Abnormal Detection Device and Abnormal Detection Method

Abstract

To accurately detect an abnormality.SOLUTION: A first generation unit generates pseudo abnormal image data in which an approximately circular image is superposed on a random position of an image of normal image data obtained by photographing a facility that includes a fluid supply unit, the facility to normally supply fluid from the fluid supply unit. A second generation unit generates a determination model for determining normality and abnormality of the facility by learning normal image data and pseudo abnormal image data. An acquisition unit acquires image data obtained by photographing the facility. A detection unit detects an abnormality of the facility, by using the determination model, from an image acquired by the acquisition unit.SELECTED DRAWING: Figure 3

Description

Technical Field

[0001] The present disclosure relates to an abnormality detection device and an abnormality detection method.

Background Art

[0002] Patent Document 1 discloses a method of generating a change inspection area image in which a feature amount is changed from an image of an inspection area of an instruction image having an inspection area, and generating a new defect addition instruction image in which the change inspection area image is arranged in the instruction image.

Prior Art Documents

Patent Documents

[0003]

Patent Document 1

Summary of the Invention

Problems to be Solved by the Invention

[0004] The present disclosure provides a technique for accurately detecting an abnormality.

Means for Solving the Problems

[0005] An abnormality detection device according to an aspect of the present disclosure includes a first generation unit, a second generation unit, an acquisition unit, and a detection unit. The first generation unit generates pseudo-abnormality image data in which a substantially circular image is synthesized at a random position of an image of normal image data obtained by photographing a facility including a liquid supply unit and supplying liquid normally from the liquid supply unit. The second generation unit performs learning on the normal image data and the pseudo-abnormality image data, and generates a determination model for determining normality and abnormality of the facility. The acquisition unit acquires image data obtained by photographing the facility. The detection unit detects an abnormality of the facility from the image data acquired by the acquisition unit using the determination model.

Effects of the Invention

[0006] According to the present disclosure, an abnormality can be accurately detected.

Brief Description of the Drawings

[0007]

Figure 1

Figure 2

Figure 3

Figure 4

Figure 5

Figure 6

Figure 7

MODE FOR CARRYING OUT THE INVENTION

[0008] Hereinafter, embodiments of the abnormality detection device and the abnormality detection method disclosed in the present application will be described in detail with reference to the drawings. Note that the abnormality detection device and the abnormality detection method disclosed are not limited by this embodiment.

[0009] By the way, the detection of anomalies using learning models such as AI (Artificial Intelligence) is being considered. In the detection of anomalies using this learning model, when only normal image data when the monitoring target is operating normally is learned by the learning model, there are cases where anomalies in a narrow range cannot be detected accurately. Therefore, it is conceivable to also let the learning model learn abnormal image data when an anomaly occurs in the monitoring target. However, in reality, there are various locations where anomalies occur in the monitoring target. Also, there are cases where anomalies rarely occur in the monitoring target. For this reason, it may be difficult to obtain sufficient abnormal image data required for learning from the abnormal image data in which an actual anomaly has occurred in the monitoring target. As a result, in the detection of anomalies using a learning model, there are cases where anomalies cannot be detected accurately.

[0010] Therefore, a technique that can accurately detect anomalies is expected.

[0011] [Embodiment] [Monitoring Target] The embodiment will be described. Hereinafter, the case where an anomaly detection device detects the occurrence of an anomaly in a monitoring target will be described as an example. Also, hereinafter, the case where the monitoring target is a coating unit that discharges a liquid will be described as an example. FIG. 1 is a schematic diagram showing an example of a monitoring target according to the embodiment. In FIG. 1, a coating unit 10 that discharges a liquid is shown as the monitoring target. The coating unit 10 is an example of a facility equipped with a liquid supply unit. The coating unit 10 is provided in a liquid processing device such as a coating device or a developing device that uses a liquid such as resist, pure water, or a solvent.

