JP2021036103A - Quality prediction method of improvement body - Google Patents

Abstract

To provide a method capable of predicting quality of an improvement body without waiting for a confirmation result of quality of the improvement body by boring after ground improvement.SOLUTION: There is provided a quality prediction method of an improvement body for predicting quality of the improvement body using a prediction model in a mechanical agitation type deep mixing and treating method. The prediction model is created by learning by a neural network using a plurality of pieces of learning data that is a combination of data concerning a method determined before construction, data concerning the foundation to be constructed, data concerning a construction condition at the time of construction, input data for learning including a ground strength constant, and output data for learning composed of a core sampling ratio. The method includes steps of: inputting, to the prediction model, the data concerning the method determined before construction, the data concerning the foundation to be constructed, the data concerning the construction condition at the time of construction, and the input data for prediction including the ground strength constant; outputting the core sampling ratio from the prediction model as the output data for prediction; and predicting quality of the improvement body.SELECTED DRAWING: None

Description

The present invention relates to a method for predicting the quality of an improved product.

As one of the ground improvement methods for deep layers using a solidifying material, a deep layer mixing treatment method by mechanical stirring (hereinafter referred to as "mechanical stirring type deep mixing treatment method") is known. In this method, the solidifying material and the soil to be improved are stirred and mixed by rotating a stirring blade to create a columnar improved body in the ground.
In the mechanical stirring deep layer mixing treatment method, the stirring blade and the improvement target are related to the special conditions such as when a layer with high adhesive strength exists in the ground to be improved with a certain thickness or more and the shape of the stirring blade. A phenomenon called "agitation" may occur in which the soil is united into a lump and rotates. When the attendance occurs, the solidifying material and the soil to be improved become insufficiently mixed, and the quality of the improved body deteriorates.

As a stirring and mixing device in which rotation is unlikely to occur, Patent Document 1 describes that a rotating shaft rotated by a driving device, an excavation head for excavating the ground, and the excavated soil and a solidifying material injected into the soil are agitated. A stirring and mixing device for deep mixing processing, which comprises a stirring blade for mixing and a rotation prevention plate rotatably attached to the rotating shaft to prevent rotation between the excavated soil and the stirring blade. In the above-mentioned rotation prevention plate, a soil mass shaving plate for shaving the soil mass made of the excavated soil that adheres to the rotating shaft and the stirring blade and rotates is detachably attached, and the soil mass shaving plate is attached to the excavation. Described is a stirring and mixing device characterized in that it is formed into a plate thickness that deforms when it hits an obstacle in the soil.
Further, as a device for detecting and monitoring the rotation, Patent Document 2 describes the single shaft and the above in a ground improvement device including a single shaft, a stirring blade, a drilling bit, and a rotation prevention blade. In order to detect the relative rotation of the rotation prevention blade, a sensor including a detected body arranged on the rotation prevention blade and a detection means arranged on the single shaft corresponding to the detected body, and a sensor. An electric wire for transmitting a relative rotation signal due to the relative rotation of the single shaft and the rotation prevention blade detected by the sensor to the upper part of the single shaft, and an electric wire arranged on the upper part of the single shaft to transmit the relative rotation signal. A rotation detection device for a ground improvement wing, which comprises a wireless means for converting into a wireless signal, is described.

JP-A-2018-150775 Japanese Unexamined Patent Publication No. 2017-66795

After checking the quality of the improved body created in the ground after the ground improvement, if the quality of the improved body is found to be poor due to poor mixing such as the occurrence of rotation, repair work is performed and the specified quality is performed. It is necessary to secure. However, when the repair work is carried out, there is a problem that the process control of the entire work is seriously delayed and the economic loss becomes extremely large.
An object of the present invention is to provide a method capable of predicting the quality of an improved body without waiting for the result of confirming the quality of the improved body by boring after the ground improvement.

