Chip detector

Abstract

A chip detection device for detecting chips generated when a workpiece is machined by a machine tool, in order to easily and accurately detect chips generated when a workpiece is machined by a machine tool, includes a mesh division unit that divides at least a portion of an area image captured of a target area for chip detection into a plurality of mesh images of a predetermined mesh size, and a chip information determination unit that determines chip information for each mesh image using determination parameters that are preset to determine chip information related to the chips.

Description

[Technical field]

[0001] The present invention relates to a chip detection device that detects chips generated when a workpiece is machined by a machine tool, a machine tool, a chip detection method, an image synthesis device for learning, and the like. [Background technology]

[0002] Conventionally, chips generated when machining a workpiece with a machine tool can cause damage to the workpiece during machining, or can cause machining defects, malfunctions of the machine, and reduced availability due to unintended operation stoppages. For this reason, as a technology for automatically detecting and removing chips generated during machining, for example, JP-A-7-108435 discloses a chip removal device that captures images of the table and workpiece and detects the position of the chips (Patent Document 1). [Prior art documents] [Patent documents]

[0003] [Patent Document 1] Japanese Patent Application Publication No. 7-108435 Summary of the Invention [Problem to be solved by the invention]

[0004] However, the chips generated during machining have countless variations in shape, color, size, and collection state depending on the machining conditions, tools used, workpiece material, etc. In addition, the in-machine environment that forms the background of the image has conditions such as structure and position, type of workpiece and jig, presence or absence of coolant (cleaning fluid), and illuminance, and there are various combinations of these conditions.

[0005] For this reason, when attempting to detect chips from an entire image as in Patent Document 1, a large number of chips and the in-machine environment are captured in the image in a complex manner with a wide variety of combinations, making it difficult to precisely and accurately detect each chip by image processing, and image processing to detect all chips requires a huge amount of calculation and time. [Means for solving the problem]

[0006] Therefore, the present invention provides a chip detection device, a machine tool, a chip detection method, an image synthesis device for learning, and the like. One aspect of the present invention is a chip detection device that detects chips generated when a workpiece is machined by a machine tool, in order to solve the problem of easily and accurately detecting chips generated when a workpiece is machined by a machine tool, and has a mesh division unit that divides at least a portion of an area image photographed of a target area for detecting the chips into a plurality of mesh images of a predetermined mesh size, and a chip information determination unit that determines the chip information for each of the mesh images using determination parameters that are preset in order to determine chip information related to the chips. Effect of the Invention

[0007] According to the present invention, chips generated when a workpiece is machined by a machine tool can be easily detected. [Brief description of the drawings]

[0008] [Figure 1] 1 is a block diagram showing an embodiment of a chip detection device and a machine tool including the same; [Diagram 2] FIG. 2 is a diagram showing a state in which an area image is divided into a plurality of mesh images in the present embodiment. [Diagram 3] 11A and 11B are diagrams showing how a plurality of partial images are synthesized to form an area image in the present embodiment. [Figure 4] 11A to 11C are diagrams illustrating a division process performed by a mesh division unit according to the present embodiment. [Diagram 5] FIG. 11 is a diagram for explaining “class” which is chip information in this embodiment. [Figure 6] FIG. 11 is a diagram illustrating a determination process using a chip determination model according to the present embodiment. [Figure 7] 13 is a diagram showing an example of a feature amount of a mesh image calculated by a chip information determination unit that employs a non-machine learning algorithm. FIG. [Figure 8] 13 is an example of a display screen showing (a) a judgment result before correction, and (b) a judgment result after correction in the teaching mode of the present embodiment. [Figure 9] 4 is a flowchart showing a chip detection method executed by the chip detection device of the present embodiment. [Figure 10] 4 is a flowchart showing a chip detection process in the present embodiment. [Figure 11] 4 is a flowchart showing a teaching process in the present embodiment. [Figure 12] 1 is a block diagram showing a configuration of a learning image synthesis device according to an embodiment of the present invention; DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0009] Hereinafter, an embodiment of a chip detection device, a machine tool including the same, and a chip detection method will be described with reference to the drawings.

[0010] The chip detection device 1 of this embodiment detects chips generated when a workpiece is machined by a machine tool 10. Specifically, as shown in Fig. 1, the chip detection device 1 controls the imaging means 11 provided on the machine tool 10 to obtain an image of a target area for detecting chips. Then, as shown in Fig. 2, the image is divided into a plurality of mesh images, and chip information related to the chips is determined for each of the mesh images. Each component will be described in detail below.

[0011] In the present invention, chip information is a concept that includes detectable information regarding chips, such as the presence or absence of chips, the amount of chips, the shape of the chips, the type of chips, or a combination of these, and in this embodiment, as will be described later, a "class" indicating the ease of cleaning the chips is determined as the chip information.

[0012] The machine tool 10 is a machine for performing machining such as cutting and grinding on workpieces such as metal, wood, stone, and resin. In this embodiment, the machine tool 10 is numerically controlled according to a drive signal output from a numerical control device serving as the chip detection device 1. In this embodiment, the chip detection device 1 and the numerical control device of the machine tool are configured as a single device, but this is not limited to this. A computer separate from the machine tool may be used as the chip detection device 1. Also, as shown in FIG. 1, the machine tool 10 has an imaging means 11 for imaging the inside of the machine, a table 12 on which the workpiece is placed, a cleaning means 13 for cleaning the chips, and an illumination means 14 for illuminating the inside of the machine.

[0013] The photographing means 11 photographs a target area for detecting chips. In this embodiment, the photographing means 11 is configured with a camera (hereinafter referred to as an ATC camera) that can be attached to the spindle of the machine tool 10 by an automatic tool changer (ATC) in the same way as a tool. Then, according to a photographing control signal from an image data acquisition unit 43 described later, an area image covering a preset target area is photographed at a preset photographing timing.

