JP2019087068A - Control device, pressure test device and control method - Google Patents

Abstract

To automatically adjust a control parameter and maintain high learning performance (learning speed and control performance after learning) even if characteristics of a control object changes.SOLUTION: A control device outputs an operation amount to a control object so that a control amount from the control object tracks a target value. The control device comprises an adaptive controller and a learning controller. The adaptive controller adaptively adjusts a control parameter according to characteristics of the control object. The learning controller corrects and outputs a current input using an input of one cycle earlier or an input of one trial earlier of repeated trials.SELECTED DRAWING: Figure 1

Description

  The present invention mainly relates to a control device using adaptive control and learning control.

  Conventionally, the control deviation is reduced by correcting the operation amount using the control deviation of one cycle before for the periodic target signal or the control deviation of one trial in repeated trials, and repeating this control. Learning control is known.

  In learning control, since learning can be performed using only the operation amount and the control deviation without using the mathematical model of the control object, a certain degree of control performance can be obtained even if the characteristics of the control object are not exactly known. However, if the characteristic of the control target fluctuates significantly, the learning speed may be slow and the control performance after learning may not be good.

  In order to prevent such a decrease in learning speed, a method has been proposed in which control parameters used for learning control are adaptively adjusted in accordance with characteristic variations of a control target. Patent Document 1 discloses a position control device including an iterative learning control circuit to which this method is applied.

  The position control device of Patent Document 1 includes a first filter to which a deviation between a target value and an output of a detection unit that detects a position of a control target is input, and a linear time-invariant second filter that cuts off a predetermined band. And an iterative learning control circuit that feeds forward a control input to the controlled object. In this position control device, the characteristic of the first filter is changed in accordance with the parameter fluctuation of the control target.

Patent No. 5279299 gazette

  However, although patent document 1 can acquire the effect which prevents the fall of learning speed by adjusting the control parameter of a 1st filter adaptively according to the characteristic fluctuation | variation of control object, the control performance after learning is It can not be improved and there is room for improvement in this respect.

  The present invention has been made in view of the above circumstances, and an object thereof is to be able to automatically adjust control parameters even when the characteristic of the control object fluctuates, and to achieve high learning performance (learning speed and control performance after learning) To be able to maintain the

  The problem to be solved by the present invention is as described above, and next, means for solving the problem and its effect will be described.

  According to a first aspect of the present invention, a control device having the following configuration is provided. That is, the control device outputs an operation amount to the control target so as to make the control amount from the control target follow the instructed target value. The controller includes an adaptive controller and a learning controller. The adaptive controller adaptively adjusts control parameters according to the characteristics of the control target. The learning controller corrects and outputs the current input using the input one cycle before or the input one trial before the repeated trial.

  Thereby, since the control parameter of the adaptive controller is adaptively adjusted according to the characteristic fluctuation of the control object, even when the characteristic of the control object fluctuates, it is possible to suppress the influence on the stability and responsiveness of the control system. Therefore, the learning speed of the learning controller and the control performance after learning can be favorably maintained without being affected by the characteristic variation of the control object.

  According to a second aspect of the present invention, the following control method is provided. That is, in this control method, the operation amount is output to the control target so that the control amount from the control target follows the instructed target value. In the control method, an adaptive controller adaptively adjusts control parameters in accordance with the characteristics of the control target. The operation amount output by the adaptive controller, the target value input to the adaptive controller, and the control are output by a learning controller that performs correction based on the input one cycle before or the input before one trial of repeated trials. The deviation from the quantity or the quantity determined by the adaptive controller to calculate the manipulated variable is corrected.

  Thereby, since the control parameter of the adaptive controller is adaptively adjusted according to the characteristic fluctuation of the control object, even when the characteristic of the control object fluctuates, it is possible to suppress the influence on the stability and responsiveness of the control system. Therefore, the learning speed of the learning controller and the control performance after learning can be favorably maintained without being affected by the characteristic variation of the control object.

