JP2000003213A - Numerical controller and fuzzy inference device applicable to the controller - Google Patents

Abstract

PROBLEM TO BE SOLVED: To easily change the know-how of a machining condition, etc., without changing a system and then to easily perform the complicated control that considers various factors by inputting an element that decides a feeding speed necessary for a rule of a function form which changes the feeding speed to calculate the necessary output value and calculating the final feeding speed based on a prescribed mathematical expression. SOLUTION: A feeding speed control part 24 includes a knowledge storage part 25 and an inference part 26. The part 25 stores a rule which changes a feeding speed in a function form, and the part 26 inputs an element to decide a machining condition necessary for the said rule of a function form. Then the part 26 calculates the necessary output value based on the inputted element and the rule of a function form and then calculates the final feeding speed based on a mathematical expression that is previously defined. Then the function to be stored in the part 25 is automatically generated based on the form of a work to be machined or to be measured, and a relevant object is controlled based on the feeding speed that is calculated at the part 26.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a numerical controller and a fuzzy inference apparatus applicable to the numerical controller.

[0002]

2. Description of the Related Art A numerical controller executes a numerical control process based on a processing program instructed from a paper tape or the like, and drives a machine tool based on a result of the process to perform a process on a work as instructed.

FIG. 34 is a block diagram of such a conventional numerical controller. The processing program read from the tape reader 11 is stored in the memory 12. When executing this machining program, the machining program is read out from the memory 12 block by block, and firstly, the processing device (CP
U) and a control device 17 incorporating a control program memory and the like
Is processed. Next, the control device 17 executes a numerical control process according to the machining program, drives the servo motor of the machine tool 1 to move the table or the tool post according to the movement command, or executes the machining via the high-power control device 13. Controls such as turning on / off the coolant of the machine 1 and normal rotation / reverse rotation / stop of the spindle. Reference numeral 16 denotes an operation panel having switches, buttons and the like for instructing return to origin, jog, and the like, and 14 denotes a manual data input operation (hereinafter referred to as MDI) for inputting various data to the control device 17 by manual operation.
, 15 is a display unit for displaying the current position of the machine and the like, and each of the elements 11 to 17 constitutes a computer numerical controller (hereinafter referred to as CNC). The control device 17 in the CNC has a computer configuration having a processing device (CPU), a control program memory and the like as described above, and the processing device (CPU) has a predetermined numerical value based on the control program and the machining program. The control process is performed to control the machine tool 1.

[0004] In general, machining by a machine tool is a removal process for removing unnecessary portions as chips by relative movement between a tool and a work material (hereinafter referred to as a work). In this removal processing, the processing efficiency is determined by the amount of chips removed per unit time. In order to increase the processing efficiency, it is necessary to increase the amount of chip removal per unit time as much as possible.However, in practice, a certain amount is required due to the limit of the load applied to the machine or tool, the required accuracy of the processing surface, etc. Be restricted. The processing conditions determine the chip discharge amount per unit time. In turning, the number of revolutions of the work per unit time, the relative feed speed of the tool to the work, and the cutting depth of the tool to the work correspond to this. In milling, the tool per unit time The rotation speed, the relative feed speed of the tool to the work, and the cutting depth of the tool to the work correspond to this. That is, in any of the turning and the milling, suitably controlling the relative feed speed of the tool to the workpiece is a very important processing element in the removal. If this is unnecessarily reduced, the processing efficiency is impaired, and the processing time is increased. If it is increased beyond the allowable value, the processing accuracy is adversely affected, and tools, machines and the like are overloaded.

FIG. 25 is a block diagram of a main part of a conventional feed speed control unit. The machining program read out one block at a time from the memory 12 in FIG. 34 and analyzed by the control device 17 is a CNC command data 20 in FIG. 25, that is, a pulse distribution processing unit 21 as a movement command and a feed speed command for each axis.
Sent to The pulse distribution processing unit 21 calculates the movement pulse per unit time of each axis from the movement command of each axis and the feed speed command, and sends it to the servo control unit 22 of each axis. The servo control unit 22 drives the servo motor 23 of the machine tool 1 by the movement pulse.

[0006]

FIG. 26 is an explanatory diagram of a program path in a corner portion.
(A) shows the programmed path and the actual tool path. Ideally, it is naturally desirable that the programmed path and the actual tool path coincide with each other. However, in practice, the two always differ at the corner P due to a delay in following the servo system. Therefore, when the tool 31 bends outside of the work 30 with respect to the work 30 at the corner P as shown in FIG. Measures such as lowering the feed speed of No. 31 and temporarily stopping at the corners are taken. Also, as shown in FIG.
Is bent inside the work 30, the work 30 is not entangled, but a large amount of uncut material is left or the tool 31
However, there is a problem that the load is suddenly increased. For this reason, as shown in the corner override function explanatory diagrams of FIGS. 27A and 27B, the commanded feed speed is reduced at the commanded ratio within the distances Le and Ls commanded before and after the corner P. Some CNCs could perform (override) processing. However, the speed only changes in two steps within the distances Le and Ls from the corner P and the other portions. In addition, the conventional machining program command is a command for changing the feed speed within one block (start and end points). Command such as lowering the feed speed than the intermediate point feed speed) cannot be performed, so that the speed within the distances Le and Ls from the corner P must be reduced to the speed at which the speed must be reduced most. The first problem is that the feed rate must be set, and the feed rate changes too rapidly.

