KR20230157582A - Apparatus and method for analyzing charge behavior inside oled device based on machine learning - Google Patents

Abstract

The present invention relates to an apparatus and method for analyzing charge behavior inside an OLED device by predicting the charge density for each internal location of an OLED device through a machine learning model based on frequency-based measurement values such as impedance, and machine learning according to the present invention. According to the OLED device internal charge behavior analysis device and method, frequency-based measured values, including impedance, can be obtained from the OLED device to be inspected, and frequency-based characteristic data can be generated by performing preprocessing on the measured values. Charge distribution data representing the charge density for each location inside the device can be derived from frequency-based characteristic data using a machine learning model. Therefore, even if charge mobility is not individually analyzed by creating a separate test device for each of the organic layers constituting the OLED device, it is possible to analyze where the charge is accumulated inside the OLED device.

Description

Machine learning-based organic light emitting diode (OLED) device internal charge behavior analysis device and method {APPARATUS AND METHOD FOR ANALYZING CHARGE BEHAVIOR INSIDE OLED DEVICE BASED ON MACHINE LEARNING}

The present invention relates to a machine learning-based device and method for analyzing the internal charge behavior of an organic light-emitting diode (OLED) device. More specifically, the internal charge behavior of an OLED device is analyzed through a machine learning model based on frequency-based measurements such as impedance. This relates to an apparatus and method for analyzing charge behavior inside a device by predicting charge density for each location.

In order for an organic light emitting diode (OLED) device to operate smoothly and emit light, the mobility of holes and electrons must be maintained high within the OLED device. However, when holes or electrons are trapped in some organic layers of an OLED device and charges are accumulated, a local electric field is formed due to the accumulated charges, which may reduce the mobility of holes and electrons and reduce the luminous efficiency of the OLED device.

In order to prevent such a decrease in the luminous efficiency of the OLED device, it may be required to analyze the charge behavior inside the device to reveal in which organic layer of the OLED device the charge is accumulated. To analyze charge behavior, conventionally, a hole-only device (HOD) and an electron-only device (EOD) were manufactured separately for each of the organic layers constituting the OLED device to measure charge mobility. Although a measuring method has been used, this conventional method may be problematic in that it requires a large amount of separate devices to be manufactured and each device must be analyzed one by one, which requires a lot of cost and time.

Patent Document 1: Registered Patent Publication No. 10-2230354 Patent Document 2: Registered Patent Publication No. 10-2293791

The technical problem to be solved by the present invention is to predict the charge density for each internal location of the OLED device without manufacturing separate analysis devices by improving the conventional inefficient method of analyzing the charge behavior inside the OLED device. To provide a device and method for analyzing charge behavior.

As a means to solve the above-described technical problem, a machine learning-based organic light emitting diode (OLED) device internal charge behavior analysis device according to some embodiments of the present invention performs alternating current testing for each of the test voltage strengths and test frequencies. A measuring unit configured to obtain frequency-based measurement values by applying a voltage to the OLED device to be inspected; Frequency-based derived variable values are calculated by performing conversion processing on the frequency-based measured values, and frequency-based characteristic data representing changes in the frequency-based derived variable values according to changes in the test voltage intensities and the test frequencies are generated. Generate, and utilize a machine learning model that is learned to predict the charge density for each location inside the OLED device from the frequency-based characteristic data, and derive charge distribution data representing the charge density for each location inside the OLED device to be inspected. processing department; and an output unit configured to output the charge distribution data; Includes.

The machine learning-based method of analyzing the internal charge behavior of an organic light-emitting diode (OLED) device according to another embodiment of the present invention applies an alternating current test voltage to the OLED device to be tested for each of the test voltage strengths and test frequencies through the measuring unit. applying and obtaining frequency-based measurements; Calculating frequency-based derived variable values by performing conversion processing on the frequency-based measured values through a processing unit; generating, through the processing unit, frequency-based characteristic data representing changes in the frequency-based derived variable values according to changes in the test voltage intensities and the test frequencies; Deriving charge distribution data representing the charge density at each internal location of the OLED device to be inspected using a machine learning model that is learned to predict the charge density at each location inside the OLED device from the frequency-based characteristic data through the processing unit. ; and outputting the charge distribution data through an output unit. Includes.