[0014] In this embodiment, an ATC camera that can be attached to and detached from the spindle of the machine tool is used as the imaging means 11, taking into consideration the advantages that it is not necessary to install it inside the machine and that the time of exposure to coolant mist and the like is relatively short. Since the ATC camera can be attached to the spindle of the machine tool, if the spindle of the machine tool is moved, the camera can be moved accordingly. Therefore, the imaging range and imaging direction of the camera can be easily changed. Although the ATC camera is used as an example to explain the imaging range and imaging direction of the camera because they can be easily changed, the present invention is not limited to this configuration. One or more fixed cameras may be installed inside the machine. This configuration has the advantages that the imaging time is relatively short and imaging can be performed even during processing.

[0015] Also, when the angle of view of the imaging means 11 is smaller than the target area, an ATC camera may capture multiple partial images while moving its main axis, or a fixed camera may capture multiple partial images by multiple fixed cameras, as shown in Fig. 3. Then, each partial image may be synthesized to form an overall area image that covers the entire target area.

[0016] The cleaning means 13 cleans and removes chips generated during machining of the workpiece. In this embodiment, the cleaning means 13 is configured with a programmable nozzle capable of controlling the spray direction of the cleaning liquid. The cleaning liquid, such as coolant liquid, is sprayed onto a desired area on the workpiece or table 12 according to a cleaning control signal from a cleaning processing unit 48, which will be described later.

[0017] In this embodiment, a programmable nozzle is used as the cleaning means 13, but the present invention is not limited to this configuration. For example, a plurality of cleaning nozzles with fixed spray directions may be provided to cover the entire target area, and cleaning liquid may be sprayed from the cleaning nozzle closest to the area that needs cleaning.

[0018] The lighting means 14 illuminates the interior of the machine so as to facilitate detection of chips. In this embodiment, the lighting means 14 is fixed inside the machine as shown in FIG. 1 and is configured as a programmable light that can be switched on and off, and whose brightness and direction of illumination can be controlled. However, the lighting means 14 is not limited to this configuration, and a light attached to the ATC camera may be used. In addition, in order to suppress reflections due to lighting, the lighting means 14 may be rotated or a polarizing filter may be attached to the camera lens of the imaging means 11.

[0019] Next, the chip detection device 1 is composed of a computer such as a numerical control device that controls the machine tool 10, and as shown in Fig. 1, it mainly has a display input means 2 that displays user's instruction input and judgment results, a storage means 3 that stores various data and functions as a working area when the arithmetic processing means 4 performs arithmetic processing, and the arithmetic processing means 4 that executes various arithmetic processing by executing the chip detection program 1a installed in the storage means 3 and functions as each of the components described below. Each of the components will be described in detail below.

[0020] It should be noted that, to reiterate, the chip detection device may be a device separate from the numerical control device that controls the machine tool 10. The chip detection device only needs to have at least a mesh dividing unit that divides at least a part of an area image obtained by photographing a target area for chip detection into a plurality of mesh images with a predetermined mesh size, and a chip information determination unit that determines chip information for each of the mesh images by using determination parameters that are preset in order to determine chip information related to the chips.

[0021] The display input means 2 is configured with a touch panel, and has both an input function of receiving instruction input from a user and a display function of displaying the results of chip determination, etc. In this embodiment, a touch panel is used as the display input means 2, but this configuration is not limited, and a display means having only a display function and an input means having only an input function may be provided separately.

[0022] The storage means 3 is composed of a hard disk, a ROM (Read Only Memory), a RAM (Random Access Memory), a flash memory, etc., and as shown in FIG. 1, has a program storage unit 31, a judgment parameter storage unit 32, a judgment result storage unit 33, a cleaning condition storage unit 34, and a cleaning count storage unit 35.

[0023] A chip detection program 1a for controlling the chip detection device 1 of this embodiment is installed in the program storage unit 31. The arithmetic processing means 4 executes the chip detection program 1a, thereby causing the chip detection device 1 as a computer to function as each of the components described below.

[0024] The use of the chip detection program 1a is not limited to the above configuration. For example, the chip detection program 1a may be stored in a non-transitory computer-readable recording medium such as a CD-ROM or USB memory, and may be read and executed directly from the recording medium. The program may also be used from an external server using a cloud computing method or an ASP (Application Service Provider) method.

[0025] The judgment parameter storage unit 32 stores judgment parameters that are set in advance to judge chip information related to chips. In this embodiment, as described later, a chip judgment model is adopted as an algorithm for judging chip information, which is a learning and inference model that is a machine learning algorithm, and learning parameters are applied to the model. For this reason, learning parameters obtained by pre-learning using teacher data consisting of a set of mesh images and chip information are stored as judgment parameters.

[0026] In addition, as described below, this embodiment has a mesh size selection unit 44 that selects the mesh size of the mesh image. For this reason, the judgment parameter storage unit 32 stores learning parameters as judgment parameters for each of a plurality of different mesh sizes. In this embodiment, the judgment parameter storage unit 32 separately stores hyperparameters that affect characteristics such as the learning efficiency of the chip judgment model.

[0027] The judgment result storage unit 33 stores the chip information judged for each mesh image as the judgment result. In this embodiment, the chip information judgment unit 46, which will be described later, judges, for each mesh image, the probability that the mesh image belongs to each class indicating the ease of cleaning the chips, as the chip information.

