WO2022230008A1

WO2022230008A1 - Numerical control device, learning device, and chattering vibration suppressing method

Info

Publication number: WO2022230008A1
Application number: PCT/JP2021/016599
Authority: WO
Inventors: 遼輔池田; 一樹高幣; 智哉藤田
Original assignee: 三菱電機株式会社
Priority date: 2021-04-26
Filing date: 2021-04-26
Publication date: 2022-11-03
Also published as: JPWO2022230008A1; DE112021007584T5; JP7179198B1; CN117120946A

Abstract

This numerical control device (1) comprises a driving command unit (12) which issues an operation command (103) to a main shaft (17) and a feed shaft (18) of a machine tool (16), and further comprises a feature amount calculation unit (10) and a vibration control unit (11). The feature amount calculation unit (10) calculates feature information (102) about chattering vibration on the basis of a sensor signal (100) that detects the vibration of a tool or an object to be machined mounted on the machine tool (16) and a main shaft operation command (101) that is an operation command for the main shaft (17). The vibration control unit (11) outputs, to the driving command unit (12), a driving correction value (104) for correcting the operation command (103), until the feature amount information (102) reaches a target range.

Description

Numerical controller, learning device, and chatter vibration suppression method

The present disclosure relates to a numerical control device, a learning device, and a chatter vibration suppression method for suppressing chatter vibration in a machine tool.

A machine tool is a mechanical device that removes and processes a workpiece into a desired shape by changing the relative position between the workpiece and the tool. Machine tools typified by milling machines and lathes perform machining by mounting a tool or an object to be machined on a spindle and rotating the spindle. During machining, vibrations called "chatter" may occur. When chatter vibration occurs, the finished surface accuracy deteriorates and it causes tool breakage.

Patent Document 1 below discloses a method of searching for and obtaining the optimum spindle speed for suppressing chatter vibration. The optimum spindle speed is calculated based on the phase difference gradient. The phase difference gradient is the ratio of the amount of change in the phase difference to the amount of change in the spindle speed, that is, the gradient of the phase difference. This phase difference is calculated by a known formula including chatter vibration, the number of tool edges, and the spindle speed as parameters representing the chatter vibration feature quantity. In Patent Document 1, the phase difference is measured while changing the spindle speed, and the phase difference gradient is calculated based on the measured phase difference.

JP 2018-118366 A

However, the technique described in Patent Document 1 has the problem that if the characteristics of the vibration system that generates chatter vibration change during the search for the optimum spindle speed, the search must be interrupted and the process must be restarted from the beginning. be. This is because the characteristics of the vibration system that generates chatter vibration easily change due to factors such as a decrease in the mass of the workpiece due to machining or a change in the machining location of the workpiece. Therefore, in the method of Patent Document 1, if the characteristics of the vibration system that causes chatter vibration change before the search for the optimum spindle speed is completed, the optimum spindle speed cannot be calculated, and chatter vibration is suppressed. is difficult.

The present disclosure has been made in view of the above, and an object thereof is to obtain a numerical controller capable of suppressing chatter vibration even when the characteristics of the vibration system change during suppression of chatter vibration. .

In order to solve the above-described problems and achieve the object, the numerical control device according to the present disclosure has a drive command unit that gives an operation command to the main shaft and the feed shaft of the machine tool, and further includes a feature amount calculation unit and vibration control. have a department. The feature amount calculator calculates chatter vibration feature amount information based on a sensor signal that detects vibration of a tool attached to a machine tool or a workpiece and a spindle operation command that is a spindle operation command. The vibration control section outputs a correction value for correcting the operation command to the drive command section until the feature amount information reaches the target range.

According to the numerical control device according to the present disclosure, it is possible to suppress chatter vibration even when the characteristics of the vibration system change during suppression of chatter vibration.

1 is a diagram showing a functional configuration of a numerical controller according to Embodiment 1; FIG. A diagram showing the functional configuration of the feature amount calculation unit shown in FIG. A diagram showing the relationship between the timing signal output by the timing signal generator shown in FIG. 2 and the spindle angle command. A diagram showing a dimensionless quantity calculated by the phase difference calculator shown in FIG. Flowchart showing a processing flow by the vibration control unit and the drive command unit shown in FIG. A diagram showing a specific operation example by the vibration control unit and the drive command unit shown in FIG. FIG. 2 is a diagram showing another configuration example for realizing the function of the vibration control unit shown in FIG. 1; Diagram showing an example configuration of an inference model using a general neural network A diagram showing a configuration example of a learning device that learns an inference model used by the vibration control unit shown in FIG. FIG. 4 shows a functional configuration of a numerical controller according to Embodiment 2; Flowchart showing the processing flow by the vibration control unit shown in FIG. FIG. 10 is a diagram showing a functional configuration of a numerical control device according to Embodiment 3; The figure which shows the structural example of the control system which implement|achieves the function of the vibration control part which concerns on Embodiment 3. The figure which shows the function structure of the numerical control apparatus which concerns on Embodiment 4. Flowchart showing the processing flow by the vibration control unit shown in FIG. FIG. 10 is a diagram showing a functional configuration of a numerical control device according to Embodiment 5; A diagram showing the functional configuration of the feature amount calculation unit shown in FIG. A diagram showing the relationship between the timing signal output by the timing signal generator shown in FIG. 17 and the spindle angle command. Block diagram showing an example of a hardware configuration for realizing the functions of the numerical controller described in the first to fifth embodiments Block diagram showing another example of hardware configuration for realizing the functions of the numerical controller described in the first to fifth embodiments

A numerical control device, a learning device, and a chatter vibration suppression method according to an embodiment of the present disclosure will be described below in detail with reference to the accompanying drawings.

Embodiment 1.
FIG. 1 is a diagram showing the functional configuration of a numerical control device 1 according to Embodiment 1. As shown in FIG. The numerical controller 1 numerically controls the machine tool 16 by giving an operation command 103 to the machine tool 16 .

The machine tool 16 has a main shaft 17 and a feed shaft 18 each provided with a motor driven by a run command 103 . In Embodiment 1, a workpiece is installed on the spindle 17 and a tool is installed on the feed shaft 18 . The machine tool 16 also outputs to the numerical controller 1 operation information 105 including at least information on the positions of the spindle 17 and the feed shaft 18 and information on speed and motor current.

