JP2022134336A - Organic el panel manufacturing device - Google Patents

Abstract

To provide an organic EL panel manufacturing device capable of improving the yield of organic EL panels compared to the prior art.SOLUTION: A neural network receives a plurality of manufacturing parameters from an input layer, passes through an upstream intermediate layer, a midstream intermediate layer, and a downstream intermediate layer, and outputs predicted values of maximum peak intensity and wavelength from an output layer. Therein, the midstream intermediate layer comprises: a first node group having four nodes respectively associated with a factor related to variation in in-plane thickness of organic EL panels, a factor related to variation in thickness of light-emitting layers depending on a manufacturing unit, a factor related to variation in path light length depending on positions of the light-emitting layer in the organic EL panel, and a factor of time variation of a ratio of RGB; and a second node group in which the respective nodes are multiplied by respective corresponding constant weighting factors. An organic EL panel manufacturing device is configured to: adjust manufacturing parameters using a machine learning model so that the predicted values of the maximum peak intensity and wavelength are within a specified range; and manufacture the organic EL panels using the adjusted manufacturing parameters.SELECTED DRAWING: Figure 3

Description

The present invention relates to an organic EL panel manufacturing apparatus.

2. Description of the Related Art In recent years, attention has been focused on organic EL panels as lighting devices, and many studies have been conducted (for example, Patent Document 1).
In the organic EL panel of Patent Document 1, an organic EL element is laminated on a transparent insulating substrate, and the organic EL element is sealed with a transparent insulating substrate and an inorganic sealing layer. An organic EL element is formed by stacking a transparent electrode layer, an organic functional layer, and a back electrode layer.
Further, in the organic EL panel described in Patent Document 1, the organic functional layers are, in order from the transparent electrode layer side, a hole injection layer, a hole transport layer, a blue fluorescent emitting layer, an electron transport layer, an electron injection layer, a connection layer, A hole-injecting layer, a hole-transporting layer, a red-green phosphorescent emitting layer, and an electron-transporting layer are laminated.

JP 2018-174162 A

By the way, in commercializing an organic EL device, the product characteristics need to satisfy product standards.
However, many of the product characteristics are in a trade-off relationship, and if the formulation of an organic EL device that satisfies one product standard is used as a reference, and the layer structure, etc., is changed in order to satisfy the product standard for other product characteristics as well, Product characteristics that met the original product specifications may also change and both may deviate from the product specifications. That is, in manufacturing an organic EL device, each manufacturing parameter interacts with each other and affects product characteristics, so it has been difficult to meet product specifications for multiple product characteristics at the same time.

Therefore, the inventors of the present invention considered to generate a correlation between manufacturing parameters and product characteristics by calculation using a neural network.
However, when all the manufacturing parameters used in manufacturing the organic EL device are input to the input layer using a neural network and a large number of product characteristics are output, the number of manufacturing parameters is too large and the calculation does not converge, resulting in a solution. I had a problem that I couldn't figure out.

SUMMARY OF THE INVENTION Accordingly, an object of the present invention is to provide an organic EL panel manufacturing apparatus capable of improving the yield of organic EL panels as compared with the conventional one.

When the inventor of the present invention compared the product characteristics of past organic EL panels with the product standards, it was found that in the product standards, not only the occurrence of dark spots, but also the chromaticity performance in the durability test was not satisfied, and it was determined as a defective product. It turned out that there were many.
In general, chromaticity performance is often determined by product standards based on color temperature and color coordinates. Particularly, when color temperature falls within a predetermined range, color coordinates also tend to fall within a certain range. That is, in order to apply chromaticity performance to product standards, it is important to keep the color temperature within a predetermined range.
This color temperature can be obtained by calculation by measuring the light intensity at each wavelength, and has a certain correlation.
Therefore, the present inventor compared and studied the relationship between the light intensity and the color temperature at each wavelength, and discovered that the light intensity in a specific wavelength region tends to have a substantially constant correlation with the color temperature. That is, the change in waveform in the wavelength region of red light emission has a substantially constant correlation with the light intensity.
It is thought that the peak in the wavelength region of this red emission is greatly affected by the position of the light-emitting interface where holes and electrons combine. , variation in path light length depending on the position of the light-emitting layer on the organic EL panel, and temporal variation in the ratio of RGB.
That is, variation in the in-plane thickness of the organic EL panel in the intermediate layer in the machine learning model, variation in the thickness of the light-emitting layer during manufacturing, variation in the light path length depending on the position of the light-emitting layer in the organic EL panel, and RGB By passing a factor corresponding to each time variation of the ratio of , the calculation can be converged even when a large number of manufacturing parameters are input to the input layer, and the intensity of the maximum peak in the wavelength region of red emission and the maximum It was thought that the peak wavelength could be output from the output layer.

One aspect of the present invention derived based on the above idea is a manufacturing department that manufactures an organic EL panel including a light-emitting layer based on a plurality of manufacturing parameters, and a machine using a machine learning model having a neural network. A machine learning unit that performs learning, and a peak acquisition unit that acquires the maximum peak intensity and wavelength in the wavelength range of 500 nm to 630 nm, the neural network has an input layer, an intermediate layer, and an output layer, The intermediate layer includes an upstream intermediate layer, a midstream intermediate layer, and a downstream intermediate layer, and the neural network receives the plurality of manufacturing parameters from the input layer and receives the upstream intermediate layer and the midstream intermediate layer. Predicted values of the maximum peak intensity and wavelength are output from the output layer via the intermediate layer and the downstream intermediate layer, and the intermediate intermediate layer includes a first node group and a second node group. , wherein the first node group has a plurality of nodes including a first node, a second node, a third node, and a fourth node, and the first node is the organic EL The second node is a factor related to the variation in thickness of the light-emitting layer formed by the manufacturing department, and the third node is a factor related to the variation in the thickness of the light-emitting layer formed by the manufacturing department. The fourth node is a factor related to the variation of the optical path length due to the position on the panel, the fourth node is a factor of the time variation of the ratio of RGB, and the second group of nodes is a weighting factor corresponding to the plurality of nodes. and each weighting factor is a constant value assigned to the plurality of nodes, and the machine learning model determines the predicted values of the maximum peak intensity and wavelength to fall within a predetermined range. The organic EL panel manufacturing apparatus adjusts a plurality of manufacturing parameters and manufactures the organic EL panel using the plurality of adjusted manufacturing parameters.

