JP7089361B2 - Command generator and command generation method - Google Patents

Description

The present invention relates to a command generator and a command generation method.

Generally, as a system that inputs a command to a plant and controls the output of the plant according to the target waveform, the output of the plant is detected and this output is fed back to deviate from the indicated value indicated by the target waveform. There is a request for a command to give to the plant by performing PI compensation and PID compensation. As described above, as a system for performing feedback control, for example, there is a system for controlling a hydraulic actuator used in a vibration tester used for testing the fatigue durability of a test piece such as a machine, a machine part or a structure. ..

In such a vibration tester, for example, a vibration test is performed by applying sinusoidal vibration to the test piece with a hydraulic actuator. The hydraulic actuator includes a cylinder having an extension side chamber and a compression side chamber internally partitioned by a piston, a rod connected to the piston and inserted into the cylinder and movable in the axial direction with respect to the cylinder, and an extension side chamber and the compression side. The chamber is equipped with a servo valve that can communicate with the hydraulic source and the tank. Such a control device that controls the hydraulic actuator as a plant performs feedback control that feeds back the displacement, speed, or acceleration of the hydraulic actuator, and drives the servo valve to control the stroke and thrust of the hydraulic actuator (for example,). See Patent Document 1).

Japanese Unexamined Patent Publication No. 11-101708

However, in the conventional system, since feedback control is performed, the output of the plant approaches the indicated value indicated by the target waveform to some extent, but the plant output is in line with the target waveform due to the response delay of the plant and the non-linear characteristics of the plant. It is not possible to make the plant operate ideally.

Therefore, an object of the present invention is to provide a command generation device and a command generation method capable of generating an input command capable of ideally operating the plant by accurately matching the output of the plant with the target waveform.

In order to achieve the above object, the command generator of the present invention is optimally suitable for outputting the target output for each predetermined time unit obtained by dissociating the target waveform to be output to the plant in the predetermined time unit. A command is obtained for each target output, and an input command is generated based on the optimum command for each target output. With the command generator configured in this way, even if the plant has response delays and non-linear characteristics, it is optimal for the plant to output the target output for each target output obtained by discretizing the target waveform. Find the optimal command.

Further, the command generator gives the plant a first sample command and a second sample command having a value different from that of the first sample command for each target output, and for each target output, the first sample command and the first sample command are used. Equipped with an optimum command search unit that obtains the optimum command corresponding to the target output based on the output of the plant for the input of the second sample command, the first sample command and the second sample command, and the target output. Optimal command An input command may be generated based on the optimum command obtained by the exploration unit. In the command generator configured in this way, since the first sample command and the second sample command are given to the plant to obtain the optimum command, the optimum command can be quickly obtained regardless of the characteristics of the plant.

Furthermore, the command generator uses an evaluation function with the plant output and the target output as parameters, and the value approaches 0 when the deviation between the plant output and the target output becomes small. An evaluation function that obtains a candidate command based on the values of the evaluation function that inputs the output of the plant for the input of the second sample command and the input of the first sample command and the second sample command, and inputs the output of the plant for the input of the candidate command. When the value of is less than the threshold value, the candidate command may be used as the optimum command. The command generator configured in this way can obtain the optimum command more quickly regardless of the plant characteristics.

Further, the evaluation function may be defined as a function considering the wasted time from the input to the output of the command to the plant. The command generator configured in this way can generate an input command capable of ideally operating the plant by matching the output of the plant with the target waveform with high accuracy.

Further, the evaluation function may be defined as a function considering the influence on the output of the plant in the command to the plant after the waste time has elapsed. The command generator configured in this way can generate an input command capable of ideally operating the plant by matching the output of the plant with the target waveform with high accuracy.

Further, when the command generator inputs the first sample command, the second sample command, or the candidate command for the target output, if there is an optimum command adopted so far, the command generator inputs the adopted optimum command. Later, a first sample command, a second sample command or a candidate command may be input. With the command generator configured in this way, when the optimum command for a certain target output is obtained, the optimum command that has been adopted up to that point is input, so the optimum command that takes into account the dynamic characteristics of the plant is obtained. obtain. Therefore, the command generator can generate an input command that can ideally operate the plant by matching the output of the plant with the target waveform with high accuracy.

Then, the command generator determines whether or not the optimum command exists in the search range between the first sample command and the second sample command based on the first sample command, the second sample command, and the candidate command. It may be determined whether or not, and if it does not exist, a process of changing the search range may be performed. The command generator configured in this way can quickly find the search range in which the optimum command can exist and quickly obtain the optimum command. Further, the command generator has a comparison result between the value of the evaluation function inputting the output of the plant with respect to the input of the first sample command and the value of the evaluation function inputting the output of the plant with respect to the input of the candidate command, and the second. Based on the comparison result between the value of the evaluation function that input the output of the plant for the input of the sample command of and the value of the evaluation function that input the output of the plant for the input of the candidate command, the optimum command is the first sample command. It may be determined whether or not it exists in the search range between the second sample command and the second sample command. Furthermore, the first sample command is X ₁ , the second sample command is X ₂ , the value of the evaluation function that inputs the output of the plant to the input of the first sample command is Q (X ₁ ), and the second sample command is Assuming that the value of the evaluation function inputting the output of the plant with respect to the input of is Q (X ₂ ), the command generator adopts a command on the X-axis and an evaluation function on the Y-axis in the process of changing the search range. The search range indicated value is the value of the intersection of the straight line passing through the coordinates (X ₁ , Q (X ₁ )) and the coordinates (X ₂ , Q (X ₂ )) on the graph and the X axis as the search range indicated value. A new first sample command and a second sample command may be obtained based on.

In addition, when the process of changing the search range cannot be performed, the command generator gives commands to the plant in order, in which the command of the value is changed stepwise from the minimum value to the maximum value among the commands that can be given to the plant. , A command in which the output of the plant is the closest to the target output may be adopted as the optimum command. With the command generator configured in this way, among the commands that can be input to the plant, the optimum command for causing the plant to output the target output can be obtained.

Further, the command generation method of the present invention disperses the target waveform in predetermined time units to obtain a target output in each predetermined time unit, and an optimum command suitable for causing the plant to output the target output for each target output. It is provided with an optimum command exploration step required for the above and an input command generation step for generating an input command based on the optimum command for each target output. In the command generation method configured in this way, even if the plant has a response delay or non-linear characteristics, the optimum command is optimal for the plant to output the target output for each target output obtained by discretizing the target waveform. Ask for.

Further, in the command generation method, in the optimum command exploration step, the plant is given a first sample command and a second sample command having a value different from that of the first sample command for each target output, and the first sample command is given for each target output. Based on the plant's output to the input of the sample command, the second sample command, the first sample command and the second sample command, and the target output, the optimum command corresponding to the target output is adopted. May be good. In the command generation method configured in this way, since the first sample command and the second sample command are given to the plant to obtain the optimum command, the optimum command can be quickly obtained regardless of the characteristics of the plant.

According to the command generator of the present invention, it is possible to generate an input command capable of ideally operating the plant by accurately matching the output of the plant with the target waveform.

It is a system block diagram of the command generator in one Embodiment. It is a graph which showed an example of a target waveform. It is a graph of the target output which discretized the target waveform. It is a graph explaining an evaluation function. It is a figure explaining the relationship between the evaluation function and a candidate directive. It is a figure explaining the relationship between the optimum command and a search range. It is a flowchart which showed an example of the processing procedure of the command generator in one Embodiment.

Hereinafter, the present invention will be described based on the embodiments shown in the figure. As shown in FIG. 1, in the description of the command generator 1 in one embodiment, the plant is used as the fluid pressure actuator A in the vibration tester, and the fluid pressure actuator A is to output a target waveform (target waveform). The command generator 1 will be described by taking the case of generating an input command as an example.

The command generator 1 of the present embodiment issues an optimum command suitable for outputting the target output for each predetermined time unit obtained by dissociating the target waveform in a predetermined time unit to the fluid pressure actuator A for each target output. The input command is generated based on the optimum command for each target output.

More specifically, the command generator 1 of the present embodiment gives the fluid pressure actuator A a first sample command, a second sample command, and a candidate command for each target output, and selects the optimum command. Part 2, the stroke sensor 3 as a detection unit that detects the stroke displacement that is the output of the fluid pressure actuator A with respect to the input of the first and second sample commands and candidate commands, and the optimum command selected for each target output. It is configured to include an input command generation unit 4 that generates an input command based on the above. Further, the command generation device 1 of the present embodiment also functions as a control device for controlling the fluid pressure actuator A.

