JP7155489B2 - Pumping control device and pumping control model learning device - Google Patents

Description

The present invention relates to a pumping control device and a pumping control model learning device.

2. Description of the Related Art Conventionally, there is known a pumping management system that pumps groundwater by automatically controlling the opening degree of an electric valve installed in a pumping well in real time (for example, Patent Document 1).

JP-A-2001-323477

In the technique described in Patent Document 1, the current groundwater level is estimated from the measured pumping amount, and the opening of the motor-operated valve is automatically controlled in real time to eliminate the difference between the current groundwater level and the set water level. In this case, various control functions are used for automatic control of the electric valve.

However, since the amount of groundwater varies depending on the area where the construction site is located and the scale of construction work, pumping control using a preset control function may not be appropriate. For this reason, in addition to the automatic control of the motor-operated valve, the set water level may be manually input many times.

SUMMARY OF THE INVENTION In consideration of the above facts, it is an object of the present invention to appropriately control pumping of groundwater based on data obtained from construction sites.

In order to achieve the above object, a pumping control apparatus of the present invention includes an acquisition unit that acquires information about a groundwater level at a construction site of a target construction work and construction conditions of the target construction work, and Further, the information on the groundwater level and the construction conditions, the information on the groundwater level at the construction site of the construction work for learning, the construction conditions of the construction work for learning, and the information on the pumping control of the groundwater in the construction work for learning A control information generation unit that generates information related to pumping control of groundwater for the target construction work based on a learned model that has been learned in advance from learning data representing the pumping generated by the control information generation unit and a control unit that controls a pump for pumping up groundwater at the construction site of the target construction work according to the control information. As a result, it is possible to appropriately control the pumping of groundwater in construction work based on the data obtained from the construction site.

The information about the groundwater level of the present invention includes the groundwater level and the target water level of the observation well at the construction site, and the construction conditions of the construction work are the number of wells for pumping groundwater at the construction site of the construction work, wells. including at least one of the distance between the It can include a control amount representing at least one of the number of revolutions and the time opening degree of the motor-operated valve. Thereby, the pump for pumping up the groundwater can be appropriately controlled based on the groundwater level and the target water level of the observation well at the construction site.

The learning data of the present invention can be created according to data obtained by a confirmation pumping test at the construction site. As a result, groundwater pumping control can be properly performed based on the data obtained by the confirmation pumping test.

The learning data may be created according to at least one of data obtained when pumping water at the construction site and data obtained when pumping water at a construction site different from the construction site. can. Accordingly, groundwater pumping control can be appropriately performed based on the data obtained when water is pumped up at the target construction site or at a site different from the target construction site.

A pumping control model learning device of the present invention is a learning device that represents information about groundwater level at a construction site of construction work for learning, construction conditions of construction work for learning, and information about pumping control of groundwater for construction work for learning. A model for outputting information on the groundwater level at the construction site of the construction work and information on the pumping control of the groundwater for the construction work from the construction conditions of the construction work based on the data for the construction work, and learning the information on the groundwater level and the above It comprises a learning unit that obtains a learned model that outputs information on the pumping control from construction conditions. It is possible to obtain a trained model that outputs information on the groundwater pumping plan for the construction work according to the conditions regarding the scale of the construction work and the conditions regarding the area of the construction work. As a result, it is possible to obtain a learned model that outputs information on groundwater pumping control for construction work according to the information on the groundwater level at the construction site and the construction conditions.

ADVANTAGE OF THE INVENTION According to this invention, the effect that pumping up of the groundwater in a construction work can be appropriately performed based on the data obtained from a construction site is acquired.

It is an explanatory view for explaining pumping up of groundwater in a construction site. It is a figure which shows an example of the data obtained by a confirmation pumping test. FIG. 4 is a diagram showing an example of the relationship between the pumping flow rate and the amount of water level drop in an observation well; BRIEF DESCRIPTION OF THE DRAWINGS It is a block diagram which shows schematic structure of the pumping control apparatus which concerns on embodiment. It is a figure which shows an example of the data set for learning. It is a figure which shows an example of the trained model used by embodiment. It is a figure which shows an example of the learning processing routine of this embodiment. It is a figure which shows an example of the pumping control processing routine of this embodiment.