According to this aspect, the machine learning model is configured with a neural network containing four fixed nodes in the middle layer on the middle stream side, and the solution is narrowed down to the maximum peak intensity and wavelength. Machine learning can be performed more efficiently than before, and the solution is less likely to diverge even if a huge amount of manufacturing parameters are used.
According to this aspect, the fixed node is a factor related to the variation in the in-plane thickness of the organic EL panel, a factor related to the variation in the thickness of the light-emitting layer formed by the manufacturing department, and a path light depending on the position of the light-emitting layer in the organic EL panel. Since these factors are related to the variation in length and the factor of time variation in the ratio of RGB, it is possible to calculate the predicted values of the maximum peak intensity and wavelength based on the position of the light emitting interface. As a result, the range of color temperatures can be roughly predicted from the predicted values of the intensity and wavelength of the maximum peak.
According to this aspect, since the organic EL panel is manufactured using the manufacturing parameters predicted by the machine learning model, the maximum peak intensity and maximum peak wavelength are kept within a predetermined range, and the color temperature is kept within a predetermined range. and the chromaticity performance can be kept within product specifications. As a result, the number of defective products can be reduced and the yield can be improved.

By the way, the variation in the in-plane thickness of the organic EL panel, the variation in the thickness of the light emitting layer formed by the manufacturing department, the variation in the light path length due to the position of the light emitting layer on the organic EL panel, and the time of the RGB ratio Variations can be ranked relative to their impact on the relationship between maximum peak intensity and maximum peak wavelength according to the inventor's rule of thumb.

Therefore, in a preferred aspect, each weighting factor assigned to the first node, the second node, the third node and the fourth node is equal to the first node, the second node, the third node and the fourth node. It is that each is set based on the correlation of the four nodes.

According to this aspect, since each weighting factor is set based on the fixed correlation of the four nodes, machine learning can be performed more accurately and efficiently than before.

A preferred aspect is that the first node group consists of the first node, the second node, the third node and the fourth node.

According to this aspect, only the fixed four nodes are passed during the calculation in the intermediate layer, so the calculation can be converged more.

A preferred aspect is that the fourth node is a factor of color temperature time variation.

According to this aspect, the calculation can be more converged.

In a preferred aspect, the organic EL panel has an organic functional layer including the light-emitting layer, a back electrode layer, and a sealing layer laminated on a transparent conductive substrate, and the production department A light-emitting functional layer forming step for forming a functional layer, a back electrode layer forming step for forming the back electrode layer, and a sealing layer forming step for forming the sealing layer are performed in this order. forming a work-in-process substrate, inspecting quality parameters of the work-in-process substrate, and further inputting the quality parameters of the work-in-process substrate into the input layer.

According to this aspect, since the quality parameter is input to the input layer in addition to the manufacturing parameter, the intensity and wavelength of the maximum peak can be predicted with higher accuracy.

According to the organic EL panel manufacturing apparatus of the present invention, the yield of organic EL panels can be improved compared to the conventional one.

1 is a configuration diagram of an apparatus for manufacturing a divided organic EL panel according to a first embodiment of the present invention; FIG. FIG. 2 is a configuration diagram schematically showing the first film forming apparatus of FIG. 1; FIG. 2 is a configuration diagram schematically showing a machine learning model in the control unit of FIG. 1; It is a top view which shows the cutting position of a large-sized organic EL panel, and a cutting position is shown by the dashed-dotted line. FIG. 2 is a cross-sectional view showing a typical layer structure of a split organic EL panel, and hatching is omitted for easy understanding.

BEST MODE FOR CARRYING OUT THE INVENTION Hereinafter, embodiments of the present invention will be described in detail.

The manufacturing apparatus 1 of the first embodiment of the present invention is used in the process of commercializing an organic EL panel, manufactures a large-sized organic EL panel 200, and divides the organic EL panel 200 from the large-sized organic EL panel 200 into divided organic EL panels having a certain chromaticity performance. It forms the panel 207 .
As shown in FIG. 1, the manufacturing apparatus 1 is mainly composed of a manufacturing section 2, a control section 3, and a panel inspection section 4, which are connected wirelessly or by wire.
Manufacturing department 2, control department 3, and panel inspection department 4 may be connected via networks, such as the Internet and an intranet.

The manufacturing department 2 manufactures the large-sized organic EL panel 200 .
As shown in FIG. 1, the manufacturing department 2 includes a pretreatment device 20, a first film forming device 21, a second film forming device 22, an electrode film forming device 23, a first sealing device 24, a second A sealing device 25 , a sheet inspection section 26 and a film sticking device 27 are provided.
The pretreatment device 20 is a cleaning unit that cleans the transparent conductive substrate 201 prior to forming each layer.
The pretreatment device 20 includes a supporting section, a plasma generator, and a gas discharging section in a vacuum chamber. While discharging the cleaning gas from the plasma generator, it is possible to generate plasma with the plasma generator and perform plasma processing on the transparent conductive substrate 201 .

As shown in FIG. 2, the first film-forming apparatus 21 applies each layer of the blue-green light-emitting unit 215 and the red-green light-emitting unit 217 of the large-sized organic EL panel 200 ( 5), specifically a vacuum deposition apparatus.
The first film forming apparatus 21 includes, as main constituent members, a film forming chamber 50, a gas discharge section 51, a first evaporation group 52, a second evaporation group 53, a first piping section 54, and a second piping section. 55 , a mixing pipe section 56 and a substrate holding section 57 .

The first film forming apparatus 21 causes the gas discharge part 51 to face the film forming surface of the in-process base material 208 held by the base material holding part 57, and in this state, the in-process base material 208 is produced from the gas discharge part 51. A desired layer can be deposited on the film surface.

