the determining of a defect type includes calculating deviations that are distance differences between at least one of electrical characteristics measured for the target battery from which the defect type is classified and respective values belonging to the learning data; and deducing a number of reference data in least order from among the calculated deviations, and classifying the defect type of the target battery with the class that has a greatest result of counting classes to which the deduced reference data respectively belong.

14. The defect type classifying method of claim 13, wherein the learning to classify the defect type includes: receiving at least one of electrical characteristics from the learning data; and learning to predict the defect type that corresponds to one of the received electrical characteristics to be one of a plurality of classes.

Description

DESCRIPTION OF THE DRAWINGS

[0021] FIG. 1 shows a block diagram pf a defect type classifying system according to an embodiment.

[0022] FIG. 2 shows a graph of a method for generating input data from measured data generated by measuring battery voltages according to an embodiment.

[0023] FIG. 3 shows an operation of a defect type predicting module according to an embodiment.

MODE FOR INVENTION

[0024] The present disclosure includes two stages so as to efficiently learn a feature value of a faulty battery. A first machine learning stage includes allowing a defect type predicting module to machine-learn a feature value of a faulty battery, and a second classifying stage includes allowing the machine-learning defect type predicting module to determine a defect cause of the faulty battery. Regarding the first machine learning stage, the defect type predicting module learns electrical feature values such as the measured voltages or currents of the faulty battery and the defect causes of the corresponding faulty battery. Regarding the second classifying stage, the electrical feature values of the defect-generated battery are measured, the measured feature values are input to the defect type predicting module, and the defect type predicting module may determine the defect cause based on the input feature values.

[0025] Hereinafter, embodiments disclosed in the present specification will be described in detail with reference to the accompanying drawings. In the present specification, the same or similar components will be denoted by the same or similar reference numerals, and an overlapped description thereof will be omitted. The terms module and unit for components used in the following description are used only in order to make the specification easier, and hence, these terms do not have meanings or roles that distinguish them from each other by themselves. In describing embodiments of the present specification, when it is determined that a detailed description of the well-known art associated with the present invention may obscure the gist of the present invention, it will be omitted. The accompanying drawings are provided only in order to allow embodiments disclosed in the present specification to be easily understood and are not to be interpreted as limiting the spirit disclosed in the present specification, and it is to be understood that the present invention includes all modifications, equivalents, and substitutions without departing from the scope and spirit of the present invention.

[0026] Terms including ordinal numbers such as first, second, and the like, will be used only to describe various components, and are not interpreted as limiting these components. The terms are only used to differentiate one component from other components.

[0027] It is to be understood that when one component is referred to as being connected or coupled to another component, it may be connected or coupled directly to another component or be connected or coupled to another component with the other component intervening therebetween. On the other hand, it is to be understood that when one component is referred to as being connected or coupled directly to another component, it may be connected to or coupled to another component without the other component intervening therebetween.

[0028] It will be further understood that terms comprises or have used in the present specification specify the presence of stated features, numerals, steps, operations, components, parts, or a combination thereof, but do not preclude the presence or addition of one or more other features, numerals, steps, operations, components, parts, or a combination thereof.

[0029] FIG. 1 shows a block diagram pf a defect type classifying system according to an embodiment.

[0030] As shown in FIG. 1, the defect type classifying system 1 includes a measurer 10, a converter 20, a defect type predicting module 30, and a database 40.

[0031] The measurer 10 may be connected to a battery 2 for detecting a defect type, may measure at least one of the electrical characteristics of the battery 2 for a predetermined test period, and may generate measured data based on a measured result, and may store the measured data. The measurer 10 may include a memory 11 for storing the measured data. The memory 11 may be a non-volatile memory for storing the measured data. When the battery for detecting a defect type is replaced, the measurer 10 may erase content stored in the memory 11 and may write new measured data.

[0032] For example, the measurer 10 may measure a battery voltage that is a voltage between a positive electrode and a negative electrode of the battery 2 for a test period, and may store measured data for indicating the battery voltage measured for the test period in the memory 11. The present disclosure is not limited to use the battery voltage so as to detect the defect type. One or more of the battery voltage, the battery current flowing to the battery, internal resistance of the battery, inductance of the battery, and capacitance of the battery may be used to detect the defect type of the battery.

[0033] The converter 20 converts the measured data and generates input data that are appropriate for the defect type predicting module 30. The input data are determined according to a value to be input to an input node of the defect type predicting module 30. The converter 20 reads the measured data from the memory 11, analyzes the read measured data, extracts characteristics of the analyzed measured data, and generates input data.

[0034] FIG. 2 shows a graph of a method for generating input data from measured data generated by measuring battery voltages according to an embodiment.

[0035] FIG. 2 shows an example for describing one of the various methods for generating input data, which does not limit the present disclosure.

[0036] As shown in FIG. 2, for the test period, the measurer 10 measures a voltage at respective ends of the battery 2, generates measured data on the battery voltage, and stores the generated measured data in the memory 11.

[0037] The converter 20 reads the measured data from the memory 11, analyzes the read measured data, extracts a maximum value, a minimum value, a mean value, and a center value that are characteristics of the measured data, and the number of peaks from the measured data, and generates input data. That is, the input data include the maximum value, the minimum value, the mean value, the center value of the battery voltage, and the number of the peaks of the battery voltage for the test period.

[0038] The defect type predicting module 30 machine-learns learning data stored in the database 40. The database 40 classifies the input data for respective defect types and stores the same, and the corresponding input data and the defect type may be updated in the database 40 each time the defect is generated in the battery. Here, the input data may include at least one of the maximum value, the minimum value, the mean value, the center value of the battery voltage, and the number of the peaks of the battery voltage.