As a result of diligent studies to solve the above problems, the present inventor has input data for learning (data on the construction method determined before construction, data on the ground to be constructed, data on construction conditions at the time of construction, and data on construction conditions. Prediction input data (construction) is added to the prediction model created by learning by a neural network using multiple training data, which is a combination of ground strength constants) and training output data (consisting of core sampling rate). Input the data related to the construction method specified earlier, the data related to the ground to be constructed, the data related to the construction conditions at the time of construction, and the data including the ground strength constant), and output the core sampling rate from the prediction model. Therefore, it was found that the above object can be achieved by the method of predicting the quality of the improved product, and the present invention has been completed.
That is, the present invention provides the following [1] to [5].
[1] In the mechanical stirring type deep mixing treatment method, it is a method of predicting the quality of the improved body using a prediction model, and the above prediction model relates to data on the construction method determined before construction and the ground to be constructed. Created by learning with a neural network using multiple learning data that is a combination of data, data related to construction conditions at the time of construction, learning input data including ground strength constants, and learning output data consisting of core sampling rates. The data related to the construction method determined before construction, the data related to the ground to be constructed, the data related to the construction conditions at the time of construction, and the input data for prediction including the ground strength constant are input to the above prediction model. , A method for predicting the quality of an improved product, which comprises outputting the core sampling rate as output data for prediction from the above prediction model and predicting the quality of the improved product.

[2] The data on the above-mentioned construction method is at least one selected from the data on the stirring and mixing device and the data obtained in the pre-mixing test, and the data on the above-mentioned ground is the maximum depth for creating the improved body and the target of construction. It is at least one selected from the layer thickness of each layer constituting the ground and the soil quality classification of the ground, and the data on the construction conditions at the time of the above construction are the torque, rotation speed, penetration speed, and pulling up at the time of stirring and mixing. The improved product according to the above [1], which is at least one selected from the speed, the discharge amount of the solidifying material slurry, and the construction depth, and the ground strength constant is at least one selected from the N value and the adhesive strength. Quality prediction method.
[3] If the core sampling rate output from the above prediction model is less than the standard value within the range of 80 to 95% set according to the soil classification of the ground to be constructed, the improved body will be used. The quality prediction method for the improved product according to the above [1] or [2], wherein it is determined that a mixing defect has occurred, and if it is equal to or higher than the above reference value, it is determined that no mixing defect has occurred in the improved product. ..
[4] At least one core obtained from a section of the improved body in which the core collection rate output from the above prediction model is an interval within the range of 0.5 to 2 m and is divided at an arbitrarily determined interval. If at least one of the output core collection rates is less than the above reference value, it is judged that a mixing defect has occurred in the improved product, and all the output core collection rates are output. The quality prediction method for an improved product according to the above [3], wherein if the core sampling rate is equal to or higher than the above-mentioned reference value, it is determined that no mixing defect has occurred in the improved product.
[5] The method for predicting the quality of the improved product according to the above [3] or [4] is performed in the mechanical stirring type deep mixing treatment method, and the core sampling rate output using the above prediction model is used to mix the improved product. When it is determined that a defect has occurred, the mechanical stirring deep layer is subjected to the second stirring and mixing accompanied by the discharge of the solidifying material slurry for the section of the improved body in which the core sampling rate less than the above reference value is obtained. Mixing process method.

According to the quality prediction method of the improved body of the present invention, the quality of the improved body can be quickly predicted at the construction stage without waiting for the confirmation result of the quality of the improved body by boring after the ground improvement.
In addition, since the quality of the improved product after curing can be predicted during construction or before the improved product is cured, when it is predicted that quality defects will occur due to poor mixing, etc. The quality of the improved product can be improved by rapidly and again stirring and mixing the solidifying material slurry (hereinafter, also referred to as “slurry discharge stirring and mixing”). As a result, it is possible to prevent a serious delay in the process management of the entire construction and an extremely large economic loss as compared with the case where the repair work is performed after the ground improvement.

The quality prediction method of the improved body of the present invention is a method of predicting the quality of the improved body using a prediction model in the mechanical stirring type deep mixing treatment method, and the prediction model is data related to the construction method determined before construction. We used a plurality of learning data, which is a combination of data related to the ground to be constructed, data related to construction conditions at the time of construction, learning input data including the ground strength constant, and learning output data consisting of the core sampling rate. It was created by learning with a neural network, and includes data on the construction method determined before construction, data on the ground to be constructed, data on construction conditions at the time of construction, and predictive input data including ground strength constants. , Is input to the created prediction model, and the core sampling rate is output as the prediction output data from the prediction model to predict the quality of the improved product.
Hereinafter, the present invention will be described in detail.