The film forming chamber 50 is a portion for forming a film on the in-process substrate 208 by a vacuum deposition method, and is connected to a vacuum pump (not shown) so that the internal space can be evacuated. In addition, the film forming chamber 50 is also connected to an exhaust pipe (not shown) so that residual material gas in the film forming chamber 50 can be exhausted.

The gas discharge part 51 is a shower plate and has a plurality of material discharge openings for discharging material gas.

As shown in FIG. 2, the first evaporator group 52 includes one or more first evaporators 60a to 60d and a first connecting pipe 61 that connects each of the first evaporators 60a to 60d and the first pipe 54. ing.
In the first evaporating units 60a to 60d, a crucible (accommodating unit) for accommodating a film forming material composed of an organic material or an inorganic material and a heating means for heating the crucible are arranged in the evaporating chamber. By gradually heating the deposited film-forming material by a heating means, the film-forming material can be vaporized to generate a material gas.
The first connection pipe 61 is provided with a temperature control device (not shown) so that the material gas can flow to the downstream first pipe portion 54 at a desired temperature.

As shown in FIG. 2, the second evaporator group 53 includes one or more second evaporators 65a to 65d and a second connection pipe 66 that connects each of the second evaporators 65a to 65d and the second pipe 55. ing.
In the second evaporating units 65a to 65d, crucibles (accommodating units), heating means for heating the crucibles, and supply units for supplying the crucibles with a minute amount of film-forming material composed of an organic material or an inorganic material are arranged in the evaporating chambers. It is The second evaporators 65a to 65d instantaneously heat the film-forming material supplied from the supply unit into the crucible by the heating means to vaporize the film-forming material and generate a material gas. ing.
The heating means is preferably a heater or a heating gas obtained by heating an inert gas.
The second connection pipe 66 is provided with a temperature control device (not shown) so that the material gas can flow to the second pipe portion 55 on the downstream side at a desired temperature.

The first pipe section 54 is a pipe that connects each of the evaporation sections 60 a to 60 d of the first evaporation group 52 and the mixing pipe section 56 . The first pipe portion 54 is provided with a temperature control device (not shown) so that the material gas can flow to the mixing pipe portion 56 on the downstream side at a desired temperature.
The second pipe section 55 is a pipe that connects each of the evaporation sections 65 a to 65 d of the second evaporation group 53 and the mixing pipe section 56 . The second pipe portion 55 is provided with a temperature control device (not shown) so that the material gas can flow to the mixing pipe portion 56 on the downstream side at a desired temperature.
The base material holding part 57 is a part that holds the in-process base material 208 , and is provided with a temperature control device (not shown) for controlling the temperature of the in-process base material 208 .

The second film-forming device 22 is a film-forming device for forming the light-emitting intermediate layer 216 of the large-sized organic EL panel 200 on the in-process substrate 208, and is specifically a vacuum deposition device.
As shown in FIG. 2, the second film-forming apparatus 22 has the same configuration as the first film-forming apparatus 21, so the same components are denoted by the same reference numerals and the description thereof is omitted.
That is, the second film forming apparatus 22 includes a film forming chamber 50, a gas discharge section 51, evaporation groups 52 and 53, piping sections 54 to 56, and a substrate holding section 57 as main constituent members. .
Then, the second film-forming apparatus 22 causes the gas discharge part 51 to face the film-forming surface of the in-process substrate 208 held by the substrate holding part 57 , and in this state, the gas discharge part 51 discharges the in-process substrate 208 . It is possible to vapor-deposit on the film-forming surface of

The electrode film-forming apparatus 23 is a film-forming apparatus for forming the back electrode layer 203 of the large-sized organic EL panel 200 on the in-process substrate 208, and is specifically a vacuum deposition apparatus.
As shown in FIG. 2, the electrode film-forming apparatus 23 has the same configuration as the first film-forming apparatus 21, so the same components are denoted by the same reference numerals and the description thereof is omitted.
The electrode film forming apparatus 23 includes a film forming chamber 50, a gas discharge section 51, evaporation groups 52 and 53, pipe sections 54 to 56, and a substrate holding section 57 as main constituent members.
Then, the electrode film-forming apparatus 23 causes the gas discharge part 51 to face the film-forming surface of the in-process substrate 208 held by the substrate holding part 57 , and in this state, the gas discharge part 51 releases the in-process substrate 208 . Vapor deposition on the film forming surface is possible.

The evaporators 65a to 65d of the second evaporator group 53 of the electrode film forming apparatus 23 include ion beam evaporators that evaporate materials by the ion beam method.
This ion beam evaporator is equipped with a hearth for storing solid materials and a beam irradiation unit in a vacuum chamber. , it is possible to generate material gas.

The first sealing device 24 is a film forming device that forms the dry sealing layer 250 on the in-process substrate 208, and is a device that forms the dry sealing layer 250 by a dry method.
Specifically, the first sealing device 24 is a plasma CVD device, and includes a gas discharge part, a substrate holding part, a plasma generator, and a temperature control device in a film deposition chamber. A dry sealing layer 250 is formed on the film forming surface of the in-process substrate 208 held by the holding unit by discharging the material gas from the gas discharging unit while generating plasma with the plasma generator toward the film forming surface. It is possible.

The second sealing device 25 is a film forming device that forms the wet sealing layer 251 on the in-process substrate 208, and is a device that forms the wet sealing layer 251 by a wet method.
The second sealing device 25 includes a precursor coating device and a baking device. The precursor coating device applies a liquid containing a precursor of the wet sealing layer 251 to the film forming surface, and the baking device bakes the liquid. Thus, the wet sealing layer 251 can be formed.

The sheet inspection unit 26 detects the current, voltage, luminance, color coordinates (CIE-x, y), color coordinates (u', v'), deviation from the blackbody orbit (Δuv), color temperature ( K), color rendering properties (Ra, R1 to R15), radiance (Le), dominant wavelength (λd), output efficiency (Pe), light intensity per wavelength (380 nm to 780 nm), presence or absence of current leakage, generation of dark spots It is a device for inspecting quality parameters such as

The film sticking device 27 is a device that sticks the protective film 206 (see FIG. 5) onto the sealing layer 205 of the substrate 208 in process.