On the other hand, as shown in FIG. 1, the fluid pressure actuator A controlled by the command generation device 1 is movably inserted into the cylinder 11 and the cylinder 11, and the inside of the cylinder 11 is the extension side chamber R1 and the compression side chamber R2. The piston 12, the piston rod 13, the pump 14, the tank 15, the extension side chamber R1 and the compression side chamber R2, which are movably inserted into the cylinder 11 and connected to the piston 12, are divided into the pump 14 and the tank 15. A servo valve 16 for controlling the differential pressure between the extension side chamber R1 and the compression side chamber R2 is provided. Further, in the present embodiment, the extension side chamber R1, the compression side chamber R2, and the tank 15 are filled with hydraulic oil as a working fluid in this example. Therefore, in the present embodiment, the fluid pressure actuator A is configured as a direct-acting hydraulic cylinder.

The servo valve 16 has an extension side supply position in which the extension side chamber R1 communicates with the pump 14 and the compression side chamber R2 communicates with the tank 15, and a compression side in which the extension side chamber R1 communicates with the tank 15 and the compression side chamber R2 communicates with the pump 14. The solenoid valve has a supply position and a neutral position in which the extension side chamber R1 and the compression side chamber R2 communicate with both the pump 14 and the tank 15, and is switched and driven by a solenoid. Then, the servo valve 16 switches to each of the above-mentioned positions according to the first and second sample commands, candidate commands, or input commands input from the command generation device 1.

In this example, the pump 14 is driven by the command generation device 1 at a constant rotation speed, and the communication state of the extension side chamber R1 and the compression side chamber R2 by the servo valve 16 to the pump 14 and the tank 15 is determined by the command generation device 1. It is controlled and the differential pressure between the extension side chamber R1 and the compression side chamber R2 is controlled. As described above, in the present embodiment, the command generation device 1 controls the stroke amount of the fluid pressure actuator A by adjusting the amount of current supplied to the solenoid of the servo valve 16. In the present embodiment, the command generation device 1 includes a pump control unit 5 in addition to the above-described configuration in order to control the pump 14 to be driven at a constant rotation speed.

Subsequently, the stroke sensor 3 as a detection unit is built in the fluid pressure actuator A in the present embodiment. As the stroke sensor 3, various stroke sensors such as a magnetostrictive type can be used. The stroke sensor 3 inputs the detected stroke displacement Y of the fluid pressure actuator A to the optimum command search unit 2.

The optimum command exploration unit 2 disperses the target waveform in a predetermined time unit and obtains an optimum command suitable for outputting the target output for each predetermined time unit to the fluid pressure actuator A for each target output. The first and second sample commands and candidate commands are input to the fluid pressure actuator A. The first and second sample commands and candidate commands input to the fluid pressure actuator A by the optimum command search unit 2 are commands instructing the fluid pressure actuator A to make a displacement stroke. Here, considering the case where the test piece is loaded with a sinusoidal vibration having a frequency of 1 Hz by the vibration tester, the command generator 1 needs to cause the fluid pressure actuator A to output the displacement stroke of the sinusoidal waveform. Therefore, in this case, the target waveform is a sinusoidal waveform as shown in FIG. When this target waveform is dissociated in an arbitrary predetermined time unit, as shown in FIG. 3, since the frequency is 1 Hz, the target waveform for 1 second is divided in a predetermined time unit, and the target waveform for each predetermined time unit is instructed. A discontinuous command group consisting of indicated values is obtained. Each indicated value obtained for each predetermined time unit indicates a target output to be output by the fluid pressure actuator A for each predetermined time unit. Therefore, in order to discretize the target waveform and obtain the target output, it is sufficient to perform a process (discretization step) of collecting the values specified by the target waveform at predetermined time intervals. In this example, the predetermined time unit is 0.05 seconds, and the target waveform, which is a set of indicated values for one second, is discretized in the predetermined time unit and divided into 20 target outputs. The predetermined time unit may be set to match the control cycle of the fluid pressure actuator A, for example, but is not limited to this, and may be set to be suitable for control.

Then, the optimum command exploration unit 2 issues a command that the fluid pressure actuator A, which is a plant, can output a displacement stroke instructed by the target output for each predetermined time unit obtained by temporally dissociating the target waveform in a predetermined time unit. The first and second sample directives are input to the plant for exploration.

In this way, the target waveform is separated in time to obtain the target output, and the first and second sample commands, which are impulse signals, are input to the fluid pressure actuator A, and the output of the fluid pressure actuator A becomes the target output. If such a command is obtained, the command becomes the optimum command for the fluid pressure actuator A to output the target output.

Therefore, the optimum command exploration unit 2 monitors the displacement stroke Y, which is the output of the fluid pressure actuator A when the first and second sample commands are input to the fluid pressure actuator A, seeks a candidate command, and obtains a candidate command. However, if the fluid pressure actuator A is suitable for outputting the target output, this candidate command is adopted as the optimum command. Specifically, for example, when the target waveform for 1 second is discreteized and 20 target outputs can be obtained, the optimum command exploration unit 2 obtains the optimum command corresponding to the first target output. Input the first and second sample commands to search for the optimum command of the first target output. The optimum command search unit 2 obtains a candidate command from the first and second sample commands input to the fluid pressure actuator A in order to search for the optimum command of the first target output. When inputting a candidate command for searching for the optimum command of the second target output, the optimum command search unit 2 has already obtained the optimum command for the first target output, so first input the optimum command. Then, after the lapse of the predetermined time, a candidate command for searching for the optimum command of the second target output is input. Then, the optimum command exploration unit 2 inputs a candidate command to the fluid pressure actuator A, determines whether the candidate command is suitable for the fluid pressure actuator A to output the target output, and if it is suitable, this candidate command. Is adopted as the optimum command.

Optimal command The search unit 2 searches for the optimum command of the second target output in order to obtain the optimum command corresponding to the second target output when the optimum command suitable for outputting the first target output is adopted. Enter the first and second sample directives for. When the optimum command search unit 2 inputs the first and second sample commands for searching the optimum command for the second target output, the optimum command for the first target output has already been obtained, so first. , Then input the first and second sample commands for searching the optimum command of the second target output after the lapse of the predetermined time.

The optimum command search unit 2 outputs the second target output from the first and second sample commands input to the fluid pressure actuator A in order to search for the optimum command of the second target output. Seek a candidate directive suitable for. When inputting a candidate command for searching for the optimum command of the second target output, the optimum command search unit 2 has already obtained the optimum command for the first target output, so first input the optimum command. Then, after the lapse of the predetermined time, a candidate command for searching for the optimum command of the second target output is input. Then, the optimum command exploration unit 2 determines whether the candidate command is suitable for the fluid pressure actuator A to output the target output, and if the candidate command is suitable for the fluid pressure actuator A to output the target output, the optimum command search unit 2 determines. This candidate command is adopted as the optimum command for the second target output.

When the optimum command search unit 2 inputs the first and second sample commands or candidate commands in order to obtain the optimum command corresponding to the nth target output (n is a positive integer of 2 or more), n-. The optimum commands for the first target output are input in order at predetermined time intervals, and after the input is completed and the predetermined time elapses, the first and second sample commands or candidate commands are input. That is, when the optimum command exploration unit 2 inputs the first and second sample commands or candidate commands to the fluid pressure actuator A, if the optimum command exploration unit 2 has an optimum command that has been adopted so far, the optimum command is issued. After inputting, the first and second sample commands or candidate commands are input to the fluid pressure actuator A. After that, the above procedure is repeated, and the input of the first and second sample commands or candidate commands and the adoption of the optimum command are repeated in sequence, and when the optimum commands for the number of target outputs obtained by dispersal are obtained, The input command generation unit 4 uses a set of data sets obtained by arranging the values of these optimum commands in the order in which they are obtained as input commands.

In this way, when there is an optimum command that has already been adopted by the optimum command exploration unit 2, if the first and second sample commands or candidate commands are input to the fluid pressure actuator A after the optimum command is input, the first and second samples commands are input. The optimum command can be obtained by taking into account the influence of the optimum command input before the input of the sample command or the candidate command.

Next, the optimum command exploration unit 2 will be described in more detail. The optimum command exploration unit 2 executes the command optimization process, which is the process of optimizing the first and second sample commands in obtaining the optimum command, and inputs the first and second samples to the fluid pressure actuator A. Search for the range that includes the optimum directive to be adopted based on the sample directive or candidate directive of. That is, the optimum command exploration unit 2 inputs the first and second sample commands or candidate commands, and optimizes the first and second sample commands or candidate commands so as to be optimal for adopting the optimum command. Perform processing.

