JP6072665B2 - Working method and working device - Google Patents

Description

  The present invention relates to a work method and a work device for performing work on a work conveyed in a state of being in contact with a conveyance device.

  When manufacturing industrial products such as automobiles, various operations such as assembly and painting by a working robot are sequentially performed while a workpiece is conveyed using a conveyance device. Recently, in order to improve work efficiency, attempts have been made to work on a workpiece without stopping conveyance.

  For example, Patent Document 1 proposes a synchronous transfer device that keeps the relative positional relationship between the two by making the working robot move in synchronization with the transfer of the workpiece.

Japanese Patent No. 4202953

  By the way, when a work is transported using the transport device, vibration due to the operation of the transport device may occur. Since this vibration is transmitted to the workpiece via the contact portion, it can be a factor that varies the relative positional relationship between the workpiece and the work robot. However, in Patent Document 1 described above, no consideration is given to vibration that may occur in the workpiece, and there is room for further improvement.

  The present invention has been made in order to solve the above-described problems, and provides a working method and a working apparatus capable of performing work while maintaining high positioning accuracy regardless of the vibration of the work accompanying the transfer operation. For the purpose.

A working method according to the present invention is a method using a working device having a working unit that performs work on a workpiece conveyed in a state of contact with the conveying device, wherein the workpiece and / or the vibration of the conveying device is used. By sequentially measuring the state at regular time intervals, a measurement process for acquiring vibration data sequences for at least m (m is an integer of 2 or more) points, and k ( k is an integer equal to or greater than 1) a prediction step of predicting the vibration state at a previous measurement time point by a point, and the predicted vibration state is converted into a vibration displacement amount of the workpiece, and the previous measurement time point is A teaching step of teaching the working position to which the vibration displacement amount is added before the passage , the prediction step using an m-th order autoregressive model expressed by a first autoregressive coefficient; Vibration data for the nearest m points Using the first calculation step of calculating the first predicted value at the previous measurement time from the data sequence and the m-th order autoregressive model expressed by the second autoregressive coefficient, the most recent (m + j) ( j is an integer satisfying 1 ≦ j ≦ m), a second calculation step of calculating a second predicted value at the previous measurement time point from the vibration data sequence for the points, and the calculated first and the calculated A combining step of predicting the vibration state by combining a second predicted value, and in the second calculation step, the second prediction is performed from a difference data sequence between vibration data whose measurement time points are different by j points. Calculate the value.

  Thus, the prediction process for predicting the vibration state at the previous measurement time point from the measured vibration data string, and the predicted vibration state is converted into the vibration displacement amount of the workpiece, before the previous measurement time point elapses. In addition, a teaching process for teaching the work position to which the vibration displacement amount is added is provided, so that the teaching can be performed in consideration of the vibration displacement amount of the workpiece. The teaching position at the time of measurement) substantially coincides with the work position of the workpiece. Thereby, it is possible to perform an operation while maintaining high positioning accuracy regardless of the vibration of the work accompanying the conveying operation.

In addition, the behavior of the vibration that occurs when feeding transportable, one high temporal correlation, tends to microscopically aperiodic. Therefore, the prediction accuracy of the vibration state is improved by using the autoregressive model.

Further, the first predicted value sensitivity to time change of vibration data was obtained from relatively high prediction method, and, by combining the second prediction value, wherein the sensitivity is obtained from relatively low Prediction The tracking error with respect to a rapid time change of vibration data is reduced.

The working device according to the present invention is a device having a working unit that performs work on a workpiece conveyed in a state of contact with the conveying device, and the vibration state of the workpiece and / or the conveying device is determined for a certain period of time. By sequentially measuring at intervals, a measurement unit that acquires vibration data sequences for at least m points, and predicting the vibration state at a measurement point ahead of k points from the vibration data sequence measured by the measurement unit. The vibration prediction unit and the vibration state predicted by the vibration prediction unit are converted into the vibration displacement amount of the workpiece, and the work position to which the vibration displacement amount is added before the previous measurement time has elapsed is e Bei teaching section for teaching the working unit, the vibration prediction unit uses the m-th order autoregressive model is represented by a first autoregressive coefficients, from the vibration data string of the latest m point fraction, said destination At the time of measurement The first (m + j) (j is an integer satisfying 1 ≦ j ≦ m, using the first calculation process for calculating the predicted value of 1 and the m-th order autoregressive model expressed by the second autoregressive coefficient The second calculation process for calculating the second predicted value at the previous measurement time point from the vibration data string for the point) and the calculated first and second predicted values are combined. In the second calculation process, the second predicted value is calculated from a difference data string between vibration data whose measurement time points are different by j points.

