JP2022085257A - Cleaning system, control device, cleaning method, control method, program, recording medium and manufacturing method of article - Google Patents

Abstract

To provide a cleaning device and method for cleaning a cleaning object under a cleaning condition which exhibits the high cleaning performance and prevents the cleaning object from being damaged according to the type of defect of the cleaning object and the like.SOLUTION: A cleaning system comprises: a cleaning device which has a cleaning unit that cleans a cleaning object; an image acquisition device which acquires an image of the cleaning object before and/or after cleaning; an image evaluation unit which evaluates a state of the cleaning object on the basis of a type of defect of the cleaning object from the image information; a control unit which controls operation of the cleaning device; and a machine learning unit which generates cleaning conditions of the cleaning device on the basis of the image acquired by the image acquisition device and a result obtained by evaluating the state of the cleaning object by the image evaluation unit. The control unit controls the operation of the cleaning device by using the cleaning conditions generated by the machine learning unit.SELECTED DRAWING: Figure 1

Description

The present invention relates to a cleaning system for cleaning an object to be cleaned, a control device, a cleaning method, a control method program, a recording medium, and a method for manufacturing an article.

Conventionally, in the cleaning of optical elements such as lenses and precision parts, high cleaning performance is required to remove stains having dimensions of micron level or less. Further, when the target material is soft or the microstructure is fragile, it is important that the object to be cleaned is not damaged such as damage.

However, the cleaning conditions that exhibit high detergency and do not damage the object to be cleaned differ depending on the material and shape of the object to be cleaned, the number of objects to be cleaned, and the like. Further, especially in ultrasonic cleaning, a plurality of cleaning conditions affect the cleaning property and damage to the object to be cleaned.

Patent Document 1 discloses a cleaning process optimizing device that optimizes cleaning conditions when cleaning an object to be cleaned. According to this device, the cleaning conditions can be optimized based on the dirt state of the cleaning target.

Japanese Unexamined Patent Publication No. 2018-153872

However, we have not optimized the cleaning conditions based on cleaning defects other than the dirty state, and we have optimized the cleaning conditions that achieve both high cleaning performance and no damage to the object to be cleaned. It was difficult to generate.

The cleaning system for solving the above problems includes a cleaning device having a cleaning unit for cleaning the object to be cleaned, an image acquisition device that acquires an image of the object to be cleaned before and / or after cleaning, and the image information. An image evaluation unit that evaluates the state of the object to be cleaned based on the type of defect of the object to be cleaned, a control unit that controls the operation of the cleaning device, and the image and the image acquired by the image acquisition device. The evaluation unit includes a machine learning unit that generates cleaning conditions for the cleaning device based on the result of evaluating the state of the object to be cleaned, and the control unit uses the cleaning conditions generated by the machine learning unit. It is characterized in that the operation of the cleaning device is controlled.

INDUSTRIAL APPLICABILITY According to the present invention, it is possible to wash an object to be cleaned under cleaning conditions that exhibit high detergency and do not damage the object to be cleaned according to the type of defect of the object to be cleaned.

It is a schematic diagram of the cleaning system which concerns on embodiment. It is the schematic of the image pickup part which concerns on embodiment. It is a schematic diagram of the washing tank which concerns on embodiment. This is an example of information and parameters handled in the present invention. It is a flow figure which shows the flow of the washing which concerns on embodiment.

An embodiment of the present invention will be described with reference to the drawings.

(System configuration)
FIG. 1 is a schematic view of the cleaning system 1 according to the embodiment. The cleaning system 1 includes a cleaning device 10, an image acquisition device 20, a control device 30 for controlling the cleaning device 10 and the image acquisition device 20, an information input device 40, a first accommodating unit 510, and a second accommodating unit 520. And. The control device 30 is communicably connected to the cleaning device 10 and the image acquisition device 20.

