JP2012200807A - Parameter automatically adjusting device of screw fastening robot - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a parameter automatically adjusting device of a screw fastening robot automatically setting a parameter of high accuracy with reduced frequencies of operation tests regardless of the degree of skill without necessitating a means for detecting a fastening torque on actual operation on lines.SOLUTION: A parameter estimation part 58 calculates the optimum value of the screw fastening parameter using a fastening torque estimation model showing the relation between a screw fastening parameter and a fastening torque learned from the history of operation tests. The frequencies of operation tests are reduced by simplification of the fastening torque estimation model and formation of test parameters based on a D optimization program and a confidence interval. Thereby, frequencies of required operation tests, that is, the number of consumed test pieces are made to be near about tens of pieces, that of an expert.

Description

  The present invention relates to a parameter automatic adjustment device for a screw tightening robot.

  Conventionally, Patent Literature 1 is known as a screw fastening robot for fastening a screw to a target workpiece. Such a screw tightening robot includes a first period of meshing, a second period of tightening, a third period of tightening for torque adjustment, and a fourth period of stabilizing the tightening torque. Execute in this order. In this case, in order to ensure the correct operation of the screw tightening robot, parameters such as the supply current value to the motor, the screw rotation speed, and the screw rotation speed are precisely set in each step from the first phase to the fourth phase. It is necessary to adjust to. In particular, these parameters require that the screw tightening torque be exactly matched to the target tightening torque, that the variation in tightening torque be small, and that the time required for parameter adjustment be short.

  However, the parameter setting in the conventional screw tightening robot is actually performed by assembling the test piece with the screw tightening robot. That is, the parameter is set based on the user's own experience and intuition from data obtained by assembling the test piece using the screw tightening robot. In this case, the consumption of the test piece depends on the skill level of the user. For example, a user with a high level of skill can set a parameter with an operation test of several tens of times using a test piece, whereas a user with a low level of skill can set a target even if an operation test is performed over 100 times. In some cases, approximate values of parameters cannot be obtained. On the other hand, it is not realistic to set a parameter only with a highly skilled user because there is not always a highly skilled user at every site. In addition, it is not realistic for a user with a low level of skill to secure a sufficient number of test pieces despite the increased consumption of test pieces as described above.

  Incidentally, Patent Document 2 has been proposed as a technique for automatically setting the parameters of these robots regardless of the skill level of the user. Patent Document 2 discloses that a load is estimated by actually operating a robot, and parameters are set according to the load. Further, Patent Document 3 discloses that a torque detection unit is provided in a screw tightening robot so that the actual tightening torque is controlled based on the torque detected by the torque detection unit rather than the tightening torque control by parameter adjustment. Has been.

  However, in the case of Patent Document 2, the estimated load corresponds to the load at the tip of the robot arm. Therefore, it is difficult to apply to a screw fastening robot that fastens a screw to a workpiece by rotating a bit. Moreover, in the case of patent document 3, a torque detection means is required as a component of a screw fastening robot. For example, torque detection means such as a torque sensor is expensive. In particular, if various parameters of the screw tightening robot can be accurately set before the actual operation on the line, the torque detecting means is not required during the robot operation on the line after the parameter adjustment. For this reason, it is desirable that the torque detection means be used only at the time of adjustment, and that the torque detection means be unnecessary when operating on the line.

Japanese Patent No. 3811999 JP-A-7-261844 JP-A-6-170662

  Accordingly, an object of the present invention is to provide a screw tightening robot that automatically sets a high-accuracy parameter with a small number of operation tests regardless of skill level without requiring a means for detecting a tightening torque during actual operation on a line. It is to provide an automatic parameter adjustment apparatus.

In the invention according to any one of claims 1 to 9, a highly accurate parameter can be automatically set with a small number of operation tests regardless of the skill level. This is based on the fact that in the invention according to any one of claims 1 to 9, a skilled person always performs an operation test in the vicinity of a parameter that is considered to be optimal while slightly changing this parameter. It is. The reason why the skilled person can adjust the screw tightening robot with less test piece consumption and less parameters is that the characteristics of the robot are efficiently grasped with less operation tests. That is, the skilled person estimates the optimum value in advance by using his own experience and intuition, and adjusts the test parameter in the vicinity of the optimum value. On the other hand, when this is automatically performed, it is conceivable that a random test parameter is first generated and an optimum value is searched. However, generating random test parameters in this way increases the consumption of motion tests, ie test pieces. Therefore, in the invention according to any one of claims 1 to 9, the operation test is performed in the vicinity of the parameter considered to be optimal while slightly changing this parameter. Thereby, it is possible to automatically set a highly accurate parameter with a small number of operation tests regardless of the skill level of the operator.
In particular, in the invention according to claim 4, 5 or 6, the test parameter is generated by using the D optimal plan or the confidence interval to imitate an expert. Thereby, the number of operation tests can be greatly reduced regardless of the skill level of the operator.

In the invention according to claim 8 or claim 9, the parameter estimation means includes a rotational speed ω ₃ in the third period for adjusting the screw tightening torque, a current τ ₄ in the fourth period for stabilizing the screw, and A coefficient serving as a parameter is estimated using the actual measurement value τ _f acquired by the detachable torque acquisition means. The relationship between the rotational speed ω ₃ in the third period and the current value τ ₄ in the fourth period is the relationship between the tightening torque τ _f detected by the torque detection means when the test piece is assembled. It is modeled by a quadratic expression including a logarithmic term. However, in this case, there are many terms included in the model, and the coefficient to be searched increases. As a result, the number of operation tests required for coefficient search, that is, the number of test pieces to be consumed also increases. On the other hand, considering the operating range of the screw tightening robot, that is, the rotational speed ω ₃ and current τ ₄ in the region where the screw tightening robot operates, it is not always necessary to strictly apply the modeled expression. That is, if the ranges of the rotational speed ω ₃ and the current τ ₄ are limited, it is not necessary to obtain the coefficients of all terms included in the model. Therefore, in the case of the invention described in claim 8, the parameter estimation means is:
τ _f = uτ ₄ + w log (ω ₃ ) + v
The coefficient u, the coefficient w, and the coefficient v are estimated from the simplified model. Thereby, the coefficients u, v, and w are acquired as sufficient parameters in the operation range of the screw tightening robot. Then, by obtaining the coefficients u, v, and w, the parameter estimation means allows the measured torque value τ _f of the tightening torque to approximate the target tightening torque τ _g and the third period of rotation speed ω ₃ and the fourth period. Estimate the current τ ₄ . At this time, since there are three types of coefficients, u, v, and w as described above, the number of operation tests necessary to converge these parameters, that is, the number of test pieces to be consumed is about several tens. And get close to skilled people. Therefore, it is possible to automatically set a highly accurate parameter with a small number of operation tests regardless of the skill level.

Further, in the invention according to claim 8 or claim 9, the screw tightening parameter in the third period and the fourth period, that is, τ _f = τ _g , is set by using the above model. The rotational speed ω ₃ and the current τ ₄ are estimated. At this time, the actual measured value τ _f of the tightening torque used in the above model is necessary only when setting these parameters. That is, by setting parameters for the screw tightening robot, the screw tightening robot adjusts the tightening torque in the third period without using the actual measured value of the tightening torque by the torque acquisition means. Therefore, when the screw tightening robot is operated in the line after setting the parameters, it is not necessary to install a torque acquisition means in each screw tightening robot. In particular, in the invention according to claim 8 or claim 9, even if the parameters of a plurality of screw tightening robots are automatically adjusted by making the torque acquisition means detachable, there is only one torque acquisition means. Good. Therefore, the parameters of a plurality of screw tightening robots can be automatically adjusted without preparing many expensive torque acquisition means.

By the way, in the model of the invention described in claim 8, the rotational speed ω ₃ in the third period is a logarithmic term. This is because the relationship with the tightening torque τ _f becomes non-linear when the rotational speed ω ₃ is small, so that this can be corrected. On the other hand, if the rotational speed ω ₃ in the third period is in a practical range, the relationship with the tightening torque τ _f is hardly affected. Therefore, in the invention according to claim 9, the parameter estimation means is further simplified as a model,
τ _f = uτ ₄ + wω ₃ + v
Are used to estimate the coefficients u, v, and w. Even if a model simplified in this way is used, parameters sufficient for the operation of the screw tightening robot can be estimated. Therefore, with a simpler process, a highly accurate parameter can be automatically set with a small number of operation tests regardless of skill level.

