JP5777184B2 - Forming machine - Google Patents

Description

  The present invention presses a forming tool against a wire, sheet metal, or other metal material to form a workpiece sequentially, and uses the correction amount calculated based on an actual measurement size of a dimension management target portion of the workpiece to determine the pressing amount of the forming tool. It relates to a forming machine to be corrected.

  Conventionally, as an example of this type of forming machine, there has been known a spring forming machine that forms a compression coil spring by pressing a forming tool against a wire and measures the spring length (natural length of the spring) of the formed compression coil spring one by one. It has been. In this spring forming machine, the spring length can be changed by changing the amount of pressing of the pitch tool, which is one of the forming tools, to the wire. Therefore, in the spring forming machine described above, the coil length is then adjusted so that the spring length approaches the target dimension by using a correction amount calculated by multiplying the deviation between the measured dimension of the spring length and the target dimension by a predetermined coefficient. The pressing amount of the pitch tool when forming the spring was corrected (for example, see Patent Document 1).

Japanese Patent Laid-Open No. 08-10883 (Claim 1)

Incidentally, the cause of the deviation between the measured dimension and the target dimension of the spring length varies, internal stress of the wire, the thermal deformation and the like various elements of the forming tool. However, it is difficult to identify. Therefore, conventionally, the coefficient to be multiplied by the deviation between the actually measured spring length and the target dimension is determined by trial and error based on the operator's intuition, so that skill is required and the above-described correction functions sufficiently. In many cases, there is a problem that the dimension of the part to be dimension-controlled in the workpiece is not stable.

  The present invention has been made in view of the above circumstances, and it is an object of the present invention to provide a forming machine capable of stabilizing the dimensions of a workpiece dimension management target without requiring an operator to be skilled.

  The inventor of the present application spectrum-analyzed changes in the spring length of a plurality of workpieces formed by a spring forming machine as an example of a forming machine, and confirmed that the change in the spring length has periodicity. The exact cause of this periodicity is unknown, but if the fluctuation period is known, control means for suppressing the fluctuation is well known. However, the change in the spring length includes many fluctuations without periodicity due to accidental errors. Therefore, the inventor of the present application performs a time series analysis based on a model of the change in the spring length, finds that the change in the spring length can be predicted by the autoregressive model, and that the correction amount can be calculated using the model formula. The present invention of claims 1 to 5 has been completed.

  In addition, the technology to control using the model formula of the autoregressive model is conventionally known. However, the combustion amount control for stabilizing the boiler temperature to the target temperature and the traveling direction of the ship are stabilized. It is used for control to stabilize continuously changing values of management objects (for example, boiler temperature and ship traveling direction), such as steering angle control, and is formed by a forming machine. There is no precedent for using a model formula of an autoregressive model for control to stabilize the value of a management target that changes intermittently, such as the size of a workpiece to be changed.

The forming machine according to the first aspect of the present invention is to form a workpiece sequentially by pressing a forming tool against a wire, sheet metal, or other metal material, and also subject to dimensional control of the workpiece that changes in accordance with the pressing amount of the forming tool against the metal material. In the forming machine that measures the dimensions of the part one by one in the dimension measurement unit and corrects the pressing amount of the forming tool using the correction amount calculated based on the measured dimensions of the part to be dimensionally controlled, the number of samples set in advance The correction amount data generated by the pseudo correction amount generation unit for generating the correction amount data at random within a preset range and the correction amount data generated by the pseudo correction amount generation unit are sequentially used. Sample data acquisition means for forming workpieces for the number of samples while sequentially correcting the pressing amount, and acquiring measured dimension data for the number of samples; Model formula identification means for identifying the model formula of the autoregressive model of the actual measurement dimension and correction amount from the data of the correction amount for the number of samples and the actual measurement dimension by a statistical method, the model formula, and the actual measurement dimensions of the latest past multiple workpieces and having features and a correction amount, the next was a correction amount calculating means for calculating a correction amount at the time of molding the work.

According to a second aspect of the present invention, in the forming machine according to the first aspect, the actually measured dimension of the _i-th workpiece is set to y _i and the correction amount is set to u _i , and the measured dimension is set to y _i and the correction amount. Assuming that autoregressive coefficients related to u _i are a _i and b _i and p and q are preset dimension numbers, the model formula is described by the following formula 1, and the model formula identifying means The a _i and b _i of the equation (1 ₎ are determined by the least square method using the correction amount and the measured dimension data, and the correction amount calculator is a correction operation of the following equation (2) obtained by modifying the equation (1). It is characterized in that the correction amount is calculated using an equation.