The cleaning device 10 includes a cleaning unit 110, a state detection unit 120, and a condition setting unit 130. The cleaning unit 110 includes at least one cleaning tank, and cleans the object to be cleaned 2 transferred to the cleaning tank. The state detection unit 120 detects the state of the cleaning unit 110 including the cleaning tank for cleaning the object to be cleaned 2. The state detection unit 120 includes a sound pressure probe for detecting an ultrasonic state (not shown), an oscilloscope, and various sensors for detecting the cleaning liquid temperature, cleaning liquid ionization degree (pH), electrical resistivity (specific resistance), dissolved gas amount, etc. of the cleaning tank. Consists of. The condition setting unit 130 sets the conditions under which the cleaning unit 110 cleans the object to be cleaned 2 in the cleaning tank. Further, the details of the information handled in the cleaning tank of the cleaning unit 110 and each unit and its parameters will be described later.

The image acquisition device 20 includes an image output unit 210, an image pickup unit 220, and a drive control unit 230. The image pickup unit 220 includes the image pickup means 221, the lighting means 222, and the respective driving means 223a and 223b, and captures an image of the object to be cleaned 2 before and / or after cleaning. FIG. 2 shows an example of an image pickup unit 220 that irradiates the object to be cleaned 2 with light and captures it as an image. The lighting means 222 is, for example, an LED, which is attached to the driving means 223b and irradiates the object to be cleaned 2 with illumination light from various directions based on the operation of the driving means 223b. The image pickup means 221 is, for example, a commercially available digital camera, which is attached to the drive means 223a and changes the image pickup direction with respect to the object to be cleaned 2 or with respect to the object to be cleaned 2 based on the operation of the drive means 223a. You can change the focal position. However, other methods may be used as long as the same result can be obtained. The drive control unit 230 controls the operation of the drive means 223a and 223b of the image pickup unit 220 under conditions according to the shape of the object to be cleaned 2 and the like. The image output unit 210 outputs the image of the object to be cleaned 2 captured by the image pickup unit 220 to the outside of the image acquisition device 20.

The control device 30 includes a control unit 310, a machine learning unit 320, an information acquisition unit 330, an image evaluation unit 340, and a storage unit 350. The control unit 310 controls the operation of the cleaning system 1 by performing information communication with each unit included in the cleaning system 1. The information acquisition unit 330 acquires information about the object to be cleaned 2 from the information input device 40. The image evaluation unit 340 evaluates the image of the object to be cleaned 2 acquired by the image pickup unit 220, inspects the object to be washed 2 and calculates a reward for the image. The machine learning unit 320 has a learning model 321 and inputs one or more of the following information to the learning model 321 to generate new cleaning conditions. The information input to the learning model 321 includes the unique information of the object to be cleaned 2, the image information acquired by the image acquisition device 20, the cleaning conditions of the cleaning device 10, the detection result detected by the state detection unit 120, and the image evaluation unit 340. Calculated rewards, etc. Further, the machine learning unit 320 calculates the unique information of the object to be cleaned 2, the image information acquired by the image acquisition device 20, the cleaning conditions of the cleaning device 10, the detection result detected by the state detection unit 120, and the image evaluation unit 340. Learn at least one of the rewards you have made. Update the learning model based on the learning results. The storage unit 350 includes the unique information of the object to be cleaned 2 when the object to be cleaned 2 is washed, the image information acquired by the image acquisition device 20, the cleaning conditions of the cleaning device 10, and the detection detected by the state detection unit 120. The result and the reward calculated by the image evaluation unit 340 are stored. The unique information of the object to be cleaned 2 stored in the storage unit 350 may be managed in association with the part number of the object to be cleaned 2. Further, as the unique information of the object to be cleaned 2, the information stored in the CAD, the parts management database, or the database on the cloud existing outside the cleaning system 1 may be used.

