Tool state learning device, tool state estimation device, control device, tool state learning method, and tool state estimation method

Abstract

The tool state is easily estimated from the image of the machined surface. The tool state learning device comprises: a storage unit that stores an arbitrary image obtained by photographing the machined surface of an arbitrary workpiece that has been cut using an arbitrary tool through a photographing device; a teacher data acquisition unit that acquires the arbitrary image stored in the storage unit as input data, and acquires the state of the tool annotated corresponding to a predetermined stage representing the degree of wear of the tool based on the arbitrary image as a label; and a tool state learning unit that performs supervised learning using the input data and the label acquired by the teacher data acquisition unit to generate a learning completion model that inputs the machined surface image of the workpiece photographed by the photographing device and outputs the state of the tool that has cut the machined surface of the workpiece.

Description

Technical Field

[0001] The present invention relates to a tool state learning device, a tool state estimating device, a control device, a tool state learning method and a tool state estimating method. Background Art

[0002] The cutting edge of the tool used in cutting processing wears out as it is used. The precision of cutting processing is affected by the wear of the tool tip, so generally, a specific service life is pre-set for each tool type, and the tool is replaced based on the total processing time and the service life.

[0003] In this regard, the following technology is known: images of machined surfaces of multiple workpieces after cutting are captured, and the life of the tool is determined based on the contrast of the captured machined surface images indicating the quality of the cutting results compared with a preset threshold value. For example, refer to Patent Document 1.

[0004] Prior art literature

[0005] Patent Literature

[0006] Patent Document 1: Japanese Patent Application Publication No. 2018-1288 Summary of the invention

[0007] Problems to be solved by the invention

[0008] When using numerical data such as contrast as a reference, it may vary depending on the workpiece or imaging conditions, and it is generally difficult to determine a threshold value for the tool replacement timing.

[0009] Therefore, it is desired to easily estimate the tool state based on an image of a machined surface.

[0010] Means for solving problems

[0011] (1) One method of the tool state learning device disclosed in the present invention comprises: a storage unit, which stores an arbitrary image obtained by photographing a machined surface of an arbitrary workpiece that has been cut using an arbitrary tool through a photographing device; a teacher data acquisition unit, which acquires the arbitrary image stored in the storage unit as input data, and acquires the state of the tool annotated corresponding to a predetermined stage representing the degree of wear of the tool based on the arbitrary image as a label; and a tool state learning unit, which uses the input data and labels acquired by the teacher data acquisition unit to perform supervised learning and generate a learning completion model, which inputs a machined surface image of the machined surface of the workpiece photographed by the photographing device and outputs the state of the tool that has cut the machined surface of the workpiece.

[0012] (2) One method of the tool state estimation device disclosed in the present invention comprises: a learning completion model, which is generated by the tool state learning device of (1), which inputs a machining surface image of the machining surface of the workpiece photographed by a camera and outputs a state of a tool that has cut the machining surface of the workpiece; an input unit, which inputs the machining surface image of the machining surface photographed by the camera; and a tool state estimation unit, which estimates the state of the tool that has cut the machining surface of the machining surface image output by the learning completion model by inputting the machining surface image input by the input unit into the learning completion model.

[0013] (3) One aspect of the control device disclosed herein includes the tool state estimating device of (2).

[0014] (4) One mode of the tool state learning method disclosed in the present invention comprises: a teacher data acquisition step, in which an arbitrary image obtained by photographing a machined surface of an arbitrary workpiece that has been cut using an arbitrary tool is acquired as input data, and the state of the tool annotated corresponding to a predetermined stage representing the degree of wear of the tool based on the arbitrary image is acquired as a label; and a tool state learning step, in which supervised learning is performed using the acquired input data and the label to generate a learning completion model, which inputs a machined surface image of the machined surface of the workpiece photographed by the photographing device and outputs the state of the tool that has cut the machined surface of the workpiece.

[0015] (5) One mode of the tool state estimation method disclosed in the present invention comprises: an input step of inputting a machined surface image of a machined surface photographed by a photographing device; and an estimation step of inputting the machined surface image input in the input step into a learning completion model generated by a tool state learning device of (1), thereby estimating the state of a tool that has been subjected to cutting processing on the machined surface of the machined surface image output by the learning completion model, wherein the learning completion model inputs the machined surface image of the workpiece that is photographed by the photographing device and outputs the state of the tool that has been subjected to cutting processing on the machined surface of the workpiece.

[0016] Effects of the Invention

[0017] According to one aspect, the tool state can be easily estimated from the image of the machined surface. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] Figure 1 This is a functional block diagram showing an example of a functional configuration of a tool state estimation system according to one embodiment.

[0019] Figure 2 This is a diagram showing an example of the relationship between the wear of the cutting edge of the tool and the machined surface of the workpiece.

[0020] Figure 3 This is a diagram showing an example of an image of a processed surface.

[0021] Figure 4 This is a diagram showing an example of the degree of wear of a tool.

[0022] Figure 5 This is a flowchart for explaining the estimation process of the tool state estimation device in the operation phase.