The control unit 3 includes a machine learning unit 30, a data accumulation unit 31, an estimation unit 32, an output control unit 33, and an examination information acquisition unit 34, as shown in FIG.
The machine learning unit 30 is a deep learning unit that performs machine learning using a machine learning model having a neural network 100 incorporating a neuron model as shown in FIG. 3 as a value function approximation algorithm. It is possible to perform supervised learning.
The neural network 100 for constructing the machine learning model of this embodiment includes an input layer 101, an intermediate layer 102, and an output layer 103, as shown in FIG. It is a deep neural network.
The neural network 100 receives each manufacturing parameter and each quality parameter from the input layer 101, passes through the intermediate layer 102, and outputs the red intensity peak p of the divided organic EL panel 207 and the wavelength λ of the intensity peak from the output layer 103. It is what is done.

The intermediate layer 102 is composed of an upstream intermediate layer 110, a midstream intermediate layer 111, and a downstream intermediate layer 112, as shown in FIG.
The upstream intermediate layer 110 is a hidden layer composed of one or more layers. Each manufacturing parameter input from the input layer 101 is multiplied by a weight w and input to each neuron (node). The upstream intermediate layer 110 of this embodiment is composed of two layers of node groups 113 and 114 .

The midstream intermediate layer 111 is composed of a midstream output layer 115 (first node group) and a midstream input layer 116 (second node group), as shown in FIG.
The midstream output layer 115 is a node group composed of nodes D1, V1, L1, and T1 output by the upstream intermediate layer 110 .
The first node D<b>1 is a factor related to the in-plane thickness variation of the large-sized organic EL panel 200 provided as a solution for the upstream intermediate layer 110 .
The second node V<b>1 is a factor related to variations in thickness of each layer forming the organic functional layer 202 by the manufacturing department 2 provided as a solution for the upstream intermediate layer 110 .
The third node L<b>1 is a factor related to the variation of the optical path length depending on the position of the organic functional layer 202 provided as a solution for the upstream intermediate layer 110 on the large-sized organic EL panel 200 .
The fourth node T1 is a factor of time-varying RGB ratio, specifically a factor of time-varying color temperature.
The midstream input layer 116 receives input from the midstream output layer 115 by multiplying weighting factors W1 to W4.
The midstream input layer 116 is a node group composed of nodes D 2 , V 2 , L 2 , and T 2 input by the midstream output layer 115 .

The downstream intermediate layer 112 is a hidden layer composed of one or more layers. Each manufacturing parameter and each quality parameter input from the intermediate input layer 116 is multiplied by a weight w, and each neuron (node ).
The downstream intermediate layer 112 of this embodiment is composed of node groups 117 and 118 in two layers.

The output layer 103 is composed of a node P that outputs the maximum peak intensity in the wavelength range of 500 nm to 630 nm to which the solution of the downstream intermediate layer 112 is input, and a node λ that outputs the peak wavelength of the maximum peak. It is a node group.

The data accumulation unit 31 stores the learning model learned by the machine learning unit 30, the set values and actual measurement values of each manufacturing parameter in the past, the inspection results of the large-format organic EL panel 200 manufactured in the past by the sheet inspection unit 26, and the aging inspection. This is a part for accumulating data such as inspection results by the unit 41 and inspection results by the durability inspection unit 43 (for example, color temperature, light intensity by wavelength, etc.).

The estimating unit 32 uses a learning model learned by the machine learning unit 30 from the manufacturing parameters and quality parameters accumulated in the data accumulation unit 31 to determine the peak intensity of the maximum peak and the wavelength of the maximum peak in the wavelength range of 500 nm to 630 nm. This is a part for estimating manufacturing parameters that fall within a range.

The output control unit 33 is a unit that can input manufacturing parameters to the manufacturing unit 2 and the machine learning unit 30 , and is a unit that can input manufacturing parameters estimated by the estimating unit 32 to the manufacturing unit 2 .

The inspection information acquisition unit 34 is a part that acquires the inspection results of the sheet inspection unit 26 of the manufacturing unit 2 and the inspection results of the panel inspection unit 4 .

The panel inspection unit 4 processes the large-sized organic EL panel 200 to form the divided organic EL panels 207 and inspects each divided organic EL panel 207 .
As shown in FIG. 1, the panel inspection section 4 is composed of a processing device 40, an aging inspection section 41, a performance inspection section 42, and a durability inspection section 43. As shown in FIG.
The processing device 40 is a part that divides the large-sized organic EL panel 200 into a plurality of divided organic EL panels 207 as shown in FIG.
The aging inspection section 41 is a section that inspects the warp of the divided organic EL panel 207 with a laser beam, and leaves it in an energized state at a predetermined temperature for a predetermined period of time after the inspection.

The performance inspection unit 42 inspects the performance of the divided organic EL panel 207, and includes current, voltage, luminance, luminous efficiency (lm/W), color coordinates (CIE-x, y), color coordinates (u′, v′), deviation from the blackbody orbit (Δuv), color temperature (K), color rendering properties (Ra, R1 to R15), radiance (Le), and color rank can be inspected.

The durability inspection section 43 inspects the durability of the divided organic EL panel 207. It is driven for a predetermined time under high temperature and high humidity conditions, and the current, voltage, luminance, luminous efficiency (lm/W), Color coordinates (CIE-x, y), color coordinates (u', v'), deviation from the blackbody orbit (Δuv), color temperature (K), color rendering properties (Ra, R1 to R15), radiance (Le ), which checks for changes in color rank.
Further, the durability inspection unit 43 is a peak acquisition unit capable of calculating the maximum peak intensity (P) and the peak wavelength (λ) of the maximum peak in the wavelength range of 500 nm to 630 nm as measured values.

Next, the divided organic EL panel 207 that is suitably manufactured by the manufacturing apparatus 1 of this embodiment will be described.