BEST MODE FOR CARRYING OUT THE INVENTION Hereinafter, embodiments of the present invention will be described in detail.

<System configuration of pumping control device according to first embodiment>

During root cutting work in construction work, groundwater springs up from the basement of the construction site. In this case, it is necessary to properly pump up groundwater that wells up from the ground.

In order to properly draw up groundwater at a construction site, it is necessary to properly control a pump for drawing up groundwater. In groundwater pumping control, pumping control is often performed using a preset control function. However, since the amount of groundwater varies depending on the area where the construction site is located and the scale of construction work, pumping control using a preset control function may not be appropriate.

Therefore, in this embodiment, pumping control is performed by utilizing data obtained from the construction site. Specifically, learning data is set using data obtained from a construction site, and a trained model is generated based on the learning data. Then, using the obtained learned model, the pumping of groundwater is automatically controlled. As a result, groundwater pumping control can be appropriately performed based on data obtained from the construction site. In addition, when there is little field performance data, analysis data by numerical analysis may be used together.

FIG. 1 shows an explanatory diagram for explaining pumping up of groundwater at a construction site. A shown in FIG. 1 is a plan view of a construction site where root cutting work has been performed. Plan view A shows pumping wells DW1, DW2, DW3 and DW4, observation wells OW1 and OW2, and impermeable wall S installed at the construction site. The pumping well DW1 and the pumping well DW2 are installed with a distance D1 therebetween. Also, the pumping well DW1 and the pumping well DW3 are installed with a distance D2 therebetween. The pumping well DW1 covers the pumping range X1, the pumping well DW2 covers the pumping range X2, the pumping well DW3 covers the pumping range X3, and the pumping well DW4 covers the pumping range X4.

B shown in FIG. 1 is a cross-sectional view showing the cross section of the construction site of the plan view A viewed from the direction C. As shown in FIG. As shown in cross-sectional view B, when root cutting is performed on the ground of the construction site, water permeation pressure represented by arrows in cross-sectional view B is generated, and groundwater flows out according to the water permeation pressure. As shown in the cross-sectional view B, the hydraulic pressure in the vertical direction is generated according to the vertical hydraulic conductivity Kv, which indicates the hydraulic conductivity in the vertical direction. Further, as shown in the cross-sectional view B, the horizontal water permeability pressure is generated according to the horizontal water permeability coefficient Kh, which represents the water permeability coefficient in the horizontal direction.

At the construction site, when the pumping wells DW1, DW2, DW3 and DW4, the observation wells OW1 and OW2, and the impermeable wall S as shown in FIG. 1 are installed, a confirmation pumping test is performed. In the confirmation pumping test, water is pumped by various pumping patterns, and data representing the relationship between the groundwater level and the pumping flow rate (hereinafter also referred to as the drainage flow rate) is obtained. Fig. 2 shows an example of data representing the relationship between groundwater level and drainage flow rate (Source: Architectural Institute of Japan, "Earth stop design guidelines", 4th edition, November 25, 2017, p.302) .

For example, as shown in FIG. 2, the discharge flow rate Q when the pumping wells DW1 and DW4 are operated in a predetermined time interval, and the discharge flow rate Q when the pumping wells DW1 to DW4 are operated in a predetermined time interval A result (upper result in FIG. 2) representing the discharge flow rate Q is obtained. Further, at this time, the groundwater level h of the observation wells OW1 and OW2 when the pumping wells DW1 and DW4 are operated, and the groundwater level h of the observation wells OW1 and OW2 when the pumping wells DW1 to DW4 are operated (results on the lower side of FIG. 2) are acquired. Then, the pump is controlled according to the test results obtained by the confirmation pumping test (the upper result and the lower result in FIG. 2).

