Machine learning device, cleaning prediction device, and cleaning system

Abstract

The present application provides a learned model that eliminates confirmation of a visual sensor in processing, and prevents accumulation of chips with a small cleaning frequency. A machine learning device includes: an input data acquisition unit that acquires input data including: an arbitrary machining condition of an arbitrary machine tool for an arbitrary workpiece, a cleaning condition for cleaning a machining machine interior after the workpiece is machined according to the machining condition by an arbitrary cleaning device, and state information indicating a degree of dirt in the machining machine interior before machining based on the machining condition; a label acquisition unit that acquires state information of the machining machine interior after machining with the machining condition included in the input data and cleaning with the cleaning condition as label data; and a learning unit that performs supervised learning using the input data acquired by the input data acquisition unit and the label data acquired by the label acquisition unit, and generates a learned model.

Description

TECHNICAL FIELD

[0001] The present application relates to a machine learning device, a cleaning prediction device, and a cleaning system. BACKGROUND

[0002] In a case where a cleaning device (for example, a device fixed inside a machine tool, or a device held by a robot or the like and movable) cleans the inside of a machine, the inside of the machine is generally cleaned at a predetermined frequency.

[0003] In Patent Literatures 1 to 3, a technique is proposed in which the amount of accumulation of chips is measured using a vision sensor and it is determined whether or not to perform cleaning (sweeping).

[0004] PRIOR ART DOCUMENTS

[0005] PATENT LITERATURES

[0006] Patent Literature 1: Japanese Patent Application Publication No. 2016-168661

[0007] Patent Literature 2: Japanese Patent Application Publication No. 2017-013182

[0008] Patent Literature 3: Japanese Patent Application Publication No. 2017-104943 SUMMARY

[0009] PROBLEMS TO BE SOLVED BY THE INVENTION

[0010] In a case where the inside of the machine is cleaned at a predetermined frequency, the frequency is determined by an operator, and is determined empirically based on the shape of the chips, the number and the amount of accumulation of the workpieces after machining.

[0011] In addition, the method of confirming the amount of accumulation of the chips by the vision sensor is also confirmed by the vision sensor each time a workpiece is machined, or after a plurality of workpieces are machined, it is determined by the operator based on experience whether or not to confirm by the vision sensor.

[0012] In addition, during the cleaning (sweeping) of the inside of the machine or during the confirmation of the amount of accumulation of the chips by the vision sensor, the workpieces cannot be machined by the machine.

[0013] Therefore, in order to improve the productivity of the workpieces, it is necessary to reduce the time spent other than machining as much as possible, and it is necessary to reduce the number of cleanings of the inside of the machine, the time of confirmation by the vision sensor as much as possible.

[0014] Therefore, it is desirable to eliminate the confirmation by the vision sensor during machining, and to generate a learned model that prevents the accumulation of chips at a very low cleaning frequency.

[0015] MEANS FOR SOLVING THE PROBLEMS

[0016] (1) One embodiment of the machine learning device of the present disclosure includes: an input data acquisition unit that acquires input data including: an arbitrary machining condition for an arbitrary machine tool with respect to an arbitrary workpiece, a cleaning condition for cleaning a machining chamber of the machine tool after machining the workpiece according to the machining condition by an arbitrary cleaning device, and state information indicating a degree of dirt in the machining chamber before machining based on the machining condition; a label acquisition unit that acquires state information of the machining chamber after machining according to the machining condition included in the input data and cleaning according to the cleaning condition as label data; and a learning unit that performs supervised learning using the input data acquired by the input data acquisition unit and the label data acquired by the label acquisition unit, and generates a learned model.

[0017] (2) One embodiment of the cleaning prediction device of the present disclosure includes: a learned model generated by the machine learning device of (1); an input unit that inputs, before machining by a machine tool, a machining condition to be performed next, a cleaning condition of a predetermined cleaning, and state information of a machining chamber of the machine tool at present; and a prediction unit that inputs the machining condition to be performed next, the cleaning condition, and the state information of the machining chamber at present input by the input unit to the learned model, and predicts state information of the machining chamber after machining according to the machining condition to be performed next and cleaning according to the cleaning condition.

[0018] (3) One embodiment of the cleaning system of the present disclosure includes the cleaning prediction device of (2) and a cleaning device.

[0019] Effects of Invention

[0020] According to one embodiment, it is possible to eliminate confirmation of a visual sensor during machining, and generate a learned model that prevents chip accumulation with a small cleaning frequency. BRIEF DESCRIPTION OF DRAWINGS

[0021] Figure 1 is a functional block diagram that shows a functional configuration example of the cleaning system of the first embodiment.

[0022] Figure 2A is a drawing that shows an example of chip accumulation in a machining chamber of a machine tool viewed from the side.

[0023] Figure 2B is a drawing that shows an example of a predetermined cleaning of a machining chamber viewed from the upper surface.

[0024] Figure 3 is a drawing that shows an example of a learned model provided to the cleaning prediction device of Figure 1

[0025] Figure 4 ​is a diagram indicating an example of the degree of dirt in the machining device.

[0026] Figure 5 is a flowchart illustrating the prediction processing of the cleaning prediction device in the use stage.

[0027] Figure 6 is a functional block diagram indicating an example of the functional structure of the cleaning system of the second embodiment.

[0028] Figure 7 is a flowchart illustrating the prediction processing of the cleaning prediction device in the use stage.

[0029] Figure 8A is a diagram indicating an example of the predetermined cleaning at the first time.

[0030] Figure 8B is a diagram indicating an example of the predetermined cleaning at the second time.

[0031] Figure 9 is a diagram indicating an example of the structure of the cleaning system.

[0032] Figure 10 is a diagram indicating an example of the structure of the cleaning system. DETAILED DESCRIPTION

[0033] <First Embodiment>

[0034] The structure of the present embodiment will be described in detail using the drawings. Here, a case is exemplified in which the cleaning timing of the cleaning for which the workpiece production number is the largest is decided in accordance with the machining condition to be performed next, the cleaning condition of the predetermined cleaning, and the degree of dirt in the machining device of the current machine tool. In addition, as described later, the present application can also be applied to a case in which the number of cleanings of the predetermined cleaning that is the least among the cleanings of the machining device after machining performed in the machining condition to be performed next is decided in accordance with the machining condition to be performed next, the cleaning condition of the predetermined cleaning, and the degree of dirt in the machining device of the current machine tool.