The divided organic EL panel 207 is obtained by dividing the large-sized organic EL panel 200 into a plurality of pieces by the processing device 40 .
The divided organic EL panel 207 is an organic EL panel in which an organic functional layer 202, a back electrode layer 203, a sealing layer 205, and a protective film 206 are laminated on a transparent conductive substrate 201, as shown in FIG.
The transparent conductive base material 201 is a plate-like substrate having a planar spread, and is capable of transmitting light in the thickness direction. The transparent conductive substrate 201 of this embodiment is a substrate in which a transparent electrode layer 211 is laminated on a transparent insulating substrate 210 .
In the divided organic EL panel 207, the overlapping portion of the transparent electrode layer 211, the organic functional layer 202, and the back electrode layer 203 constitutes an organic EL element. By applying a voltage, the intervening organic functional layer 202 can emit light.

The transparent insulating substrate 210 is not particularly limited as long as it has transparency and insulating properties. For example, a transparent substrate such as a glass substrate, a transparent resin film, or the like can be used.
The transparent electrode layer 211 is not particularly limited as long as it has transparency and conductivity. For example, indium tin oxide (ITO), indium zinc oxide (IZO), tin oxide (SnO ₂ ). , zinc oxide (ZnO) can be used.

The organic functional layer 202 is a light-emitting functional layer composed of a blue-green light-emitting unit 215, a light-emitting intermediate layer 216, and a red-green light-emitting unit 217 in order from the transparent conductive substrate 201 side, as shown in FIG.
The blue-green light-emitting unit 215 is a light-emitting unit that emits blue-green light when lit, and includes a hole injection layer 220, a hole transport layer 221, a blue light-emitting layer 222, and an electron transport layer 223 in this order from the transparent conductive substrate 201 side. consists of
The light-emitting intermediate layer 216 is composed of an electron injection layer 230 and a connection layer 231 in order from the blue-green light-emitting unit 215 side.
The red-green light-emitting unit 217 is a light-emitting unit that lights red-green when lit. , an electron transport layer 239 and an electron injection layer 240 .

The back electrode layer 203 is an electrode layer paired with the transparent electrode layer 211, and is a reflective electrode layer that reflects light emitted from the organic functional layer 202 toward the transparent electrode layer 211 side.
As the back electrode layer 203, for example, a metal layer such as silver or aluminum can be used.

The sealing layer 205 is a layer that seals the organic EL element together with the transparent insulating substrate 210 .
The sealing layer 205 of this embodiment is composed of a dry sealing layer 250 and a wet sealing layer 251 .
The dry sealing layer 250 is a sealing layer formed by a dry method such as a CVD method or a PVD method using a CVD device, a PVD device, or the like. For example, an inorganic sealing layer such as silicon oxide or silicon nitride is used. can.
The wet sealing layer 251 is a sealing layer formed by a wet method such as a spray method or a casting method by applying a liquid and baking it. Inorganic sealing layers such as silicon oxide formed can be used.

The protective film 206 protects the organic EL element from external force.
The protective film 206 is an elastic film having elasticity, and is capable of relieving the external force when it is applied.
The protective film 206 is an adhesive film having an adhesive material on its surface, and is a waterproof film having waterproof properties.

Next, the manufacturing process of the organic EL panel using the manufacturing apparatus 1 will be described.

The manufacturing process of the organic EL panel includes a manufacturing process in which the large-sized organic EL panel 200 is manufactured by the manufacturing department 2, and a process in which the large-sized organic EL panel 200 is processed and divided into divided organic EL panels 207, and the divided organic EL panels 207 are manufactured. It includes a quality confirmation process to confirm quality.

In the manufacturing process, first, the transparent conductive substrate 201 is introduced into the pretreatment device 20, and subjected to plasma treatment to clean the transparent conductive substrate 201 (pretreatment step).

At this time, manufacturing parameters such as the plasma processing time used in the pretreatment process, the high-frequency power of the plasma generator, and the position on the large-sized organic EL panel 200 after manufacturing are stored in the data storage unit 31 .

After the pretreatment process is finished, a light-emitting functional layer forming process for forming the organic functional layer 202 is performed.
The light-emitting functional layer forming process includes a first light-emitting unit forming process, an intermediate layer forming process, and a second light-emitting unit forming process.
Specifically, first, the transparent conductive base material 201 is introduced into the first film forming apparatus 21, and the blue liquid is deposited on the transparent electrode layer 211 of the transparent conductive base material 201 by the vacuum deposition method in the first film forming apparatus 21. A green light emitting unit 215 is formed (first light emitting unit forming step).

At this time, the film forming time and film forming rate of each layer 220 to 223 of the blue-green light emitting unit 215 used in the first light emitting unit film forming process, the pressure and temperature of the film forming chamber 50, the temperature of the gas discharge section 51, the exhaust Piping temperature, material gas flow rate and heating temperature in each piping section 54 to 56, material supply amount, material remaining amount, and heating temperature in each evaporating section 60a to 60d of the first evaporating group 52, second evaporating group Material supply amount, remaining amount of material, and heating temperature of each evaporating unit 65a to 65d of 53, film forming standby from completion of pretreatment in pretreatment device 20 to start of film formation in first film forming device 21 Manufacturing parameters such as the time and film deposition position corresponding to the large-sized organic EL panel 200 after manufacturing are stored in the data storage unit 31 .

When the first light-emitting unit film-forming process is completed, the transparent conductive substrate 201 on which the blue-green light-emitting unit 215 is formed (hereinafter referred to as the transparent conductive substrate on which each layer is formed until the large-sized organic EL panel 200 is completed) 201 are collectively referred to as an in-process base material 208) is introduced into the second film-forming apparatus 22, and the second film-forming apparatus 22 deposits a light-emitting intermediate layer 216 on the blue-green light-emitting unit 215 of the in-process base material 208 by vacuum deposition. is formed (intermediate layer forming step).