For example, as shown in FIG. 3, as the pumping flow rate Q increases, the water level decrease h in the observation well increases. In this case, the appropriate pumping flow rate to be pumped by the pump is the critical flow rate Qx. Pumping with a pumping flow rate greater than the critical flow rate Qx is not appropriate because the speed at which groundwater rises is lower than the speed at which groundwater is pumped, resulting in rapid pumping.

Therefore, in this embodiment, in order to appropriately pump up groundwater at a construction site, learning data is set from data obtained by a confirmation pumping test, and a trained model for controlling a pump for pumping up groundwater is generated. . As a result, groundwater pumping control can be appropriately performed based on data obtained from the construction site.

Specifically, after the pumping wells, observation wells, and impermeable walls are installed, water is pumped by various pumping patterns, and the relationship between the groundwater level and pumping flow rate (drainage flow rate) as shown in Fig. 2 above. Get data that represents the relationship between Then, learning data is set based on the data representing the relationship between the groundwater level and the pumping flow rate, and the model is learned based on the learning data.

FIG. 4 is a block diagram showing an example of the configuration of the pumping control device 100 according to the embodiment of the present invention. The pumping control device 100 can be functionally represented by a configuration including a data reception unit 10, a computer 20, and a pump 30, as shown in FIG.

The data receiving unit 10 receives learning data representing information about the groundwater level at the construction site of the learning construction work, the construction conditions of the learning construction work, and information about the groundwater pumping control of the learning construction work. accept. In the first embodiment, learning data is created according to data obtained from a confirmation pumping test at a construction site. In addition, below, the information regarding a groundwater level is only called "groundwater level information." Further, the information on pumping control of groundwater is simply referred to as "pumping control information".

The data receiving unit 10 also receives the groundwater level information at the construction site of the target construction work and the construction conditions of the target construction work. The data reception unit 10 is implemented by, for example, a keyboard, a mouse, or an input/output device that receives input from an external device.

In this embodiment, a case where the groundwater level information is the groundwater level and target water level of an observation well at a construction site will be described as an example. In this embodiment, the construction conditions include the number of wells, the distance between the wells, the hydraulic conductivity, the information about the impermeable wall (for example, the type of impermeable wall and the length of the impermeable wall), and the An example will be described in which the plane area is the flat area of the Also, a case where the pumping control information is the number of rotations and the pumping flow rate of a pump for pumping groundwater will be described as an example.

It should be noted that the pumping control information in the learning data is created manually in advance according to the test result data obtained by the confirmation pumping test. For example, a construction worker at a construction site may obtain pumping control information at specific groundwater level information (e.g., groundwater level and target water level) based on the relationship between the groundwater level and the target water level in the test results of a confirmation pumping test. (For example, pumping diversion and pump rotation speed) are created as learning data and input to the computer 20 via the data reception unit 10 .

The computer 20 includes a CPU (Central Processing Unit), a ROM (Read Only Memory) that stores programs and the like for realizing each processing routine, a RAM (Random Access Memory) that temporarily stores data, and a memory as storage means. , a network interface, etc. The computer 20 functionally includes an acquisition unit 22, a learning data storage unit 23, a learning unit 24, a trained model storage unit 25, a control information generation unit 26, and a control unit 28. .

Acquisition unit 22 acquires a plurality of pieces of learning data received by data reception unit 10 . The acquisition unit 22 also acquires the groundwater level information and construction conditions of the target construction work received by the data reception unit 10 .

A plurality of pieces of learning data acquired by the acquisition unit 22 are stored in the learning data storage unit 23 . For example, as shown in FIG. 5, groundwater level information, construction conditions, and pumping control information are stored in association with each other.

In the learning data with the actual ID "00001" shown in FIG. Permeability coefficient is "E1", type of impermeable wall is "F1", length of impermeable wall is "G1", and plane area of area surrounded by impermeable wall is "Z1" , the number of revolutions of the pump for pumping up the groundwater is "H1" and the pumping flow rate is "I1". In this case, for example, the types of impermeable walls are represented by numerical values. For example, the type x1 of the impermeable wall is represented by "1", and the type x2 of the impermeable wall is represented by "2".