[0012] FIG. 1 shows a target surface S onto which liquid is to be discharged. The target surface S is, for example, a surface for applying liquid on the upper surface of a substrate such as a semiconductor wafer. The coating unit 10 is disposed around the target surface S. The coating unit 10 includes an arm 11 and a support portion 12. The arm 11 is supported by the support portion 12. The arm 11 is arranged such that its tip 11a is positioned above the target surface S. A nozzle 13, which is a liquid discharge port, is provided at the tip 11a of the arm 11. The nozzle 13 is connected via a pipe (not shown) provided inside the arm 11 and the support portion 12 to a supply system (not shown) including a liquid supply source, a valve, and a metering pump. The coating unit 10 discharges the liquid supplied from the supply system from the nozzle 13. Note that the coating unit 10 may be provided with a moving mechanism and be configured to move the arm 11 in the horizontal and vertical directions. Further, the coating unit 10 may be configured to be movable between a coating position where the tip 11a of the arm 11 is on the target surface S by rotating the support portion 12 and a standby position outside the target surface S.

[0013] A camera 20 is disposed around the coating unit 10. The camera 20 is arranged such that the arm 11, the nozzle 13, and the target surface S are within its field of view, and is capable of taking an image of the arm 11 and the nozzle 13 in a side view. The camera 20 captures the status of the arm 11 having the nozzle 13 and the target surface S during the coating process in which the coating unit 10 discharges liquid onto the target surface S at a predetermined frame rate (for example, 30 fps).

[0014] FIG. 2 is a diagram showing an example of an image obtained by photographing a monitoring target according to an embodiment. FIG. 2 shows an example of an image captured by the camera 20. In the image shown in FIG. 2, the arm 11, the nozzle 13, and the target surface S are depicted. In FIG. 2, liquid is being discharged from the nozzle 13 and applied to the target surface S.

[0015] The image data of the image captured by the camera 20 is output to an abnormality detection device, and abnormality detection is performed.

[0016] [Configuration of Abnormality Detection Device] Next, the anomaly detection device will be described in detail. FIG. 3 is a block diagram showing a schematic configuration of an anomaly detection device 50 according to an embodiment. The anomaly detection device 50 is, for example, a computer such as a personal computer or a server computer. The anomaly detection device 50 includes an external I / F (interface) unit 51, a display unit 52, an input unit 53, a storage unit 54, and a control unit 55. Note that the anomaly detection device 50 may have various functional units that a known computer has in addition to the functional units shown in FIG. 3.

[0017] The external I / F unit 51 is an interface for inputting and outputting information with other devices. For example, the external I / F unit 51 is a communication interface such as a USB (Universal Serial Bus) port or a LAN port. The external I / F unit 51 receives the image data of the image captured by the camera 20.

[0018] The display unit 52 is a display device for displaying various information. Examples of the display unit 52 include display devices such as an LCD (Liquid Crystal Display) and a CRT (Cathode Ray Tube). The display unit 52 displays various information.

[0019] The input unit 53 is an input device for inputting various information. For example, examples of the input unit 53 include input devices such as a mouse and a keyboard. The input unit 53 receives an operation input from an administrator or the like and inputs operation information indicating the received operation content to the control unit 55.

[0020] The storage unit 54 is a storage device for storing various data. For example, the storage unit 54 is a storage device such as a hard disk, an SSD (Solid State Drive), or an optical disk. Note that the storage unit 54 may be a semiconductor memory such as a RAM (Random Access Memory), a flash memory, or an NVSRAM (Non Volatile Static Random Access Memory) in which data can be rewritten.

[0021] The storage unit 54 stores the OS (Operating System) and various programs executed by the control unit 55. For example, the storage unit 54 stores various programs including a program for executing the generation process and the abnormality detection process described later. Further, the storage unit 54 stores various data used in the programs executed by the control unit 55. For example, the storage unit 54 stores the model generation data 60 and the model data 61. Note that the storage unit 54 can also store other data in addition to the data exemplified above.

[0022] The model generation data 60 is data used for generating a determination model described later. The model generation data 60 includes various data used for generating the determination model. For example, the model generation data 60 includes a plurality of normal image data 60a, a plurality of actual abnormal image data 60b, and a plurality of pseudo-abnormal image data 60c.

[0023] The normal image data 60a is image data of an image obtained by photographing a monitoring target that is operating normally. For example, the normal image data 60a is image data of an image in which the coating unit 10 discharges a liquid without causing abnormalities such as liquid dripping or liquid leakage. Here, "liquid dripping" refers to a state in which liquid droplets hang from the nozzle 13 or the arm 11. Also, "liquid leakage" refers to a state in which liquid drips from the nozzle 13 or the arm 11. For example, a plurality of image data obtained by photographing a series of operations in which the coating unit 10 discharges a liquid without causing an abnormality with the camera 20 are each stored as the normal image data 60a.