The block diagram which shows the parameter automatic adjustment apparatus of the screw fastening robot by one Embodiment Schematic diagram showing an example of a screw tightening robot The fragmentary sectional view which expanded the principal part of the screw fastening robot shown in FIG. Schematic showing the flow of operation of the screw tightening robot Schematic which shows the initial data input into the parameter automatic adjustment apparatus by one Embodiment Schematic which shows the data which the parameter automatic adjustment apparatus by one Embodiment acquires from a screw fastening robot Schematic which shows the data which the parameter automatic adjustment apparatus by one Embodiment produces | generates by internal processing Schematic showing the current change in the third phase Schematic which shows the data which the parameter automatic adjustment apparatus by one Embodiment outputs Schematic showing the flow of initial characteristic investigation processing in the automatic parameter adjustment apparatus according to one embodiment Schematic showing the relationship between the amount of rotation in the second period and the amount of stable rotation in the third period Schematic diagram showing the result of linear regression of the relationship between the amount of rotation in the second period and the amount of stable rotation in the third period Explanatory drawing for demonstrating the procedure which calculates | requires the relationship between the rotation amount in a 2nd period, and the stable rotation amount in a 3rd period by robust linear regression Schematic which shows the example which tests the 4th period electric current value-the 3rd period speed by case 1 in the parameter automatic adjustment device by one embodiment. Schematic which shows the example which tests the 4th period electric current value-the 3rd period speed by case 2 in the parameter automatic adjustment device by one embodiment. Schematic which shows the example which tests the 4th period electric current value-the 3rd period speed by case 3 in the parameter automatic adjustment device by one embodiment. Schematic showing the flow of parameter adjustment processing in the automatic parameter adjustment device according to one embodiment Schematic showing the relationship between the amount of rotation in the second period, the amount of stable rotation in the third period, and the amount of rotation in the second period Schematic showing the relationship between rotation speed and tightening time in the third phase Schematic showing the coefficients for calculating the current value in the third period, the seating current value in the third period, and the judgment current value in the fourth period The schematic diagram which shows the example of a display of the optimal parameter of the 4th period electric current value calculated with the parameter automatic adjustment apparatus by one Embodiment The schematic diagram which shows the example of a display of the optimal parameter of 2nd rotation amount calculated with the parameter automatic adjustment apparatus by one Embodiment

Hereinafter, an embodiment of an automatic parameter adjustment apparatus for a screw tightening robot will be described with reference to the drawings.
(Screw tightening robot)
First, a screw tightening robot that is a target for setting parameters by the automatic parameter adjustment device will be described.
As shown in FIG. 2, the screw fastening robot 10 includes a head portion 11, a stay 12, and a connector box 13. As shown in FIGS. 2 and 3, the head unit 11 includes a linear drive mechanism unit 14, a rotary drive mechanism unit 15, and a bit 16. The linear drive mechanism unit 14 drives the head unit 11 to reciprocate linearly with respect to the stay 12. In the case illustrated in FIG. 2, the linear drive mechanism unit 14 drives the head unit 11 in the vertical direction of FIG. 2. For example, as shown in FIG. 3, the linear drive mechanism unit 14 includes a motor 17, a rack 18, a pinion 19, and the like that are power sources. Thereby, the head unit 11 moves in the vertical direction in FIG. 2 by energizing the motor 17. The rotation drive mechanism unit 15 corresponds to a drive unit in claims, and includes a motor 21 and a shaft 22 as shown in FIG. The motor 21 has a rotating shaft 23 connected to the shaft 22 and rotationally drives the shaft 22 by supplying electric power. The bit 16 is connected to the end of the shaft 22 opposite to the motor 21 and is driven to rotate by the shaft 22. The head unit 11 has an adsorption pipe 24. The suction pipe 24 covers the outer peripheral side of the bit 16, and the tip on the bit 16 side is open. The suction pipe 24 is connected to the suction portion 26 via the pipe 25 on the side opposite to the open tip. The suction unit 26 decompresses the inside of the suction pipe 24 via the pipe 25. Thereby, a screw (not shown) to be screwed is attracted to the tip of the suction pipe 24 and attached to the bit 16.

  As shown by a broken line in FIG. 3, a torque sensor 27 can be attached to the rotating shaft 23 of the motor 21. The torque sensor 27 is detachably attached to the rotating shaft 23 of the motor 21. The torque sensor 27 is attached to the head unit 11 when parameters are set by a parameter automatic adjustment device to be described later, and is removed from the head unit 11 when parameter setting is completed. This torque sensor 27 corresponds to the actually measured tightening torque acquisition means in the claims. The motor 21 and the suction unit 26 of the head unit 11 are covered with a cover 28.

Next, the operation of the screw fastening robot 10 having the above configuration will be described.
As shown in FIG. 1, the screw fastening robot 10 fastens the screw 34 into the screw hole 33 of the workpiece 32 placed on the base 31. At this time, as shown in FIG. 4, the screw fastening robot 10 fastens the screw 34 to the workpiece 32 in the first, second, third, and fourth steps.
In the first stage, the screw fastening robot 10 meshes the screw 34 with the screw hole 33. That is, the screw tightening robot 10 aligns the screw 34 attached to the tip of the bit 16 by the decompression of the suction part 26 with the axis of the screw hole 33 of the workpiece 32. Then, the screw fastening robot 10 rotates the screw 34 attached to the bit 16 to mesh the screw 34 and the screw hole 33. At this time, the rotation speed of the bit 16 is set to a medium speed that is lower than the second period and higher than the third period. Thereby, the positive meshing between the screw 34 and the screw hole 33 is achieved. Moreover, the force which presses the screw 34 against the workpiece | work 32 is set small compared with the 4th period from the other 2nd period. Thereby, the sticking of the screw | thread 34 accompanying the inclination of the screw | thread 34 and unnecessary biting is avoided.

  When the first period ends, the screw fastening robot 10 proceeds to the second period in which the screw 34 meshed with the screw hole 33 in the first period is screwed into the screw hole 33. The screw tightening robot 10 rotates the bit 16 at a higher speed than the other first period and third period, and screws the screw 34 engaged with the screw hole 33 in the first period into the screw hole 33. Thereby, the screw 34 is screwed into the screw hole 33 at high speed, and the time required for screwing is shortened. Further, the force for pressing the screw 34 against the work 32 is set to be stronger than the first period and smaller than the third period and the fourth period. Thereby, the screw 34 is screwed into the screw hole 33 stably.

When the second period ends, the screw tightening robot 10 moves to a third period in which the screw 34 screwed in the second period is tightened and the torque applied to the screw 34 is adjusted to a preset tightening torque. The screw tightening robot 10 rotates the bit 16 at a lower speed than the other first and second periods, and tightens the screw 34 screwed at a high speed in the second period. Thereby, the screw 34 screwed into the screw hole 33 is adjusted with high accuracy to a preset tightening torque. Moreover, the force which presses the screw 34 against the workpiece | work 32 is set larger than the 1st period and the 2nd period. Thereby, the screw 34 is prevented from being crushed.
When the third period ends, the screw tightening robot 10 stops the tightening of the screw 34 whose tightening torque has been adjusted in the third period, and shifts to the fourth period in which the tightening torque applied to the screw 34 is stabilized. The screw fastening robot 10 stops the rotation of the bit 16 and stops the screw 34 from being screwed into the workpiece 32. Moreover, the force which presses the screw 34 against the workpiece | work 32 is set strongly similarly to the 3rd term. Thereby, like the third period, the screw 34 is prevented from being crushed.

  As described above, the screw tightening robot 10 sequentially tightens the screw 34 into the screw hole 33 of the work 32 by sequentially executing the first period to the fourth period. The operation of the screw tightening robot 10 is executed as a series of flows according to an instruction from the controller 36 connected to the screw tightening robot 10 as shown in FIG. When adjusting parameters with the automatic parameter adjustment device, the screw tightening robot 10 assembles a test piece including the screw 34 and the workpiece 32. The test piece is the same member as the screw 34 and the workpiece 32 that are actually assembled by the screw fastening robot 10.

(Automatic parameter adjustment device)
Next, an automatic parameter adjustment device that automatically adjusts the control parameters of the screw tightening robot having the above-described configuration will be described.
As shown in FIG. 1, the parameter automatic adjustment device 40 is communicably connected to the controller 36 of the screw tightening robot 10. The parameter automatic adjustment device 40 is realized in software by executing a parameter automatic adjustment program in a personal computer having a control unit 41 including a CPU, RAM, and ROM (not shown), for example. A personal computer constituting the parameter automatic adjustment device 40 includes an input unit 42 such as a mouse and a keyboard, an output unit 43 such as a display and a printer, a storage unit 44 such as an HDD and an EEPROM, and a communication interface 45 that performs communication with a controller. have.