A sensor 19 is attached to the machine tool 16 . The sensor 19 outputs to the numerical controller 1 a sensor signal 100 that detects the vibration of the tool or workpiece. Sensor 19 is attached to the structure of spindle 17 or feed shaft 18 of machine tool 16 . The position where the sensor 19 is attached may be any position that can detect the vibration of the tool or the object to be processed, but it is more preferable if it is near the point where the tool and the object to be processed contact each other.

Any type of sensor 19 may be used as long as it can detect the vibration of a tool or an object to be processed. Examples of sensors 19 are displacement sensors, velocity sensors, acceleration sensors, angular velocity sensors. Further, the sensor 19 may be a force sensor that detects cutting reaction force, or a microphone that detects cutting noise during machining. Alternatively, instead of the sensor 19, at least one of the positions, speeds, and motor currents of the spindle 17 and feed shaft 18 included in the motion information 105 is used to detect vibration of the tool or workpiece during machining. Information indicating the detected vibration may be output as the sensor signal 100 .

The numerical control device 1 has a feature amount calculator 10 , a vibration controller 11 , and a drive commander 12 . The drive command unit 12 receives motion information 105 from the machine tool 16 and gives a drive command 103 to the spindle 17 and the feed shaft 18 of the machine tool 16 based on the received motion information 105 .

The drive command unit 12 further receives the drive correction value 104 from the vibration control unit 11. The drive command unit 12 outputs a drive command 103 generated based on the motion information 105 and the drive correction value 104 to the machine tool 16 and the vibration control unit 11 . Further, the drive command unit 12 outputs the spindle run command 101 included in the generated run command 103 to the feature quantity calculator 10 . The spindle operation command 101 is a signal including at least a spindle angle command, which is an angle command to the spindle 17 .

The spindle operation command 101 may include a spindle speed command, a spindle angle, or a spindle speed in addition to the spindle angle command. A spindle speed command is a speed command to the spindle 17 . The spindle angle is the actual angle of the spindle 17 relative to the spindle angle command. Spindle speed is the actual speed of the spindle 17 relative to the spindle speed command. The operation command 103 includes at least one of a spindle angle command and a spindle speed command, and at least one of a feed axis position command and a feed shaft speed command. A feed axis position command is a position command to the feed shaft 18 . A feed axis speed command is a speed command for the feed shaft 18 .

The feature quantity calculation unit 10 receives the sensor signal 100 from the machine tool 16 and receives the spindle operation command 101 from the drive command unit 12 . The feature quantity calculator 10 calculates feature quantity information 102 based on the sensor signal 100 and the spindle operation command 101 and outputs the calculated feature quantity information 102 to the vibration controller 11 . The feature quantity information 102 includes at least phase difference information, which is a feature quantity of chatter vibration.

FIG. 2 is a diagram showing the functional configuration of the feature quantity calculation unit 10 shown in FIG. The feature amount calculator 10 has a sensor signal processor 13 , a phase difference calculator 15 , and a timing signal generator 14 .

The sensor signal processing unit 13 generates a plurality of types of state quantities based on the sensor signal 100 and outputs a state quantity signal 110 indicating the generated plurality of types of state quantities to the phase difference calculation unit 15 . The state quantity signal 110 is a signal containing a first state quantity and a second state quantity. Specifically, the state quantity signal 110 is a signal including a state quantity represented by the sensor signal 100 and a state quantity obtained by subjecting the sensor signal 100 to time differentiation or time integration. For example, the first state quantity is the sensor signal 100, and the second state quantity is a state quantity obtained by time-differentiating the sensor signal 100 once. The first state quantity itself may also be a state quantity obtained by time differentiation.

The sensor signal processing unit 13 uses the sensor signal 100 and the signal obtained by differentiating the sensor signal 100 once with respect to time as a time-series signal at the same time as the state quantity signal 110 . Here, time differentiation means processing for calculating the amount of change in the state quantity per unit time, and time integration means processing for calculating the cumulative amount for each unit time.

The difference between the number of times of time differentiation of the first state quantity and the number of times of time differentiation of the second state quantity may be any combination as long as it is an odd number. For example, assume that the first state quantity is a state quantity obtained by time-differentiating the sensor signal 100 P times, and the second state quantity is a state quantity obtained by time-differentiating the sensor signal 100 Q times. Here, P and Q are integers. In this case, the difference between P and Q should be an odd number.

Furthermore, the dimension of the first state quantity or the second state quantity may be a dimension obtained by time-integrating the sensor signal 100 . As a specific example, the first state quantity may be acceleration and the second state quantity may be velocity. Alternatively, the first state quantity may be jerk, which is the differential value of acceleration, and the position, which is the integral value of velocity, may be the second state quantity. The sensor signal processing unit 13 includes, for example, a sensor signal 100 and at least one of a state quantity obtained by time differentiation of the sensor signal 100 and a state quantity obtained by time integration of the sensor signal 100. More than one type of state quantity can be generated.

When the timing signal generation unit 14 determines that the spindle angle command has passed a predetermined angle based on the spindle operation command 101 output by the drive command unit 12 , the timing signal generation unit 14 outputs a timing signal 111 to the phase difference calculation unit 15 . . Details of the method for generating the timing signal 111 will now be described.

FIG. 3 is a diagram showing the relationship between the timing signal 111 output by the timing signal generator 14 shown in FIG. 2 and the spindle angle command. The spindle angle command is included in the spindle operation command 101 output by the drive command section 12 . As shown in FIG. 3, the spindle angle command is a signal that takes a value from 0 [rad] to 2π [rad] and returns to 0 [rad] when reaching 360 [rad]. The timing signal generator 14 outputs the timing signal 111 each time the spindle angle command passes through the set angle φ1, and does not output the timing signal 111 otherwise. As a result, the timing signal 111 is periodically output at timing synchronized with the rotation of the spindle 17 of the machine tool 16 . The angle φ1 may be any angle as long as it is determined at one point while the main shaft 17 rotates once. The angle φ1 can be, for example, the angle of the main shaft 17 when the main shaft 17 is oriented. Here, the orientation of the spindle 17 is a reference angle at which the spindle 17 is stopped.

Alternatively, the timing signal generator 14 can also generate the timing signal 111 using the spindle speed command included in the spindle operation command 101 . In this case, the timing signal generator 14 outputs the timing signal 111 every time T1 calculated from the spindle speed command S [rpm] using the following equation (1) with the initial time t0 as the reference time. do. Also, the timing signal generator 14 may use both the spindle angle command and the spindle speed command.