In order to obtain the optimum command corresponding to one target output, the optimum command exploration unit 2 first obtains two first and second sample commands X ₁ , X ₂ (where X ₂ > X ₁ ) having different values. ). The first sample command X ₁ is a command having a value that is considered to be output by the fluid pressure actuator A with a value smaller than the value specified by the target output. The second sample command X ₂ is a command having a value that is considered to be output by the fluid pressure actuator A in a value larger than the value specified by the target output. If the optimum command exploration unit 2 is made to grasp the relationship between the command value and the output of the fluid pressure actuator A in advance, the optimum command exploration unit 2 will have two large and small first and second samples with respect to the target output. Directives X ₁ and X ₂ are easily obtained.

The optimum command exploration unit 2 has displacement strokes Y ₁ and Y, which are outputs when the fluid pressure actuator A is driven with respect to the first and second sample commands X ₁ and X ₂ having two different values initially input. Based on ₂ , it is determined whether or not the optimum command exists in the range between the first sample command X ₁ and the second sample command X ₂ .

The optimum command exploration unit 2 uses an evaluation function Q (X) with the target output as U and the output (displacement stroke) Y with respect to the input of the sample command X and the target output U as parameters in the determination. The evaluation function Q (X) is a function for evaluating the degree of coincidence between the output Y of the fluid pressure actuator A and the target output U when the command X is input to the fluid pressure actuator A. The evaluation function Q (X) is a function that takes a value closer to 0 as the degree of coincidence between the output Y of the fluid pressure actuator A and the target output U increases. Therefore, the evaluation function Q (X) takes the minimum value for a sample command of a certain value. On the other hand, the farther the value of the sample command is from the value of the sample command that takes the minimum value, the larger the value of the evaluation function Q (X) becomes. Since the output (displacement stroke) Y is the output of the fluid pressure actuator A with respect to the input of the sample command X, it can be expressed by the transfer function from the sample command X to the output Y. Therefore, the evaluation function Q (X) is the sample command X. Can be expressed by a function with. Therefore, the evaluation function Q (X) is expressed by incorporating the transfer function from the input of the sample command X to the output Y of the fluid pressure actuator A, and the input X is used as the parameter of the evaluation function Q (X) instead of the output Y. May be good. However, since the fluid pressure actuator A, which is an actual plant, has non-linear characteristics, it is necessary to prepare a number of transfer functions according to the conditions and determine which transfer function to use each time. It doesn't become. Therefore, in the present embodiment, the stroke sensor 3 as the detection unit detects the output (displacement stroke) Y and inputs it to the evaluation function Q (X), which is a simple calculation without using a plurality of transfer functions. Can execute the operation of the evaluation function Q (X) with.

The optimum command exploration unit 2 obtains the values Q (X ₁ ) and Q (X ₂ ) of the evaluation functions Q (X) corresponding to the first and second sample commands X ₁ and X ₂ , and the first and second sample commands X 1 and X 2. It is determined whether or not there is an unknown command having a high degree of coincidence between the output Y and the target _output U in the range between the _second sample commands X1 and X2. It should be noted that the values Q (X ₁ ) and Q (X ₂ ) indicate that they are the values of the evaluation functions Q (X) corresponding to the first and second sample commands X ₁ and X ₂ , and the same applies to the following. Treat to.

As shown in FIG. 4, the evaluation function Q (X) is basically an evaluation function whose value approaches 0 when the deviation between the output Y of the fluid pressure actuator A and the target output U with respect to the input of the sample command X becomes small. It has become. Therefore, the optimum command exploration unit 2 adopts a sample command in which the value of the evaluation function is equal to or less than the threshold value as the optimum command. Therefore, consider the case where the evaluation function Q (X) takes the minimum value with a certain command X _at and the command X _at is the optimum command. When the first sample command X ₁ is smaller than the command X _at and the second sample command X ₂ is larger than the command X _at , as shown in FIG. 4, between the first and second sample commands X ₁ and X ₂ . There is an optimal directive for. Therefore, the optimum command may be searched for in the range between the _first and _second sample commands X1 and X2.

On the other hand, if the command X _at is smaller than the first sample command X ₁ and the second sample command X ₂ , the optimum command is in the range between the first sample command X ₁ and the second sample command X ₂ . It should be in the range shifted to the left side in FIG. Further, when the command X _at is larger than the first sample command X ₁ and the second sample command X ₂ , the optimum command is the range between the first sample command X ₁ and the second sample command X ₂ . It should be in the range shifted to the right in FIG. Therefore, in such a case, it is sufficient to reset the range in which the optimum command is considered to exist and search for the optimum command again in that range. As described above, the command generator 1 of the present embodiment is optimized by using the _first and _second sample commands X1 and X2 as values that define the upper limit and the lower limit of the search range for obtaining the optimum command. Ask for a directive.

Specifically, the optimum command exploration unit 2 is based on the values of the first and second sample commands X ₁ and X ₂ and the values Q (X ₁ ) and Q (X ₂ ), and the first and second samples. Candidate command X ₃ of a certain value within the range between sample commands X ₁ and X ₂ is obtained, and this candidate command X ₃ is input to the fluid pressure actuator A to obtain the value Q (X ₃ ). Then, the optimum command exploration unit 2 compares this value Q (X ₃ ) with the threshold value δ. Specifically, the optimum command exploration unit _{2 calculates X 3 =} Q (X ₁ ) / {Q (X ₁ ) + Q (X ₂ )} × (X ₂ -X ₁ ) + X ₁ . Ask. In this way, when X ₃ is obtained, as shown in FIG. 5, assuming that the evaluation function Q (X) draws a downwardly convex parabola within the search range between the sample commands X ₁ and X ₂ , the candidate command is a candidate command. X ₃ is between the first and second sample directives X ₁ and X ₂ (X ₁ <X ₃ <X ₂ ), and the value of the evaluation function Q (X) is the minimum value or the value near the minimum value. It is expected to be a directive to do.

Then, as a result of this comparison, the optimum command exploration unit 2 determines that the value Q (X ₃ ) of the evaluation function in the candidate command X ₃ is smaller than the threshold value δ, that is, when Q (X ₃ ) <δ, the candidate command X. The deviation between the output _{Y3 of the fluid pressure actuator A and the target output U with respect to the input of 3 becomes} sufficiently small. The threshold value δ can be set arbitrarily, but in the above comparison _, it is set to a numerical value that can be determined so that the deviation between the output _Y3 and the target output U of the fluid pressure actuator A with respect to the input of the candidate command X3 is sufficiently small.

As described above, when the evaluation function Q (X ₃ ) becomes a very small value in the search range defined by the first and second sample directives X ₁ and X ₂ , the fluid pressure actuator A with respect to the input of the candidate directive X ₃ Since the deviation between the output Y ₃ and the target output U of the above is sufficiently small, the candidate command X ₃ is a candidate command sufficient to be adopted as the optimum command. Therefore, in this case, the optimum command exploration unit 2 adopts the candidate command X ₃ as the optimum command.

On the other hand, when the value Q (X ₃ ) of the evaluation function in the candidate command X ₃ is equal to or greater than the threshold value δ, that is, when Q (X ₃ ) ≧ δ, the output Y of the fluid pressure actuator A with respect to the input of the candidate command X _{3 3} and the target output U and the deviation are large, and the candidate command X ₃ is not suitable as the optimum command. Therefore, in this case, the optimum command exploration unit 2 does not adopt the candidate command X ₃ as the optimum command, and the optimum command exploration unit 2 uses the values and values Q (values Q) of the first and second sample commands X ₁ and X ₂ . Based on X ₁ ) and Q (X ₂ ), the search range in which the command X _at , which is the optimum command, is considered to exist is set as follows.

First, in the case of Q (X ₁ ) ≧ Q (X ₃ ) and Q (X ₂ ) ≧ Q (X ₃ ) (Case 1), the first and second sample directives X ₁ and X ₂ are defined. Since the evaluation function Q (X) takes the minimum value within the search range, there is a command X _at that is the optimum command within this search range. Therefore, in this case, the command X _at , which is the optimum command, exists in the search range, but the candidate command X ₃ obtained last time is not the optimum command.