The information input device 40 is a user interface for inputting unique information of the object to be cleaned 2 to the information acquisition unit 330. The information input device 40 includes a monitor, a keyboard, a mouse, and the like. As the unique information of the object to be cleaned 2, the information input from the information input device 40 can be used, or the information already stored in the storage unit 350 can be used. Further, the information to be input is not limited to the unique information of the object to be cleaned 2, and any information useful for the operation of the cleaning system 1 and the cleaning system 1 can be input.

The uncleaned object 2 is housed in the first accommodating portion 510, and the uncleaned object 2 is the cleaning unit 110 of the cleaning device 10 or the image pickup unit 220 of the image acquisition device 20 depending on the timing of cleaning. Will be reprinted in. The second accommodating unit 520 accommodates the unit to be cleaned whose image information has been acquired by the image pickup unit 220 of the image acquisition device 20 after cleaning.

With the above configuration, the unwashed object 2 housed in the first storage unit 510 is housed in the first storage unit 510 to the cleaning unit 110, from the cleaning unit 110 to the image pickup unit 220, and from the image pickup unit 220 to the second storage unit 220. Cleaning is completed through the process of sequential transfer to section 520. The transfer between the configurations may be performed manually by a person or automatically by a transfer device (not shown).

(Washing tank)
Here, the cleaning tank when the cleaning device 10 is an ultrasonic cleaning device will be described. The cleaning tank of the cleaning unit 110 has an ultrasonic oscillator at either the bottom or the side surface. A cleaning liquid is put into the cleaning tank, and ultrasonic vibration is applied to the cleaning liquid to clean the object to be cleaned 2 transferred into the tank. The cleaning parameters are cleaned by the control from the condition setting unit 130. Typical cleaning parameters include frequency and output as ultrasonic conditions, and these are cleaned in the optimum range, but the frequency and output range can be selected according to the purpose and size. For example, when removing small-sized adhered stains, a high frequency is often selected as the frequency, and in the case of large stains or strong stains, a low frequency is often selected. Further, by having oscillators with different oscillation frequencies in one tank, it is possible to select the frequency and transmit at the same time. Further, the cleaning time, the temperature of the cleaning liquid, and the like are also important factors that determine the cleaning performance, and these are determined by the conditions and purposes such as the type of dirt and the heat resistance of the object to be cleaned 2.

Further, when the object 2 to be cleaned is a glass lens, the glass lens obtains optical characteristics such as a refractive index and dispersibility, so that the material composition is diverse. Therefore, the hardness and chemical resistance of the object to be cleaned 2 differ depending on the material composition. For example, in the case of glass containing a large amount of fluorine, the object to be cleaned 2 is generally soft and easily scratched. On the other hand, various stains such as abrasives, inorganic deposits such as glass powder, processing oil, sebum, dust, etc. are often attached, and it is necessary to select a cleaning method and parameters suitable for each stain. ..

Examples of the cleaning liquid to be used include those called alkaline or acidic water-based cleaning agents, quasi-water-based cleaning agents, hydrocarbon-based cleaning agents, alcohol-based cleaning agents, and the like. In addition, it may be washed with water, hydrogen water containing hydrogen, or functional water such as fine bubble water containing fine bubbles. Generally, when cleaning with a cleaning agent, a rinsing step is required after the cleaning agent step. After that, it is dried in a drying step. Each process is often performed in a plurality of tanks, and the number of tanks may be a multi-tank type of 10 or more tanks in some cases.