[0023] Figure 6 It is a diagram showing an example of the configuration of a tool state estimation system.

[0024] Figure 7 It is a diagram showing an example of the configuration of a tool state estimation system. DETAILED DESCRIPTION

[0025] Hereinafter, one embodiment of the present disclosure will be described using the drawings.

[0026] <One embodiment>

[0027] Figure 1 1 is a functional block diagram showing an example of a functional configuration of a tool state estimation system according to one embodiment. Figure 1 As shown, the tool state estimation system 1 includes an imaging device 10 , a tool state estimation device 20 , and a tool state learning device 30 .

[0028] The photographing device 10, the tool state estimation device 20, and the tool state learning device 30 may also be directly connected to each other via a connection interface not shown. In addition, the photographing device 10, the tool state estimation device 20, and the tool state learning device 30 may also be connected to each other via a network not shown, such as a LAN (Local Area Network), the Internet, etc. In this case, the photographing device 10, the tool state estimation device 20, and the tool state learning device 30 have a communication unit not shown for communicating with each other through such a connection. In addition, as described later, the tool state estimation device 20 may also include the tool state learning device 30.

[0029] The imaging device 10 is a digital camera or the like, and is disposed on a machine tool (not shown) for example, to capture an image of a machined surface of a workpiece W that has been subjected to cutting by the machine tool (not shown). The imaging device 10 outputs the captured image of the machined surface of the workpiece W to a tool state estimation device 20 described later, based on a control instruction from a control device (not shown) that controls the cutting of the machine tool (not shown).

[0030] The tool state estimation device 20 acquires the image of the machined surface of the workpiece W captured by the imaging device 10 during the operation phase. Furthermore, the tool state estimation device 20 inputs the acquired image of the machined surface of the workpiece W into the learning completed model provided by the tool state learning device 30 described later. Thus, the tool state estimation device 20 can estimate, for example, the degree of wear of the tool tip (hereinafter also referred to as "tool wear") as the state of the tool after cutting the machined surface of the workpiece W.

[0031] Furthermore, the tool state estimating device 20 may output the image of the machined surface of the workpiece W acquired from the imaging device 10 to the tool state learning device 30 during the learning phase.

[0032] In addition, as will be described later, the tool state estimating device 20 may be included in a control device (not shown).

[0033] Before describing the tool state estimation device 20 , the “degree of tool wear” and machine learning for generating a learned model will be described.

[0034] <About tool wear>

[0035] As described above, the “tool wear degree” indicates the degree of wear of the cutting edge of the tool used in the cutting process.

[0036] Figure 2 : is a diagram showing an example of the relationship between the wear of the tool tip and the machined surface of the workpiece W. Figure 2 For example, the front slicing process is shown. Figure 2 As shown in the figure, as the number of cutting processes (i.e., the cumulative use time of the tool) increases from left to right, the tool tip wears, and the cutting marks on the machined surface of the workpiece W gradually change. For example, when the workpiece W is aluminum, as the tool tip wears, the machined surface of the workpiece W gradually becomes brighter. For example, when the workpiece W is a casting, as the tool tip wears, the machined surface of the workpiece W gradually becomes darker.

[0037] For example, Figure 2 As shown on the left side of the figure, a new tool has no wear on the cutting edge and is in the best sharpness state, so the "tool wear degree" is "0%". Figure 2 As shown, as the tool is used, the tool tip wears, and therefore, the "tool wear degree" is a value between "0%" and "100%".

[0038] Furthermore, the “degree of tool wear” is a percentage value in the range of “0%” to “100%”, but may be a value in the range of “0” to “1”, for example.

[0039] As described later, an operator determines an annotation of the degree of tool wear based on an image of a machined surface of an arbitrary workpiece captured by the imaging device 10 .

[0040] In addition, in the present embodiment, the tool state estimation device 20 estimates "tool wear" as the state of the tool that has been subjected to cutting processing, but the present invention is not limited to this. For example, the tool state estimation device 20 may estimate any one of two or more stages such as "normal" and "abnormal" as the state of the tool that has been subjected to cutting processing. In this case, "normal" means, for example, Figure 2 The tool state that can perform cutting with appropriate processing accuracy like the two tools shown on the left side of the figure. In addition, "abnormal" means, for example, Figure 2 Like the two tools shown on the right, these are tools that are close to replacement time.

[0041] <Tool status learning device 30>

[0042] The tool state learning device 30 acquires in advance, as input data, an image obtained by, for example, capturing an image of a machined surface of an arbitrary workpiece cut using an arbitrary tool using the imaging device 10 .

[0043] Furthermore, the tool state learning device 30 acquires, based on the acquired image of input data, data indicating the state of the tool after cutting the machined surface, determined from the machined surface of the workpiece in the image, as a label (correct answer).

[0044] The tool state learning device 30 performs supervised learning based on the acquired input data and training data of the label set, and constructs a learning completion model described later.

[0045] Thus, the tool state learning device 30 can provide the constructed learning model to the tool state estimating device 20 .