According to a third aspect of the present invention, in the forming machine according to the first aspect, the actually measured dimension of the _i- shaped workpiece is set to y _i and the correction amount is set to u _i , and the measured dimension is set to y _i and the correction amount. If the autoregressive coefficients related to u _i are a _i and b _i, and the number of dimensions is p and q, and the upper limit values M and N of the number of dimensions p and q, the model equation is described by In the model formula identifying means, M × N models obtained by substituting any integer of 1 to M for the number of dimensions p and substituting any one of 1 to N for the number of dimensions q of Equation 1. Model equation candidate generation means for determining a _i and b _i of the equations by the least square method, and the dimensions and measured dimensions of the dimension management target portions of a plurality of workpieces calculated by each of these M × N model equations A model formula that calculates the sum of squares of the differences and minimizes the sum of the squares And a model formula determining means for determining a model formula to be used by the correction amount calculating means, and the correction amount calculating means is a correction calculation of the following formula 2 obtained by modifying the model formula of the formula 1 determined by the model formula determining means. It is characterized in that the correction amount is calculated using an equation.

  According to a fourth aspect of the present invention, in the forming machine according to any one of the first to third aspects, a plurality of latest measured dimensions and correction amount data updated every time a workpiece is formed is used. It has a feature in that it is equipped with a model formula updating means for re-identifying the model formula by a statistical method.

According to a fifth aspect of the present invention, in the forming machine according to the fourth aspect , there is provided a process capability index calculating means for calculating a process capability index with respect to the upper limit standard and the lower limit standard of the dimension of the dimension management target portion set in advance. A feature is that the model formula update means re-identifies the model formula when the process capability index deteriorates beyond a preset threshold value .

In the forming machine according to the first aspect of the present invention, by using a model equation of an autoregressive model of the latest measured dimensions of a plurality of past works and the correction amount at the time of forming the work, Since the correction amount with respect to the pressing amount is calculated, it becomes possible to control the pressing amount of the forming tool while suppressing the influence of variations in forming, and the dimensions of the dimension management target portion in the workpiece are stabilized. Moreover, since the forming machine itself of the present invention identifies the model expression of the autoregressive model, no skill is required.

Specifically, as in the invention of claim 2, the model equation of the autoregressive model is set in advance with the measured dimension as y _i , the correction amount as u _i , and the autoregressive coefficients as a _i and b _i. If the number of dimensions is p and q, the model formula is described by the following [Equation 1]. Then, the model equation can be identified by determining a _i and b _i of the model equation of [Equation 1] by the least square method using the correction amount for the number of samples and the data of the actually measured dimensions. Further, the correction amount can be calculated by the following equation [2] obtained by modifying the model equation [1].

Further, as in the invention of claim 3, the upper limit values M and N of the dimension numbers p and q are set in advance, and any integer of 1 to M is substituted for p and q in [Equation 1] is set. The accuracy of identification is improved because the model equation having the optimum number of dimensions is determined from a _i and b _{i of the} M × N model equations obtained by substituting in any integer of 1 to N, and the workpiece is dimensionally controlled. The dimensional stability of the part is increased.

  In the forming machine according to claims 4 and 5, the model formula is re-identified by a statistical method using a plurality of latest measured dimensions and correction amount data updated every time the workpiece is formed. The dimensional stability of the site to be managed is further increased.

The front view of the spring molding machine which concerns on 1st Embodiment of this invention. Side view of spring forming machine An enlarged perspective view of a part of the spring forming machine Conceptual diagram of a spring molding machine controller Conceptual diagram showing the coil spring forming process Spring forming program flowchart Flow chart of correction control preparation process Flow chart of correction formula coefficient determination processing Flow chart of the spring forming program of the second embodiment Flowchart of the spring forming program of the third embodiment Flow chart of correction control preparation process Flow chart of correction formula generation processing Flowchart of the spring forming program of the fourth embodiment Flowchart of correction control preparation process in the spring forming program of the fourth embodiment Flowchart of correction formula coefficient determination processing according to a modification of the present invention

[First Embodiment]
Hereinafter, an embodiment of the present invention will be described with reference to FIGS. FIG. 1 shows a spring forming machine 10 as a forming machine of the present invention. The spring forming machine 10 is configured by assembling a wire rod feeding device 50, a pair of vertical linear motion mechanisms 30, 30, a pair of inclined linear motion mechanisms 40, 40, and the like to a base 11 standing upright.