　T1=60/S　...(1)

In the above description, the timing signal generator 14 uses the spindle operation command 101 to generate the timing signal 111, but it is not limited to this. The timing signal generator 14 may generate the timing signal 111 using a feed axis operation command, which is an operation command for the feed shaft 18 , instead of or in addition to the spindle operation command 101 .

The phase difference calculator 15 receives the timing signal 111 output by the timing signal generator 14 and the state quantity signal 110 output by the sensor signal processor 13 . Based on the timing signal 111 and the state quantity signal 110 , the phase difference calculator 15 normalizes each of the plurality of types of state quantities included in the state quantity signal 110 to make them dimensionless for each timing signal 111 . The phase difference calculator 15 calculates a phase difference indicating a difference between phases of the dimensionless state quantity signals in the state space composed of the dimensionless state quantity, and stores the calculated phase difference as the feature quantity information 102 . , and is output to the vibration control unit 11 . Details of the state space referred to here will be described later.

The phase difference calculator 15 normalizes each of the plurality of types of state quantities included in the state quantity signal 110 by dividing each state quantity by a predetermined maximum value of each state quantity. The maximum value used here may be the maximum value of each state quantity obtained in a previous machining experiment, or the maximum value of each state quantity obtained in a previous simulation.

Assume that the phase difference calculator 15 receives the timing signal 111 at time t1 and time t2 later than time t1. At this time, the phase difference calculator 15 normalizes the state quantity signal 110 at the time t1, and sets the normalized value as the first dimensionless quantity N1. At time t2, the phase difference calculator 15 normalizes the state quantity signal 110 in the same manner as at time t1, and sets the normalized value as the second dimensionless quantity N2.

FIG. 4 is a diagram showing dimensionless quantities calculated by the phase difference calculator 15 shown in FIG. In FIG. 4, the first dimensionless quantity N1 and the second dimensionless quantity N2 are represented in the state space represented by the first state quantity and the second state quantity included in the state quantity signal 110. It is shown. Both the first and second state quantities are vector quantities. In FIG. 4, the first horizontal axis represents the direction of the first state quantity, and the second vertical axis represents the direction of the second state quantity. 2 are orthogonal to each other. Therefore, as shown in FIG. 4, a vector space can be defined in which the first axis is the horizontal axis and the second axis is the vertical axis. In this paper, this vector space is called "state space".

The phase difference calculator 15 calculates the angle θ is calculated as the phase difference between the first dimensionless quantity N1 and the second dimensionless quantity N2. The phase difference calculator 15 performs the above calculation for each timing signal 111, and calculates the phase difference between the dimensionless quantity of the current process and the dimensionless quantity of the previous process. The phase difference calculator 15 outputs feature amount information 102 including at least the calculated phase difference to the vibration controller 11 . In the first embodiment, the phase difference takes a value between -π [rad] and +π [rad].

The vibration control unit 11 receives feature amount information 102 from the feature amount calculation unit 10 and receives an operation command 103 from the drive command unit 12 . The vibration control unit 11 generates a drive correction value 104 based on the feature quantity information 102 and the operation command 103 and outputs the generated drive correction value 104 to the drive command unit 12 . A drive correction value 104 is a correction value for correcting the operation command 103 .

FIG. 5 is a flow chart showing the processing flow by the vibration control section 11 and the drive command section 12 shown in FIG. FIG. 5 illustrates a processing flow in which the phase difference among the information included in the feature amount information 102 is used as a determination criterion, and the spindle speed command among various commands included in the operation command 103 is used as a control object. . The vibration control unit 11 corrects the operation command 103 in the following steps.

First, in step S101, it is determined whether or not the phase difference is outside the phase difference target range. The phase difference target range is a phase difference target range having values within a preset range. In the first embodiment, as an example, the phase difference target value is set to 0 radian, and the phase difference target range is set to a range from -0.1 [rad] to +0.1 [rad]. The phase difference target range may be set to a value sufficiently larger than the noise of the phase difference.

In step S101, if the phase difference is within the phase difference target range (step S101, No), the process proceeds to step S102 and the current spindle speed is held. That is, in step S102, the drive command unit 12 outputs the original spindle speed command without changing the current spindle speed command. On the other hand, if the phase difference is outside the phase difference target range (step S101, Yes), the process proceeds to step S103.

In step S103, it is determined whether or not the phase difference is greater than or equal to the phase difference threshold. The phase difference threshold is an arbitrary preset value. In Embodiment 1, as an example, 0 "rad" is set as the phase difference threshold. If it is less than the threshold (step S103, No), the process proceeds to step S105.

In step S103, "Yes" is determined when the phase difference and the phase difference threshold are equal, but "No" may be determined. That is, if the phase difference is equal to the phase difference threshold value, it may be determined as "Yes" or "No".

At step S104, the spindle speed command is decreased by a predetermined width. That is, in step S104, when it is determined in step S103 that the phase difference is equal to or greater than the phase difference threshold value, the value of the spindle speed command is decreased by a predetermined width. On the other hand, in step S105, the spindle speed command is increased by a predetermined width. That is, in step S105, when the phase difference is determined to be less than the phase difference threshold value in step S103, the value of the spindle speed command is increased by a predetermined width. The value to be increased or decreased may be a ratio to the current spindle speed command, or may be a fixed value that does not depend on the spindle speed command.

In step S106, it is determined whether or not the processing has ended. When the machining is finished (step S106, Yes), the vibration control unit 11 finishes the processing of FIG. On the other hand, if the processing has not ended (step S106, No), the process returns to step S101. The processing shown in FIG. 5 is continuously executed while processing is being performed.

In FIG. 5, the phase difference included in the feature amount information 102 is used as the determination criterion, and the controlled object is the spindle speed command included in the operation command 103. However, if the feature amount information 102 is other than the phase difference, , the same flow can be used when the controlled object is other than the spindle speed command. That is, the vibration control unit 11 performs control to output the driving correction value 104 for correcting the driving command 103 output by the driving command unit 12 until the feature amount information 102 reaches the target range. Further, the drive command unit 12 corrects the drive command 103 based on the drive correction value 104 .

