JP2003084806A - Device and method for supporting construction of control model - Google Patents

Abstract

PROBLEM TO BE SOLVED: To construct the control model of high performance in a case when a controlled object is controlled on the basis of the control model. SOLUTION: Plural data stored in a database 121 for constructing the control model are divided into plural data groups free from the statistical difference, by a data dividing means 114, and stored in each of divided database 122. The control model is constructed by a model constructing means 115 on the basis of one of the divided data groups, and the constructed control model is evaluated by a model evaluating means 116 on the basis of the other data.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a control system for steel, electric power, general industry, etc., and efficiently constructs a high-precision control model when data is collected from the control target to construct a control model. About the method of doing.

[0002]

2. Description of the Related Art As a conventional technique for collecting data from a control target and using the same on a computer, as disclosed in Japanese Patent Laid-Open No. 6-168222, a means for collecting control information of a control target is provided, and the collected data is stored in an independent memory. There is a method of performing a simulation repeatedly using this data.

As a conventional method for constructing a control model from data obtained from a controlled object, as disclosed in Japanese Patent Laid-Open No. 5-296923, a control amount error obtained from a controlled object and a tuning amount of a model for eliminating the error are controlled. Store some cases of relationships,
There was a method of tuning a control model with a neural network that identified this relationship.

[0004]

The above prior art has the following problems.

The method described in Japanese Patent Application Laid-Open No. 6-168222 considers the collection of data from a control target and the form in which the data is stored in a computer. In addition, no consideration has been given to a method for improving the efficiency of this operation or a method for improving the accuracy of the constructed control model. Furthermore, since it is not considered to compensate for the delay relationship between the input and output data of the control model, if it is necessary to construct the input and output of the control model using data at different times, this must be compensated manually. There was a case that had to be done. Further, since no consideration is given to compressing the stored data in a form that does not hinder the construction of the control model, there has been a problem that the amount of data stored in the computer becomes enormous when data collection is repeated.

In the method described in JP-A-5-296923,
Consideration has been given to improving the accuracy of the control model, but it does not consider how general the data obtained from the control object is.For example, if the data contains noise Has a problem that the accuracy of the constructed control model is reduced. Furthermore, since consideration is not given to the evaluation of the validity of the constructed control model, when the control model is poor in generality, the control model may be used as it is for actual control, so that the control performance may be reduced. Furthermore, no consideration is given to verifying that the stored data has sufficient quantity and quality to construct a general-purpose control model in each part of the input / output area required for control model construction. Therefore, the accuracy of the output with respect to the input in a specific area of the control model may be significantly reduced.

[0007] A first object of the present invention is to provide a control model construction support apparatus which can accurately evaluate the validity of a constructed control model.

[0008] A second object of the present invention is to provide a method for controlling input / output of a control model based on data collected from the control target when the input / output of the control model needs to be constructed using data at different times. It is another object of the present invention to provide a control model construction support apparatus for compensating delay of each input and constructing a database which can be directly applied to construction of a control model.

A third object of the present invention is to provide a system in which, when there is an area where data accumulation is insufficient in each part of an operation area of a control model, this is reported to an operator to thereby improve the local performance of the control model. It is an object of the present invention to provide a control model construction support device that avoids the decrease of the control model.

As a fourth object, when the data obtained from the control object is enormous, the size of the database is optimized in a form in which loss of statistical properties is minimized, and a control model is created using the optimized contents of the database. It is an object of the present invention to provide a control model construction support device for constructing a model.

A fifth object of the present invention is to provide an easy-to-use control model construction support apparatus which provides an environment where an operator can select and execute various control model construction support means in an interactive manner.

[0012]

A first object of the present invention is to separate data collected from a control target as input data and output data to a control model simulating the control target, and to separate the input data and the output data. Data collection means for storing a plurality of data in the first database in association with each other;
A second data division means for dividing a plurality of data stored in the first database into a plurality of data groups which are not statistically separated, and a plurality of databases each storing the separated plurality of data groups; And a model construction means for constructing a control model that simulates the control target based on data stored in the first or second database, and the control model constructed by the model construction means This can be achieved by providing a model evaluation unit for evaluating based on data stored in the first or second database.