  The wire rod feeding device 50 includes a pair of feed rollers 51 and 52 that are arranged on the front surface of the base 11 and are adjacent to each other in a vertically aligned state. A wire guide 12 (generally called “quill”) extends on a common tangent line between the pair of feed rollers 51 and 52, and a wire insertion hole (not shown) extends through the wire guide 12. Is formed through. A wire rod 90 (corresponding to the “metal material” of the present invention) is sandwiched between the feed rollers 51 and 52, and a pair of feed rollers 51 and 52 are connected by a feeding servo motor 54 (see FIG. 4). By rotating symmetrically, the wire rod 90 is fed to the right side in FIG. 1 and fed out from the tip of the wire rod guide 12.

  As shown in FIG. 1, a mandrel 13 protrudes from the front surface 11 </ b> A of the base 11 in front of the wire guide 12 with respect to the wire guide 12. As shown in FIG. 3, the mandrel 13 has a bar shape with a semicircular cross section, and the flat side surface 13 </ b> A faces the wire guide 12 side.

  As shown in FIG. 1, the pair of inclined linear motion mechanisms 40, 40 are attached to additional plates 41, 41 fixed to the front surface 11 </ b> A of the base 11. Then, one inclined linear motion mechanism 40 extends from the mandrel 13 toward the diagonally upward vertical side opposite to the wire feeding device 50, and the other inclined linear motion mechanism 40 is below the one inclined linear motion mechanism 40. It is arrange | positioned and it extends in the diagonally upward horizontal direction as it leaves | separates from the mandrel 13. As shown in FIG.

  Each inclined linear motion mechanism 40 is provided with a slider 43 that linearly moves in the direction in which they extend, and a servo motor 46 is connected to the slider 43 via a crank slider mechanism 40K. That is, an eccentric shaft 45J protruding from a position away from the center of rotation is provided on a rotating disk 45 that is rotationally driven by a servo motor 46, and the eccentric shaft 45J and the slider 43 are connected by a link member 42. The rotation output of the servo motor 46 is converted into the linear motion of the slider 43.

  The coiling tool 16 is fixed to the mandrel 13 side of the slider 43 of each inclined linear motion mechanism 40. As shown in FIG. 3, the coiling tool 16 provided in one inclined linear motion mechanism 40 is abutted against the mandrel 13 from obliquely above, and the coiling tool 16 provided in the other inclined linear motion mechanism 40 is It is abutted against the gold 13 obliquely from below. Further, a round groove 16 </ b> M is formed on the tip surface of the coiling tool 16. The wire rod 90 fed from the wire rod feeding device 50 is guided to the circular groove 16M of each coiling tool 16, and the wire rod 90 is formed (formed) into an arc shape so as to surround the mandrel 13, and the coil spring 91 is formed. (Corresponding to “work” of the present invention), and the coil spring 91 grows away from the base 11 as shown in FIG.

  As shown in FIGS. 1 and 2, the up-and-down linear motion mechanisms 30 and 30 are arranged symmetrically vertically with the mandrel 13 interposed therebetween. Each vertical translation mechanism 30 includes a slider 31 that moves up and down, and a servo motor 35 is connected to the slider 31 via a crank slider mechanism 30K, similarly to the tilt translation mechanism 40 described above. Further, the cutting tool 15 is fixed to the lower end of the slider 31 of the upper vertical translation mechanism 30, and when the cutting tool 15 is lowered together with the slider 31, the wire 90 is cut between the core bar 13.

  On the other hand, a pitch tool 14 is fixed to the upper end portion of the slider 31 of the lower vertical linear motion mechanism 30. As shown in FIG. 3, the tip of the pitch tool 14 is wedge-shaped and has an inclined surface 14A, and this inclined surface 14A contacts the wire 90 constituting the coil spring 91 from the base 11 side in the winding axis direction. It touches. When the pitch tool 14 is moved upward and the pressing amount of the pitch tool 14 against the wire 90 is increased, the pitch and the spring length of the coil spring 91 are increased, and the pitch tool 14 is moved downward to the wire 90. When the pressing amount of the pitch tool 14 is reduced, the pitch and the spring length of the coil spring 91 are reduced.