FIG. 6 is a diagram showing a specific operation example by the vibration control section 11 and the drive command section 12 shown in FIG. FIG. 6 is a diagram showing the relationship between the phase difference and the spindle speed at each time. The upper part of FIG. 6 shows the phase difference, and the lower part shows the spindle speed. Moreover, the horizontal axis of FIG. 6 represents time.

Between time t60 and time t61, the phase difference takes a value within the phase difference target range, so the spindle speed is a constant value. Between time t61 and time t62, the phase difference is out of the phase difference target range and has a value less than 0 [rad], which is the phase difference threshold. Therefore, the spindle speed increases by a predetermined set value.

Next, between time t62 and time t63, the phase difference takes a value within the phase difference target range, so the spindle speed becomes a constant value. Between time t63 and time t64, the phase difference is out of the phase difference target range and exceeds the phase difference threshold value of 0 [rad]. Therefore, the spindle speed is decreased by a predetermined set value. After time t64, the phase difference takes a value within the phase difference target range, so the spindle speed takes a constant value.

Thus, the drive command unit 12 changes the drive command 103 according to the value of the phase difference. At this time, the vibration control section 11 outputs a drive correction value 104 based on the value of the phase difference to the drive command section 12 .

As described above, according to Embodiment 1, time-series data of a plurality of types of state quantities are generated based on the sensor signal 100 that detects the vibration of the tool attached to the machine tool 16 or the workpiece. be. In the state space, dimensionless quantities indicating a plurality of types of state quantities are generated for each timing signal 111, and a phase difference indicating an angle θ, which is a difference between the phases of the dimensionless quantities, is calculated. The operation command 103 is corrected based on the value of this phase difference.

In the method of Embodiment 1, the phase difference is calculated for each timing signal 111, and the operation command 103 is corrected based on the calculated phase difference. That is, in the method of Embodiment 1, the spindle speed can be corrected without measuring the phase difference in advance for spindle speeds under a plurality of conditions. Therefore, even if the characteristics of chatter vibration change during machining of the workpiece, chatter vibration can be suppressed.

As another example, the feature amount calculation unit 10 may calculate the phase difference based on the following formulas (2) and (3).

　k={60×f/(n×S)}...(2)

Phase difference = (k - [k]) x 2π ... (3)

　In the above formula (2), f represents the frequency of chatter vibration, and n represents the number of teeth of the tool. Here, the chatter vibration frequency f can be obtained by performing frequency analysis represented by Fast Fourier Transform (FFT) on the sensor signal 100 and calculating the frequency at which the gain peaks. can. In the above equation (2), the function [k] is a function that converts the value of "k" into an integer, and rounds the value of "k" below the decimal point toward zero.

As described above, according to the numerical control device 1 according to Embodiment 1 and the method of suppressing chatter vibration executed using the numerical control device 1, the vibration of the tool or the workpiece attached to the machine tool 16 can be suppressed. Chatter vibration feature quantity information 102 is calculated based on the detected sensor signal 100 and a spindle operation command 101 that is an operation command for the spindle 17 . Then, a drive correction value 104, which is a correction value for correcting the operation command 103, is output to the drive command unit 12 until the feature amount information 102 reaches the target range. As a result, chatter vibration can be suppressed even when the characteristics of the vibration system change during the suppression of chatter vibration. Further, in the method of the first embodiment, optimization of the drive command value, which has been conventionally performed, is not performed. As a result, the chatter vibration suppression time can be shortened, and the chatter vibration suppression can be speeded up.

As described above, the method of Embodiment 1 can calculate the feature amount information 102 without using frequency analysis represented by FFT. Therefore, it is possible to shorten the time from the occurrence of chatter vibration to the suppression of chatter vibration. As a result, it is possible to speed up the suppression of chatter vibration.

Although the machine tool 16 according to Embodiment 1 has a configuration in which the workpiece is installed on the spindle 17, the configuration is not limited to this. As an alternative configuration, a similar effect can be obtained even if a tool is installed on the spindle 17, as typified by a milling machine and a lathe.

Also, in Embodiment 1, one sensor 19 is used. In this case, the processing described in Embodiment 1 can be performed for all installed sensors 19 to determine the occurrence of chatter vibration. When the sensors 19 are installed at a plurality of locations, even if chatter vibration occurs at a plurality of locations during machining, the chatter vibration can be suppressed.

As another example, the vibration control unit 11 may suppress chatter vibration using an inference model that has undergone machine learning in advance for suppressing chatter vibration. An example of this is shown in FIG. That is, FIG. 7 is a diagram showing another configuration example for realizing the function of the vibration control section 11 shown in FIG.

In FIG. 7, the vibration control section 11 has an information observation section 201 and an inference section 202 . The feature amount information 102 and the operation command 103 are input to the vibration control unit 11 .

The information observation unit 201 observes the feature quantity information 102 and the driving command 103 for a predetermined number of samplings as time-series data, and generates an inference data set 203 based on the time-series data. The inference unit 202 inputs the inference data set 203 generated by the information observation unit 201 to the inference model that has undergone machine learning in advance so as to output the drive correction value 104 that suppresses chatter vibration. A correction value 104 is output.

The inference unit 202 may use any algorithmic inference model. As an example, an inference model using a neural network will be described. FIG. 8 is a diagram showing a configuration example of an inference model using a general neural network.

In the example shown in FIG. 8, the neural network has input layers x1, x2, . . . , xn with n neurons, hidden layers y1, y2, . and an output layer z1 having Although FIG. 8 shows an example in which there is one output layer, two or more output layers may be provided.

The input layers x1, x2, . . . , xn are connected to the intermediate layers y1, y2, . The connection between the input layer and the intermediate layer shown in FIG. 8 is an example, and each input layer x1, x2, . . . , xn may be connected to any intermediate layer y1, y2, .

In the case of a three-layer neural network as shown in FIG. 8, when a plurality of inputs are input to the input layers x1, x2, . , . . . , ym. The values input to the intermediate layers y1, y2, . . . , ym are further multiplied by weights B1 to Bb, input to the output layer z1, and output from the output layer z1. The subscripts a and b are natural numbers, and in the example shown in FIG. 8, a=n and b=m. It goes without saying that the output results vary depending on the values of the weights A1-Aa and B1-Bb.