A second object of the present invention is to separate data collected from a control target as input data and output data to a control model simulating the control target, and based on a time at which the separated output data is detected. Determining the time at which input data to be associated with the output data is collected, and storing the input data collected at the determined time and the output data in pairs in the first database; This can be achieved by providing means and model construction means for constructing a control model that simulates the control target based on the input data and the output data stored as a pair in the first database.

A third object of the present invention is to provide a control model construction support apparatus which stores a plurality of data collected from a control target in a database and simulates the control target based on the plurality of data stored in the database. The input / output area of the control model is divided into a plurality of partial areas, a plurality of data stored in the database are separated according to the divided partial areas, and as a result of the separation, an area smaller than a predetermined number of data is provided to the operator. This can be achieved by providing a data quantification means for informing.

A fourth object of the present invention is to provide a control model construction support apparatus which stores a plurality of data collected from a control target in a database and simulates the control target based on the plurality of data stored in the database. This can be achieved by determining a combination of a small number of data representing the data from the data stored in the database, and providing data compression means for storing the combination of the small number of data in the database.

A fifth object of the present invention is to separate data collected from a control object as input data and output data to a control model simulating the control object, based on a time at which the separated output data is detected. Determining the time at which input data to be associated with the output data is collected, and storing the input data collected at the determined time and the output data in pairs in the first database; Means, data dividing means for separating a plurality of data stored in the first database into a plurality of data groups which are not statistically separated, and a plurality of databases each storing the separated plurality of data groups. And a control model that simulates the control target based on data stored in the first or second database. Model construction means for constructing, model evaluation means for evaluating the control model constructed by the model construction means based on data stored in the first or second database, and input / output area of the control model Is divided into a plurality of partial areas, a plurality of pieces of data stored in the database are separated according to the divided partial areas, and as a result of the separation, an area smaller than a predetermined number of data is notified to an operator. And determining a combination of a small number of data representing the data from the data stored in the first or second database, and storing the combination of the small number of data in the first or second database. Data collection means, data collection means, data division means,
This can be achieved by providing the model construction means, the model evaluation means, the data quantification means, and the data compression means with management means for independently managing activation and termination.

[0017]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS Embodiments of the present invention will be described below in detail with reference to the drawings.

FIG. 1 shows the configuration of a control model construction support apparatus realized by the present invention. First, the overall configuration will be described, and then details of each unit will be described. Control model construction support device 1
00 denotes arithmetic means 110 in which various tasks are stored,
Task activating means 101 provided in the computing means 110 for selectively activating various computations, and storage means 12 for storing control model construction data, constructed control models, and the like.
0, a communication interface 102 for exchanging signals with the outside
Consists of Further, the control model construction support device 100
Is provided with a man-machine device 140 for input and output by an operator. Further, it is connected to the control device 130 and the I / O 160 via the network 150, and outputs the constructed control model to the control device 130 via the network 150. Further, the control target 170 is controlled via the I / O 160.
Observe the state of, and import necessary data. The control device 130 detects the state of the control target 170 via the I / O 160 and the network 150 and uses the control model constructed by the control model construction support device 100 to control the control target 1.
And outputs a signal for controlling 70. Control target 170
Receives the control signal output from the control device 130 via the I / O 160, and operates according to it.

In the present embodiment, a heating furnace plant of a hot rolling line will be described as an example of the control object 170. Heating furnace 18
0 is a plant that heats the steel sheet to a high temperature before rolling.
A steel plate 190 called a slab is inserted, and the burner 181 is inserted.
After heating at １８184 to raise the temperature to about 1100 ° C., the slab 192 is output as a high-temperature slab 192. The heating furnace is generally composed of about four furnace zones (pre-tropical zone, first heating zone, second heating zone, uniform tropical zone) as shown in the figure. The sensor 171 measures the slab temperature θ ₀ before entering the furnace, and the sensors 172 to 175
Measures the temperatures t _{1 to} t ₄ of the respective furnace zones. In addition, sensor 176 measures the slab temperature θ _out upon exiting the furnace.