[0035] Figure 1 is a functional block diagram indicating an example of the functional structure of the cleaning system of the first embodiment. As shown in Figure 1 , the cleaning system 1 has a machine tool 10, a cleaning prediction device 20, and a machine learning device 30.

[0036] The machine tool 10, the cleaning prediction device 20, and the machine learning device 30 can also be directly connected to each other via a connection interface not shown. In addition, the machine tool 10, the cleaning prediction device 20, and the machine learning device 30 can also be connected to each other via a network not shown such as a LAN (Local Area Network), the Internet, or the like. In this case, the machine tool 10, the cleaning prediction device 20, and the machine learning device 30 are provided with a communication section not shown for communicating with each other through the connection. In addition, as described later, the machine tool 10 can also include the cleaning prediction device 20 and the machine learning device 30.

[0037] The machine tool 10 is a machine tool known to those skilled in the art, and is assembled with a cleaning device 101, a measurement device 102, and a control device 103. The machine tool 10 is operated based on an operation instruction of the control device 103. In addition, the cleaning device 101 and the control device 103 can also be devices independent of the machine tool 10.

[0038] The cleaning device 101 is a cleaning device known to those skilled in the art. For example, the cleaning device 101 sprays a cleaning liquid such as a coolant from a cleaning nozzle (not shown) included in the cleaning device 101, based on a control instruction generated by the control device 103 executing a cleaning program, to operate a pump (not shown) included in the cleaning device 101, thereby cleaning the machining chamber of the machine tool 10 with a predetermined cleaning.

[0039] Here, the predetermined cleaning refers to cleaning in which cleaning conditions (for example, an angle of the cleaning nozzle, an amount of spraying, a location in the machining chamber, a cleaning path, a cleaning time, and the like) for cleaning the machining chamber according to a type of the machine tool 10 and a machining content are set in advance in the cleaning program. Also, in the cleaning conditions, a number of cleanings of the predetermined cleaning is set as a unit of basic cleaning of the predetermined cleaning.

[0040] Figure 2A is a view showing an example of a state of accumulation of chips in the machining chamber of the machine tool 10 viewed from the side. Figure 2B is a view showing an example of the predetermined cleaning of the machining chamber viewed from the upper surface. In addition, in Figure 2B , the lower side is the door side of the machine tool 10.

[0041] As shown in Figure 2B , the cleaning device 101, for example, repeatedly performs the predetermined cleaning in which the cleaning nozzle is moved on the cleaning paths of (1) to (6) according to the cleaning program for the number of cleanings included in the cleaning conditions, thereby removing the chips in the machining chamber of the machine tool 10.

[0042] In addition, the cleaning device 101 can be installed to a robot (not shown) or built in a robot (not shown) to clean the machining chamber of the machine tool 10. In this case, the machine tool 10 can include a robot control device (not shown).

[0043] The measuring device 102 can measure, for example, the degree of dirt in the machining chamber of the machine tool 10 as a whole or at least one point of interest in the machining chamber after the workpiece is machined by the machine tool 10 from a predetermined position. Specifically, the measuring device 102 can calculate the difference in the amount (area, volume, etc.) of each portion of the machining chamber to be cleaned, the proportion (%) of the difference in brightness, etc. based on the difference from the measurement value of the whole or at least one point of interest in the machining chamber obtained by measuring the machining chamber to which no dirt is attached in advance, using a known method (for example, Japanese Patent Application Publication No. 2018-153872, etc.), as the degree of dirt in the machining chamber. In addition, as the measuring device 102, a visual sensor, an infrared laser, an ultrasonic wave meter, etc. can be used.

[0044] The control device 103 is a numerical control device known to those skilled in the art, generates an operation command based on the control information, and transmits the generated operation command to the machine tool 10, the cleaning device 101, and the measuring device 102. Thereby, the control device 103 controls the operation of the machine tool 10, the cleaning device 101, and the measuring device 102. In addition, the control device 103 can output the control information to the cleaning prediction device 20 via a communication section of the machine tool 10 not shown. In addition, the control information includes the machining program, the cleaning program, and the values of the parameters set in the control device 103.

[0045] In addition, the control device 103 can output, to the cleaning prediction device 20 via the communication section of the machine tool 10 not shown, state information indicating the degree of dirt in the machining chamber of the machine tool 10 currently measured by the measuring device 102.

[0046] In the operation stage, the cleaning prediction device 20 can acquire the machining condition to be performed next, the cleaning condition of the predetermined cleaning, and the state information of the machining chamber of the machine tool 10 currently from the control information of the control device 103 before the machine tool 10 machines the workpiece. The cleaning prediction device 20 inputs the acquired machining condition, cleaning condition, and state information of the machining chamber to the learned model provided by the machine learning device 30 described later, and thereby can predict the state information of the machining chamber of the machine tool 10 after machining is performed with the machining condition to be performed next and cleaning is performed with the cleaning condition.

[0047] In addition, a main cause of the degree of dirt in the machining device is the chips that are generated in large quantities and accumulated as a result of machining the workpiece using the tool of the machine tool 10. Therefore, in order to take into account the generation of chips, the type of the machine tool 10, the material of the workpiece (ferrous or non-ferrous), the type of the cutting tool, the spindle speed, the feed speed, the depth of cut, and the number of workpieces produced can also be included in the machining conditions. In addition, the type of the cutting tool is related to the shape of the chips, and the spindle speed, the feed speed, and the depth of cut are related to the amount of chips.

[0048] In addition, as described later, the degree of dirt in the machining device is included in the state information.

[0049] Here, before explaining the cleaning prediction device 20, the degree of dirt in the machining device and the machine learning for generating the learned model are explained.