The prediction model used in the present invention includes data on a construction method determined before construction, data on the ground to be constructed, data on construction conditions at the time of construction, input data for learning including ground strength constants, and core sampling. It was created by training with a neural network using a plurality of training data, which is a combination of training output data consisting of rates.

Examples of data related to the construction method determined before construction used as input data for learning include data related to a stirring and mixing device, data obtained in a pre-mixing test, and the like. These may be used individually by 1 type, and may be used in combination of 2 or more type.
Examples of the data regarding the stirring and mixing device include the stirring diameter, the type of the stirring blade, the shape of the stirring blade, and the like in the stirring and mixing device to be used in the deep mixing processing method. These may be used individually by 1 type, and may be used in combination of 2 or more type.
The data obtained in the pre-blending test is actually based on the soil collected from the ground to be improved, the solidifying material, water, and various admixtures that are blended as needed before construction. This data is obtained when preparing a test body and examining the type and amount of solidifying material required to obtain an improved body satisfying the target strength. More specifically, the target strength (for example, the in-situ improved 28-day strength); the target fluidity (for example, the flow value and shear resistance (vane) immediately after stirring and mixing the solidifying material, water, and soil. (Shear), liquid limit test index); type of solidifying material that can satisfy the target strength and target fluidity, and the core sampling rate can meet the target value (for example, 90% or more). Addition amount of solidifying material, adjusted moisture content: {(natural moisture content of soil + mass of added water) / (dry mass of soil)}, type of admixture, admixture addition amount, water addition amount, and Examples include the water ratio of the solidifying material. These may be used individually by 1 type, and may be used in combination of 2 or more type.

Examples of data related to the ground to be constructed, which is used as input data for learning, are the maximum depth for creating an improved body, the layer thickness of each layer constituting the ground to be constructed, and the soil classification of the above ground. (Including soil classification of each layer that constitutes the ground). These may be used individually by 1 type, and may be used in combination of 2 or more type.
It should be noted that the data on the ground does not include the ground strength constant described later.

The data related to the construction conditions at the time of construction, which is used as the input data for learning, is the data obtained at the time of construction.
Specific examples thereof include torque during stirring and mixing, rotation speed, penetration speed, pulling speed, solidifying material slurry discharge amount, and construction depth of the stirring blade. These data are data whose numerical values can be changed as appropriate depending on the construction situation. These may be used individually by 1 type, and may be used in combination of 2 or more type.

Examples of the ground strength constant used as input data for learning include N value, adhesive strength, and the like. These may be used individually by 1 type, and may be used in combination of 2 or more type.
The ground strength constant can be measured by a standard penetration test, a triaxial compression test, a uniaxial compression test, a vane shear test and the like.
By using the data related to the construction method determined before the construction as the input data for learning, the quality of the improved body can be predicted with high accuracy.

In the present specification, the core sampling rate refers to the soil and solidifying material to be improved for a specific section (specific length: also referred to as "unit interval") of the core collected by the sampler. The total length of the proportion of the portion solidified by stirring and mixing is expressed as a percentage.
The core collection rate used as the output data for learning includes the core collection rate of the total length of the improved body finally obtained, and the core collection rate for each unit interval of the entire improved body finally obtained (for example, the obtained improved body). The specific position of the improved body (for example, all core sampling rates of each unit interval divided at intervals of 0.5 to 2 m (for example, 1 m)) from the ground surface. It is possible to use the core sampling rate of one or more unit intervals at a specific length (depth) from the ground surface side of the improved body. These may be appropriately selected according to the purpose.

The neural network used in the present invention is not particularly limited, and commercially available software, customized software, or programmed using open source software can be used.
Further, the neural network may be a hierarchical neural network having an intermediate layer between the input layer and the output layer.
Learning by the neural network includes the above-mentioned data on the construction method determined before construction, data on the ground to be constructed, data on construction conditions at the time of construction, input data for learning including ground strength constants, and core sampling rate. It may be performed according to a conventionally known general learning method of a neural network by using a plurality of learning data which is a combination of learning output data consisting of the above.