According to the machine learning-based OLED device internal charge behavior analysis device and method according to the present invention, frequency-based measurement values, including impedance, can be obtained from the OLED device to be inspected, and preprocessing processes are performed on the measurement values to determine the frequency. Based characteristic data can be generated, and charge distribution data representing the charge density for each location inside the device can be derived from frequency-based characteristic data using a machine learning model. Therefore, even if charge mobility is not individually analyzed by creating a separate test device for each of the organic layers constituting the OLED device, it is possible to analyze where the charge is accumulated inside the OLED device.

Figure 1 is a diagram to explain how charge mobility is reduced by charges accumulated in some organic layers of an OLED device.
Figure 2 is a diagram showing exemplary data for evaluating existing OLED devices.
Figure 3 is a diagram for explaining how the machine learning-based charge behavior analysis device inside an organic light-emitting diode (OLED) device operates according to the present invention.
Figure 4 is a diagram for explaining the elements that constitute the machine learning-based internal charge behavior analysis device of an organic light-emitting diode (OLED) device according to the present invention.
Figure 5 is a diagram for explaining the types of frequency-based measurement values measured from the OLED device to be inspected according to the present invention.
Figure 6 is a diagram illustrating the process of generating data input to a machine learning model by performing preprocessing processes on frequency-based measurement values according to the present invention.
Figure 7 is a diagram for explaining examples of frequency-based measurement values and frequency-based characteristic data generated correspondingly according to the present invention.
Figure 8 is a diagram for explaining a machine learning model implemented in the form of a convolutional neural network (CNN) according to the present invention.
Figure 9 is a diagram for comparing the performance of a machine learning model implemented in the form of a convolutional neural network (CNN) according to the present invention and the performance of a conventional model.
Figure 10 is a diagram showing implementation examples of OLED devices to be inspected according to the present invention and evaluation data for them.
Figure 11 is a diagram for explaining charge distribution data showing the charge density at each internal location of the OLED device to be inspected according to the present invention.
Figure 12 is a diagram for explaining the steps of configuring the method of analyzing charge behavior inside a machine learning-based organic light-emitting diode (OLED) device according to the present invention.

Hereinafter, embodiments of the present invention will be described in detail with reference to the drawings. The description below is only intended to specify embodiments and is not intended to limit or limit the scope of rights according to the present invention. What a person skilled in the art related to the present invention can easily infer from the detailed description and examples of the invention should be construed as falling within the scope of rights according to the present invention.

The terms used in the present invention are described as general terms widely used in the technical field related to the present invention, but the meaning of the terms used in the present invention is the intention of the technician working in the field, the emergence of new technology, examination standards, or precedents. It may vary depending on etc. Some terms may be arbitrarily selected by the applicant, in which case the meaning of the arbitrarily selected terms will be explained in detail. Terms used in the present invention should be construed not only in their dictionary meaning, but in a meaning that reflects the overall context of the specification.

Terms such as 'consists of' or 'includes' used in the present invention should not necessarily be interpreted as including all of the components or steps described in the specification, and if some components or steps are not included, And if additional components or steps are further included, they should also be interpreted as intended from the corresponding terms.

Terms containing ordinal numbers such as 'first' or 'second' used in the present invention may be used to describe various components or steps, but the components or steps should not be limited by the ordinal numbers. . Terms containing ordinal numbers should be interpreted only to distinguish one component or step from other components or steps.

Hereinafter, embodiments of the present invention will be described in detail with reference to the drawings. Detailed descriptions of matters widely known to those skilled in the art related to the present invention will be omitted.

Figure 1 is a diagram to explain how charge mobility is reduced by charges accumulated in some organic layers of an OLED device.

Referring to FIG. 1, an OLED device 10 may be shown to illustrate how charge mobility is reduced by charges accumulated in some organic layers.

In the OLED device 10, holes and electrons may move between an anode formed of indium tin oxide (ITO) and a LiF/Al cathode to emit light. For example, the OLED device 10 may include at least a hole injection layer (HIL), a hole transport layer (HTL), an emission layer (EML), and an electron transport layer (ETL).