The configuration of the cleaning unit will be described in detail with reference to FIG. FIG. 3 is a view of a three-tank type cleaning unit provided with a cleaning agent tank 111, a rinsing tank 112, and a drying tank 113 from the left. In the first step, the object to be cleaned 2 is immersed in the cleaning liquid 401 contained in the first cleaning tank 111 to remove stains. In the next step, the object to be cleaned 2 is immersed in the rinse solution 402 contained in the second rinse tank 112, and the cleaning solution 401 adhering to the object to be cleaned 2 is rinsed in the detergent tank 111. In the next step, the object to be cleaned 2 is immersed in the ultrapure water 403 placed in the third drying tank 113, the object to be cleaned 2 is pulled up, and the object to be cleaned 2 is dried with warm air or a heated atmosphere. In the illustrated warm pure water drying method, ultrapure water having an electric conductivity of about 0.1 S / m is used, and the ultrapure water is often heated to about 50 ° C. In addition to this, the drying may be performed using a fluorine-based solvent, drying using alcohol, drying by a physical method such as air blow or spin, or the like. An ultrasonic oscillator unit 404 may be installed in each tank, and the ultrasonic oscillation is set to conditions suitable for the work.

Even in the case of the above-mentioned three-tank type cleaning, there are various factors related to cleaning. These factors must be set to optimal conditions. However, even if the cleaning conditions are initially set optimally, continuous cleaning may cause changes in the liquid temperature, changes in the liquid due to evaporation, fluctuations in pH, and changes in the amount of dissolved gas in the liquid. Therefore, the ultrasonic intensity and chemical influence will change. Therefore, it is not easy to maintain both detergency and damageless cleaning.

(Handling information and parameters)
Next, the information handled in the present invention and its parameters will be described. The information handled in the present invention can be mainly classified into the following four types. That is, the detection result detected by the state detection unit 120 at the time of cleaning, the image information of the object to be cleaned 2, the unique information of the object to be cleaned 2 not including the image information, and the information of the cleaning conditions generated by the machine learning unit 320. .. Typical parameters of each are shown in FIG. However, this parameter is typical, and these vary depending on the content to be cleaned, and the present invention is not limited to the items described here.

The detection results detected by the state detection unit 120 during ultrasonic cleaning include sound pressure, which is an index of cavitation strength, actually oscillated ultrasonic frequency and oscillation waveform, and ultrasonic elements. There is a sound wave oscillation time. These can be detected with a sound pressure probe, an oscilloscope, etc., and the information captured as a signal by measurement can be processed. The frequency is about 20 kHz to 5 MHz in general ultrasonic cleaning, and the frequency and waveform in that range are measured. The main information on the liquid state includes temperature, cleaning liquid ionization degree (pH), electrical resistivity (specific resistance), and amount of dissolved gas, and each detection method may be a general method.

All of these parameters are factors that affect detergency and damage, but among them, sound pressure is the detergency and has a large effect on detergency in ultrasonic cleaning, which is an important factor for controlling cleaning conditions. .. This sound pressure is determined not only by the output applied to the ultrasonic vibrator, but also by the temperature of the cleaning liquid and the amount of dissolved gas in the cleaning liquid. In particular, the amount of dissolved gas decreases due to ultrasonic waves when washing is continuously performed. This is an important factor because the sound pressure intensity generated changes even if the same ultrasonic wave is applied. However, since the amount of dissolved gas cannot be easily adjusted from the outside like the temperature, the change in sound pressure due to the change in dissolved gas must be covered by other ultrasonic parameters so that the cleaning result is optimized.

Next, the image information of the object to be cleaned 2 includes the degree of stain, the position of stain, the number of scratches, the shape of scratches, the degree of cloudiness, and the respective defective positions. These are elements that can be acquired by the image pickup unit 220 and can be evaluated by the image evaluation unit 340.

The unique information of the object to be cleaned 2 does not include image information and varies depending on what the object to be cleaned 2 looks like, and is mounted on the hardness, size, thickness, shape complexity, number, and cleaning basket. Or, if it is installed, what kind of form it is. For example, when considering an optical lens, the outer diameter of the lens, the curvature (R), the thickness of the edge, the material, the number of pieces to be washed at one time, the mounting form in the basket, and the like. These are input to the information acquisition unit 330 from the outside of the cleaning system.