[0046] The tool state learning device 30 will be described in detail.

[0047] like Figure 1 As shown, the tool state learning device 30 includes a control unit 31, a storage unit 32, and an image display unit 33. In addition, the control unit 31 includes a teacher data acquisition unit 311, a tool state learning unit 312, and an image region identification unit 313.

[0048] The image display unit 33 is a liquid crystal display or the like, and displays an image captured by the imaging device 10 in accordance with a control instruction from the control unit 31 described later.

[0049] <Storage Unit 32>

[0050] The storage unit 32 is a RAM (Random Access Memory), a HDD (Hard Disk Drive), etc. The storage unit 32 stores the teacher data 321 and the learning completion model 250 .

[0051] The teacher data 321 stores an image of a machined surface of an arbitrary workpiece cut using an arbitrary tool, which is acquired from the imaging device 10 by a teacher data acquisition unit 311 described later via a communication unit (not shown).

[0052] In addition, the teacher data 321 stores data indicating the state (“normal” or “abnormal”) of the tool after cutting the machined surface, determined based on the machined surface of the workpiece in the stored image, as a label.

[0053] The learned model 250 is a learned model constructed by a tool state learning unit 312 described later.

[0054] <Control Unit 31>

[0055] The control unit 31 includes a CPU (Central Processing Unit), a ROM (Read Only Memory), a RAM, a CMOS (Complementary Metal-Oxide-Semiconductor) memory, etc., which are configured to be able to communicate with each other via a bus, which is well known to those skilled in the art.

[0056] The CPU is a processor that controls the entire tool state learning device 30. The CPU reads the system program and application program stored in the ROM via the bus, and controls the entire tool state learning device 30 according to the system program and application program. Figure 1 As shown, the control unit 31 is configured to realize the functions of the teacher data acquisition unit 311, the tool state learning unit 312, and the image area determination unit 313. Various data such as temporary calculation data and display data are stored in the RAM. In addition, the CMOS is backed up by a battery (not shown) and is configured as a non-volatile memory that maintains the storage state even if the power supply of the tool state learning device 30 is turned off.

[0057] In the learning phase, the teacher data acquisition unit 311 acquires, via a communication unit (not shown), an image of a machined surface of an arbitrary workpiece cut using an arbitrary tool, captured by the imaging device 10 as input data.

[0058] In addition, the teacher data acquisition unit 311 displays the acquired image on the image display unit 33, for example. The operator annotates the tool state, that is, "the wear degree of the tool", based on the cutting marks, glossiness, or dullness in the machined surface of the displayed image. The teacher data acquisition unit 311 acquires the wear degree of the tool annotated by the operator as a label via an input device (not shown) such as a keyboard and a touch panel included in the tool state learning device 30.

[0059] The training data acquisition unit 311 stores the acquired input data of the image and the label of the wear degree of the tool in the training data 321 of the storage unit 32 as training data.

[0060] The tool state learning unit 312 obtains a set of input data and labels as training data from the teacher data 321 of the storage unit 32. The tool state learning unit 312 uses the obtained training data to perform supervised learning, thereby constructing a learning completion model 250 that inputs an image captured by the imaging device 10 and outputs the state of the tool after cutting the processing surface of the image, that is, "tool wear degree".

[0061] In addition, in the present embodiment, the tool state learning unit 312 may perform machine learning according to a neural network including a multi-layer neural network, for example, a CNN (Convolutional Neural Network) may be applied.

[0062] Then, the tool state learning unit 312 provides the constructed learning model 250 to the tool state estimating device 20 .

[0063] In addition, it is preferable to prepare a large amount of training data for supervised learning. For example, the training data may be acquired from the imaging device 10 in various places actually operated in a customer's factory or the like.

[0064] In addition, when new training data is obtained after the learned model 250 is constructed, the tool state learning unit 312 may further perform supervised learning on the learned model 250 to update the once constructed learned model 250 .

[0065] This allows training data to be automatically obtained from the normal imaging operation of the imaging device 10 , thereby improving the estimation accuracy of the tool state on a daily basis.

[0066] The above-mentioned supervised learning can be performed through online learning, batch learning, or small batch learning.

[0067] Online learning is a learning method that performs supervised learning every time a newly machined workpiece W is photographed by the camera 10 to create training data. Batch learning is a learning method that performs supervised learning based on training data of a predetermined number of images collected by the camera 10 of the newly machined workpiece W. Furthermore, mini-batch learning is a learning method between online learning and batch learning, and is a learning method that performs supervised learning every time a certain amount of training data is accumulated.

[0068] In addition, the tool state learning unit 312 can also use, for example, input data of tool information and training data of a group of labels of tool wear annotated by an operator to construct a learning completion model 250, wherein the tool information includes an image of a machined surface of any workpiece that has been cut using any tool, and at least one of the material (e.g., carbon tool steel, ceramic, etc.), shape (e.g., turning tool, front slice, etc.), tool diameter, tool length, and tool state before processing of the tool used in the cutting process, and the learning completion model 250 inputs an image of the machined surface captured by the camera 10 and tool information of the tool that has been cut on the machined surface and outputs the state of the tool.