  The parameter automatic adjustment device 40 executes automatic parameter adjustment based on various data acquired from the screw fastening robot 10 by executing a parameter automatic adjustment program in the control unit 41. Specifically, the parameter automatic adjustment device 40 includes a third period rotational speed acquisition unit 51, a fourth period supply current acquisition unit 52, and a measured tightening torque acquisition unit 53 that acquire various data from the controller 36 of the screw tightening robot 10. have. The parameter automatic adjustment device 40 includes a rotation speed input unit 54, a target torque input unit 55, and a second period initial rotation amount setting unit 56 that receive input of data set by the user. Further, the parameter automatic adjustment device 40 includes a constant determination unit 57 and a parameter estimation unit 58. The third period rotation speed acquisition unit 51 acquires the rotation speed of the bit 16 in the third period of the screw fastening robot 10, that is, the rotation speed of the motor 21 from the controller 36. The rotation speed of the motor 21 is acquired using, for example, a current supplied to the motor 21, a voltage applied to the motor 21, or a device that directly detects the rotation speed of the motor 21 or the bit 16. The fourth period supply current acquisition unit 52 acquires the supply current to the motor 21 in the fourth period of the screw fastening robot 10 from the controller 36. The measured tightening torque acquisition unit 53 is configured in cooperation with the torque sensor 27, and acquires an actual measured value of the tightening torque detected by the torque sensor 27 from the controller 36. The rotational speed input unit 54, the target torque input unit 55, and the second period initial rotation amount setting unit 56 receive data input from the user that is input via the input unit 42 using the display screen of the output unit 43. . The constant determination unit 57 and the parameter estimation unit 58 are realized by software in the parameter automatic adjustment device 40 by executing a parameter automatic adjustment program. The parameter estimation unit 58 constitutes a test parameter generation unit 61, a tightening torque estimation model generation unit 62, and a tightening torque learning unit 63.

The third period rotational speed acquisition unit 51 acquires the rotational speed ω3 at which the screw 34 is fastened to the workpiece 32 in the third period. Specifically, the third period rotational speed acquisition unit 51 determines whether the third period from the current supplied to the motor 21, the voltage applied to the motor 21, or the rotation number directly detected by the rotation number sensor 71. The rotational speed ω3 at which the screw 34 is fastened to the work 32 is acquired. The fourth period supply current acquisition unit 52 acquires the current τ ₄ supplied to the motor 21 of the rotation drive mechanism unit 15 in the fourth period. The measured tightening torque acquisition unit 53 is a torque applied to the screw 34 when the screw 34 is screwed into the screw hole 33 of the workpiece 32 from the torque sensor 27 described above, that is, the rotating shaft of the motor 21 connected to the bit 16 via the shaft 22. The torque applied to 23 is acquired.

The rotational speed input unit 54 accepts input of arbitrary high-reliability rotational speed ω _3c1 , medium-reliable rotational speed ω _3c2 and low-reliable rotational speed ω _3c3 as rotational speeds for tightening the screw 34 in the third period. The high reliability rotation speed ω _3c1 , medium reliability rotation speed ω _3c2, and low reliability rotation speed ω _3c3 are all arbitrary values that can be set by the user. The target torque input unit 55 receives an input of a target tightening torque τ _g that is a target value of the tightening torque in the third period. This target tightening torque τ _{g is} also a target value of a user's arbitrary tightening torque expected from the screw tightening robot 10 in the third period. Constant determining section 57, to the target tightening torque tau _g, calculates the constants Cτ, including any error range set in advance. The second period initial rotation amount setting unit 56 sets the initial rotation amount in the second period to “one time” when the test piece is first assembled in the screw tightening robot 10 to acquire the initial characteristics of the screw tightening robot 10. To do. The parameter estimation unit 58 estimates parameters for adjusting the screw tightening robot 10 based on the various data acquired above and the various values set by the input from the user. Each time the test piece is assembled, the test parameter generation unit 61 generates a test parameter related to the tightening torque of the screw 34. Each time the test piece is assembled, the storage unit 44 stores a history in which the test parameter generated by the test parameter generation unit 61 and the actual torque value τ _f acquired by the actual tightening torque acquisition unit 53 are associated with each other. The tightening torque estimation model generation unit 62 generates a tightening torque estimation model in which the tightening torque is estimated based on the screw tightening parameters of the screw tightening robot 10. The tightening torque learning unit 63 learns the tightening torque estimation model generated by the tightening torque estimation model generation unit 62 using the test parameters stored in the storage unit 44 and the history of the actually measured torque value τ _f and learns the tightening torque after learning. Generate an estimation model. The parameter estimation unit 58 estimates the optimum value of the screw tightening parameter in the screw tightening robot 10 using the post-learning tightening torque estimation model learned by the tightening torque learning unit 63. Detailed processing of the parameter estimation unit 58 will be described later.

(Parameter adjustment processing)
Hereinafter, the flow of parameter adjustment processing by the parameter automatic adjustment device 40 having the above-described configuration will be described in detail.
<Definition of symbols and constants>
The symbols and constants used in the following description are defined as follows.

  The constants defined above are those used in this embodiment, and can be arbitrarily changed depending on the characteristics of the screw fastening robot 10 and the types of the target screw 34 and workpiece 32. In the present specification and claims, there is a difference in typeface between characters displayed in the document and characters displayed in the mathematical image, but the difference in typeface is the same unless otherwise specified. It shall be treated as a character. In the following explanation, “First period” is “1 period”, “Second period” is “2 period”, “Third period” is “3 period” and “Fourth period” is “4 period”. May be expressed in a simplified manner, but they shall be treated as having the same significance. Furthermore, the characters shown on the right shoulder of each coefficient C are not “power numbers” but “characters” for identifying each coefficient.

<Initial setting data>
The user inputs various initial data shown in FIG. 5 when performing parameter adjustment using the parameter automatic adjustment device 40. The initial data is input from the input unit 42 of the personal computer using an input screen displayed on the output unit 43 such as a display of the personal computer constituting the parameter automatic adjustment device 40. In FIG. 5, “call” means a standard of the screw 34 such as M3 or M4. “Target tightening torque τ _g ” corresponds to “target value of tightening torque”. The “adjustment mode” is an arbitrarily selectable mode prepared when parameters are adjusted using the parameter automatic adjustment device 40. In the case of the present embodiment, the parameter automatic adjustment device 40 provides “accuracy priority mode”, “time priority mode”, and “speed fixed mode”. The “accuracy priority mode” is a mode in which priority is given to the accuracy of the rotation speed in the third period so that the adjusted tightening torque approximates the target value. The “time priority mode” is a mode in which the rotation speed of the third period is prioritized so that the average tightening time is shortened. The “speed fixed mode” is a mode for fixing the rotation speed in the third period. The “third period speed candidate” is a speed candidate in the third period. In this embodiment, three arbitrary speed candidates can be set. In this case, “3rd phase speed candidate 1” is set in all the adjustment modes described above. On the other hand, in the “accuracy priority mode”, all of three speed candidates “third period speed candidate 1”, “third period speed candidate 2”, and “third period speed candidate 3” are set. In this case, “3rd phase speed candidate 1” corresponds to the 1st speed candidate and has the highest expected value within the range assumed in the 3rd period, that is, the arbitrary speed that seems to be most suitable as the rotation speed in the 3rd period. Is set as the value of. On the other hand, “3rd period speed candidate 2” corresponds to the second speed candidate, and is set as an arbitrary value with higher safety than the “3rd period speed candidate 1”, although the expected value is lower. Similarly, “3rd period speed candidate 3” is set as an arbitrary value that has a lower safety value than “3rd period speed candidate 2” but has higher safety.

<Operation test result of screw tightening robot>
By executing an operation test of the screw tightening robot 10, the parameter automatic adjustment device 40 acquires data shown in FIG. 6 from the screw tightening robot 10. The “tightening torque” shown in FIG. 6 is a torque applied to the fastening of the screw 34 obtained by an operation test of the screw fastening robot 10. This “tightening torque” is acquired from the torque sensor 27 to the actually measured tightening torque acquisition unit 53. The “tightening torque” may be input by manually measuring the additional tightening torque or the loosening torque after executing the tightening operation by the screw tightening robot 10. “Control current data” is a current waveform of the rotating shaft 23 of the motor 21 that rotates the bit 16. “Control step data” is data indicating which “period” the screw fastening operation is currently in. “Tightening time” is the time required for screw tightening by the screw tightening robot 10 in the operation test.

<Data obtained by processing operation test results>
The parameter automatic adjustment device 40 generates the following data shown in FIG. 7 by internal processing using the data acquired from the screw fastening robot 10. Here, the third period stable section is a section in the third period in which the washer attached to the screw 34 is not affected by the seating on the end surface of the workpiece 32. “Third period stable time” shown in FIG. 7 means the time required for the third period stable section. Further, the “third period stable rotation amount” means the rotation amount of the third period stable section.

The parameter automatic adjustment device 40 acquires, as the third period maximum current value “τ ₃ ^max ”, the value at which the “control current data” is maximum in the range where the “control step data” is in the third period. Here, in the third period, the parameter automatic adjustment device 40 ^sets the section of τ ₃ ≦ τ ₃ ^max × C ^StTh as the third period stable section. Moreover, the parameter automatic adjustment device 40 sets the time of the third period stable section as “the third period stable time t _s3 ” as described above.

  If the third period is divided into a plurality of sections, the longest section is set as the third stable section. For example, in the second period, the motor 21 that rotationally drives the screw 34 rotates at a high speed, and thus a large amount of current is supplied. For this reason, the current in the second period is maintained at the start of the third period, and a short interval may be calculated in the initial period of the third period. As described above, by setting the longest section as the third period stable section, the influence of this short section on the following processing is eliminated.