The information on the cleaning conditions generated by the machine learning unit 320 includes the oscillating ultrasonic frequency, output, oscillation time, liquid temperature, tank selection in the case of a multi-tank type, the position of the object to be cleaned 2 in the tank, and the swing distance. , Swing speed, etc. These are learning models 321 in which the detection results detected by the state detection unit 120 during cleaning, the appearance information of the object to be cleaned 2 to be observed, and the unique information of the object to be cleaned 2 acquired by the information acquisition unit 330 are input and updated. Is generated by the machine learning unit 320.

(How to clean the object to be cleaned)
A flow for cleaning the object to be cleaned 2 will be described with reference to FIGS. 1 and 5. The present invention relates to precision cleaning for removing fine stains on optical elements and precision parts including lenses made of glass or plastic as a raw material. When performing precision cleaning, ultrasonic cleaning is often used. Therefore, the cleaning method of the present invention will be described below using ultrasonic cleaning, which is particularly effective when performing precision cleaning. However, even when cleaning is performed by a method other than ultrasonic waves, the present invention is also effective in other cleaning methods, only the cleaning control items are different.

First, the uncleaned object 2 to be cleaned is transferred from the first storage unit 510 to the cleaning unit 110 of the cleaning device 10, and predetermined cleaning conditions are set by the cleaning condition setting unit 130 (step S001). When the image of the object to be cleaned 2 before cleaning is acquired, the uncleaned object 2 to be cleaned is transferred from the first storage unit 510 to the image pickup unit 220 of the image acquisition device 20 before step S001. , The appearance image of the object 2 to be cleaned is acquired. Next, the object to be cleaned 2 is washed under the set conditions (step S002). At this time, the state detection unit 120 detects the state of the cleaning unit 110. When the washing is completed, the washed object 2 to be washed is transferred to the image pickup unit 220 of the image acquisition device 20. Further, the state detection unit 120 inputs the detected detection result to the control unit 310 of the control device 30.

When the object to be cleaned 2 is transferred to the image pickup unit 220, the drive control unit 230 controls the image pickup unit 220 according to the drive conditions set by the control unit 310 of the control device 30, and an external image of the object to be cleaned 2 is obtained. (Step S003). When the image pickup is completed, the image output unit 210 inputs the acquired image information of the object to be cleaned 2 to the control unit 30. In addition, the object to be cleaned 2 from which the image has been acquired is transferred to the second accommodating portion 520.

Next, the control unit 310 inputs the acquired image information of the object to be cleaned 2 to the image evaluation unit 340, and the image evaluation unit 340 evaluates the state of the object to be cleaned 2 (step S004). Specifically, the types of defects such as stain residue, scratches, cloudiness, and sprinkling in the image information of the object to be cleaned 2 are analyzed, and the reward as an evaluation result is calculated based on the number, size, and position of the defects. .. The reward is given when the evaluation result is good, for example, when the defect has a small effect on the quality of the product 2 to be cleaned, when the degree of defect such as residual dirt, scratches, cloudiness, and sprinkling is small, and the like. Positive rewards are calculated according to the degree. On the contrary, when the evaluation result is not good, for example, when the defect has a large influence on the quality of the product 2 to be cleaned, when the degree of defect such as residual stain, scratch, cloudiness, and sprinkling is large, etc. Negative rewards are calculated according to the degree. Further, the appearance inspection may be performed using the image of the object to be cleaned 2. Information on detergency and defects can be evaluated on the image, and pass / fail determination (determination of OK or NG) of the object to be cleaned 2 as a product can be performed. The defect information analyzed by the image evaluation unit 340 and the calculated reward are included in the image information of the object to be cleaned 2, and the image information of the object to be cleaned 2 is input to the machine learning unit 320.

Further, the control unit 310 inputs the cleaning conditions set by the condition setting unit 130, the detection result detected by the state detection unit 120, and the unique information of the object to be cleaned 2 acquired by the information acquisition unit 330 to the machine learning unit 320. (Step S005). The unique information of the object to be cleaned 2 may be input once during the cleaning of one type of object 2 before the first type of object 2 to be cleaned performs step S005.