As shown, charge accumulation of holes or electrons may occur in the light emitting layer (EML) of the OLED device 10 or other organic material layers. The formation of an electric field due to such charge accumulation can adversely affect the mobility of holes or electrons, thereby reducing OLED light emission performance. Therefore, in order to improve OLED performance, it may be required to determine in which organic layer of the OLED device 10 the charge is accumulated, that is, the charge behavior inside the device.

Figure 2 is a diagram showing exemplary data for evaluating existing OLED devices.

Referring to Figure 2, voltage (V) - current density (J) graph (21), voltage - luminance graph (22), luminance - quantum efficiency graph (23), and luminance - power efficiency for four types of OLED devices. Graph 24 may be shown.

As shown in the voltage (V)-current density (J) graph (21), even if the voltage applied to OLED devices increases, there may be a limit to the increase in current density due to lower charge mobility. As shown in the voltage-luminance graph (22), the OLED luminance may increase as the applied voltage increases, but as shown in the luminance-quantum efficiency graph (23) and the luminance-power efficiency graph (24), as luminance increases, the OLED device decreases. The luminous efficiency may be lowered. Therefore, in order to improve the luminous efficiency of the OLED device, it may be required to analyze the charge behavior inside the OLED device.

Figure 3 is a diagram for explaining how the machine learning-based charge behavior analysis device inside an organic light-emitting diode (OLED) device operates according to the present invention.

Referring to FIG. 3, the machine learning-based organic light emitting diode (OLED) device internal charge behavior analysis device 200 receives measurement values obtained from the OLED device 100 to be inspected and analyzes the interior of the OLED device 100 to be inspected. Charge distribution data 300 indicating charge behavior can be output.

Unlike conventional charge behavior analysis methods, the machine learning-based organic light-emitting diode (OLED) device internal charge behavior analysis device 200 produces a HOD or EOD for each OLED organic material layer and performs the process of analyzing each device one by one. Even without going through the process, the charge behavior inside the device can be analyzed using the impedance characteristics and machine learning model of the OLED device 100 to be inspected.

Figure 4 is a diagram for explaining the elements that constitute the machine learning-based internal charge behavior analysis device of an organic light-emitting diode (OLED) device according to the present invention.

Referring to FIG. 4 , the machine learning-based organic light emitting diode (OLED) device internal charge behavior analysis device 200 may include a measurement unit 210, a processing unit 220, and an output unit 230.

The measurement unit 210 may have a structure for measuring frequency-based measurement values, including impedance, from the OLED device 100 to be inspected. For example, the measuring unit 210 may include an impedance measuring device. The processing unit 220 may have a structure for processing data such as frequency-based measurement values. For example, the processing unit 220 may include a general-purpose memory and processor. The output unit 230 may have a structure for outputting charge distribution data 300. For example, the output unit 230 may include a display, a monitor, etc.

The measurement unit 210 may be configured to obtain frequency-based measurement values by applying an alternating current test voltage to the OLED device 100 to be tested for each of the test voltage intensities and test frequencies.

For example, 200 alternating current test voltages are applied to the OLED device 100 under test for each of 10 test voltage intensities, i.e., 10 direct current voltages and 20 test frequencies, and 200 frequency-based measurement values are obtained. It can be. For example, frequency-based measurements may refer to frequency-based measurements such as impedance or capacitance.

The processing unit 220 may be configured to calculate frequency-based derived variable values by performing conversion processing on frequency-based measurement values.

For example, if the frequency-based measurements are impedance (Z) measurements, conversion processing on the impedance measurements yields frequency-based derived variable values such as admittance (1/Z) or modulus (-iωZ). It can be.

The processing unit 220 may be configured to generate frequency-based characteristic data indicating changes in frequency-based induced variable values according to changes in test voltage strengths and test frequencies.

For example, if the frequency-based derived variable values are modulus variable values and the alternating test voltage has 10 test voltage strengths and 20 test frequencies, the modulus variable values for each of the 200 voltage strength-frequency cases are Frequency-based characteristic data indicating what values it has can be generated. For example, the frequency-based characteristic data may be an image of 200 modulus variable values plotted in two dimensions.