Next, the operation of each unit of the control model construction support apparatus 100 will be described in detail. The task activating unit 101 activates and manages a task stored in the arithmetic unit 110 according to information input by the operator from the man-machine device 140. FIG. 2 shows a process executed by the task activation unit 101. Usually it performs a process of detecting whether the input signals via the operator man-machine device 140 in S _2-1. In the present embodiment, an example is shown in which this signal is received by interruption, but periodic detection processing may be performed. Enters the interrupt routine interrupt occurs, interprets the commands of the interrupt in S _2-2, in the computing means 110 activates the task instructed by the operator. In the present embodiment, the arithmetic means 110 includes a data collection means 111, a data quantification means 112, a data compression means 113, a data division means 114, a model construction means 115, and a model evaluation means 11.
6, an example including each task of the model transfer means 117 is shown. Similarly, the storage unit 120 includes a database 121 for storing data collected from the control target 170, a second database 122 for storing divided data, and a control model storage unit 123 for storing a control model constructed using these. Consists of

FIG. 3 shows a processing algorithm of the data collection means 111. The data collection unit 111 is the task activation unit 1
Execution is started according to a start command from 01. First through the communication I / F102 at S _3-1, incorporate as a detection value of the sensor provided to the controlled object 170 a signal corresponding to the input and output of the control model. In this embodiment, θ ₀ , t _{1 to} t ₄ , and θ _out are taken in according to FIG. Hereinafter, the input of the control model is θ ₀ , t _{1 to} t ₄ , the slab thickness B given in advance as a specification, and the output of the control model is θ
Here is an example of _out . It is necessary to reproduce these input / output relationships as a control model. An example of modeling a relationship is shown. For the slab 192 for which θ _out was detected in S _3-2 ,
The time when the slab is inserted into the heating furnace and the time when the slab is heated in each furnace zone are calculated from the in-furnace time of each furnace zone. The heating furnace 180 is composed of four furnace zones, and each in-furnace time from the pre-tropical zone to the solitary zone can be obtained from the speed at which the steel sheet is conveyed and the length of each furnace, and these are defined as ΔT ₁ to ΔT ₄ . If the time when θ _out is detected is T _out ,
The time T _in when the slab is inserted into the heating furnace,

[0022]

[Number 1] T _in = T _out - can be calculated by _{(ΔT 1 + ΔT 2 + ΔT} 3 + ΔT 4) ... ( number 1). Also, the time when the slab was heated in each furnace zone is the pre-tropical T _pre , the first heating zone _{Th 1} , the second heating zone _{Th 2} ,
For each of the tropics T _unif

[0023]

[Number 2] _{T pre = T out - (ΔT} 1/2 + ΔT 2 + ΔT 3 + ΔT 4) ... ( number 2)

[0024]

[Number 3] _{T h1 = T out - (ΔT} 2/2 + ΔT 3 + ΔT 4) ... ( number 3)

[0025]

[Number 4] _{T h2 = T out - (ΔT} 3/2 + ΔT 4) ... ( number 4)

[0026]

[Number 5] _{T unif = T out -ΔT 4/} 2 ... can be approximated by (number 5). In S _3-3, this way time T _in which is calculated, T _pre, in association with the _{T h1,} T _h2, T unif were collected respectively θ _0, t ₁ ~t _4, further collecting slab thickness B The control model construction data is generated as a set of pairs in association with the obtained θ _out . Then, in S _3-4 , the constructed data pair is stored in the first database 121.

FIG. 4 shows a configuration example of the first database 121. Many sets of θ ₀ , t _{1 to} t ₄ , B which are inputs of the control model and θ _out which is an output of the control model are stored.

The data quantification means 112 searches the contents of the first database 121, quantifies statistical properties and data gaps, and notifies the operator. The operator determines whether or not the contents of the first database 121 are sufficient to construct a good control model based on the notified result.