Further, although the possibility of ^occurrence is small, when the coefficient C ^StTh is small, there is a possibility that the third period stable section may be divided along with the change of the current due to the load during screwing. This can be detected by the following procedure. That is, normally, the required time t ₃ in the third period and the time of the stable section in the third period, that is, the third period stable time t _s3 are not significantly different. Therefore, as _{t s3 / t 3 <C x} , it is possible to detect the division of the third phase stable section. In this case, the parameter automatic adjustment device 40 is updated as, for example, C ^StTh = C ^StTh + (1−C ^StTh ) / 2, and the operation test of the screw tightening robot 10 is performed again or adjustment is performed by notifying that C ^StTh is low. May be terminated with “NG”.

The parameter automatic adjustment device 40 calculates “three-phase stable rotation amount R _s3 ” by the following equation.
3rd period stable rotation amount R _s3 = 3rd period stable time t _s3 × 3rd period rotation speed ω ₃ / motor speed coefficient C ^ms
Here, the coefficients are set as follows based on FIG. FIG. 8 is a schematic diagram showing a change in the current τ ₃ in the third period.

・ Three-period stable interval calculation coefficient [-]: C ^StTh = 0.33
The value of C ^StTh is determined by the influence of a washer attached to the screw 34 without detecting a natural load change naturally generated during the screwing operation by the screw tightening robot 10, that is, a ripple-like change in FIG. It is a value that can eliminate an increase in current due to. In the present embodiment, C ^StTh = 0.33 is empirically set.

-Motor speed coefficient [ms]: C ^ms = 1200
This motor speed coefficient C ^ms corresponds to the time (ms) required for one rotation of the motor 21 when the rotation speed of the motor 21 is 1 (%). Therefore, the motor speed coefficient C ^ms is a value that is automatically determined when the type or type of the motor 21 is determined. In the present embodiment, C ^ms = 1200 is set as described above.

<Parameters that the parameter automatic adjustment device outputs to the screw tightening robot>
The parameter automatic adjustment device 40 outputs the following data shown in FIG. 9 to the screw fastening robot 10. Specifically, the parameter automatic adjustment device 40 includes the “rotation amount in the second period”, the “limit value of the screw tightening torque in the third period”, the “seating determination torque in the third period”, and the “screw in the third period”. “Tightening speed”, “Screw tightening torque limit value in the fourth period”, and “Screw tightening torque determination value in the fourth period” are output. These parameters are not different from the parameters output by the known screw tightening robot 10.

<Test history data>
The parameter automatic adjustment device 40 accumulates the acquired data and the output data in the storage unit 44 as a history each time an operation test is executed in the screw tightening robot 10.

<Flow of initial characteristic investigation process>
When the initial setting data is input, the parameter automatic adjustment device 40 proceeds to the initial characteristic investigation process. The initial characteristic investigation process will be described with reference to FIG. The initial characteristic investigation process is a process for acquiring the initial characteristic of the screw tightening robot 10 by first assembling the test piece in the screw tightening robot 10. Here, the test piece is a test member including the screw 34 and the workpiece 32 to be assembled by actually applying the screw fastening robot 10 as described above.

  The automatic parameter adjustment device 40 operates the screw tightening robot 10 using parameters with higher safety in the initial characteristic checking process, and checks the initial characteristics of the screw tightening robot 10. That is, immediately after the start of parameter adjustment using the automatic parameter adjustment device 40, that is, when the test piece is assembled by first operating the screw fastening robot 10, the amount of screwing the screw 34 into the workpiece 32 is unknown. Therefore, it is unclear how much the amount of rotation in the second period should be set. Here, the rotation speed in the second period is higher than that in the other periods as shown in FIG. For this reason, if the amount of rotation in the second period is too large, the screwed screw 34 is finally seated on the work 32 with a large impact. As a result, not only the workpiece 32 and the screw 34 but also the screw fastening robot 10 may be damaged. Therefore, the automatic parameter adjustment device 40 first executes a rotation amount characteristic test in the initial characteristic process, and gradually increases the rotation amount in the second period. Then, the parameter automatic adjustment device 40 roughly sets the rotation amount in the second period, and then investigates the relationship between the current and speed and the tightening torque in the second period.

  When the process proceeds to the initial characteristic investigation process, the parameter automatic adjustment device 40 executes “rotation amount characteristic test 1” and acquires parameters in this “rotation amount characteristic test 1” (S101). Then, the parameter automatic adjustment device 40 tests the operation of the screw fastening robot 10 (S102), and determines whether or not the “rotation amount characteristic test 1” is completed (S103).

In “rotation amount characteristic test 1”, the parameter automatic adjustment device 40 tests the operation of the screw tightening robot 10 in a state where the rotation amount in the second period is minimized. That is, in the “rotation amount characteristic test 1”, the second period initial rotation amount setting unit 56 sets the initial value of the rotation amount in the second period to “one time”. In addition, if the initial value of the rotation amount in the second period is “0”, it is synonymous with the direct transition from the first period to the third period and the disappearance of the second period. Therefore, the characteristics of the screw tightening robot 10 without the second period are different from the characteristics of the screw tightening robot 10 that is actually used, and parameter adjustment becomes difficult. Therefore, the second period initial rotation amount setting unit 56 sets the initial value of the rotation amount in the second period to “one time”. In the “rotation amount characteristic test 1”, the automatic parameter adjustment device 40 uses “third period speed candidate 1” shown in FIG. 5 as the rotation speed in the third period, and the current value in the fourth period is set in advance. Use the default values specified. In the present embodiment, no special screw is assumed such that the amount of rotation of the screw fastening robot 10 in the second period is “less than once”.
Therefore, the test parameter generation unit 61 sets the test parameters in the “rotation amount test 1” as follows.

  The parameter automatic adjustment device 40 executes “rotation amount characteristic test 1” in S101, and when the accuracy of the “average value of the three-phase stable rotation amount” obtained becomes sufficiently high, “rotation amount characteristic test 1” is executed. finish. Here, as a standard for ensuring accuracy, there is generally a method in which point estimation of an average value is performed and accuracy is ensured when an estimated interval is equal to or less than a preset threshold value. However, the average value of the stable rotation amount in the third period, that is, the “average value of the third period stable rotation amount” hardly changes even if the dispersion changes the workpiece 32 or the screw 34. Therefore, the estimated interval is substantially determined by the number of tests. Therefore, in the case of the present embodiment, the parameter automatic adjustment device 40 executes the “rotation amount characteristic test” when the test number is “3 times”, that is, the test piece is assembled “3 times” by the “rotation amount characteristic test 1” in S102. 1 ”is completed. For example, when the test is “NG” as in the case where the engagement between the screw 34 and the work 32 in the first period cannot be achieved, the number of tests is not counted.

  When testing the operation of the screw tightening robot 10 in S102, the parameter automatic adjustment device 40 first transfers various parameters to the screw tightening robot 10. In this case, the parameter automatic adjustment device 40 transfers all parameters set in the transfer stage to the screw tightening robot 10. The screw tightening robot 10 operates using various parameters transferred from the parameter automatic adjustment device 40 and executes screwing of the screw 34 into the workpiece 32. At this time, the parameter automatic adjustment device 40 acquires the current waveform of the motor 21 and the tightening torque detected by the torque sensor 27 from the operating screw tightening robot 10. The parameter automatic adjustment device 40 stores the acquired current waveform and tightening torque in the storage unit 44.

  When the parameter automatic adjustment device 40 determines that “rotation amount characteristic test 1” is completed in S103 (S103: Yes), it executes “rotation amount characteristic test 2” and sets the parameters in this “rotation amount characteristic test 2”. Obtain (S104). Then, the parameter automatic adjustment device 40 tests the operation of the screw fastening robot 10 (S105), and determines whether or not the “rotation amount characteristic test 2” is completed (S106). If the parameter automatic adjustment device 40 determines in S103 that “rotation amount characteristic test 1” has not been completed (S103: No), it returns to S101 and repeatedly executes “rotation amount characteristic test 1”.

In “rotation amount characteristic test 2”, the parameter automatic adjustment device 40 sets the rotation amount in the second period to be larger than that of “rotation amount characteristic test 1”. That is, after setting the initial value of the rotation amount in the second period to “1” in “Rotation amount characteristic test 1” and testing the operation, in “Rotation amount characteristic test 2”, “Rotation amount characteristic test 1” The initial value of the rotation amount in the second period is set to be larger than “Rotation amount characteristic test 1” using the acquired parameters. The rotation amount of the second period set in the “rotation amount characteristic test 2” is calculated based on the following concept. Here, when C ^r23 = −1.2 is set as the conversion coefficient C ^r23 from the rotation amount in the second period to the stable rotation amount in the third period as described above, the rotation amount in the second period and the third period The relationship with the stable rotation amount at can be calculated as follows.