  According to the present invention, control parameters can be automatically adjusted and high learning performance (learning speed and control performance after learning) can be maintained even when the characteristic of the control target changes.

FIG. 1 is a block diagram showing a schematic configuration of an adaptive learning control device according to a first embodiment of the present invention. FIG. 2 is a block diagram showing a plurality of configuration examples of a learning controller. FIG. 1 is a block diagram showing a configuration of an adaptive learning control device according to a first embodiment. The block diagram which shows the composition which performs PFC from the output of controlled object and the output of a reference model control part, adding PFC. The block diagram which shows the structure of the adaptive learning control apparatus of 2nd Embodiment. The block diagram which shows the structure of the adaptive learning control apparatus of 3rd Embodiment. The block diagram which shows the structure of the adaptive learning control apparatus of 4th Embodiment. The graph which shows the simulation result of general learning control. FIG. 7 is a graph showing simulation results in a prior art configuration of adaptively adjusting a learning filter. The graph which shows the simulation result in the composition of this embodiment. The figure which compares and shows the structure of a prior art, and this invention by the analogy which used PI control system. The schematic diagram which shows the structure of a pressure | voltage resistant test apparatus provided with an adaptive learning control apparatus. FIG. 8 is a block diagram showing a configuration of an adaptive learning control device of a modified example.

  Next, embodiments of the present invention will be described with reference to the drawings. FIG. 1 is a block diagram showing a schematic configuration of an adaptive learning control device 1 according to a first embodiment of the present invention. FIG. 2 is a block diagram showing a plurality of configuration examples of the learning controller 3.

  An adaptive learning control device (control device) 1 shown in FIG. 1 is used to cause a control amount y from the control target 4 to follow a target value r set in advance so as to have a periodic waveform. The adaptive learning control device 1 can be configured, for example, from a computer such as a microcontroller.

  The adaptive learning control device 1 includes an adaptive controller 2 and a learning controller 3.

  The adaptive controller 2 performs control of causing the control amount y from the control target 4 to follow a target value r instructed from a control device (for example, the upper controller 51 described later illustrated in FIG. 12) including a computer or the like. . Although the details will be described later, the control parameters of the adaptive controller 2 of FIG. 1 can be adaptively adjusted in accordance with the characteristic variation of the controlled object 4. The manipulated variable us generated by the adaptive controller 2 is output to the learning controller 3.

  The learning controller 3 corrects the operation amount us currently output by the adaptive controller 2 using the operation amount us' output by the adaptive controller 2 one cycle earlier, and outputs the operation amount u to the control target 4 Do.

  The learning controller 3 can be appropriately selected and used from, for example, the learning controllers 3a, 3b and 3c having the three types of configurations shown in FIG. 2 in accordance with the required responsiveness and stability. Each of the learning controllers 3a, 3b, 3c has a stabilization compensator 31 and a time compensator 32.

  The stabilization compensator 31 is configured, for example, as a low pass filter. The stabilization compensator 31 passes low frequency components having frequencies lower than a predetermined cutoff frequency to the input signal, and attenuates high frequency components having frequencies higher than the cutoff frequency. Cut off frequency components. The stabilization compensator 31 can be configured, for example, as an FIR (Finite Impulse Response) filter.

  The stabilization compensator 31 can prevent the manipulated variable u from increasing and diverging with respect to the change of the target value r which can not be followed by the apparatus (hardware).

  The time compensator 32 is configured to delay the input signal by a predetermined time and to output the delayed signal. The predetermined time may be determined in consideration of a phase delay or the like by the stabilization compensator 31 so that the learning controller 3 can perform correction using a signal corresponding to a signal input one cycle earlier. it can.

  In the embodiment shown in FIG. 1, the learning controller 3 configured in this way uses the operation amount corresponding to the operation amount us ′ input one cycle earlier, and the adaptive controller 2 becomes the control target 4. The process of correcting the given operation amount us is repeated. Thereby, the waveform of the control amount y can be made close to the waveform of the target value r.

