Prediction device, prediction system, prediction method, and program

Abstract

To improve prediction accuracy of the quality of a discharge film.SOLUTION: The present disclosure is a prediction device that predicts a quality of a film formed by an ejected liquid ejected using a liquid ejecting apparatus, the prediction device including an acquisition unit that acquires a learned model learned using a data set including a calculated value of a physical quantity representing a flight state of the ejected liquid or a hydrodynamic dimensionless number calculated from a geometric parameter representing a flow path shape of the liquid ejecting apparatus, and a quality result of the film formed by the ejected liquid, and a prediction unit that predicts the quality of the film formed by the ejected liquid using the learned model.SELECTED DRAWING: Figure 1

Description

[Technical Field]

[0001] The present disclosure relates to a prediction device, a prediction system, a prediction method, and a program.

[0002] In recent years, research has been conducted into methods for predicting the liquid ejection state (e.g., ejection amount, ejection speed, etc.) through machine learning using numerical data related to the physical properties of the ejected liquid and the ejection conditions in the formation of images and films by ejecting liquid. The physical properties of the liquid include viscosity and surface tension, and the numerical data related to the ejection conditions include the ejection pressure, the shape of the pulse voltage applied to the piezoelectric element, the nozzle open time, the nozzle shape, etc.

[0003] For example, a machine learning method has been proposed that learns the relationship between ejection parameters, including an ejection state value indicating the ejection state of a liquid ejection device and an ejection result value indicating the ejection result of the ejected liquid, as explanatory variables for machine learning, and image quality judgment results (see Patent Document 1).

[0004] The ejection status value, which indicates the ink flight status, is a value that quantifies, for example, whether or not the ink has been ejected, the amount of ink ejected, the ejection speed, or the number of satellites, and the ejection results include whether or not the ink has landed, the amount, shape, and area of ​​the ink that has landed, etc.

[0005] A value that quantifies the quality or level of image quality is generated as the image quality judgment result, and the quality or level of image quality is judged based on the degree of ink bleeding, the degree of ink coagulation, or the presence or absence of missing dots. Summary of the Invention [Problem to be solved by the invention]

[0006] However, the method described in Patent Document 1 involves obtaining image quality judgment results that judge the ejection conditions and printed image quality, and then adjusting the ejection conditions using a trained model to improve image quality, which results in a problem of low accuracy in predicting the quality of the ejected film.

[0007] Therefore, the present disclosure has been made in consideration of the above circumstances, and aims to improve the prediction accuracy of the quality of the discharged film. [Means for solving the problem]

[0008] The present disclosure relates to a prediction device for predicting the quality of a film formed by a discharge liquid discharged using a liquid discharge device, an acquisition unit that acquires a trained model that has been trained using a data set that includes a calculated value of a physical quantity that represents the flight state of the discharged liquid, or a hydrodynamic dimensionless number that is calculated from a geometric parameter that represents the flow path shape of the liquid discharge device, and a quality result of a film formed by the discharged liquid; a prediction unit that predicts the quality of a film formed by the discharged liquid using the trained model; The prediction device is provided with: [Effects of the Invention]

[0009] As described above, the present disclosure provides the effect of improving the prediction accuracy of the quality of the discharged film. [Brief explanation of the drawings]

[0010] [Figure 1] FIG. 1 is a diagram illustrating an example of a prediction system according to an embodiment of the present invention. [Figure 2] FIG. 2 is a block diagram showing an example of a hardware configuration of a computer according to the present embodiment. [Figure 3] FIG. 10 is a diagram illustrating an example of a method for creating a data set in this embodiment. [Figure 4] 1 shows the evaluation results of the prediction accuracy of the prediction device of Example 1. [Figure 5] 10 shows the evaluation results of the prediction accuracy of the prediction device of Example 2. [Figure 6] 10 shows the evaluation results of the prediction accuracy of the prediction device of Example 3. [Figure 7] 10 shows the evaluation results of the prediction accuracy of the prediction device of Comparative Example 1. DETAILED DESCRIPTION OF THE INVENTION

[0011] [Prediction device and prediction method] The prediction device disclosed herein is a prediction device that predicts the quality of a film formed by a discharged liquid discharged using a liquid discharge device, and includes an acquisition unit that acquires a trained model trained using a dataset including a calculated value of a physical quantity that represents the flight state of the discharged liquid or a hydrodynamic dimensionless number calculated from geometric parameters that represent the flow path shape of the liquid discharge device, and the quality results of the film formed by the discharged liquid, and a prediction unit that predicts the quality of the film formed by the discharged liquid using the trained model.