[0024] The actual abnormal image data 60b is image data of an image obtained by photographing a monitoring target in which an abnormality has occurred. For example, the actual abnormal image data 60b is image data of an image when the coating unit 10 causes abnormalities such as liquid dripping or liquid leakage. For example, image data obtained by photographing with the camera 20 when an abnormality occurs in the coating unit 10 are each stored as the actual abnormal image data 60b.

[0025] The suspected abnormal image data 60c is image data that pseudo-generates an image when an abnormality occurs. Details of the suspected abnormal image data 60c will be described later.

[0026] The model data 61 is data that stores a determination model generated using a learning model such as AI.

[0027] The control unit 55 is a device that controls the abnormality detection device 50. As the control unit 55, an electronic circuit such as a CPU (Central Processing Unit) or an MPU (Micro Processing Unit), or an integrated circuit such as an ASIC (Application Specific Integrated Circuit) or an FPGA (Field Programmable Gate Array) can be adopted. The control unit 55 has an internal memory for storing programs and control data that define various processing procedures, and executes various processes based on these. The control unit 55 functions as various processing units when various programs operate. For example, the control unit 55 includes a first generation unit 70, a reception unit 71, a second generation unit 72, an acquisition unit 73, a detection unit 74, and an output unit 75.

[0028] By the way, in the detection of anomalies using a learning model, if only normal image data 60a is used for learning by the learning model, there are cases where anomalies in a narrow range cannot be accurately detected. Therefore, it is conceivable to also use the actual abnormal image data 60b when an anomaly occurs in the monitoring target for learning by the learning model. However, in reality, there are various locations where anomalies occur in the monitoring target, and there are also cases where anomalies rarely occur in the monitoring target. For example, the coating unit 10 may experience anomalies such as liquid dripping or dropping from the nozzle 13. In addition, the coating unit 10 may experience anomalies such as liquid dripping or dropping from the arm 11 due to liquid leakage from the pipes inside the arm 11. The coating unit 10 is designed so as not to cause anomalies such as liquid dripping or dropping. For this reason, it is rare for the coating unit 10 to experience anomalies. For example, cases where anomalies occur in the arm 11 are rare. For this reason, it may be difficult to obtain abnormal image data in which an actual anomaly has occurred in the monitoring target. For example, it is difficult to obtain image data in which anomalies such as liquid dripping or dropping are actually occurring at various locations on the arm 11. As a result, in the detection of anomalies using a learning model, there are cases where anomalies cannot be accurately detected.

[0029] Therefore, the first generation unit 70 generates pseudo-abnormal image data 60c by modifying a part of the image of the normal image data 60a. For example, the first generation unit 70 generates pseudo-abnormal image data 60c by synthesizing a substantially circular image at a random position in the image of the normal image data 60a. Substantially circular includes circular and elliptical. As an example, the first generation unit 70 performs random erasing on the image of the normal image data 60a and generates pseudo-abnormal image data 60c by synthesizing a substantially circular image at a random position. Here, random erasing is a technique for usually increasing teacher data that is normal by generating an image in which a random partial rectangular region of the image serving as teacher data is masked for the purpose of improving the robustness of the determination. On the other hand, in the present embodiment, random erasing is used to synthesize a substantially circular image having a shape similar to anomalies such as liquid dripping or dropping on the image of the normal image data 60a to generate pseudo-abnormal pseudo-abnormal image data 60c.

[0030] The first generation unit 70 generates a plurality of pseudo-abnormal image data with a part of the image changed for each of a plurality of normal image data obtained by photographing a series of operations of a monitoring target operating normally. For example, the first generation unit 70 generates a plurality of pseudo-abnormal image data 60c with a part of the image changed for each of a plurality of normal image data 60a obtained by photographing a series of operations in which the coating unit 10 discharges liquid without generating an abnormality. For example, in the present embodiment, a plurality of image data photographed by the camera 20 for 10 times of a series of operations in which the coating unit 10 discharges liquid without generating an abnormality are stored as normal image data 60a respectively. The first generation unit 70 generates pseudo-abnormal image data 60c in which a substantially circular image is synthesized at one random position in the image of one normal image data 60a with respect to the normal image data 60a of a series of operations for one time out of 10 times.