  The adaptive learning control device 1 includes a CPU, a ROM, a RAM and the like, and the above-mentioned ROM stores a program for realizing the control method of the present invention. The cooperation of the hardware and software allows the adaptive learning control device 1 to operate as the adaptive controller 2 and the learning controller 3 or the like. However, the adaptive controller 2 and the learning controller 3 may be configured by analog or digital circuits.

  As is known, the learning controller 3 has better learning speed and control performance after learning as the characteristics of the control system before adding learning control are better. In this regard, in the adaptive learning control device 1 of the present embodiment, the learning controller 3 is added to the adaptive controller 2 capable of maintaining control performance even when the characteristic of the control target changes. Thereby, the learning speed of the learning controller 3 and the control performance after learning can be improved.

  Subsequently, an example of the configuration of the adaptive controller 2 in the case of using simple adaptive control will be described in detail with reference to FIG. FIG. 3 is a block diagram showing the configuration of the adaptive learning control device 1 according to the first embodiment.

  As shown in FIG. 3, the adaptive controller 2 includes a reference model control unit 20, a first adaptive feedforward control unit (adaptive feedforward control unit) 21, and a second adaptive feedforward control unit (adaptive feedforward control unit). 22), an adaptive feedback control unit 23, a parallel feed forward compensator (PFC) 24, a first subtractor 25, a second subtracter 26, a first adder 27, and a second addition. And a vessel 28.

  The reference model control unit 20 is configured to apply a reference model Gm designed to generate a predetermined (ideal) response. The reference model Gm can be represented by, for example, a transfer function such as a first-order lag system or a second-order lag system. The reference model control unit 20 outputs the generated reference output ym to the first subtractor 25. Further, the reference model control unit 20 outputs the state quantity xm of the model to the second adaptive feedforward control unit 22.

  The first subtractor 25 subtracts the control amount y input to the adaptive controller 2 from the reference output ym input from the reference model control unit 20. The first subtractor 25 outputs the obtained result, the deviation e, to the second subtractor 26.

  The second subtractor 26 subtracts the compensation value yf output from the PFC 24 from the deviation e input from the first subtractor 25. The second subtractor 26 outputs the obtained feedback deviation ea to the adaptive feedback control unit 23.

  The first adaptive feedforward control unit 21 multiplies the input target value r by the adjusted first feedforward gain Kr. The first adaptive feedforward control unit 21 outputs the obtained multiplication result ur to the first adder 27.

  The second adaptive feedforward control unit 22 multiplies the state amount xm of the reference model Gm input from the reference model control unit 20 by the adjusted second feedforward gain Kx. The second adaptive feedforward control unit 22 outputs the obtained multiplication result ux to the first adder 27.

  The adaptive feedback control unit 23 multiplies the feedback deviation ea input from the second subtractor 26 by the adjusted feedback gain Ke. The adaptive feedback control unit 23 outputs the obtained manipulated variable ue, which is the multiplication result, to the second adder 28 and to the PFC 24.

  The PFC 24 is configured to generate a pseudo compensation value yf based on the output of the adaptive feedback control unit 23, and to output the pseudo compensation value yf to the second subtractor 26. The circuit shown in FIG. 3 is the circuit shown in FIG. 4, that is, the output of the PFC is added to the output of the control target 4 and the output of the PFC having the same configuration is used as the output of the reference model control unit 20. It is equivalent to a circuit that additionally performs dynamic compensation. As a result, even if a response delay occurs in the actual control target 4, the stability of the control system is ensured by making it appear that there is no response delay in the expansion control target in which the PFC is added to the control target 4. Can. Further, the output of the control target 4 can be made to follow the output of the reference model Gm (reference output ym), not the output of the expansion control target, by the above-described dynamic compensation.

  The gains Kr, Kx, Ke of the first adaptive feedforward control unit 21, the second adaptive feedforward control unit 22, and the adaptive feedback control unit 23 respectively output the output of the control object 4 as the output of the reference model Gm (reference output ym) Are adjusted adaptively to make them follow. The control gain parameters (control parameters) Kr, Kx, Ke can be determined as shown in the following equation (1) by an adaptive adjustment rule known in simple adaptive control, for example, an adaptive adjustment rule of proportional and integral. .