[0028] For example, when three classes are set, namely "Class 0" with no chips, "Class 1" with few chips, and "Class 2" with many chips, the probability P0 of "Class 0", the probability P1 of "Class 1", and the probability P2 of "Class 2" are determined. Therefore, the determination result storage unit 33 stores the above-mentioned probabilities P0, P1, and P2 for each class as the determination result in association with each mesh image.

[0029] The judgment result is not limited to the probability of each class, and the class with the highest probability may simply be stored as the judgment result. Also, the more classes there are, the more difficult it becomes to make judgments and the harder it is to improve performance, so it is preferable to set the number to around a few. An example of the simplest classes would be two classes: "Class 0" where there are no cuttings, and "Class 1" where there are cuttings.

[0030] The cleaning condition storage unit 34 stores cleaning conditions for determining whether or not cleaning of chips is necessary. In this embodiment, the cleaning conditions are set by combining the probability of each class with the position of the mesh area corresponding to each mesh image. For example, when the above-mentioned three classes are set, an efficient and high-performance determination algorithm is realized by evaluating the following cleaning conditions (1) to (4) in descending order.

[0031] (1) Among P0, P1, and P2, the maximum value is P2: Cleaning is performed. (2) P1+P2≧99%: Cleaning is performed. (3) The mesh area is on the table 12 and P1+P2≧P0: Cleaning is performed. (4) The above conditions (1) to (3) are not satisfied: Do not clean.

[0032] In this embodiment, the cleaning conditions are set by combining the chip information and the position of the mesh area, but the present invention is not limited to this and may be set using the chip information. For example, when the most simplified two classes described above are set, the cleaning conditions may be set such that, regardless of the position of the mesh area, cleaning is not performed if the class is 0 (no chips) and cleaning is performed if the class is 1 (chips).

[0033] The cleaning count storage unit 35 stores the number of cleanings performed consecutively to clean chips. In this embodiment, the cleaning count storage unit 35 stores the number of consecutive cleanings, which is the number of consecutive cleanings performed in the target area, and the number of local consecutive cleanings, which is the number of consecutive cleanings performed in the same mesh area. These cleaning counts are used to avoid a situation in which the automatic cleaning operation never ends, as described below.

[0034] Next, the calculation processing means 4 is composed of a CPU (Central Processing Unit) and the like, and by executing the chip detection program 1a installed in the storage means 3, it functions as a target area setting section 41, a drive control section 42, an image data acquisition section 43, a mesh size selection section 44, a mesh division section 45, a chip information determination section 46, a cleaning necessity determination section 47, a cleaning processing section 48, a teaching processing section 49, and an additional learning section 50, as shown in Fig. 1. Each component section will be described in more detail below.

[0035] The target area setting unit 41 sets a target area in which chips are to be detected. In this embodiment, the target area setting unit 41 is configured to set any target area such as the entire inside of the machine, on the table 12, on the protector, around the workpiece, a part of these, or a combination of these, in response to an instruction input by a user using the display input means 2.

[0036] In this embodiment, the target area setting unit 41 sets the target area manually by the user, but the present invention is not limited to this configuration and may set the target area automatically. Specifically, the target area setting unit 41 may acquire the shape of the workpiece in each machining process by referring to the machining program, and automatically set the target area so that the entire workpiece is included, or may automatically set only the machining path as the target area.

[0037] The drive control unit 42 drives and controls the motors of each axis provided in the machine tool 10. In this embodiment, the drive control unit 42 generates drive control signals for causing the machine tool 10 to perform cutting or the like based on a machining program, and outputs the signals to the motors of each axis. When an ATC camera is used as the imaging means 11, the drive control unit 42 controls an automatic tool changer (ATC) (not shown) to cause the ATC camera to change the tool.

[0038] The image data acquisition unit 43 acquires an image of the entire area of ​​the target area. In this embodiment, the image data acquisition unit 43 outputs an image capture control signal to the imaging means 11 at a preset imaging timing to acquire an image of the entire area. When the imaging means 11 acquires a partial image smaller than the target area, the imaging means 11 captures multiple partial images to cover the target area, and combines them to acquire an image of the entire area.

[0039] The timing of photographing may be set manually by the user or automatically according to a predetermined set condition. The set condition may be between processes, when a tool is replaced, at a certain time interval, when a predetermined machining amount is reached, etc. The machining amount can be estimated from the machining program, etc.

[0040] The mesh size selection unit 44 selects a mesh size when dividing an image of an entire area into a plurality of mesh images. In this embodiment, the mesh size selection unit 44 selects one or a plurality of mesh sizes as candidates from among preset mesh sizes based on a predetermined criterion.

[0041] In the present invention, the detection of chips is simplified and made more accurate by dividing the image of the entire area into a mesh image. Therefore, the mesh size is an important factor that affects the accuracy of chip detection, and it is preferable to consider the following characteristics of the chips when selecting the mesh size.

[0042] Specifically, as described above, there are many variations and patterns in chips, so no two chips are exactly the same on a two-dimensional image. However, the edges (ridges) that make up the outer shape of the chips have a common feature of being relatively complex and irregular. And these features are not lost even when a large number of chips pile up and overlap, making it difficult to recognize the individual shapes.

[0043] On the other hand, the workpiece and the jig that are the background of the chips generally have almost no irregularity like the chips, and when compared on the same scale as the chips, the edge shape is simple. Therefore, in this embodiment, it is preferable that the mesh size selection unit 44 selects a mesh size that allows the shape of the chips to be recognized based on the sizes of the chips, the workpiece, and the jig, and that simplifies the edge shapes of the workpiece and the jig that are the background of the chips in the mesh image. By selecting such a mesh size, it becomes easier to distinguish the edge shapes of the workpiece and the jig from the chips, and erroneous detection is reduced.