The processing unit 220 utilizes a machine learning model that is learned to predict the charge density for each position inside the OLED device from frequency-based characteristic data to derive charge distribution data representing the charge density for each position inside the OLED device 100 to be inspected. It can be configured.

The machine learning model may be a neural network model that is learned in advance to predict the charge density for each location inside the OLED device from frequency-based characteristic data. By inputting frequency-based characteristic data into a pre-trained machine learning model, charge distribution data representing the charge density at each internal location of the OLED device 100 to be inspected can be inferred. For example, the frequency-based feature data may be 2D image data, and the machine learning model may be a convolutional neural network (CNN) model.

The output unit 230 may be configured to output charge distribution data. For example, the charge distribution data may be a graph showing the density of holes or charges according to the distance from the anode in the OLED device, and a distance-charge density graph corresponding to the charge distribution data is output through the output unit 230. It can be.

Figure 5 is a diagram for explaining the types of frequency-based measurement values measured from the OLED device to be inspected according to the present invention.

Referring to FIG. 5 , a graph 520 may be shown to explain the types of frequency-based induced variable values 510 and the qualitative meaning of the modulus measurement value determined through the equivalent circuit.

As in type 510 of frequency-based derived variable values, the frequency-based measured value may be an impedance (Z) measurement, and the frequency-based derived variable value derived from the impedance (Z) measurement may be the impedance (Z) itself, the admittance It may be any one of (Y; 1/Z), capacitance (C), modulus (M; -iωZ), and phase difference (ϕ). Meanwhile, as shown, the impedance (Z) may have the values of Z' corresponding to the real part Re(Z) and Z'' corresponding to the imaginary part Im(Z).

Regarding the modulus (M) among the types of induced variable values 510, the graph 520 shows the imaginary part (Im(Z)) of the measured modulus (M) for the equivalent circuits of five types of OLED devices at the frequency. It can be depicted according to changes. According to the measurement results, in order to increase the charge mobility, the lower the value of the equivalent resistance (R2) in the equivalent circuit, the higher the value of the imaginary part (Im(Z)) of the modulus (M) in the high frequency region on the right side of the x-axis in the graph 520. can be lowered, and as the value of the equivalent capacitance (C2) is lowered in the equivalent circuit to reduce charge accumulation, the value of the imaginary part (Im(Z)) of the modulus (M) in the high frequency region on the right side of the x-axis in the graph 520 increases. It can be confirmed that it is increasing. By utilizing the qualitative characteristics of the modulus (M) value, it may be possible to analyze the charge density at each location inside the OLED device using a machine learning model.

Meanwhile, the machine learning model used to analyze the internal charge behavior of OLED devices can be implemented in the form of a convolutional neural network (CNN). That is, the frequency-based characteristic data may be image data based on frequency-based derived variable values, and the machine learning model is a convolutional neural network that is learned to predict the charge density for each internal location of the OLED device 100 to be inspected from the image data. It could be (CNN).

Figure 6 is a diagram illustrating the process of generating data input to a machine learning model by performing preprocessing processes on frequency-based measurement values according to the present invention.

Referring to FIG. 6, processes 610, 620, 630, and 640 of generating data input to a machine learning model by performing preprocessing processes on frequency-based measurement values can be shown.

In the first process 610, 20 frequency-based measurements are generated for each of the 10 test voltage strengths (V1, ..., V10), for example, a total of 200 impedance (Z) values. can be measured. In the second process 620, plotting may be performed with the frequency and voltage intensity (direct current voltage) as the X-axis and Y-axis and the impedance (Z) value as the Z-axis value. Variables on the

In relation to the plotting of the second process 620, the types of frequency-based induced variable values include impedance (Z), admittance (Y; 1/Z), capacitance (C), modulus (M; -iωZ), and phase difference (ϕ ), and the image data may be a 2D floating image in which the magnitudes of frequency-based induced variable values are plotted against inspection voltage strengths and inspection frequencies.

In addition, in relation to the plotting of the second process 620, the magnitude of the frequency-based derived variable values (Z, Y, C, M) is the absolute value of the frequency-based derived variable values (Z, Y, C, M) (Abs), real part (Re), and imaginary part (Im), the It may be another one, and the floating value Z-axis may be the magnitude of the frequency-based derived variable values (Z, Y, C, M).