[0007] FIG. 28 is an explanatory view of a drilling process in which a drill tool 31 performs a drilling process on a work 30. FIG. 28A shows a case where the tool 31 starts cutting the work 30. When the tool 31 comes into contact with the work 30, the feed speed is reduced, and the tool 31 is completely inserted into the work 30. If the feed rate is increased in this state, more suitable processing can be performed. This is because, when the tool 31 is brought into contact with the work 30 at a normal feed speed for processing the work 30, a load is suddenly applied to the tool 31 and the tool 31 is broken or a position shift occurs. is there. Therefore, in general, the tool 31 is positioned to a point a slightly before the work 30, the feed speed is reduced to a point b at which the tool 31 is completely inserted into the work 30, and the normal work 30 is moved from the point b. Process at the feed rate to be processed.

FIG. 28 (b) shows that the tool 31 is
This is an example of a case where a hole penetrating through 0 is formed. In this case, the tool 31 is fed at a feed rate at which a normal workpiece 30 is machined.
When the tool 31 penetrates through the work 30, burrs may be formed at the bottom of the hole of the work 30. , From the point c, the feed speed is reduced to perform the processing.

29 (a) shows the case where the surface of the work 30 in contact with the tool 31 is inclined, and FIG. 29 (b) shows the case where the tool 3 is drilled.
1 is a case where the surface of the bottom of the hole that penetrates is inclined. In this case, if the feed rate is not reduced particularly when the workpiece 30 comes into contact with the workpiece 30 or penetrates the workpiece 30, the accuracy of drilling is reduced, and the risk of breakage of the tool is increased. In order to control the feed rate in this way, the conventional machining program command is:
Since it was not possible to issue a command to change the feed speed within one block (eg, a command to lower the feed speed between the start point and end point than the feed speed at the intermediate point), it was necessary to divide the block into a plurality of blocks. Among them, there is a second problem that the feed speed must be set in consideration of the worst case in the block.

FIG. 30 (a) shows a case where a work 30 such as a formed material is machined by a tool 31, and the tool works the work as shown in FIG. There are parts and parts that are not processed. Therefore, in order to increase the processing efficiency, the distance between a and e should be originally shown in FIG.
As shown in FIG. 3A, the conventional machining program command is to change the feed speed within one block (for example, to make the feed speed between the start point and the end point slower than the feed speed at the intermediate point). 30), the machining block was changed to ab and b as shown in FIG.
It is processed by dividing it into four parts, bc, cd, and de. In this case, when the tool 31 comes into contact with the work 30 at the point (b, d) at which processing is started on the work, the feed speed is reduced to reduce the impact on the tool 31 and reduce the work speed.
It is desirable to reduce the feed speed of the tool 31 so as not to cause burrs on the work 30 even at a point (point c) where the tool 31 comes out of the tool. However, since the block is further divided, FIG. Processing as is, b, c,
There is a third problem in that if there is a problem in consideration of the feed speed at point d, the entire feed speed must be reduced for machining.

[0011] FIG.
Is a case where the middle part is machined by the tool 31. The part ab is a part that cuts into the work and the part that gradually applies a load to the tool. Is a part where a constant load is always applied, and c-
The part d is a part where the load of the tool gradually decreases.
Conventional machining program commands cannot change the feed speed in one block (commands such as lowering the feed speed between the start point and the end point than the feed speed at the intermediate point). The feed rate is determined in consideration of the feed rate of the part, but at this feed rate, if the load of the tool is too abruptly applied at the part a-b and adversely affects the tool, the part a-b is unavoidable. There is a fourth problem that it is necessary to specify the feed speed by lowering the feed speed in consideration of the feed speed.

The measurement explanatory diagram of FIG. 32 shows a case where the work 30 is measured by the measuring tool 31. The position of the work is measured by bringing the sensor tool 31 into contact with the work 30. In this case, the conventional machining program command cannot change the feed speed within one block (command to make the feed speed between the start point and the end point slower than the feed speed at the intermediate point). The program is divided into blocks so that the tool is sent at a relatively high speed to point a before point a, and the tool is sent at a low measurement speed from point a to point b. Thus, near the measurement point (ab)
In order to divide the block and to reduce the feed speed during this period (ab), it takes a long time to perform measurement.
There was a problem.

FIG. 33 is an explanatory diagram of control in setting an entry-prohibited area. An area 32 where entry of a tool 31 is prohibited is provided, and it is always monitored whether the tool 31 enters this area 32. Has a
The function of stopping the tool at a point is explained. In this case, the conventional machining program command cannot issue a command to change the feed speed within one block (a command to make the feed speed between the start point and the end point slower than the feed speed at the intermediate point). There is a sixth problem that the tool is fed at the specified speed until the vehicle 31 enters the no-go area 32, so that the no-go area must be specified slightly larger for safety. .

In general, the feed speed of a tool greatly depends on the relative relationship between the material of the work and the material of the tool.
There is a seventh problem in that when a tool is changed to another tool during machining and the materials of the two tools are different, it is necessary to modify the CNC machining program to change the feed rate.

For each of the first to seventh problems,
When trying to control by applying control in the form of a function,
There are cases where it is desired to change the contents of the knowledge storage unit set in advance. However, there is an eighth problem that any change in the content to be changed must be made one by one, which is troublesome.

Further, for each of the above first to seventh problems,
FIG. 8 shows a case where control is attempted by applying fuzzy control.
Conventionally, when defining a rule in the example of FIG.
Rules were defined as indicated by R3, and the membership functions A1, A2, A3, B1, B2, and B3 had to be defined separately. For this reason, a dedicated man-machine interface that defines the membership function is prepared,
The ninth problem is that a large storage area for storing the membership function inside the system is required.

