JPH1086039A - Numerical controller having automatic machining condition preparation function - Google Patents

Abstract

PROBLEM TO BE SOLVED: To automatically determine optimum machining conditions in response to a variety of input condition data. SOLUTION: A theoretical formula, etc., is used to determine reference values for machining condition data from input condition data (102). The input condition data are inputted by a neural network to output a quantity of correction relative to the reference value for the machining condition data (102 and 104). Such quantity of correction is added to the reference value for the machining conditions to compute actual, optimum machining condition data (108). The coupling factor of the neural network is learned assuming a variety of input condition data as an input and a quantity of correction for machining condition data which is most suitable for such data as a teacher signal. When the output of the neural network is not appropriate, the coupling factor is learned assuming an appropriate output at that moment as a teacher signal. Since the experience as well as the instinct of an operator are memorized in the form of a coupling factor of the neural network, an appropriate quantity of correction for a machining condition data corresponding to a variety of input condition data is automatically determined.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a numerical controller having a function for automatically creating processing condition data.

[0002]

2. Description of the Related Art Conventionally, for example, in a grinding process using a numerically controlled grinding machine, a rough grinding, a fine grinding, a fine grinding, etc. The number of rotations of the workpiece for each grinding mode, the grinding start position,
2. Description of the Related Art There is known an apparatus for automatically determining processing conditions such as a grinding feed speed and a feed stop time after grinding by a computer.

[0003]

However, in practice,
It is empirically known that automatically determined processing condition data does not become optimal data for various input condition data. In such a case, the operator looks at the automatically determined processing condition data, and if it is not appropriate, corrects the value based on the operator's experience and intuition, or actually performs processing with the processing condition data. The operator judges the processed result and corrects the processing condition data based on experience and intuition.

[0004] How to correct which processing condition data should be corrected from the input condition data depends on the experience and intuition of the operator, and the correct correction has not always been performed. For this reason, it has been necessary to repeatedly perform the work of processing the workpiece with the corrected processing condition data and correcting the processing condition data again according to the processing result.

SUMMARY OF THE INVENTION The present invention has been made to solve the above-mentioned problem, and an object of the present invention is to automatically determine an optimum machining condition for a combination of various actual input condition data. Is to do so.

[0006]

As shown in FIG. 9, the configuration of the present invention for solving the above-mentioned problems includes an input condition data storage means for storing input condition data, and a reference value for the processing condition data. A reference value calculating means for calculating based on the data, a neural network which receives input condition data as input, and outputs a correction amount for each processing condition data with respect to a reference value, and a processing amount of processing condition data with a correction amount output from the neural network. Correction means for correcting the reference value to obtain the processing condition data.

[0007]

In the present invention, the reference value of the processing condition data is determined by the reference value calculation means from the input condition data by using a theoretical formula or the like. Further, the input condition data is input by the neural network, and the correction amount for the reference value of the processing condition data is output. Then, by the correction amount calculating means,
The correction amount is added to the reference value of the processing condition, and the actual optimum processing condition data is calculated. The coupling coefficients of the neural network are learned by inputting various input condition data and using a correction amount of the processing condition data optimal for the input condition data as a teacher signal. The coupling coefficient is
If the output of the neural network is not appropriate, the appropriate output at that time is learned as a teacher signal. As described above, since the experience and intuition of the worker are stored in the form of the coupling coefficient of the neural network, an appropriate correction amount of the processing condition data corresponding to various input condition data is automatically determined.

[0008]

According to the present invention, input condition data is input,
The correction amount for the reference value of the processing condition data is calculated by a neural network which outputs the correction amount for each processing condition data, and the processing value is obtained by correcting the reference value with the correction amount. Therefore,
It becomes possible to automatically determine the processing condition data that accurately reflects the correction of the processing condition data performed based on the experience and intuition of the operator.