The learning unit 24 acquires a learned model by learning a model for outputting pumping control information from groundwater level information and construction conditions based on a plurality of pieces of learning data stored in the learning data storage unit 23. . Since the learning data is data obtained from a confirmation pumping test at the target construction site, information about the target construction site is reflected in the trained model.

As shown in FIG. 6, the trained model of this embodiment includes the groundwater level of the observation wells, the target water level of the observation wells, the number of wells, the distance between the wells, the hydraulic conductivity, the length of the impermeable walls, and the The type of water wall and the plane area of the region surrounded by the impermeable wall are used as input data, and the pumping flow rate and pump rotation speed are used as output data. For example, as shown in FIG. 6, a neural network can be used as an example of a model, and deep learning can be used as an example of a learning algorithm.

The learned model storage unit 25 stores the learned model obtained by the learning unit 24 .

The control information generation unit 26 controls the pumping of groundwater for the target construction work based on the groundwater level information and construction conditions acquired by the acquisition unit 22 and the learned model stored in the learned model storage unit 25. Generate information. In this embodiment, as the pumping control information, a control amount representing the pumping flow rate and the number of revolutions of the pump is generated.

As shown in FIG. 6, when the groundwater level information and construction conditions are input to the learned model, pumping control information is output. As a result, information about what kind of pumping control is appropriate is output from the learned model.

From the learned model, as the pump rotation speed in the pumping control information, for example, the probability that the pump rotation speed is A is 60%, the probability that the pump rotation speed is B is 30%, the pump rotation speed Information in the form of 10% probability that is C is output. Similarly, as the pumping flow rate in the pumping control information, for example, the probability that the pumping flow rate is x is 50%, the probability that the pumping flow rate is y is 20%, and the probability that the pumping flow rate is z is 30%. Information in the form of Alternatively, for example, an actual value may be output as the number of rotations of the pump or the pumping flow rate.

The control unit 28 controls the pump 30 for pumping up groundwater at the construction site of the target construction work according to the pumping control information generated by the control information generating unit 26 .

For example, the control unit 28 selects the pump rotation speed with the highest probability from the pumping control information output from the learned model, and controls the pump 30 to achieve the selected rotation speed. Alternatively, the control unit 28 selects the pumping flow rate with the highest probability from the pumping control information output from the learned model, and controls the pump 30 so as to achieve the selected pumping flow rate. Note that the control unit 28 controls the pump 30 according to the fail-safe function, for example, controls the pump 30 so that the groundwater level does not drop too much.

The pump 30 is installed at the construction site of the target construction work, and pumps up the groundwater that wells up at the construction site after the root cutting work is performed. The pump 30 is driven according to a control signal from the control section 28 and pumps up groundwater.

<Action of pumping control device>

Next, the action of the pumping control device 100 will be described. The pumping control device 100 executes a learning process routine and a pumping control process routine.

<Learning processing routine>

When the data reception unit 10 of the pumping control device 100 receives input of a plurality of learning data, the data are stored in the learning data storage unit 23 . Then, when the computer 20 of the pumping control device 100 receives the instruction signal for executing the learning process, it executes the learning process routine shown in FIG.

In step S<b>100 , the learning section 24 acquires a plurality of learning data stored in the learning data storage section 23 .

In step S102, the learning unit 24 learns a model for outputting pumping control information from groundwater level information and construction conditions based on the plurality of learning data acquired in step S100, and outputs the learned model. Generate.

In step S104, the learning unit 24 stores the learned model generated in step S102 in the learned model storage unit 25, and ends the learning processing routine.

<Pumping control processing routine>

When the learned model is stored in the learned model storage unit 25 and the data reception unit 10 receives the target groundwater level information and construction conditions, the pumping control device 100 executes the pumping control processing routine shown in FIG. . The groundwater level in the groundwater level information is the groundwater level in the observation well. Also, the target water level in the groundwater level information is the target water level in the observation well.