In the third process 630, a frequency-based measurement value may be converted into a frequency-based derived variable value, for example, an impedance (Z) value may be converted into a modulus (M) value. Meanwhile, the order in which the second process 620 and the third process 630 are performed may be changed. In the fourth process 640, a grayscale image can be generated by converting frequency-based derived variable values into black and white shades according to numerical values.

Regarding the grayscale image of the fourth process 640, the size of the frequency-based derived variable values (Z, Y, C, M) may be the imaginary part (Im(M)) of the modulus (M), and the 2D The floating image may be a 2D grayscale modulus plotting image that gradually expresses the size of the imaginary part (Im(M)) of the modulus (M) from white to black.

Figure 7 is a diagram for explaining examples of frequency-based measurement values and frequency-based characteristic data generated correspondingly according to the present invention.

Referring to FIG. 7 , frequency-based measurement values 711 and 712 and frequency-based characteristic data 712 and 722 generated corresponding thereto may be shown as an example.

Frequency-based measurements 711 may be 200 measurements measured from the first OLED device for 10 test voltage strengths and 20 test frequencies, and frequency-based measurements 721 may be 200 measurements taken from the first OLED device for 10 test voltage strengths and 20 test frequencies. There may be 200 measurements taken from the second OLED device for voltage strengths and 20 test frequencies.

The 200 measurements of frequency-based measurements 711 can be plotted together with the frequency-based characteristic data 712 and converted to a grayscale image, and the 200 measurements of frequency-based measurements 721 can be converted to a grayscale image. It can be plotted like the characteristic data 722 and converted to a grayscale image. The frequency-based characteristic data 712 and 722 generated in this way can be processed by a machine learning model, and charge distribution data representing the charge density for each location inside the OLED device can be derived therefrom.

Figure 8 is a diagram for explaining a machine learning model implemented in the form of a convolutional neural network (CNN) according to the present invention.

Referring to FIG. 8, the model structure (a) and prediction performance (b, c) of convolutional neural network (CNN) and fully connected neural network (FC-DNN) as types of machine learning models can be shown.

As shown in (a) of FIG. 8, unlike the structure of a fully connected neural network (FC-DNN), a convolutional neural network (CNN) may include a plurality of convolutional layers and pooling layers. Therefore, the convolutional neural network (CNN) can extract features from frequency-based characteristic data (2D floating image or 2D grayscale modulus floating image), and based on this, make inferences for learning purposes, in this case the OLED device to be inspected (100 ) can be performed to derive charge distribution data representing the charge density for each internal location.

As shown in Figures 8 (b) and (c), a convolutional neural network (CNN) can show higher performance than a conventional fully connected neural network (FC-DNN). As in (b), it can be confirmed that the correlation coefficient (Pearson correlation coefficient), which indicates the correlation between the result value predicted by the model and the actual value, is higher in the convolutional neural network (CNN), and as in (c), the correlation coefficient of the model is higher. It can be seen that the RMS error, which represents the error, is lower in convolutional neural networks (CNNs).

Figure 9 is a diagram for comparing the performance of a machine learning model implemented in the form of a convolutional neural network (CNN) according to the present invention and the performance of a conventional model.

Referring to FIG. 9, graphs 910 and 920 may be shown for comparing machine learning model performance of a convolutional neural network (CNN) and a fully connected neural network (FC-DNN).

The graphs 910 and 920 may represent actual measured values on the horizontal axis and the predicted values of the machine learning model on the vertical axis. Therefore, it can be confirmed that the more the measured value and the predicted value are the same, that is, the better the fit to the y=x graph, the more high-performance the model is. Looking at the box portions of the graphs 910 and 920, it can be seen that the CNN model of the graphs 910 fits the graph of y=x better than the DNN model of the graphs 920, It can be confirmed that the CNN model has higher prediction performance than the DNN model.

Figure 10 is a diagram showing implementation examples of OLED devices to be inspected according to the present invention and evaluation data for them.

Referring to FIG. 10, three exemplary types of OLED devices to be inspected (a) and evaluation data (b, c, d) for them can be shown.