  FIG. 4 shows a controller 60 for controlling the spring molding machine 10. The controller 60 includes a servo amplifier of each servo motor described above and a control circuit 60C to which these servo amplifiers are connected. The control circuit 60C includes a CPU 61 and a main storage device 64. The main storage device 64 includes specifications of the coil spring 91 (pitch, coil diameter, number of turns, spring length, wire diameter, wire length, etc.). The spring forming program PG1 (see FIGS. 6 to 8) is stored. When the CPU 61 of the main storage device 64 executes the spring forming program PG1, the servo motors 35, 46, 54 described above are driven and controlled, and the coil springs 91 are sequentially formed from the wire 90.

  More specifically, when the spring forming process (S12) in the spring forming program PG1 (FIGS. 6 to 8) is executed, the wire feeding device 50 is configured to store the “wire length” stored in the main storage device 64. The wire 90 is fed. Then, the fed wire rod 90 abuts against the inner surfaces of the round grooves 16M and 16M in the coiling tools 16 and 16, and slides against the inclined surface 14A of the pitch tool 14 to form the coil spring 91. Specifically, in the initial stage of the spring forming operation, as shown in FIGS. 5A to 5B, the pitch tool 14 is disposed at the first forming position slightly elevated from the initial position, and in this state, 1 is set. The wire rod 90 is fed to form the end winding 91A.

  Next, as shown in FIGS. 5C to 5D, the pitch tool 14 rises from the first molding position to the second molding position, and in that state, a plurality of turns of wire rod 90 is fed to form a coil. The main body 91B is formed, the pitch tool 14 is lowered again to the first forming position, and as shown in FIG. 5 (E), one volume of wire rod 90 is fed, and the end winding 91C on the rear end side is moved. Molded. Then, after the pitch tool 14 is retracted to the standby position separated from the coil spring 91, the cutting process (S14) is executed, and the cutting tool 15 is lowered from the initial position as shown in FIG. 91 is separated from the subsequent wire rod 90 and then returned to the initial position, and the pitch tool 14 also returns to the initial position (see FIG. 5A).

  In order to detect the spring length of the coil spring 91, the spring molding machine 10 is provided with a spring length detector 20 shown in FIG. Although there are a capacitance type and an image processing type for detecting the spring length, the spring length detector 20 of this embodiment is a capacitance type, and the tip of the support arm 17 protruding forward from the front surface of the base 11. And is arranged in front of the mandrel 13. The spring length detector 20 has a detection chip 24 attached to the tip thereof, and the detection chip 24 has a disk shape having substantially the same diameter as the coil diameter of the coil spring 91.

The spring length detector 20 outputs a signal obtained by converting the capacitance value into a voltage value in accordance with the gap amount between the detection chip 24 and the coil spring 91, and a spring length detection circuit 20A provided in the controller 60. In addition, a detection voltage signal is applied. The relationship between the gap amount and the detected voltage signal and the detection chip 24 and the coil spring 91 to the main memory 64, and the reference voltage are stored in advance, from the reference gap amount detected voltage signal at CPU61 Convert to deviation. Based on this deviation amount, the axial length of the coil spring 91 (hereinafter referred to as “spring length”, referred to as “dimension management target part” in the present embodiment) is detected. More specifically, after the end of the spring forming process (S12), the spring length detection process (S13) of the spring forming program PG1 is executed, the data conversion from the detected voltage signal to the spring length is completed, A cutting process (S14) is executed. In the present embodiment, the spring length detection unit 20 and the controller 60 constitute a “dimension measurement unit” according to the present invention.

  In the spring forming process (S12) in the spring forming program PG1 described above, the position of the pitch tool 14 when forming the coil main body 91B is corrected based on the detection result of the spring length detection process (S13). That is, pitch tool position data for positioning control of the pitch tool 14 is stored in the main storage device 64, and the upper position from the initial position of the pitch tool 14 determined by the pitch tool position data X for the second molding position is stored. Correction is performed by adding a correction amount of the movement distance of the pitch tool 14 determined by the correction data to the movement amount (ie, the pressing amount of the pitch tool 14 against the wire 90). In the spring molding machine 10 of the present embodiment, a correction calculation formula (see [Expression 2] described later) for calculating the correction data is generated by a statistical method described below.

  Specifically, when the spring forming program PG1 is executed as shown in FIG. 6, a correction control preparation process (S10) for generating a correction arithmetic expression is executed. In this correction control preparation process (S10), as shown in FIG. 7, in order to obtain statistical data of the spring length of the coil spring 91, an input of the number of samples L is obtained (S21). Therefore, when an arbitrary integer (for example, 100) is input to the number of samples L, for example, data of L correction amounts that vary within a predetermined range set in advance using an M-sequence random number is generated. It is sequentially stored in the mold correction amount data u [n] (n = 1 to L) (S22). The CPU 61 executing step S22 corresponds to the “pseudo correction amount generation means” of the present invention.