As the above-mentioned learning input data and learning output data used as learning data, various data (actual measurement values) obtained when an improved body is actually created in the past may be used. Further, as the learning input data or the learning output data, the presence or absence of the occurrence of the rotation obtained by mounting a detector capable of detecting the rotation on the stirring blade may be used.
The learning data is data that is not directly related to the improved body that predicts the quality in the present invention, and is prepared in advance for the purpose of learning.
The number of learning data (combination of learning input data and learning output data) used for learning is not particularly limited, and may follow a conventionally known general learning method of a neural network.

In addition, a part of a plurality of training data (for example, 20 to 30%) obtained when the improved body was actually created in the past is used as test data for confirming the reliability of the obtained prediction model. You may use it. Specifically, after creating a prediction model, input data for prediction obtained from test data is input to the prediction model, and the core sampling rate (prediction value) obtained from the prediction model and the test data are obtained. By comparing the core sampling rate (actual measurement value), the reliability of the obtained prediction model can be estimated.
Learning by the neural network is performed while appropriately changing the type of data used as training data, the number of training data, the number of times of learning, and the like, and is performed until a highly reliable prediction model is obtained.

In the prediction model obtained by the above learning, data on the construction method determined before construction, data on the ground to be constructed, data on construction conditions at the time of construction, and input data for prediction including ground strength constants are input. By doing so, the core collection rate can be obtained as the output data for prediction from the prediction model.
Forecast input data (data on the construction method determined before construction, data on the ground to be constructed, data on construction conditions at the time of construction, and ground strength constant) and output data for prediction (core sampling rate) Examples thereof include the same as the learning input data and the learning output data described in paragraphs 0011 to 0015.
Further, as the prediction input data input to the prediction model, the same type of training input data as that used when creating the prediction model is used.
The input data for prediction is data obtained when the deep mixing treatment method is actually applied to the ground to be improved.

The core collection rate obtained as the output data for prediction may be one or a plurality. For example, as the output data for prediction, the core collection rate of the total length of the improved body finally obtained and the core collection rate for each unit interval of the entire improved body (for example, the obtained improved body is 0.5 from the ground surface). The interval within the range of ~ 2 m and arbitrarily determined (for example, 1 m) may be output for all core sampling rates of each unit interval divided by the interval), and a specific position (improvement) of the improved body may be output. The core sampling rate of one or more unit intervals at a specific length (depth) from the surface side of the body may be output. These may be appropriately selected according to the purpose.

According to the quality prediction method of the improved body of the present invention, in the mechanical stirring type deep layer mixing treatment method, the prediction model created in advance includes data on the construction method determined before construction, data on the ground to be constructed, and data at the time of construction. By inputting the data related to the construction conditions and the input data for prediction including the ground strength constant, the core sampling rate is output as the output data for prediction from the above prediction model, and the quality of the improved body (for example, mixed) is obtained. Whether or not a defect has occurred) can be predicted quickly with high accuracy.
If the core sampling rate output from the prediction model does not meet the specific reference value (less than the reference value), it can be determined that a mixing defect has occurred in the obtained improved product. Further, when the core collection rate satisfies a specific reference value (or more than the reference value), it can be determined that no mixing defect has occurred in the obtained improved product.

Further, when the output core collection rate is a plurality of core collection rates (for example, the core collection rate obtained from a plurality of arbitrarily determined sections of the improved body), among the output core collection rates, If at least one core sampling rate is less than the standard value, it can be determined that a mixing failure has occurred in the improved body, and if all the output core sampling rates are equal to or higher than the standard value, the improved body It can be determined that no mixing failure has occurred.
In addition, the occurrence of poor mixing means that there are places where the solidifying material and the improved soil are not sufficiently mixed (more specifically, there is a rotation when stirring and mixing). Say.