FIG. 9 is a diagram showing a configuration example of a learning device 300 that learns an inference model used by the vibration control unit 11 shown in FIG. The learning device 300 has a learning data acquisition unit 301 and a learning processing unit 302 . The learning device 300 can be applied to a numerical control device 1 having a drive command section 12 that gives an operation command 103 to a spindle 17 and a feed shaft 18 of a machine tool 16 as shown in FIG.

The learning data acquisition unit 301 acquires a learning data set 304 in which the feature amount information 102 and the operation command 103 acquired during actual machining are associated with chatter vibration presence/absence information 303 indicating the presence or absence of chatter vibration. The chatter vibration presence/absence information 303 included in the learning data set 304 can be expressed using different numerical values depending on whether or not chatter vibration has occurred. For example, it can be represented by a numerical value such as "1" when chatter vibration occurs, and "0" when it does not occur. Further, the chatter vibration presence/absence information 303 can use, for example, the result determined by evaluating the machined surface after machining.

The learning processing unit 302 has a learning data set 304 and an inference model 305 . The learning processing unit 302 performs learning using the feature amount information 102 and the driving command 103 included in the learning data set 304 as input data. Specifically, the learning processing unit 302 performs so-called supervised learning so that the output of the inference model 305 matches the numerical value of the chatter vibration presence/absence information 303 . The inference model 305 outputs an output value corresponding to the presence or absence of chatter vibration when the feature quantity information 102 and the operation command 103 are input.

The learning processing unit 302 inputs the feature amount information 102 and the operation command 103 to the input layer, and weights A1 to Aa and B1 to Bb so that the values output from the output layer approach numerical values indicating the presence or absence of chatter vibration. to adjust. Learning of the inference model 305 is performed by adjusting the weights A1 to Aa and B1 to Bb.

The learning processing unit 302 can use the error backpropagation method as a supervised learning method used when the inference model 305 is learned. In addition, in order to improve the generalization performance of the inference model 305, the learning processing unit 302 employs methods such as "dropout" that randomly excludes neurons during learning, and "early stopping" that monitors errors and terminates learning early. may be used. The learning processing unit 302 outputs the learned inference model 305 as a learned inference model. The learned inference model output by the learning processing unit 302 can be used in the inference unit 202 shown in FIG.

As described above, the learning device 300 according to Embodiment 1 has the inference model 305 and the learning dataset 304 . In the learning data set 304, chatter vibration presence/absence information 303 indicating the presence or absence of chatter vibration, feature amount information 102, and operation command 103 are associated and held. The feature amount information 102 , the operation command 103 and the chatter vibration presence/absence information 303 are input to the inference model 305 . The inference model 305 performs machine learning so that when the feature amount information 102 and the operation command 103 are input, an output value corresponding to the presence or absence of chatter vibration is output. If such a learning device 300 is applied to the numerical control device 1, it is possible to output the drive correction value 104 using the inference model 305 that has undergone machine learning in advance, so that chatter vibration can be suppressed at high speed.

In addition, although the case where a neural network is used for the inference model 305 has been described above, it is not limited to this. The inference model 305 may be configured using other known methods such as RNN (Recurrent Neural Network), LSTM (Long Short-Term Memory), SVM (Support Vector Machine), and the like.

Embodiment 2.
FIG. 10 is a diagram showing the functional configuration of a numerical control device 2 according to Embodiment 2. As shown in FIG. The numerical controller 2 has a feature quantity calculator 10 , a vibration controller 11 - 1 , a drive commander 12 and an input device 20 . In the following, parts different from the numerical control device 1 will be mainly described.

The numerical controller 2 has an input device 20 in addition to the configuration described for the numerical controller 1 according to the first embodiment. The operator inputs information on the upper limit value and the lower limit value of the operation command 103 to be corrected in the vibration control section 11-1 to the input device 20 before processing.

The input device 20 outputs restriction information 106 including at least the upper limit value and the lower limit value of the input operation command 103 to the vibration control section 11-1. Instead of this method, the vibration control unit 11-1 may refer to a machining program in which the upper limit value and the lower limit value of the operation command 103 are described and acquire them.

The vibration control unit 11 - 1 receives the feature amount information 102 from the feature amount calculation unit 10 , the operation command 103 from the drive command unit 12 , and the restriction information 106 from the input device 20 . Vibration control section 11 - 1 generates drive correction value 104 based on feature amount information 102 , operation command 103 and limit information 106 , and outputs generated drive correction value 104 to drive command section 12 .

FIG. 11 is a flow chart showing the processing flow by the vibration control section 11-1 shown in FIG. The vibration control section 11-1 corrects the operation command in the following steps. In FIG. 11, the same reference numerals are given to the same or equivalent processes as those shown in FIG. 11, similarly to FIG. 5, illustrates a processing flow in which the phase difference is used as a determination criterion and the spindle speed command is controlled. Hereinafter, the parts different from FIG. 5 will be mainly described.

The processing of steps S101 to S105 is the same as or equivalent to that of FIG. 5, and the description is omitted here.

In step S201, the spindle speed command processed in step S104 is limited within a predetermined lower limit of the spindle speed command. Specifically, when the spindle speed command processed in step S104 is less than the lower limit of the spindle speed command (step S201, No), the process proceeds to step S202 to change the spindle speed command to the lower limit of the spindle speed command. After that, the process proceeds to step S106. If the spindle speed command is equal to or higher than the lower limit value of the spindle speed command (step S201, Yes), the process advances to step S106 without changing the spindle speed command.

Also, in step S203, the spindle speed command processed in step S105 is limited within the range of the predetermined upper limit of the spindle speed command. Specifically, when the spindle speed command processed in step S105 exceeds the upper limit of the spindle speed command (step S203, No), the process advances to step S204 to set the spindle speed command to the upper limit of the spindle speed command. After changing, the process proceeds to step S106. If the spindle speed command is equal to or lower than the upper limit of the spindle speed command (step S203, Yes), the process proceeds to step S106 without changing the spindle speed command.

In step S106, it is determined whether or not the processing has ended. If the machining is completed (step S106, Yes), the vibration control section 11-1 terminates the processing of FIG. On the other hand, if the processing has not ended (step S106, No), the process returns to step S101. The processing shown in FIG. 11 is continuously executed while processing is being performed.

FIG. 11 illustrates the case where the phase difference included in the feature amount information 102 is used as the determination criterion, and the controlled object is the spindle speed command included in the operation command 103. However, if the feature amount information 102 is other than the phase difference, , the same flow can be used when the controlled object is other than the spindle speed command.