FIG. 5 shows the structure of the data quantification means 112. It is activated by the task activation means 101. Processing method determining means 501, first to n-th plurality of quantifying means 502
To 505. A plurality of quantification means 502 to 505 are provided as necessary, but one may be provided if unnecessary. The processing method determining means 501 selects which of the first to n-th quantifying means 502 to 505 is to be used according to an instruction from the user via the task activating means 101. The calculation result of the selected quantification means is displayed on the man-machine device 140.

FIG. 6 shows an example of the processing of the quantification means. In FIG. 6, the entire region of the control model is divided into n partial regions of regions 1 to n, and the first database 1
21 shows an example in which the number of data stored in 21 is counted and reported to the operator. In _S6-1 , a pair of data is extracted from the first database 121. It is determined whether or not θ _{out of} the data extracted in S _6-2 is smaller than 1090 ° C. If smaller, it adds 1 to the number of data corresponding to the area 1 in S _6-3. If it is higher than 1090 ° C., similarly, θ _out is higher than 1090 ° C. and 110 in S _6-4.
It is determined whether it is 0 ° C. or less. If the determination result is Yes, 1 is added to the number of data corresponding to the area 2 in _S6-5 .
Such operations are sequentially performed, and the first database 121
Calculate how many data pairs belong to each region.
To end the check and post-processing that was performed on a pair of all of the data in the S _6-11.

FIG. 7 shows an example of the quantification result notified by the man-machine device 140. The number of data counted for each area is displayed.
(Add the data of 1100 ℃ <θ _out ≦ 1110 ℃ from the viewpoint of improving the accuracy of the control model) . "To indicate that it is desirable to add data. According to this result, the operator sets the target temperature Θ _out of the slab 192 output from the heating furnace 180 to 1100 ° C. <Θ _out ≦ 1110.
By setting the temperature to ° C., data can be added to improve the accuracy of the control model. In addition to the processing performed by the quantification means 502 to 505, the variances of the parameters θ ₀ , t _{1 to} t ₄ , B, and θ _out are quantified and stored in the first database 121 according to the size. Various methods, such as a method of notifying the degree of separation of existing data, can be considered.

FIG. 8 shows the structure of the data compression means 113. The data compression means 113 includes a first database 121
Is compressed to a predetermined size in such a manner that loss of information is minimized. Here, as an example, the data compression means 1
13 shows an embodiment in the case of comprising a quantization network 801 and a quantization algorithm 802. In this embodiment,
When the data stored in the database 121 shown in FIG. 1 is M, an example will be described in which the data is compressed into p data (p <M) that is well represented. Quantization network 801 is I ₁ ~I _n input (input in the present embodiment is θ _0, t ₁ ~t _4, for B, n = 6) threshold for outputting the input neurons 805 and constants capture It comprises an input layer 803 composed of neurons 806, a quantized neuron layer 804 composed of p quantized neurons 807, and a synapse 808 for transmitting signals between the input layer 803 and the quantized neuron layer 804. The input neuron 805 outputs the value of the input signal as it is,
The quantization neuron 807 outputs the output of the input neuron 805 and the value W of the synapse weight of the synapse 808 connected thereto.
The product of _ij is output according to the following equation.

[0033]

(Equation 6)

The quantization algorithm 802 inputs inputs to the quantization network 801 one after another, and the quantization neuron having the largest output outputs the synapse weight in such a direction as to output a larger value with respect to the corresponding input. Update the value.

FIG. 9 shows an algorithm executed by the quantization algorithm 802. First, in _S9-1 , database 1
The data is extracted from the data 21 and input to the quantization network 801. When compressing the data shown in FIG.
(30,610,1020, ..., 1100), (27,
612, 980,..., 1100) are sequentially input. Next, in S _9-2 , each quantization neuron 807
Is calculated based on (Equation 6), and output values O ₁ to
To calculate the O _p. In S _9-3, among values of O ₁ ~ O _p detects the maximum one. Assuming that O _j is the largest,
The _weights W _{1j to} W _{n + 1 · j} of the synapses connecting the quantization neuron j and the neuron of the input layer 803 are updated. W _{1j to} W _{n · j} which are synapse loads corresponding to the input neuron 805
According to (Equation 7), the threshold neuron 80
For the synapse load W _{n + 1 · j} corresponding to _No. 6, (Equation 8)
Calculates a new value.