The machine learning unit 320 is the information obtained by cleaning the object to be cleaned 2 under the cleaning conditions set by the condition setting unit 130. The detection result detected by the state detection unit 120 and the subject acquired by the information acquisition unit 330. The unique information of the cleaning object 2 and the image information of the cleaning object 2 are learned. By learning, the learning model 321 is updated so that an appropriate cleaning condition having a high reward value can be calculated (step S006).

The machine learning unit 320 adds the unique information of the object to be cleaned 2, the image information of the object to be cleaned 2, the cleaning conditions, the detection result detected by the state detection unit, and the reward calculated by the image evaluation unit to the updated learning model 321. At least one of the above is input to generate a new cleaning condition (step S007).

After the new cleaning conditions are generated, it is determined whether or not the object to be cleaned 2 remains to be cleaned (step S008). When the object to be cleaned 2 remains, the process returns to step S001, and the object to be cleaned 2 is transferred to the cleaning unit 110 using the cleaning conditions generated by the machine learning unit 320. The object to be cleaned 2 is washed. When step S008 is reached again after cleaning, when the object to be cleaned 2 does not remain and the cleaning of all the objects to be cleaned 2 is completed, the series of cleaning of the objects to be cleaned 2 is completed.

By repeatedly updating the learning model 321 (step S006) by cleaning the object to be cleaned 2 in the above series, a learning model capable of calculating more appropriate cleaning conditions is constructed. A known algorithm can be used for such learning. For example, Q-learning, neural networks, and the like.

When the object to be cleaned 2 is continuously washed, the state of the cleaning liquid changes and the cleaning performance changes. In the case of ultrasonic cleaning, the effect of ultrasonic waves on the object to be cleaned 2 itself and the dirt adhering to the object to be cleaned 2 differs depending on the temperature of the cleaning liquid and the amount of dissolved gas. Therefore, if the cleaning is continuously performed under certain cleaning conditions, the cleaning property changes with the change in the state of the cleaning liquid, so that good cleaning cannot be continued. However, according to the present invention, it is possible to maintain the detergency and continue good cleaning by generating cleaning conditions according to the condition of the object to be cleaned 2 and the condition of the cleaning liquid and the like.

In the above flow, the learning model was updated (step 006) every time one object 2 to be cleaned was washed. However, when a learning model capable of calculating appropriate cleaning conditions is constructed, step 007 may be performed after step 005 without updating the learning model 321 (step S006).

Further, in the above flow, for the first object to be cleaned 2 of the type to be cleaned for the first time, predetermined cleaning conditions are set by the cleaning condition setting unit 130, and the second and subsequent objects to be cleaned 2 are set. On the other hand, cleaning was performed under the cleaning conditions generated by the machine learning unit 320. However, if the learning model 321 has already been repeatedly updated, even if the first object to be cleaned 2 of the type to be cleaned for the first time is cleaned under the cleaning conditions generated by the machine learning unit 320. good. In this case, the machine learning unit 320 generates cleaning conditions using the input unique information of the object to be cleaned 2 and the cleaning conditions of the similar object to be cleaned 2 stored in the storage unit 350 in the past. ..

The present invention supplies a program to the system via a network or storage medium that implements one or more of the functions of the embodiments described above. One or more processors in the computer of the system can realize the process of reading and executing a program. In this case, the program itself read from the recording medium realizes the function of the above-described embodiment, and the program itself and the recording medium on which the program is recorded constitute the present invention.

As a computer-readable recording medium for supplying a program, a flexible disk, a hard disk, an optical disk, a magneto-optical disk, a CD-ROM, a CD-R, a magnetic tape, a non-volatile memory card, a ROM, or the like is used. Can be done. Further, the program in the present embodiment may be downloaded via the net object to be cleaned 2 and executed by a computer.