In step S<b>200 , the control information generation unit 26 acquires the groundwater level information and construction conditions of the target construction work received by the data reception unit 10 .

In step S<b>202 , the control information generation unit 26 reads the learned model stored in the learned model storage unit 25 .

In step S204, the control information generation unit 26 inputs the groundwater level information and construction conditions of the construction work acquired in step S200 to the learned model read out in step S202, and generates pumping control information. Generate.

In step S206, the control unit 28 controls the pump 30 according to the pumping control information generated in step S204.

The pump 30 pumps up groundwater according to a control signal output from the control section 28 .

As described in detail above, in the first embodiment, the groundwater level information and construction conditions of the target construction work and the learned model learned in advance from the learning data are used to determine the groundwater level of the target construction work. to generate pumping control information for Then, in this embodiment, the pump for pumping groundwater at the construction site of the target construction work is controlled according to the pumping control information. As a result, groundwater pumping control can be appropriately performed based on data obtained from the construction site. In addition, it is possible to appropriately control a pump for pumping up groundwater based on the groundwater level and the set water level of the observation well at the construction site. Moreover, groundwater pumping control can be performed appropriately based on the data obtained by the confirmatory pumping test.

Further, in the first embodiment, based on the learning data, a model for outputting the pumping control information is learned from the groundwater level information and the construction conditions at the construction site of the construction work, and a learned model is obtained. As a result, it is possible to obtain a learned model that outputs groundwater pumping control information for construction work in accordance with groundwater level information and construction conditions at the construction site.

<System configuration of pumping control device according to second embodiment>

Next, a second embodiment will be described. The second embodiment differs from the first embodiment in that learning data is created based on data obtained when water is pumped up at a construction site. Specifically, in the second embodiment, learning data is generated based on data obtained while the pump 30 is operating, and the learned model is sequentially learned. The configuration of the second embodiment is the same as that of the first embodiment, so the same reference numerals are given and the description is omitted.

For example, a construction worker at a construction site acquires groundwater level data in an observation well while the pump 30 is operating, and determines whether the groundwater level in the observation well matches the target water level. to confirm. Then, when the groundwater level in the observation well and the target water level do not match, the person in charge of construction causes the trained model to learn further, and feeds back the trained model.

For example, if the groundwater level information W and the construction condition G are the groundwater level information W and the pump 30 is driven at the rotation speed S, the groundwater level in the observation well and the target water level do not match. Learning data is generated based on the number of revolutions S' at which the groundwater level and the target water level of are matched. Specifically, when it is desirable to drive the pump 30 at the rotation speed S′ when the groundwater level information W and the construction condition G are present, the construction worker uses the groundwater level information W, the construction condition G, and the rotation speed S′. The pumping control information including is created as new learning data.

Then, the person in charge of construction inputs new learning data to the computer 20 via the data receiving section 10 . The data reception unit 10 receives new learning data.

The acquiring unit 22 of the second embodiment acquires new learning data accepted by the data accepting unit 10 and stores it in the learning data storage unit 23 .

The learning unit 24 of the second embodiment relearns the trained model stored in the trained model storage unit 25 based on new learning data stored in the learning data storage unit 23 .

The control information generation unit 26 of the second embodiment reads out the learned model re-learned by the new learning data from the learned model storage unit 25 . Then, the control information generation unit 26 inputs the groundwater level information and construction conditions acquired by the acquisition unit 22 to the re-learned model, and generates pumping control information.

The controller 28 of the second embodiment controls driving of the pump 30 according to the pumping control information generated by the control information generator 26 .

As described in detail above, in the second embodiment, a learned model is learned based on learning data created according to data obtained when water is pumped up at a construction site. As a result, groundwater pumping control can be appropriately performed based on the data obtained when actually pumping water. Also, even if the data obtained from the confirmation pumping test is small, by adding learning data during the period, it is possible to obtain a trained model that performs appropriate pumping control. Therefore, since the learned model is updated one by one, it is possible to appropriately control the pumping of groundwater.