[0012] The predictor receives as explanatory variables the discharge conditions set in the liquid discharge device, the mass Mj of the discharged liquid droplets obtained from the flying droplet observation unit and droplet mass measurement unit, and the flight velocity Vj of the discharged liquid, as well as a dimensionless fluid dynamic number calculated from the geometric parameters representing the physical properties of the discharged liquid and the shape of the flow path. Next, the machine learning device separates the input data into training data and test data. A trained model is created using the training data with omission defects and contamination defects as objective variables. Test data is input into the created machine learning program, which predicts omissions and contamination, and verifies whether the actual omissions and contamination match the predictions.

[0013] The learning model can be logistic regression, light GBM, support vector machine, decision tree, random forest, neural network, etc. Next, the model design is revised using validation data, and hyperparameter tuning is performed. Finally, the model is finally evaluated using test data. Using this trained model, it becomes possible to predict defects and stains in the quality of the ejected film.

[0014] The prediction method disclosed herein is a prediction device that predicts the quality of a film formed by a discharged liquid discharged using a liquid discharge device, and includes an acquisition unit that acquires a trained model trained using a dataset that includes a calculated value of a physical quantity that represents the flight state of the discharged liquid or a hydrodynamic dimensionless number calculated from geometric parameters that represent the flow path shape of the liquid discharge device, and the quality results of the film formed by the discharged liquid, and a prediction unit that predicts the quality of the film formed by the discharged liquid using the trained model.

[0015] According to the present disclosure, the quality of the discharged film can be predicted from explanatory variables including calculated values ​​of physical quantities representing the state of the droplets in flight and parameters representing the flow path shape, or from fluid dynamic dimensionless numbers calculated from the physical properties of the discharged liquid, such as the inertial force of the discharged droplets flying at high speed through the air and the surface tension that the discharged droplets receive from the air interface, and geometric parameters representing the nozzle shape. This allows for a dramatic improvement in prediction accuracy, as predictions take into account the behavior of the droplets as they are discharged from the nozzle and the air resistance that the droplets experience in flight.

[0016] The physical quantity representing the flight state of the ejected liquid is not particularly limited and can be appropriately selected depending on the purpose, and examples thereof include the flight speed Vj (m / sec) of the ejected liquid, the mass Mj (μg) of a droplet of the ejected liquid, and the mass of the ejected liquid.

[0017] The calculated value of the physical quantity representing the flight state of the discharged liquid is a parameter that indicates the state in which the droplets discharged from the discharge holes are flying. Examples include the kinetic energy expressed by the following formula (1), the momentum expressed by the formula (2), and the density expressed by the formula (3). In particular, by using the kinetic energy of the droplets, the momentum of the droplets can be used in the calculation, which is related to the spread of the droplets upon impact and can improve the prediction accuracy of the film formation process.

[0018] Kinetic energy = Vj × Mj Equation (1)

[0019] Momentum = Vj 2×Mj...Equation (2)

[0020] Density = Mass of discharged liquid (g) / Mj (μg)...Equation (3)

[0021] Generally, droplets ejected from a nozzle with a certain mass and initial velocity travel a parabolic path due to the effects of gravity and air resistance. However, droplet diameters typically range from several microns to several hundred microns, initial velocities range from several meters per second to several dozen meters per second, and impact locations are typically within 10 mm of the nozzle, making the effects of gravity and air resistance negligible. The nozzle size is typically preferably several microns to 100 μm in diameter; in this example, a nozzle with a diameter of 50 μm was used. The droplet shape during flight can be controlled by the nozzle diameter, ejection rate, and ejection conditions, but it is often roughly spherical or accompanied by ligaments. When ligaments are present, droplets may merge with the leading droplet due to surface tension, or the ligaments may separate and merge into multiple droplets before flight. Therefore, physical quantities that describe the flight state include the droplet mass, flight velocity, droplet volume, maximum droplet diameter, ligament length, and flight time. In particular, the mass Mj of the droplet of the ejected liquid and the flight velocity Vj of the ejected liquid are preferably used to express the state of ejection. The calculated value expressing the flight state is not limited to the momentum, density, and kinetic energy of the droplet, but it is also possible to use the ratio or product of two or more physical quantities selected from the mass of the droplet of the ejected liquid, the flight velocity of the ejected liquid, the droplet volume, the maximum droplet diameter, the ligament length, the flight time, and the like, which are physical quantities expressing the flight state.