[0050] <DEGREE OF DIRT IN MACHINING DEVICE>

[0051] As described above, the degree of dirt in the machining device refers to the amount of difference (area, volume, etc.) of each portion of the cleaning target in the machining device calculated by the measuring device 102 based on the difference in the measurement values of the entire machining device or at least one point of interest obtained by measuring the machining device in advance without dirt attached, the proportion (%) of the brightness of the difference, and the like. That is, the degree of dirt in the machining device of the machine tool 10 immediately after cleaning is "0%". Also, the degree of dirt in the machining device becomes a value corresponding to the amount of difference (area, volume, etc.) of each portion of the cleaning target in the machining device measured by the measuring device 102 as the machining of the workpiece by the machine tool 10 is repeated, the proportion (%) of the brightness of the difference. In addition, the maximum allowable limit value set in advance is set to γ.

[0052] <LEARNED MODEL>

[0053] The machine learning device 30, for example, obtains, in advance, as input data, any machining condition of any workpiece in machining by any machine tool, a cleaning condition for cleaning the machining device after machining the workpiece according to the machining condition by any cleaning device, and state information indicating the degree of dirt in the machining device before machining based on the machining condition.

[0054] In addition, the machine learning device 30 obtains, as a label (correct answer), state information (degree of dirt in the machining device) of the machining device after machining with the machining condition in the obtained input data and cleaning with the cleaning condition.

[0055] The machine learning device 30 performs supervised learning on the training data of the combination of the obtained input data and the label, and constructs a learned model described later.

[0056] Thus, the machine learning device 30 can provide the constructed learned model to the cleaning prediction device 20 .

[0057] The machine learning device 30 will be described in detail.

[0058] like Figure 1 As shown, the machine learning device 30 includes an input data acquisition unit 301 , a label acquisition unit 302 , a learning unit 303 , and a storage unit 304 .

[0059] During the learning phase, the input data acquisition unit 301 acquires, via a communication unit (not shown), from the control device 103 or the like, arbitrary machining conditions for an arbitrary workpiece being machined by an arbitrary machine tool, cleaning conditions for cleaning the interior of the machining machine using an arbitrary cleaning device after machining the workpiece according to the machining conditions, and status information indicating the degree of contamination within the machining machine before machining according to the machining conditions. The input data acquisition unit 301 outputs the acquired input data to the storage unit 304.

[0060] The label acquisition unit 302 acquires data indicating the degree of dirt in the processing machine after processing under the processing conditions in the input data and cleaning under the cleaning conditions as label data (correct answer data), and outputs the acquired label data to the storage unit 304 .

[0061] The learning unit 303 accepts the combination of the above-mentioned input data and labels as training data, and uses the accepted training data to perform supervised learning, thereby constructing a learned model 250. The learned model 250 predicts the degree of dirt in the processing machine after processing under the next processing conditions and cleaning under the cleaning conditions based on the next processing conditions, cleaning conditions (for example, the number of cleanings to be performed), and the current degree of dirt in the processing machine.

[0062] Then, the learning unit 303 provides the constructed learned model 250 to the cleaning prediction device 20 .

[0063] Furthermore, it is preferable to prepare a plurality of training data for supervised learning. For example, the training data may be obtained from the control devices 103 of the machine tools 10 actually operated at various locations such as a customer's factory.

[0064] Figure 3 It means to Figure 1 FIG. 2 is an example of a learned model 250 provided by the cleaning prediction device 20. Figure 3As illustrated, the learned model 250 exemplifies a multilayer neural network that takes, as an input layer, the kind of machine tool, the material of workpiece, the machining condition to be performed next, the cleaning condition of the number of cleaning times of predetermined cleaning, and the state information of the degree of dirt in the machining machine at present, and takes, as an output layer, data indicating the "degree of dirt" in the machining machine after machining is performed in the machining condition to be performed next and cleaning is performed in the cleaning condition.

[0065] Here, as described above, in the machining condition to be performed next, the kind of machine tool, the material of workpiece (ferrous or non-ferrous), the kind of cutting tool, the spindle speed, the feed speed, the cutting amount, and the workpiece production quantity are included.

[0066] In addition, as the cleaning condition, the number of cleaning times of predetermined cleaning is set, but it can be the cleaning time (= (cleaning time of predetermined cleaning) x (number of cleaning times)).

[0067] In addition, the learning unit 303 can further perform supervised learning on the learned model 250 to update the once-constructed learned model 250 in a case where new training data is obtained after the learned model 250 is constructed.

[0068] The above-described supervised learning can be performed by online learning, can be performed by batch learning, or can be performed by mini-batch learning.

[0069] Online learning refers to a learning method in which supervised learning is performed immediately when training data is generated as the machine tool 10 performs machining of a workpiece. In addition, batch learning refers to a learning method in which a plurality of training data corresponding to repetition is collected during a period in which training data is generated as the machine tool 10 repeatedly performs machining of a workpiece, and supervised learning is performed using all of the collected training data. Furthermore, mini-batch learning refers to a learning method in which supervised learning is performed each time training data of a certain degree intermediate between online learning and batch learning is accumulated.

[0070] The storage unit 304 is a RAM (Random Access Memory) or the like, and stores input data obtained by the input data obtaining unit 301, label data obtained by the label obtaining unit 302, the learned model 250 constructed by the learning unit 303, and the like.

[0071] The above describes machine learning for generating the learned model 250 possessed by the cleaning prediction device 20.

[0072] Next, the cleaning prediction device 20 in the application stage will be described.

[0073] <Cleaning Prediction Device 20 in Application Stage>

[0074] like Figure 1 As shown, the cleaning prediction device 20 in the operation phase is configured to include an input unit 201 , a prediction unit 202 , a decision unit 203 , a notification unit 204 , and a storage unit 205 .

[0075] In addition, in order to achieve Figure 1 The cleaning prediction device 20 includes a CPU (Central Processing Unit) and other arithmetic processing devices (not shown). Furthermore, the cleaning prediction device 20 includes a ROM (Read Only Memory) storing various control programs, an auxiliary storage device (not shown) such as an HDD, and a main storage device (not shown) such as a RAM for temporarily storing data required after the arithmetic processing device executes the programs.