According to the working method and the working device of the present invention, the vibration state at the previous measurement time point is predicted from the measured vibration data string, the vibration state is converted into the vibration displacement amount of the workpiece, and the previous measurement time point is Before the time elapses, the work position to which the vibration displacement amount is added is taught to the work unit, so that it is possible to teach taking into account the vibration displacement amount of the workpiece. The teaching position at the time of measurement) substantially coincides with the work position of the workpiece. Thereby, it is possible to perform an operation while maintaining high positioning accuracy regardless of the vibration of the work accompanying the conveying operation.
In addition, the behavior of vibration generated during conveyance has a high temporal correlation, but tends to be non-periodic microscopically. Therefore, the prediction accuracy of the vibration state is improved by using the autoregressive model. Further, by combining the first predicted value obtained from the prediction method having relatively high sensitivity to the temporal change of the vibration data, and the second predicted value obtained from the prediction method having relatively low sensitivity, The tracking error with respect to a rapid time change of vibration data is reduced.

It is a whole block diagram of the conveyance system which enforces the working method concerning the present invention. It is a schematic diagram of the conveyance system shown in FIG. It is a side view of the conveyance system concerning this embodiment. It is a top view of the conveyance system shown in FIG. It is a side view of the working device shown in FIG. It is a rear view of the working device shown in FIG. It is a flowchart with which it uses for operation | movement description of the working apparatus shown in FIG.1, FIG5 and FIG.6. It is a graph which shows the prediction result of a vibration data sequence. It is a chart figure which shows the flow of operation | movement of a working device. It is a graph which shows the characteristic of the tracking error in each prediction model.

  Hereinafter, the working method according to the present invention will be described in detail with reference to the accompanying drawings by giving preferred embodiments in relation to a working device for carrying out the working method.

[Overall configuration of transfer system 10]
FIG. 1 is an overall block diagram of a transport system 10 that implements a working method according to the present invention. FIG. 2 is a schematic diagram of the transport system 10 shown in FIG.

  As illustrated in FIG. 1, the transport system 10 includes a work device 20 that performs a predetermined work on the work W, and a transport device 30 that transports the work W that is a work target. In the example of FIG. 2, the car body of the automobile after the painting process is used as the workpiece W, and a hanger conveyor that conveys the workpiece W in a state of being hung from above is used as the conveying device 30.

  The work device 20 is, for example, a device that performs a work of removing a door temporarily attached for painting from the work W. Specifically, the work device 20 basically includes a moving unit 40, a measurement unit 50, a work unit 60, and a control unit 70.

  The moving unit 40 moves (including tracking of the workpiece W) along the conveyance direction of the workpiece W under the control of the control unit 70. In the example of FIG. 2, the moving unit 40 is a carriage 41, 42 that can move along the rails 43, 44 (see FIG. 4) in parallel with the conveyance direction (or tracking direction) of the workpiece W.

  The measurement unit 50 sequentially measures the vibration state of the workpiece W and / or the conveyance device 30 that is generated in a state where it is in contact with the conveyance device 30 that is operating at a constant time interval Δt. In the example of FIG. 2, the measurement unit 50 includes sensor devices 52 and 53 installed on the upper surface of the carriage 41 and sensor devices 54 and 55 installed on the upper surface of the carriage 42.

In addition, the measurement unit 50 cooperates with the control unit 70 to sequentially measure the vibration data x _m indicating the vibration state of the measurement position Pm on the workpiece W, and obtain a time series of the vibration data x _m. And sequentially output to the buffer memory 51 side. The buffer memory 51 is a storage device that stores the vibration data x _m from the measurement unit 50 in a first-in first-out (FIFO) format. Hereinafter, the time series of the vibration data x _m is referred to as a vibration data string {x (n)}.