The specific reference value is preferably a value in the range of 80 to 95%, more preferably 85 to 95%, and is a value set according to the soil classification of the ground to be constructed. Further, the above reference value may be appropriately changed according to the size of the unit interval of the improved body from which the core collection rate can be obtained, the quality requirement of the target improved body, and the like.
When the present invention is used in the field of building foundation (when the ground on which the building is constructed is improved before the building is constructed), the individual columnar improved bodies formed in the ground are almost alone. Since it is likely that a state that supports the load of a building built on the improved ground is likely to be supported and the quality requirements of the improved body are often strict, the above standard value is, for example, the core sampling rate is the total length of the improved body. When obtained from, 95% if the target ground is sandy soil, 90% if the target ground is cohesive soil, loam, organic soil, or highly organic soil. is there. Further, when the core sampling rate is within the range of 0.5 to 2 m and is obtained from an arbitrarily determined (for example, 1 m) unit interval, the above reference value is that the target ground is sandy soil. If the target ground is cohesive soil, loam, organic soil, or highly organic soil, it is 85%.
On the other hand, when the present invention is applied when a plurality of columnar improved bodies existing within the range of ground improvement in the field of civil engineering are combined with the original ground to resist external forces such as earth pressure. Compared with the case of using in the above-mentioned building foundation field, the quality requirement for the improved body is not strict, and the above-mentioned standard value is often relaxed.
In the present specification, the terms "less than the reference value" and "greater than or equal to the reference value" are used for convenience in order to divide them into two categories based on a specific value. , Each can be replaced with the words "below the reference value" and "exceeding the reference value", respectively.

When the above-mentioned quality prediction method for the improved product is performed in the mechanical stirring type deep mixing treatment method and it is determined from the output core sampling rate that a mixing defect has occurred in the improved product, the standard value of the improved product is satisfied. High quality (mixing failure occurs) by performing multiple slurry discharge stirring mixing (stirring mixing accompanied by discharge of solidifying material slurry) in the section where no core sampling rate (for example, less than 80%) was obtained. It is possible to create an improved body (not).
The above-mentioned quality prediction method of the improved body may be performed after the improved body is formed in the ground and before the improved body is completely cured (preferably immediately after the improved body is formed). It may be carried out at any time during the preparation (during stirring and mixing).

In the mechanical stirring type deep mixing treatment method, stirring and mixing accompanied by discharge of the solidifying material slurry is usually performed only when the stirring blade is penetrated or pulled up.
Specifically, when the stirring blade is penetrated to the maximum depth for creating the improved body, when stirring and mixing is performed with the discharge of the solidifying material slurry, when the stirring blade that has reached the maximum depth is pulled up to the ground surface, Although stirring and mixing are performed, the solidifying material slurry is not discharged. Further, in the mechanical stirring high-pressure injection combined construction method, which is one of the mechanical stirring type deep layer mixing treatment methods, although stirring is performed when the stirring blade is penetrated to the depth (maximum depth) for creating the improved body, the solidifying material slurry is used. Is not discharged (injected), but when the stirring blade that has reached the maximum depth is pulled up to the ground surface while stirring and mixing, stirring and mixing is performed with discharge (injection) of the solidifying material slurry.

As a method of performing the above prediction method at any time during the construction of the improved body (during stirring and mixing), the above prediction method is appropriately performed when the construction conditions change at the construction site (for example, when the soil quality of the ground changes). A method of performing the above prediction method periodically while stirring and mixing, and the like.
As a method of periodically performing the above prediction method while stirring and mixing, the stirring blade for stirring and mixing may perform a specific depth (length) every time it penetrates or is pulled up, and a specific time elapses. You may go every time you do.
The specific depth may be arbitrarily determined, for example, a depth corresponding to a unit interval in the core sampling rate (usually, a depth arbitrarily determined from the range of 0.5 to 2 m) or a unit interval in the core sampling rate. A depth shorter than the depth corresponding to (for example, a depth corresponding to 1/10 to 1/2 of the length of the unit interval) and the like can be mentioned. When the specific depth is made shorter than the unit interval, a part of the unit section of the improved body from which the core collection rate is obtained overlaps with a part of the unit section of the improved body from which the other core collection rate is obtained. However, this allows a large amount of data on core uptake at various positions in the improved product.
The specific time may be arbitrarily set, for example, 1 minute.
For the purpose of performing the above-mentioned quality prediction method for the improved product while stirring and mixing, various detection devices are provided in the stirring and mixing device so that data on the construction conditions at the time of construction can be obtained in real time by wire or wirelessly. A computer may be used to perform a quality prediction method for the improved product in the cab of the stirring / mixing device or in a facility in the vicinity.