From the above equation, an intermediate value between the “average value of the third period stable rotation amount” and C ^rm3 can be considered as the rotation amount in the second period at which it can be safely tested. Therefore, the amount of rotation in the second phase is

And set.
Here, the third rotation amount margin C ^rm3 will be verified.
When adjusting the amount of rotation in the second phase, it is necessary to satisfy the following three requirements.
Condition 1: The screw 34 should not be seated on the workpiece 32 in the second period.
As described above, the rotation speed in the second period is high. Therefore, if the screw 34 is seated on the workpiece 32 in the second period, the screw fastening robot 10 is damaged without being limited to the workpiece 32 and the screw 34. On the other hand, in order to calculate the limit value at which the screw 34 is seated on the workpiece 32, it is necessary to take into account the variation in the stable rotation amount in the third period, that is, the "third period stable rotation amount: R _s3 ". When the screw 34 is screwed into the workpiece 32, an error of at least one rotation occurs due to the meshing in the first stage. Therefore, 0.5 rotation is the minimum value of error in either the workpiece 32 or the screw 34. Further, variation in characteristics of the screw tightening robot 10 is added to the rotation error. In the case of the present embodiment, an error due to variation in characteristics of the screw tightening robot 10 is set to 0.7 rotation as an experimental value.

・ Condition 2: Ensure stable rotation in the third period.
If the moment of rotation in the second period remains and the third period is passed to the fourth period, the variation in tightening torque increases. Therefore, it is desirable to secure a certain amount or more of the stable rotation amount in the third period, that is, the “third period stable rotation amount”. If the stable rotation amount in the third period is experimentally ensured by 0.5 rotation, the variation in the tightening torque is reduced. Therefore, in the present embodiment, 0.5 rotation is secured as the stable rotation amount in the third period.
Condition 3: Increase the amount of rotation in the second phase while satisfying Condition 1 and Condition 2.

It is desirable to shorten the time required for screw tightening through the first period to the fourth period in the screw tightening robot 10 as much as possible. Therefore, the larger the amount of rotation in the second period, the better.
From the above condition 1 to condition 3, in the present embodiment, the “third period stable rotation amount margin” which is the margin of the stable rotation amount in the third period is
C ^rm3 = 0.5 + 0.7 + 0.5 = 1.7
Set to.

Next, the conversion coefficient ^Cr23 from the rotation amount in the second period to the stable rotation amount in the third period will be verified.
Qualitatively, when the rotation amount R ₂ in the second period increases, the stable rotation amount R _s3 in the third period decreases by the increase in the rotation amount R ₂ . That is, it is reasonable that C ^r23 = −1. However, in practice, for example, C ^r23 = −1 does not ^occur due to slippage between the screw 34 and the workpiece 32, hysteresis of the screw fastening robot 10, or the like. That is, when data is collected experimentally, the conversion coefficient C ^r23 becomes a value deviated from “−1” as shown in FIG. Here, in order to verify the conversion coefficient C ^r23 , the relationship between the rotation amount R ₂ in the second period and the stable rotation amount R _s3 in the third period is obtained and shown in FIG. As shown in FIG. 12, when these relationships are robust linear regression and linear regression, the slope corresponding to the conversion coefficient C ^r23 becomes C ^r23 = −1.18≈−1.2. Yes. Therefore, in the present embodiment, considering this tendency, C ^r23 = −1.2 is set.
For the above reasons, the test parameter generation unit 61 sets the test parameters in the “rotation amount test 2” as follows.

  The parameter automatic adjustment device 40 executes “rotation amount characteristic test 2” three times in S104 and S105 in the same manner as “rotation amount characteristic test 1”. When the parameter automatic adjustment device 40 determines that the “rotation amount characteristic test 2” has been completed in S106 (S106: Yes), the process proceeds to generation of a test parameter for the current-speed characteristic (S107). Then, the parameter automatic adjustment device 40 tests the operation of the screw fastening robot 10 (S108), and determines whether or not the generation of the test parameter for the current-speed characteristic is completed (S109). If the parameter automatic adjustment device 40 determines in S106 that “rotation amount characteristic test 2” has not been completed (S106: No), it returns to S104 and repeatedly executes “rotation amount characteristic test 2”. If the automatic parameter adjustment device 40 determines in S109 that the generation of the current-speed characteristic test parameter has been completed (S109: Yes), it completes the initial characteristic investigation process. On the other hand, if the parameter automatic adjustment device 40 determines in S109 that the generation of the test parameter of the current-speed characteristic is not completed (S109: No), the process returns to S107 and repeats the test parameter generation process.

  The current-speed characteristic test (hereinafter referred to as “current-speed test”) will be described in detail. In the current-speed test, the parameter automatic adjustment device 40 investigates the relationship between the current and the rotational speed and the tightening torque while slightly changing in the vicinity of the default parameters set in advance. The purpose of this investigation is to prevent the regression analysis from being degraded in the subsequent parameter adjustment process. According to this purpose, since there are three parameters for regression analysis, it is sufficient to acquire the relationship at least at three points. In the present embodiment, the relationship is acquired at four points in order to improve so-called robustness.

As the test parameters in the current-speed test, the rotation amount R ₂ in the second period and the fourth period current value-third period speed shown below are used.

The rotation amount in the second period is expressed by the following equation 1 from the results of “rotation amount characteristic test 1” in S101 and “rotation amount characteristic test 2” in S104 and the results of the current-speed test in S107. The model R _s3 of the third period stable rotation amount shown is learned, and R ₂ is determined so that R _s3 = C ^rm3 .

Specifically, the rotation amount in the second phase, based on a result of the "rotation amount characteristic test 1" and "rotation amount characteristic test 2", seeking a ^I, b ^I from Formula 1. That is, the coefficient set to a fixed value in the “rotation amount characteristic test 1” and “rotation amount characteristic test 2” is obtained by robust linear regression as shown in FIG. Then, R ₂ is set so that the third stable rotation amount R _s3 and C ^rm3 are equal. Specifically, the following formula 2 is assumed.

  The fourth period current value-the third period speed is tested while being slightly changed in the vicinity of a default parameter set in advance. However, the results of testing using the default parameters in the “rotation amount characteristic test 1” and “rotation amount characteristic test 2” performed previously are obtained. Therefore, it is classified into the following three cases with reference to the test results already obtained in “Rotation amount characteristic test 1” and “Rotation amount characteristic test 2”.

● Case 1: When all the tightening torques in “Rotation amount characteristic test 1” and “Rotation amount characteristic test 2” are smaller than the target tightening torque ● Case 2: “Rotation amount characteristic test 1” and “Rotation amount characteristic test 2” When all of the tightening torques in "" are larger than the target tightening torque ● Case 3: When neither Case 1 nor Case 2 applies

  In case 1, it is clear that the current value in the fourth period has insufficient tightening torque in the default parameters. Therefore, the parameter automatic adjustment device 40 executes the current-speed test in a range obtained by multiplying the default parameter by 1.0 to 1.2.

Further, in case 2, it is clear that the current value in the fourth period has an excessive tightening torque in the default parameters. Therefore, the parameter automatic adjustment device 40 executes the current-speed test in a range in which the default parameter is multiplied by 0.85 to 1.0.
Further, in case 3, the current value in the fourth period includes a case where the tightening torque is insufficient and a case where it is excessive in the default parameters. Therefore, the parameter automatic adjustment device 40 executes the current-speed test in a range in which default parameters are multiplied by 0.85 to 1.15.

On the other hand, the automatic parameter adjustment device 40 slightly changes the rotation speed around the rotation speed in the third period centering on ω ₃ = ω _3c1 which is considered to be most reliable. Here, when the accuracy priority mode is selected, three types of candidates are set as the rotation speed ω ₃ in the third period. In this case, the parameter automatic adjustment device 40 performs a current-speed test using all candidates. Thereby, in case 1, the parameter automatic adjustment device 40 executes the test four times, that is, the operation test of the screw tightening robot 10 in S108 using the above-described parameter conditions as shown in FIG. Similarly, in the case 2, the automatic parameter adjustment device 40 executes four tests using the above-described parameter conditions as shown in FIG. 15, and in the case 3, the above-described parameter as shown in FIG. 16. Perform four tests using conditions.
If the parameter automatic adjustment device 40 determines in S109 that the operation test of the screw tightening robot 10 in S108 has been executed four times, it determines that the current-speed characteristic test has been completed.

<Parameter adjustment processing>
When the initial characteristic investigation process described above is completed, the parameter automatic adjustment device 40 proceeds to the parameter adjustment process. The parameter adjustment process will be described based on FIG. When the automatic parameter adjustment device 40 completes the initial characteristic investigation process shown in FIG. 10, the parameter automatic adjustment apparatus 40 executes the parameter adjustment process using the various parameters acquired in the initial characteristic investigation process and the already input initial setting data.