[0031] Here, for the monitoring target, there may be a case where the location where an abnormality occurs is biased within a specific range. Also, there may be a case where the size of the abnormal portion in the image is biased to be equal to or less than a specific size. For example, the coating unit 10 generates abnormalities such as liquid dripping and dripping from the nozzle 13 or the arm 11. Also, liquid dripping and dripping occur in a small and narrow range.

[0032] Therefore, the reception unit 71 receives the specification of the range for synthesizing the image and the size of the image to be synthesized. For example, the reception unit 71 causes the display unit 52 to display a screen for specifying the range for synthesizing the image and the size of the image to be synthesized, and receives the specification of the range for synthesizing the image and the size of the image to be synthesized from the input unit 53.

[0033] The first generation unit 70 generates pseudo-abnormal image data 60c by synthesizing a substantially circular image at a random position within the specified range and having a size equal to or smaller than the size specified by the reception unit 71. For example, when detecting liquid dripping or bottle dropping from the nozzle 13 or the arm 11 as an abnormality, the administrator specifies the range including the nozzle 13 or the arm 11 and the maximum size of the liquid droplets of the liquid dripping or bottle dropping. The first generation unit 70 generates pseudo-abnormal image data 60c by synthesizing a substantially circular image having a size equal to or smaller than the specified size at a random position within the specified range including the specified nozzle 13 or arm 11.

[0034] FIG. 4 is a diagram showing an example of the pseudo-abnormal image data 60c according to the embodiment. FIG. 4 shows a pseudo-abnormal image in which a substantially circular image 80 in white (white-outlined) is synthesized with the image shown in FIG. 2. In FIG. 4, a substantially circular image 80 in white is synthesized near the lower surface of the arm 11.

[0035] Note that the color of the image 80 to be synthesized is not limited to white. The color of the image 80 to be synthesized may be determined in advance. Also, the color of the image 80 to be synthesized may be determined from the normal image data 60a. For example, the first generation unit 70 may generate pseudo-abnormal image data 60c by synthesizing a substantially circular image 80 having a luminance equal to a predetermined ratio (e.g., 20%) of the luminance values from the top when the luminance values of the pixels of the image of the normal image data 60a are arranged in descending order. Thus, by determining the color of the image 80 to be synthesized from the normal image data 60a, the image 80 can be synthesized with a color similar to the actual image. Thus, by synthesizing the image 80 with a color similar to the actual image, an abnormality that actually occurs can be accurately detected.

[0036] The second generation unit 72 performs learning on a plurality of normal image data 60a, a plurality of actual abnormal image data 60b, and a plurality of pseudo-abnormal image data 60c, and generates a determination model for determining whether the monitoring target is normal or abnormal. For example, the second generation unit 72 attaches additional information indicating normality to the plurality of normal image data 60a, and attaches additional information indicating abnormality to the plurality of actual abnormal image data 60b and the plurality of pseudo-abnormal image data 60c. For example, the determination model is a model that outputs image data when the image data is input. The second generation unit 72 performs learning, for example, by Opposite Learning, to increase the reproducibility of normal image data and decrease the reproducibility of abnormal image data, and generates a determination model that outputs image data with high reproducibility for normal image data. The determination model can learn rare abnormalities that actually occur from the pseudo-abnormal image data 60c by performing learning on the pseudo-abnormal image data 60c. Therefore, the determination accuracy of the determination model is improved. Note that the second generation unit 72 may generate a determination model by performing learning on the plurality of normal image data 60a and the plurality of pseudo-abnormal image data 60c without using the plurality of actual abnormal image data 60b. However, by generating a determination model that has learned the actual abnormal image data 60b in addition to the normal image data 60a and the pseudo-abnormal image data 60c, the determination model can increase the difference in reproducibility between normal data and abnormal data, and the determination accuracy is improved.

[0037] The second generation unit 72 stores the data of the generated determination model in the storage unit 54 as model data 61.

[0038] The acquisition unit 73 acquires image data obtained by photographing the monitoring target. For example, the acquisition unit 73 acquires the image data output from the camera 20 via the external I / F unit 51.