[0022] The fluid dynamic dimensionless numbers calculated from the geometric parameters that represent the flow path shape include the forces acting on the fluid due to inertia, pressure, viscosity, gravitational acceleration, surface tension, and elasticity, and are expressed as follows, where L is the characteristic length that represents the characteristics of the flow in the system, V is the speed of the fluid in the flow path or during flight, and ρ is the density: These numbers are used to fluid-dynamically analyze the motion of a model using the ratio or product of these forces.

[0023] Inertial force:Fi=ρV 2 L 2

[0024] Pressure force: Fp=pL 2

[0025] Viscous force: Fμ=μVL

[0026] Force due to gravitational acceleration: Fg=ρL 3 g (g indicates the acceleration due to gravity)

[0027] Surface tension force: Fσ=σL

[0028] Elastic force: Fk=KL 2 (K is the bulk modulus)

[0029] Examples of fluid dynamic dimensionless numbers include the capillary number, the Ohnesorge number, the Reynolds number, and the Weber number. The fluid dynamic dimensionless numbers include the capillary number, which indicates the competition between inertial force and surface tension, the Ohnesorge number, which indicates how much the behavior of a fluid is affected by viscous force and surface tension relative to inertial force, the Reynolds number, which indicates the boundary between laminar flow and turbulent flow of a fluid, and the Weber number, which indicates how much resistance a fluid has to gravity. This allows the behavior of a droplet at the time of ejection and during flight to be quantitatively incorporated into calculations based on the inertial force, viscous force, surface tension, etc., that the droplet experiences, thereby further improving prediction accuracy.

[0030] Generally, the capillary number, Ohnesorge number, Reynolds number, and Weber number, which are dimensionless hydrodynamic numbers, are defined as follows:

[0031] Ohnesorge number: On = inertia force x surface tension / viscous force = μ / √ρσL...Equation (4) (In equation (4), μ is the viscosity of the fluid, ρ is the density of the fluid, σ is the surface tension of the fluid, and L is a specific length scale.)

[0032] Examples of specific length scales include the radius of a tube, the diameter of a capillary, and the characteristic length of a flow.

[0033] The Ohnesorge number indicates how much the behavior of a fluid is affected by viscous and surface tension forces relative to inertial forces.

[0034] Capillary number: Ca = viscous force / surface tension = μV / σ Equation (5) (In equation (5), μ is the viscosity of the fluid, V is the velocity of the fluid, and σ is the surface tension of the fluid.)

[0035] The capillary number indicates the competition between inertial forces and surface tension. When this number is small, surface tension is dominant and capillary action is caused by the capillary effect. On the other hand, when the capillary number is large, inertial forces are dominant and the behavior of the fluid approaches that caused by viscosity.

[0036] Reynolds number: Re = inertia force / viscous force = ρVL / μ...Equation (6) (In equation (6), ρ is the density of the fluid, V is the velocity of the fluid, L is a specific length scale, and μ is the viscosity of the fluid.)

[0037] The Reynolds number represents the boundary between laminar and turbulent fluid flow; when Re << 1, viscous forces are dominant and the flow is called laminar. When Re >> 1, inertial forces are dominant and the flow is called turbulent, a very complex state.

[0038] Weber number: We = inertial force / surface tension = ρV 2 L / σ···Eq.(7) (In equation (7), ρ is the density of the fluid, V is the velocity of the fluid, L is a specific length scale, and σ is the surface tension of the fluid.)

[0039] The Weber number indicates how much resistance a fluid has to gravity. Specifically, when the Weber number is large, surface tension dominates over gravity, making the shapes of droplets and bubbles more stable. On the other hand, when the Weber number is small, gravity dominates and the fluid is more likely to change shape.