According to the second embodiment, the vibration control unit 11-1 receives the limit information 106 including the upper limit value and the lower limit value of the operation command 103, and the vibration control unit 11-1 controls the input upper limit value and lower limit value. Increase or decrease the run command 103 within the range between This prevents the vibration control unit 11-1 from outputting to the drive command unit 12 the drive correction value 104 outside the cutting conditions determined by the tool and machine tool. As a result, it is possible to prevent the corrected operation command 103 from becoming a value unexpected by the operator. As a result, it is possible to obtain the further effect of preventing the occurrence of unintended processing defects.

Embodiment 3.
FIG. 12 is a diagram showing the functional configuration of a numerical control device 3 according to Embodiment 3. As shown in FIG. 12, the numerical controller 1 is replaced with the numerical controller 3, and the vibration controller 11 is replaced with the vibration controller 11-2. In addition to the operation command 103, operation information 105 is input to the vibration control unit 11-2. Other configurations are the same as or equivalent to those in FIG. 1, and the same or equivalent components are denoted by the same reference numerals. In the following, portions different from the numerical control device 1 will be mainly described.

The vibration control unit 11-2 receives the feature amount information 102 from the feature amount calculation unit 10, the operation command 103 from the drive command unit 12, and the motion information 105 from the machine tool 16. Vibration control section 11 - 2 generates drive correction value 104 based on feature quantity information 102 , operation command 103 and motion information 105 , and outputs generated drive correction value 104 to drive command section 12 .

In the third embodiment, the vibration control unit 11-2 performs control so that the feature amount information 102 reaches the feature amount target value set in advance. Specifically, a control system is configured as shown in FIG. FIG. 13 is a diagram showing a configuration example of a control system that implements the functions of the vibration control section 11-2 according to the third embodiment. FIG. 13 shows, as an example, the configuration of a control system that performs PID (Proportional Integral Differential) control.

First, a feature amount deviation 125 is generated by subtracting the feature amount information 102 from the feature amount target value. Here, "0" is set as an example of the feature quantity target value. In Embodiment 3, the feature quantity information 102 is the phase difference calculated by the phase difference calculator 15 in FIG. It is generally known that the phase difference is 0 when chatter vibration does not occur. Therefore, chatter vibration can be suppressed by setting the feature amount target value to "0".

The feature quantity deviation 125 is branched into three, one of which is multiplied by the proportional gain 122 . One of the remaining two is multiplied by integral gain 123 after passing through integrator 120 . That is, the time integral value of the feature quantity deviation 125 is multiplied by the integral gain 123 . Also, the other of the remaining two is multiplied by a differential gain 124 after passing through a differentiator 121 . That is, the time differential value of the feature amount deviation 125 is multiplied by the differential gain 124 . These three types of gain values are added to obtain the feature amount correction amount 126 . The feature amount correction amount 126 is further added to the spindle speed and output to the drive command unit 12 as the drive correction value 104 .

The proportional gain 122, the integral gain 123, and the differential gain 124 are values set in advance and given before processing. Also, these three types of gain values can be adjusted by performing trial machining. By appropriately setting these three types of gain values, chatter vibration can be suppressed at a higher speed than in the first embodiment.

As described above, according to the third embodiment, the vibration control unit 11-2 includes a control system for matching the feature amount information 102 with the feature amount target value. It can be performed faster.

In the control system of FIG. 13, the feature amount deviation 125 is multiplied by the proportional gain 122, the time integral value of the feature amount deviation 125 is multiplied by the integral gain 123, and the time differential value of the feature amount deviation 125 is multiplied by the differential gain 124. Although multiplying, it is not limited to this configuration. At least one of the feature amount deviation 125, the time integral value of the feature amount deviation 125, and the time differential value of the feature amount deviation 125 may be multiplied by the gain, and the desired characteristics of the control system or the desired characteristics may be obtained. It is possible to construct a similar control system.

Embodiment 4.
FIG. 14 is a diagram showing the functional configuration of a numerical controller 4 according to Embodiment 4. As shown in FIG. 14, the numerical controller 1 is replaced with the numerical controller 4, and the vibration controller 11 is replaced with the vibration controller 11-3. Also, the drive correction value 104 input to the drive command unit 12 is replaced with a drive correction value 104-1. Other configurations are the same as or equivalent to those in FIG. 1, and the same or equivalent components are denoted by the same reference numerals. In the following, portions different from the numerical control device 1 will be mainly described.

The vibration control unit 11-3 receives the feature amount information 102 from the feature amount calculation unit 10 and receives the operation command 103 from the drive command unit 12. The vibration control unit 11-3 generates a drive correction value 104-1 based on the feature amount information 102 and the operation command 103, and outputs the generated drive correction value 104-1 to the drive command unit 12.

FIG. 15 is a flow chart showing the processing flow by the vibration control section 11-3 shown in FIG. The vibration control section 11-3 corrects the operation command in the following steps. In addition, in FIG. 15, the same reference numerals are given to the same or equivalent processes as those shown in FIG. In addition, FIG. 15 illustrates a processing flow in which the phase difference is used as a determination criterion, and the main shaft speed command and the feed axis speed command, which is the speed command of the feed shaft 18, are controlled. Hereinafter, the parts different from FIG. 5 will be mainly described.

In step S301, after changing the spindle speed command in step S104 or step S105, the feed axis speed command is changed. Specifically, in step S301, the feed shaft speed command is changed so that the feed amount per rotation of the main shaft 17 does not change. Here, the feed amount per rotation of the main shaft 17 is a value obtained by dividing the feed shaft speed, which is the speed of the feed shaft 18, by the main shaft speed. For example, if the spindle speed command is decreased by 10% in step S104, the feed shaft speed command is also decreased by 10% at the same rate. In step S105, when the spindle speed command is increased by 10%, the feed shaft speed command is also increased by 10% at the same rate.

In step S106, it is determined whether or not the processing has ended. If the machining is finished (step S106, Yes), the vibration control section 11-3 finishes the processing of FIG. On the other hand, if the processing has not ended (step S106, No), the process returns to step S101. The processing shown in FIG. 15 is continuously executed while processing is being performed.

Note that FIG. 15 illustrates the case where the phase difference included in the feature amount information 102 is used as the determination criterion, but processing in a similar flow is possible even when the feature amount information 102 is other than the phase difference.