At this time, the film forming time of each layer 230, 231 of the light emitting intermediate layer 216 used in the intermediate layer forming process, the pressure and temperature of the film forming chamber 50 of the second film forming apparatus 22, the flow rate of the material gas, each host material Film formation rate, film thickness, and heating time, film formation standby time from the end of film formation in the first film formation apparatus 21 to the start of film formation in the second film formation apparatus 22, large format after production Manufacturing parameters such as positions on the organic EL panel 200 are accumulated in the data accumulation unit 31 .

When the intermediate layer forming process is completed, the in-process substrate 208 on which the light-emitting intermediate layer 216 has been formed is again introduced into the first film forming apparatus 21, and the in-process substrate 208 is deposited in the first film forming apparatus 21 by vacuum deposition. A red-green light-emitting unit 217 is formed on the light-emitting intermediate layer 216 (second light-emitting unit forming step). That is, in the present embodiment, the in-process substrate 208 formed into a film by the first film forming apparatus 21 is formed into a film by the second film forming apparatus 22 and then formed into a film by the first film forming apparatus 21 again.

At this time, the film forming time and film forming rate of each layer 235 to 240 of the red and green light emitting unit 217 used in the second light emitting unit film forming process, the pressure and temperature of the film forming chamber 50, the temperature of the gas discharge section 51, the exhaust Pipe temperature, material gas flow rate and heating temperature in each pipe part 54 to 56, material supply amount, remaining material amount, and heating temperature in each evaporation part 60a to 60d of the first evaporation group 52, second evaporation group 53 Material supply amount, remaining amount of material, heating temperature of each evaporating unit 65a to 65d, film forming from the end of film forming in the second film forming apparatus 22 to the start of film forming in the first film forming apparatus 21 Manufacturing parameters such as the waiting time and the position on the large-sized organic EL panel 200 after manufacturing are accumulated in the data accumulation unit 31 .

When the second light-emitting unit film-forming process is completed and the light-emitting functional layer film-forming process is completed, the process shifts to the back surface electrode layer-forming process, and the in-process substrate 208 on which the red and green light-emitting units 217 are formed is placed in an electrode film forming apparatus. 23, and the back electrode layer 203 is formed on the red-green light emitting unit 217 of the in-process base material 208 by a vacuum vapor deposition method (back electrode layer forming step).

At this time, the film forming time, the film forming rate, the film thickness, the supply amount and remaining amount of the material, and the completion of the film forming in the first film forming apparatus 21 of the back electrode layer 203 used in the back electrode layer forming process are described. Manufacturing parameters such as the film deposition standby time from the start of film deposition to the start of film deposition in the electrode film deposition device 23 and the position on the large-sized organic EL panel 200 after manufacturing are stored in the data storage unit 31 .

After the back electrode layer forming process is completed, a sealing layer forming process for forming the sealing layer 205 is performed.
The sealing layer forming process includes a first sealing layer forming process and a second sealing layer forming process.
Specifically, first, the in-process substrate 208 on which the back electrode layer 203 is formed is introduced into the first sealing device 24, and the dry sealing layer 250 is formed on the back electrode layer 203 of the in-process substrate 208 by the plasma CVD method. is formed (first sealing layer forming step).

At this time, the film forming time of the dry sealing layer 250 used in the first sealing layer forming process, the pressure and temperature in the film forming chamber 50, the gas flow rate of each material constituting the dry sealing layer 250, the electrode Manufacturing parameters such as the film deposition standby time from the end of film deposition in the film deposition device 23 to the start of film deposition in the first sealing device 24 and the position on the large-sized organic EL panel 200 after manufacturing are stored in the data. It is accumulated in the accumulation unit 31 .

After the first sealing layer forming process is completed, the in-process base material 208 on which the dry sealing layer 250 is formed is introduced into the second sealing device 25, and the solution containing the precursor of the wet sealing layer 251 is applied. It is coated on the dry sealing layer 250 of the base material 208, and the in-process base material 208 is introduced into a baking device and baked to form a wet sealing layer 251 on the dry sealing layer 250 (second sealing layer film-forming process).

At this time, the baking time of the wet sealing layer 251 used in the second sealing layer forming process, the content of the precursor of the wet sealing layer 251, and the Film formation standby time until film formation in the second sealing device 25 is started, baking standby time from the end of film formation in the second sealing device 25 to the start of baking in the baking device, manufacturing Manufacturing parameters such as positions in the large-sized organic EL panel 200 are accumulated in the data accumulation unit 31 .

When the second sealing layer forming process is completed and the sealing layer forming process is completed, the sheet inspection unit 26 checks the in-process base material 208 on which the sealing layer 205 is formed, such as leakage current. 208 quality characteristics (quality parameters) are inspected (sheet inspection process, manufacturing side inspection process).

At this time, the current, voltage, luminance, color coordinates (CIE-x, y), color coordinates (u′, v′), deviation from the black body orbit (Δuv ), color temperature (K), color rendering properties (Ra, R1 to R15), radiance (Le), dominant wavelength (λd), output efficiency (Pe), light intensity per wavelength (380 nm to 780 nm), presence or absence of current leakage , occurrence of dark spots, etc. are stored in the data storage unit 31 .

When the sheet inspection process is completed, the protective film 206 is adhered to the in-process substrate 208 by the film applying device 27 (protective film applying process), and the large-sized organic EL panel 200 is completed.

At this time, manufacturing parameters such as the thickness of the protective film 206 and the hardness of the protective film 206 used in the protective film attaching process are stored in the data storage unit 31 .

Subsequently, in the quality confirmation process, first, the processing device 40 divides the large-sized organic EL panel 200 into a plurality of divided organic EL panels 207 (dividing process).

At this time, as shown in FIG. 4, the divided organic EL panels 207 are equally divided so that each divided organic EL panel 207 has approximately the same size.
At this time, quality parameters such as the position and size of each divided organic EL panel 207 used in the large-sized organic EL panel 200 are accumulated in the data accumulation unit 31 .

When the division process is completed, warp of the division organic EL panel 207 is inspected by a laser beam, and after the inspection, it is left in a state of being energized at a prescribed temperature for a prescribed period of time (aging process).