[0076] Furthermore, in the cleaning prediction device 20, the calculation 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 calculation processing based on these OS and application software. Based on the calculation results, the cleaning prediction device 20 controls each hardware. In this way, the cleaning prediction device 20 realizes the cleaning prediction device 20 based on the cleaning prediction result. Figure 1 That is, the cleaning prediction device 20 can be realized by the cooperation of hardware and software.

[0077] Before the machine tool 10 begins machining a workpiece, the input unit 201 inputs, for example, the next machining conditions, cleaning conditions with a predetermined cleaning frequency, and current status information on the degree of contamination within the machining center based on control information from the control device 103. The input unit 201 outputs the acquired status information on the next machining conditions, cleaning conditions with a predetermined cleaning frequency, and current status information on the degree of contamination within the machining center to the prediction unit 202.

[0078] The prediction unit 202 inputs the processing conditions to be performed next, the cleaning conditions of the cleaning times to be scheduled, and the current state information of the degree of dirt in the processing machine into the prediction unit 202. Figure 3 The learned model 250 predicts the state information of the “degree of dirt” in the processing machine after processing under the processing conditions to be performed next and cleaning under the cleaning conditions.

[0079] Furthermore, in order for the decision unit 203 (described later) to determine the cleaning timing that maximizes the number of workpieces produced, the number of cleaning times in the cleaning conditions is set to "0." Specifically, the prediction unit 202 predicts the degree of contamination within the processing machine before the cleaning timing at which cleaning is required.

[0080] The decision section 203 decides the cleaning timing at which the number of workpieces produced is the largest, based on the state information of the degree of dirt in the machining device predicted by the prediction section 202.

[0081] Specifically, the decision section 203 determines whether it is the cleaning timing at which cleaning of the machining device is required, based on a comparison of the predicted value of the degree of dirt in the machining device with a threshold value set in advance.

[0082] Figure 4 is a graph showing an example of the degree of dirt in the machining device. As shown in Figure 4 for example, the degree of dirt in the machining device is "0%" immediately after cleaning, and increases as the machining device 10 repeatedly machines workpieces. For example, in a case where the degree of dirt in the machining device is less than a predetermined ratio a, the machining device 10 is able to machine workpieces with the desired machining quality. On the other hand, in a case where the degree of dirt is a predetermined ratio a or more, the machining device 10 is unable to machine workpieces with the desired machining quality.

[0083] In the following description, the degree of dirt in the machining device, i.e., the predetermined ratio a, is set as the threshold value. Further, the predetermined ratio a (hereinafter, also referred to as "threshold value a") can be appropriately set in accordance with the environment of the factory in which the machining device 10 is installed, the machining contents of the machining device 10, and the like.

[0084] Further, as will be described later, in a case where the degree of dirt in the machining device is less than a predetermined ratio β, it can be considered that there is no cutting chips in the machining device, and the degree of dirt in the machining device, i.e., the predetermined ratio β, is set as a threshold value indicating a state in which the machining device does not need to be cleaned and there are no cutting chips. Here, γ > a > β.

[0085] The decision section 203 determines whether the degree of dirt in the machining device predicted by the prediction section 202 is less than the threshold value a. The decision section 203, for example, in a case where the predicted degree of dirt in the machining device is less than the threshold value a, causes the number of workpieces produced for each of the machining conditions to be performed next to be increased by one (a predetermined number), causes the prediction section 202 to predict the degree of dirt in the machining device after machining is performed with the machining conditions to be performed next after the number of workpieces produced is increased and cleaning is performed with the cleaning condition, and decides the cleaning timing at which the number of workpieces produced is the largest within a range in which the predicted degree of dirt in the machining device is less than the threshold value a.

[0086] On the other hand, the decision section 203, in a case where the predicted degree of dirt in the machining device is equal to or greater than the threshold value a, causes the workpiece production number of the machining condition to be performed next to be reduced by one (a predetermined number) each time, causes the prediction section 202 to predict the degree of dirt in the machining device after machining is performed with the machining condition to be performed next after the workpiece production number is reduced and cleaning is performed with the cleaning condition, and decides the cleaning timing in which the workpiece production number is the largest in the range in which the predicted degree of dirt in the machining device is less than the threshold value a.

[0087] The notification section 204 can also output the cleaning timing in which the workpiece production number is the largest, decided by the decision section 203, to an output device (not shown) such as a liquid crystal display included in the machine tool 10 and / or the control device 103.

[0088] Thus, the notification section 204 can recommend the cleaning timing in which the workpiece production number is the largest, decided by the decision section 203, to the user (operator). In addition, the notification section 204 can also notify by voice via a speaker (not shown).

[0089] The storage section 205 is a ROM, HDD, or the like, and can also store the learned model 250 together with various control programs.

[0090] < Prediction processing of the cleaning prediction device 20 in the application stage >

[0091] Next, the operation of the prediction processing of the cleaning prediction device 20 of the present embodiment will be described.

[0092] Figure 5 is a flowchart illustrating the prediction processing of the cleaning prediction device 20 in the application stage.

[0093] In step Sll, before the machining of the workpiece is performed by the machine tool 10, the input section 201 inputs the machining condition to be performed next, the cleaning condition of the cleaning number "0" of the predetermined cleaning, and the state information of the degree of dirt in the machining device at present from the control information of the control device 103.

[0094] In step S12, the prediction section 202 inputs the machining condition to be performed next, the cleaning condition of the cleaning number "0" of the predetermined cleaning, and the state information of the degree of dirt in the machining device at present to the learned model 250, and predicts the state information of the "degree of dirt" in the machining device after machining is performed with the machining condition to be performed next and cleaning is performed with the cleaning condition.

[0095] In step S13, the decision section 203 decides the cleaning timing in which the number of workpieces is the largest in the range in which the predicted value of the degree of contamination in the machine after machining and cleaning is predicted in step S12 is less than the threshold value a. For example, in a case where the predicted degree of contamination in the machine is less than the threshold value a, the decision section 203 causes the prediction section 202 to predict the degree of contamination in the machine after machining under the machining condition to be performed next with the number of workpieces increased by one (a predetermined number) and cleaning under the cleaning condition, and decides the cleaning timing in which the number of workpieces is the largest in the range in which the predicted degree of contamination in the machine is less than the threshold value a.