  The working unit 60 performs a predetermined work on the workpiece W under the control of the control unit 70. In the example of FIG. 2, the working unit 60 is the robots 61 and 62 installed on the upper surface of the carriage 41 and the robots 63 and 64 installed on the upper surface of the carriage 42.

  The control unit 70 is a unit that comprehensively controls each component including the moving unit 40, the measurement unit 50, the buffer memory 51, and the working unit 60. Specifically, the control unit 70 includes a tracking control unit 71 that performs tracking control by the moving unit 40, a vibration prediction unit 72 that predicts the vibration state of the workpiece W from the vibration data string {x (n)}, A position teaching unit 73 that teaches the position to the working unit 60.

[Specific Configuration of Transport System 10]
Next, an embodiment of the transport system 10 that implements the present invention will be described in detail with reference to FIGS. 3 is a side view of the transport system 10, and FIG. 4 is a plan view of the transport system 10 shown in FIG. 5 is a side view of the working device 20 shown in FIG. 3, and FIG. 6 is a rear view of the working device 20 shown in FIG.

  As shown in FIGS. 3 and 4, the transport system 10 includes a working device 20 that performs a predetermined work on the work W, and a transport device 30 that transports the work W. In the figure, a point between point A and point B indicates a station that performs work on the workpiece W.

  The conveyance device 30 constitutes a part of a production line for a car body (work W) of an automobile, and here is a hanger conveyor. Specifically, the transport device 30 includes a support rail 31 provided along the transport path, and a hanger 32 that moves while being suspended from the support rail 31. The support rail 31 is provided with a chain (not shown). The hanger 32 is pulled by the chain being guided by the support rail 31 and moving.

  As shown in FIG. 4, the work device 20 includes carriages 41 and 42 and rails 43 and 44. The rails 43 and 44 are provided along the conveyance path of the workpiece W and regulate the moving direction of the carriages 41 and 42. Each of the carriages 41 and 42 includes a motor (not shown), and moves from the point A to the point B in accordance with a pulse signal supplied from the control unit 70. Specifically, the carriages 41 and 42 start tracking (synchronous movement) of the workpiece W from the point A and move to the point B. Thereafter, when arriving at the point B, the tracking of the work W is ended, the work W is moved to the point A, and the tracking of the work W reaching the next is started. For convenience of illustration, in FIGS. 3 and 4, the distance between the front and rear workpieces W and the distance between the points A and B are simplified as appropriate.

  As shown in FIGS. 5 and 6, on the upper surfaces of the carriages 41 and 42, sensor devices 52 to 55 as the measurement unit 50, robots 61 to 64 as the working unit 60, and a control unit 75 as the control unit 70. Is installed.

  The sensor devices 52 to 55 are installed at positions below the workpiece W on the upper surfaces of the carriages 41 and 42, and measure the vibration state at the measurement position Pm (for example, the formation position of the welding hole) on the bottom surface of the workpiece W. To do. For example, the sensor devices 52 to 55 are configured to be able to measure the three-dimensional position of the measurement position Pm from the distance to the bottom surface of the workpiece W and the horizontal position.

  The robots 61 to 64 are work robots that are installed on the upper surfaces of the carriages 41 and 42 and perform predetermined work on the workpieces W tracked by the carriages 41 and 42 from the side surfaces. The robots 61 to 64 are configured by, for example, multi-joint manipulators in which a plurality of joints rotate independently.

  The control unit 75 as the control unit 70 executes various controls such as movement of the carriages 41 and 42 and operations of the robots 61 to 64. The control unit 75 may be configured by one or more devices. For example, the number of devices corresponding to the carts 41 and 42 and the robots 61 to 64 may be provided, and the carts 41 and 42 and the robots 61 to 64 may be configured to be controllable by only one device.

  The tracking control unit 71 in the control unit 75 detects the position of the hanger 32 based on an output signal from an encoder (not shown) provided in the transport device 30 and uses this position as a clue to determine the relative position with respect to the workpiece W. The tracking and movement of the carriages 41 and 42 are controlled so as to keep the relationship substantially constant.