Further, for each of the first to seventh problems,
When attempting to perform control by applying fuzzy control, it is necessary to perform fuzzy inference following cutting of a machine tool.
For this reason, the fuzzy inference must be performed at a very high speed, and if the fuzzy inference is executed by software, it takes too much time. For this reason, it is necessary to mount a dedicated fuzzy chip or the like on the CNC and perform hardware processing, and there is a tenth problem that the cost is increased.

In general, some rules are very important and some rules are not so important. That is, there is a difference in importance between rules. However, the conventional fuzzy inference has an eleventh problem that all set rules are treated equivalently.

The present invention has been made in order to solve the above first to eighth problems. In order to control the feed speed, the present invention does not divide a plurality of blocks into one block. It is possible to control the increase and decrease of the feed rate based on the set rules, and furthermore, it is possible to freely change the set rules, so that the processing know-how possessed by the operator can be incorporated, and the knowledge storage set in advance It is an object of the present invention to obtain a numerical control device capable of easily changing the contents of a unit.

Further, the present invention has been made to solve the tenth problem as described above. When there are many rules set, the inference is performed only on the rules that need to be determined, whereby the inference is performed. It is an object of the present invention to obtain a fuzzy inference device capable of reducing time.

The present invention has been made to solve the ninth problem described above, and does not require a large storage area for storing membership functions, and performs fuzzy inference by software processing. It is an object of the present invention to obtain a fuzzy inference device that can be performed at high speed.

Further, the present invention provides the eleventh aspect as described above.
The purpose of the present invention is to provide a fuzzy inference apparatus that can reflect the importance of a rule in a conclusion.

[0023]

A numerical controller according to the present invention is a numerical controller capable of changing a feed rate, wherein one or more rules for changing the feed rate are described in a functional form. Input the elements to determine the processing conditions required for the function format rules stored in this knowledge storage unit and the function format rules stored in this knowledge storage unit. An inference unit that calculates the final output speed by calculating the calculated output value based on a previously defined mathematical expression, and a function stored in the knowledge storage unit, the shape of the workpiece to be processed or Means for automatically generating the workpiece based on the shape of the workpiece to be measured, and controlling the controlled object based on the feed speed obtained by the inference unit.

Further, the fuzzy inference apparatus according to the present invention comprises:
A knowledge storage unit that stores a plurality of rules for changing control conditions in the form of a production rule, and a fuzzy inference unit that infers a desired control condition from the rules stored in the knowledge storage unit. From the plurality of rules stored in the knowledge storage unit, only the rule including the membership function corresponding to the input value is extracted,
A means for inferring using only the extracted rules is provided.

Further, the fuzzy inference device according to the present invention
A knowledge storage unit in which a rule for changing the control condition is described in the form of a production rule; and a fuzzy inference unit for inferring a desirable control condition from the rule stored in the knowledge storage unit. The membership functions used for all are isosceles triangles,
This is the definition of the membership function directly in the above rule.

Further, in the fuzzy inference device according to the present invention, in the fuzzy inference device described above, the knowledge storage unit stores a rule that directly defines the importance βi of a production rule for changing a control condition. Then, when performing the fuzzy inference, the inference result is obtained using the importance βi defined in the above rules, the area Si and the center of gravity position Li of the result of each rule.

[0027]

DESCRIPTION OF THE PREFERRED EMBODIMENTS Embodiment 1 Next, a first embodiment of the present invention will be described with reference to the drawings. In FIG.
Is the same or corresponding part as the conventional device, and 24 is a feed speed control unit. The feed speed control unit includes a knowledge storage unit 25 and an inference unit 26.

Next, the operation will be described. A plurality of rules for changing the feed speed of the corner as shown in FIG. 2 are described in the knowledge storage unit 25, and method 1 is to reduce the feed speed of the tool as the tool approaches the corner. . Conventionally, as shown in FIGS. 27 and (a) and (b), the feed speed is reduced to a certain value at which the tool approaches from the corner P to a certain distance Le, and the tool moves away from the corner P to a certain distance Ls. The feed rate is simply switched by the threshold value of the distance from the corner part to return to the original feed rate, but in the embodiment of the present invention,
As shown in FIG. 2, the function 1 that defines the deceleration rate can be freely changed according to the distance of the tool approaching the corner. Method 2 shows a rule for correcting the reduction rate of the feed speed of the tool according to the inclination angle of the corner. Generally, as the inclination of the corner is steeper (closer to 0 degrees), the speed is greatly reduced.
Since deceleration is small, an average deceleration rate is set in function 1 of method 1 and correction is performed according to the inclination of the corner by function 2 of method 2 in the corner feed set in the example of FIG. It is an example of a rule about control.

FIG. 3 shows the knowledge storage unit 2 in the inference unit 26.
6 is a flowchart showing a procedure for actually controlling the feed speed of a corner using the rule described in FIG. First, the inference unit 26 reads the method 1 from the knowledge storage unit 25 (steps 200 to 201), obtains the distance between the tool and the corner, which is data necessary for the method 1, and gives this as input data (step 202). . The feed rate deceleration rate Z1 is extracted according to this distance (step 20).
3). In this case, the function 1 is used to extract the feed speed reduction rate of the tool corresponding to the distance from the corner.
Similarly, the deceleration rate Z2 of the feed speed is extracted from the inclination angle of the corner portion by the method 2 (steps 204 to 205,
201-203). Next, in the case of this embodiment, N = 2
Therefore, step 205 becomes YES, and the two deceleration rates Z1 and Z2 obtained by the two rules are combined (step 206), and the feed speed Fo of the tool is determined (step 207). The above combination takes the product of the respective values as shown in Equation 1 in this case.