At this time, the quality parameters such as the amount of warpage of the split organic EL panel 207 inspected in the aging process are accumulated in the data accumulation unit 31 .

After the aging process is finished, the performance inspection unit 42 evaluates the performance of the divided organic EL panel 207 (performance evaluation process).

At this time, current, voltage, luminance, luminous efficiency (lm/W), color coordinates (CIE-x, y), color coordinates (u', v'), black Quality parameters such as deviation from body trajectory (Δuv), color temperature (K), color rendering properties (Ra, R1 to R15), radiance (Le), and color rank are stored in data storage unit 31 .

When the performance evaluation process is finished, the durability inspection unit 43 drives the divided organic EL panel 207 for a predetermined time under high temperature and high humidity conditions (durability test), and evaluates the performance of the divided organic EL panel 207 before and after driving ( durability evaluation process).

At this time, the current, voltage, luminance, luminous efficiency (lm/W), color coordinates (CIE-x, y), color coordinates (u′, v′), deviation from blackbody orbit (Δuv), color temperature (K), color rendering properties (Ra, R1 to R15), radiance (Le), change in color rank, temperature conditions in durability test, humidity conditions , test time, etc. are stored in the data storage unit 31 .
Further, the maximum peak intensity (P) and the maximum peak wavelength (λ) in the wavelength range of 500 nm to 630 nm are calculated by calculation, and the maximum peak intensity (P) and the maximum peak wavelength (λ) are calculated. It is accumulated in the data accumulation unit 31 as actual measurement data.

After that, the divided organic EL panel 207 is processed and commercialized, but since the processes after the durability evaluation process are the same as the conventional process, the description thereof is omitted.

Next, a machine learning operation by the machine learning unit 30 will be described.

The machine learning unit 30 performs machine learning with a machine learning model using past manufacturing parameters, quality parameters, and measured data used or inspected in the manufacturing process and quality confirmation process described above.
In the neural network 100 that constitutes the machine learning model, as shown in FIG. It is input to each node N1 of the most upstream node group 113 in the upstream intermediate layer 110 of the intermediate layer 102 . Each node N1 outputs a feature vector, and the feature vector is input in the upstream intermediate layer 110 after each node N2 of the downstream node group 114 is multiplied by a corresponding weight w2. The above processing is repeated as necessary.
Each node N2 of the node group 114 on the most downstream side in the upstream intermediate layer 110 outputs a feature vector, and the feature vector corresponds to the nodes D1, V1, L1, and T1 of the intermediate layer 111. It is input after being multiplied by weight w3.

Each of the nodes D1, V1, L1 and T1 of the midstream side intermediate layer 111 outputs a feature vector, which is multiplied by weighting coefficients W1 to W4 and input to each of the nodes D2, V2, L2 and T2. Each of the nodes D2, V2, L2, and T2 outputs a feature vector, and the feature vector is applied to each node N3 of the downstream intermediate layer 112 by a corresponding weight w4 and then input.

At this time, the weighting coefficients W1 to W4 are fixed values and are calculated based on the correlation among the nodes D1, V1, L1 and T1. That is, each node D1, V1, L1, T1 and each node D2, V2, L2, T2 have a one-to-one correlation.

Each node N3 of the most upstream node group 117 of the downstream intermediate layer 112 outputs a feature vector, and the feature vector corresponds to each node N4 of the downstream node group 118 in the downstream intermediate layer 112. It is input after being multiplied by weight w5. The above processing is repeated as necessary.
Then, each node N4 of the node group 118 on the most downstream side in the downstream intermediate layer 112 outputs a feature vector, and the feature vector is multiplied by the weight w6 corresponding to the nodes P and λ of the output layer 103. The red intensity peak of the divided organic EL panel 207 and the wavelength of the intensity peak are output.

Note that the weights w1 to w6 can be learned by the error backpropagation method, and the error backpropagation method is such that for each neuron (node), the output y when the input x is input and the true output y (teacher ) to adjust (learn) each weight w.
That is, the output value (P, λ) from the output layer 103 when the manufacturing parameter and the quality parameter are input to each neuron (node) of the input layer 101, and the measured value (P, λ) in the durability evaluation process The respective weights w1 to w6 are adjusted so that the difference between .

When the construction of the machine learning model based on the neural network 100 is completed as described above, the intensity peak in the red wavelength region of the divided organic EL panel 207 and the predicted value of the wavelength of the intensity peak are obtained using the machine learning model. Each manufacturing parameter is set so as to fall within the range, and using each set manufacturing parameter, the manufacturing department 2 manufactures the large-sized organic EL panel 200 in the manufacturing process, and the panel inspection department 4 in the quality confirmation process. divides the large-sized organic EL panel 200 to form divided organic EL panels 207 .

According to the organic EL panel manufacturing apparatus 1 of the present embodiment, the neural network 100 including four fixed nodes D1, V1, L1, and T1 in the midstream side intermediate layer 111 constitutes a machine learning model, and the solution is is narrowed down to the maximum peak intensity (P) and maximum peak wavelength (λ), machine learning can be performed more efficiently than before, and a solution can be obtained even using a huge amount of manufacturing parameters. Hard to dissipate.

According to the organic EL panel manufacturing apparatus 1 of the present embodiment, the fixed nodes D1, V1, L1, and T1 are the factors related to the in-plane thickness variation of the large-sized organic EL panel 200, and the light emitting layer is formed by the manufacturing department 2. A factor related to thickness variation, a factor related to variation in path length due to the position of the light emitting layer in the organic EL panel, and a factor related to time variation of the RGB ratio, so the maximum peak intensity based on the position of the light emitting interface Predicted values for (P) and wavelength (λ) can be calculated. As a result, the range of color temperatures can be roughly predicted from the predicted values of the maximum peak intensity (P) and wavelength (λ).

According to the organic EL panel manufacturing apparatus 1 of the present embodiment, since the divided organic EL panel 207 is manufactured using the manufacturing parameters predicted by the machine learning model, the maximum peak intensity (P) and the maximum peak wavelength ( λ) can be kept within a predetermined range, the color temperature can be kept within a predetermined range, and the chromaticity performance can be kept within the product standard. As a result, the number of defective products can be reduced and the yield can be improved.