[0096] On the other hand, in a case where the predicted degree of contamination in the machine is equal to or more than the threshold value a, the decision section 203 causes the prediction section 202 to predict the degree of contamination in the machine after machining under the machining condition to be performed next with the number of workpieces decreased by one (a predetermined number) and cleaning under the cleaning condition, and decides the cleaning timing in which the number of workpieces is the largest in the range in which the predicted degree of contamination in the machine is less than the threshold value a.

[0097] In step S14, the notification section 204 notifies the cleaning timing in which the number of workpieces is the largest, which is decided in step S13.

[0098] According to the above, the cleaning prediction device 20 of the first embodiment inputs the machining condition to be performed next involved in machining of a workpiece by the machine tool 10, the cleaning condition of the predetermined number of cleanings of cleaning, and the state information including the degree of contamination in the machine at present to the learned model 250 before the machining, predicts the state information of the degree of contamination in the machine after machining and cleaning. Then, the cleaning prediction device 20 decides the cleaning timing in which the number of workpieces is the largest based on comparison of the predicted value of the degree of contamination in the machine predicted with the threshold value.

[0099] Thus, the cleaning prediction device 20 eliminates confirmation of the vision sensor during machining, and can prevent accumulation of chips with a very small frequency of cleaning.

[0100] In addition, the cleaning prediction device 20 uses the learned model 250, and thus the operator does not need to judge necessity of cleaning in the machine, and can reduce the burden on the operator.

[0101] The first embodiment has been described above.

[0102] <Second Embodiment>

[0103] Next, the second embodiment will be described. As described above, the cleaning prediction device 20 of the first embodiment determines the cleaning timing at which the workpiece production number is the largest, based on the machining condition to be performed next, the cleaning condition of the scheduled cleaning, and the degree of contamination in the machining machine of the current machine tool. In contrast, the cleaning prediction device 20a of the second embodiment differs from the first embodiment in that it determines the cleaning number of the scheduled cleaning that is the least among cleanings in the machining machine after machining performed with the machining condition to be performed next, based on the machining condition to be performed next, the cleaning condition of the scheduled cleaning, and the degree of contamination in the machining machine of the current machine tool. In addition, the predicted value of the degree of contamination in the machining machine after machining performed according to the machining condition is less than the threshold value a. In addition, by setting the machining condition in which the workpiece production number is the largest in a range in which the degree of contamination in the machining machine is less than the threshold value a, it is possible to determine the cleaning number of the scheduled cleaning that is the least among cleanings in the machining machine at the cleaning timing at which the workpiece production number is the largest.

[0104] Thus, the cleaning prediction device 20a of the second embodiment eliminates the confirmation of the vision sensor during machining, and it is possible to prevent the accumulation of chips with a very small cleaning frequency.

[0105] Next, the second embodiment will be described.

[0106] Figure 6 is a functional block diagram that shows an example of the functional structure of the cleaning system of the second embodiment. In addition, elements having the same functions as the elements of the cleaning system 1 of the first embodiment are labeled with the same reference numerals, and detailed description will be omitted. Figure 1

[0107] As shown in Figure 6 , the cleaning system 1A has a machine tool 10, a cleaning prediction device 20a, and a machine learning device 30.

[0108] The machine tool 10 includes a cleaning device 101, a measurement device 102, and a control device 103.

[0109] The cleaning device 101, the measurement device 102, and the control device 103 have the same functions as the cleaning device 101, the measurement device 102, and the control device 103 of the first embodiment.

[0110] The machine learning device 30 includes an input data acquisition section 301, a label acquisition section 302, a learning section 303, and a storage section 304.

[0111] The input data acquisition section 301, the label acquisition section 302, the learning section 303, and the storage section 304 have the same functions as the input data acquisition section 301, the label acquisition section 302, the learning section 303, and the storage section 304 of the first embodiment.

[0112] ​The cleaning prediction device 20a includes an input section 201, a prediction section 202, a decision section 203a, a notification section 204, and a storage section 205.

[0113] The input section 201, the prediction section 202, the notification section 204, and the storage section 205 have the same functions as the input section 201, the prediction section 202, the notification section 204, and the storage section 205 of the first embodiment.

[0114] The decision section 203a decides the minimum number of cleanings of the predetermined cleaning in the cleaning of the machining device after the machining under the machining condition to be performed next, based on the state information of the degree of the dirt in the machining device after the machining and the cleaning predicted by the prediction section 202.

[0115] Specifically, the decision section 203a determines whether there is a chip in the machining device based on a comparison between the predicted value of the degree of the dirt in the machining device and the threshold value β that indicates that there is no chip in the machining device. The decision section 203a determines that there is a chip in the machining device when the degree of the dirt in the machining device predicted by the prediction section 202 is equal to or greater than the threshold value β. In this case, the decision section 203a, for example, increases the number of cleanings of the predetermined cleaning in the input cleaning condition by one (a predetermined number) each time, causes the prediction section 202 to predict the degree of the dirt in the machining device under the cleaning condition after the number of cleanings of the predetermined cleaning is increased, and decides the minimum number of cleanings of the predetermined cleaning in the cleaning of the machining device after the machining under the machining condition to be performed next, in a range in which the predicted degree of the dirt in the machining device is less than the threshold value β.

[0116] On the other hand, the decision section 203a determines that there is no chip in the machining device when the degree of the dirt in the machining device predicted by the prediction section 202 is less than the threshold value β. In this case, the decision section 203a, for example, decreases the number of cleanings of the predetermined cleaning in the input cleaning condition by one (a predetermined number) each time, causes the prediction section 202 to predict the degree of the dirt in the machining device under the cleaning condition after the number of cleanings of the predetermined cleaning is decreased, and decides the minimum number of cleanings of the predetermined cleaning in the cleaning of the machining device after the machining under the machining condition to be performed next, in a range in which the predicted degree of the dirt in the machining device is less than the threshold value β.