In this case ₁ , the search range is reset by newly obtaining two sample directives X 1'and X ₂ '(where X ₂ '> X ₁ ') having different values in the following three cases. Then, the candidate directive X ₃ ', in which the value of the evaluation function Q (X) is expected to be the minimum value or a value near the minimum value, is obtained again.

In case 1, when Q (X ₁ ) = Q (X ₂ ), the first and second sample directives X 1'and X 2'are issued by X ₁ '= X ₃ -η × (X ₁ '= X ₃ -η ×, respectively. It is calculated by calculating X ₃ -X ₁ ) and X ₂ '= X ₃ + η × (X ₂ -X ₃ ). However, η is an arbitrary value satisfying 0 <η <1. Therefore, in case 1, when Q (X ₁ ) = Q (X ₂ ), the first and second sample directives X ₁ so as to satisfy X ₁ '> X ₁ and X ₂ > X ₂ '. Since', X _2'are determined, the command X _at , which is the optimum command, is searched in the search range in which the search range defined by the first and second sample commands X ₁ and X ₂ is narrowed. When the optimum command search unit ₂ obtains the first and second sample commands X 1'and X 2'in this way, it is within the search range between the _first and _{second sample commands X 1'and X 2 '} . Find the candidate directive X _3'with . Specifically, the optimum command exploration unit 2 sets X _{3'as X 3 '} = Q (X ₁ ') / {Q (X ₁ ') + Q (X ₂ ')} x (X _{2'-X 1 '} . ) + X _1'is calculated. Thus, assuming that the evaluation function Q (X) draws a downwardly convex parabola within the search range between the _first and _{second sample directives X 1'and X 2 '} , when X 3'is obtained. The candidate directive X _3'is between the first and _{second sample directives X 1'and X 2 '} , and it is expected that the value of the evaluation function Q (X) is the minimum value or a value near the minimum value. Next, the optimum command exploration unit 2 inputs the candidate command X _3'to the fluid pressure actuator A to obtain the value Q (X ₃ '). Then, the optimum command exploration unit 2 compares this value Q (X ₃ ') with the threshold value δ. As a result of this comparison, the optimum command exploration unit 2 adopts the candidate command X _3'as the optimum command when Q (X ₃ ') <δ.

In case 1, when Q (X ₁ )> Q (X ₂ ), the directive X _at , which is the optimum command, is the search range defined by the first and second sample directives X ₁ , X ₂ . It is expected to exist in a range biased toward the _second sample command X2 side from the center. Therefore, in this case, the optimum command exploration unit 2 issues the _first and second sample commands X 1'and X ₂ ', respectively, as X ₁ '= X ₃ -η × (X ₂ -X ₃ ) and X ₂ respectively. '= X ₃ + η × (X ₂ -X ₃ ) is calculated and calculated. However, η is an arbitrary value satisfying 0 <η <1. Therefore, in case 1, when Q (X ₁ )> Q (X ₂ ), the optimum command search unit 2 sets the width of η × (X ₂ -X ₃ ) on both sides of X ₃ as the center. Search for the command X _at , which is the optimum command within the search range. When the optimum command search unit ₂ obtains the first and second sample commands X 1'and X 2'in this way, it is within the search range between the _first and _{second sample commands X 1'and X 2 '} . Find the candidate directive X _3'with . Specifically, the optimum command exploration unit 2 sets X _{3'as X 3 '} = Q (X ₁ ') / {Q (X ₁ ') + Q (X ₂ ')} x (X _{2'-X 1 '} . ) + X _1'is calculated. Thus, when X _3'is obtained, the evaluation function Q (X) produces a downwardly convex parabola within the search range between the _first and _{second sample commands X 1'and X 2 '} . Assuming to draw, it is expected that the value of the evaluation function Q (X) is the minimum value or the value near the minimum value between the first and _{second sample commands X 1'and X 2 '} . Next, the optimum command exploration unit 2 inputs the candidate command X _3'to the fluid pressure actuator A to obtain the value Q (X ₃ '). Then, the optimum command exploration unit 2 compares this value Q (X ₃ ') with the threshold value δ. As a result of this comparison, the optimum command exploration unit 2 adopts the candidate command X _3'as the optimum command when Q (X ₃ ') <δ.

In case 1, when Q (X ₁ ) <Q (X ₂ ), the command X _at , which is the optimum command, is first than the center of the search range defined by the sample commands X ₁ and X ₂ . It is expected to exist in a range biased toward the sample command X1 _side . Therefore, in this case, the optimum command exploration unit ₂ issues the sample commands X 1'and X ₂ ', respectively, to X ₁ '= X ₃ -η × (X ₃ -X ₁ ) and X ₂ '= X ₃ + η ×, respectively. Calculated by calculating (X ₃ -X ₁ ). However, η is an arbitrary value satisfying 0 <η <1. Therefore, in case 1, when Q (X ₁ ) <Q (X ₂ ), the optimum command search unit 2 sets the width of η × (X ₃ -X ₁ ) on both sides of X ₃ as the center. Search for the command X _at , which is the optimum command within the search range. When the optimum command search unit ₂ obtains the first and second sample commands X 1'and X 2'in this way, it is within the search range between the _first and _{second sample commands X 1'and X 2 '} . Find the candidate directive X _3'with . Specifically, the optimum command exploration unit 2 sets X _{3'as X 3 '} = Q (X ₁ ') / {Q (X ₁ ') + Q (X ₂ ')} x (X _{2'-X 1 '} . ) + X _1'is calculated. Thus, assuming that the evaluation function Q (X) draws a downwardly convex parabola within the search range between the _first and _{second sample directives X 1'and X 2 '} , when X 3'is obtained. Candidate directive X _3'is between the first and _{second sample directives X 1'and X 2 '} , and it is expected that the value of the evaluation function Q (X) will be the minimum value or the value near the minimum group. Next, the optimum command exploration unit 2 inputs the candidate command X _3'to the fluid pressure actuator A to obtain the value Q (X ₃ '). Then, the optimum command exploration unit 2 compares this value Q (X ₃ ') with the threshold value δ. As a result of this comparison, the optimum command exploration unit 2 adopts the candidate command X _3'as the optimum command when Q (X ₃ ') <δ.

On the other hand, if Q (X ₃ ') ≥ δ in any of the three cases in Case 1, the optimum command exploration unit 2 determines that the candidate command X _3'is not suitable as the optimum command. Therefore, in this case, the optimum command exploration unit 2 does not adopt the candidate command X _3'as the optimum command, and the optimum command exploration unit 2 has Q (X ₁ ') ≧ Q (X ₃ ') and Q ( It is determined whether or not X ₂ ') ≥ Q (X ₃ ') is satisfied. That is, the values Q (X ₁ '), Q (X ₂ '), Q (X ₃ ) of the evaluation function Q (X) for the first and _second sample directives X ₁ ', X 2'and the candidate directive X ₃ '. It is determined whether or not') meets the above-mentioned condition of case 1. As a result, when the above-mentioned values Q (X ₁ '), Q (X ₂ '), and Q (X ₃ ') meet the conditions of Case 1, the first case is according to the above-mentioned three cases of Case 1. Based on the first and second sample directives X 1'and X _{2'and the candidate directive X 3 '} , the _{first and second sample directives X 1 "} and X ₂ " are set to redefine the search range by the above operation. do. Then, the optimum command exploration unit 2 obtains the candidate command X ₃ "from the obtained first and second sample commands X ₁ " and X ₂ ", and the value Q (X ₃ ") is the threshold value as compared with the threshold value δ. Determine if it is less than δ.

In this way, even if the optimum command exploration unit 2 repeats the same processing, the value of the evaluation function Q (X) does not become less than the threshold value δ, and the value Q (X ₃ ) obtained last time and the Q (X ₃ ) obtained this time do not fall. ) Is less than the threshold value δ, the optimum command exploration unit 2 adopts the last candidate command X ₃ as the optimum command. As described above, in the present embodiment, if the condition that the difference between the value Q (X ₃ ) obtained last time and the Q (X ₃ ) obtained this time is less than the threshold value δ is satisfied, the case is satisfied. The process of 1 is completed and the optimum command is adopted. Instead of this, if the value of the evaluation function Q (X) does not fall below the threshold value δ even after repeating the process of case 1 a predetermined number of times, the optimum command search unit ₂ of the candidate command X3 obtained so far. The minimum value may be adopted as the optimum command. The predetermined number of times can be set arbitrarily.