According to the fourth embodiment, the vibration control section 11-3 outputs a correction value for correcting the feed axis speed command to the drive command section 12 in synchronization with the increase or decrease of the spindle speed command. In this control, the vibration control unit 11-3 changes the spindle speed command and the feed axis speed command according to the phase difference value so that the feed amount per rotation of the spindle 17 does not change. do. This control makes the feed amount per rotation of the main shaft 17 constant, so that the intervals between cutter marks generated on the machined surface are made uniform. As a result, in addition to the effect of the first embodiment, the effect of improving the machined surface quality can be obtained.

Embodiment 5.
FIG. 16 is a diagram showing the functional configuration of a numerical control device 5 according to Embodiment 5. As shown in FIG. 16, the numerical controller 1 is replaced with the numerical controller 5, and the feature amount calculator 10 is replaced with the feature amount calculator 10-1. Further, in FIG. 16, a tool information recording unit 21 for generating tool information 113 is added. In addition to the sensor signal 100 and the spindle operation command 101, the tool information 113 is input to the feature amount calculation unit 10-1. Other configurations are the same as or equivalent to those in FIG. 1, and the same or equivalent components are denoted by the same reference numerals. In the following, parts different from the numerical control device 1 will be mainly described.

The tool information recording unit 21 records tool information 113, which is information about the tools installed in the machine tool 16, and outputs the recorded tool information 113 to the feature amount calculation unit 10-1. The tool information 113 includes at least information on the number of blades of the tool. The tool information 113 may further include information on the types of tools such as end mills and cutting tools, and information on tool shapes such as tool lengths and tool diameters. When a turning tool is installed, the tool information recording unit 21 outputs the tool information 113 assuming that the tool has one blade. Moreover, even when a turning tool is installed, the number of cutting edges is the same as the number of tools when a plurality of tools are used for cutting at the same time. For example, when the machine tool 16 has a lower tool post and an upper tool post, and machining is performed using a turning tool installed on each tool post, the tool information recording unit 21 assumes that the number of blades is two, and the tool information 113 is output.

The feature quantity calculation unit 10-1 receives the sensor signal 100 from the machine tool 16, receives the spindle operation command 101 from the drive command unit 12, and receives the tool information 113 from the tool information recording unit 21. Based on the sensor signal 100, the spindle operation command 101, and the tool information 113, the feature amount calculation unit 10-1 generates feature amount information 102 including at least a phase difference, which is a feature amount of chatter vibration, and calculates the generated feature amount information. 102 is output to the vibration control unit 11 .

FIG. 17 is a diagram showing the functional configuration of the feature amount calculation unit 10-1 shown in FIG. The feature quantity calculator 10-1 has a sensor signal processor 13, a phase difference calculator 15, and a timing signal generator 14-1.

Based on the spindle operation command 101 output by the drive command unit 12 and the tool information 113 output by the tool information recording unit 21, the timing signal generation unit 14-1 detects when the spindle angle, which is the angle of the spindle 17, has passed the set value. Then, the timing signal 111 is output to the phase difference calculator 15 at the determined timing. A method of generating the timing signal 111 by the timing signal generator 14-1 will be described in detail below. In the following description, it is assumed that a tool having α blades is installed on the spindle 17 . However, α is a natural number of 2 or more.

FIG. 18 is a diagram showing the relationship between the timing signal 111 output by the timing signal generator 14-1 shown in FIG. 17 and the spindle angle command. As shown in FIG. 18, the spindle angle command is a signal that takes a value from 0 [rad] to 2π [rad] and returns to 0 [rad] when reaching 360 [rad]. The timing signal 111 is a signal that is output each time the spindle angle command passes through the set angle _φβ . Here, β is a natural number equal to or greater than 2 and equal to or smaller than α. That is, the timing signal 111 is output α times during one rotation of the spindle 17 . FIG. 18 shows an example of α=2, in which the timing signal 111 is output every time the spindle angle command passes through the set angles φ1 and φ2, and otherwise the timing signal 111 is not output. It is sufficient that the angle φ1 is set at only one point during one rotation of the main shaft 17, and any angle may be set. However, there is a constraint between φ _β and φ _β−1 as shown in the following equation (4).

φ _β −φ _β−1 = 2π/α (4)

As another example, the timing signal generator 14-1 may generate the timing signal 111 using the spindle speed command included in the spindle operation command 101. In this case, the timing signal generator 14-1 outputs the timing signal 111 every time T2 calculated from the spindle speed command S [rpm] using the following equation (5).

　T2=60/(S×α)　...(5)

As described above, the numerical control device 5 according to Embodiment 5 has the tool information recording unit 21 for recording information about the number of blades of the tool, and the feature amount calculation unit 10-1 considers the number of blades of the tool. Then, the feature amount information 102 is calculated. Based on the tool information 113 acquired from the tool information recording unit 21, the timing signal generation unit 14-1 generates a timing signal 111 and outputs it to the phase difference calculation unit 15 each time the main shaft 17 rotates. As a result, even if the tool has two or more blades, it is possible to suppress chatter vibration.

Next, the hardware configuration for realizing the functions of the numerical controllers 1 to 5 described in Embodiments 1 to 5 will be described with reference to FIGS. 19 and 20. FIG. FIG. 19 is a block diagram showing an example of a hardware configuration that implements the functions of the numerical controllers 1-5 described in the first to fifth embodiments. FIG. 20 is a block diagram showing another example of the hardware configuration that implements the functions of the numerical controllers 1-5 described in the first to fifth embodiments.

When realizing the functions of the numerical controllers 1 to 5, as shown in FIG. 19, there are a processor 701 that performs calculations, a memory 702 that stores programs read by the processor 701, and a signal input/output unit. The configuration may include an interface 704 .

The processor 701 may be a computing means called microprocessor, microcomputer, CPU (Central Processing Unit), or DSP (Digital Signal Processor). The memory 702 includes non-volatile or volatile semiconductor memories such as RAM (Random Access Memory), ROM (Read Only Memory), flash memory, EPROM (Erasable Programmable ROM), EEPROM (registered trademark) (Electrically EPROM), Magnetic discs, flexible discs, optical discs, compact discs, mini discs, and DVDs (Digital Versatile Discs) can be exemplified.