[0117] <Prediction processing of cleaning prediction device 20a in use stage>

[0118] Next, the operation of the prediction processing of the cleaning prediction device 20a of the present embodiment will be described.

[0119] Figure 7 is a flowchart illustrating the prediction processing of the cleaning prediction device 20a in the use stage.

[0120] In step S21, before the machine tool 10 performs processing of a workpiece, the input section 201 inputs, from the control information of the control device 103, a processing condition to be performed next, a cleaning condition of a cleaning number of times of a predetermined cleaning, and state information of a degree of dirt in the machine tool at present.

[0121] In step S22, the prediction section 202 inputs, to the learned model 250, the processing condition to be performed next, the cleaning condition of the cleaning number of times of the predetermined cleaning, and the state information of the degree of dirt in the machine tool at present, and predicts the state information of the "degree of dirt" in the machine tool after processing is performed in the processing condition to be performed next and cleaning is performed in the cleaning condition.

[0122] In step S23, the decision section 203a decides, in a range where the predicted value of the degree of dirt in the machine tool predicted in step S22 is smaller than the threshold value β, the cleaning number of times of the predetermined cleaning that is the least in cleaning in the machine tool after processing is performed in the processing condition to be performed next. For example, the decision section 203a makes the cleaning number of times of the predetermined cleaning in the cleaning condition input in step S21 increase by one (a predetermined number) at a time, makes the prediction section 202 predict the degree of dirt in the machine tool under the cleaning condition after the cleaning number of times of the predetermined cleaning is increased, and decides the cleaning number of times of the predetermined cleaning that is the least in cleaning in the machine tool after processing is performed in the processing condition to be performed next in a range where the predicted degree of dirt in the machine tool is smaller than the threshold value β. On the other hand, the decision section 203a makes the cleaning number of times of the predetermined cleaning in the cleaning condition input in step S21 decrease by one (a predetermined number) at a time in a case where the degree of dirt in the machine tool predicted by the prediction section 202 is smaller than the threshold value β, makes the prediction section 202 predict the degree of dirt in the machine tool under the cleaning condition after the cleaning number of times of the predetermined cleaning is decreased, and decides the cleaning number of times of the predetermined cleaning that is the least in cleaning in the machine tool after processing is performed in the processing condition to be performed next in a range where the predicted degree of dirt in the machine tool is smaller than the threshold value β.

[0123] In step S24, the notification section 204 notifies the cleaning number of times of the predetermined cleaning that is the least decided in step S23.

[0124] According to the above, the cleaning prediction device 20a of the second embodiment inputs the next-to-be-performed forming condition involved in the machining of the workpiece by the machine tool 10, the cleaning condition of the predetermined number of cleanings of cleaning, and the state information including the degree of contamination in the machining tool at present to the learned model 250 before the machining of the workpiece by the machine tool 10, and predicts the state information of the degree of contamination in the machining tool after the machining and cleaning. Then, the cleaning prediction device 20a determines the minimum number of cleanings of cleaning in the machining tool after the machining by the next-to-be-performed machining condition based on the comparison of the predicted value of the degree of contamination in the machining tool predicted with the threshold value.

[0125] Thus, the cleaning prediction device 20a can eliminate the confirmation of the vision sensor during the machining, and prevent the accumulation of chips with a very small frequency of cleaning.

[0126] In addition, the cleaning prediction device 20a uses the learned model 250, and thus the operator does not need to judge the necessity of cleaning in the machining tool, and the burden on the operator can be reduced.

[0127] The second embodiment has been described above.

[0128] The first and second embodiments have been described above, but the cleaning prediction device 20, 20a and the machine learning device 30 are not limited to the above-described embodiments, and include variations, modifications, and the like within a range capable of achieving the object.

[0129] <Variant Example 1>

[0130] In the above-described first and second embodiments, the machine learning device 30 is exemplified as a device different from the machine tool 10, the control device 103, and the cleaning prediction device 20, 20a, but the machine tool 10, the control device 103, or the cleaning prediction device 20, 20a can be provided with part or all of the functions of the machine learning device 30.

[0131] <Variant Example 2>

[0132] In addition, for example, in the above-described first and second embodiments, the cleaning prediction device 20, 20a is exemplified as a device different from the machine tool 10 and the control device 103, but the machine tool 10 or the control device 103 can be provided with part or all of the functions of the cleaning prediction device 20, 20a.

[0133] Alternatively, a server may include part or all of the input unit 201, prediction unit 202, decision unit 203, notification unit 204, and storage unit 205 of the cleaning prediction device 20. Furthermore, for example, a server may include part or all of the input unit 201, prediction unit 202, decision unit 203a, notification unit 204, and storage unit 205 of the cleaning prediction device 20a. Furthermore, the functions of the cleaning prediction devices 20 and 20a may be implemented on the cloud using virtual server functionality or the like.

[0134] Furthermore, the cleaning prediction devices 20 and 20a may be a distributed processing system in which the functions of the cleaning prediction devices 20 and 20a are appropriately distributed across a plurality of servers.

[0135] <Variation 3>

[0136] In addition, for example, in the second embodiment described above, Figure 2B As shown, the control device 103 sets the predetermined cleaning content for each cleaning cycle to be the same in the cleaning program, causing the cleaning device 101 to perform the predetermined cleaning cycles. However, this is not limiting. For example, the chips in the processing machine of the machine tool 10 decrease as the number of cleaning cycles increases. Therefore, the control device 103 may also change the predetermined cleaning content for each cleaning cycle in the cleaning program.

[0137] Figure 8A This is a diagram showing an example of the first scheduled cleaning. Figure 8B This is a diagram showing an example of the second scheduled cleaning.

[0138] like Figure 8A As shown, the control device 103 performs a predetermined cleaning process by moving the cleaning nozzle of the cleaning device 101 along the cleaning path (1) to (4) according to the cleaning program to remove the chips in the processing machine. The control device 103 causes the measuring device 102 to measure the inside of the processing machine after the first predetermined cleaning. The control device 103 can also be based on the measurement results of the measuring device 102, for example, Figure 8B As shown, when chips remain only in the lower left portion of the processing machine, a cleaning path for removing only the chips in this portion is set in the cleaning program as the second scheduled cleaning.