  For example, on condition that the start button provided on the console 69 (see FIG. 4) of the controller 60 is turned on (YES in S23), the above-described spring forming process (S12) and spring length detecting process (S13). The same processing as the cutting processing (S14), the spring forming processing (S24), the spring length detection processing (S25), and the cutting processing (S26) is repeatedly executed L times (NO loop of S28). ), L coil springs 91 are formed.

  At this time, in the spring forming process (S24) executed repeatedly, the L correction amount data u [n] (n = 1 to L) are sequentially used to correct the pitch tool position data X. The In addition, the detection result of the spring length by the spring length detection process (S25) is sequentially stored in, for example, array type spring length data y [n] (n = 1 to L). The CPU 61 when executing the above-described steps S24 to S28 corresponds to the “sample data acquisition means” of the present invention.

  Next, using the spring length data y [n] and the correction amount data u [n] (n = 1 to L) for the L number of samples, the following [Equation 1] is performed in the correction formula coefficient determination process (S29). ] Is identified.

In particular, the actual dimensions of the coil spring 91 molded into i pieces prior to the correction amount and u _i with the y _i, the autoregressive coefficients of them actually measured dimensions y _i and correction amount u _i a _i, If b _i is set and p and q are the preset number of dimensions, the model expression shown in the above [Expression 1] can be established statistically. The actual measurement dimension y _i of the coil spring 91 may be the entire length of the coil spring 91, or may be a deviation of the actual dimension of the coil spring 91 from the target value of the coil spring 91. Further, the correction amount u _i may be NC data that specifies the rotation amount of the servo motor, or may be a linear motion amount that the pitch tool 14 moves. Here, in order to simplify subsequent calculations, the measured dimension y _i of the coil spring 91 is assumed to be a deviation from the target value 0. Further , the correction amount u _i is a linear motion amount to which the pitch tool 14 moves.

Then, in the correction formula coefficient determination process (S29), using the upper limit values M and N of the dimension numbers p and q set in advance (for example, both M and N are about “10”), the above [Equation 1] M × N model formulas obtained by substituting any integer of 1 to M for the number of dimensions p of the model formula and substituting any one of 1 to N for the number of dimensions q of [Formula 1]. Is generated. Then, the autoregressive coefficients a _i and b _i of each of the M × N model equations are obtained by using the spring length data y [n] and the correction amount data u [n] (n = 1 to L) for the above-described L samples. ) And stored in the main storage device 64 (S500). The CPU 61 executing the process of step S500 corresponds to the “model formula candidate generation unit” according to the present invention.

Next, the predicted spring length y ′ _i (i = 1 to l) of the coil springs 91 for the number of samples L using each of M × N model equations in which the autoregressive coefficients a _i and b _i are specified. ) Is calculated and stored in the main storage device 64 (S501). When calculating the predicted spring length y ′ _i (i = 1 to 1), if the number of samples reaches the dimension numbers p and q, the previous predicted spring length y ′ _i is calculated as “0”. To do. Specifically, when the number of dimensions p and q are both “10”, when the 12th predicted spring length y ′ ₁₂ is calculated, the actually measured dimension y _i and the correction amount for the last 10 dimensions are calculated. two of the u _i uses the data of the actually measured dimension _{y i} and correction amount _{u i,} the one remaining 8, calculates the predicted spring length y _'12 the measured dimension _{y i} and correction amount _{u i} as "0" To do.

Next, a difference sum of squares J between the predicted spring length y ′ _i and the actually measured dimension y _i is calculated for each of the M × N model formulas (S502). Then, using the model formula having the smallest difference sum of squares J, a correction calculation formula of the following [Equation 2] is generated (S503), the correction formula coefficient determination process (S29) is skipped, and shown in FIG. Also exits from the correction control preparation process (S10). The CPU 61 when executing the correction formula coefficient determination process (S29) corresponds to the “model formula identification unit” according to the present invention, and the CPU 61 when executing the processes of the above-described steps S502 and S503. Corresponds to the “model formula determining means” according to the present invention.