As shown, the three types of OLED devices to be inspected (a) may be composed of a plurality of organic layers disposed between an ITO anode and a LiF/Al cathode. For example, the three types of OLED devices to be inspected (a) may include a BCFA layer, an NPB layer, an NPB:POT2T:Ir dopant layer, a POT2T layer, and a TPBi/ZADN layer, and the three types of devices (Device 1, 2, and 3), there may be some differences in composition ratio or type of TPBi/ZADN.

The evaluation data (b, c, d) is a voltage (V)-current density (J) graph, voltage-luminance graph (b), luminance-quantum efficiency graph, and A luminance-power efficiency graph (c) and the power efficiency and operating voltage (d) of three types of OLED devices to be inspected (a) can be shown. In particular, the evaluation data (b, c) may refer to the evaluation data of the general OLED device equally mentioned in FIG. 2. For the three types of OLED devices (a) subject to inspection, the charge density at each internal location can be analyzed using a machine learning model, as will be described later.

Figure 11 is a diagram for explaining charge distribution data showing the charge density at each internal location of the OLED device to be inspected according to the present invention.

Referring to FIG. 11, fingerprint images (charge distribution data; 2D floating image; 2D grayscale modulus plotting image; 1111, 1121 and 1131) may be shown, and charge distribution data 1112, 1122 and 1132 predicted through a machine learning model for the fingerprint images 1111, 1121 and 1131 may be shown.

As shown in the fingerprint images (1111, 1121, and 1131), the three types of OLED devices (Devices 1, 2, and 3) to be inspected are formed with some different structures, and the fingerprint images (1111, 1121, and 1131) are formed in different structures. 1131) can also take different forms.

The charge distribution data 1112, 1122, and 1132 can represent the internal charge behavior of the three types of OLED devices to be inspected (Device 1, 2, and 3), that is, the charge density at each location inside the device. As shown, the charge distribution data 1112, 1122, and 1132 can represent the charge density by location inside the device, that is, by distance from the anode, for holes shown in red from the left and electrons shown in blue from the right. there is. According to the charge distribution data (1112, 1122, 1132), the density of holes and electrons is high in the EML layer, which suggests that there is charge accumulation in the EML layer, which is interpreted to reflect the charge accumulation structure observed in Figure 1. It can be.

In other words, even if separate devices such as HOD/EOD for each organic material layer are not manufactured and analyzed one by one, charge distribution data (1112, 1122, 1132) can be predicted, so it can be efficiently analyzed in terms of cost and time to determine where charges are accumulated in the OLED device 100 to be inspected and have a high charge density.

Figure 12 is a diagram for explaining the steps of configuring the method of analyzing charge behavior inside a machine learning-based organic light-emitting diode (OLED) device according to the present invention.

Referring to FIG. 12, a machine learning-based method 1200 for analyzing internal charge behavior of an organic light emitting diode (OLED) device may include steps 1210 to 1250. However, it is not limited to this, and other general-purpose steps may be further included in the method 1200.

The method 1200 may consist of steps processed in time series in the machine learning-based OLED device internal charge behavior analysis device 200. Therefore, even if the content is omitted below, the content described above for the device 200 can be equally applied to the method 1200.

In step 1210, the device 200 applies an alternating current test voltage to the OLED device 100 to be tested for each of the test voltage strengths and test frequencies through the measurement unit 210 to produce frequency-based measurement values. It can be obtained.

In step 1220, the device 200, through the processing unit 220, may perform conversion processing on the frequency-based measurement values to calculate frequency-based derived variable values.

In step 1230, the device 200, through the processor 220, may generate frequency-based characteristic data indicating changes in frequency-based induced variable values according to changes in test voltage strengths and test frequencies.

In step 1240, the device 200, through the processing unit 220, uses a machine learning model that is learned to predict the charge density for each location inside the OLED device from frequency-based characteristic data to inspect the OLED device 100. Charge distribution data representing the charge density for each internal location can be derived.

In step 1250, the device 200 may output charge distribution data through the output unit 230.

Meanwhile, the machine learning-based OLED device internal charge behavior analysis method 1200 may be recorded on a computer-readable recording medium in which at least one program or software including instructions for executing the method is recorded.