The memory 702 stores programs for executing the functions of the numerical controllers 1-5. The processor 701 transmits and receives necessary information via the interface 704, the processor 701 executes the program stored in the memory 702, and the processor 701 refers to the table stored in the memory 702, thereby performing the above-described processing. It can be carried out. The computation results by processor 701 can be stored in memory 702 .

Further, when realizing the functions of the numerical controllers 1 to 5, the processing circuit 703 shown in FIG. 20 can also be used. The processing circuit 703 corresponds to a single circuit, a composite circuit, an ASIC (Application Specific Integrated Circuit), an FPGA (Field-Programmable Gate Array), or a combination thereof. Information to be input to the processing circuit 703 and information to be output from the processing circuit 703 can be obtained via the interface 704 .

It should be noted that part of the processing in the numerical controllers 1 to 5 may be performed by the processing circuit 703, and the processing not performed by the processing circuit 703 may be performed by the processor 701 and the memory 702.

The configurations shown in the above embodiments are only examples, and can be combined with other known techniques, or can be combined with other embodiments, without departing from the scope of the invention. It is also possible to omit or change part of the configuration.

1, 2, 3, 4, 5 Numerical control device, 10, 10-1 feature quantity calculation unit, 11, 11-1, 11-2, 11-3 vibration control unit, 12 drive command unit, 13 sensor signal processing unit , 14, 14-1 timing signal generator, 15 phase difference calculator, 16 machine tool, 17 spindle, 18 feed axis, 19 sensor, 20 input device, 21 tool information recorder, 100 sensor signal, 101 spindle operation command, 102 feature amount information, 103 operation command, 104, 104-1 drive correction value, 105 operation information, 106 limit information, 110 state amount signal, 111 timing signal, 113 tool information, 120 integrator, 121 differentiator, 122 proportional gain , 123 integral gain, 124 differential gain, 125 feature amount deviation, 126 feature amount correction amount, 201 information observation unit, 202 inference unit, 203 inference data set, 300 learning device, 301 learning data acquisition unit, 302 learning processing unit, 303 Chatter vibration presence/absence information, 304 Learning data set, 305 Inference model, 701 Processor, 702 Memory, 703 Processing circuit, 704 Interface.

Claims

A numerical control device having a drive command unit that gives drive commands to a spindle and a feed shaft of a machine tool,
a feature amount calculation unit that calculates feature amount information of chatter vibration based on a sensor signal that detects vibration of a tool attached to the machine tool or an object to be machined and a spindle operation command that is an operation command for the spindle;
a vibration control unit that outputs a correction value for correcting the operation command to the drive command unit until the feature amount information reaches a target range;
A numerical control device comprising:
The feature amount calculation unit is
a sensor signal processing unit that generates a plurality of types of state quantities based on the sensor signal;
a timing signal generator that periodically generates and outputs a timing signal at a timing synchronized with the rotation of the main shaft;
a phase difference indicating a difference between phases of the dimensionless state quantity signals in a state space composed of the dimensionless state quantities by normalizing the plurality of types of state quantities for each of the timing signals; a phase difference calculator that calculates
The numerical controller according to claim 1, characterized by comprising:
further comprising a tool information recording unit that records information of a tool attached to the machine tool;
3. The numerical controller according to claim 2, wherein the timing signal generation unit outputs the timing signal each time the spindle rotates based on the tool information acquired from the tool information recording unit. .
The sensor signal processing unit calculates the sensor signal, a state quantity obtained by calculating a change amount of the sensor signal per unit time, and a cumulative amount of the sensor signal per unit time. 4. The numerical controller according to claim 2, wherein two or more types of state quantities are generated among the state quantities.
The timing signal generator generates the timing signal based on at least one of a spindle angle command, a spindle speed command, a spindle angle that is an actual angle of the spindle, and a spindle speed that is an actual speed of the spindle. 5. The numerical controller according to any one of claims 2 to 4, which generates a signal.
The operation command includes a spindle speed command, which is a speed command for the spindle, and a feed axis speed command, which is a speed command for the feed axis,
6. The vibration control section outputs a correction value for correcting the feed axis speed command to the drive command section in synchronization with the increase or decrease of the spindle speed command. Numerical controller according to the item.
The vibration control unit changes the spindle speed command and the feed axis speed command in accordance with the phase difference value so that the feed amount per rotation of the spindle does not change. 7. A numerical controller according to claim 6.
Restriction information including an upper limit value and a lower limit value of the operation command is input to the vibration control unit,
The numerical value according to any one of claims 1 to 7, wherein the vibration control unit increases or decreases the operation command within a range between the input upper limit value and the lower limit value. Control device.
The vibration control unit corrects driving based on a value obtained by multiplying at least one of a deviation between a target value of the feature amount information and the feature amount information, a time integral value of the deviation, and a time differential value of the deviation by a gain. The numerical controller according to any one of claims 1 to 8, which generates a value.
The vibration control unit is
an information observation unit that observes the spindle operation command and the feature amount information as time-series data and generates an inference data set based on the observed time-series data;
an inference unit that outputs the correction value by inputting the inference data set generated by the information observation unit to an inference model that has undergone machine learning in advance;
The numerical controller according to any one of claims 1 to 9, characterized by comprising:
A learning device configured to be applicable to a numerical control device having a drive command section that gives drive commands to a spindle and a feed shaft of a machine tool,
a feature amount calculation unit that calculates feature amount information of chatter vibration based on a sensor signal that detects vibration of a tool attached to the machine tool or an object to be machined and a spindle operation command that is an operation command for the spindle;
a learning data acquisition unit that acquires a learning data set in which chatter vibration presence/absence information indicating the presence or absence of chatter vibration is associated with the feature amount information and the operation command;
The data set for learning acquired by the learning data acquiring unit and an inference model are provided, and when the feature amount information and the operation command are input to the inference model, the presence or absence of the chatter vibration is determined from the inference model. A learning processing unit that performs machine learning of the inference model so that an output value corresponding to is output;
A learning device comprising:
A method for suppressing chatter vibration executed by using a numerical control device having a drive command unit that gives drive commands to a spindle and a feed shaft of a machine tool, comprising:
calculating feature quantity information of the chatter vibration based on a sensor signal that detects vibration of a tool or an object to be machined attached to the machine tool and a spindle operation command that is an operation command for the spindle;
and outputting a correction value for correcting the operation command to the drive command unit until the feature amount information reaches a target range.