[0039] The detection unit 74 detects an abnormality of a monitoring target from the image data acquired by the acquisition unit 73 using the determination model of the model data 61. FIG. 5 is a diagram schematically showing an example of abnormality detection according to the embodiment. For example, the detection unit 74 inputs the acquired image data into the determination model. The determination model outputs the image data. When the output image data is normal image data, the reproducibility of the input image data is high, and when it is abnormal image data, the reproducibility is low. The detection unit 74 compares the output image data with the input image data to obtain a difference. For example, the detection unit 74 obtains the change in the pixel value of each pixel as the difference. The detection unit 74 determines that it is normal when the obtained difference is within a predetermined threshold value, and determines that it is abnormal when the difference exceeds the threshold value, thereby detecting an abnormality.

[0040] Note that the determination model may be the following model. For example, the second generation unit 72 attaches additional information indicating normality to a plurality of normal image data 60a, and attaches additional information indicating abnormality to a plurality of actual abnormal image data 60b and a plurality of pseudo-abnormal image data 60c. Then, the second generation unit 72 performs machine learning such as deep learning using the plurality of normal image data 60a, the plurality of actual abnormal image data 60b, and the plurality of pseudo-abnormal image data 60c, and generates a determination model that determines and outputs whether the input image data is normal or abnormal. The detection unit 74 inputs the acquired image data into the determination model. The determination model outputs whether the input image data is normal or abnormal. The detection unit 74 detects an abnormality using the normal and abnormal output results of the determination model. Also in this case, the second generation unit 72 may perform learning of the plurality of normal image data 60a and the plurality of pseudo-abnormal image data 60c without using the plurality of actual abnormal image data 60b to generate a determination model. However, by generating a determination model learned from the actual abnormal image data 60b in addition to the normal image data 60a and the pseudo-abnormal image data 60c, the generated determination model improves the determination accuracy of normality and abnormality.

[0041] By the way, noise or the like may temporarily occur in the image data captured by the camera 20, and it may be detected that an abnormality has occurred in the coating unit 10 even though the coating unit 10 is normal. When a series of operations in which the coating unit 10 applies a liquid are photographed by the camera 20, abnormalities such as dripping and splashing are detected in a plurality of pieces of image data. Therefore, when the detection unit 74 determines that there is an abnormality in a predetermined number of continuously photographed image data, the detection unit 74 may detect that an abnormality has occurred in the monitoring target. For example, when the detection unit 74 determines that there is an abnormality in three continuously photographed image data, the detection unit 74 detects that an abnormality has occurred in the coating unit 10. Thereby, the determination accuracy of the abnormality is improved.

[0042] The output unit 75 outputs the detection result of the detection unit 74. For example, when an abnormality is detected by the detection unit 74, the output unit 75 outputs a message indicating that an abnormality has occurred to the display unit 52. Note that the output unit 75 may output the data of the determination result of the detection unit 74 to another device. For example, when the detection unit 74 determines that there is an abnormality, the output unit 75 may output data indicating that an abnormality has occurred to a management device that manages the device on which the coating unit 10 is mounted.

[0043] [Flow of processing] Next, the flow of various processes performed by the abnormality detection device 50 according to the embodiment will be described. First, the flow of the generation process in which the abnormality detection device 50 according to the embodiment generates a determination model will be described. FIG. 6 is a flowchart showing an example of the flow of the generation process according to the embodiment.

[0044] The first generation unit 70 generates pseudo-abnormal image data 60c in which a part of the image of the normal image data 60a is changed (step S10). For example, the first generation unit 70 generates pseudo-abnormal image data 60c in which a substantially circular image is synthesized at a random position in the image of the normal image data 60a.

[0045] The second generation unit 72 performs learning on the normal image data 60a, the actual abnormal image data 60b, and the pseudo-abnormal image data 60c, and generates a determination model for determining the normality / abnormality of the monitoring target (step S11). The second generation unit 72 stores the data of the generated determination model in the storage unit 54 as model data 61 (step S12), and ends the process.

[0046] Next, the flow of the abnormality detection process in which the abnormality detection device 50 according to the embodiment detects an abnormality using the determination model will be described. FIG. 7 is a flowchart showing an example of the flow of the abnormality detection process according to the embodiment.

[0047] The acquisition unit 73 acquires image data obtained by photographing the monitoring target (step S20). For example, the acquisition unit 73 acquires the image data output from the camera 20 via the external I / F unit 51.

[0048] The detection unit 74 detects an abnormality of the monitoring target from the image data acquired by the acquisition unit 73 using the determination model of the model data 61 (step S21).