Examples of computer-readable recording media include magnetic media such as hard disks, floppy disks, and magnetic tapes, optical media such as CD-ROMs and DVDs, and floptical disks. Magneto-optical media and hardware devices specifically configured to store and perform program instructions, such as ROM, RAM, flash memory, etc. may be included. Examples of program instructions may include machine language code, such as that created by a compiler, as well as high-level language code that can be executed by a computer using an interpreter, etc.

Although the embodiments of the present invention have been described in detail above, the scope of rights according to the present invention is not limited thereto, and various modifications and improvements made by those skilled in the art using the basic concept of the present invention described in the following claims can also be made according to the present invention. It should be construed as being included in the scope of rights pursuant to.

100: OLED device to be inspected
200: OLED device internal charge behavior analysis device
210: measuring unit
220: processing unit
230: output unit
300: Charge distribution data

Claims (6)

In a machine learning-based device for analyzing the internal charge behavior of an organic light-emitting diode (OLED) device,
A measuring unit configured to obtain frequency-based measurement values by applying an alternating current test voltage to the OLED device to be tested for each of the test voltage strengths and test frequencies;
Perform conversion processing on the frequency-based measured values to calculate frequency-based derived variable values,
Generating frequency-based characteristic data representing changes in the frequency-based derived variable values according to changes in the test voltage strengths and test frequencies, and
A processing unit configured to derive charge distribution data representing the charge density at each internal location of the OLED device to be inspected by utilizing a machine learning model that is learned to predict the charge density at each location inside the OLED device from the frequency-based characteristic data; and
an output unit configured to output the charge distribution data; A machine learning-based OLED device internal charge behavior analysis device including. According to paragraph 1,
The frequency-based characteristic data is image data based on the frequency-based derived variable values,
The machine learning model is a convolutional neural network (CNN) that is learned to predict the charge density for each internal location of the OLED device to be inspected from the image data. A machine learning-based OLED device internal charge behavior analysis device. According to paragraph 2,
The types of frequency-based induced variable values are one of impedance (Z), admittance (Y; 1/Z), capacitance (C), modulus (M; -iωZ), and phase difference (ϕ),
The image data is a 2D floating image in which the magnitude of the frequency-based induced variable values is plotted against the inspection voltage intensities and the inspection frequencies. A machine learning-based OLED device internal charge behavior analysis device. According to paragraph 3,
The magnitude of the frequency-based derived variable values (Z, Y, C, M) is the absolute value (Abs), real part (Re), and imaginary part of the frequency-based derived variable values (Z, Y, C, M). (Im) is any one of
The X-axis of the 2D floating image is one of the inspection voltage intensity and the inspection frequency, the Y-axis is the other one of the inspection voltage intensity and the inspection frequency, and the floating value Z-axis is the frequency-based induced variable values (Z, Y, C, M ), a machine learning-based device for analyzing the internal charge behavior of OLED devices. According to paragraph 4,
The magnitude of the frequency-based derived variable values (Z, Y, C, M) is the imaginary part (Im(M)) of the modulus (M),
The 2D floating image is a 2D grayscale modulus plotting image that gradually expresses the size of the imaginary part (Im(M)) of the modulus (M) from white to black. A machine learning-based OLED device internal charge behavior analysis device. In a machine learning-based method for analyzing the internal charge behavior of an organic light-emitting diode (OLED) device,
Obtaining frequency-based measurement values by applying an alternating current test voltage to the OLED device to be tested for each of the test voltage intensities and test frequencies through a measurement unit;
Calculating frequency-based derived variable values by performing conversion processing on the frequency-based measured values through a processing unit;
generating, through the processing unit, frequency-based characteristic data representing changes in the frequency-based derived variable values according to changes in the test voltage intensities and the test frequencies;
Deriving charge distribution data representing the charge density at each internal location of the OLED device to be inspected using a machine learning model that is learned to predict the charge density at each location inside the OLED device from the frequency-based characteristic data through the processing unit. ; and
outputting the charge distribution data through an output unit; A method of analyzing charge behavior inside an OLED device based on machine learning, including.