[0009]

DESCRIPTION OF THE PREFERRED EMBODIMENTS Hereinafter, the present invention will be described based on specific embodiments. (1) Configuration of Grinding Machine FIG. 1 is a configuration diagram showing the overall mechanical configuration of a numerically controlled grinding machine having the processing condition automatic determination device according to the present invention.

Reference numeral 50 denotes a grinding machine. On a bed 51 of the grinding machine 50, a table 52 which folds with respect to the bed 51 is provided. The table 52 is moved in the left-right direction in the drawing by driving the table feed motor 53. A headstock 54 is placed on the table 52.
And a tailstock 56 are provided.
The tailstock 56 has a tailstock shaft 57. The workpiece W is supported by the spindle 55 and the tailstock 57, and the spindle 5
5 is rotated. The rotation of the spindle 55 is performed by a spindle motor 59 arranged on the headstock 54.

On the other hand, a grinding wheel 60 for grinding the workpiece W is supported by a drive shaft of a grinding wheel drive motor 62 provided on a grinding wheel stand 61. The movement of the wheel head 61 is controlled by a wheel head feed motor 63 in the vertical direction in the drawing. The numerical controller 30 is provided to drive and control the table feeding motor 53, the grinding wheel head feeding motor 63, the spindle motor 59, the grinding wheel drive motor 62 and the like.

(2) Configuration of Numerical Control Unit Numerical control unit 30 is mainly composed of C as shown in FIG.
It comprises a PU 31, a ROM 32, a RAM 33, and an IF (interface). NC in RAM33
An NC data area 331 for storing a program, an input condition data area (input condition data storage means) 332 for storing input condition data such as a finish diameter, an allowance, a rigidity coefficient, a part type, a rough finish division, and an operator name; There are provided a processing condition data storage area 333 for storing processing condition data such as a rotation speed, a grinding start position, a grinding feed speed, and a feed stop time after grinding, and a coupling coefficient area 334 for storing coupling coefficients of a neural network. The RAM 33 is backed up by a battery, and stores the learned coupling coefficient.

The ROM 32 has a control program area 321 storing a control program for operating the numerically controlled grinding machine in accordance with the NC data, and an automatic determination program area 322 storing a main automatic determination program for determining machining condition data. Neural network area 323 storing a neural network operation program
And a learning program area 324 storing a program for learning a coupling coefficient of the neural network.

An operation panel 20 is attached to the numerical controller 30 via an IF 34. A keyboard 2 for inputting data is provided on an operation panel 21 of the operation panel 20.
2 and a CRT display device 23 for displaying data.

(3) Operation Next, the processing procedure of the CPU 31 used in the present embodiment will be described with reference to a flowchart.

1. Input of Input Condition Data FIG. 3 is a flowchart of a main program for automatically generating machining condition data from input condition data.

In step 100, input condition data input from the keyboard 22 is read and stored in the input condition data area 332 of the RAM 33. In this embodiment,
The input condition data includes finish diameter data D ₁ , machining allowance data D ₂ , stiffness coefficient data D ₃ , part type data D ₄ , rough finish division data D ₅ , A worker name data D ₆ , B worker name data
D _7, a C worker name data D _8.

2. Calculation of reference value of processing condition data In the next step 102 (reference value calculating means), the reference value V of the processing condition data is obtained from the input condition data (D _{1 to} D ₈ ) and other input condition data. ₁ ~V ₁₁ is calculated.

In this embodiment, the processing condition data includes
Rough rotation speed data K ₁ , fine grinding speed data K ₂ , fine grinding speed data K ₃ , rough grinding start diameter data K ₄ , fine grinding start diameter data K ₅ ,
Fine grinding start diameter data K ₆ , coarse grinding feed speed data K ₇ , fine grinding feed speed data K ₈ , fine grinding feed speed data K ₉ , coarse grinding feed stop time data K ₁₀ , fine grinding feed stop time data K It is composed of ₁₁ and.