[0036]

W _ij = W _ij + α · (I _i −W _ij ) (7) (i = 1,..., N)

[0037]

(Equation 8)

Here, W _ij and W _{n + 1 · j} are the values of the synapse load corresponding to the updated input neuron and threshold neuron, respectively, and α is a constant. The equation for updating the synapse load is expressed by a vector (W _1j ,..., W _nj ) and a vector (I ₁ ,
..., it is sufficient to correspond to the process to increase the similarity of I _n), such update equation are considered also some other (7) (8). In _S9-5 , the end of the process is determined. The termination may be determined by repeating S _{9-1 to} S _9-4 a certain number of times, but all the synapse load update amounts corresponding to the data pairs extracted from the database 121 have become equal to or less than the certain value. It may be determined by the following. Processing burr to S _9-1 If not completed, one after another extract a set of data, and repeats the same processing. When the above processing is completed, the M data pairs stored in the first database 121 can be represented by the synapse loads of the p quantized neurons. Then, in S _9-6 , the synapse weight corresponding to the p quantized neurons,

[0039]

(Equation 9)

By copying the data into the database 121, M data can be represented by a small number of p representative data, and the contents can be compressed. By such processing, the size of the first database 121 can be kept appropriate. In this embodiment, the data compression means 11
3 performs quantization on the data stored in advance in the first database 121 and updates the content by copying the result to the first database 121. A method of directly taking in the quantization result from the communication I / F 102 and outputting the quantization result to the first database 121 may be used.

FIG. 10 shows the processing of the data dividing means 114. The data dividing means 114 stores the first database 12
1 is divided into a plurality of databases that are not statistically separated, and these are stored in the second database 122. In this embodiment, an example in which the database is divided into two databases will be described. First, at S _10-1, rearranging according to the magnitude of the value of the data stored in the first database 121 theta _out. Next, in _S10-2 , the data is arranged and extracted accordingly, and is alternately distributed to the divided database 1 and the divided database 2 of the second database 122. That is, data division is performed in such a manner that odd-numbered data pairs are stored in the divided database and even-numbered data pairs are stored in the divided database 2. Split database 11101
The configuration of the second divided database 122 composed of the divided database 21102 and the divided database 21102 is, for example, in the form shown in FIG. After confirming that the end of the distribution of all the data in the first database 121 in S _10-3, S
_In order to clarify that there is no difference between the statistical properties of the two databases in _10-4 , θ ₀ , t _{1 to} t ₄ , B, θ
_{For out} , statistics are calculated and compared. As a statistic, an average value, a variance, a median value, and the like are obtained for each parameter, and it is determined that there is no difference in statistical properties when the difference between the two databases is within a predetermined range. In step _S10-5 , the comparison result is determined. If there is no difference in statistical properties, it is determined that good data division has been performed.
The process ends. If there is gap is divided into a split database 11101 in S _10-6 database 2110
The data sets stored in the data base 2 and the previously obtained statistics are displayed, and the data is exchanged between the divided database 1 and the divided database 2 by the operator's selection, and the processing after _S10-4 is repeated.

FIG. 12 shows the processing of the model construction means 115. The model construction means 115 includes a first database 12
1, divided database 11101, divided database 21
102, a control model is constructed in which θ ₀ , t _{1 to} t ₄ , and B are input and θ _out is output from the stored data pairs. Various forms of the control model can be considered, such as the use of a multilayer neural network. In this embodiment, a regression model will be described as an example. In _S12-1 of FIG. 12, the input of the control model is determined. Input is θ ₀ , t _{1 to} t ₄ , B
With consideration to be well approximated the relationship theta _out with, for example, _{ln (θ 0), ln (} t 1), ln (t 2), ln
(t ₃ ), (t ₄ ) ² , and B are determined. Next, S _12-2
A control model is constructed as shown in (Equation 10) using the data stored in the database.