[0139] Thereby, the cleaning prediction device 20a can further shorten the cleaning time.

[0140] <Variation 4>

[0141] Furthermore, for example, in the first and second embodiments described above, the cleaning prediction devices 20 and 20a use the learned model 250 provided by the machine learning device 30 to predict the degree of dirt inside the processing machine after processing under the next processing conditions obtained from one machine tool 10 and cleaning under the cleaning conditions, but the present invention is not limited to this. For example, Figure 9 As shown, the server 50 may store the learned model 250 generated by the machine learning device 30 and share the learned model 250 with m cleaning prediction devices 20B(1) to 20B(m) connected to the network 60 (m is an integer greater than or equal to 2). This allows the learned model 250 to be applied even when a new machine tool and cleaning prediction device are deployed.

[0142] Furthermore, each of the cleaning prediction devices 20B( 1 ) to 20B(m) is connected to each of the machine tools 10A( 1 ) to 10A(m).

[0143] In addition, the machine tools 10A(1) to 10A(m) correspond to Figure 1 The cleaning prediction devices 20B(1) to 20B(m) correspond to the machine tool 10. Figure 1 Cleaning prediction device 20 or Figure 6 Cleaning prediction device 20a.

[0144] Or, as Figure 10 As shown, the server 50 can also operate as, for example, the cleaning prediction device 20, 20a, and predict the degree of contamination in each machine tool 10A(1) to 10A(m) connected to the network 60 after machining under the next machining conditions and cleaning under the cleaning conditions. This allows the learned model 250 to be applied even when a new machine tool is deployed.

[0145] The functions of the cleaning prediction devices 20 and 20a and the machine learning device 30 in the first and second embodiments can be realized by hardware, software, or a combination thereof. "Realization by software" here means that a computer reads a program and executes it.

[0146] Each of the constituent parts included in the cleaning prediction device 20, 20a and the machine learning device 30 can be realized by hardware including an electronic circuit and the like, software, or a combination thereof. In the case of being realized by software, a program constituting the software is installed in a computer. In addition, these programs can be recorded in a removable medium and distributed to users, or can be distributed by being downloaded to a computer of a user via a network. In addition, in the case of being constituted by hardware, a part or all of the functions of each of the constituent parts included in the above-described devices can be constituted by, for example, an integrated circuit (IC) such as an application specific integrated circuit (ASIC), a gate array, a field programmable gate array (FPGA), a complex programmable logic device (CPLD), and the like.

[0147] The program can be stored using various types of non-transitory computer-readable media and provided to a computer. The non-transitory computer-readable medium includes various types of tangible storage media. Examples of the non-transitory computer-readable medium include a magnetic recording medium (e.g., a floppy disk, a magnetic tape, a hard disk drive), an opto-magnetic recording medium (e.g., a magneto-optical disk), a CD-ROM (Read Only Memory), a CD-R, a CD-R / W, a semiconductor memory (e.g., a mask ROM, a programmable ROM (PROM), an erasable PROM (EPROM), a flash ROM, a RAM). In addition, the program can also be provided to a computer through various types of transitory computer-readable media. Examples of the transitory computer-readable medium include an electric signal, an optical signal, and an electromagnetic wave. The transitory computer-readable medium can provide the program to a computer via a wired communication path such as an electric wire and an optical fiber, or a wireless communication path.

[0148] In addition, the step of describing the program recorded in the recording medium includes not only processing performed in time series according to the order thereof, but also processing performed in parallel or individually, without necessarily being performed in time series.

[0149] In other words, the machine learning device, the cleaning prediction device, and the cleaning system of the present disclosure can adopt various embodiments having the following structures.

[0150] (1) The machine learning device 30 of the present disclosure includes: an input data acquisition unit 301 that acquires input data including: an arbitrary machining condition of an arbitrary machine tool for an arbitrary workpiece, a cleaning condition for cleaning a machining chamber after machining the workpiece according to the machining condition by an arbitrary cleaning device, and state information indicating a degree of dirt in the machining chamber before machining based on the machining condition; a label acquisition unit 302 that acquires, as label data, state information of the machining chamber after machining with the machining condition included in the input data and cleaning with the cleaning condition; and a learning unit 303 that performs supervised learning using the input data acquired by the input data acquisition unit 301 and the label data acquired by the label acquisition unit 302, and generates a learned model 250.

[0151] According to the machine learning device 30, it is possible to eliminate the confirmation of the visual sensor during machining, and generate a learned model that prevents the accumulation of chips with a very low cleaning frequency.

[0152] (2) In the machine learning device 30 described in (1), the machining condition includes at least a type of machine tool, a material of the workpiece, a type of cutting tool, a spindle speed, a feed speed, a cutting amount, and a workpiece production quantity indicating a number of workpieces to be machined, and the cleaning condition includes at least a cleaning frequency of the predetermined cleaning.

[0153] Thus, the machine learning device 30 can generate a learned model 250 that can accurately predict the degree of dirt in the machining chamber of the machine tool 10.

[0154] (3) The cleaning prediction device 20 of the present disclosure includes: the learned model 250 generated by the machine learning device 30 described in (1) or (2); an input unit 201 that inputs, before machining by the machine tool 10, a machining condition to be performed next, a cleaning condition of the predetermined cleaning, and state information of the machining chamber of the machine tool 10 at present; and a prediction unit 202 that inputs the machining condition to be performed next, the cleaning condition, and the state information of the machining chamber at present, which are input by the input unit 201, to the learned model 250, and predicts state information of the machining chamber after machining with the machining condition to be performed next and cleaning with the cleaning condition.

[0155] According to the cleaning prediction device 20, it is possible to eliminate the confirmation of the visual sensor during machining, and prevent the accumulation of chips with a very low cleaning frequency.