Next, when the value Q (X ₃ ) of the evaluation function in the candidate directive X ₃ is equal to or greater than the threshold value δ, that is, when Q (X ₃ ) ≧ δ, Q (X ₁ ) <Q (X ₃ ). Or if Q (X ₂ ) <Q (X ₃ ) (Case 2), the evaluation function Q (X) is within the search range defined by the first and second sample directives X ₁ and X ₂ . ) Does not have a command to take the minimum value.

In this case 2, it can be considered that X _at <X ₁ or X _at > X ₂ , so that the first and _second sample directives X _1'and X 2'with different values are newly (however, X). ₂ '> X ₁ ') is obtained, the search range is reset, and the candidate directive X ₃ ', in which the value of the evaluation function Q (X) is expected to be the minimum value or a value near the minimum value, is obtained again.

Specifically, the search range instruction value X _REF is obtained in order to specify the search range in which the optimum command exists. When the search range instruction value X _REF takes a command on the X axis and the evaluation function Q (X) on the Y axis, the coordinates (X ₁ , Q (X ₁ )) on the evaluation function Q (X) and the coordinates Let it be the X coordinate of the intersection of the straight line connecting (X ₂ , Q (X ₂ )) and the X axis. If the command X _at , which is the optimum command, exists on the left side of the previous search range, if the search range instruction value X _REF is obtained as described above, there is a possibility that the optimum command exists around this search range instruction value. high. As shown in FIG. 6, for example, when Q (X ₁ ) <Q (X ₃ ) <Q (X ₂ ), the intersection of the above-mentioned straight line (in FIG. 6) and the X axis is searched for the search range indicated value X _REF . Then, it is highly possible that the evaluation function Q (X) becomes the minimum value around the search range indicated value X _REF .

Therefore, in Case 2, the optimum command exploration unit 2 first has X _REF = {Q (X ₂ ) × X ₁ −Q (X ₁ ) × X ₂ } / {Q (X ₁ ) −Q (X ₂ ). The search range instruction value X _REF is obtained by performing the operation of + α}. Note that α is a very small value that satisfies α << 1, so that division can be executed even if Q (X ₁ ) −Q (X ₂ ) becomes 0 in the calculation of the search range indicated value X _REF . However, if it is unnecessary, the α term may be abolished.

Then, the optimum command search unit 2 uses the search range indicated value X _REF obtained in this way, and divides into the following three cases into new first and _{second sample commands X 1'and X 2 '} . Ask for. First, when X _REF > X ₂ , the optimum command search unit 2 sets X 1'as X ₃ and calculates X ₂ '= ₂ × X _REF −X ₃ , respectively, and the first and second samples, respectively. Find the directives X _{1'and X 2 '} . In this case, when the sample command is taken on the X axis and the evaluation function is taken on the Y axis, the search range of the original first and second sample commands X ₁ and X ₂ is shifted in the direction of increasing the value on the X axis. It is considered that there is an optimum command in the range. Therefore, the search range of the new _first and _second sample directives X1'and X2' obtained by the above calculation is relative to the search range of the original _first and _second sample directives X1 and X2. Is shifted to the right on the X-axis. Then, when new first and second sample commands X 1'and X _2'are obtained, candidate command X _3'is within the search range between the _first and _{second sample commands X 1'and X 2 '} . Ask for. Specifically, the optimum command exploration unit 2 sets X _{3'as X 3 '} = Q (X ₁ ') / {Q (X ₁ ') + Q (X ₂ ')} x (X _{2'-X 1 '} . ) + X _1'is calculated. Thus, when X _3'is obtained, the evaluation function Q (X) produces a downwardly convex parabola within the search range between the _first and _{second sample commands X 1'and X 2 '} . Assuming to draw, it is expected that the value of the evaluation function Q (X) is the minimum value or the value near the minimum value between the first and _{second sample commands X 1'and X 2 '} . Next, the optimum command exploration unit 2 inputs the candidate command X _3'to the fluid pressure actuator A to obtain the value Q (X ₃ '). Then, the optimum command exploration unit 2 compares this value Q (X ₃ ') with the threshold value δ. As a result of this comparison, the optimum command exploration unit 2 adopts the candidate command X _3'as the optimum command when Q (X ₃ ') <δ.

Next, when X _REF <X ₁ , the optimum command search unit 2 calculates X ₁ '= ₂ × X _REF −X ₃ , and sets X 2'as X ₃ , and sets the sample commands X _1'and X, respectively. Ask for ₂ '. In this case, when the command is taken on the X-axis and the evaluation function is taken on the Y-axis, the range shifted from the search range of the original _first and _second sample commands X1 and X2 in the direction of reducing the value on the X-axis. It is considered that there is an optimum command for. Therefore, the search range of the new _first and _second sample directives X1'and X2' obtained by the above calculation is relative to the search range of the original _first and _second sample directives X1 and X2. Is shifted to the left on the X-axis. Then, when new first and second sample commands X 1'and X _2'are obtained, candidate command X _3'is within the search range between the _first and _{second sample commands X 1'and X 2 '} . Ask for. Specifically, the optimum command exploration unit 2 sets X _{3'as X 3 '} = Q (X ₁ ') / {Q (X ₁ ') + Q (X ₂ ')} x (X _{2'-X 1 '} . ) + X _1'is calculated. Thus, when X _3'is obtained, the evaluation function Q (X) produces a downwardly convex parabola within the search range between the _first and _{second sample commands X 1'and X 2 '} . Assuming to draw, it is expected that the value of the evaluation function Q (X) is the minimum value or the value near the minimum value between the first and _{second sample commands X 1'and X 2 '} . Next, the optimum command exploration unit 2 inputs the candidate command X3'to the fluid pressure actuator A to obtain the value Q ₍ X3'). Then, the optimum command exploration unit 2 compares this value Q (X ₃ ') with the threshold value δ. As a result of this comparison, the optimum command exploration unit 2 adopts the candidate command X _3'as the optimum command when Q (X ₃ ') <δ. If the search range instruction value X _REF is less than the threshold value δ when the search range instruction value X _REF is obtained, the search range instruction value X _REF is adopted as the optimum command, and the new first and first orders are adopted. It is also possible to omit the process of obtaining the _{second sample directive X 1'and X 2 '} .

On the other hand, in Case 2, when the search range indicated value X _REF is X ₁ ≤ X _REF ≤ X ₂ , the situation is such that an appropriate search range indicated value X _REF cannot be obtained for some reason. In such a case, the optimum command exploration unit 2 performs the optimum command extraction process. In the optimum command extraction process, the optimum command exploration unit 2 gives commands having different values to the fluid pressure actuator A in order, and the output Y of the fluid pressure actuator A is the value closest to the target output U. Is a process that adopts the command as the optimum command. Specifically, when the optimum command extraction process is executed, the optimum command exploration unit 2 sequentially changes the command of the value from the minimum value to the maximum value that can be given to the fluid pressure actuator A. It is given to the fluid pressure actuator A. Therefore, for example, when the minimum value of the command is -1000 and the maximum value is 1000 and the value is changed by 1, the optimum command exploration unit 2 changes the value by 1 in order from the minimum value of -1000 to the maximum. Enter 2001 commands up to the value 1000. If the optimum command has already been adopted before the optimum command extraction process, the optimum command exploration unit 2 inputs a command for obtaining the optimum command following the input of the adopted optimum command. Then, the optimum command exploration unit 2 monitors the output Y of the fluid pressure actuator A with respect to the input of one command, and adopts the command that minimizes the deviation between the output Y and the target output U as the optimum command. That is, the optimum command exploration unit 2 inputs the adopted optimum command and the command for obtaining the optimum command, and performs the process of obtaining the difference between the output Y and the target output U 2001 times, and obtains 2001 units. The command for which the smallest difference is obtained is the optimum command.

On the other hand, except for the case where the optimum command extraction process is performed in the case 2, if Q (X ₃ ') ≥ δ, the optimum command search unit 2 states that the candidate command X 3'is not suitable as the optimum command. judge. Therefore, in this case, the optimum command exploration unit 2 does not adopt the candidate command X3'as the optimum command, and the optimum command exploration unit 2 has Q (X ₁ ') ≧ Q (X ₃ ') and Q (X). ₂ ') ≧ Q (X ₃ ') is determined. That is, the optimum command exploration unit 2 has the values Q (X ₁ ') and Q (X ₂ ') of the evaluation function Q (X) for the first and _second sample commands X ₁ ', X 2'and the candidate command X ₃ '. It is determined which of the above-mentioned cases 1 and 2 the conditions of') and Q (X ₃ ') are satisfied. As a result, when the above-mentioned values Q (X ₁ '), Q (X ₂ '), and Q (X ₃ ') meet the conditions of the case 1, the optimum command search unit 2 performs the above-mentioned case. The process in case 1 is performed. On the other hand, as a result, when the above-mentioned values Q (X ₁ '), Q (X ₂ '), and Q (X ₃ ') meet the conditions of Case 2, the optimum command exploration unit 2 describes the above. The processing in the case of Case 2 is performed.