  When the process proceeds to the parameter adjustment process, the test parameter generation unit 61 generates a test parameter in the second period (S201). Then, the test parameter generation unit 61 generates the test parameter of the rotation speed in the third period and the current value in the fourth period (S202), and then the current value in the third period and the screw 34 in the third period become the workpiece 32. Test parameters for the seating current value when seated and the determination current value in the fourth period are generated (S203). When the test parameters are generated in S201 to S203, the parameter automatic adjustment device 40 transfers the generated parameters to the screw tightening robot 10 (S204), and executes an operation test in the screw tightening robot 10 (S205). Then, the parameter automatic adjustment device 40 acquires the measured values of the current waveform and the tightening torque from the screw tightening robot 10 (S206). In the case of the present embodiment, the parameter automatic adjustment device 40 repeatedly performs the operation test process in the screw tightening robot 10 from S204 to S206 five times. Accordingly, when the automatic parameter adjustment device 40 acquires the current waveform and the tightening torque from the screw tightening robot 10, it determines whether or not the number of executions of the operation test has reached “5 times” (S207).

  If the parameter automatic adjustment device 40 determines in S207 that the operation test of the screw tightening robot 10 has not been repeatedly executed "5 times" (S207: No), the process returns to S204 and repeats the processing from S204 to S206. On the other hand, when the parameter automatic adjustment device 40 determines in S207 that the operation test of the screw tightening robot 10 has been repeatedly executed "5 times" (S207: Yes), it sets the optimum parameter and informs the convergence state of the optimum parameter. (S208). Then, the parameter automatic adjustment device 40 determines whether or not there is an instruction to interrupt the process in the middle of the process (S209). If there is no instruction to interrupt the process (S209: No), whether the parameter adjustment process is completed. It is determined whether or not (S210). The parameter automatic adjustment device 40 ends the parameter adjustment process when there is an instruction to interrupt the process at S209 (S209: Yes) and when it is determined that the parameter adjustment process is completed at S210 (S210: Yes). On the other hand, if the parameter automatic adjustment device determines that the parameter adjustment process is not completed in S210 (S210: No), the process returns to S201 and repeats the process.

Details of the parameter adjustment process will be described below.
In the parameter adjustment process shown in FIG. 17, the rotation amount in the second period is set in common in all adjustment modes.
The test parameter generation unit 61 calculates the test parameter for the rotation amount in the second period in S201 of the parameter adjustment process according to the following procedure.
First, the automatic parameter adjustment device 40 learns the rotation amount in the third period using Equation 3 based on all past test results including the initial characteristic investigation process, and calculates the coefficients a ^s and b ^s . Then, the automatic parameter adjustment device 40 sets the rotation amount in the second period based on Expression 4 so that the stable rotation amount in the third period, that is, the third period stable rotation amount R _s3 becomes equal to C ^rm3 . That is, as shown in FIG. 18, a correlation is obtained between the rotation amount R ₂ in the second period and the stable rotation amount R _s3 in the third period.

  The test parameter generation unit 61 calculates the test parameters of the third-period rotation speed and the current value in the fourth period in S202 of the parameter adjustment process according to the following procedure. Here, the test parameters for the rotation speed of the third period and the current value of the fourth period generated in S202 are “accuracy priority mode-normal”, “accuracy priority mode-simple”, “time priority mode”, and “fixed speed”. -"Simple mode" can be arbitrarily selected. Hereinafter, these will be described for each mode.

● When “Accuracy Priority Mode – Normal” When calculating the tightening torque τ _f for the rotation speed ω _{3 in} the third period and the current value τ _{4 in} the fourth period, the following model is established between them. . This model is created from the viewpoint of minimizing the square error for experimentally collected data.

However, when there are many coefficients a ₁ to a ₇ as in this model, it is necessary to repeat the operation test of the screw tightening robot 10 using the test piece many times. Therefore, in order to reduce the coefficient, the tightening torque estimation model generation unit 62 generates the following equation as a model 1.

Here, the reason that the term “log” is included in the above equation is that when the rotational speed ω ₃ in the third period is small, it is approximated nonlinearly. Therefore, when the rotational speed ω ₃ in the third period is relatively large and the influence can be eliminated, the tightening torque estimation model generation unit 62 may generate the following equation as the model 2.

  Here, the tightening torque learning unit 63 learns the coefficients u, w, and v from the data acquired by the operation test for assembling the test piece executed in the screw tightening robot 10 using any of the above formulas. The coefficients u, w, v represent the characteristics of the screw fastening robot 10, the screw 34, and the workpiece 32.

  When the operation test of the screw fastening robot 10 is executed, a test piece composed of the workpiece 32 and the screw 34 is consumed for each operation test. Therefore, it is desirable that the parameter adjustment process is completed with as few operation tests as possible. That is, it is necessary to select test parameters so as to efficiently improve the estimation accuracy of the coefficients u, w, and v. Therefore, in the present embodiment, the following concept is introduced.

(1) Improve accuracy only in the vicinity of the optimum parameter Obtaining the optimum parameter means obtaining τ ₄ and ω _{3 such} that τ _f = τ _g . Therefore, in the above model, it is sufficient that the accuracy is high only in the vicinity of τ _g . Therefore, it is desirable to increase the number of operation tests in a region where τ _f = τ _g is highly likely to hold. In particular, the above model is an approximate expression in which the coefficients are limited to three, u, v, and w. Therefore, when obtaining a wide range of characteristics, the accuracy is significantly reduced in a region far from τ _g .

(2) Introducing a general optimal design in linear regression The D optimal design is used as a general method for increasing the accuracy of coefficients in linear regression.
In the case of the present embodiment, the D optimum plan in (2) is applied to the region where τ _f = τ _g described in (1) is highly likely to be satisfied, that is, the range limited by the confidence interval, and the test level To decide. Specifically, the parameter estimation unit 58 calculates a test parameter by the following procedure.
First, the parameter estimation unit 58 calculates confidence intervals for the coefficients u, w, and v. By using the upper limit value and the lower limit value of the confidence interval, the parameter range that is highly likely to hold τ _f = τ _g is limited. This confidence interval is successively narrowed every time the operation test is repeatedly executed in the screw tightening robot 10. The specific procedure is as follows.

Parameter estimation unit 58, the current value tau ₄ in the rotation speed [omega] 3, and the fourth stage in the third period acquired in initial characteristics investigation process shown in FIG. 10, the coefficient using the model 1 of the u, v, and w Ask. Note that model 2 may be used instead of model 1.
Then, in order to increase the accuracy in the vicinity of τ _g , the range of the test parameter in the confidence interval is limited using a prediction distribution by linear regression. Here, the number of operations n of the screw tightening robot 10 is n = 1, 2,...

And It should be noted that “n” shown on the right shoulder of the character in the above formula is not a “power number” but for identifying the above-mentioned number of operations “n”. The coefficient w (vector) and the next test parameter x are respectively

And
At this time, the learned w can be learned as follows. “∧” indicates that learning has been completed.

  At this time, the posterior prediction distribution is calculated from the following equation based on a statistical method. Here, “p” corresponds to the number of coefficients, and in the case of model 1 and model 2, “p = 3”.

  According to the above formula, the confidence interval of the predicted response of 100 (1-α)% is shown as follows.

t is a t distribution and the degree of freedom is np-1. Among these, “n” is the number of data, and “p” is the number of coefficients. Also, setting α = 0.05 means that it is included in the confidence interval set with a probability of 95%. In the present embodiment, the condition of the next test parameter is that the target tightening torque τ _g is included in the range of the above-described confidence interval. Specifically, the next test parameter x is limited to satisfy the following expression.

  In the confidence interval set as described above, the D optimal plan is calculated. In particular,

It is desirable to obtain a sample that reduces the size of the sample. Therefore,

Was the next test parameter. However, the rotational speed ω ₃ in the third period is approximated to the “third period speed candidate 1”, “third period speed candidate 2”, and “third period speed candidate 3” using the nearest neighbor method.
Regarding the setting of the above confidence interval, “Pattern recognition and machine learning-Statistical prediction by Bayesian theory: by CM Bishop” is referred to. For D-optimal design, refer to “Experimental Design Methodology-From Basic Method to Response Surface Method Taguchi Method Optimal Design: Hide Yamada”.

By the way, depending on the characteristics of the screw fastening robot 10, the influence of the rotational speed ω ₃ in the third period may be small. As described above, when the influence of the rotational speed ω ₃ in the third period is small, the restriction due to the confidence interval expands in the ω ₃ axis direction, and an appropriate result may not be obtained. In this case, the rotational speed ω ₃ in the third period may be excluded from the model 1 and the model 2, and τ _f = uτ ₄ + v. Then, as described below, the calculation of the current value τ _{4 in} the fourth period may be simplified by reducing the variables with p = 2.

In this case, the rotational speed ω ₃ in the third period is determined so that the number of operation tests is equal at the following three levels.

  In the above, the operation test in the screw tightening robot 10 is executed by setting the same parameter five times as one set. Thereby, the standard deviation σ is calculated for each parameter.

Next, conditions for completing the generation of the test parameters for the rotation speed ω ₃ in the third period and the current value τ ₄ in the fourth period will be described.
The optimum parameter is calculated by the following formula.

  By applying this to the following equation, the confidence interval is calculated.