[0044] The mesh size selection unit 44 is not limited to the above configuration, and may be manually selected by the user from among a plurality of mesh sizes set in advance. However, since the mesh size must be fixed in the chip determination model for determining chip information, when a plurality of mesh sizes are prepared, a chip determination model trained with mesh images of each mesh size is required. On the other hand, if the mesh size used is always fixed, there is no need to provide the mesh size selection unit 44.

[0045] The mesh division unit 45 divides the area image into a plurality of mesh images. In this embodiment, the mesh division unit 45 divides the area image acquired by the image data acquisition unit 43 into a plurality of mesh images by the mesh size selected by the mesh size selection unit 44. In this embodiment, as shown in FIG. 2, a checkerboard or lattice-shaped mesh is adopted, but the shape is not limited to this, and the shape may be a rhombus, a triangle, a honeycomb, or the like. In this embodiment, the entire area image is divided into mesh images, but the shape is not limited to this, and the area where chip detection is not required may not be divided, and only at least a part of the area image may be divided. In addition, the captured image of the area and the lattice-shaped mesh information may be stored in association with each other separately. In other words, the mesh division unit 45 is not limited to generating individual mesh images, but may overlap the captured image and the lattice information to process the area of ​​the captured image divided by the lattice as a mesh area. In this case, the mesh area corresponds to the above-mentioned mesh image.

[0046] In addition, in this embodiment, when the image data acquisition unit 43 acquires an area image by synthesizing multiple partial images, the mesh division unit 45 is configured to execute one of the following processes during division processing, as shown in Figure 4. (i) If there is no overlap at the boundary between one partial image and another partial image, the boundary line of the mesh image is made to coincide with the boundary line of each partial image. (ii) If there is an overlap at the boundary between one partial image and another partial image, a mesh image including the boundary is cut out from only one of the partial images.

[0047] By executing the above process (i) or (ii) by the mesh division unit 45, multiple different partial images are not mixed in the same mesh image, which reduces false detections due to differences in brightness between partial images, positional deviations, etc.

[0048] The chip information determination unit 46 determines chip information for each of the mesh images. In this embodiment, the chip information determination unit 46 determines chip information for each of the mesh images divided by the mesh division unit 45 using the determination parameters read from the determination parameter storage unit 32, and stores the determined chip information in the determination result storage unit 33.

[0049] In this embodiment, the chip information determination unit 46 is configured to determine a "class" indicating the ease of cleaning the chips as the chip information. This class is determined by comprehensively considering the amount, density, size, length, shape, etc. of the chips. Specifically, as shown in Fig. 5, the chips are classified into a lower class when they are in a state where they are easier to clean, such as when they are few, scattered, small, short, or difficult to get caught, and the like, and the chips are classified into a higher class when they are in a state where they are more difficult to clean, such as when they are many, dense, large, long, or difficult to get caught.

[0050] In this embodiment, the chip information determination unit 46 is configured with a chip determination model that is obtained by applying learning parameters to a learning and inference model that is a machine learning algorithm. Specifically, among deep learning methods that are relatively resistant to environmental changes, a Convolutional Neural Network (CNN) is used, which is a model that is particularly specialized for image class classification.

[0051] When determining chip information using the above-mentioned chip determination model, first, the chip information determination unit 46 performs preprocessing for each mesh image to improve the determination accuracy, such as noise removal processing, image size conversion processing, etc. Next, as shown in Fig. 6, the chip information determination unit 46 inputs the mesh image to a learned chip determination model to which learning parameters prepared in advance in the determination parameter storage unit 32 are applied, and outputs the probability of each class as chip information for each mesh image.

[0052] In this embodiment, a convolutional neural network is used as the machine learning algorithm, but the invention is not limited thereto, and other machine learning algorithms may be used. For example, when using a machine learning algorithm that uses input data other than images, such as a support vector machine (SVM), a filter process or the like is performed on the mesh images to calculate the shape feature values ​​of each mesh image. Then, the shape feature values ​​may be input to a learned chip determination model to determine the class of each mesh image.

[0053] In addition, in this embodiment, a machine learning algorithm is adopted as the chip information determination unit 46, but the present invention is not limited thereto, and a non-machine learning algorithm such as a deterministic algorithm may be used. For example, the mesh image tends to become more complex and have more high image frequency components as the number of chips increases. Therefore, as shown in FIG. 7, the chip information determination unit 46 executes frequency analysis such as Fast Fourier Transform (FFT) to calculate the spectral statistics of each mesh image as a feature amount. Then, the class of chip information for each mesh image may be determined based on the magnitude relationship between the feature amount and a threshold value previously set as a determination parameter.

[0054] However, the fast Fourier transform is highly dependent on the environment, and if small screw holes, splashes of coolant, etc., other than the cutting chips are captured in the mesh image, the high frequency components increase, making it easier to make a mistaken judgment. For this reason, an area image without cutting chips may be prepared in advance, and the difference image between this and the area image to be judged may be divided into meshes and subjected to the fast Fourier transform to eliminate the environment-dependent components.

[0055] Furthermore, in this embodiment, the chip information determination unit 46 determines the chip information using a single algorithm, but is not limited to this configuration, and may make a comprehensive determination based on the determination results using multiple algorithms.

[0056] When multiple mesh sizes are selected by the mesh size selection unit 44, the chip information determination unit 46 performs determination for all mesh sizes. Then, the chip information for each mesh size is stored in the determination result storage unit 33 as a determination result.

[0057] The cleaning necessity determination unit 47 determines whether or not cleaning of chips is required for each mesh image. In this embodiment, the cleaning necessity determination unit 47 compares the cleaning conditions stored in the cleaning condition storage unit 34 with the chip information of each mesh image stored in the determination result storage unit 33. Then, it is determined that cleaning is required for mesh images that satisfy the cleaning conditions.