Z = Z1 * Z2 *... * Zn (1)

The feed rate is determined by the specified feed rate F
Is multiplied by the feed speed deceleration rate obtained by Expression 1 to calculate a corrected feed speed Fo. Fo = F * (100−Z) / 100 Z is the synthesized deceleration rate, and the unit is%. Combining multiple results in this way and processing conditions (tool feed rate)
, It is possible to realize complicated control based on a plurality of rules. Also, the knowledge storage unit 2
By making the 5 and the inference unit 26 independent, a more complicated rule can be defined. In the above-described embodiment, the rules described in the knowledge storage unit 25 are described in a free format as shown in FIG. 2, but the rules may be described using a format of a function and an operation expression of the function. For example, in the case of the example of FIG. 2, functions 1 and 2 are defined, and Expression 2 for calculating the results of these functions is defined as follows.

F = F1 * F2 (2)

Equation 2 is a function 1 (F1) and a function 2 (F2)
Is calculated, and the result is calculated as the final result.

The inference unit 26 calculates the operation result of each function as shown in FIG. 3, and synthesizes the result by a defined equation. Equation 2 in step 206 is the defined equation in this case.

For example, when the following arithmetic expression (Equation 3) is defined, an average of the results calculated by the functions 1, 2, 3 is taken, and the value is multiplied by the result calculated by the function 4. Means to get the final result. Thus, it is also possible to define rules by freely defined functions and expressions that define their operation methods. F = (F1 + F2 + F3) / 3 * F4 (3)

Embodiment 2 Next, a second embodiment of the present invention will be described with reference to the drawings. In FIG. 1, the knowledge storage unit 2
After describing the conditions to determine the rule for changing the feed speed by sending the rules stored in 5 and the contents to be executed when the conditions of the antecedents are not satisfied or satisfied, It is described in the subject section. That is, a rule is described in the form of a so-called production rule in the form of if antecedent part then consequent part, and the value described in the rule is expressed in the form of a membership function. Thus, the knowledge storage unit 25 has a structure in which macro general-purpose knowledge is described by rules, and micro-specific knowledge is expressed by a membership function. The inference unit 26 performs fuzzy inference on the given membership function based on the rules described in the knowledge storage unit 25 to derive a conclusion.

In general, fuzzy inference in fuzzy control often takes the center of gravity of the inference result by the minimax composition rule of fuzzy relation, and is called a minimax composition centroid method. As shown in FIG. 36, (1) given the assumptions x0 and y0, calculate the fitness a of each rule; (2) obtain the inference result Ci * for each rule (3) each rule C0 is obtained by integrating the inference results by
As the weighted center of gravity, the inference result Z0 of each rule as a whole
The inference is made in three stages. In addition, the fuzzy set C
As a method of interpreting 0, instead of truncating Ci by ai to obtain Ci *, instead of obtaining Ci *, as a method of defuzzifying C0, a median (median) is calculated instead of the center of gravity, Various methods such as a height method for selecting an element of a set of tables giving a value have been proposed. However, according to past experience, among these many methods, the minimax composite centroid has been proposed. The method is known to give very good results. FIGS. 4 and 5 are explanatory diagrams of rules provided in the knowledge storage unit 25 according to the present invention.
A rule for controlling the feed rate of the hole drill shown in FIGS. 28A and 29A is described.

In FIG. 4, R1 to R5 indicate rules, and in this embodiment, five rules are combined to draw a conclusion. In the case of FIG. 28A, the POS is a tool 3
1 is the distance between the work 30 and the tool 31
Is + before contact with, and-after contact. 0 indicates a contact point. A1 is a membership function shown in FIG. 5, and this function reduces the feed speed of the tool from shortly before the tool comes into contact with the work, and until the tool completely enters the work for a while after the contact with the work. It is an instruction to keep the feed speed lowered, and to return the feed speed to the original speed after the feed is completely stopped.

ANG in FIG. 4 shows the inclination of the surface where the tool comes into contact with the workpiece.
B5. FIG. 5 shows the membership functions of B1 to B5.

FEED in FIG. 4 is a tool feed speed reduction rate and indicates how much the tool feed speed is reduced. The FEED is classified into five stages C1 to C5. In FIG.
4 shows membership functions of C1 to C5.

According to the rule shown in FIG. 4, the feed rate of the tool is reduced according to the inclination of the work surface with which the tool comes into contact immediately before the tool comes into contact with the work from immediately before the tool comes into contact with the work. means.

One example of the inference method is performed as follows. Taking rule 1 as an example, the distance between the tool and the workpiece is obtained, and the degree of conformity is evaluated using the membership function shown in FIG. 5-A1. Also, the inclination of the surface where the tool contacts the workpiece is obtained, and the degree of conformity is evaluated using the membership function of FIG. 5-B1. Since the rule in the antecedent part is an AND condition, the rule with the smaller degree of conformity is adopted, and FIG.
Find the result using the membership function of C1. Similarly, results are obtained for rules 2 to 5, and the results are combined to obtain a conclusion. As an example of the method of taking the sum, a max-min logical product was used, and the center of gravity method was used as a combining method. Using the conclusions derived in this way,
The feed rate F of the tool is corrected as in Expression 4, and this is set as the feed rate of the tool. F0 = F * (100−Z) / 100 (4) F0 is the corrected feed rate of the tool F is the commanded feed rate Z is the deceleration rate of the tool derived from fuzzy inference.