The plurality of times of slurry discharge stirring mixing is, for example, a stirring blade for stirring and mixing after performing slurry discharge stirring mixing in a section where a core sampling rate that does not satisfy the reference value (for example, less than 80%) is obtained. Is pulled up by the section where the core sampling rate was obtained (for example, 1 m), and then the stirring blade is pulled up by the section where the core sampling rate was obtained (for example, in the mechanical stirring high-pressure injection combined method). It is carried out by 1 m) intrusion, and then (while pulling up again) by discharging, stirring and mixing the slurry. The plurality of times is usually 2 times (2 times in total including the first slurry discharge, stirring and mixing for one section).
When re-penetrating (or pulling up) the stirring blade, the slurry discharge stirring mixing conditions may be changed as appropriate. For example, the discharge amount of the solidifying material slurry may be changed for the purpose of increasing the fluidity at the time of mixing the soil and the solidifying material while satisfying the target strength.
Further, after changing the slurry discharge / stirring / mixing conditions and before performing the slurry discharge / stirring / mixing again under the changed conditions, the changed slurry discharge / stirring / mixing conditions and the like are input to the above prediction model, and the output core sampling rate is obtained. After confirming that the value satisfies the standard value, the penetration (or increase) may be performed again.

Specific example of a method of performing a quality prediction method of an improved product while stirring and mixing, and performing slurry discharge stirring and mixing a plurality of times in a section where a core sampling rate that does not meet the standard value (for example, less than 80%) is obtained. Examples thereof include the following methods (1) to (4).
(1) A method of outputting all the core sampling rates for each unit interval of the entire improved body to be created as the core sampling rate. All the core collection rates for each unit section (for example, the section divided at 1 m intervals from the ground surface) of the entire improved body constructed in 1 are output.
Among the obtained plurality of core collection rates, in the section where the core collection rate that does not meet the reference value (for example, less than 80%) is obtained, mixing failure (for example, attendance) occurs (for example, rotation). Alternatively, it can be determined that it will occur). In the above section, by performing the slurry discharge, stirring and mixing a plurality of times (usually twice), it is possible to eliminate the poorly mixed portion of the obtained structure.
This method is preferable in that it is possible to predict to some extent in advance at what depth the mixing failure will occur.

(2) Method of outputting the core sampling rate of the unit interval each time the slurry discharge, stirring and mixing are performed in the unit interval In the method, the method is used from a specific position (for example, from the ground surface or from an arbitrarily determined depth). Every time the slurry discharge / stirring / mixing is performed in a unit interval (for example, every time the slurry discharge / stirring / mixing is performed in 1 m, in other words, every time the penetration of the stirring blade advances by 1 m), the core sampling rate in the unit interval is output.
If the obtained core sampling rate does not meet the reference value (for example, less than 80%), it is determined that a mixing defect (for example, a circumstance) has occurred in the section where the slurry was discharged, stirred and mixed. it can.
In this case, after stopping the stirring and mixing and pulling up the stirring blade by a unit interval (for example, 1 m), the slurry discharge stirring and mixing is performed while penetrating the stirring blade again.

(3) A method of periodically outputting the core sampling rate of a unit interval when pulling up the stirring blade after the slurry discharge / stirring / mixing is completed. In this method, the slurry discharge / stirring / mixing is performed up to the maximum depth for creating an improved body. After that, when the stirring blade that has reached the maximum depth is pulled up to the ground surface while performing stirring and mixing without discharging the solidifying material slurry, the core sampling rate is output for each unit interval of pulling up.
When the obtained core collection rate does not satisfy the reference value (for example, less than 80%), it can be determined that a mixing defect (for example, a rotation) has occurred in the section where the pulling was performed earlier.
In this case, the pulling is stopped, and the slurry is discharged, stirred and mixed for a unit interval (for example, 1 m) while penetrating the stirring blade again.