In this way, the optimum command exploration unit 2 performs the above-mentioned processing (optimum command exploration step) to disperse the target waveform in predetermined time units, and obtains each target output for each predetermined time unit. The optimum command, which is the optimum command for all targets, is adopted, and the process ends when the optimum command is adopted for all target outputs.

The optimum command adopted by the optimum command exploration unit 2 in this way is input to the input command generation unit 4, and when the optimum commands for the number of target outputs obtained by dispersal are obtained, the input command generation unit 4 receives the optimum command. , A set of data sets obtained by arranging the values of these optimum commands in the order in which they are obtained is used as an input command (input command generation step). When the input command is obtained in this way, the output Y of the fluid pressure actuator A is a set of optimum commands that are the optimum commands for outputting the target output U. Therefore, when an input command is input to the fluid pressure actuator A, the output Y of the fluid pressure actuator A accurately matches the target output U, so that the displacement of the fluid pressure actuator A can be controlled to be the displacement of the target waveform.

As an example, the evaluation function Q (X) may be defined by the following mathematical formula. Y (T + b) is the output of the fluid pressure actuator A when a predetermined time b has elapsed after the input of the sample signal X ₃ , and Y (T + b + 1) is the output of the fluid pressure actuator A for a predetermined time ( b + 1) after the input of the sample signal X ₃ . Assuming that the output of the fluid pressure actuator A at the time elapsed, U (T + b) is the target output corresponding to the output Y (T + b), and U (T + b + 1) is the target output corresponding to the output Y (T + b + 1), this implementation is carried out. In the form, the evaluation function Q (X) is defined by Q (X) = | a ₁ × {Y (T + b) -U (T + b)} + a ₂ × {Y (T + b + 1) -U (T + b + 1)} |. .. b is a positive integer arbitrarily determined, a ₁ and a ₂ are coefficients arbitrarily determined, and one of a ₁ and a ₂ is set to 0, and the evaluation function Q (X) is used after the lapse of wasted time. Only one response may be considered.

In general, there is wasted time for the system to output a command input, it takes time to respond, and there is a delay in the system, and when a command is input, the command is output. Affect for a while. As described above, in the fluid pressure actuator A, it takes time to output the command input of the optimum command search unit 2, and the command affects the subsequent output of the fluid pressure actuator A. Therefore, in the present embodiment, the evaluation function Q (X) is defined as described above in consideration of the wasted time until the influence of the input of the command appears on the output and the time when the command affects the output. ing. Therefore, b is a value for setting the wasted time, and if the next output Y and the target output U after the elapse of the wasted time are added to the evaluation function Q (X), the influence of the input on the output. This is to evaluate. Therefore, the coefficients a ₁ and a ₂ may be determined with reference to the time constant of the fluid pressure actuator A, etc., but it can be understood that they are values that set the degree of influence on the output with respect to the input of the command. When the time constant of the fluid pressure actuator A is large, the influence of a longer time than the above-mentioned definition formula may be taken into consideration in the definition formula of the evaluation function Q (X). That is, in the term of a ₂ × {Y (T + b + 1) -U (T + b + 1)}, then the length of time to be considered is an × {Y (T + b + _n ) -U (T + b + n)} (where n = 3, 4, 5 ...) may be added.

If the evaluation function Q (X) is defined by the above equation, the wasted time from the input to the output of the command of the fluid pressure actuator A and the influence of the command on the output can be taken into consideration, and this evaluation function Q (X) can be used. The optimum command adopted by using it is a command that matches the characteristics of the actual system. From the above, the optimum command obtained by using this evaluation function Q (X) can match the output of the fluid pressure actuator A, which is a system, with the target output with high accuracy. The evaluation function Q (X) may be defined by a formula other than the above formula, and the absolute value of the deviation between the output Y of the fluid pressure actuator A and the target output U with respect to the input is simply obtained from the command and used. It may be an evaluation function Q (X).

The processing of the command generation device 1 up to the above will be described with reference to the flowchart shown in FIG. The optimum command exploration unit 2 first discretizes the target waveform in time and obtains a target output at predetermined time intervals (step S1). Subsequently, the optimum command exploration unit 2 obtains the _first and second sample commands X1 and X2 having _two different values from the target output (step S2). Further, the candidate commands X3 _are obtained from the obtained _first and _second sample commands X1 and X2 (step S3).

Subsequently, the optimum command exploration unit 2 determines whether or not the candidate command X ₃ is less than the threshold value δ (step S4). As a result, when the candidate command X ₃ is less than the threshold value δ, the optimum command exploration unit 2 adopts the candidate command X 3 as the optimum command because the candidate command X ₃ is suitable as the optimum command (step S18). .. On the other hand, when the candidate command X ₃ is equal to or greater than the threshold value δ, the optimum command exploration unit 2 does not find the candidate command X ₃ suitable as the optimum command. In this case, when the optimum command search unit 2 requests the candidate command X ₃ more than once in this routine, the Q (X ₃ ) for the previous candidate command X ₃ and the Q (X) for the candidate command X ₃ obtained this time. It is determined whether the deviation ε from ₃ ) is less than the threshold value δ (step S5). As a result, when the deviation between the Q (X ₃ ) with respect to the previous candidate command X ₃ and the Q (X ₃ ) with respect to the candidate command X ₃ obtained this time is less than the threshold value δ, the optimum command search unit 2 has obtained this time. The candidate command X ₃ is set as the optimum command (step S18).

On the other hand, when the deviation between the Q (X ₃ ) with respect to the previous candidate command X ₃ and the Q (X ₃ ) with respect to the candidate command X ₃ obtained this time is the threshold value δ or more, the candidate command X ₃ is not suitable as the optimum command. .. Therefore, the optimum command exploration unit 2 determines whether Q (X ₁ ), Q (X ₂ ), and Q (X ₃ ) correspond to the case 1 or the case 2 by the following procedure.

The optimum command search unit 2 is optimal between the first and second sample commands X ₁ and X ₂ when Q (X ₁ ) ≧ Q (X ₃ ) and Q (X ₂ ) ≧ Q (X ₃ ). It is determined that the case 1 in which the command exists is applicable (step S6). If Q (X ₁ ), Q (X ₂ ) and Q (X ₃ ) meet the conditions of Case 1, the search range is changed in three cases. Therefore, the optimum command exploration unit 2 performs a process of determining which of these three cases is the situation. First, the optimum command exploration unit 2 determines whether Q (X ₁ ) = Q (X ₃ ) (step S7), and if Q (X ₁ ) = Q (X ₃ ), a new one. The _first and _second sample commands X1 and X2 are reset (step S8). In this case, the optimum command search unit 2 calculates X ₁ = X ₃ -η × (X ₃ -X ₁ ) and X ₂ = X ₃ + η × (X ₂ -X ₃ ) to perform the first and second operations. Update sample commands X ₁ and X ₂ , respectively. When the sample command is reset in this way, the optimum command search unit 2 shifts to the process of step S3 and uses the reset _first and _second sample commands X1 and X2 as candidates. _Ask for Directive X3.

If Q (X ₁ ) = Q (X ₃ ), it is determined whether or not Q (X ₁ )> Q (X ₂ ) (step S9). If Q (X ₁ )> Q (X ₂ ), the new first and second sample commands X ₁ and X ₂ are reset (step S10). In this case, the optimum command search unit 2 calculates X ₁ = X ₃ -η × (X ₂ -X ₃ ) and X ₂ = X ₃ + η × (X ₂ -X ₃ ) to perform the first and second operations. Update sample commands X ₁ and X ₂ , respectively. When the sample command is reset in this way, the optimum command search unit 2 shifts to the process of step S3 and uses the reset _first and _second sample commands X1 and X2 as candidates. _Ask for Directive X3.