[0069] In this case, a tool management table (not shown) for managing all tools installed on a machine tool (not shown) may also be stored in a storage unit such as a HDD included in the machine tool (not shown) or the control device (not shown). The tool state learning device 30 may also obtain at least one piece of tool information including the material, shape, tool diameter, tool length, tool state before machining, etc. of the tool from the tool management table (not shown) via the machine tool (not shown) or the control device (not shown) based on the tool number set in the machining program.

[0070] In addition, the tool state learning unit 312 can also use, for example, input data of workpiece information and training data of a group of labels of tool wear annotated by an operator to construct a learning completion model 250, wherein the workpiece information includes an image of a machined surface of an arbitrary workpiece that has been cut using an arbitrary tool, tool information of the tool used in the cutting process, and the material of the workpiece W that is the processing object, etc. The learning completion model 250 inputs an image of the machined surface captured by the photographing device 10, tool information of the tool that has performed cutting on the machined surface, and workpiece information and outputs the state of the tool.

[0071] This can improve the estimation accuracy of the learned model 250 .

[0072] In addition, the tool state learning unit 312 may construct the learning completion model 250 for each tool.

[0073] When the tool state learning unit 312 performs supervised learning using training data including an image, the image area determination unit 313 masks the image area other than the image area of ​​the machined surface of the workpiece W in the image in order to eliminate the influence from areas other than the image area.

[0074] Figure 3 is a diagram showing an example of an image of a processed surface. Figure 3 As shown, in the image of the processed surface, together with the image area SR of the processed surface indicated by the shadow, there is included the image area BR of the background such as the worktable on which the workpiece W is arranged, and the image area AR of the processed surface processed by other tools outside the image area SR. However, depending on the irradiation of light on the processed surface of the workpiece W, in the image captured by the imaging device 10, the position and shape of the processed surface of the image area SR may become unclear due to the influence of reflection from the image area AR outside the image area SR of the processed surface and the image area BR of the background.

[0075] Therefore, the image region determination unit 313 performs image matching processing using CAD data and training data obtained from a CAD / CAM device (not shown) to determine the image region SR of the processing surface. In addition, the image region determination unit 313 may mask the image region AR other than the determined image region SR of the processing surface and the image region BR of the background, and output the masked image to the tool state learning unit 312.

[0076] Thus, the tool state learning device 30 performs supervised learning using an image in which the image regions AR and BR other than the image region SR of the machined surface are masked, thereby being able to improve the estimation accuracy of the constructed learning model 250 .

[0077] Alternatively, the image region specifying unit 313 may specify the image region SR of the machining surface based on, for example, an input operation by an operator via an input device (not shown) of the tool state learning device 30 while displaying the image on the image display unit 33 .

[0078] In the above, the machine learning for generating the learned model 250 included in the tool state estimation device 20 has been described.

[0079] Next, the tool state estimating device 20 in the operation phase will be described.

[0080] <Tool state estimation device 20 in operation phase>

[0081] like Figure 1 As shown, the tool state estimating device 20 in the operation phase includes an input unit 201 , a tool state estimating unit 202 , a determining unit 203 , a notifying unit 204 , and a storage unit 205 .

[0082] In addition, in order to achieve Figure 1 In order to realize the operation of the functional blocks, the tool state estimation device 20 has a CPU and other arithmetic processing devices (not shown). In addition, the tool state estimation device 20 has a ROM (Read Only Memory) for storing various control programs, an auxiliary storage device (not shown) such as a HDD, and a main storage device (not shown) such as a RAM for storing data temporarily required when the arithmetic processing device executes the program.

[0083] Furthermore, in the tool state estimation device 20, the operation processing device reads the OS and application software from the auxiliary storage device, expands the read OS and application software in the main storage device, and performs operation processing based on these OS and application software. Based on the operation result, the tool state estimation device 20 controls each hardware. In this way, the tool state estimation device 20 realizes the operation based on the OS and application software. Figure 1 That is, the tool state estimation device 20 can be implemented by the cooperation of hardware and software.

[0084] The input unit 201 inputs an image of the machined surface of the workpiece W captured by the imaging device 10 from the imaging device 10 . The input unit 201 outputs the input image to the tool state estimating unit 202 .

[0085] In addition, if Figure 3 As shown, when a learning completion model 250 is generated using an image in which an image area AR other than an image area SR of the processing surface and an image area BR of the background are masked, the input unit 201 preferably masks the input image when outputting the image to the tool state estimation unit 202, in the same manner as the image area determination unit 313.

[0086] The tool state estimation unit 202 inputs the image input by the input unit 201 to the learned model 250. Thus, the tool state estimation unit 202 can estimate the "tool wear degree" after cutting the machined surface of the input image as the tool state based on the output of the learned model 250.