ここで、γ_Pr，γ_Ir，γ_Px，γ_Ix，γ_Ieは、ゲインの調整スピードを決める設計パラメータである。σは、Ｋｅの発散を防ぐためのフィードバックゲインであって、偏差が大きいときに大きく設定される。

Here,? _Pr ,? _Ir ,? _{Px ,? Ix} and? _Ie are design parameters that determine the speed of adjusting the gain. σ is a feedback gain for preventing the divergence of Ke, and is set large when the deviation is large.

  The first adder 27 adds the multiplication result ur which is the output of the first adaptive feedforward control unit 21 and the multiplication result ux which is the output of the second adaptive feedforward control unit 22. The first adder 27 outputs the obtained addition result uf to the second adder 28.

  The second adder 28 adds the addition result uf that is the output of the first adder 27 and the operation amount ue that is the output of the adaptive feedback control unit 23. The obtained addition result is output to the learning controller 3 as the above-mentioned operation amount us.

  With the above configuration, in the adaptive learning control device 1, the stability and the tracking performance of the control system can be favorably ensured by the adaptive controller 2. As a result, even when the characteristic of the control target 4 changes, the learning speed of the learning controller 3 and the control performance after learning can be favorably maintained.

  Next, a second embodiment will be described. FIG. 5 is a block diagram showing the configuration of the adaptive learning control device 1x of the second embodiment. In the description of the present embodiment, members that are the same as or similar to the above-described embodiment may be assigned the same reference numerals in the drawings, and descriptions thereof may be omitted.

  As shown in FIG. 5, in the adaptive learning control device 1x of this embodiment, a learning controller 3 is added to the inside of the adaptive controller 2x. Specifically, the learning controller 3 is provided between the adaptive feedback control unit 23 of the adaptive controller 2 x and the second adder 28. Therefore, the operation amount ue output from the adaptive feedback control unit 23 is input to the learning controller 3.

  The learning controller 3 can be appropriately selected and used from, for example, the three types of configurations shown in FIG. 2 as in the first embodiment described above. The same applies to the third and fourth embodiments described later.

  As shown in FIG. 5, the learning controller 3 included in the adaptive controller 2x corrects the operation amount ue output from the adaptive feedback control unit 23 using the operation amount ue 'one cycle before. The manipulated variable ue to be corrected is an amount obtained by the adaptive controller 2x to calculate the manipulated variable u output from the adaptive controller 2x. The correction operation amount ul obtained by the learning controller 3 is input to the second adder 28.

  The second adder 28 adds the addition result uf, which is the output of the first adder 27, and the correction operation amount ul, which is the output of the learning controller 3. The obtained addition result is output from the adaptive controller 2x (the adaptive learning control device 1x) to the control target 4 as the above-described manipulated variable u.

  The adaptive learning control device 1x of this embodiment can also obtain the same effect as the adaptive learning control device 1 of the first embodiment.

  Next, a third embodiment will be described. FIG. 6 is a block diagram showing the configuration of an adaptive learning control device 1y according to the third embodiment. In the description of the present embodiment, members that are the same as or similar to the above-described embodiment may be assigned the same reference numerals in the drawings, and descriptions thereof may be omitted.

  In this embodiment, a modification is added to the first embodiment (FIG. 2), and the learning controller 3 is added to the inside of the adaptive controller 2 as in the second embodiment. As shown in FIG. 6, the learning controller 3 is added between the first subtractor 25 and the second subtractor 26. The learning controller 3 learns the deviation e between the reference output ym output from the reference model Gm and the control amount y input to the adaptive controller 2, and corrects the deviation e. The deviation e can be said to be an amount determined by the adaptive controller 2y to calculate the manipulated variable u output from the adaptive controller 2y. The learning controller 3 corrects the present deviation e using the deviation e ′ of one cycle before output from the first subtractor 25. The corrected deviation el obtained by the learning controller 3 is input to the second subtractor 26. The configuration other than that described above is the same as that of the above-described first embodiment.