The calculation of this reference value is performed as follows. In the rotation speed data for each grinding mode, the peripheral speed of the workpiece is determined in advance by a function of the required roughness of the workpiece with respect to the peripheral speed of the grinding wheel at a certain grinding. The peripheral speed of the workpiece is calculated from the commanded surface roughness of the workpiece, and the rotational speed data of the workpiece is calculated based on the peripheral speed of the workpiece and the diameter of the workpiece.

In the grinding feed speed data in each grinding mode, the depth of cut per revolution of the workpiece of the grinding wheel is determined in advance as a function of the dimensional tolerance. The depth of cut is calculated from the commanded dimensional tolerance in each grinding mode, and the grinding feed speed data is calculated from the rotation speed data. In the grinding start diameter data of each grinding mode, a standard feed amount is set for each grinding mode, and is calculated based on the relationship between the commanded finishing diameter and the feed amount.

The post-grinding feed stop time data determines the number of revolutions of the workpiece at which the feed is stopped, depending on whether or not the fixed-size grinding is performed and whether or not the process is divided and ground. The feed stop rotation speed after grinding is determined from the input condition data,
The post-grinding feed stop time is calculated using the rotation speed data.

3. Non-feedback Neural Network Next, in step 104, the non-feedback neural network is started, and input condition data D _{1 to} D ₈ are inputted, and each processing condition data (K _{1 to} K ₁₁₎ ) Each primary correction amount δ ₁
～Deruta ₁₁ is calculated.

FIG. 4 shows a non-feedback type neural network.
The configuration shown in FIG. In this embodiment, the neural network has a three-layer structure including an input layer, an intermediate layer, and an output layer. Neural networks are well known,
Each element of the second intermediate layer and the third output layer is defined as an element that performs an operation of the following equation.

The output O ⁱ _j of the j-th element of the i-layer is calculated by the following equation. Here, i ≧ 2.

[Formula 1] O ⁱ _j = f (I ⁱ _j ) (1)

[Number 2] _{^{I i j = k ΣW i-}} 1 k, i j · O i-1 k + V i j ... (2)

[Equation 3] f (x) = 1 / {1 + exp (-x)} (3)

[0026] However, V ⁱ _j bias of the j th processing element of the i-th layer, W ^i-1 _k, ⁱ _j is the first i-1 layer k-th element and the j of the i-th layer The coupling coefficient between the first element and O ¹ _j represents the output value of the j-th element in the first layer. That is, since it is the first layer, the input is output as it is without performing any operation,
It is also the input value of the J-th element in the input layer (first layer).
Therefore,

[0027]

O ¹ _j = D _j (4) where D _j is input condition data input to the j-th element of the input layer.

The specific operation of the non-feedback neural network is executed according to the procedure shown in FIG. In step 200, each element of the intermediate layer (second layer) is connected to the input layer (first layer).
The output values D _{1 to} D ₈ from each element of the layer are input, and the product-sum operation of the following equation is performed. For the j-th element in the second layer, it is calculated by the following equation. In this embodiment, the bias is zero.

[Expression 5] I ² _j = _{k = 1} Σ ⁸ W ¹ _k, ² _j · D _k (5)

Next, in step 202, the output of each element of the intermediate layer (second layer) is calculated by the following equation using the sigmoid function of the product-sum function value of the input value of equation (5).
The output value of the j-th element in the second layer is calculated by the following equation.

O ² _j = f (I ² _j ) = 1 / ｛1 + exp (−I ² _j )… (6) The output value O ² _j is the value of each element of the output layer (third layer). Input value.

Next, in step 204, a product-sum operation of the input values of each element of the output layer (third layer) is executed.

[Formula 7] I ³ _j = _{k = 1} Σ ¹⁰ W ² _k, ³ _j · O ² _k (7)

Next, in step 206, the output value of each element in the output layer is calculated by the sigmoid function, as in the case of the equation (6). This output value is primary correction amount [delta] _j of the processing condition data. That is, the primary correction amount δ _j is obtained by the following equation.