Similarly, in the case of FIGS. 28 (b) and 29 (b), rules such as those shown in FIGS. 4 and 5 can be defined. In this case, the POS is the distance with respect to the surface when the tool 31 penetrates the work 30 in FIG. 28A, and the inclination of this surface is ANG.

In the case of FIG. 30, the control is performed so that the feed speed of the tool is corrected in accordance with the moving direction of the tool, and the tool moves in the moving direction 0 in parallel with the workpiece, that is, b- a case where cutting is performed between c and a direction in which the moving direction + is cut into the work, that is, a case where cutting is performed between a and b, and a moving direction-is a direction in which the work escapes from the work;
That is, it is assumed that cutting is performed between cd. Using the moving direction of the tool as input data, a conclusion is extracted using the membership function shown in FIG. 7 based on the rule shown in FIG. From the tool feed speed deceleration rate thus obtained, the corrected feed speed of the tool is calculated using Expression 4.

In this case, when the moving direction of the tool is-, that is, when the tool moves in the direction of escaping from the workpiece, the deceleration rate becomes-, and the corrected feed speed of the tool is instructed. It will increase from the feed rate.

In the example shown in FIG. 33, the distance between the tool 31 and the no-go area 32 is extracted, this distance is used as input data, and the deceleration is performed by the membership function shown in FIG. 9 based on the rule shown in FIG. Extract the rate. Using the extracted deceleration rate, the feed speed of the tool corrected by Expression 4 is extracted.

Here, the distance L between the tool 31 and the no-go area 32 is determined according to each position 1 to 8 with respect to the no-go area 32 as shown in FIG. The distance L as shown in the flowchart of FIG.
Is extracted.

In the flowchart of FIG. 11, the position of the tool is X, Y, the upper limit of the entry prohibited area in the X-axis direction is XL, the lower limit is XS, the upper limit in the Y-axis direction is YL, and the lower limit is YS. In step 300, X is used to determine in which of areas 1 to 8 the tool is shown in FIG.
Find L, XS, YL, YS. Next, step 301
, Classification is performed according to the values of XL, XS, YL, and YS, and a distance L between the tool 31 and the entry-prohibited area 32 is extracted according to the classification result (step 302).

Embodiment 3 Next, a third embodiment of the present invention will be described with reference to the drawings. As shown in FIG. 30 (a), when processing is performed on a molding material, it is desired to increase or decrease the feed rate in accordance with the shape of the work, so that the function itself is automatically generated from the shape of the work. C with built-in automatic program
In some NCs, the shape of a material is input in advance, and a function is generated based on the shape of the material. FIG. 12 shows this automatically generated function. Contact the work according to the shape of the work (a
Point) Deceleration before this point, a little after point a, and then return to the original speed, decelerate again before coming out of the work (point b),
In a portion where the work does not exist (between bc), it accelerates to the allowable limit, and decelerates before coming into contact with the work again (point c).
After a short time after passing the point c, the speed is restored.

The method of generating the function shown in FIG. 12 will be described with reference to the flowchart shown in FIG. The entry part to the work is determined (step 400), and the acceleration / deceleration pattern of the entry part of the work is set (steps 401 to 403). First, the speed is reduced to Z1% from a position L1 away from the work end surface to a position L2 away from the work end surface (step 401). Next, Z1% is maintained at a position where the work enters L3 (step 402). The original feed speed is returned to the position where the work enters L4 (step 403). According to the above steps, an acceleration / deceleration pattern is obtained in which the speed is reduced immediately before the tool comes into contact with the work, and the normal feed speed is obtained at a position where the work has entered a predetermined distance. Next, in the case where the tool separates from the work, a part that separates from the work is determined (step 404), and an acceleration / deceleration pattern of the part that separates from the work is set (step 405).
407). First, the speed is reduced to Z2% from the position L5 to L6 before the position away from the work (step 4).
05). Next, after moving away from the work, Z
Leave it at 2% (step 406). L8 from work
Accelerate to Z3% at a distance. Through the above steps, an acceleration / deceleration pattern is obtained in which the deceleration is performed immediately before the tool separates from the workpiece, and accelerates to the specified ratio after the tool is separated from the workpiece by a certain distance.

Using the function generated as described above, the feed speed of the tool is corrected at the time of actual machining. That is,
The acceleration / deceleration ratio is extracted from the designated feed speed F using the function generated as shown in FIG. 12, and the corrected feed speed Fo is obtained as shown in Expression 5. Here, Z is an acceleration / deceleration ratio obtained from the function of FIG.

F0 = F * (100 + Z) / 100 (5)

In the above embodiment, an example is shown in which the acceleration / deceleration ratio is used on the vertical axis of the function, but an actual feed speed may be used instead of the acceleration / deceleration ratio.
In this case, Equation 5 becomes Equation 6. F0 = Z (6) Here, Z is the corrected feed speed itself obtained from the function.

In the case of the measurement operation shown in FIG. 32, in order to increase the measurement accuracy, there is a case where the work 30 is once measured with the tool 31 and then measured again to measure again.
In this case as well, as shown in FIG. 14, by automatically generating a function indicating the acceleration / deceleration pattern of the second feed speed from the result of the first measurement, more accurate and lean measurement can be performed. In FIG. 14, a function 1 is a deceleration pattern of the tool feed speed in the first measurement. The deceleration band is widened for the first measurement, and the deceleration rate is also small to save the measurement time.