  Next, a fourth embodiment will be described. FIG. 7 is a block diagram showing a configuration of an adaptive learning control device 1z of the fourth embodiment. In the description of the present embodiment, members that are the same as or similar to the above-described embodiment may be assigned the same reference numerals in the drawings, and descriptions thereof may be omitted.

  In this embodiment, a modification is added to the first embodiment (FIG. 2), and the learning controller 3 is added to the inside of the adaptive controller 2 as in the second embodiment. As shown in FIG. 7, the learning controller 3 is added between the second subtractor 26 and the adaptive feedback control unit 23. The learning controller 3 learns the deviation ea inputted to the adaptive controller 2 and corrects the deviation ea. The deviation ea can be said to be an amount determined by the adaptive controller 2z to calculate the manipulated variable u output from the adaptive controller 2z. The learning controller 3 corrects the present deviation ea using the deviation ea 'of one cycle before output from the second subtractor 26. The corrected deviation el obtained by the learning controller 3 is input to the adaptive feedback control unit 23. The configuration other than that described above is the same as that of the above-described first embodiment.

  Next, the effects of the control of the present invention will be described with reference to the simulation results shown in FIG. 8 to FIG.

  In this simulation experiment, changes in the control amount were examined when general learning control, learning control of the prior art (the above-mentioned Patent Document 1), and learning control of the present invention (the second embodiment) were respectively performed. . In any of the simulations, a trapezoidal waveform with one cycle of 10 seconds was used as the target value.

  In the graphs from FIG. 8 to FIG. 10, the target value and the control amount are shown on the upper side, and the deviation (control deviation) between the target value and the control amount is shown on the lower side. In any of the upper and lower graphs, the horizontal axis is time. In the simulation calculation, the control performance of the control system is intentionally lowered before the time of zero seconds, thereby simulating that the characteristic of the controlled object has largely changed at the time of zero seconds. The time is supported by the upper and lower graphs. In the graph on the right side, the details of the transition of the values for one cycle when sufficient time has elapsed in the graph on the left side are shown in a stretched form in the time axis direction.

  FIG. 8 shows simulation results of general learning control (in other words, when only a learning controller is used). From the graph at the upper left of FIG. 8, it can be seen that, in general learning control, it takes about 10 cycles of time for the control amount to sufficiently approach the target value.

  FIG. 9 shows the simulation result of the prior art (the configuration corresponding to Patent Document 1). From the graph at the upper left of FIG. 9, it can be seen that in the configuration of the prior art, the control amount approaches the target value sufficiently if about 5 cycles, which is about half of general learning control, elapse.

  FIG. 10 is a simulation result of the present invention (the second embodiment described above). As shown in the graph at the upper left of FIG. 10, when using the adaptive learning control device 1x according to the present embodiment, it is 1/2 to 2 of the prior art, and 2 to 3 cycles which is about 1/4 compared to general learning control. The control amount can be sufficiently brought close to the target value.

  Then, comparing the lower right graph in FIG. 10 with the lower right graphs in FIG. 8 and FIG. 9, according to the control of the present invention, the control deviation after the control amount follows the target value is generally learned. It can be seen that it is extremely small compared to control and prior art learning control.

  From the above simulation results, it was confirmed that the control of the present invention has quick learning speed and excellent post-learning control performance (that is, accurate tracking characteristics).

  Next, with reference to FIG. 11, the difference between the above-mentioned prior art and the control of the present invention will be supplementarily explained by an analogy using a PI control system.

  In the servo system, the principle that "in order to follow the target waveform without steady deviation, a compensation element (a compensator with the same pole as the target waveform) must be provided in the feedback loop". It is known that there is. This principle is called the internal model principle.