[0040] The droplet flight observation unit is installed at a position several millimeters to several tens of millimeters from the liquid ejection head, and takes pictures from a direction approximately perpendicular to the droplet flight direction.A strobe is activated in synchronization with the droplets ejected at a constant frequency, and by taking pictures for several seconds, droplets flying at equal intervals are photographed.The image determination unit measures the distance between droplets, making it possible to measure the flight speed Vj of the ejected liquid.

[0041] The droplet mass measurement unit ejects a predetermined number of droplets under certain conditions from a downward-facing liquid ejection head into a container filled with silicone oil, measures the mass of the silicone oil and the ejected liquid into the silicone oil, and calculates the average droplet mass My using a mass calculation unit.

[0042] [Prediction System] The prediction system of the present disclosure is a prediction system including a prediction device and a liquid ejection device.

[0043] Hereinafter, an embodiment of a prediction system including a prediction device according to the present disclosure will be described with reference to the drawings.

[0044] 1, the prediction system includes a prediction device (hereinafter, sometimes referred to as a "machine learning device") and a liquid ejection device. The prediction device and the liquid ejection device are connected to each other so as to be able to communicate data with each other via a communication network such as a LAN (Local Area Network) or the Internet. The prediction device is an information processing device such as a personal computer, a workstation, or a server that predicts the quality of a film formed by a liquid discharged using a liquid discharge device.

[0045] The device has an arithmetic circuit that controls the entire device, a recording device that stores data, a drive circuit for driving a liquid ejection head having piezoelectric elements corresponding to a large number of nozzles, a detection circuit that detects the state of the drive circuit and synchronizes with the control circuit, a control circuit that monitors and controls the state of the drive circuit and stage movement, and a liquid pressure-feeding mechanism that controls the supply of liquid.Furthermore, the control circuit controls a flying droplet observation unit and a droplet mass measurement unit to measure the flight speed and droplet mass of the ejected liquid.

[0046] There are various types of liquid ejection devices, including spray, dispenser, and inkjet types. In particular, inkjet types have an advantage over other types in that they can precisely control the speed and size of droplets ejected by bringing an inkjet head close to the application medium, applying pressure to the ink in the individual nozzles and flow paths, and controlling the time of ejection or non-ejection for each nozzle, making it possible to draw letters, images, and other images.

[0047] In order to produce high-quality images using an inkjet liquid ejection device, it is extremely important to control the flight state of the droplets. For this reason, as in the present invention, it is necessary to predict the quality of the ejected film from explanatory variables that include the ejection conditions of the liquid ejection device and calculated values ​​of physical quantities that represent the flight state of the ejected liquid.

[0048] Inkjet printers include, but are not limited to, piezoelectric types that eject ink by deforming a piezoelectric element, thermal types that generate bubbles in the ink by heating it with a heater and eject it from the nozzle, and valve types that eject pressurized ink by opening and closing the nozzle cover.

[0049] In this embodiment, a valve type inkjet printer will be used for explanation.

[0050] The ejection conditions for a valve-type inkjet printer include the liquid pressure applied to the ejected liquid, the valve open time and valve lift amount driven by a piezoelectric element, frequency, stage speed, and gap (distance between the nozzle and the substrate).

[0051] <Hardware configuration> The hardware configuration of the prediction system in this embodiment will be described with reference to FIG.

[0052] Computer The prediction device and the terminal liquid ejection device in this embodiment are realized by, for example, a computer. Fig. 2 is a block diagram showing an example of the hardware configuration of a computer in this embodiment.

[0053] As shown in Figure 2, the computer has a CPU (Central Processing Unit), ROM (Read Only Memory), RAM (Random Access Memory), HDD (Hard Disk Drive), input device, display device, communication I / F (Interface), and external I / F. The CPU, ROM, and RAM form what is known as a computer. Each piece of computer hardware is connected to one another via a bus line. The input device and display device may also be connected to the external I / F for use.

[0054] A CPU is a computing device that controls and functions the entire computer by reading programs and data from storage devices such as ROM or HDD into RAM and executing the processes.