  The vibration prediction unit 72 in the control unit 75 predicts the behavior of the measurement position Pm based on the output signals (vibration data string {x (n)}) from the sensor devices 52 to 55. As described above, since the relative positional relationship between the workpiece W and the carriages 41 and 42 is kept substantially constant, the behavior of the measurement position Pm substantially matches the vibration state of the workpiece W.

  The position teaching unit 73 in the control unit 75 outputs information for teaching the work positions Pt1 and Pt2 (hereinafter referred to as teaching information) to the robots 61 to 64 side. The teaching information includes, for example, a work position Pt1 on the local coordinate system with reference to the robot 61, and a vibration displacement amount Δx described later.

[Operation of working device 20]
Then, operation | movement of the working apparatus 20 is demonstrated in detail, referring FIGS. 7-9. First, a specific example of the process for predicting the vibration state of the workpiece W will be described with reference to the flowchart of FIG. Here, one robot 61 will be described as an example, but the other robots 62 to 64 operate similarly.

In step S1, the measurement unit 50 acquires the current vibration data x _m by measuring the vibration state of the workpiece W and / or the robot 61 at a constant time interval Δt (for example, equivalent to Δt = 20 [ms]). To do. The measurement unit 50 acquires, for example, the uniaxial coordinates (position in the transport direction) of the measurement position Pm on the workpiece W.

  In step S <b> 2, the control unit 70 updates the vibration data sequence {x (n)} at (m + j + 1) points stored in the buffer memory 51 in a first-in first-out (FIFO) format. n is an arbitrary integer in the range [−j, m]. A larger value of n indicates newer data, and a smaller value of n indicates older data. In other words, x (m) is the latest data, while x (−j) is the oldest data.

  Further, j corresponds to the phase shift number of the vibration data string {x (n)}, and is an integer satisfying 1 ≦ j ≦ m (for example, j = 8). m corresponds to the order of the autoregressive model (hereinafter referred to as “AR order”), and is an integer of 2 or more (for example, m = 12).

In step S3, the vibration prediction unit 72 reads out the vibration data string {x (n)} updated in step S2, and then the first autoregressive coefficient (specifically, in the vibration data string {x (n)}). The first calculation process is executed by calculating the coefficient {a _i }). The first autoregressive model includes vibration data strings {x (n)} (n = 0, 1,..., M) for the most recent m points and coefficients {a _i } (i = 1, 2,. m) is given by the following equation (1).

As understood from the equation (1), the matrix expressing the state space of the model is an upper triangular matrix, and the contribution of the vibration data sequence {x (n)} in the range of n <0 is all 0. Keep this in mind. Here, by calculating the coefficient {a _i } using the Gauss-Jordan sweep-out method, the calculation process can be speeded up.

  In addition, the behavior of vibration generated when the workpiece W is conveyed has a high temporal correlation, but tends to be non-periodic microscopically. Therefore, the prediction accuracy of the vibration state is improved by using the autoregressive model described above.

In step S4, the vibration prediction unit 72 calculates vibration data x (m + k) ahead by k points using the coefficient {a _i } calculated in step S3, and determines it as the first predicted value x _OS . Specifically, the vibration prediction unit 72 repeats the prediction process of vibration data x (m + 1), x (m + 2),... K times one point at a time using the coefficient {a _i } in the first prediction model. Predict x (m + k). k is an integer of 1 or more, and can be arbitrarily set in consideration of the immediate responsiveness of the work device 20.

In step S5, the vibration prediction unit 72 reads out the vibration data string {x (n)} updated in step S2, and then the second autoregressive coefficient (specifically, the vibration data string {x (n)}). The “second calculation process” is executed by calculating the coefficient {b _i }). The second autoregressive model includes a vibration data string {x (n)} (n = −j, −j + 1,..., M) for the most recent (m + j) points and coefficients {b _i } (i = 1, 2,..., M) is given by the following equation (2).

  As understood from the equation (2), the present model has a difference data string {x (n) −x (n) between vibration data x (n) and x (n−j) whose measurement time points are different by j points. Note that this is an autoregressive model based on -j)}. Similarly to the equation (1), the matrix expressing the state space of this model is an upper triangular matrix, and the difference data string {x (n) −x (n−j)} in the range of n <0 is used. All contributions are zero.