[0087] In addition, the tool state estimation device 20 can also output the wear degree of the tool estimated by the tool state estimation unit 202 to a machine tool or control device not shown in the figure, and update the tool state before processing in the tool management table (not shown) stored in the storage unit of the machine tool or control device not shown in the figure.

[0088] The decision unit 203 determines whether to replace the tool based on the degree of wear of the tool estimated by the tool state estimation unit 202 .

[0089] More specifically, the decision unit 203 determines the optimal timing for tool replacement based on a comparison between the estimated value of the degree of tool wear and a preset threshold value.

[0090] Figure 4 is a diagram showing an example of tool wear. Figure 2 As described in the , the tool tip wears as the number of cutting operations (i.e., the cumulative time the tool is used) increases, so the wear degree of the tool increases. For example, when the wear degree of the tool is "80%" or less, the tool can perform cutting with appropriate processing accuracy. On the other hand, when the wear degree of the tool exceeds "80%", it is close to the time for tool replacement, and the operator needs to replace the tool.

[0091] In the following description, the wear degree "80%" is set as the threshold value α. The threshold value α may be appropriately set according to the material of the workpiece W to be processed, the environment in which a machine tool (not shown) is installed, and the like.

[0092] The decision unit 203 determines whether the wear degree of the tool estimated by the tool state estimation unit 202 is equal to or less than the threshold value α. The decision unit 203 determines not to replace the tool when the estimated wear degree of the tool is equal to or less than the threshold value α.

[0093] On the other hand, when the estimated degree of wear of the tool exceeds the threshold value α, the determination unit 203 determines it as the time to replace the tool.

[0094] Furthermore, the determination unit 203 compares the estimated degree of wear of the tool with one threshold value α, but may compare the estimated degree of wear of the tool with two or more threshold values ​​set in advance.

[0095] Specifically, the decision unit 203 may compare the estimated tool wear with a threshold value α1 such as "80%" and a threshold value α2 such as "90%". In this case, the decision unit 203 may also decide that it is not necessary to replace the tool when the estimated tool wear is below the threshold value α1. In addition, the decision unit 203 may also decide to output an alarm indicating that the tool replacement period is approaching and that the tool replacement is not necessary but the tool replacement is imminent when the estimated tool wear exceeds the threshold value α1 and is below the threshold value α2. In addition, the decision unit 203 may also decide to output an alarm to the operator to replace the tool immediately when the estimated tool wear exceeds the threshold value α2.

[0096] Thus, the method of notifying the alarm can be changed in detail according to the degree of wear of the tool estimated by the learned model 250 .

[0097] When receiving the decision of tool replacement from the decision unit 203, the notification unit 204 may output a tool replacement instruction to an output device (not shown) such as a liquid crystal display included in the machine tool and / or the control device (not shown). Furthermore, the machine tool (not shown) may automatically replace the tool for which the replacement instruction is given by an automatic tool changer (not shown) included in the machine tool, based on a tool management table (not shown) stored in the storage unit of the machine tool or the control device (not shown).

[0098] Furthermore, the notification unit 204 may make a notification by sound via a speaker (not shown) or by lighting a rotating light or a signal light (not shown) or the like.

[0099] The storage unit 205 is a ROM, a HDD, or the like, and may store the learned model 250 together with various control programs.

[0100] <Estimation Processing by the Tool State Estimation Device 20 in Operation Phase>

[0101] Next, the operation of the estimation process of the tool state estimation device 20 according to the present embodiment will be described.

[0102] Figure 4 This is a flowchart for explaining the estimation process of the tool state estimation device 20 in the operation phase. The process shown here can be executed each time one image of the machined surface of the workpiece W is input, or each time 10 or 100 images of the machined surface of the workpiece W are input at equal intervals.

[0103] In step S11 , the input unit 201 inputs an image of the machined surface of the workpiece W captured by the imaging device 10 from the imaging device 10 .

[0104] In step S12 , the tool state estimating unit 202 estimates the degree of wear of the tool by inputting the image input in step S11 to the learned model 250 .

[0105] In step S13, the decision unit 203 determines whether the wear degree of the tool estimated in step S12 is less than the threshold value α. When the wear degree of the tool is less than the threshold value α, the decision unit 203 determines that the tool does not need to be replaced, and decides not to replace the tool. Then, the processing ends. On the other hand, when the wear degree of the tool exceeds the threshold value α, the decision unit 203 determines that the tool needs to be replaced, and decides to replace the tool. Then, the processing enters step S14.

[0106] In step S14 , the notification unit 204 outputs a tool replacement instruction to an output device (not shown) of a machine tool and / or a control device (not shown), based on the tool replacement decision received from the determination unit 203 .

[0107] As described above, the tool state estimating device 20 according to one embodiment inputs the image of the machined surface of the workpiece W captured by the imaging device 10 into the learned model 250 , thereby estimating the wear degree of the tool as the tool state.

[0108] Thus, the tool state estimating device 20 can easily estimate the tool state (that is, the degree of wear of the tool) based on the image of the machined surface, and can appropriately determine the tool replacement timing regardless of the operator's experience.