  Thereby, if the following formula is established, it can be determined that the parameter generation is completed.

Note that the confidence interval is roughly inversely proportional to the square root of the number of tests. Therefore, more simply, the parameter generation may be completed when the number of tests reaches a preset number of tests. Generally, it is desirable to set so that there are about 30 operation tests in the screw tightening robot 10.
When the parameter generation is completed, the optimum parameter is set in S208.

The parameter estimation unit 58 performs an operation test of the screw tightening robot 10 by setting 5 times as a set in S204 to S206, and calculates a standard deviation σ _f for each set. Based on this result, σ _f = rω ₃ + s is learned, and coefficients r and s are obtained.
The parameter estimation unit 58 selects the speed candidate with the fastest speed among the three kinds of speed candidates in the third period, satisfying σ _f = rω ₃ + s <σ _g . If any speed candidate does not satisfy this condition, the parameter estimation unit 58 selects the slowest speed candidate in consideration of safety. In this case, the parameter automatic adjustment device 40 notifies the output unit 43 such as a display that the target accuracy cannot be obtained for all speed candidates.

In the case of “accuracy priority mode-simple” In the above-mentioned “accuracy priority mode-normal”, complicated calculation is performed, so that the time required for executing automatic parameter adjustment becomes longer. Therefore, simple automatic parameter adjustment can be performed with priority given to accuracy by the following procedure. That is, the confidence interval is considered to be inversely proportional to the square root of the number of operation tests as described above. Therefore, when performing an operation test several tens of times, the confidence interval does not change significantly except for the initial stage. Therefore, simple calculation is possible by setting the confidence interval to the following fixed value. Also, for the D optimal plan, in many cases, the ends of the confidence intervals generally tend to be selected. Therefore, simple calculation is possible by sequentially selecting the end points of the confidence interval.

The constant “Cτ” shown above is preferably a value that does not deviate significantly from the calculated value of the confidence interval. The constant “Cτ” is calculated by the constant determination unit 57. That is, the constant determination unit 57 sets a constant “Cτ” based on the following equation, including a predetermined error range in the target tightening torque τ _g input from the target torque input unit 55.

As described above, even when the calculation is simply performed, the coefficients u and v are calculated by learning τ _f = uτ ₄ + v using the current value τ ₄ obtained in the fourth period. Here, at least four levels are selected in order from level 1 so that the number of operation tests is equal to the above nine levels. Then, for the same parameter, an operation test is executed by setting 5 times as one set in S204 to S206. The parameter generation completion conditions and optimum parameter settings are the same as in the case of “accuracy priority mode-normal” described above.

● In “time priority mode” When calculating in the time priority mode, t _f = g / ω ₃ + h is learned using the already obtained rotation speed ω ₃ in the third period, and the coefficients g and h are calculated. To do. If the target tightening time is t _g , the rotational speed ω ₃ in the third period is shown below.

In this case, the current value τ ₄ in the fourth period can be set from any level 1 to level 3 as current value candidates. Further, in the “time priority mode”, since the standard deviation σ need not be calculated, an operation test is executed once for each identical parameter. A relationship as shown in FIG. 19 is established between the rotation speed in the third period and the tightening time in the third period.
The parameter generation completion conditions are the same as in the case of the “accuracy priority mode—normal” described above.
On the other hand, the optimum parameter is set as follows.

● In the case of “Fixed speed mode-simple” In “Fixed speed mode-simple”, parameters are generated by further simplifying “Accuracy priority mode-simple”. Also in this case, τ _f = uτ ₄ + v is learned using the current value τ ₄ already acquired in the fourth period, and the coefficients u and v are calculated. Since the calculation of the standard deviation σ is not required, one operation test is executed in S204 to S206 with the same parameters for the following three levels. Since the rotation speed in the third period is fixed, ω ₃ = ω _3c1 . The parameter generation completion conditions are the same as in the case of the “accuracy priority mode-simple” described above.
On the other hand, the optimum parameter is set as follows.

The automatic parameter adjustment device 40 calculates the test parameters of the current value in the third period, the seating current value in the third period, and the determination current value in the fourth period in S203 of the parameter adjustment process according to the following procedure.
The fourth period current value is determined by the characteristics of the motor 21. Specifically, when the number of operations of the screw tightening robot 10 is small, such as at the beginning of the automatic adjustment process, the parameter estimation unit 58 calculates the fourth period current value by linear transformation. This value is defined as the default parameter. This value indicates the relationship between the target tightening torque τ _g and the current of the motor 21 as described below, and is determined by the characteristics of the motor 21 as described above.

The current value τ ₃ in the third period, the sitting current value τ _h3 in the third period, and the determination current value τ _h4 in the fourth period are calculated by linear transformation based on the current values in the fourth period. In the present embodiment, the coefficient C for calculating these values is obtained experimentally and set as shown in FIG. Here, the seating current value τ _h3 in the third period is a current value when the head of the screw 34 tightened in the third period contacts the end face of the workpiece 32. The determination current value τ _h4 in the fourth period is a current value when it is determined that screwing of the screw 34 is stopped in the fourth period.

  Various parameters are generated in S201 to S203 by the procedure as described above, and when the operation test of the screw tightening robot 10 is executed in S204 to S206 based on the generated parameters, the parameter automatic adjustment device 40 determines in S207. When the obtained optimum parameter is present, it is displayed on the output unit 43 such as a display, and when the convergence state is inappropriate, that fact is displayed on the output unit 43 such as a display. Thereby, the parameter adjustment process ends.

By the way, the optimum parameter obtained in accordance with the above-described procedure gradually converges to a specific value every time the number of times that the test piece is assembled increases. Therefore, each time the optimum parameter is updated by assembling the test piece, the parameter automatic adjustment device 40 outputs the updated optimum parameter to the output unit 43 such as a display as shown in FIGS. In the graphs shown in FIG. 21 and FIG. 22, the horizontal axis represents the number of times the test piece is assembled, and the vertical axis represents the optimum parameter for each number of times of assembly. 21 and 22, the lower diagram is an enlarged scale of the vertical axis of the upper diagram. Thereby, the user can recognize the transition of the optimum parameter that accompanies the repetition of the operation test of the screw fastening robot 10 by assembling the test piece. As a result, particularly in the case of an expert, by confirming the display of the output unit 43, it is possible to determine whether to stop or continue the parameter adjustment.
In the above description, the range of the test parameter is not particularly described, but it is desirable to set the upper limit value and the lower limit value of the test parameter in advance in order to ensure safety.

In the embodiment described above, the parameter estimation unit 58 estimates a parameter using a simplified model. In other words, the parameter estimation unit 58 uses the simplified model to determine the third rotation speed ω ₃ and the fourth rotation current τ _{4 in which} the actual measurement value τ _f of the tightening torque approximates the target tightening torque τ _g. Is estimated. At this time, since there are three types of coefficients to be obtained, the number of operation tests necessary for converging these values, that is, the number of test pieces to be consumed is about several tens, which is close to a skilled person. Therefore, it is possible to automatically set a highly accurate parameter with a small number of operation tests regardless of the skill level.

In one embodiment, the parameter estimation unit 58 estimates the screw tightening parameters in the third and fourth periods, that is, the rotation speed ω ₃ and the current τ _{4 at} which τ _f = τ _g using a simplified model. is doing. At this time, the actual measured value τ _f of the tightening torque used in the simplified model is required only when setting these parameters. That is, by setting parameters for the screw tightening robot 10, the screw tightening robot 10 adjusts the tightening torque in the third period without using the actual measured value of the tightening torque by the torque sensor 27 when operating in the line. Therefore, when the screw tightening robot 10 is operated in the line after setting the parameters, it is not necessary to install expensive torque sensors 27 in all the screw tightening robots 10 provided in the line. In particular, when the parameters of the plurality of screw tightening robots 10 are automatically adjusted by making the torque sensor 27 detachable from the screw tightening robot 10 as in the embodiment, only one torque sensor 27 is required. Therefore, the parameters of the plurality of screw tightening robots 10 can be automatically adjusted without preparing many expensive torque sensors 27.

  Furthermore, in one embodiment, when “accuracy priority mode—simple”, the preset speed candidates are three fixed values. The parameter estimation unit 58 executes an operation test of the screw tightening robot 10 by selecting four levels from nine levels set corresponding to the three speed candidates. Therefore, the number of operation tests necessary to converge the parameters, that is, the number of consumed test pieces is about several tens, which is close to a skilled person. For example, when the operation test is executed five times for four levels as in the embodiment, the parameters converge in about 20 operation tests in total. Therefore, it is possible to automatically set a highly accurate parameter with a small number of operation tests regardless of the skill level.

  Further, in one embodiment, the second period initial rotation amount setting unit 56 tests the operation of the screw tightening robot 10 with the rotation amount in the second period minimized when acquiring the initial characteristics of the screw tightening robot 10. . That is, the second period initial rotation amount setting unit 56 sets the initial value of the rotation amount in the second period to “one time”. Thereby, even when the screwing amount of the screw 34 with respect to the workpiece 32 is unknown, the rotation amount in the second period in the initial characteristic investigation process does not become excessive. Therefore, it is possible to avoid damage to the test piece and the screw tightening robot 10 due to excessive movement when acquiring the initial characteristics.