[0058] The cleaning processing unit 48 controls the cleaning means 13 to clean the chips. In this embodiment, the cleaning processing unit 48 calculates the position (actual position within the machine tool 10) of a mesh area corresponding to a mesh image determined to require cleaning, and causes the cleaning means 13 to spray cleaning liquid toward the mesh area.

[0059] The position of the mesh area can be specified, for example, by making an image memory (not shown) that stores an area image correspond to a coordinate system within the machine tool 10. In this embodiment, the cleaning means 13 is configured with a programmable nozzle, and therefore sprays coolant liquid onto all mesh areas determined to require cleaning while changing the spray direction.

[0060] In this embodiment, the cleaning processor 48 updates the number of consecutive cleanings performed in the target area and the number of local consecutive cleanings performed in the same mesh area every time the cleaning operation is performed, both of which are stored in the cleaning count memory 35. The cleaning processor 48 performs cleaning when the number of consecutive cleanings and the number of local consecutive cleanings are less than a predetermined threshold. On the other hand, when at least one of the number of consecutive cleanings and the number of local consecutive cleanings reaches a predetermined threshold, the cleaning processor 48 executes a predetermined abnormality process to avoid a situation in which the automatic cleaning operation never ends.

[0061] The abnormality process is not particularly limited as long as it is a process for avoiding a situation in which the automatic cleaning operation never ends, and may, for example, stop the operation of the machine tool 10 and sound an alarm. Alternatively, the fact that the threshold has been reached may be recorded and notified, and cleaning of that mesh area may be suspended for a "cleaning suspension period" previously set by the user, while cleaning of other mesh areas and operation of the machine tool 10 may continue.

[0062] The threshold value for the number of consecutive cleanings and the threshold value for the number of consecutive local cleanings can be set individually by the user. As a method for notifying the above alarm, for example, a warning sound can be output from a speaker (not shown), but the method is not limited to this as long as it can notify the user. Specifically, a message can be displayed on the display input means 2, or a light (not shown) can be made to blink or light up. If the user is not around the machine tool 10, an error message can be notified to the user's mobile terminal or the like.

[0063] The teaching processing unit 49 confirms and corrects the judgment results by the chip information judgment unit 46. In this embodiment, the teaching processing unit 49 first reads out the judgment results of each mesh image from the judgment result storage unit 33, and displays them on the display input means 2 as shown in Fig. 8(a). Specifically, different colors or the like are applied to different judgment results, and each mesh image is displayed in an identifiable manner.

[0064] When displaying the judgment results, the chip information judgment unit 46 calculates the judgment probability for each mesh image, and for mesh images for which the judgment probability of any class is not very high, the judgment result is likely to be incorrect (the judgment result is not reliable), and the mesh images may be highlighted to assist the user in correcting the judgment. For example, in the example shown in Fig. 8(a), the images are classified into the following four groups according to the values ​​of the judgment probability P0 for class 0 (no chips) and the judgment probability P1 for class 1 (chips present).

[0065] P1 is 0-30% (P0 is 70-100%): Class 0 (no chips) - "confident" P1 is 30-50% (P0 is 50-70%): "Not confident" P1 is 50-70% (P0 is 30-50%): "Not confident" P1 is 70-100% (P0 is 0-30%): Class 1 (with chips) "confident"

[0066] Next, when the user who has checked the above judgment result touches and selects any mesh image that has been erroneously judged, and inputs cutting chip information as the correct judgment result, the teaching processing unit 49 acquires the judgment result and corrects the judgment result stored in the judgment result storage unit 33. Then, as shown in Fig. 8(b), the display input means 2 displays the corrected judgment result.

[0067] The additional learning unit 50 performs additional learning using new teacher data. In this embodiment, the additional learning unit 50 acquires learning parameters reflecting the correct judgment result by performing additional learning on the teacher data, which is a set of the judgment result (correct chip information) corrected by the teaching processing unit 49 and a mesh image, in the chip judgment model. Then, the additional learning unit 50 updates the judgment parameters in the judgment parameter storage unit 32 with the learning parameters as new judgment parameters.

[0068] In this embodiment, the additional learning unit 50 is provided in the local chip detection device 1, and additional learning is performed in the background, but the present invention is not limited to this configuration. That is, teacher data may be uploaded to a learning server on the network for additional learning, and the obtained learning parameters may be downloaded to be used as new judgment parameters.

[0069] In this embodiment, an automatic cleaning mode and a teaching mode are prepared as operation modes for the cleaning process by the chip detection device 1, and the device can be set to either operation mode. The automatic cleaning mode is a mode in which the cleaning process is automatically performed using the judgment result by the chip information judgment unit 46 as is. The teaching mode is a mode in which the cleaning process is performed using the judgment result corrected by the teaching processing unit 49.

[0070] Next, the operation of the chip detection device 1 of this embodiment and the machine tool 10 equipped with the same, as well as a chip detection method, will be described with reference to Fig. 9. In the following description, a case will be described in which the chips are cleaned after they are detected.

[0071] When chips are detected using the chip detection device 1 of this embodiment and the machine tool 10 equipped with the same, judgment parameters are stored in advance in the judgment parameter storage unit 32. At this time, first, one or more sample images of chips are prepared, and the sample images are divided into a predetermined mesh size. As a result, a large number of mesh images are obtained as training data for training the chip judgment model. Then, the training data is trained into the chip judgment model, and the optimized judgment parameters are stored in the judgment parameter storage unit 32.

[0072] In a system using machine learning, such as the chip detection device 1 of the present embodiment, it is necessary to prepare a large amount of training data for the learning model in advance and have it trained before operation. However, the task of preparing the training data is very time-consuming and laborious, which is a problem that acts as a barrier to introduction.