[0049] The output unit 75 outputs the detection result of the detection unit 74 (step S22). The process ends.

[0050] As described above, the abnormality detection device 50 according to the embodiment includes a first generation unit 70, a second generation unit 72, an acquisition unit 73, and a detection unit 74. The first generation unit 70 generates pseudo-abnormal image data 60c in which a substantially circular image is synthesized at a random position of an image of the normal image data 60a obtained by photographing a facility (coating unit 10) that normally supplies liquid from a liquid supply unit (nozzle 13). The second generation unit 72 performs learning on the normal image data 60a and the pseudo-abnormal image data 60c, and generates a determination model for determining the normality / abnormality of the facility. The acquisition unit 73 acquires image data obtained by photographing the facility. The detection unit 74 detects an abnormality of the facility from the image data acquired by the acquisition unit 73 using the determination model. As a result, the abnormality detection device 50 can learn an abnormality that rarely occurs in reality from the pseudo-abnormal image data 60c by the determination model, and thus can accurately detect the abnormality.

[0051] Further, in addition to the normal image data 60a and the pseudo-abnormal image data 60c, the second generation unit 72 performs learning on the actual abnormal image data 60b obtained by photographing the facility (coating unit 10) in which an abnormality has occurred, and generates a determination model. As a result, since the abnormality detection device 50 can learn the actually occurring abnormality from the actual abnormal image data 60b by the determination model, the abnormality can be detected with higher accuracy.

[0052] Also, the facility (coating unit 10) is an arm 11 provided with a nozzle 13 for discharging a liquid or a pipe. The first generation unit 70 generates pseudo-abnormal image data 60c by synthesizing a substantially circular image 80 at a random position in the image of the normal image data 60a. As a result, the abnormality detection device 50 can detect abnormalities such as dripping and dropping more accurately.

[0053] The abnormality detection device 50 further includes a reception unit 71. The reception unit 71 receives a designation of a range for synthesizing an image and / or a size of the image to be synthesized. The first generation unit 70 generates pseudo-abnormal image data 60c by synthesizing a substantially circular image at a random position within the designated range and having a size equal to or smaller than the size designated by the reception unit 71. As a result, the abnormality detection device 50 can accurately detect an abnormality by appropriately designating the range for synthesizing an image and the size of the image to be synthesized according to the abnormality to be detected.

[0054] The first generation unit 70 generates pseudo-abnormal image data 60c by synthesizing a substantially circular image having a luminance value of a predetermined ratio from the top when the luminance values of the pixels of the image of the normal image data 60a are arranged in descending order. As a result, the abnormality detection device 50 can accurately detect the actually occurring abnormality.

[0055] In addition, the first generation unit 70 generates a plurality of pseudo-abnormal image data 60c in which a substantially circular image is synthesized at random positions of the images for each of a plurality of normal image data 60a obtained by photographing a series of operations of equipment (coating unit 10) that operates normally. The second generation unit 72 performs learning on the plurality of normal image data 60a and the plurality of pseudo-abnormal image data 60c, and generates a determination model for determining normality and abnormality of the equipment. The acquisition unit 73 acquires a plurality of image data obtained by photographing a series of operations of the equipment. The detection unit 74 determines the normality and abnormality of the plurality of image data acquired by the acquisition unit 73 using the determination model, and when any of the image data is determined to be abnormal, detects that an abnormality has occurred in the equipment. Thereby, since the abnormality detection device 50 can cause the determination model to learn a series of operations of the equipment, it can accurately detect an abnormality that has occurred during a series of operations of the equipment.

[0056] In addition, when the detection unit 74 determines that there is an abnormality in a predetermined number of continuously photographed image data, it detects that an abnormality has occurred in the equipment. Thereby, the abnormality detection device 50 improves the determination accuracy of the abnormality.

[0057] As described above, the embodiments have been described. However, the embodiments disclosed this time should be considered as illustrative in all respects and not restrictive. In fact, the above-described embodiments can be embodied in various forms. Also, the above-described embodiments may be omitted, replaced, or changed in various forms without departing from the scope of the claims and their gist.

[0058] For example, in the above-described embodiment, the case where the equipment to be monitored is the coating unit 10 has been described as an example. However, it is not limited to this. The equipment may be any as long as it includes a liquid supply unit that discharges (supplies) a liquid.