Δ _j = f (I ³ _j ) = 1 / {1 + exp (−I ³ _j )} (8)

4. Feedback Neural Network Next, the process proceeds to step 106 in FIG. 3, where the operation by the feedback neural network is executed. As shown in FIG. 6, the structure of the feedback neural network is such that the output of each element in the output layer is input to all the elements in the intermediate layer. The operation of the feedback type neural network is executed according to the procedure shown in FIG.

In step 300, a product-sum operation of the input values of the hidden layer is executed. At this time, the inputs of each element in the intermediate layer are a total of 19 inputs, 8 outputs from the input layer and 11 outputs from the output layer.

That is, the following equation is calculated.

Equation 9] the ^{_{I 2 j = k = 1 Σ}} 8 W 1 k, 2 j · D 1 k + k = 1 Σ 11 W 3 k, 2 j · δ k ... (9).

Next, in step 302, the output of the intermediate layer is calculated by the equation (6) using the sigmoid function of the product sum operation result of the input values. In step 304, a product-sum operation of the input values is executed in each element of the output layer by the above equation (7). Then, in step 306, the sigmoid function product-sum function value of the input value by the following equation, the final correction amount delta _j of the processing condition data is calculated.

[0036]

Δ _j = f (I ³ _j ) = 1 / {1 + exp (−I ³ _j )} (10) In this way, the correction amounts Δ _{1 to} Δ _{11 of} all the processing condition data are obtained.
Is calculated, the process returns to step 108 of FIG.

5. In step 108 (correction means) of processing condition data, reference values V _{1 to} V _{11 of} the processing condition data obtained in step 102 and correction amounts Δ _{1 to} Δ ₁₁ obtained in step 106 are obtained. by the sum of the processing conditions data K ₁ ~K ₁₁ is obtained.

6. Learning of Coupling Coefficient Next, a method of learning the coupling coefficient of the neural network will be described. The coupling coefficient is executed by a well-known back propagation method for the feedback neural network shown in FIG.

In step 400 of FIG. 8, a learning signal of each element in the output layer is calculated by the following equation.

Y ³ _j = (T _j −Δ _j ) · f ^′ (I ³ _j ) (11) where T _j is a desired output (teacher signal) of the element of the j-th output layer with respect to the input signal. ) Is a signal given from outside and given by a skilled worker.

Next, in step 402, the learning signal of the intermediate layer is calculated by the following equation.

Equation 12] _{^{Y 2 j = f '(I}} 2 j) · k = 1 Σ 11 Y 3 k · W 2 j, 3 k ... (12) Next, in step 404, the coupling coefficient of the elements of the output layer Is corrected. The correction amount is obtained by the following equation.

Δω ² _i, ³ _j (t) = P · Y ³ _j · f (I ² _i ) + Q · Δω ² _i, ³ _j (t−1) (13)

Here, Δω ² _i, ³ _j (t) is the amount of change in the t-th calculation of the coupling coefficient between the j-th element in the output layer and the i-th element in the intermediate layer. Also, Δω ² _i, ³ _j (t-1)
Is the previous correction amount of the coupling coefficient. P and Q are proportional constants. Therefore, the coupling coefficient is

## EQU14 _{## The} corrected coupling coefficient is obtained from W ² _j, ³ _j + Δω ² _i, ³ _j (t) → W ² _j, ³ _j (14)

Next, the routine proceeds to step 406, where the coupling coefficient of each element of the intermediate layer is corrected. The correction amount of the coupling coefficient is obtained by the following equation, as in the case of the output layer.