[0073] In this regard, the teacher data used in this embodiment is a mesh image and its class, and a large amount of teacher data is required to obtain a certain level of performance and versatility. However, the teacher data is easy to prepare for the following reasons. (1) A large number of mesh images, which are learning units, can be obtained from one area image. (2) Even for the same area image, different teacher data can be easily increased by simply performing simple processing such as shifting or rotating the division boundary of the mesh image. (3) Object detection algorithms such as Object Detection and Semantic Segmentation require position information of all chips reflected in the area image. In contrast, in this embodiment, by dividing the area into mesh sizes necessary and sufficient for cleaning, there is no need to specify the exact positions of the chips in the area image, and the process can be reduced to simple class classification. (4) When class classification is used as chip information, teacher data is created by simply assigning one numerical value (class) to each mesh image.

[0074] This makes it possible to input teaching data for detecting chips in at least a specific environment by simply capturing a few area images. In other words, it is possible to optimize for individual machines, environments, and processes with a relatively small number of teaching operations.

[0075] Next, in detecting chips, first, the target area setting unit 41 sets a target area where chips are to be detected (step S1), and then the drive control unit 42 controls the machine tool 10 to machine the workpiece (step S2). As a result, as shown in Fig. 1, chips are scattered on the table 12 and the workpiece, so the image data acquisition unit 43 acquires an area image of the target area from the imaging means 11 (step S3).

[0076] When the area image is acquired, the chip detection process according to the present invention is executed (step S4). The chip detection process according to step S4 will be described below with reference to FIG.

[0077] In the chip detection process, first, the mesh size selection unit 44 selects a mesh size (step S21). In this embodiment, the mesh size selection unit 44 selects one or more candidates that are likely to correspond to a mesh size that is large enough to recognize the shape of the chip and that simplifies the edge shape of the workpiece or jig that is the background of the chip. This makes it easier to distinguish the chip from the background and suppresses false detection, thereby improving detection accuracy.

[0078] Next, the mesh division unit 45 divides the area image into a plurality of mesh images using any of the mesh sizes selected by the mesh size selection unit 44 (step S22). As a result, when using the chip information, it is only necessary to specify the mesh area corresponding to each mesh image, and it is not necessary to specify the exact positions of all chips. This significantly reduces the amount of calculation required for chip detection.

[0079] In addition, in this embodiment, when multiple partial images are synthesized to obtain an area image, the mesh division unit 45 executes the above process (i) or (ii) depending on whether there is overlap at the boundary between the partial images. This prevents multiple different partial images from being mixed in the same mesh image. This reduces false detections caused by differences in brightness between partial images, positional deviations, and the like.

[0080] Next, the chip information determination unit 46 determines the chip information for one mesh image (step S23), and repeats this until the determination of all mesh images in the target area is completed (step S24: NO). This simplifies and improves the accuracy of the chip detection process, since it is only necessary to determine the chip information for mesh images divided into a predetermined mesh size.

[0081] When the determination is completed for all mesh images in the target area (step S24: YES), the chip information determination unit 46 determines whether the determination is completed for all mesh sizes in the target area (step S25). If there is a mesh size that has not yet been determined (step S25: NO), the process returns to step S21 and the division process (step S22) and the determination process (step S23) for that mesh size are repeated.

[0082] On the other hand, when the judgment for all mesh sizes is completed (step S25: YES), the chip information judgment unit 46 stores the judgment results for each mesh size in the judgment result storage unit 33 (step S26), and this process ends.

[0083] 9, it is determined whether the operation mode is set to the teaching mode (step S5), and if it is set to the automatic cleaning mode (step S5: NO), the process proceeds to step 7 described below. On the other hand, if it is set to the teaching mode (step S5: YES), the teaching processing unit 49 executes the teaching process (step S6). The teaching process in step S6 will be described below with reference to FIG. 11.

[0084] In the teaching process, first, the teaching processing unit 49 displays the judgment result on the display input means 2 (step S31) and accepts the selection of a mesh image by the user. At this time, if there are judgment results for multiple mesh sizes, the user is allowed to visually select the optimal mesh size, and by using this mesh size, the accuracy of the chip information judgment is improved. Next, it is determined whether or not any mesh image has been selected (step S32), and if no mesh image has been selected (step S32: NO), the selection of the mesh image is accepted unless the teaching mode is terminated (step S35: NO).

[0085] On the other hand, when any mesh image is selected by the user (step S32: YES), the teaching processing unit 49 prompts the user to input correct chip information for the selected mesh image, and acquires the chip information as a judgment result (step S33). As a result, the correct judgment result confirmed and corrected by the user's visual inspection is stored in the judgment result storage unit 33 (step S34), and the judgment result is displayed on the display input means 2 (step S35).

[0086] Thereafter, when an instruction to end the teaching mode is given (step S36: YES), the additional learning unit 50 starts additional learning in the background at a predetermined timing, acquires judgment parameters reflecting the corrected judgment results, and updates the judgment parameters in the judgment parameter storage unit 32 with the judgment parameters (step S37). By executing the teaching process as described above, the accuracy of the judgment results is improved, and by updating the judgment parameters for individual machines, environments, and processes, the chip judgment performance is optimized.

[0087] 9, after the teaching process (step S6) is performed, or in the automatic cleaning mode (step S5: NO), the cleaning necessity determination unit 47 determines whether or not cleaning of the chips is necessary (step S7). In this embodiment, the cleaning conditions are set by combining the chip information and the position of the mesh area, so that the cleaning process is performed efficiently according to the amount and position of the detected chips.