[0109] Although one embodiment has been described above, the tool state estimating device 20 and the tool state learning device 30 are not limited to the above-described embodiment, and include modifications and improvements within a range that can achieve the purpose.

[0110] <Variation 1>

[0111] In the above-mentioned embodiment, the tool state learning device 30 is illustrated as a device different from the tool state estimation device 20, the unillustrated machine tool or the unillustrated control device, but the tool state estimation device 20, the unillustrated machine tool or the unillustrated control device may also have part or all of the functions of the tool state learning device 30.

[0112] <Variation 2>

[0113] In the above-mentioned embodiment, the tool state estimation device 20 is illustrated as a device different from the unillustrated machine tool and the unillustrated control device. However, the unillustrated machine tool or the unillustrated control device may have some or all of the functions of the tool state estimation device 20.

[0114] Alternatively, for example, the server may include part or all of the input unit 201, tool state estimation unit 202, determination unit 203, notification unit 204, and storage unit 205 of the tool state estimation device 20. In addition, each function of the tool state estimation device 20 may be implemented on the cloud using a virtual server function or the like.

[0115] Furthermore, the tool state estimation device 20 may be a distributed processing system in which the functions of the tool state estimation device 20 are appropriately distributed among a plurality of servers.

[0116] <Variation 3>

[0117] In addition, for example, in the above-mentioned embodiment, the tool state estimation device 20 uses the learning model 250 provided by the tool state learning device 30 to estimate the wear of the tool as the tool state based on the image of the processing surface captured by the imaging device 10, but the present invention is not limited to this. Figure 6As shown, the server 50 may store the learning model 250 generated by the tool state learning device 30, and the m tool state estimation devices 20A(1)-20A(m) connected to the network 60 may download the learning model 250 from the server 50 (m is an integer greater than or equal to 2). Thus, even if a new machine tool and a tool state estimation device are installed, the learning model 250 can be applied.

[0118] Furthermore, each of the tool state estimation devices 20A(1)-20A(m) is connected to each of the machine tools 15(1)-15(m).

[0119] In addition, each of the machine tools 15(1)-15(m) includes Figure 1 The tool state estimation devices 20A(1)-20A(m) are each configured to have a camera 10 similar to the camera 10 and a control device not shown. Figure 1 The tool state estimating device 20 corresponds to.

[0120] Or, if Figure 7 As shown, the server 50 may also operate as a tool state estimation device 20, for example, to estimate the tool state for each of the machine tools 15 (1) to 15 (m) connected to the network 60 based on the image of the processing surface captured by the imaging device of each machine tool 15 (i). In addition, i is an integer from 1 to m. Thus, even if a new machine tool is installed, the learning completion model 250 can be applied.

[0121] In addition, each function included in the tool state estimation device 20 and the tool state learning device 30 in one embodiment can be realized by hardware, software, or a combination thereof. Here, realization by software means realization by a computer reading a program and executing it.

[0122] The various components included in the tool state estimation device 20 and the tool state learning device 30 can be implemented by hardware including electronic circuits, software, or a combination thereof. In the case of implementation by software, the program constituting the software is installed in a computer. In addition, these programs can be recorded on a removable medium and distributed to users, or they can be distributed by downloading to the user's computer via a network. In addition, in the case of being composed of hardware, for example, a part or all of the functions of the various components included in the above-mentioned device can be constituted by an integrated circuit (IC) such as ASIC (Application Specific Integrated Circuit), gate array, FPGA (Field Programmable Gate Array), CPLD (Complex Programmable Logic Device), etc.

[0123] Various types of non-transitory computer readable media (Non-transitory computer readable medium) can be used to store the program and provide it to the computer. Non-transitory computer readable media include various types of tangible recording media (Tangible storage medium). Examples of non-transitory computer readable media include: magnetic recording media (e.g., floppy disks, magnetic tapes, hard disk drives), magneto-optical recording media (e.g., optical magnetic disks), CD-ROM (Read Only Memory), CD-R, CD-R / W, semiconductor memory (e.g., mask ROM, PROM (Programmable ROM), EPROM (Erasable PROM), flash ROM, RAM). In addition, the program can also be supplied to the computer through various types of transient computer readable media (Transitory computer readable medium). Examples of transient computer readable media include electrical signals, optical signals, and electromagnetic waves. Transient computer readable media can supply the program to the computer via wired communication paths such as wires and optical fibers or wireless communication paths.

[0124] Furthermore, the steps describing the program recorded in the recording medium naturally include processes that are performed in time series according to the order, and also include processes that are not necessarily performed in time series but are executed in parallel or individually.

[0125] In other words, the tool state learning device, tool state estimating device, control device, tool state learning method, and tool state estimating method of the present disclosure can take various embodiments having the following structures.