  For example, an arbitrary stepped target waveform can be generated by giving the integrator (1 / s) an appropriate initial value. Correspondingly, it is necessary to incorporate an integrator into the feedback loop in order to null the steady state deviation with respect to the stepped target waveform. PI control system is a typical example.

  If this concept is applied, in order to make the steady state deviation zero with respect to the target waveform whose period is L, a mechanism that generates a periodic function whose period is L may be included in the feedback loop. In fact, the above-mentioned learning controller 3 has such a property because an output based on an input one cycle before is performed. That is, it can be said that the learning controller 3 when making a control amount follow a periodic target waveform corresponds to an integrator in a PI control system to which a stepped target waveform is given.

  If the above prior art (patent document 1) and this embodiment are compared to a PI control system, the prior art adaptively adjusts only the integral term as shown in FIG. Is an adaptive adjustment of both the proportional term and the integral term as shown in FIG. 11 (b), and it can be said that there is an essential difference.

  Also, in general, when the characteristic of the control target fluctuates significantly, the response can be further improved by adjusting not only the integral term but also the proportional term. In this respect, it can be said that the present invention has a clear advantage over the prior art.

  Then, with reference to FIG. 12, the application example which applied the adaptive learning control method demonstrated above to the pressure | voltage resistant testing apparatus is demonstrated. FIG. 12 is a schematic view showing a configuration of a pressure resistance test device 5 provided with the adaptive learning control device 1.

  The pressure resistance test apparatus (pressure test apparatus) 5 of this application example shown in FIG. 12 is for evaluating the internal pressure fatigue resistance and the like by repeatedly applying a pressure to the specimen 60. Although various things, such as piping, a valve | bulb, the outer shell of a machine, etc., can be considered as the test object 60, in FIG. 12, the example which tests a pressure vessel is shown.

  The pressure resistance test device 5 includes a pressure control controller (pressure control device) 50, a host controller 51, a hydraulic unit 52, a pressure booster 53, and a pressure sensor 54.

  The pressure controller 50 sets a target pressure waveform (target value r) input from the host controller 51 and an actual pressure value (control amount y) input from the pressure sensor 54 that detects the pressure in the specimen 60. Based on this, the operation amount u is output to the hydraulic unit 52.

  In this application example, the adaptive learning control device 1 of the first embodiment is used as the pressure controller 50 (however, adaptive learning control devices 1x, 1y, 1z of other embodiments can also be used). The whole including the hydraulic unit 52, the pressure intensifier 53 and the test object 60 corresponds to the control target 4 by the adaptive learning control device 1.

  The hydraulic unit 52 drives the pressure intensifier 53 based on the operation amount u input from the pressure controller 50. The hydraulic unit 52 includes, for example, a solenoid valve or the like that switches pressure oil discharged from a pump (not shown).

  The pressure booster 53 includes a cylinder or the like driven by the hydraulic unit 52. The pressure intensifier 53 increases or decreases the pressure applied to the sample 60 by discharging or sucking the pressure medium.

  The pressure sensor 54 actually detects the pressure in the sample 60, and outputs a pressure value which is a detected value to the pressure controller 50 as a control amount y.

  The pressure resistance test apparatus 5 can repeatedly apply pressure to the specimen 60 according to a waveform such as a sine wave, a triangular wave, or a trapezoidal wave as shown by a small graph in FIG. The pressure controller 50 can cause the pressure (that is, the control amount y) in the sample 60 to follow the fluctuating target value r quickly and accurately by performing the above-described adaptive learning control.

  As described above, various specimens having different capacities and the like can be considered as the specimen 60 to be subjected to the test. Since the change in capacity changes the ease of pressure increase, the control performance may be deteriorated in the conventional method. In this respect, the pressure resistance test apparatus 5 of FIG. 12 is actually added to the sample 60 because the pressure control controller 50 has the above-described excellent control performance even if the sample 60 is replaced with another one. The pressure can be made to follow the target pressure waveform quickly and accurately. Thereby, various test pieces 60 can be tested without time and effort.