[0059] Also, in the above-described embodiment, the case of detecting abnormalities such as liquid dripping and bottle dropping has been described as an example. However, it is not limited to this. The abnormality to be detected may be any.

[0060] In the above-described embodiment, the case where the determination model is generated in the abnormality detection device 50 has been described as an example. However, the present invention is not limited to this. The determination model may be generated by another device and stored in the storage unit 54 of the abnormality detection device 50. The abnormality detection device 50 may detect an abnormality in the facility from the acquired image data using the determination model generated by another device and stored in the storage unit 54.

[0061] It should be considered that the embodiments disclosed this time are illustrative in all respects and not restrictive. In fact, the above-described embodiments can be embodied in various forms. Further, the above embodiments may be omitted, replaced, or changed in various forms without departing from the scope and spirit of the appended claims.

Explanation of Reference Numerals

[0062] 10 Coating Unit 11 Arm 12 Support Part 13 Nozzle 20 Camera 50 Abnormality Detection Device 51 External I / F 52 Display Unit 53 Input Unit 54 Storage Unit 55 Control Unit 60 Model Generation Data 60a Normal Image Data 60b Actual Abnormality Image Data 60c Suspected Abnormality Image Data 61 Model Data 70 First Generation Unit 71 Reception Unit 72 Second Generation Unit 73 Acquisition Unit 74 Detection Unit 75 Output Unit S Target Surface

Claims

1. An equipment provided with a liquid supply unit, wherein a first generation unit generates pseudo-abnormal image data by synthesizing a substantially circular image with luminance values of pixels of an image of the normal image data arranged in descending order of luminance at random positions within a specified range of the image of the normal image data obtained by photographing the equipment where no liquid dripping or bottle dropping occurs; a second generation unit that learns the normal image data and the pseudo-abnormal image data and generates a determination model for determining whether the equipment is normal or abnormal; an acquisition unit that acquires image data obtained by photographing the equipment; a detection unit that detects an abnormality of the equipment from the image data acquired by the acquisition unit using the determination model; An abnormality detection device having the above.

2. The second generation unit generates the determination model by learning, in addition to the normal image data and the pseudo-abnormal image data, actual abnormal image data obtained by photographing the equipment where an abnormality has occurred. The abnormality detection device according to Claim 1.

3. The liquid supply unit comprises an arm provided with a nozzle or a pipe for discharging liquid. The abnormality detection device according to Claim 1 or 2.

4. It further has a reception unit that receives designation of a range for synthesizing an image and a size of the image to be synthesized, and the first generation unit generates pseudo-abnormal image data by synthesizing the substantially circular image at random positions within the specified range and with a size equal to or smaller than the size specified by the reception unit. The abnormality detection device according to any one of Claims 1 to 3.

5. The first generation unit generates a plurality of pseudo-abnormal image data by synthesizing the substantially circular image at random positions of an image for each of a plurality of normal image data obtained by photographing a series of operations of the equipment operating normally, the second generation unit generates a determination model for determining whether the equipment is normal or abnormal by learning the plurality of normal image data and the plurality of pseudo-abnormal image data, the acquisition unit acquires a plurality of image data obtained by photographing a series of operations of the equipment, and the detection unit determines whether the plurality of image data acquired by the acquisition unit is normal or abnormal using the determination model, and if any of the image data is determined to be abnormal, it detects that an abnormality has occurred in the equipment. The abnormality detection device according to any one of Claims 1 to 4.

6. If the detection unit determines that an abnormality has occurred in a predetermined number of continuously photographed image data, it detects that an abnormality has occurred in the equipment. The abnormality detection device according to Claim 5.

7. Equipment provided with a liquid supply unit, a normal image data obtained by photographing the equipment without liquid dripping and bottle dropping occurring, and a substantially circular image obtained by synthesizing luminance values of pixels of the image of the normal image data as luminance in descending order of luminance values at random positions within a designated range of the image of the normal image data. A determination model generated by performing learning of the pseudo-abnormal image data, a storage unit that stores the determination model for determining normality and abnormality of the equipment, An acquisition unit that acquires image data obtained by photographing the equipment, A detection unit that detects an abnormality of the equipment from the image data acquired by the acquisition unit using the determination model, An abnormality detection device having the above components.

8. The liquid supply unit comprises an arm provided with a nozzle or a pipe for discharging liquid The abnormality detection device according to claim 7.