[0039] The defect type predicting module 30 may input the input data stored for respective defect types to the input node, and may perform machine learning for predicting the defect type. The period for the defect type predicting module 30 to perform machine learning may be determined according to an amount of the learning data updated into the database 40. For example, each time a meaningful number of learning data are updated to the database 40, the defect type predicting module 30 may include the updated learning data and may perform machine learning.

[0040] The defect type predicting module 30 may be realized as a processor for performing a program having realized one of a supervised machine learning algorithm, an unsupervised machine learning algorithm, a semi-supervised machine learning algorithm, and a reinforced machine learning algorithm. When the defect type predicting module 30 is a processor for performing a program having realized a neural network algorithm, the defect type predicting module 30 may perform neural network learning to classify the defect type with label-assigned learning data. The defect type predicting module 30 may repeatedly perform neural network learning for predicting the defect type with the greatest probability from among a plurality of defect types that may be generated to the classify target based on the learning data. The defect type predicting module 30 may adjust weight values between activation functions of respective nodes of the neural network and the nodes so that the defect type predicted through the repeated neural network learning may approach the actual defect type.

[0041] The defect type predicting module 30 built through the above-noted machine learning may determine the defect type based on the input data. For example, when the input data are at least one of the maximum value, the minimum value, the center value, and the mean value of the battery voltage, and the number of peaks of the battery voltage, the defect type predicting module 30 may apply the respective elements configuring the input data to the corresponding input node and may determine one of a plurality of defect types. The number of the input nodes may be determined according to the number of the elements configuring the input data. The defect types may be classified according to causes of the short-circuit defect of the battery, and for example, it may include tab folded, broken, stabbed, and wrinkling of a separation film. The broken and the stabbed may respectively signify that parts of elements of the battery cell are broken or stabbed.

[0042] FIG. 3 shows an operation of a defect type predicting module according to an embodiment.

[0043] The defect type predicting module may be realized with a KNN classified model (K-Nearest-Neighbor classification) from among the machine learning models. In addition, regarding the classifying method described with reference to FIG. 3, the input data include the maximum value of the battery voltage and the number of the peaks of the battery voltage.

[0044] In the machine learning stage, the defect type predicting module 30 receives the maximum value of the battery voltage that is a value of the X-axis and the number of the peaks that is the number of the peaks of the battery voltage that is a value of the Y-axis, and is learned to output one of stabbed and broken as the defect cause. As shown in FIG. 3, the defect type predicting module 30 may perform learning on the batteries (class A) in which the cause of defect marked with stars is stabbed and the batteries (class B) in which the cause of defect marked with triangles is broken.

[0045] When finishing the machine learning, the defect type predicting module 30 may receive the maximum value Xi of the battery voltage corresponding to the position marked with quadrangles and the number Yi of the peaks of the battery voltage in FIG. 3, and may predict the cause of a defect of the battery according to the corresponding received value and the number.

[0046] In detail, the defect type predicting module 30 may receive the maximum value Xi of the battery voltage measured from the battery (hereinafter, target battery) for classifying the defect type and the number Yi of the peaks of the battery voltage.

[0047] As shown in FIG. 3 (b), the defect type predicting module 30 may calculate distance differences (hereinafter, deviations) between a position (hereinafter, target position) defined by the maximum value Xi of the battery voltage and the number Yi of the peaks of the battery voltage and all the positions (hereinafter, compared positions) corresponding to the learned data. The defect type predicting module 30 may deduce a predetermined number of reference data in least an order from among a plurality of deviations between the target position and the compared positions. For example, the defect type predicting module 30 may calculate first and second distance differences in the X-axis and the Y-axis between the target position and one of the compared positions, may perform a square root process on a value generated by squaring the calculated first and second distance differences and adding the same, and may then calculate deviations. The defect type predicting module 30 calculate a plurality of deviations between the target position and the compared positions according to the above-described method. The defect type predicting module 30 may arrange the calculated deviations in ascending order, may deduce a predetermined number of deviations in a least order, and may determine data of the compared positions corresponding to the deduced deviations to be reference data. However, the method for calculating deviations is not limited to the above-described method in the present disclosure.

[0048] As shown in FIG. 3 (c), the defect type predicting module 30 may position the reference data deduced from among the class A and the class B in a dotted circle. The defect type predicting module 30 may count the classes to which the respective reference data belong, and may classify the defect type of the target battery with the class that has a greater counting result. Referring to FIG. 3 (c), as the counting result of the class A is 1 and the counting result of the class B is 2, the defect type of the target battery may be predicted to be the class B (broken).

[0049] Regarding the example described with reference to FIG. 3, a dimension on the defect type is determined with the X-axis and the Y-axis, and types of the axis may increase such as the X-axis, the Y-axis, the Z-axis, . . . according to the number of the input data, so the number of the dimensions on the defect types may increase.

[0050] According to the above-described method, when the maximum value of the battery voltage of the target battery is 37V and the number of the peaks of the battery voltage is 3, the defect type predicting module 30 may classify the target battery into the class B (cause of defect: broken). In another way, when the maximum value of the battery voltage of the target battery is 47 V and the number of the peaks of the battery voltage is 5, the defect type predicting module 30 may classify the target battery into the class A (cause of defect: stabbed).

[0051] While this invention has been described in connection with what is presently considered to be practical embodiments, it is to be understood that the invention is not limited to the disclosed embodiments, but, on the contrary, is intended to cover various modifications and equivalent arrangements included within the spirit and scope of the appended claims.