Further, the function of the present embodiment is not limited to the realization by executing the program read by the computer. It is also included that the OS (operating system) or the like running on the computer performs a part or all of the actual processing based on the instruction of the program code, and the processing realizes the function of the above-described embodiment. ..

Further, the program read from the recording medium may be written in the memory provided in the function expansion board inserted in the computer or the function expansion unit connected to the computer. A case is also included in which the function expansion board, the CPU provided in the function expansion unit, or the like performs a part or all of the actual processing based on the instruction of the program, and the function of the present embodiment is realized by the processing. The present invention supplies a program that realizes one or more functions of the above-described embodiment to a system or apparatus via a net object to be cleaned 2 or a storage medium. One or more processors in the computer of the system can realize the process of reading and executing a program. Further, the control of the present invention can also be realized by a hardware circuit (for example, ASIC) that realizes one or more functions.

(Other embodiments)
The present invention is not limited to the embodiments described above, and many modifications can be made within the technical idea of the present invention. In addition, the effects described in the embodiments of the present invention merely list suitable effects arising from the present invention, and the effects according to the present invention are not limited to those described in the embodiments of the present invention.

In addition to the above ultrasonic cleaning method, there is a method of removing stains by utilizing the physical force of a brush or the like and the chemical force of a cleaning agent. Cleaning parameters for brush cleaning include the applied force, the number of rubbing times, and the brush shape. In the case of chemical cleaning, there are properties such as liquid concentration and solubility. However, in either case, if the force acting on the removal is too large, the object to be cleaned 2 may be scratched or damaged by chemical deterioration or the like.

Regardless of which cleaning method is used, it is necessary to inspect the object to be cleaned 2 in order to confirm the detergency and the presence or absence of damage after cleaning. Especially in ultrasonic cleaning, damage is often seen due to the cleaning strength being too strong. For example, in glass lens cleaning, if the glass hardness is low, scratches called latent scratches due to cavitation or scratches may occur. There is. When an alkaline cleaning agent is used, the glass surface may be chemically altered to cause fogging.

In order to prevent them, cleaning conditions must be set so that there is no residual stain and no damage to the object to be cleaned 2, but good cleaning conditions are the material of the object to be cleaned 2, the number of items to be cleaned, and the object to be cleaned 2. It depends on the factors of the object to be cleaned 2, such as the shape and the like. Further, the range of suitable cleaning conditions varies depending on the type and degree of dirt adhering to the object to be cleaned 2.

(Manufacturing method of goods)
The cleaning system according to the embodiment described above can also be used in a method for manufacturing an article. The method for producing an article of the present embodiment includes a step of cleaning the object to be cleaned 2 using the cleaning system described above, and a step of performing a predetermined treatment on the washed object 2 to be cleaned. The predetermined process is, for example, at least one selected from processing, cutting, transporting, assembling, and inspection.

1 Cleaning system 10 Cleaning device 110 Cleaning unit 120 State detection unit 130 Condition setting unit 20 Image acquisition device 210 Image output unit 220 Imaging unit 230 Drive control unit 30 Control device 310 Control unit 320 Machine learning unit 321 Learning model 330 Information acquisition unit 340 Image evaluation unit 350 Storage unit 40 Information input unit 510 1st storage unit 520 2nd storage unit

Claims (20)