[0126] (1) The tool state learning device 30 disclosed in the present invention comprises: a storage unit 32, which stores an arbitrary image obtained by photographing the machined surface of an arbitrary workpiece that has been cut using an arbitrary tool through the photographing device 10; a teacher data acquisition unit 311, which acquires an arbitrary image stored in the storage unit 32 as input data, and acquires the state of the tool annotated corresponding to a predetermined stage representing the degree of wear of the tool based on the arbitrary image as a label; and a tool state learning unit 312, which uses the input data and labels acquired by the teacher data acquisition unit 311 to perform supervised learning and generate a learning completion model 250, which inputs a machined surface image of the machined surface of the workpiece W photographed by the photographing device 10 and outputs the state of the tool that has been cut on the machined surface of the workpiece W.

[0127] According to the tool state learning device 30 , it is possible to generate a learning model 250 that outputs the tool state with high accuracy based on an image of a machined surface.

[0128] (2) The tool state learning device 30 described in (1) may further include an image display unit 33 that displays an arbitrary image in order to annotate the state of the tool in accordance with a predetermined stage.

[0129] Thus, the tool state learning device 30 can acquire, as a label, the state of the tool annotated in correspondence with the predetermined stage indicating the degree of wear of the tool.

[0130] (3) In the tool state learning device 30 described in (1) or (2), it may also have: an image area determination unit 313, which masks the image areas AR and BR other than the image area SR of the processing surface in any image, and the teacher data acquisition unit 311 obtains the masked arbitrary image as input data.

[0131] Thus, the tool state learning device 30 can improve the estimation accuracy of the learned model 250 by performing supervised learning using the masked image.

[0132] (4) In the tool state learning device 30 recorded in any one of (1) to (3), the tool state learning unit 312 may use the processing surface image newly captured by the photographing device 10 as input data, and use the state of the tool output from the learning completion model 250 for the input processing surface image as a label to perform supervised learning, thereby updating the learning completion model 250.

[0133] Thereby, the tool state learning device 30 can improve the estimation accuracy of the learned model 250 .

[0134] (5) In the tool state learning device 30 described in any one of (1) to (3), the teacher data acquisition unit 311 may acquire an arbitrary image and at least one piece of tool information including the material, shape and state of the tool before processing as input data, and the tool state learning unit 312 may generate a learning completion model, which inputs a processing surface image captured by the photographing device 10 and tool information of the tool after cutting the processing surface of the processing surface image and outputs the state of the tool.

[0135] Thereby, the tool state learning device 30 can improve the estimation accuracy of the learned model 250 .

[0136] (6) The tool state estimation device 20 disclosed in the present invention comprises: a learning completion model 250, which is generated by the tool state learning device 30 recorded in any one of (1) to (4), and inputs a machined surface image of the machined surface of the workpiece W photographed by the photographing device 10, and outputs a state of a tool that has cut the machined surface of the workpiece W; an input unit 201, which inputs the machined surface image of the machined surface photographed by the photographing device 10; and a tool state estimation unit 202, which infers the state of the tool that has cut the machined surface of the machined surface image output by the learning completion model 250 by inputting the machined surface image input by the input unit 201 into the learning completion model 250.

[0137] According to the tool state estimating device 20, the tool state can be easily estimated based on the image of the machined surface.

[0138] (7) The tool state estimating device 20 described in (6) may further include a determination unit 203 that determines whether it is time to replace the tool based on the state of the tool estimated by the tool state estimating unit 202 .

[0139] Thus, the tool state estimating device 20 can notify the operator of the tool replacement time with high accuracy.

[0140] (8) In the tool state estimation device 20 described in (6) or (7), the learned model 250 may be provided in a server 50 that is connected to the tool state estimation device 20 via the network 60 so as to be accessible thereto.

[0141] Thus, the tool state estimation device 20 can apply the learned model 250 even if a new imaging device 10 and tool state estimation device 20 are installed.

[0142] (9) The tool state estimating device 20 described in any one of (6) to (8) may include the tool state learning device 30 described in any one of (1) to (5).

[0143] Thus, the tool state estimating device 20 can obtain the same effects as any of the above-mentioned (1) to (8).

[0144] (10) The control device of the present disclosure may include the tool state estimating device 20 described in any one of (6) to (9).

[0145] According to this control device, the same effect as any one of the above-mentioned (1) to (9) can be obtained.

[0146] (11) The tool state learning method disclosed in the present invention comprises: a teacher data acquisition step, in which an arbitrary image obtained by photographing the machined surface of an arbitrary workpiece that has been cut using an arbitrary tool is acquired as input data, and the state of the tool annotated corresponding to a predetermined stage representing the degree of wear of the tool based on the arbitrary image is acquired as a label; and a tool state learning step, in which supervised learning is performed using the acquired input data and labels to generate a learning completion model, which inputs a machined surface image of the machined surface of the workpiece photographed by the photographing device 10 and outputs the state of the tool that has cut the machined surface of the workpiece.

[0147] This tool state learning method can achieve the same effect as (1) above.

[0148] (12) The tool state estimation method disclosed in the present invention comprises: an input step of inputting a machined surface image of a machined surface photographed by a photographing device 10; and an estimation step of inputting the machined surface image inputted in the input step into a learning completion model 250 generated by a tool state learning device 30 described in any one of (1) to (4), which inputs a machined surface image of a workpiece W photographed by a photographing device 10 and outputs a state of a tool having cut the machined surface of the workpiece W, thereby estimating the state of the tool having cut the machined surface of the machined surface image outputted by the learning completion model 250.