<System configuration of pumping control device according to third embodiment>

Next, a third embodiment will be described. In the third embodiment, learning data is created according to data obtained at a construction site different from the construction site of the target construction work, and a trained model is generated according to the learning data. It differs from the first and second embodiments. The configuration of the third embodiment is the same as that of the first embodiment, so the same reference numerals are used and the description thereof is omitted.

If the construction site conditions of the target construction work are similar to the conditions of other construction sites, it may be possible to properly pump up groundwater by performing similar pumping control.

Therefore, in the third embodiment, learning data is set according to data obtained from a construction site different from the construction site of the target construction work, and based on the learning data obtained from the data of the different construction site, , to generate a trained model.

For example, in the third embodiment, a combination of groundwater level information and construction conditions of a construction site where construction work was performed in the past and pumping control information at the construction site is set as learning data. In this case, for example, the learning data is provided with ground information (for example, stratum structure at a construction site, etc.) and construction scale information (for example, excavation depth, etc.).

The data receiving unit 10 of the third embodiment receives learning data to which ground information and construction scale information have been added, and which is for a construction site different from the target construction site.

The acquisition unit 22 of the third embodiment acquires learning data of different construction sites received by the data receiving unit 10 and stores the learning data in the learning data storage unit 23 .

The learning unit 24 of the second embodiment creates a model for outputting pumping control information from groundwater level information and construction conditions based on learning data for different construction sites stored in the learning data storage unit 23. Let it learn and get a trained model.

When learning a trained model, the learning unit 24 selects learning data of a construction site similar to the construction site of the target construction work, and generates a trained model according to the selected learning data. do. In this case, the learning unit 24 determines whether or not the target construction site is similar to other construction sites based on the ground information and the construction scale information. Then, the learning unit 24 stores the generated learned model in the learned model storage unit 25 .

The control information generation unit 26 of the third embodiment reads the learned model from the learned model storage unit 25 . Then, the control information generation unit 26 inputs the groundwater level information and construction conditions acquired by the acquisition unit 22 to the learned model, and generates pumping control information.

The controller 28 of the third embodiment controls driving of the pump 30 according to the pumping control information generated by the control information generator 26 .

As described in detail above, in the third embodiment, learning data is created according to data obtained when water is pumped up at a construction site different from the construction site of the target construction work. Then, in the third embodiment, a trained model is obtained based on learning data obtained from data of different construction sites. As a result, a learned model can be obtained based on data obtained when actually pumping water at another construction site similar to the target construction site, and groundwater pumping can be appropriately controlled. Also, even if the data obtained from the confirmation pumping test is small, by adding learning data for different construction sites, it is possible to obtain a trained model that performs appropriate pumping control.

The present invention is not limited to the above-described embodiments, and various modifications and applications are possible without departing from the gist of the present invention.

For example, in the above embodiment, a case where a neural network model as an example of a model is learned by deep learning has been described as an example, but the present invention is not limited to this. For example, a model different from the neural network model may be learned by a learning method different from deep learning.

Further, in the above embodiment, the case where the pumping control device performs the learning process and the pumping control process has been described as an example, but it is not limited to this. For example, a system may be configured by a pumping control model learning device that performs learning processing and a pumping control device that performs pumping control processing.

Further, in the above embodiment, the groundwater level of the observation well and the target water level of the observation well are used as an example of the groundwater level information. any information. Similarly, the construction conditions may be any information as long as it relates to the construction conditions of the construction work, and the pumping control information may be any information as long as it is related to pumping control. good too.

In each of the above-described embodiments, the case where the pumping diversion of the pump 30 and the rotation speed of the pump 30 are output as the pumping control information has been described as an example, but the present invention is not limited to this. For example, the time opening degree of the electric valve of the pump 30 may be used as the pumping control information. In this case, the control unit 28 controls the time opening degree of the electric valve.