[0043]

Equation 10] _{θ out = a 0 + a 1} ln (t 1) + a 2 ln (t 2) + a 3 ln (t 3) + a 4 t 4 2 + a 5 B ... ( Equation 10) where a ₀ ~a ₅ Can be uniquely determined by a widely known linear regression analysis method (for example, “Regression Analysis and Principal Component Analysis” by Nikkagiren). In this embodiment, the construction of the regression model has been described as an example. However, a model is constructed by a multilayer neural network having θ ₀ , t _{1 to} t ₄ , and B as input and θ _out as output.
There are various methods for constructing a control model.

Next, the processing of the model evaluation means 116 will be described.
The model evaluation unit 116 evaluates the likelihood of the control model constructed using the divided database 11101 using the divided database 21102, and conversely, determines the likelihood of the control model constructed using the divided database 21102 in the first database 1101. And evaluates the result, and notifies the operator of the result. First, a model is constructed using the model construction means 115 in _S13-1 . The above processing is the same as in FIG. As a result, for example, a control model such as (Equation 10) can be constructed. In _S13-2 , i is set to 1.
In _S13-3 , the i-th data pair is extracted from the second database 1102, and the input sets θ _{0 to} B are extracted. S
Type them in _13-4 was constructed control model equation, and estimates theta _out of the resulting theta _out as an output, the theta _out of pairs of data extracted difference Delta] E _i, is calculated by equation (11).

[0045]

ΔE _i = | θ _out −θ _out | (Equation 11) The processing of S _{13-3 to} S _13-4 is divided into divided databases 21102.
Repeat until there are no more unprocessed data pairs. S _13-7
Are calculated, and the model error E _total is calculated according to (Equation 12).

[0046]

(Equation 12)

[0047] to determine the end of the process in S _13-8, to return the process to the S _13-1 again. If it is not finished, by changing the input of the control model, S _13-1 ~S _13-7 repeat. As a result,
Model errors E _total for various combinations of inputs can be calculated. Man-machine device 1 displaying this in FIG.
40 shows an example of a screen. As shown in FIG. 14, the model error E _total corresponding to each model is displayed, and the operator can select the control model with the smallest model error. still,
In this case, the model with the smallest E _total may be obtained by the model evaluation means 116, and the one with the smallest E _total may be automatically determined as the control model.

The model transfer means 117 transmits the selected model to the communication I / F and the network 1 in accordance with an instruction from the operator transmitted and received via the task activating means 101.
The data is transferred to the control device 130 via the control unit 50.

Next, as a second embodiment of the present invention, FIG.
The data normalizing means 150 as a function of the arithmetic means 110
1 shows an embodiment with 1. The data normalizing means 1501 is incorporated as needed in parallel with the tasks of the other arithmetic means shown in FIG. In the case where the control model is configured by a multilayer neural network, the input / output data is normally used after being normalized to 0 to 1 or the like. , The process is started. The data normalizing means 1501
The data of each variable stored in the database 121 is normalized to a certain range, and the result is normalized.
02 is stored. The normalized database 1502 is similarly used for model construction, and division into a plurality of databases by the data dividing means 114 is similarly performed.

FIG. 16 shows a processing algorithm of the data normalizing means 1501. In this embodiment, a case where the value is normalized to a range of 0 to 1 is shown. First database 1 in _S16-1
The maximum values _Dmax and _Dmin of 21 variables are calculated. Next, in _S16-2 , the value of this variable is

[0051]

D _i = (D _i −D _min ) / (D _max −D _min ) (Equation 13) The value is updated to a new value. Thus, Di has a minimum value of 0 and a maximum value of 1, and all data have values between 0 and 1. Determining whether there is a variable that is not normalized by S _16-3, repeated S _16-1 to S _16-2 processing until the variable which is not normalized is eliminated in some cases.