In step S6, the vibration prediction unit 72 calculates vibration data x (m + k) ahead by k points using the coefficient {b _i } calculated in step S5, and determines the second predicted value x _US . Specifically, the vibration prediction unit 72 uses the coefficient {b _i } in the second prediction model, and uses the difference data x (m + 1) −x (m−j + 1), x (m + 2) −x (m−j + 2). ),... Is repeated k times one point to predict x (m + k) −x (m−j + k). Then, x (m + k) is obtained by adding known x (m−j + k) to the difference data.

In step S <b> 7, the vibration prediction unit 72 executes a “combination process” that combines the first predicted value x _OS calculated in step S <b _{> 4} and the second predicted value x _US calculated in step S <b> 6. Thus, a predicted value x _p ahead of k points is calculated. The predicted value x _p is calculated as an average of two values, for example, x _p = (x _OS + x _US ) / 2. Hereinafter, this prediction model may be referred to as an “AR prediction combination model”.

  FIG. 8 is a graph showing a prediction result of the vibration data string {x (n)}. The horizontal axis of the graph is time (unit: s), and the vertical axis of the graph is the work position (unit: mm) as the vibration data string {x (n)}.

The graph indicated by the solid line corresponds to the actual measurement value (x _m ), and the graph indicated by the broken line corresponds to the first predicted value x _OS based on the coefficient {a _i }. As understood from this figure, in the first prediction model, there is a timing delay for predicting / reflecting the inflection point (in other words, the maximum point or the minimum point) of the measurement value (x _m ). Overshoot occurs near the maximum point.

Also, the lightly filled area corresponds to the range covered by the second predicted value x _US based on the coefficient {b _i }. More specifically, this corresponds to a region surrounded by an envelope in the second predicted value x _US when the shift number j is changed in the range of j = 1 to 12. As understood from this figure, in the second prediction model, the next change tendency (for example, a decreasing tendency) is immediately given before the inflection point level (for example, the maximum point) of the measurement value (x _m ) is reached. Therefore, undershoot occurs near the maximum point.

On the other hand, the graph indicated by the alternate long and short dash line corresponds to a predicted value x _p obtained by combining the first predicted value x _OS and the second predicted value x _US . As can be understood from the figure, in the AR prediction combined model, the first predicted value x _OS obtained from the prediction method having relatively high sensitivity to the temporal change of the vibration data x _m and the sensitivity is relatively low. By combining the second predicted value x _US obtained from the prediction method, a tracking error with respect to a rapid time change of the vibration data x _m is reduced. As a result, a tracking error of about 4 mm occurs in the normal autoregressive model, whereas it can be suppressed to about 1.5 mm in the AR prediction combined model.

In step S8 of FIG. 7, the position teaching unit 73 appropriately adjusts the gain of the predicted value x _p calculated in step S7. This adjustment is performed in order to suppress the occurrence of a tracking error due to a sudden change in the phase component.

In step S <b> 9, the position teaching unit 73 converts the predicted value x _p adjusted in step S <b> 8 into a vibration displacement amount Δx of the workpiece W by executing coordinate conversion with the robot 61 as a reference. Here, the vibration displacement amount Δx corresponds to the deviation between the ideal value and the actual measurement value of the measurement position Pm.

  In step S10, the position teaching unit 73 outputs teaching information reflecting the vibration displacement amount Δx converted in step S9 to the robot 61 side. More specifically, the position teaching unit 73 outputs new coordinates obtained by adding the vibration displacement amount Δx to the coordinates of the known work position Pt1. As described above, the operation of the work device 20 related to the prediction process is completed.

FIG. 9 is a chart showing the flow of operation of the work device 20. In this example, there are five steps (or “processing” contents of the working device 20), specifically, (1) “vibration acquisition” step of acquiring the vibration data x _m , and (2) vibration data x (m + k). (Vibration prediction) step, (3) “position teaching” step of teaching the working positions Pt1 and Pt2 to the working unit 60, (4) “preparation” step by the working unit 60, and (5) the working unit 60. The "work" process by is described.

  Hereinafter, an example will be described in which Δt = 1, m = 12, and k = 9, and the work is performed on the workpiece W at t = 20 and 25.