Δω ¹ _i, ² _j (t) = P · Y ² _j · f (I ¹ _i ) + Q · Δω ¹ _i, ² _j (t−1) (15) Accordingly, the coupling coefficient is

The corrected coupling coefficient is obtained by the following equation: W ¹ _j, ² _j + Δω ¹ _i, ² _j (t) → W ¹ _j, ² _j (16)

Next, in step 408, it is determined whether or not the correction amount of the coupling coefficient has become a predetermined value or less, and it is determined whether or not the coupling coefficient has converged. If the coupling coefficient has not converged, the process returns to step 400, and the same operation is repeated using the newly corrected coupling coefficient, so that the coupling coefficient is corrected again. By repeating such operations,
Learning for a certain teacher signal is completed.

The learning of the coupling coefficient of the neural network is performed repeatedly by changing input condition data in various ways and using a teacher signal for an output corresponding to the input condition data.

As described above, in this embodiment, the primary correction amount is obtained by the non-feedback neural network, and the primary correction amount is fed back by the feedback neural network to calculate the final correction amount. I have to. Therefore, the influence of the correction amount of the processing condition data due to the mutual relationship between the processing condition data is also considered. As a result, it becomes possible to automatically determine the processing condition data that accurately reflects the correction of the processing condition data performed based on the experience and intuition of the operator.

In the above embodiment, the correction operation is performed by the feedback neural network. However, the reference value may be corrected using the output (primary correction amount) of the non-feedback neural network as the correction amount. . In this case, the learning of the neural network is performed on the non-feedback neural network, but is the same as the learning method shown in FIG.

[Brief description of the drawings]

FIG. 1 is a configuration diagram showing the overall mechanical configuration of a numerically controlled grinding machine having an automatic processing condition determination device according to a specific embodiment of the present invention.

FIG. 2 is a block diagram showing an electrical configuration of a numerical controller and an operation panel according to the apparatus of the embodiment.

FIG. 3 is an exemplary flowchart showing a main processing procedure of a CPU used in the apparatus of the embodiment.

FIG. 4 is a structural diagram showing a structure of a non-feedback neural network according to the embodiment.

FIG. 5 is a flowchart showing a calculation procedure of the non-feedback neural network.

FIG. 6 is a structural diagram showing the structure of a feedback neural network.

FIG. 7 is a flowchart showing a processing procedure of the feedback neural network.

FIG. 8 is a flowchart showing a learning procedure of the neural network.

FIG. 9 is a block diagram showing the configuration of the present invention.

[Explanation of symbols]

 DESCRIPTION OF SYMBOLS 20 ... Operation panel 21 ... Operation panel 22 ... Keyboard 23 ... CRT display device 30 ... Numerical control device 31 ... CPU 32 ... ROM 33 ... RAM 50 ... Grinding machine 51 ... Bed 52 ... Table 53 ... Table feed motor 54 ... Headstock 55 ... spindle 56 ... tailstock 57 ... tailstock 59 ... spindle motor 60 ... grinding wheel 61 ... grinding wheel drive 62 ... grinding wheel drive motor 63 ... motor for feeding the grinding wheel head W ... workpiece

Continuing from the front page (72) Inventor Moriaki Sakakura 1-1-1, Asahi-cho, Kariya-shi, Aichi, Japan Toyota Machine Works Co., Ltd. (72) Inventor Shiho 1-1-1, Asahi-cho, Kariya-shi, Aichi Prefecture Toyota Machine Works Co., Ltd.

Claims (1)

[Claims] When a workpiece is machined by a numerically controlled machine tool, machining condition data such as the number of revolutions of the workpiece, a machining start position, a tool feed speed, etc. in each machining mode are converted into a workpiece finish diameter, a machining allowance, and a rigidity. An apparatus for automatically determining from input condition data such as coefficients and tool types, an input condition data storage unit for storing the input condition data, and a reference value calculation for calculating a reference value of the processing condition data based on the input condition data. Means, a neural network that receives the input condition data as input, and outputs a correction amount for a reference value of each of the processing condition data, and corrects the reference value of the processing condition data with the correction amount output from the neural network. A numerical control device having a processing condition automatic creation function, comprising: a correction unit for obtaining processing condition data by performing the processing.