In this embodiment, the control model construction support device 10
0 and the control device 130 are configured separately, and the network 150
, But may be integrated. Further, the task of the arithmetic means 110 may be selected as necessary, and it is easy to define a task having another function. In this embodiment, a heating furnace plant is described as an example of the control target 170. However, the present invention can be widely used for other applications in which the relationship between the input and output of the control target is modeled and used for control. In addition, an example in which the vector quantization method is applied as the processing of the data compression means has been described. The same effect can be realized by performing

[0053]

According to the present invention, even if it is necessary to construct the input and output of the control model by using data at different times by providing the data collecting means, it is necessary to perform this operation manually. There is no. Further, by providing the data quantification means, it is possible to immediately know which region of data should be newly collected from the control target in order to construct a good control model. Further, by providing the data dividing means and the model evaluation means, the validity of the constructed control model can be easily verified before using it for actual control.
Also, by providing the data compression means, even if the data stored in the database becomes enormous over time, the size can be kept at an appropriate size. Further, by providing the task starting means, the operator can efficiently select and use various support means provided for constructing the control model. In the embodiment comparing the superiority of the control model, an example was shown in which one of the regression equations having different shapes was clarified.However, the superiority of the multilayer neural nets having different numbers of hidden layers and hidden neurons was compared. , Comparison of regression model and neural network,
The evaluation of how far back the input from the present should be used for prediction in the time-series model can be realized in a similar form.

[Brief description of the drawings]

FIG. 1 is a control model construction support apparatus realized in the present invention.

FIG. 2 is a processing algorithm of a task starting unit.

FIG. 3 is a processing algorithm of a data collection unit.

FIG. 4 is a configuration diagram of a database.

FIG. 5 is a processing configuration diagram of a data quantification unit.

FIG. 6 is a processing algorithm of a data quantification unit.

FIG. 7 is a display example of a man-machine device.

FIG. 8 is a processing configuration diagram of a data compression unit.

FIG. 9 is a processing diagram of a quantization algorithm.

FIG. 10 is a processing algorithm of a data dividing unit.

FIG. 11 is a configuration diagram of a divided database.

FIG. 12 is a processing algorithm of a model construction unit.

FIG. 13 is a processing algorithm of a model evaluation unit.

FIG. 14 is a display example of a man-machine device.

FIG. 15 is a configuration diagram of a control model construction support device provided with data normalization means.

FIG. 16 is a processing algorithm of a data normalizing unit.

[Explanation of symbols]

Reference Signs List 100: Control model construction support device, 101: Task activation means, 111: Data collection means, 112: Data quantification means, 113: Data compression means, 114: Data division means, 115: Model construction means, 116: Model evaluation means 117: model transfer means, 121: first database, 122: second database, 130: control device,
140: man-machine device, 170: control target.

Continuation of front page    (72) Inventor Shohei Fukuoka             5-2-1 Omika-cho, Hitachi City, Ibaraki Prefecture             Hitachi, Ltd. Omika Plant (72) Inventor Masato Yoshida             5-2-1 Omika-cho, Hitachi City, Ibaraki Prefecture             Hitachi, Ltd. Omika Plant (72) Inventor Yoichi Sugita             7-2-1, Omika-cho, Hitachi City, Ibaraki Prefecture             Hitachi, Ltd., Power & Electrical Development Division (72) Inventor Yasuo Morooka             7-2-1, Omika-cho, Hitachi City, Ibaraki Prefecture             Hitachi, Ltd., Power & Electrical Development Division F-term (reference) 5H004 GA40 GB03 GB04 HB01 KD70

Claims (11)