  As described above, the adaptive learning control devices 1, 1x, 1y, 1z of the above embodiments operate the control target 4 so that the control amount y from the control target 4 follows the instructed target value r. Output the quantity u. The adaptive learning control devices 1, 1x, 1y, 1z include adaptive controllers 2, 2x, 2y, 2z and a learning controller 3. The adaptive controllers 2, 2x, 2y, 2z adaptively adjust the control gain parameters Kr, Kx, Ke in accordance with the characteristics of the control target 4. The learning controller 3 corrects and outputs the present inputs us, ue, e and ea using the inputs us ', ue', e 'and ea' one cycle earlier.

  As a result, the control gain parameters Kr, Kx, Ke of the adaptive controllers 2, 2x, 2y, 2z are adaptively adjusted in accordance with the characteristic fluctuation of the controlled object 4, so that the control by the fluctuation of the characteristic of the controlled object 4 The influence on the stability and responsiveness of the system can be suppressed. Therefore, the learning speed of the learning controller 3 and the control performance after learning can be favorably maintained without being affected by the characteristic variation of the control target 4.

  Further, in the adaptive learning control device 1 according to the first embodiment, the learning controller 3 corrects the operation amount us output from the adaptive controller 2.

  As a result, the control system before input to the learning controller 3 can have high stability against characteristic variations of the control target 4 and good tracking characteristics for the target value r. Further, the influence of the fluctuation of the characteristics of the control target 4 on the control system is absorbed by the adaptive controller 2. Therefore, even when the characteristic of the control target 4 changes, the learning characteristic of the learning controller 3 can be maintained well.

  On the other hand, in the adaptive learning control device 1x of the second embodiment, an adaptive feedback control unit 23 is provided in the adaptive controller 2x. The learning controller 3 corrects the operation amount ue output from the adaptive feedback control unit 23. The manipulated variable ue is an amount obtained by the adaptive controller 2x to calculate the manipulated variable u output from the adaptive controller 2x.

  In the adaptive learning control devices 1y and 1z of the third embodiment and the fourth embodiment, the learning controller 3 performs control that the adaptive controllers 2y and 2z obtain based on the control amount y input to the adaptive controllers 2y and 2z. Correct the deviations e and ea. The control deviations e and ea are quantities obtained by the adaptive controllers 2 y and 2 z in order to calculate the manipulated variables u output from the adaptive controllers 2 y and 2 z.

  Also in this configuration, when the characteristic of the control target 4 changes, the learning characteristic of the learning controller 3 can be favorably maintained.

  Further, in the above-described adaptive learning control devices 1, 1x, 1y, 1z, the adaptive controllers 2, 2x, 2y, 2z include the reference model control unit 20, the first adaptive feedforward control unit 21, and the second adaptive feed. It is configured as a simple adaptive controller configured of a forward control unit 22, an adaptive feedback control unit 23, and a PFC 24. The reference model control unit 20 applies a reference model Gm that gives a predetermined response. The first adaptive feedforward control unit 21 feedsforward the target value r. The second adaptive feedforward control unit 22 feedsforward the state quantity xm of the reference model Gm. The adaptive feedback control unit 23 receives the control deviation ea of the enlargement system obtained by subtracting the output yf of the PFC 24 from the deviation e between the output of the reference model control unit 20 and the control amount y from the control target 4.

  Thereby, good control performance in the adaptive controllers 2, 2x, 2y, 2z can be realized.

  Further, the pressure resistance test device 5 of FIG. 12 includes the above-mentioned adaptive learning control devices 1, 1x, 1y, 1z.

  As a result, even if the characteristics of the control target 4 are significantly changed along with various replacements of the test body 60 for the pressure resistance test, the adaptive learning control devices 1, 1x, 1y, 1z realize good control. can do.