Further, in the first embodiment, the learning data is created according to the confirmation pumping test, in the second embodiment, the learning data is created according to the data during pumping of groundwater, and in the third embodiment, In the embodiments, the learning data is created according to the data of other construction sites. There may be. For example, a trained model may be generated by combining multiple types of learning data.

Specifically, in the learning data in the third embodiment, if there is no learning data of the construction site corresponding to the ground information and construction scale information similar to the target construction site, the first implementation It is also possible to use the learning data of the morphological confirmation pumping test and the learning data during pumping of groundwater in the second embodiment.

Further, when there are abundant learning data of other construction sites in the third embodiment, after learning the trained model with the learning data of other construction sites, The learned model may be fitted to the target construction site using the learning data from the confirmation pumping test and the learning data during pumping of groundwater in the second embodiment.

Also, a trained model generated for the target construction site can be used at other construction sites. In this case, the number of wells, the distance between wells, the hydraulic conductivity, the type and length of impermeable walls, and the area surrounded by impermeable walls, which are the construction conditions in the input data, Plane area is set as a different variable and applies to other construction sites as well.

In addition, for the number of wells and the like, which are the input data of the learned model, information regarding the layout of the wells may be input together. In this case, information about the arrangement of wells may be read as image data.

In the above description, the program is pre-stored (installed) in a storage unit (not shown), but the program may be recorded on any recording medium such as a CD-ROM, DVD-ROM, micro SD card, or the like. It is also possible to provide it in the form provided.

10 data reception unit 20 computer 22 acquisition unit 23 learning data storage unit 24 learning unit 25 learned model storage unit 26 control information generation unit 28 control unit 29 control information generation unit 30 pump 100 pumping control device

Claims (6)

Information about the groundwater level at the construction site of the target construction work and the construction conditions of the target construction work, which include at least one of the distance between wells and the flat area of the area surrounded by impermeable walls an acquisition unit that acquires
The information on the groundwater level and the construction conditions acquired by the acquisition unit, the information on the groundwater level at the construction site of the construction work for learning and the construction conditions for the construction work for learning, and the distance between the wells and a learned model pre-learned from learning data representing construction conditions including at least one of the flat area of the area surrounded by the impermeable wall and information on groundwater pumping control of the construction work for learning a control information generating unit that generates information on pumping control of groundwater for the construction work of the object, based on
a control unit for controlling a pump for pumping up groundwater at a construction site of the target construction work according to the information about the pumping control generated by the control information generating unit;
pumping control equipment including; The information about the groundwater level includes the groundwater level and the target water level of the observation well at the construction site,
The construction conditions of the construction work further include at least one of information on the number of wells for pumping groundwater , hydraulic conductivity , and impermeable walls at the construction site of the construction work,
The information on the pumping control includes a control amount representing at least one of the pumping flow rate, the rotation speed of the pump for pumping the groundwater, and the time opening degree of the electric valve,
The pumping control device according to claim 1. The learning data is created according to data obtained by a confirmation pumping test at the construction site,
The pumping control device according to claim 1 or 2. The learning data is created according to at least one of data obtained when pumping water at the construction site and data obtained when pumping water at a construction site different from the construction site.
The pumping control device according to claim 1 or 2. The trained model is learned based on learning data of another construction site different from the construction site of the target construction work, and then according to data obtained by a confirmation pumping test at the construction site of the target construction work. and learning data created according to the data obtained during pumping of groundwater at the construction site of the target construction work,
The pumping control device according to claim 1 or 2. Information on the groundwater level at the construction site of the construction work for learning and construction conditions for the construction work for learning, which include at least one of the distance between wells and the flat area of the area surrounded by impermeable walls Information on the groundwater level at the construction site of the construction work and the construction conditions of the construction work, based on the learning data representing the conditions and the information on the pumping control of the groundwater in the construction work for learning, and the distance between the wells and the flat area of the area surrounded by the impermeable wall, learning a model for outputting information on groundwater pumping control for construction work from construction conditions including at least one of the information on the groundwater level and the construction conditions a pumping control model learning device including a learning unit that obtains a trained model that outputs information on the pumping control from a learning unit.

Images