[0055] ROM is an example of a non-volatile semiconductor memory (storage device) that can retain programs and data even when the power is turned off. ROM functions as a main storage device that stores various programs and data necessary for the CPU to execute various programs installed on the HDD. Specifically, ROM stores boot programs such as the BIOS (Basic Input / Output System) and EFI (Extensible Firmware Interface) that are executed when the computer starts up, as well as data such as OS (Operating System) settings and network settings.

[0056] RAM is an example of volatile semiconductor memory (storage device) in which programs and data are erased when the power is turned off. Examples of RAM include DRAM (Dynamic Random Access Memory) and SRAM (Static Random Access Memory). RAM provides a working area where various programs installed on the HDD are expanded when executed by the CPU.

[0057] An HDD is an example of a non-volatile storage device that stores programs and data. The programs and data stored on an HDD include the OS, which is the basic software that controls the entire computer, and applications that provide various functions on the OS. Note that instead of an HDD, a computer may use a storage device that uses flash memory as a storage medium (e.g., an SSD (Solid State Drive)).

[0058] The input device is a touch panel used by the user to input various signals, operation keys or buttons, a keyboard or mouse, a microphone for inputting sound data such as voice, etc.

[0059] The display device is composed of a display such as a liquid crystal display or an organic electroluminescence (EL) display for displaying a screen, a speaker for outputting sound data such as voice, and the like.

[0060] The communication I / F is an interface that connects to a communication network and enables a computer to communicate data.

[0061] The external I / F is an interface with an external device, such as a drive device.

[0062] A drive unit is a device into which a recording medium is inserted. The recording medium referred to here includes media that record information optically, electrically, or magnetically, such as CD-ROMs, flexible disks, and magneto-optical disks. Recording media may also include semiconductor memories that record information electrically, such as ROMs and flash memories. This allows a computer to read and / or write data from and to the recording medium via an external interface.

[0063] The various programs to be installed on the HDD are installed, for example, by setting the distributed recording medium in a drive device connected to an external I / F and reading the various programs recorded on the recording medium by the drive device. Alternatively, the various programs to be installed on the HDD may be installed by downloading them via the communication I / F from a communication network or a network different from the communication network.

[0064] 3 is a diagram showing an example of a method for creating a dataset in this embodiment. The method for creating a dataset is executed by the prediction device in this embodiment, and the specific procedure is as follows: steps S21 to S26.

[0065] In step S21, the moving speed of the stage moving mechanism is set by the arithmetic unit and the control circuit based on the data retrieved from the storage device.

[0066] In step S22, the physical property values ​​of the ink used for printing are input, ρ: ink density, μ: ink viscosity, and σ: ink surface tension.

[0067] In step S23, the frequency input to the piezoelectric element of the liquid ejection head, OPT (valve open time), and liquid pressure are set by the arithmetic unit and the drive circuit based on the data called from the storage device.

[0068] In step S24, printing is performed at the stage speed set in step S21 and the discharge conditions set in step S23.

[0069] In step S25, a stroboscopic photograph is taken by the flying droplet observation unit, and is transferred to the control circuit to measure the flying velocity Vj of the ejected liquid.

[0070] In step S26, the ejection-formed film quality measurement unit observes defects such as voids where sufficient ink is not adhered within the drawn line of the ejection-formed film, and stain defects scattered around the drawn line, and judges whether the film is good or bad.

[0071] The droplet mass measurement unit measures the mass Mj of the droplets of the ejected liquid for a predetermined number of droplets, and the mass calculation unit calculates the droplet mass per dot. A data set is generated from the calculated Vj, Mj, quality results (missing, staining), and the values ​​ρ (ink density), μ (ink viscosity), and σ (ink surface tension) input in step S22.

[0072] 〔program〕 The prediction program disclosed herein causes a computer to execute the following steps: acquire a trained model trained using a dataset including calculated values ​​of physical quantities representing the flight state of the ejected liquid or dimensionless hydrodynamic numbers calculated from geometric parameters representing the flow path shape of the liquid ejection device, and quality results of a film formed by the ejected liquid; and predict the quality of the film formed by the ejected liquid using the trained model. In other words, when explanatory variables are input according to the generated trained model, it is possible to predict print quality caused by defects such as omissions and stains, which are the objective variables. [Example]

[0073] The present embodiment will be described in more detail below with reference to examples, although the present embodiment is not limited to the following examples.