  Furthermore, in one embodiment, the parameter automatic adjustment device 40 performs assembly of the test piece while gradually increasing the rotation amount after “1 time” is set as the initial value of the rotation amount in the second period. Thereby, the parameter in the second period is adjusted without depending on the parameter adjustment depending on the experience and intuition of the expert. In addition, since the parameters are set using data obtained by assembling each test piece, the number of operation tests, that is, the number of test pieces to be consumed is reduced. Therefore, it is possible to automatically set a highly accurate parameter with a small number of operation tests regardless of the skill level.

(Other embodiments)
In the embodiment described above, the example in which the torque sensor 27 that detects the tightening torque of the screw tightening robot 10 is detachably attached to the head unit 11 has been described. However, the torque sensor 27 may be configured to be attached to the workpiece 32 and to detect the tightening torque of the screw 34 attached to the screw hole 33 of the workpiece 32. Moreover, although the example which selects four levels from nine levels was demonstrated in "accuracy priority mode-simple", you may select four or more levels. Further, at this time, “5 times” is described as an example of the number of operation tests. However, the number of operation tests is not limited to “5” as long as the convergence of each parameter is ensured. Furthermore, the initial value of the rotation amount in the second period set by the second period initial rotation amount setting unit 56 is not limited to “one time”. That is, the initial value of the rotation amount in the second period can be set to “twice” or more as long as safety is ensured.
The present invention described above is not limited to the above-described embodiment, and can be applied to various embodiments without departing from the gist thereof.

  In the drawing, 10 is a screw tightening robot, 15 is a rotation drive mechanism (drive unit), 16 is a bit, 27 is a torque sensor (measured tightening torque acquisition means), 33 is a screw hole, 34 is a screw, and 40 is a parameter automatic adjustment. The apparatus, 51 is a third period rotational speed acquisition unit (third period rotational speed acquisition means), 52 is a fourth period supply current acquisition part (fourth period supply current acquisition means), 53 is a torque acquisition part (measured tightening torque acquisition) Means), 54 is a rotation speed input section (rotation speed input means), 55 is a target torque input section (target torque input means), 56 is a second period initial rotation amount setting section (second period initial rotation amount setting means), 57 is a constant determination unit (constant determination unit), 58 is a parameter estimation unit, 61 is a test parameter generation unit (test parameter generation unit), 62 is a tightening torque estimation model generation unit (tightening torque estimation model generation unit) 63 shows tightening torque learning part (tightening torque learning unit).

Claims (9)

A bit in contact with the screw, and a drive unit that rotationally drives the bit with electric power,
A first phase in which a screw is engaged with a screw hole, a second phase in which the screw meshed in the first phase is screwed into the screw hole at a higher speed than in the first phase, and the screw screwed in in the second phase. A third period in which the torque applied to the screw is tightened at a lower speed than the second period and adjusted to a preset tightening torque, and the screwing of the screw is stopped and the tightening torque is adjusted in the third period. In the screw tightening robot for performing screw tightening including the fourth period for stabilizing the screw,
A parameter automatic adjustment device for a screw tightening robot that automatically adjusts control parameters of the screw tightening robot by assembling a plurality of test pieces including the screw and an object in which the screw hole is formed,
Test parameter generation means for generating a test parameter related to the tightening torque of the screw each time the test piece is assembled;
Storage means for storing a history by associating the test parameter generated by the test parameter generation means and the actual torque value τ _f acquired by the actual tightening torque acquisition means each time the test piece is assembled;
A tightening torque estimation model generating means for generating a tightening torque estimation model in which the tightening torque is estimated based on a screw tightening parameter of the screw tightening robot;
The tightening torque for learning the tightening torque estimation model generated by the tightening torque estimation model generating means and generating the post-learning tightening torque estimation model using the test parameter and the history of the actually measured torque value τ _f stored in the storage unit Learning means,
Parameter estimation means for estimating an optimum value of a screw tightening parameter in the screw tightening robot using the post-learning tightening torque estimation model learned by the tightening torque learning means;
A parameter automatic adjustment device comprising: The screw tightening parameter of the screw tightening robot is the rotation speed ω ₃ acquired by the third period rotation speed acquisition unit and the current τ ₄ acquired by the fourth period supply current acquisition unit. Automatic parameter adjustment device. When the tightening torque estimation model has a coefficient u, a coefficient w, and a coefficient v,
τ _f = uτ ₄ + w log (ω ₃ ) + v
The parameter automatic adjustment device according to claim 2, wherein When the tightening torque estimation model has a coefficient u, a coefficient w, and a coefficient v,
τ _f = uτ ₄ + wω ₃ + v
The parameter automatic adjustment device according to claim 2, wherein   The test parameter generation means generates a test parameter which is a solution of the D optimal plan using a tightening torque estimation model learned based on a history already accumulated in the storage means when the test piece is assembled. The parameter automatic adjustment device according to any one of claims 1 to 4.   When assembling the test piece, the test parameter generation means includes a test parameter generation range including an optimal test parameter using a tightening torque estimation model learned based on a history already accumulated in the storage means. The parameter automatic adjustment device according to any one of claims 1 to 4, wherein the parameter is limited to a confidence interval.   When assembling the test piece, the test parameter generation means includes a test parameter generation range including an optimal test parameter using a tightening torque estimation model learned based on a history already accumulated in the storage means. The parameter automatic adjustment device according to claim 1, wherein the parameter automatic adjustment device generates a test parameter that is limited to the confidence interval and is a solution of the D optimal plan. A bit in contact with the screw, and a drive unit that rotationally drives the bit with electric power,
A first phase in which a screw is engaged with a screw hole, a second phase in which the screw meshed in the first phase is screwed into the screw hole at a higher speed than in the first phase, and the screw screwed in in the second phase. A third period in which the torque applied to the screw is tightened at a lower speed than the second period and adjusted to a preset tightening torque, and the screwing of the screw is stopped and the tightening torque is adjusted in the third period. In the screw tightening robot for performing screw tightening including the fourth period for stabilizing the screw,
A parameter automatic adjustment device for a screw tightening robot that automatically adjusts control parameters of the screw tightening robot by assembling a plurality of test pieces including the screw and an object in which the screw hole is formed,
A third period rotational speed acquisition means for acquiring a rotational speed ω ₃ for tightening the screw in the third period each time the test piece is assembled;
Each time the test piece is assembled, the fourth period supply current acquisition means for acquiring the current τ ₄ supplied to the drive unit in the fourth period,
Measured tightening torque acquisition means that is detachably provided on the screw tightening robot and detects the measured value τ _f of the tightening torque in the third period when the test piece is assembled.
Using the rotation speed ω ₃ acquired by the third period rotation speed acquisition means, the current τ ₄ acquired by the fourth period supply current acquisition means, and the measured value τ _f acquired by the torque acquisition means, When the coefficient u, the coefficient w, and the coefficient v are set,
τ _f = uτ ₄ + w log (ω ₃ ) + v
Parameter estimating means for estimating the coefficient u, the coefficient w and the coefficient v, and estimating screw tightening parameters in the third period and the fourth period;
A parameter automatic adjustment device comprising: A bit in contact with the screw, and a drive unit that rotationally drives the bit with electric power,
A first phase in which a screw is engaged with a screw hole, a second phase in which the screw meshed in the first phase is screwed into the screw hole at a higher speed than in the first phase, and the screw screwed in in the second phase. A third period in which the torque applied to the screw is tightened at a lower speed than the second period and adjusted to a preset tightening torque, and the screwing of the screw is stopped and the tightening torque is adjusted in the third period. In the screw tightening robot for performing screw tightening including the fourth period for stabilizing the screw,
A parameter automatic adjustment device for automatically adjusting control parameters of the screw tightening robot by assembling a plurality of test pieces including the screw and an object in which the screw hole is formed,
A third period rotational speed acquisition means for acquiring a rotational speed ω ₃ for tightening the screw in the third period each time the test piece is assembled;
Each time the test piece is assembled, the fourth period supply current acquisition means for acquiring the current τ ₄ supplied to the drive unit in the fourth period,
Measured tightening torque acquisition means that is detachably provided on the screw tightening robot and detects the measured value τ _f of the tightening torque in the third period when the test piece is assembled.
Using the rotation speed ω ₃ acquired by the third period rotation speed acquisition means, the current τ ₄ acquired by the fourth period supply current acquisition means, and the measured value τ _f acquired by the torque acquisition means, When the coefficient u, the coefficient w, and the coefficient v are set,
τ _f = uτ ₄ + wω ₃ + v
Parameter estimating means for estimating the coefficient u, the coefficient w and the coefficient v, and estimating screw tightening parameters in the third period and the fourth period;
A parameter automatic adjustment device comprising:

Images