In general, the measurement accuracy decreases in proportion to the feed speed at the time of measurement. By the first measurement, the actual position of the work is estimated in consideration of the coasting amount of the measurement sensor and the like. A deceleration pattern such as the function 2 in FIG. 14 is automatically generated in consideration of a clearance value predetermined for the estimated position of the work. In the function 2, the measurement time is shortened by narrowing the deceleration band, and the measurement accuracy is increased by increasing the deceleration rate. FIGS. 15 and 16 show examples of automatically correcting a fuzzy membership function.
The membership function is changed based on the result of the second measurement. The fuzzy control rule in this case is if POS is Pithen FEED is Fi (i = 1 to 3). The tool feed speed (FEED) is changed according to the position (POS) of the tool. The membership functions at the first time are P1 to P3 and F1 to F in FIG.
As shown in FIG. 3, the width for determining that the workpiece has approached the work is widened (P1), and the membership function in the second measurement of the tool is corrected according to the expected position P 'of the work as shown in FIG. The width for determining that the workpiece has approached the work is reduced (P1), and the feed speed (F1) of the tool is set to a lower value.
The feed speeds (F2, F3) of the tool in the portions (P2, P3) away from the workpiece are set to be higher.

Embodiment 4 Next, a fourth embodiment of the present invention will be described with reference to the drawings. Generally, the feed rate of a tool largely depends on the material of a workpiece to be machined and the material of a tool to be machined. Therefore, it is conceivable to determine the standard feed speed and correct the feed speed based on the combination of the materials of the work and the tool. FIG. 17 shows rules for this. Here, Ti (i = 0 to 9) indicates the material of the tool, and Wz (z = 0 to 9) indicates the material of the work. However, when both TOOL and WORK are finely classified (10 levels) as shown in FIG. 17, the number of rules becomes very large. In the case of FIG. 17, the number of rules reaches 100 as shown in FIG. For this reason, when it is attempted to perform an operation on all the rules and extract the result, it takes a very long inference time. Therefore, the rule is executed only for TOOL and WORK corresponding to the material hardness of a given tool or work.

For example, when the hardness of the material of a tool or a work is properly given by a numerical value, only those whose membership function shown in FIG. 17 corresponds to this value are inferred. When the membership function of T5 is a value of hardness α to β, if the hardness K of the given tool is α ≦ K ≦ β, it is determined that T5 is applicable. Further, when the material hardness of the tool or the work is ambiguous and the hardness itself is given by a membership function, T0 to T9, W within the range of the hardness taken by the given membership function.
It is determined that the membership functions 0 to W9 are applicable. Assuming that the arrow portions of the membership functions of "TOOL" and "WORK" in FIG. 17 are the facts given to this membership function, the membership function whose domain is the given fact is given by "TOOL". Are ゛ T4 ”, ゛ T5” and ゛ WOR
In the case of "K", they are $ W6 "and $ W7".

FIG. 19 shows the result of extracting the corresponding membership function as described above.
However, in the WORK, since W6 to W7 correspond, only the rules F64, F65, F74, and F75 are executed to draw a conclusion. FIG. 20 is a flowchart showing an algorithm for extracting a rule to be executed. In FIG. 20, i indicates the number of the rule, and j indicates the number of the antecedent of each rule. The number of rules is n, and the number of antecedents of each rule is mi. i =
1 (step 601), and then j = 1 (step 602). The region (α, β) of the membership function of the j-th antecedent part of the i-th rule is extracted (step 603). The region of the membership function is a range of values of the defined membership function in the horizontal axis direction. It is checked whether the value Dij (fact) given to the membership function exists in the area extracted in step 603 (step 604). If Dij is not a specific value, for example, if Dij is also a membership function, it is determined whether or not there is an overlapping portion between Dij and (α, β).

If the determination in step 604 is YES,
The value of j is increased by 1 (step 605). j is m
It is checked whether it is a value within i, that is, whether there is a membership function of the antecedent part that has not been determined by the i-th rule (step 606). If the determination in step 606 is YES, the process returns to step 603; otherwise, the process proceeds to step 607. If YES is determined in the step 604, the execution flag of the i-th rule is turned off (step 608). If NO is determined in step 606, the execution flag of the i-th rule is turned ON (step 607). The rule execution flag is used to distinguish whether each rule should be executed or not.
When performing fuzzy inference, it is assumed that only rules for which this flag is ON are executed. In the processing of steps 602 to 608, if there is a rule in which no value Dij given to the membership function exists in the membership function area of the antecedent part of each rule, the rule is MAX_M
Since the value is always 0 as long as the inference is performed using the IN method, the rule execution flag is set to OFF as no inference is necessary. Next, the value of i is increased by 1 (step 6).
09). It is checked whether i is a value within n, that is, whether all the rules have been determined (step 610). If the determination in step 610 is YES, that is,
If there are still rules to be determined, step 60
Returning to step 2, if NO, the process ends.

Embodiment 5 FIG. Next, Embodiment 5 of the present invention
Will be described with reference to FIG. Generally, the shape of the fuzzy membership function has various shapes (51) as shown in FIG.
To 56) are conceivable. For this reason, conventionally, when defining rules in the example of FIG. 8, rules are defined as indicated by R1 to R3 in FIG. 8, and membership functions A1, A2,
A3, B1, B2 and B3 had to be defined anew. For this reason, a dedicated man-machine interface for defining the membership function was prepared, and a large storage area for storing the membership function inside the system was required. Therefore, in the present invention, the membership function is set to a specific shape pattern, and the definition of the membership function can be performed simultaneously with the definition of the rule. Reference numeral 57 in FIG. 21 shows an example of such a case, in which the membership functions are all defined by isosceles triangles, and Li indicates the length of the center position li of the base half of the isosceles triangle. In this case, the shape of the membership function is the center of the shape 57 (that is, the position of the center of gravity) Li and 1/1 of the base.
Since the length can be defined by two lengths li, it is only necessary to input Li and li. Therefore, it is possible to simultaneously define the rule and the membership function for the rule in FIG. 8 as shown in FIG.