A cleaning device having a cleaning unit for cleaning the object to be cleaned,
An image acquisition device that acquires an image of the object to be cleaned before and / or after cleaning.
An image evaluation unit that evaluates the state of the object to be cleaned based on the type of defect of the object to be cleaned from the image information.
A control unit that controls the operation of the cleaning device,
A machine learning unit that generates cleaning conditions for the cleaning device based on the image acquired by the image acquisition device and the result of the image evaluation unit evaluating the state of the object to be cleaned.
Equipped with
The control unit is a cleaning system characterized in that the operation of the cleaning device is controlled by using the cleaning conditions generated by the machine learning unit. The cleaning system according to claim 1, wherein the type of defect evaluated by the image evaluation unit from the image information includes at least one of dirt, scratches, cloudiness, and sprinkling. The cleaning system according to claim 1 or 2, wherein the image evaluation unit further evaluates the state of the object to be cleaned based on the position of a defect of the object to be cleaned from the image information. The image evaluation unit calculates a reward based on the result of evaluating the state of the object to be washed.
The cleaning system according to any one of claims 1 to 3, wherein the machine learning unit generates the cleaning conditions of the cleaning device based on the reward calculated by the image evaluation unit. The cleaning system according to claim 4, wherein the image evaluation unit calculates a positive reward when the degree of defect is small as a result of evaluating the state of the object to be cleaned. The cleaning system according to claim 4, wherein the image evaluation unit calculates a negative reward when the degree of defect is large as a result of evaluating the state of the object to be cleaned. The cleaning system according to any one of claims 1 to 6, wherein the machine learning unit further adds cleaning conditions when the object to be cleaned is added to generate cleaning conditions for the cleaning device. The cleaning device has a state detection unit that detects the state of the cleaning unit.
The cleaning system according to any one of claims 1 to 7, wherein the machine learning unit generates the cleaning conditions of the cleaning device based on the state of the cleaning unit detected by the state detection unit. The cleaning system according to any one of claims 1 to 8, wherein the machine learning unit generates cleaning conditions for the cleaning device based on the unique information of the object to be cleaned. An information acquisition unit for acquiring the unique information of the object to be cleaned is provided.
The cleaning system according to claim 9, wherein the unique information of the object to be cleaned is acquired by the information acquisition unit. The cleaning system according to claim 9 or 10, wherein the control unit further controls the operation of the image acquisition device based on the unique information of the object to be cleaned. The machine learning unit further has a learning model.
The machine learning unit evaluates the image information acquired by the image acquisition device, the image evaluation unit evaluates the state of the object to be cleaned, the reward calculated by the image evaluation unit, and the cleaning conditions of the cleaning device. 1. The cleaning system according to item 1. The cleaning system according to any one of claims 1 to 12, wherein the cleaning unit has an ultrasonic vibrator. The cleaning system according to any one of claims 1 to 13, wherein the object to be cleaned is a lens. The object to be cleaned is washed according to the image information of the object to be cleaned obtained before and / or after cleaning and the result of evaluating the state of the object to be cleaned based on the type of defect from the image information. Equipped with a machine learning unit, which produces cleaning conditions,
A control device that controls cleaning of the object to be cleaned based on the cleaning conditions generated by the machine learning unit. A cleaning device having a cleaning unit for cleaning the object to be cleaned,
An image acquisition device that acquires image information of the object to be cleaned before and / or after cleaning.
An image evaluation unit that evaluates the state of the object to be cleaned based on the type of defect from the image information,
A control unit that controls the operation of the cleaning device,
A machine learning unit that generates cleaning conditions for the cleaning device based on the image information acquired by the image acquisition device and the result of the image evaluation unit evaluating the state of the object to be cleaned.
With a cleaning system equipped with
A cleaning method comprising controlling the cleaning of the object to be cleaned based on the cleaning conditions generated by the machine learning unit. Image information of the object to be washed acquired before and / or after washing,
As a result of evaluating the state of the object to be cleaned based on the type of defect from the image information,
A control method for controlling the cleaning of the object to be cleaned, which generates a cleaning condition for cleaning the object to be cleaned based on the above. A program for causing a computer to execute the cleaning method according to claim 16 or the control method according to 17. A computer-readable recording medium on which the program according to claim 18 is recorded. A step of cleaning an object to be cleaned using the cleaning system according to any one of claims 1 to 14.
A step of performing a predetermined treatment on the washed object to be washed, and
A method of manufacturing an article comprising.

Images