If Q (X ₁ ) = Q (X ₃ ) and Q (X ₁ )> Q (X ₂ ), then Q (X ₁ ) <Q (X ₂ ), so the new first and second _The sample commands X1 and X2 of the above _are reset (step S11). In this case, the optimum command search unit 2 calculates X ₁ = X ₃ -η × (X ₃ − X ₁ ) and X ₂ = X ₃ + η × (X ₃ − X ₁ ) to perform the first and second operations. Update sample commands X ₁ and X ₂ , respectively. When the sample command is reset in this way _{, the optimum command search unit 2 shifts to the process of step S3 and obtains the candidate command X 3 using the} reset sample commands X1 and X2.

If Q (X ₁ ) ≧ Q (X ₃ ) and Q (X ₂ ) ≧ Q (X ₃ ) are not determined in step S6, the first and second sample commands X ₁ and X ₂ are determined. This corresponds to case 2 in which the optimum command exists in the search range and the range deviated from the search range. Therefore, the optimum command search unit 2 obtains the search range instruction value X _REF for specifying the search range in which the optimum command exists (step S12).

Then, the optimum command search unit 2 performs processing in three cases according to the search range instruction value X _REF . Therefore, the optimum command exploration unit 2 determines whether or not X _REF > X ₂ (step S13). As a result of this judgment, when X _REF > X ₂ , it is considered that the optimum command exists in the range shifted in the direction of a large value from the search range of the original first and second sample commands X ₁ and X ₂ . , The optimum command exploration unit ₂ resets the new _first and second sample commands X1 and X2 (step S14). In this case, the optimum command exploration unit 2 calculates X ₁ = X ₃ and X ₂ = 2 × X _REF −X ₃ to update the first and second sample commands X ₁ and X ₂ . When the first and second sample commands X ₁ and X ₂ are reset in this way, the optimum command search unit 2 shifts to the process of step S3 and resets the sample commands X ₁ and X. Candidate directive _X3 is obtained using ₂ .

On the other hand, when the determination result in step S13 is not X _REF > X ₂ , the optimum command exploration unit 2 determines whether or not X _REF <X ₁ (step S15). As a result of this judgment, when X _REF <X ₁ , it is considered that the optimum command exists in the range shifted in the direction of a small value from the search range of the original first and second sample commands X ₁ and X ₂ . , The optimum command exploration unit ₂ resets the new _first and second sample commands X1 and X2 (step S16). In this case, the optimum command exploration unit 2 calculates X ₁ = 2 × X _REF −X ₃ and X ₂ = X ₃ to update the first and second sample commands X ₁ and X ₂ . When the first and second sample commands X ₁ and X ₂ are reset in this way, the optimum command search unit 2 shifts to the process of step S3 and resets the sample commands X ₁ and X. Candidate directive _X3 is obtained using ₂ .

Further, if it is determined in steps S13 and S15 that neither X _REF > X ₂ nor X _REF <X ₁ is satisfied, then X ₁ ≤ X _REF ≤ X ₂ . In this case, the appropriate search range instruction value X _REF cannot be obtained for some reason. Therefore, in this case, the optimum command exploration unit 2 performs the above-mentioned optimum command extraction process to obtain the optimum command (step S17).

By the above series of processing, the optimum command is adopted for the target output. Then, the command generation device 1 repeatedly executes this process, and when the optimum command is obtained for all the target outputs, the collective data set obtained by arranging the optimum command in the order in which the optimum command is obtained is used as an input command. As described above, the command generation device 1 can sequentially obtain the optimum command and generate an input command by performing the above-mentioned processing.

The optimum command search unit 2 and the input command generation unit 4 in the command generation device 1 may be realized by executing a program for executing the above-mentioned processing by an arithmetic processing unit such as a CPU (Central Processing Unit). ..

As described above, the command generator 1 of the present invention dissects the target waveform to be output to the fluid pressure actuator (plant) A in predetermined time units, and obtains the target output for each predetermined time unit in the fluid pressure actuator (plant). An optimum command suitable for output to A is obtained for each target output, and an input command is generated based on the optimum command for each target output. In the command generator 1 configured in this way, even if the fluid pressure actuator (plant) A has a response delay or non-linear characteristics, the fluid pressure actuator (plant) is obtained for each target output obtained by dissociating the target waveform. A finds the optimum command for outputting the target output. Therefore, if an input command is input to the fluid pressure actuator (plant) A, the output of the fluid pressure actuator (plant) A accurately follows the target waveform. Therefore, according to the command generator 1 of the present invention, even if the fluid pressure actuator (plant) A has a response delay or a non-linear characteristic, the output of the fluid pressure actuator (plant) A is accurately matched to the target waveform to make the fluid fluid. It is possible to generate an input command that can ideally operate the pressure actuator (plant) A.

Further, the command generation device 1 of the present embodiment gives the first sample command X 1 and the second sample command X 2 to the fluid pressure actuator (plant) A for each target output, and gives the first sample command X ₁ and the second sample command X ₂ for each target output. The output of the fluid pressure actuator (plant) A and the target output with respect to the inputs of one sample command X ₁ , the second sample command X ₂ , the first sample command X ₁ and the second sample command X ₂ . The optimum command exploration unit 2 for obtaining the optimum command corresponding to the target output is provided, and an input command is generated based on the optimum command obtained by the optimum command exploration unit 2 for each target output. In the command generator 1 configured in this way, the first sample command X ₁ and the second sample command X ₂ are given to the fluid pressure actuator (plant) A to obtain the optimum command, so that the fluid pressure actuator (plant) The optimum command can be obtained quickly regardless of the characteristics of A. That is, as a method of obtaining the optimum command, as in the case of the optimum command extraction process, all the commands of all values that can be input to the plant by the command generator 1 are input to the fluid pressure actuator (plant) A, and the input commands are input. It is possible to adopt a method in which the command closest to the target output obtained by dissociating the output of the fluid pressure actuator (plant) A is set as the optimum command, but it takes time. On the other hand, the first sample command X ₁ and the second sample command X ₂ are given to the fluid pressure actuator (plant) A, and the output of the fluid pressure actuator (plant) A and the first sample command X ₁ are given. And when the optimum command is obtained based on the second sample command X ₂ and the target output, the command that is likely to be the optimum command can be hit, so that the optimum command can be obtained quickly. Further, since the output when the first sample command X ₁ and the second sample command X ₂ are given to the fluid pressure actuator (plant) A is taken into consideration, the optimum command can be obtained regardless of the plant characteristics.

Further, in the command generation device 1 of the present embodiment, a value is obtained when the deviation between the output of the fluid pressure actuator (plant) A and the target output becomes small with the output of the fluid pressure actuator (plant) A and the target output as parameters. A fluid pressure actuator (plant) for the inputs of the first sample command X _{1, the second sample command X 2, the first sample command X 1 and the} second sample command X ₂ , using an evaluation function that approaches 0. ) A candidate command X ₃ is obtained based on the value of the evaluation function input of the output of A, and a candidate is obtained when the value of the evaluation function input of the output of the fluid pressure actuator (plant) A with respect to the input of the candidate command X ₃ is less than the threshold value. Command _X3 is the optimum command. The command generator 1 configured in this way is a fluid pressure actuator for the inputs of the first sample command X ₁ , the second sample command X ₂ , the first sample command X ₁ and the second sample command X ₂ . (Plant) Since the candidate command X ₃ is obtained based on the value of the evaluation function input from the output of A, the candidate command X ₃ that is likely to be adopted as the optimum command can be obtained. Therefore, in the command generator 1 of the present embodiment, the optimum command can be obtained more quickly regardless of the plant characteristics.

Further, in the command generation device 1 of the present embodiment, the evaluation function is defined as a function in consideration of the wasted time from the input to the output of the command to the fluid pressure actuator (plant) A, so that the actual fluid pressure actuator ( Plant) A can obtain the ideological optimum command. Therefore, in the command generation device 1 of the present embodiment, an input command capable of ideally operating the fluid pressure actuator (plant) A by matching the output of the fluid pressure actuator (plant) A with the target waveform with high accuracy is generated. can.

Further, in the command generation device 1 of the present embodiment, the evaluation function is defined as a function in which the command to the fluid pressure actuator (plant) A considers the influence on the output of the fluid pressure actuator (plant) A after the lapse of wasted time. Therefore, the ideological optimum command can be obtained for the actual fluid pressure actuator (plant) A. Therefore, in the command generation device 1 of the present embodiment, an input command capable of ideally operating the fluid pressure actuator (plant) A by matching the output of the fluid pressure actuator (plant) A with the target waveform with high accuracy is generated. can.