  At t = 0 to 11, the vibration acquisition process (OP1) is sequentially executed. This process corresponds to steps S1 and S2 in FIG. Note that the vibration acquisition process is continued from the start time to the end time of tracking the workpiece W.

  At t = 11-14, the first vibration prediction step (OP2) is executed. This process corresponds to steps S3 to S7 in FIG. For example, in the case of the first autoregressive model, the vibration data string {x (n)} measured in the range of t = 0 to 11 is used.

  At t = 14 to 16, the first position teaching step (OP3) is executed. This process corresponds to steps S3 to S7 in FIG. Thereafter, at t = 16, the first preparation step (OP4) and the second vibration prediction step (OP2) are started in parallel.

  By performing the first preparation step (OP4) at t = 16-19, the robot 61 is driven and controlled to be in a position / posture accessible to the work position Pt1. Thereafter (t ≧ 19), it waits until time t = 20 when the first work process is started.

  At t = 20-21, the first work process (OP5) is executed. On the other hand, after the first vibration prediction step (OP2) is executed in the range of t = 16 to 19, the robot 61 is in a waiting state until the operation of the robot 61 is completed, that is, in the range of t = 19 to 21. .

  At t = 21 to 23, the second position teaching step (OP3) is executed. After that, after the second preparation process (OP4) is executed at t = 23 to 24, the process waits until time t = 25 when the second work process is started.

[Effects of this embodiment]
As described above, the working device 20 is a device having the working unit 60 that performs work on the workpiece W that is transported in a state of being in contact with the transporting device 30. Then, the work device 20 measures the vibration state of the workpiece W and / or the transfer device 30 at a certain time interval Δt, thereby obtaining a vibration data sequence {x (n)} for at least m points. Unit 50, vibration prediction unit 72 that predicts a vibration state at a measurement time point ahead of k points (for example, k = 9) from vibration data string {x (n)}, and the predicted vibration state as vibration of workpiece W A position teaching unit 73 is provided that teaches the working unit 60 the working positions Pt1 and Pt2 to which the vibration displacement amount Δx has been added before the previous measurement time has passed.

  In this work method, the vibration state of the workpiece W and / or the conveyance device 30 is sequentially measured at a constant time interval Δt to obtain at least m points of vibration data strings {x (n)}. The process [Steps S1 and S2], the prediction process [Steps S3 to S7] for predicting the vibration state at the measurement point ahead of k points from the vibration data string {x (n)}, and the predicted vibration state by the work W Is provided with teaching steps [steps S9 and S10] for teaching the working unit 60 the work positions Pt1 and Pt2 to which the vibration displacement amount Δx is added before the previous measurement time has passed.

  In this way, the vibration state at the previous measurement time is predicted from the measured vibration data string {x (n)}, the vibration state is converted into the vibration displacement amount Δx of the workpiece W, and the previous measurement time elapses. Since the work positions Pt1 and Pt2 to which the vibration displacement amount Δx is added are previously taught to the work unit 60, it is possible to teach in consideration of the vibration displacement amount Δx of the workpiece W. The teaching position at the time (previous measurement time) substantially coincides with the work positions Pt1 and Pt2 of the workpiece W. As a result, regardless of the vibration of the workpiece W that accompanies the transfer operation, it is possible to perform work while maintaining high positioning accuracy.

FIG. 10 is a graph showing the characteristic of the tracking error in each prediction model. The horizontal axis of the graph is acceleration (unit: mm / s ² ), and the vertical axis of the graph is tracking error (unit: mm). This “following error” corresponds to the absolute value of the difference between the maximum values between the actual value and the predicted value, as in the case of FIG.

  In this figure, the tracking error when an arbitrary acceleration is applied and sudden vibration is generated in the work W while the work W is being conveyed at a constant speed is plotted. Triangular plots show measured values using AR prediction (more specifically, predicted combined model), and diamond-shaped plots show measured values using PID control. The solid line is a regression line related to “AR prediction”, and the broken line is a regression line related to “PID control”. As understood from this graph, the slope of the straight line in “AR prediction” is about half that of “PID control”. That is, by adopting “AR prediction”, the tracking error is reduced to about half.