[0074] <Discharge liquid (ink production)> A total of 45 types of inks, including inks A and B, were prepared by adjusting the types and mixing ratios of solvents, binders, surfactants, viscosity modifiers, and water. The physical properties of the prepared inks, such as surface tension, density, and shear viscosity, are shown in Table 1 below.

[0075] [Table 1]

[0076] <Film formation> Using 45 types of ink, films were formed under the discharge conditions and discharge results shown in Table 2 below, and the quality (missing and staining) of the formed films was evaluated visually. As a result, 370 data sets were created for the physical quantities (discharge results) that represent the flight state of the discharged liquid, their calculated values, discharge conditions, and quality (missing and staining) for the 45 types of ink.

[0077] [Table 2]

[0078] Example 1 Of the 45 ink datasets (370 entries) created, 30 ink datasets (250 entries) were used as training data, and a trained model for predicting film quality was constructed using LightGBM as the machine learning program.

[0079] The prediction accuracy of the constructed trained model was evaluated using a dataset of 15 types of ink (120 items) as test data. The results are shown in Figure 4.

[0080] The explanatory variables are the liquid pressure, which is the discharge condition of the discharge device, the nozzle open time OPT, and the kinetic energy of the droplet (Mj × Vj) as a calculated value of the physical quantity that represents the flight state. 2 The physical properties of the ejected liquid were the surface tension, density, and viscosity of the ejected liquid. The objective variables were the omission and contamination of the ejected liquid under each condition.

[0081] <Example 2> A trained model was constructed in the same manner as in Example 1, except that the liquid pressure, which is the discharge condition of the discharge device, the nozzle open time OPT, and the momentum of the droplet (Mj × Vj) as the calculated value of the physical quantity representing the flight state were used as explanatory variables in Example 1. The results are shown in Figure 5.

[0082] Example 3 A trained model was constructed in the same manner as in Example 1, except that the liquid pressure, nozzle open time OPT, and capillary number, which are the discharge conditions of the discharge device, and which is one of the hydrodynamic dimensionless numbers, were used as explanatory variables, and the prediction accuracy was evaluated. The results are shown in Figure 6.

[0083] <Comparative Example 1> A trained model was constructed in the same manner as in Example 1, except that the discharge conditions of the discharge device, namely, the liquid pressure, the nozzle open time OPT, and the physical quantities during flight (Mj and Vj), were used as explanatory variables, and the prediction accuracy was evaluated. The results are shown in Figure 7.

[0084] In Examples 1 to 3, it was possible to predict staining and detachment with a high probability, as shown in Figures 4 to 6. On the other hand, in Comparative Example 1, the prediction accuracy was poor, as shown in Figure 7, and it was not possible to predict staining and detachment.