[0156] (4) In the cleaning prediction device 20 described in (3), it is also possible to include: a decision unit 203 that decides a cleaning timing at which the workpiece production quantity is the largest, based on a comparison of a predicted value included in the state information of the machining chamber predicted by the prediction unit 202 and a threshold value set in advance.

[0157] Thus, the cleaning prediction device 20 can greatly reduce the time taken for cleaning.

[0158] (5) In the cleaning prediction device 20a described in (3), a decision section 203 can be provided, which decides the minimum number of cleanings scheduled in cleaning of the processing machine after processing under the processing conditions to be performed next, based on a comparison of the predicted value included in the state information in the processing machine predicted by the prediction section 202 and a threshold value set in advance.

[0159] Thus, the cleaning prediction device 20a can greatly reduce the time taken for cleaning.

[0160] (6) In the cleaning prediction device 20a described in (3), the scheduled cleaning changes the cleaning content according to the number of cleanings decided.

[0161] Thus, the cleaning prediction device 20a can further shorten the cleaning time.

[0162] (7) In any one of the cleaning prediction devices 20, 20a described in (3) to (6), a server 50 can be provided, which is connected in a manner that enables access to the learned model 250 from the cleaning prediction device 20, 20a via a network 60.

[0163] Thus, the cleaning prediction device 20, 20a can apply the learned model 250 even if a new machine tool 10 is configured.

[0164] (8) In any one of the cleaning prediction devices 20, 20a described in (3) to (7), the machine learning device 30 described in (1) or (2) can be provided.

[0165] Thus, the cleaning prediction device 20, 20a can have the same effect as any one of (1) to (7).

[0166] (9) The cleaning system 1, 1A of the present disclosure includes the cleaning prediction device 20, 20a described in any one of (3) to (8) and a cleaning device 101.

[0167] The cleaning system 1, 1A can have the same effect as any one of (1) to (8).

[0168] (10) In the cleaning system 1, 1A described in (9), the cleaning device 101 can be provided in the processing machine of the machine tool 10.

[0169] Thus, the cleaning system 1, 1A can have the same effect as any one of (1) to (8).

[0170] (11) In the cleaning system 1, 1A described in (9), the cleaning device 101 can also be mounted to a robot or built into a robot.

[0171] Thus, the cleaning system 1, 1A can have the same effects as any one of (1) to (8).

[0172] Explanation of Symbols

[0173] 1, 1A Cleaning system

[0174] 10 Machine tool

[0175] 101 Cleaning device

[0176] 102 Measuring device

[0177] 103 Control device

[0178] 20, 20a Cleaning prediction device

[0179] 201 Input section

[0180] 202 Prediction section

[0181] 203 Decision section

[0182] 204 Notification section

[0183] 205 Storage section

[0184] 250 Learned model

[0185] 30 Machine learning device

[0186] 301 Input data acquisition section

[0187] 302 Label acquisition section

[0188] 303 Learning section

[0189] 304 Storage section

Claims

1. A machine learning device, comprising: Possessing: an input data acquisition unit that acquires input data including: an arbitrary machining condition of an arbitrary machine tool for an arbitrary workpiece, a cleaning condition for cleaning a machining machine interior after machining the workpiece according to the machining condition by an arbitrary cleaning device, and state information indicating a degree of dirt in the machining machine interior before machining based on the machining condition; a label acquisition unit that acquires state information of the machining machine interior after machining with the machining condition included in the input data and cleaning with the cleaning condition as label data; and a learning unit that performs supervised learning using the input data acquired by the input data acquisition unit and the label data acquired by the label acquisition unit, and generates a learned model that predicts a degree of dirt in the machining machine interior after machining with a next-to-be-performed machining condition and cleaning with a predetermined cleaning cleaning condition based on the next-to-be-performed machining condition, the predetermined cleaning cleaning condition, and a current degree of dirt in the machining machine interior.

2. The machine learning device according to claim 1, characterized in that the machining condition includes at least a kind of machine tool, a material of workpiece, a kind of cutting tool, a spindle speed, a feed speed, a cutting amount, and a workpiece production quantity indicating a number of workpieces to be machined, the cleaning condition includes at least a cleaning number of the predetermined cleaning.

3. A cleaning prediction device characterized by comprising: Possessing: a learned model generated by the machine learning device according to claim 1 or 2; an input unit that inputs a next-to-be-performed machining condition, a predetermined cleaning cleaning condition, and state information of a machining machine interior of a machine tool before machining by the machine tool; and a prediction unit that inputs the next-to-be-performed machining condition, the cleaning condition, and the state information of the machining machine interior of the machine tool to the learned model, and predicts state information of the machining machine interior after machining with the next-to-be-performed machining condition and cleaning with the cleaning condition.

4. The cleaning prediction device according to claim 3, characterized in that the cleaning prediction device possesses: a decision unit that decides a cleaning timing of a workpiece production quantity most based on a comparison of a predicted value included in the state information of the machining machine interior predicted by the prediction unit and a threshold value set in advance.

5. The cleaning prediction device according to claim 3, characterized in that the cleaning prediction device possesses: a decision unit that decides a cleaning number of the predetermined cleaning least in cleaning of the machining machine interior after machining with the next-to-be-performed machining condition based on a comparison of a predicted value included in the state information of the machining machine interior predicted by the prediction unit and a threshold value set in advance.

6. The cleaning prediction device according to claim 5, characterized in that the predetermined cleaning changes a cleaning content according to a number of the decided cleaning number.

7. The cleaning prediction device according to claim 3, characterized in that The cleaning prediction device is provided with a server connected in a manner that the learned model is accessible from the cleaning prediction device via a network.

8. The cleaning prediction device according to claim 3, characterized in that, The cleaning prediction device is provided with the machine learning device according to claim 1 or 2.

9. A cleaning system characterized by, provided with: The cleaning prediction device according to any one of claims 3 to 8; and A cleaning device.

10. The cleaning system according to claim 9, characterized in that, The cleaning device is provided in a machining machine of the machine tool.

11. The cleaning system according to claim 9, characterized in that, The cleaning device is mounted to a robot, or built in the robot.

Images