In FIG. 22, A1, A2, A3, B
1, B2 and B3 are the centers of each membership function,
a1, a2, a3, b1, b2, and b3 indicate the length of half the base of each membership function. Also,
Other examples of membership function shape patterns include:
A case where a dead zone is provided as shown in FIG. 21 or a shape which is asymmetrical as shown in FIG. In this case, a parameter is added to change the membership function to (Li, li, m
It can be expressed in i).

Embodiment 6 FIG. Next, Embodiment 6 of the present invention
Will be described with reference to FIG. The shape of the generally defined membership function is mostly the shape indicated by 57 in FIG.
Therefore, when the shape pattern of the membership function is an isosceles triangle such as 57, when drawing a conclusion from the consequent part of the rule, as shown in FIG. 23, the fitness α (0 ≦ α ≦ 1)
When the inference result is obtained for the membership function 61 in the consequent part, the position of the center of gravity is an isosceles triangle.
Since the area value Si is trapezoidal, Si = (lj + li) * α. Here, lj is half the length of the upper base of the trapezoid, l i is half the length of the lower base, and α is the height. Since li: lj = 1: (1−α), lj = (1−α) li. Therefore, Si = (li + (1-α) li) * α = (2-α) * α * li. Therefore, the value of Si can be easily calculated from α and l i, and a conclusion can be drawn at high speed.

Embodiment 7 FIG. Next, a seventh embodiment of the present invention will be described. In conventional fuzzy inference, when summing up the inference results of each rule as shown in FIG.
In this method, the shapes of the membership functions resulting from the respective rules are superimposed, and the center of gravity of the obtained shape is obtained. FIG. 35 shows the results of the three rules, MAX_
The shape of the three membership functions is synthesized using the MIN method. When the conclusions of the rules are combined as described above, it is necessary to combine the shapes to extract a new shape, and it takes a long time to process. In the example of FIG. 35, the result of the rule 2 completely affects the final conclusion. No longer. That is,
In FIG. 35, even if the center position of the membership function of the consequent part of Rule 2 is slightly shifted, the final conclusion is not affected at all. Since such a case occurs, there is also a problem that it is difficult to adjust the membership function in the conventional method. Therefore, in the present invention, the final conclusion L is obtained by inferring Equation 7 from the barycenter position Li and the area Si resulting from each rule. Here, n is the number of rules. By doing so, the result of each rule always has an effect on the final conclusion, and since the shape is not synthesized, the processing for fuzzy inference can be performed at high speed. By combining with the method of the tenth invention, further high-speed processing becomes possible.

[0064]

(Equation 2)

Embodiment 8 FIG. Next, an eighth embodiment of the present invention will be described. Conventionally, in fuzzy inference, all rules are treated equally, and the importance of the rules cannot be reflected in the conclusion. In practice, however, it is necessary to consider the importance of the rules when drawing the final conclusion. Therefore, in the present invention, the importance of each rule is defined by, for example, VAL as shown in FIG. Important rules give VAL a large value and less important rules give a small value. Let βi be the importance of each rule, and determine the final conclusion L from the center of gravity position Li and the area Si as
It is determined by inference as follows. Here, n is the number of rules. By doing so, fuzzy inference can be performed in consideration of the importance of each rule. In addition, the importance βi is allowed to have a negative value, which makes it possible to set a negative rule.

[0066]

(Equation 3)

[0067]

As described above, according to the present invention, the rule for changing the feed rate is defined in advance in the form of a function, and the element for determining the feed rate required for the rule in the function form is defined. And the required output value is calculated based on the input element and the rules of the function form, and the calculated value is calculated based on a pre-defined formula to obtain the final feed speed. Know-how such as conditions can be easily changed without changing the system, and a complicated control considering various factors can be easily realized in order to draw a final conclusion. Conventionally, the feed speed can be changed only in a stepwise manner. However, according to the present invention, the change of the feed speed can be designated in the form of a function, so that the feed speed can be changed smoothly. According to the present invention, the function stored in the knowledge storage unit can be arbitrarily changed on the numerical control side. Therefore, by setting an optimum function in accordance with a machining situation, further optimum control becomes possible. . Thereby, when basic knowledge is set in the knowledge storage unit and the function value needs to be changed according to a certain rule, the processing can be performed without the operator changing the function each time. .

Further, according to the present invention, the rules stored in the knowledge storage unit are not uniformly executed, and only the rules determined to be required to be executed are inferred. It is possible to increase the speed of inference.

According to the present invention, the fuzzy rule itself and its membership function can be simultaneously and easily defined, the storage area of the membership function can be reduced, and the conclusion of each rule can be inferred at high speed.

Further, according to the present invention, it is possible to derive a final conclusion in consideration of the importance of each rule, and it is possible to perform more appropriate inference by adjusting the importance. Together with the adjustment of the membership function, more ideal rules can be set. In addition, when a negative value is entered for the importance, the area value at the conclusion of the rule becomes negative, which means that a negative rule has been set. This makes it possible to set a rule such as if A is A1 then Bis not B1, which could not be defined by the conventional method. In addition, since the importance of the rule is described in the "fuzzy rule",
When tuning the fuzzy rule, the importance can be easily changed in consideration of the result of the fuzzy control.

[Brief description of the drawings]

FIG. 1 is a main block diagram of a numerical control device according to the present invention.