[0088] For example, the cleaning conditions may be set so that only mesh areas corresponding to mesh images determined to be class 2 (a lot of chips) among the above-mentioned three classes are cleaned, or the cleaning conditions may be set so that areas outside the table 12 are not cleaned. By setting such cleaning conditions, the time required for cleaning and the consumption of cleaning liquid are reduced.

[0089] If the result of the determination in step S7 is that cleaning is not necessary (step S7: NO), the cleaning processing unit 48 initializes the number of continuous cleanings and the number of local continuous cleanings (step S13). Then, unless an instruction to end processing is given (step S15: NO), the process returns to step S1 and the subsequent processes are repeated. This makes it possible to process the workpiece in a state where the cutting chips have been removed.

[0090] On the other hand, if the result of the determination in step S7 is that cleaning is necessary (step S7: YES), the cleaning processing unit 48 determines whether or not the number of consecutive cleanings has reached a predetermined threshold value (step S8). If the result of the determination is that the number of consecutive cleanings has reached the threshold value (step S8: YES), the cleaning processing unit 48 executes abnormality processing without performing a cleaning operation (step S14). This avoids a situation in which chips that should have been cleaned have simply moved to another mesh area and not actually been removed, and the automatic cleaning operation never ends.

[0091] Even if the result of the determination in step S8 is that the number of continuous cleanings is less than the threshold (step S8: YES), the cleaning processing unit 48 separately determines whether or not the number of local continuous cleanings has reached a predetermined threshold (step S9). Even if the result of the determination is that the number of local continuous cleanings has reached the threshold (step S9: YES), the cleaning processing unit 48 executes abnormality processing without performing a cleaning operation (step S14). This avoids cases where chips that are difficult to remove by cleaning because they are stuck to or caught on the workpiece or the like are tried to be cleaned multiple times, or cases where screws of a jig or dirt inside the machine are mistakenly detected as chips and cleaning is attempted forever.

[0092] On the other hand, if both the number of continuous cleanings and the number of local continuous cleanings are less than the threshold (step S8: NO, step S9: NO), the cleaning processing unit 48 calculates the mesh areas corresponding to each of the mesh images determined to require cleaning (step S10), and cleans each mesh area by the cleaning means 13 (step S11). This allows the chips to be automatically and efficiently cleaned according to the chip information. In addition, since only the mesh areas that require cleaning are cleaned, the amount of cleaning liquid sprayed is reduced, which is more economical and environmentally friendly than cleaning the entire interior of the machine.

[0093] Finally, the cleaning processing unit 48 updates the number of continuous cleanings and the number of local continuous cleanings stored in the cleaning count storage unit 35 (step S12), and then returns to step S3 to repeat the subsequent processes. As a result, unless the number of continuous cleanings and the number of local continuous cleanings reach the threshold value, the chip information is judged and cleaning is performed according to the cleaning conditions.

[0094] In this embodiment, when an instruction to end the processing is given (step S15: YES), this process ends.

[0095] The chip detection device 1, chip detection program 1a, and chip detection method of the present embodiment as described above provide the following effects. 1. Chips generated when a workpiece is machined by the machine tool 10 can be easily and highly accurately detected. 2. Chips can be detected with high accuracy using machine learning. 3. Chips can be easily detected using image features. 4. The judgment parameters can be optimized according to the usage environment and individual machining, and the chip detection performance can be improved. 5. The influence of the background of the chips can be reduced, and erroneous detection can be suppressed. 6. False detection can be suppressed when multiple partial images are combined and used as an area image. 7. Chips can be automatically and efficiently washed according to the chip information. 8. Chips can be efficiently washed according to the amount and position of the detected chips. 9. It is possible to prevent the washing operation from being repeated indefinitely due to erroneous detection of chips, etc.

[0096] The chip detection device 1, the machine tool 10 including the same, and the chip detection method according to the present invention are not limited to the above-described embodiment, and can be modified as appropriate.

[0097] For example, in the above-described embodiment, the chips are washed using the chip information, but the present invention is not limited to this configuration, and the chip detection device 1 does not need to have a washing function. If the chip washing function is not provided, the washing condition storage unit 34, the washing count storage unit 35, the washing necessity determination unit 47, and the washing processing unit 48 are unnecessary, and in the flowchart shown in Fig. 9, the processes from step S7 to step S14 are unnecessary.

[0098] In addition to being used as information for cleaning chips, the chip information can be used in various other situations. For example, if the chip information can identify a mesh area where a large amount of chips or chips tend to accumulate frequently, the chip information can be used as reference information for setting cleaning conditions such as cleaning the mesh area early or constantly. Similarly, the information on mesh areas where chips tend to accumulate can be used by machine tool vendors to design machines that are less likely to accumulate chips.

[0099] In the above-described embodiment, the chip detector 1 has a teaching mode, but this is not an essential configuration. If the teaching mode is not required, there is no need to provide a teaching processing unit 49, and the processes of steps S5 and S6 in the flowchart shown in Fig. 9 are not required. Also, the teaching process shown in Fig. 11 is not required.

[0100] Furthermore, in the above-described embodiment, the cleaning operation is controlled using both the continuous cleaning count and the local continuous cleaning count, but the present invention is not limited to this configuration, and the cleaning operation may be controlled using only one of them. In this case, in the flowchart shown in FIG. 9, the process of either step S8 or step S9 is not necessary.

[0101] In the above-described embodiment, the learning parameters are used as the judgment parameters, but the present invention is not limited to this configuration and may be threshold values ​​for classifying the chip information based on various feature quantities possessed by the mesh image. The feature quantities are not particularly limited as long as they are feature quantities capable of expressing the features of the chips, and various feature quantities used in image processing, such as the degree of complexity of the edge components in the mesh image, the proportion of linear components in the edge components in the mesh image, the brightness of the mesh image, and brightness statistics, may be used.