9. Equipment provided with a liquid supply unit, a first generation unit that generates pseudo-abnormal image data by synthesizing a substantially circular image with luminance values of a predetermined ratio from the top when the luminance values of pixels of the image of the normal image data are arranged in descending order at random positions of the normal image data obtained by photographing the equipment without liquid dripping and bottle dropping occurring, A second generation unit that performs learning of the normal image data and the pseudo-abnormal image data and generates a determination model for determining normality and abnormality of the equipment, An acquisition unit that acquires image data obtained by photographing the equipment, A detection unit that detects an abnormality of the equipment from the image data acquired by the acquisition unit using the determination model, and the first generation unit generates a plurality of pseudo-abnormal image data by synthesizing the substantially circular image at random positions of the image for each of a plurality of normal image data obtained by photographing a series of operations of the equipment operating normally, The second generation unit performs learning of the plurality of normal image data and the plurality of pseudo-abnormal image data and generates a determination model for determining normality and abnormality of the equipment, The acquisition unit acquires a plurality of image data obtained by photographing a series of operations of the equipment, The detection unit determines normality and abnormality of the plurality of image data acquired by the acquisition unit using the determination model, and when any of the image data is determined to be abnormal, detects that an abnormality has occurred in the equipment An abnormality detection device.

10. When the detection unit determines that there is an abnormality in a predetermined number of continuously photographed image data, it detects that an abnormality has occurred in the equipment The abnormality detection device according to claim 9. ​

11. A facility equipped with a liquid supply unit, the step of generating pseudo-abnormal image data by synthesizing a substantially circular image with luminance values of a predetermined percentage from the top when the luminance values of the pixels of the image of the normal image data are arranged in descending order at random positions within a specified range of the image of the normal image data obtained by photographing the facility where no liquid leakage or dripping occurs, the step of performing learning on the normal image data and the pseudo-abnormal image data to generate a determination model for determining the normality or abnormality of the facility, the step of acquiring image data obtained by photographing the facility, the step of detecting an abnormality of the facility from the acquired image data using the determination model, An abnormality detection method having the above steps.

12. The liquid supply unit consists of an arm provided with a nozzle or a pipe for discharging liquid The abnormality detection method according to Claim 11.

13. Receiving specifications for the range of image synthesis and the size of the image to be synthesized, In the step of generating the pseudo-abnormal image data, pseudo-abnormal image data is generated by synthesizing the substantially circular image at a random position within the specified range and with a size equal to or smaller than the specified size. The abnormality detection method according to Claim 11 or 12.

14. The step of acquiring image data obtained by photographing a facility equipped with a liquid supply unit, Using a determination model generated by performing learning on the normal image data obtained by photographing the facility where no liquid leakage or dripping occurs and the pseudo-abnormal image data obtained by synthesizing a substantially circular image with luminance values of a predetermined percentage from the top when the luminance values of the pixels of the image of the normal image data are arranged in descending order at random positions within a specified range of the image of the normal image data, and detecting an abnormality of the facility from the acquired image data using the determination model for determining the normality or abnormality of the facility, An abnormality detection method having the above steps.

15. The liquid supply unit consists of an arm provided with a nozzle or a pipe for discharging liquid The abnormality detection method according to Claim 14.

16. A facility equipped with a liquid supply unit, the step of acquiring normal image data obtained by photographing a series of operations of the facility that operates normally without liquid leakage or dripping occurring, For each of a plurality of normal image data obtained by photographing a series of operations of the equipment operating normally, a substantially circular image is synthesized with luminance values of a predetermined ratio from the top when the luminance values of the pixels of the image of the normal image data are arranged in descending order at random positions in the image, thereby generating a plurality of pseudo-abnormal image data; Performing learning on the plurality of normal image data and the plurality of pseudo-abnormal image data to generate a determination model for determining normality and abnormality of the equipment; Obtaining a plurality of image data obtained by photographing a series of operations of the equipment; Using the determination model, determining normality and abnormality of the obtained plurality of image data, and if any of the image data is determined to be abnormal, detecting that an abnormality has occurred in the equipment; An abnormality detection method comprising the steps of:

17. In the detecting step, if it is determined that there is an abnormality in a predetermined number of the image data continuously photographed, it is detected that an abnormality has occurred in the equipment. The abnormality detection method according to claim 16.

Images