[Supplement]
In addition, this invention is not limited to embodiment mentioned above, Of course, it can change freely in the range which does not deviate from the main point of this invention.

  For example, in the above-described embodiment, the carts 41 and 42 and the rails 43 and 44 are described as examples of the moving unit 40. However, the moving unit 40 is not limited thereto, and may be realized by other configurations. The moving unit 40 may track the workpiece W along the conveyance direction by expanding and contracting and bending the arm of the working unit 60 (robots 61 to 64) that performs work on the workpiece W.

  In the above-described embodiment, the sensor devices 52 to 55 are described as an example of the measuring unit 50. However, various configurations may be adopted as long as vibration generated in the workpiece W and / or the conveyance device 30 can be detected. Good. For example, various sensors including an acceleration sensor, an angular velocity sensor, and a geomagnetic sensor may be attached to the hanger 32 that conveys the workpiece W. Various sensors may be attached to the workpiece W itself.

  Moreover, the content of the workpiece | work W, the conveying apparatus 30, or a work process is not restricted to the example of above-described embodiment. For example, various transport systems such as a slat conveyor and a belt conveyor can be applied as the transport device 30 in addition to the above-described hanger conveyor.

  Further, the vibration state prediction method is not limited to the autoregressive model, and various known prediction algorithms including a linear prediction method and pattern matching may be employed under the condition that the positioning accuracy is sufficiently maintained.

DESCRIPTION OF SYMBOLS 10 ... Conveying system 20 ... Working apparatus 30 ... Conveying apparatus 40 ... Moving part 50 ... Measuring part 60 ... Working part 70 ... Control part 71 ... Tracking control part 72 ... Vibration prediction part 73 ... Position teaching part Pm ... Measurement position Pt1, Pt2 ... Work position W ... Work

Claims (2)

It is a working method using a working device having a working unit that performs work on a workpiece conveyed in a state in contact with the conveying device,
A measurement step of acquiring vibration data strings for at least m (m is an integer of 2 or more) points by sequentially measuring the vibration state of the workpiece and / or the transfer device at a constant time interval;
A prediction step of predicting the vibration state at the previous measurement time point by k (k is an integer of 1 or more) from the measured vibration data sequence;
A teaching step of converting the predicted vibration state into a vibration displacement amount of the workpiece and teaching the work position to which the vibration displacement amount is added before the previous measurement time has elapsed. ,
The prediction step includes
A first calculation step of calculating a first predicted value at the previous measurement time point from a vibration data string for the most recent m points using an m-th order autoregressive model expressed by a first autoregressive coefficient; ,
Using the m-th order autoregressive model expressed by the second autoregressive coefficient, from the vibration data string for the most recent (m + j) (j is an integer satisfying 1 ≦ j ≦ m), A second calculation step of calculating a second predicted value at the time of the measurement;
A combining step of predicting the vibration state by combining the calculated first and second predicted values;
Including
In the second calculating step, the second predicted value is calculated from a difference data string between vibration data having different measurement time points by j points . A working device having a working unit that performs work on a workpiece conveyed in a state in contact with the conveying device,
A measurement unit that acquires vibration data strings for at least m (m is an integer of 2 or more) points by sequentially measuring the vibration state of the workpiece and / or the transfer device at a constant time interval;
A vibration prediction unit that predicts the vibration state at the previous measurement time point by k (k is an integer of 1 or more) from the vibration data string measured by the measurement unit;
The vibration state predicted by the vibration prediction unit is converted into a vibration displacement amount of the workpiece, and the work position to which the vibration displacement amount is added is taught to the work unit before the previous measurement time has elapsed. A teaching unit and
The vibration prediction unit
A first calculation process for calculating a first predicted value at the previous measurement time point from a vibration data string for the latest m points using an m-th order autoregressive model expressed by a first autoregressive coefficient; ,
Using the m-th order autoregressive model expressed by the second autoregressive coefficient, from the vibration data string for the most recent (m + j) (j is an integer satisfying 1 ≦ j ≦ m), A second calculation process for calculating a second predicted value at the time of measurement of
A combining process for predicting the vibration state by combining the calculated first and second predicted values;
Run
In the second calculation process, the second predicted value is calculated from a difference data string between vibration data having different measurement time points by j points .

Images