  Further, in the above-described adaptive learning control devices 1, 1x, 1y, 1z, the control amount is controlled to the control target 4 so that the control amount y from the control target 4 follows the instructed target value r by the following control method. Output. That is, the control gain parameters Kr, Kx, Ke are adaptively adjusted in accordance with the characteristics of the control target 4 by the adaptive control system used by the adaptive controller 2. In the first embodiment, the operation amount u output from the adaptive controller 2 is corrected by the learning controller 3 in which correction based on the input one cycle before is performed. In the second embodiment, the learning controller 3 corrects the operation amount ue determined by the adaptive controller 2x in order to calculate the operation amount u output from the adaptive controller 2x. In the third and fourth embodiments, the control deviations e and e determined by the adaptive controllers 2 y and 2 z are corrected by the learning controller 3 in order to calculate the manipulated variables u output from the adaptive controllers 2 y and 2 z. Do.

  As a result, good control performance that can exhibit quick learning speed and accurate tracking characteristics can be realized with a simple configuration and control algorithm.

  Next, modifications of the above embodiment will be described. FIG. 13 is a block diagram showing a configuration of the adaptive learning control device 1a of the modification. In the description of the present modification, the same or similar members as or to those of the above-described embodiment may be denoted by the same reference numerals as those of the embodiment, and the description thereof may be omitted.

  The adaptive learning control device 1a of this modification is configured by adding a learning controller 3 in front of (on the input side of) the adaptive controller 2 as shown in FIG. This learning controller 3 can also be appropriately selected and used from the learning controllers 3a, 3b, and 3c having the three types of configurations shown in FIG. 2 as in the above-described embodiment.

  The learning controller 3 corrects the currently input deviation e using the signal corresponding to the deviation e ′ between the target value r and the control amount y input one cycle earlier, and sets it as the deviation el as the adaptive controller 2. Output to As a result, the deviation e between the control value y and the target value r input from the control target 4 to the adaptive controller 2 is corrected by the learning controller 3.

  The same effect as the above-mentioned embodiment can be acquired also by this modification.

  Although the preferred embodiment and modification of the present invention have been described above, the above configuration can be modified as follows, for example.

  The above-mentioned adaptive learning control devices 1, 1x, 1y, 1z, 1a can also be applied to iterative learning control in a system capable of repeating trials. In this case, the learning controller 3 may be configured to correct and output the current input based on the input before one repeat trial.

  The adaptive learning control devices 1, 1x, 1y, 1z, 1a are not limited to the pressure applied to the test object 60 of the pressure resistance test device 5, and control various other things such as the position of the tip of the arm type robot, for example. It can be used for

1 adaptive learning control device 2 adaptive controller 3 learning controller 4 control target r target value u operation amount y control amount Kr, Kx, Ke control gain parameter (control parameter)

Claims (5)

A control device that outputs an operation amount to a control target so that a control amount from a control target follows an instructed target value,
An adaptive controller that adaptively adjusts control parameters according to the characteristics of the control object;
A learning controller that corrects and outputs the current input using the input one cycle earlier or the input one trial before the repeated trial;
A control device comprising: The control device according to claim 1, wherein
The learning controller may calculate the manipulated variable output from the adaptive controller, a deviation between the target value and the controlled variable input to the adaptive controller, or the adaptive controller to calculate the manipulated variable. A control device characterized by correcting an amount obtained by The control device according to claim 1 or 2, wherein
The adaptive controller
A reference model control unit that applies a reference model that gives a predetermined response;
Adaptive feedforward control,
An adaptive feedback control unit that receives a deviation between an output of the reference model control unit and the control amount from the control target;
Parallel feedforward compensator,
A control device characterized by being a simple adaptive controller comprising:   A pressure test apparatus comprising the control device according to any one of claims 1 to 3. A control method for outputting an operation amount to a control target so that a control amount from a control target follows an instructed target value,
An adaptive controller adaptively adjusts control parameters according to the characteristics of the control object,
The operation amount output by the adaptive controller, the target value input to the adaptive controller, and the control are output by a learning controller that performs correction based on the input one cycle before or the input before one trial of repeated trials. A control method comprising: correcting a deviation from an amount or an amount determined by the adaptive controller to calculate the manipulated variable.

Images