After exiting from the correction control preparation process (S10), as shown in FIG. 6, the above-described spring forming process (S12), spring length detecting process (S13), and cutting process (S14) are repeated unless a forming stop request is made. It is executed (NO loop of S15). At that time, before performing the spring forming process (S12), by the correction calculation equation above [Equation 2], then the correction operation of the correction amount u _i of the time of molding the coil spring 91 is above [Equation 2] It is calculated by an equation (S11). That is, the measured dimension y _i and the correction amount u _i of the coil spring 91 corresponding to the latest latest number of dimensions p and q, which are updated each time the coil spring 91 is molded, are expressed in the above correction equation of [ Equation 2]. substituting corrected amount u _n it is calculated, the amount of pressing against the wire 90 of the pitch tool 14 when the then molding the coil spring 91 in the correction amount u _n is corrected. The CPU 61 executing step S11 corresponds to the “correction amount calculation means” of the present invention.

  It should be noted that, as the above-described molding stop request, when a stop button (not shown) provided on the console 69 is turned ON, the coil spring 91 reaches a preset production number, or an abnormality occurs. The spring molding machine 10 stops (YES in S15, S16), and the spring molding program PG1 ends.

  As described above, in the spring molding machine 10 according to the present embodiment, the model equation of the autoregressive model (see the above [Equation 1]) of the actually measured dimensions of the coil spring 91 and the correction amount when the coil spring 91 is molded is used. Then, since the correction amount for the pressing amount of the pitch tool 14 when the coil spring 91 is formed next is calculated, it becomes possible to control the pressing amount of the pitch tool 14 while suppressing the influence of variations in forming. The dimension of the dimension management target part of is stable. In addition, since the spring forming machine 10 itself identifies the model expression of the autoregressive model, the skill of the operator is not required.

[Second Embodiment]
The spring forming program PG2 executed by the spring forming machine according to the present embodiment is shown in FIG. 9, and is updated each time the coil spring 91 is formed. _The difference from the first embodiment is that the update process (S17) for updating the correction equation of [Formula 2] is performed using _i and the correction amount u _i . Specifically, in the update process (S17), the above-described correction is performed by sequentially replacing the actually measured dimension y _i and the correction amount u _i for the L number of samples with the latest one each time the coil spring 91 is formed. The control preparation process (S10) is executed, and the model expression of [Expression 1] and the correction operation expression of [Expression 2] are regenerated. Other configurations are the same as those in the first embodiment.

  According to the spring molding machine of the present embodiment, the above-mentioned model formula [Numerical formula 1] and the above [Numeric formula] are used by using a plurality of latest measured dimensions and correction amount data updated every time the coil spring 91 is molded. 2] is regenerated, so that the dimensional stability of the part to be dimension-controlled of the workpiece is further increased.

In the update process (S17), the number of dimensions of the model formula determined in the correction control preparation process (S10) may be adopted as it is, and only the autoregressive coefficients a _i and b _i may be determined again.

[Third Embodiment]
In the first and second embodiments, each time the spring forming programs PG1 and PG2 are executed, the optimum values of the dimensional numbers p and q in the model equation of [Equation 1] are expressed by the evaluation equation shown in FIG. However, in the present embodiment, the optimum values of the dimension numbers p and q are obtained in advance, and only the autoregressive coefficients a _i and b _i are obtained one by one with the dimension number fixed. Specifically, in the spring forming machine according to the present embodiment, assuming that the optimum value of the dimension number p is 1 and the optimum value of the dimension number q is 2, a model expression that is a general expression of the above [Equation 1] is as follows: In addition to the equation [3], the correction equation, which is the general equation of [equation 2], is set as the following [equation 4]. The optimum values of the dimension numbers p and q are experimentally obtained by the same method as the correction control preparation process (S10) of the first embodiment.

  The spring forming program PG3 of the present embodiment is shown in FIG. 10, and the spring forming program PG2 of the second embodiment is a correction control preparation process (S10 ′) and a correction formula generation process (S11 ′). Only the update process (S17 ′) is different. As shown in FIG. 11, when the correction control preparation process (S10 ′) is executed, an arbitrary input is performed as in steps S21 and S22 of the correction control preparation process (S10) of the first and second embodiments. Correction amount data u [n] (1 to L) is randomly generated for the number L. When the start button is turned on (YES in step 23), the loop process including steps S24 to 28, 200, and 201 is repeated L times.

  In the loop process, a coefficient calculation process (S200), which will be described in detail later, is first executed, and then the coil spring 91 is formed in the same manner as steps S24 to S27 of the correction control preparation process (S10) of the first embodiment. After measuring the spring length and cutting the coil spring 91 from the subsequent wire, the loop counter C is incremented. Each time the coil spring 91 is formed by repeating the loop processing, the n = 1 to L-th correction amount data u [n] are used one by one in ascending order of the number of n, and the wire 90 of the pitch tool 14 is used. Correct the pressing amount.