Further, in the command generator 1 of the present embodiment, when the first sample command X ₁ , the second sample command X ₂ or the candidate command X ₃ is input to the target output, the optimum command adopted up to that point is input. If there is, after inputting the adopted optimum command, input the first sample command X ₁ , the second sample command X ₂ or the candidate command X ₃ . In the command generator 1 configured in this way, when the optimum command for a certain target output is obtained, the optimum command that has been adopted up to that point is input, so that the dynamic characteristics of the fluid pressure actuator (plant) A are dynamic. It is possible to obtain the optimum command in consideration of. Therefore, in the command generation device 1 of the embodiment, it is possible to generate an input command capable of ideally operating the fluid pressure actuator (plant) A by matching the output of the fluid pressure actuator (plant) A with the target waveform with high accuracy. ..

Then, in the command generator 1 of the present embodiment, the optimum command is the first sample command X ₁ and the second sample command X 1 based on the first sample command X ₁ , the second sample command X ₂ , and the candidate command X ₃ . It is determined whether or not it exists in the search range between the sample command X ₂ and the sample command X 2, and if it does not exist, a process of changing the search range is performed. Therefore, in the command generation device 1 of the present embodiment, the search range in which the optimum command can exist can be quickly found, and the optimum command can be quickly obtained.

Further, in the command generation device 1 of the present embodiment, the command of the value among the commands that can be given to the fluid pressure actuator (plant) A by the optimum command exploration unit 2 is changed stepwise from the minimum value to the maximum value. The commands may be sequentially given to the fluid pressure actuator (plant) A, and the command whose output of the fluid pressure actuator (plant) A is closest to the target output may be adopted as the optimum command. In the command generation device 1 configured in this way, among the commands that can be input to the fluid pressure actuator (plant) A, the optimum command for causing the fluid pressure actuator (plant) A to output the target output can be obtained.

Further, the command generation method of the present invention is suitable for a dispersal step of discriminating a target waveform in a predetermined time unit to obtain a target output in a predetermined time unit and for causing the fluid pressure actuator (plant) A to output a target output. It includes an optimum command exploration step for obtaining an optimum command for each target output, and an input command generation step for generating an input command based on the optimum command for each target output. In the command generation method configured in this way, even if the fluid pressure actuator (plant) A has a response delay or non-linear characteristics, the fluid pressure actuator (plant) A is obtained for each target output obtained by dissociating the target waveform. Finds the optimum command to output the target output. Therefore, if an input command is input to the fluid pressure actuator (plant) A, the output of the fluid pressure actuator (plant) A accurately follows the target waveform. Therefore, according to the command generator 1 of the present invention, even if the fluid pressure actuator (plant) A has a response delay or a non-linear characteristic, the output of the fluid pressure actuator (plant) A is accurately matched to the target waveform to make the fluid fluid. It is possible to generate an input command that can ideally operate the pressure actuator (plant) A.

Further, in the command generation method of the present embodiment, in the optimum command exploration step, the _first sample command X1 and the _second sample command X2 are given to the fluid pressure actuator (plant) A for each target output, and the target output is obtained. For each, the output of the fluid pressure actuator (plant) A to the inputs of the first sample command X ₁ , the second sample command X ₂ , and the first sample command X ₁ and the second sample command X ₂ . , Adopt the optimum command corresponding to the target output based on the target output. In the command generation method configured in this way, the _first sample command X1 and the _second sample command X2 are given to the fluid pressure actuator (plant) A to obtain the optimum command, so that the fluid pressure actuator (plant) A is obtained. The optimum command can be obtained quickly regardless of the characteristics of.

Further, in the command generation method of the present embodiment, in the optimum command exploration unit step, the command of the value among the commands that can be given to the fluid pressure actuator (plant) A is changed stepwise from the minimum value to the maximum value. The commands may be given to the fluid pressure actuator (plant) A in order, and the command whose output of the fluid pressure actuator (plant) A is closest to the target output may be adopted as the optimum command. In the command generation method configured in this way, among the commands that can be input to the fluid pressure actuator (plant) A, the optimum command for causing the fluid pressure actuator (plant) A to output the target output can be obtained.

The plant is not limited to the fluid pressure actuator A, and the command generator 1 can be used for various plants. Further, the command generation device 1 may be incorporated in a control device that controls the plant.

This concludes the description of the embodiments of the present invention, but the scope of the present invention is not limited to the details themselves shown or described.

1 ... Command generator, 2 ... Optimal command search unit, A ... Fluid pressure actuator (plant)

Claims (10)

The optimum command suitable for outputting the target output for each predetermined time unit obtained by dissociating the target waveform in a predetermined time unit to the plant is obtained for each target output, and the optimum command for each target output is obtained. A command generator that generates the input command to output the target waveform to the plant based on the above.
For each target output, the plant is given a first sample command and a second sample command having a value different from that of the first sample command, and for each target output, the first sample command and the first sample command are given. Optimal command exploration to obtain the optimum command corresponding to the target output based on the output of the plant with respect to the input of the second sample command, the first sample command and the second sample command, and the target output. With a part
The input command is generated based on the optimum command obtained by the optimum command exploration unit.
A command generator characterized by that. The optimum command exploration unit
Using the output of the plant and the target output as parameters, an evaluation function whose value approaches 0 when the deviation between the output of the plant and the target output becomes small is used.
Candidate commands are issued based on the first sample command, the second sample command, and the value of the evaluation function inputting the output of the plant with respect to the input of the first sample command and the second sample command. Ask,
The command generation device according to claim 1 , wherein when the value of the evaluation function inputting the output of the plant with respect to the input of the candidate command becomes less than the threshold value, the candidate command is set as the optimum command. The command generation device according to claim 2 , wherein the evaluation function is a function in consideration of wasted time from input to output of a command to the plant. The command generation device according to claim 3 , wherein the evaluation function is a function in which a command to the plant takes into consideration the influence on the output of the plant after the lapse of the wasted time. When the first sample command, the second sample command, or the candidate command for the target output is input, the optimum command search unit has already adopted the optimum command if there is an optimum command adopted so far. The command generator according to any one of claims 2 to 4 , wherein after inputting a command, the first sample command, the second sample command, or the candidate command is input. Based on the first sample command, the second sample command, and the candidate command, the optimum command search unit searches for the optimum command between the first sample command and the second sample command. The command generator according to any one of claims 2 to 5 , wherein it determines whether or not it exists in the range, and if it does not exist, performs a process of changing the search range. The optimum command exploration unit
The comparison result between the value of the evaluation function inputting the output of the plant with respect to the input of the first sample command and the value of the evaluation function inputting the output of the plant with respect to the input of the candidate command, and the second sample. Based on the comparison result between the value of the evaluation function inputting the output of the plant with respect to the input of the command and the value of the evaluation function inputting the output of the plant with respect to the input of the candidate command, the optimum command is the first. Whether or not it exists in the search range between the sample command of the above and the second sample command of the above.
The command generator according to claim 6. X the first sample directive ₁ , X the second sample directive ₂ , The value of the evaluation function inputting the output of the plant with respect to the input of the first sample command is Q (X). ₁ ), The value of the evaluation function inputting the output of the plant with respect to the input of the second sample command is Q (X). ₂ ) Then
The optimum command exploration unit
In the process of changing the search range
Coordinates on the graph (X) with commands on the X-axis and evaluation functions on the Y-axis ₁ , Q (X) ₁ )) And coordinates (X ₂ , Q (X) ₂ )) And the value of the intersection of the straight line passing through the X-axis as the search range instruction value, and a new first sample command and a second sample command are obtained based on the search range instruction value.
The command generator according to claim 6. When the process of changing the search range cannot be performed, the optimum command search unit sequentially issues commands in which the command of the value is changed stepwise from the minimum value to the maximum value among the commands that can be given to the plant. The command generator according to claim 6 , wherein a command given to the plant and having the output of the plant closest to the target output is adopted as the optimum command. It is a command generation method that generates an input command that causes the plant to output the target waveform.
A discretization step of discretizing the target waveform in predetermined time units to obtain a target output in each predetermined time unit.
An optimum command exploration step for obtaining an optimum command suitable for causing the plant to output the target output for each target output, and
It is provided with an input command generation step for generating the input command based on the optimum command for each target output .
In the optimum command exploration step, the plant is given a first sample command and a second sample command having a value different from that of the first sample command for each target output, and the first sample command is given for each target output. Optimal corresponding to the target output based on the plant's output to the input of the sample command, the second sample command, the first sample command and the second sample command, and the target output. Adopt the directive
A command generation method characterized by that.

Images