According to the organic EL panel manufacturing apparatus 1 of the present embodiment, each weighting factor W1 to W4 assigned to each node D1, V1, L1, T1 is based on the correlation of each node D1, V1, L1, T1. Since each is set, machine learning can be performed more accurately and efficiently than before.

According to the organic EL panel manufacturing apparatus 1 of the present embodiment, since the weighting coefficients W1 to W4 in the midstream-side intermediate layer 111 in the machine learning model are constant values, the weighting coefficients W1 to W4 are constant across the midstream-side intermediate layer 111 by the error back propagation method. , the weights w1 to w6 can be adjusted.

Although the upstream intermediate layer 110 is composed of two layers in the above embodiment, the present invention is not limited to this. The upstream intermediate layer 110 may be one layer, or may be three or more layers.

Although the downstream intermediate layer 112 is composed of two layers in the above embodiment, the present invention is not limited to this. The downstream intermediate layer 112 may be one layer, or may be three layers or more.

In the above-described embodiment, the midstream side intermediate layer 111 is formed of four nodes D1, V1, L1, T1, but the present invention is not limited to this. Other neurons than the four nodes D1, V1, L1, T1 may be included in the layer.

In the above-described embodiment, the input layer 101 receives the manufacturing parameters and quality parameters in the manufacturing process and the quality parameters in the quality checking process, but the present invention is not limited to this.
For example, a machine learning model may be constructed by inputting only the manufacturing parameters in the manufacturing process to the input layer 101 .
Alternatively, a machine learning model may be constructed by inputting only manufacturing parameters and quality parameters in the manufacturing process into the input layer 101 .
According to these, at the stage of manufacturing the large-sized organic EL panel 200 in the manufacturing department 2, the node P that outputs the maximum peak intensity in the wavelength range of 500 nm to 630 nm and the node λ that outputs the peak wavelength of the maximum peak can be predicted.

In the above-described embodiment, the panel inspection unit 4 divides the large-sized organic EL panel 200 into a plurality of divided organic EL panels 207, but the present invention is not limited to this. In the panel inspection unit 4 , the large-sized organic EL panel 200 may be inspected as it is without dividing the large-sized organic EL panel 200 .
In this case, the object to be inspected is the large-sized organic EL panel 200, and the maximum peak intensity (P) and the maximum peak wavelength (λ) in the wavelength range of 500 nm to 630 nm are those of the large-sized organic EL panel 200.

As long as the above-described embodiments are within the technical scope of the present invention, each component can be freely replaced or added between the embodiments.

REFERENCE SIGNS LIST 1 manufacturing device 2 manufacturing department 3 control unit 30 machine learning unit 43 durability inspection unit (peak acquisition unit)
100 Neural network 101 Input layer 102 Intermediate layer 103 Output layer 110 Upstream intermediate layer 111 Midstream intermediate layer 112 Downstream intermediate layer 115 Midstream output layer (first node group)
116 Midstream side input layer (second node group)
200 Large-sized organic EL panel 201 Transparent conductive base material 202 Organic functional layer 203 Back electrode layer 205 Sealing layer 207 Divided organic EL panel 208 In-process base material 222 Blue light-emitting layer (light-emitting layer)
237 red-green light-emitting layer (light-emitting layer)

Claims (5)

a manufacturing department that manufactures an organic EL panel including a light-emitting layer based on a plurality of manufacturing parameters;
A machine learning unit that performs machine learning using a machine learning model having a neural network;
Having a peak acquisition unit that acquires the maximum peak intensity and wavelength in the wavelength range of 500 nm to 630 nm,
said neural network has an input layer, an intermediate layer and an output layer, said intermediate layer including an upstream intermediate layer, a midstream intermediate layer and a downstream intermediate layer;
The neural network receives the plurality of manufacturing parameters from the input layer, passes through the upstream intermediate layer, the midstream intermediate layer, and the downstream intermediate layer, and outputs the maximum peak intensity and wavelength from the output layer. which outputs the predicted value of
The midstream side intermediate layer has a first node group and a second node group,
the first node group has a plurality of nodes including a first node, a second node, a third node, and a fourth node;
The first node is a factor related to variations in in-plane thickness of the organic EL panel,
The second node is a factor related to variations in the thickness of the light-emitting layer formed by the manufacturing department,
the third node is a factor related to variation in path light length depending on the position of the light-emitting layer on the organic EL panel;
the fourth node is a time-varying factor of RGB ratio;
the second group of nodes is obtained by multiplying the plurality of nodes by corresponding weighting factors;
each weighting factor is a constant value assigned to the plurality of nodes;
Using the machine learning model, the plurality of manufacturing parameters are adjusted so that the predicted values of the maximum peak intensity and wavelength fall within a predetermined range, and the organic EL panel is manufactured using the plurality of adjusted manufacturing parameters. , Organic EL panel manufacturing equipment. Each weighting factor assigned to the first node, the second node, the third node, and the fourth node depends on the correlation between the first node, the second node, the third node, and the fourth node. 2. The apparatus for manufacturing an organic EL panel according to claim 1, wherein the manufacturing apparatus of the organic EL panel is set based on the 3. The organic EL panel manufacturing apparatus according to claim 1, wherein said first node group includes said first node, said second node, said third node, and said fourth node. . 4. The apparatus for manufacturing an organic EL panel according to claim 1, wherein said fourth node is a time-varying factor of color temperature. In the organic EL panel, an organic functional layer including the light-emitting layer, a back electrode layer, and a sealing layer are laminated on a transparent conductive substrate,
The manufacturing department includes a light-emitting functional layer forming process for forming the organic functional layer, a back electrode layer forming process for forming the back electrode layer, and a sealing layer forming process for forming the sealing layer. performing the steps in this order to form a work-in-process substrate and inspecting the quality parameters of the work-in-process substrate;
5. The apparatus for manufacturing an organic EL panel according to claim 1, further inputting said quality parameter of said in-process base material to said input layer.

Images