[0149] This tool state estimation method can achieve the same effect as the above-mentioned (6).

[0150] Explanation of symbols

[0151] 1 Tool status estimation system,

[0152] 10. Camera

[0153] 20 Tool status estimation device,

[0154] 201 Input unit,

[0155] 202 Tool status estimation unit,

[0156] 203 Decision Department,

[0157] 204 Notification Department,

[0158] 205 Storage Department,

[0159] 250 Learning completion model,

[0160] 30 tool status learning device,

[0161] 311 Teacher Data Acquisition Department,

[0162] 312 Tool Status Learning Department,

[0163] 313 image area determination unit,

[0164] 32 Storage Department,

[0165] 321 Teacher Data,

[0166] 33 Image display unit.

Claims

1. A tool state learning device, characterized in that: have: a storage unit storing an arbitrary image obtained by photographing a processed surface of an arbitrary workpiece cut by using an arbitrary tool by means of an imaging device; a teacher data acquisition unit that acquires the arbitrary image stored in the storage unit as input data, and acquires the state of the tool annotated corresponding to a predetermined stage indicating the degree of wear of the tool based on the arbitrary image as a label; and a tool state learning unit that performs supervised learning using the input data and labels obtained by the teacher data acquisition unit to generate a learning completion model that inputs a machined surface image of the machined surface of the workpiece captured by the camera and outputs a state of a tool that has cut the machined surface of the workpiece, The processed surface image is captured after processing.

2. The tool state learning device according to claim 1, characterized in that: The tool state learning device further includes an image display unit configured to display the arbitrary image in order to annotate the state of the tool in correspondence with the predetermined stage.

3. The tool state learning device according to claim 1 or 2, characterized in that: The tool state learning device further comprises: an image region determining unit which masks an image region other than an image region of the processing surface in the arbitrary image; The teacher data acquisition unit acquires the masked arbitrary image as input data.

4. The tool state learning device according to claim 1 or 2, characterized in that: The tool state learning unit uses the processing surface image newly captured by the imaging device as input data, and uses the state of the tool output from the learning completion model for the newly captured processing surface image as a label to perform the supervised learning and update the learning completion model.

5. The tool state learning device according to claim 1 or 2, characterized in that: The teacher data acquisition unit acquires the arbitrary image and tool information including at least one of the material and shape of the tool and the tool state before processing as input data, The tool state learning unit generates a learning model that receives as input the machined surface image captured by the imaging device and tool information of a tool that has cut the machined surface in the machined surface image, and outputs the state of the tool.

6. A tool state estimation device, characterized in that: have: A learning completion model, which is generated by a tool state learning device according to any one of claims 1 to 4, inputting a machined surface image of a machined surface of a workpiece photographed by a photographing device and outputting a state of a tool that has cut the machined surface of the workpiece; an input unit that inputs a processed surface image of the processed surface photographed by the photographing device; and A tool state estimating unit estimates a state of a tool that has cut a machining surface of the machining surface image output by the learning model by inputting the machining surface image input by the input unit into the learned model.

7. The tool state estimation device according to claim 6, characterized in that: The tool state estimating device includes a determination unit that determines whether it is time to replace the tool based on the state of the tool estimated by the tool state estimating unit.

8. The tool state estimation device according to claim 6 or 7, characterized in that: The learning completion model is stored in a server that is connected to the tool state estimation device in a manner accessible via a network.

9. The tool state estimation device according to claim 6 or 7, characterized in that: The tool state estimation device includes the tool state learning device according to any one of claims 1 to 5.

10. A control device, characterized in that: have: The tool state estimating device according to any one of claims 6 to 9.

11. A tool state learning method, characterized in that: have: A teacher data acquisition step, acquiring as input data an arbitrary image obtained by photographing a machined surface of an arbitrary workpiece cut by an arbitrary tool using a photographing device, and acquiring as a label the state of the tool annotated corresponding to a predetermined stage indicating the degree of wear of the tool based on the arbitrary image; and a tool state learning step, performing supervised learning using the acquired input data and the labels to generate a learning completion model, the learning completion model inputting a machined surface image of the machined surface of the workpiece captured by the camera and outputting a state of a tool that has cut the machined surface of the workpiece, The processed surface image is captured after processing.

12. A tool state estimation method, characterized in that: have: An input step of inputting a processing surface image of the processing surface photographed by a photographing device; and An estimating step, inputting the machined surface image input in the input step into a learning completion model generated by a tool state learning device as described in any one of claims 1 to 4, thereby estimating the state of the tool that has performed cutting processing on the machined surface of the machined surface image output by the learning completion model, wherein the learning completion model inputs the machined surface image of the workpiece captured by the photographing device and outputs the state of the tool that has performed cutting processing on the machined surface of the workpiece.

Images