[Claims] 1. A system according to claim 1, wherein data collected from the control target is separated as input data and output data to a control model simulating the control target, and the input data is associated with the output data to store a plurality of data in a first database. Data collection means for storing the plurality of data stored in the first database into a plurality of data groups which are not statistically separated; and storing the plurality of separated data groups respectively. A second database having a plurality of databases to perform, a model construction means for constructing a control model that simulates the control target based on the data stored in the first or second database; Model evaluation for evaluating the constructed control model based on data stored in the first or second database A control model construction support device comprising means. 2. The model construction means constructs a control model that simulates the control target based on a plurality of data stored in one of the plurality of databases of the second database. The evaluation means evaluates the control model based on data stored in another database which is not used by the model construction means among a plurality of databases of the second database. Item 1. The model construction support device according to item 1. 3. The model constructing means constructs a plurality of control models that simulate the control object based on a plurality of data stored in the plurality of databases of the second database. The control model is evaluated based on data stored in another database that is not used by the model construction means among the plurality of databases of the second database, and the plurality of constructed control models are evaluated. The model construction support device according to claim 1, wherein one control model is determined. 4. The data collection means determines a time at which input data to be associated with the output data is collected based on a time at which the separated output data is detected, and at the determined time. 2. The control model construction support apparatus according to claim 1, wherein the collected input data and the output data are stored in pairs in the first database. 5. Separating data collected from a control target as input data and output data to a control model simulating the control target, and separating the output data based on a time at which the separated output data is detected. Data collection means for determining the time at which the input data to be associated is collected, storing the input data and the output data collected at the determined time as a pair in the first database, A control model construction support apparatus comprising: a model construction unit configured to construct a control model that simulates the control target based on the input data and the output data stored as a pair in one database. 6. A first database which separates data collected from a control target as input data and output data to a control model simulating the control target, and associates the input data with the output data to a first database. Data collection means for storing the plurality of data stored in the first database into a plurality of data groups which are not statistically separated; and storing the plurality of separated data groups respectively. A second database having a plurality of databases, and a model construction means for constructing a control model that simulates the control target based on the data stored in the first or second database. Control model construction support device. 7. The data dividing means calculates statistics such as an average value, a variance value, a median value, and a standard deviation for each of the plurality of separated data groups, and sets a difference between the calculated values in advance. 7. The control model construction support apparatus according to claim 6, wherein when the value is equal to or more than the set value, a part of data stored in the second database is exchanged between the plurality of databases. 8. A control model construction support apparatus for storing a plurality of data collected from a control target in a database and simulating the control target based on the plurality of data stored in the database, comprising: Data quantification for dividing an area into a plurality of partial areas, separating a plurality of data stored in the database according to the divided partial areas, and notifying an operator of an area smaller than a predetermined number of data as a result of the separation. A control model construction support device comprising means. 9. A control model construction support device for storing a plurality of data collected from a control target in a database and simulating the control target based on the plurality of data stored in the database, wherein the control model construction support device stores the plurality of data in the database. A control model construction support apparatus, comprising: data compression means for determining a combination of a small number of data representative of the data from the stored data and storing the combination of the small number of data in the database. 10. Separating data collected from a control target as input data and output data to a control model simulating the control target, and separating the output data based on a time at which the separated output data is detected. Data collection means for determining the time at which the input data to be associated is collected, storing the input data and the output data collected at the determined time as a pair in the first database, Data dividing means for separating a plurality of data stored in one database into a plurality of data groups having no statistical separation, and a second database having a plurality of databases respectively storing the plurality of separated data groups And a model for constructing a control model that simulates the control target based on data stored in the first or second database. Construction means; model evaluation means for evaluating the control model constructed by the model construction means based on data stored in the first or second database; and a plurality of input / output areas of the control model. Data quantifying means for dividing into a plurality of partial areas, separating a plurality of data stored in the database according to the divided partial areas and notifying an operator of an area smaller than a predetermined number of data as a result of the separation; Data compression means for determining a combination of a small number of data representing the data from the data stored in the first or second database, and storing the combination of the small number of data in the first or second database The collection means, the data division means, the model construction means, the model evaluation means, and the data Capacity means, control model configuration supporting apparatus being characterized in that a management unit is used to start and end manage independently the data compression means. 11. A method according to claim 1, wherein data collected from the control target is separated as input data and output data to a control model simulating the control target, and the input data is associated with the output data to store a plurality of data in a first database. The plurality of data stored in the first database are separated into a plurality of data groups that are not statistically separated, and the separated plurality of data groups are stored in a second database having a plurality of databases. Respectively, constructing a control model that simulates the control target based on the data stored in the first or second database, and constructing the control model constructed by the model constructing means into the first or the second model. A control model construction support method characterized by performing an evaluation based on data stored in a second database.