[0102] The threshold value as a judgment parameter may be set by optimizing the threshold value by calculating the feature score for each mesh image obtained by dividing the sample image and aggregating the scores for each classification of cutting information, or may be set by combining multiple feature values.

[0103] In addition, in the above-described embodiment, the chip information is determined as a "class" indicating the ease of cleaning the chips, but it is not limited to this and may be the presence or absence of chips, the amount of chips, the shape of the chips, the type of chips, or a combination of these, etc. For example, if the amount and shape of the chips are used as the chip information, it is possible to set a cleaning condition such that even in a mesh area with a small amount of chips, chips with a shape that is difficult to clean are cleaned.

[0104] Furthermore, in the above-described embodiment, the chip detection device 1 is configured to be able to additionally learn new teacher data locally, but the present invention is not limited to this configuration, and teacher data may be transmitted to a learning device connected to the chip detection device 1 via a network for additional learning. Also, teacher data for additional learning may be collected from multiple chip detection devices 1 in a learning device, and learning parameters (judgment parameters) obtained by additional learning may be shared to improve versatility.

[0105] Furthermore, learning images for deriving the judgment parameters used in the above-mentioned chip detection device 1 may be generated by a learning image synthesis device 1200 as shown in Fig. 12. The learning image synthesis device 1200 has a chip image storage unit 1201 that stores chip images representing various chips, and a background image storage unit 1202 that stores background images of various machine interiors. Furthermore, the learning image synthesis device 1200 has an image generation unit 1203 that randomly synthesizes the chip images and the background images to generate learning images.

[0106] The chip image may be an image of actual chips without a background. The background image may be an image of the actual interior of the machine. The image generating unit 1203 synthesizes the chip images with a photo of the interior of the machine, randomly but as realistically as possible, to create multiple environmental patterns and automatically create a large amount of training data. This makes it easy to prepare training data.

[0107] Furthermore, the image generating unit 1203 creates scratches and the like on the background image. Prepare as many patterns as possible.

[0108] The training images generated by the training image synthesis device 1200 are provided to a training device 1250 and stored in a training image storage unit 1251. Then, the training images are used to derive training parameters by a determination parameter derivation unit 1252. The determination parameters are provided to a chip determination model derivation unit 1253 and used to generate a chip determination model. The learning image synthesis device 1200 may provide the chip determination model to the chip detection device 1 via the network 1260.

[0109] The learning image synthesis device 1200 and the learning device 1250 may be provided within the chip detection device 1. In that case, however, the chip detection device 1 needs to be equipped with resources such as a high-speed GPU.

[0110] Furthermore, an actual image of the inside of the machine tool may be sent to the learning image synthesis device 1200 via the network 1260 and stored in the background image storage unit 1202. In this case, it becomes possible to generate more accurate teacher data (learning images).

[0111] The chip image storage unit 1201 preferably stores images of various chips having different colors, shapes, and sizes, and further preferably stores images of chips with and without coolant.

[0112] The background image stored in the background image storage unit 1202 is classified into areas with a lot of chips, areas with a small amount of chips, and areas with no chips. When combining with the chip image, the weighting for each area may be changed.

[0113] The image generating unit 1203 may add processing such as peeling paint, scratches, stains, etc. to the background image and then combine it with the chip image. When combining, the chip image may be resized or rotated. It is preferable that the image generating unit 1203 generates multiple types of learning images by changing the overall color, brightness, lighting condition, etc. of the combined image.

[0114] At the time of synthesis, the image generating unit 1203 prepares and pastes answer data representing the presence or absence of cutting chips for each region of the mesh image.

[0115] The learning image synthesis device 1200 creates a large amount of training data by combining background images of various models with various chip images, changing the weighting and density, and applying various effects (brightness, coolant mist application). The learning image synthesis device 1200 may generate learning images by classifying them by model.

[0116] According to the above configuration, training data can be generated very efficiently.

[0117] In the above-described embodiment, the chip detection device 1 is realized as one function of the numerical control device, but the present invention is not limited to this configuration. In other words, the chip detection device 1 may be configured by a computer that is separate and independent from the numerical control device.

[0118] The above-described embodiment is illustrative and not restrictive in all respects, and those skilled in the art can appropriately modify and change the embodiment. The scope of the present invention is indicated by the claims, not the above-described embodiment. Furthermore, the scope of the present invention includes modifications from the embodiment within the scope equivalent to the claims.

[0119] This application claims priority based on Japanese Patent Application No. 2019-031127, filed on February 25, 2019, the disclosure of which is incorporated herein in its entirety.

Claims

[Claim 1] A chip detection device for detecting chips generated when a workpiece is machined by a machine tool, a mesh division unit that divides at least a part of an area image obtained by photographing a target area for detecting the chips into a plurality of identical mesh images having a mesh size that allows analysis of edges constituting the outline of each chip from a state in which a large number of chips are piled up and overlapped, and simplifies analysis of the edge shape of the background; a chip information determination unit that performs an analysis of the mesh image including an analysis of the edge and an analysis of the background, calculates one feature value for the entire mesh image, and determines the chip information of the mesh image based on a determination parameter that is preset for determining chip information related to the chip and the feature value; having the judgment parameters are learning parameters for classifying feature quantities calculated from a mesh image into classes of the chip information, the learning parameters being obtained by prior machine learning using teacher data consisting of a set of a mesh image and the chip information, The chip detection device, wherein the chip information determination unit inputs a mesh image into a chip determination model to which the determination parameters have been applied, and determines, for each mesh image of the same size at the mesh size, chip information corresponding to a class classified based on features calculated from the input mesh image.

Images