Further, the correction control preparation process of the present embodiment (S10 '), each time of molding the coil spring 91, by performing an update process (S201), and the latest actual dimension y _n past the measured dimension y of the coil spring 91 Place, the measured dimension y _n until that was current past, shifts to the latest actual dimension y _n-1 to the second past. As a result, each time the coil spring 91 is formed, the latest measured dimensions y _n and y _n−1 of the latest two spring lengths are updated. Further, the correction amount data u _n-1 and un _-2 corresponding to the actually measured dimensions y _n and y _n-1 of the two spring lengths are also shifted one by one.

Then, returning to the first coefficient calculation process (S200) of the loop process, the autoregressive coefficients a _i and b _i of the model equation [Equation 3] are changed to the two latest measured dimensions y _n and y _n−1 in the past, It is obtained by the least square method using _two correction amount data u _n-1 and u _n-2 corresponding to them.

When the coefficient calculation process (S200) is first executed immediately after the start button is turned on (S23), the two latest measured dimensions y _n and y _{n−1 in} the past are not measured and two correction amount data are set. u _n−1 and u _n−2 do not exist, and when step S200 is executed for the second time, the latest measured dimension y _n−1 of the previous latest is not measured and the two correction amount data u _{n −1} and un _-2 are not present, and when step S200 is executed for the third time, one correction amount data un _-2 is not present. On the other hand, the coefficient calculation process (S200) is performed with the unmeasured measured dimension y and the nonexistent correction amount data u set to “0”.

When the number of moldings of the coil spring 91 reaches L and the correction control preparation process (S10 ′) is completed, a loop process including steps S11 ′, S12 to S15, and S17 ′ shown in FIG. 10 is performed. The correction formula generation process (S11 ′) during the loop process is shown in FIG. 12, and the correction formula update process (S211) is added to the model formula update process (S200) in the correction control preparation process (S10 ′). ). The update process (S17 '), the correction control preparation process (S10' updating in) and (S201) has become the same. That is, by the loop process including steps S11 ′, S12 to S15, and S17 ′ shown in FIG. 10, the correction calculation formula is updated together with the model formula every time the coil spring 91 is formed, and the correction calculation formula is obtained by the updated correction calculation formula. The next coil spring 91 is formed using the correction amount data. As described above, this embodiment can achieve the same effects as those of the second embodiment.

[Fourth Embodiment]
The spring forming program PG4 executed by the spring forming machine of the present embodiment is shown in FIGS. 13 to 14, and the update process (S17) is performed on the condition that the process capability index _Cpk has deteriorated beyond the threshold. This is different from the second embodiment.

Specifically, as shown in FIG. 14, the correction control preparation process of spring forming program PG4 of the present embodiment (S100), next to the input of the number of samples L (S21), the lower limit threshold of the process capability index _{C pk} LCL And the upper limit UL and lower limit LL of the spring length of the coil spring 91 and the number CN of coil springs 91 required to start the determination of the process capability index _Cpk (hereinafter referred to as “determination start number CN”). An input is requested (S101 to S103). Other processes of the correction control preparation process (S100) are the same as the correction control preparation process (S10) described in the first embodiment, and the correction calculation of the above [Equation 2] is performed as in the first embodiment. An expression is generated.

Then, in the spring forming program PG4, as shown in FIG. 13, the number of coil springs 91 formed is counted by the counter D (S110), and while the count number of the counter D has not reached the determination start number CN. (NO in S 111), without calculating the process capability index _{C pk,} as in step S11~S14 of the first embodiment, point by point correction amount by generated by correction control preparation process (S100) correction calculation equation The coil springs 91 are sequentially formed while calculating and correcting the pressing amount of the pitch tool 14 against the wire 90.

When the count number of the counter D reaches the determination start number CN (YES in S111), the measured dimension data of the coil spring 91 corresponding to the latest latest determination start number CN, the upper limit standard UL, and the lower limit standard LL Then, the process capability index C _pk is calculated (S112). When the process capability index C _pk deteriorates beyond the lower limit threshold LCL (YES in S113), the updating process (S17) described in the second embodiment is performed, and the model of the above [Equation 1] is performed. The counter D is reset after regenerating the equation and the correction equation of the above [Equation 2], and the process returns to the formation of the coil spring 91. On the other hand, when the process capability index _Cpk is larger than the lower limit threshold LCL (NO in S113), the counter D is reset without performing the update process (S17), and the process returns to the formation of the coil spring 91.