[0085] The present invention includes, for example, the following aspects. <1> A prediction device for predicting the quality of a film formed by a discharge liquid discharged using a liquid discharge device, comprising: an acquisition unit that acquires a trained model that has been trained using a data set that includes a calculated value of a physical quantity that represents the flight state of the discharged liquid, or a hydrodynamic dimensionless number that is calculated from a geometric parameter that represents the flow path shape of the liquid discharge device, and a quality result of a film formed by the discharged liquid; a prediction unit that predicts the quality of a film formed by the discharged liquid using the trained model; A prediction device comprising: <2> The calculated value of the physical quantity is kinetic energy represented by the following formula (1): The prediction device according to claim 1. Kinetic energy = Vj × Mj Equation (1) (In equation (1), Vj represents the flight speed (m / sec) of the ejected liquid, and Mj represents the mass (μg) of the droplet of the ejected liquid.) <3> The calculated value of the physical quantity is a momentum expressed by the following formula (2): the above <1> The prediction device according to claim 1. Kinetic energy = Vj 2 (m / sec)×Mj(μg)······································Formula (2) (In equation (2), Vj represents the flight speed (m / sec) of the ejected liquid, and Mj represents the mass (μg) of the droplet of the ejected liquid.) <4> The hydrodynamic dimensionless number is at least one of a Capillary number, an Ohnesorge number, a Reynolds number, and a Weber number. the above <1> from <3> 10. The prediction device according to claim 9, <5> the prediction unit predicts the quality of a film formed by the discharge liquid using a trained model trained using a data set including discharge conditions of the liquid discharge device. the above <1> from <4> 10. The prediction device according to claim 9, <6> the prediction unit predicts the quality of the film formed by the ejection liquid using a trained model trained using a data set including the physical properties of the ejection liquid. the above <1> from <5> 10. The prediction device according to claim 9, <7> the above <1> from <6> A prediction system comprising: the prediction device according to any one of claims 1 to 4; and a liquid ejection device. <8> A method for predicting the quality of a film formed by a liquid discharged using a liquid discharge device, comprising: The computer a step of acquiring a trained model trained using a dataset including a calculated value of a physical quantity representing the flight state of the discharged liquid, or a fluid dynamic dimensionless number calculated from a geometric parameter representing the flow path shape of the liquid discharge device, and a quality result of a film formed by the discharged liquid; a step of predicting the quality of a film formed by the ejection liquid using the trained model; A prediction method characterized by performing the following. <9> On the computer, a step of acquiring a trained model trained using a dataset including a calculated value of a physical quantity representing the flight state of the discharged liquid, or a fluid dynamic dimensionless number calculated from a geometric parameter representing the flow path shape of the liquid discharge device, and a quality result of a film formed by the discharged liquid; a step of predicting the quality of a film formed by a discharged liquid using the trained model; A program to execute. [Prior art documents] [Patent documents]

[0086] [Patent Document 1] Japanese Patent Publication No. 2022-49858

Claims

1. A prediction device for predicting the quality of a film formed by a discharge liquid discharged using a liquid discharge device, comprising: an acquisition unit that acquires a trained model that has been trained using a data set that includes a calculated value of a physical quantity that represents the flight state of the discharged liquid, or a hydrodynamic dimensionless number that is calculated from a geometric parameter that represents the flow path shape of the liquid discharge device, and a quality result of a film formed by the discharged liquid; a prediction unit that predicts the quality of a film formed by the discharged liquid using the trained model; A prediction device comprising:

2. The calculated value of the physical quantity is kinetic energy represented by the following formula (1): The prediction device according to claim 1 . Kinetic energy = Vj × Mj Equation (1) (In formula (1), Vj represents the flight speed (m / sec) of the ejected liquid, and Mj represents the mass (μg) of the droplet of the ejected liquid.)

3. The calculated value of the physical quantity is a momentum expressed by the following formula (2): The prediction device according to claim 1 . Kinetic energy = Vj ２ (m / sec)×Mj (μg)...Formula (2) (In formula (2), Vj represents the flight speed (m / sec) of the ejected liquid, and Mj represents the mass (μg) of the droplet of the ejected liquid.)

4. The hydrodynamic dimensionless number is at least one of a Capillary number, an Ohnesorge number, a Reynolds number, and a Weber number. The prediction device according to claim 1 .

5. the prediction unit predicts the quality of a film formed by the discharge liquid using a trained model trained using a data set including discharge conditions of the liquid discharge device. The prediction device according to claim 1 .

6. the prediction unit predicts the quality of the film formed by the ejection liquid using a trained model trained using a data set including the physical properties of the ejection liquid. The prediction device according to claim 1 .

7. A prediction system comprising: the prediction device according to claim 1 ; and a liquid ejection device.

8. A method for predicting the quality of a film formed by a liquid discharged using a liquid discharge device, comprising: The computer a step of acquiring a trained model trained using a dataset including a calculated value of a physical quantity representing the flight state of the discharged liquid, or a fluid dynamic dimensionless number calculated from a geometric parameter representing the flow path shape of the liquid discharge device, and a quality result of a film formed by the discharged liquid; a step of predicting the quality of a film formed by the ejection liquid using the trained model; A prediction method characterized by performing the following.

9. On the computer, a step of acquiring a trained model trained using a dataset including a calculated value of a physical quantity representing the flight state of the discharged liquid, or a hydrodynamic dimensionless number calculated from a geometric parameter representing the flow path shape of the liquid discharge device, and a quality result of a film formed by the discharged liquid; a step of predicting the quality of a film formed by a discharged liquid using the trained model; A program to execute.

Images