FIG. 2 is a diagram illustrating an example of a feed speed control rule set in a knowledge unit according to the present invention.

FIG. 3 is a flowchart relating to processing of an inference unit according to the present invention.

FIG. 4 is a diagram showing an example of a feed speed control rule set in a knowledge unit according to the present invention.

FIG. 5 is a diagram illustrating an example of a membership function of feed speed control set in a knowledge unit according to the present invention.

FIG. 6 is a diagram showing an example of a feed speed control rule set in a knowledge unit according to the present invention.

FIG. 7 is a diagram illustrating an example of a membership function of feed speed control set in a knowledge unit according to the present invention.

FIG. 8 is a diagram showing an example of a feed speed control rule set in a knowledge unit according to the present invention.

FIG. 9 is a diagram illustrating an example of a membership function of feed rate control set in the knowledge unit according to the present invention.

FIG. 10 is an explanatory diagram for extracting a distance between an entry-prohibited area and a tool according to the present invention.

FIG. 11 is a flowchart for extracting a distance between a tool and a tool according to the present invention.

FIG. 12 is a diagram showing an example of a function stored in the knowledge unit according to the present invention.

FIG. 13 is a flowchart for creating a function of a knowledge unit according to the present invention.

FIG. 14 is an explanatory diagram relating to creation of a function of a knowledge unit at the time of measurement according to the present invention.

FIG. 15 is an explanatory diagram relating to creation of a function of a knowledge unit at the time of measurement according to the present invention.

FIG. 16 is an explanatory diagram relating to creation of a function of a knowledge unit during measurement according to the present invention.

FIG. 17 is a diagram showing an example of setting data of a knowledge unit for correcting a feed speed based on a tool and a material of a work according to the present invention.

FIG. 18 is a diagram illustrating an example of a rule matrix of a knowledge unit for correcting a feed speed based on a tool and a material of a work according to the present invention.

FIG. 19 is a diagram showing an example of executing only a corresponding rule from a rule matrix of a knowledge unit for correcting a feed rate from a tool and a material of a work according to the present invention.

FIG. 20 is a flowchart of extracting rules to be executed according to the present invention.

FIG. 21 is a diagram showing a form of a membership function according to the present invention.

FIG. 22 is a diagram showing a rule setting method according to the present invention.

FIG. 23 is a diagram showing an operation method of a membership function according to the present invention.

FIG. 24 is a diagram showing a rule setting method according to the present invention.

FIG. 25 is a block diagram of a main part of a conventional feed speed control unit.

FIG. 26 is an explanatory diagram of a program path in a conventional corner unit.

FIG. 27 is an explanatory diagram of a conventional corner override function.

FIG. 28 is an explanatory view of a conventional drilling process.

FIG. 29 is an explanatory view of a conventional drilling process for a tapered portion.

FIG. 30 is an explanatory view of processing of a conventional molding material work.

FIG. 31 is an explanatory diagram of processing of a conventional intermediate mold.

FIG. 32 is an explanatory diagram of a conventional measurement.

FIG. 33 is an explanatory diagram of control in setting a no-entry area in the related art.

FIG. 34 is a block diagram of a conventional numerical control device.

FIG. 35 is a diagram showing a conventional membership function synthesis method using the MAX_MIN method.

FIG. 36 is a diagram showing a conventional method of calculating a membership function.

FIG. 37 is a diagram showing a form of a conventional membership function.

[Explanation of symbols]

 REFERENCE SIGNS LIST 1 machine tool 11 tape reader 12 memory 13 high power control device 14 MDI 15 display unit 16 operation panel 17 control device 20 CNC command data 21 pulse distribution processing unit 22 servo control unit 23 servo motor 24 feed speed processing unit 25 knowledge storage unit 26 inference Part 30 Work 31 Tool 32 Entry prohibited area

Claims (4)

[Claims] In a numerical control device capable of changing a feed rate, a knowledge storage unit for storing one or more rules for changing the feed speed described in a functional form, and a knowledge storage unit Input the elements for determining the processing conditions required for the function-type rules stored in the, and obtain the required output values based on the input elements and the function-type rules, and define the obtained values in advance. An inference unit that calculates the final feed rate by calculating based on the mathematical formula that has been performed, and a unit that automatically generates a function stored in the knowledge storage unit based on the shape of the workpiece to be processed or the workpiece shape to be measured. A numerical control device for controlling the controlled object based on the feed speed obtained by the inference unit. 2. A knowledge storage unit for storing a plurality of rules for changing control conditions in the form of a production rule, and a fuzzy inference unit for inferring a desired control condition from the rules stored in the knowledge storage unit. Means for extracting only rules including a membership function corresponding to an input value from a plurality of rules stored in the knowledge storage unit, and providing means for performing inference using only the extracted rules. A fuzzy inference apparatus characterized in that: 3. A fuzzy inference unit comprising: a knowledge storage unit in which a rule for changing a control condition is described in the form of a production rule; and a fuzzy inference unit for inferring a desired control condition from the rule stored in the knowledge storage unit. A fuzzy inference apparatus characterized in that the membership functions used for the consequent part of the inference are all isosceles triangles, and the membership functions are directly defined in the rules. 4. In the knowledge storage unit, the production rule for changing the control condition is directly assigned to the importance β of the rule.
i is stored, and when performing fuzzy inference, the importance βi defined in the above rule
The fuzzy inference apparatus according to claim 2 or 3, wherein the inference result is obtained from the following equation using the area Si and the center of gravity position Li of the result of each rule. (Equation 1)