[Other Embodiments]
The present invention is not limited to the above-described embodiment. For example, the embodiments described below are also included in the technical scope of the present invention, and various other than the following can be made without departing from the scope of the invention. It can be changed and implemented.

  (1) In the first embodiment, the upper limit values M and N of the dimension numbers p and q are set in advance. However, the operator may arbitrarily input M and N.

(2) In the first to third embodiments, the spring forming machine is taken as an example of the forming machine. However, the present invention may be applied to a forming machine that forms a workpiece by bending a sheet metal into a predetermined shape. Good.

  (3) Further, for example, as in the correction formula coefficient determination process (S29 ′) shown in FIG. 15, a term of the delay number K is provided in the correspondence between the actually measured dimension y and the correction amount u, and the delay The number K may be configured to be arbitrarily input by the operator.

10 Spring forming machine
14 Pitch tool (forming tool)
17 Support arm 20 Spring length detection part (dimension measurement part)
60 Controller (Dimension measuring section)
90 Wire material (metal material)
91 Coil spring (work)
PG1, PG2, PG3, PG4 Spring forming program

Claims (5)

The workpiece is sequentially formed by pressing the forming tool against a wire rod, sheet metal, or other metal material, and the dimension of the dimension management target part of the workpiece that changes according to the pressing amount of the forming tool against the metal material is used as a dimension measuring unit. In the forming machine that corrects the pressing amount of the forming tool using the correction amount calculated based on the actual measurement dimensions of the part to be dimensionally controlled,
Pseudo correction amount generation means for generating data of the correction amount for a preset number of samples so as to vary randomly within a preset range;
The workpieces for the number of samples are formed while sequentially correcting the pressing amount of the forming tool by sequentially using the correction amount data generated by the pseudo correction amount generating means, and the actual measurement for the number of samples. Sample data acquisition means for acquiring dimension data;
A model formula identifying means for identifying a model formula of the autoregressive model of the measured dimension and the correction amount by a statistical method from the correction amount and the measured dimension data for the number of samples;
Said model equation, and characterized in that from said measured dimensions and the correction amount of the most recent plurality of the workpiece past, then a correction amount calculating means for calculating the correction amount at the time of molding the workpiece Forming machine. The measured dimension of the workpiece formed i times before is set to y _i , the correction amount is set to u _i, and the autoregressive coefficient relating the measured size to y _i and the correction amount to u _i are a _i , b _{Assuming i} and p and q as the preset number of dimensions , the model formula is described by the following formula 1,
The model formula identification means determines the a _i and b _i of the model formula of Formula 1 by a least square method using the correction amount for the number of samples and the data of the measured dimensions,
2. The forming machine according to claim 1, wherein the correction amount calculation unit calculates the correction amount using a correction calculation formula of the following formula 2 obtained by modifying the model formula of the formula 1.

The measured dimension of the workpiece formed i times before is set to y _i , the correction amount is set to u _i, and the autoregressive coefficient relating the measured size to y _i and the correction amount to u _i are a _i , b Assuming that _i is the number of dimensions , p and q, and the upper limit values M and N of the number of dimensions p and q, the model equation is described by the following equation 1,
In the model formula identification means, M × which is obtained by substituting any integer of 1 to M for the dimension number p and substituting any integer of 1 to N for the dimension number q of the equation 1. Model expression candidate generation means for determining the a _i and b _i of the N model expressions by a least square method ;
The sum of squares of the differences between the dimensions of the dimension management target portions of the plurality of workpieces calculated by the M × N model formulas and the measured dimensions is calculated, and the sum of the squares becomes the smallest. A model formula determining means for determining the model formula to be the model formula used in the correction amount calculating means,
2. The forming according to claim 1, wherein the correction amount calculation unit calculates the correction amount by a correction calculation formula of the following formula 2 obtained by modifying the model formula of the formula 1 determined by the model formula determination unit. Machine.

  Characterized in that it comprises a model formula update means for re-identifying the model formula by a statistical method using a plurality of the latest measured dimensions and the correction amount data updated every time the workpiece is molded. The forming machine according to any one of claims 1 to 3. A process capability index calculating means for calculating a process capability index for the upper limit standard and the lower limit standard of the dimension of the dimension management target portion set in advance,
5. The forming machine according to claim 4, wherein when the process capability index deteriorates exceeding a preset threshold value, the model formula update unit re-identifies the model formula .

Images