(4) Method for periodically outputting the core sampling rate of the unit space starting from the position where stirring and mixing are performed In the method, the unit interval (for example, stirring and mixing is performed) starting from the position where stirring and mixing are performed. The core sampling rate (1 m section from the position toward the direction in which the improved body was created) is output.
In the method (4), the stirring mixture may be a stirring mixture with the discharge of the solidifying material slurry, or a stirring mixing without discharging the solidifying material slurry.
The position of stirring and mixing may be appropriately determined within the range in which stirring and mixing are performed. For example, the position (depth) of the upper end of the stirring blade during mixing and stirring may be used, and the position of the lower end of the stirring blade (depth). ) May be.
When the obtained core sampling rate does not meet the reference value (for example, less than 80%), mixing failure (for example, rotation) occurs in the unit interval starting from the position where stirring and mixing are currently performed. Can be judged.
In this case, the stirring and mixing are stopped, the stirring blade is pulled up (or penetrated) by a unit interval (for example, 1 m), and then the slurry discharge stirring and mixing is performed while penetrating (or pulling up) the stirring blade again.

In the method (4) above, the core sampling rate may be output periodically while stirring and mixing. More specifically, it may be performed at an arbitrarily determined specific time interval or at a specific depth (length) in which the stirring blade for stirring and mixing is arbitrarily determined.
An example of an arbitrarily determined time-by-time example is an arbitrarily determined time-by-hour (for example, one-minute-by-minute) interval between 0.5 minutes and 1 hour.
As an example for each specific depth in which the stirring blade for stirring and mixing is arbitrarily determined, the stirring blade is arbitrarily set between 40 to 120 cm (preferably 60 to 110 cm, more preferably 80 to 100 cm). Every time it penetrates the depth specified in (for example, every 50 cm). Further, the specific depth may be shorter than the depth corresponding to the unit interval in the core sampling rate.

Here, when the core sampling rate is actually measured by boring, the numerical value varies depending on the position of the unit section in which the core sampling rate is obtained in the improved body and the size and position of the location where the mixing failure occurs. May occur. The specific reference value described above is usually set on the safe side in consideration of the above variation.
According to the method (4) above, by using a neural network to periodically output the core collection rate at a high frequency, it is possible to obtain a large amount of data on the core intake rate at various positions of the improved body. As a result, it is possible to prevent a misjudgment of mixing failure due to the above variation and to produce an improved body of higher quality.

Claims (5)

It is a method of predicting the quality of the improved product using a prediction model in the mechanical stirring type deep mixing process method.
The above prediction model consists of learning consisting of data on the construction method determined before construction, data on the ground to be constructed, data on construction conditions at the time of construction, input data for learning including ground strength constants, and core sampling rate. It was created by training with a neural network using multiple training data that is a combination of output data for use.
Data on the construction method determined before construction, data on the ground to be constructed, data on construction conditions at the time of construction, and input data for prediction including the ground strength constant are input to the above prediction model, and from the above prediction model. , A method for predicting the quality of an improved product, which comprises outputting the core sampling rate as output data for prediction and predicting the quality of the improved product. The data on the above method is at least one selected from the data on the stirring and mixing device and the data obtained in the pre-blending test.
The data on the above ground is at least one selected from the maximum depth for creating an improved body, the layer thickness of each layer constituting the ground to be constructed, and the soil classification of the above ground.
The data regarding the construction conditions at the time of construction is at least one selected from the torque, rotation speed, penetration speed, pulling speed, solidifying material slurry discharge amount, and construction depth at the time of stirring and mixing.
The quality prediction method for an improved product according to claim 1, wherein the ground strength constant is at least one selected from the N value and the adhesive strength. If the core sampling rate output from the above prediction model is less than the standard value within the range of 80 to 95% set according to the soil classification of the ground to be constructed, there is a mixing defect in the above improved body. The quality prediction method for an improved product according to claim 1 or 2, wherein it is determined that the improved product has occurred, and if it is equal to or higher than the above reference value, it is determined that no mixing defect has occurred in the improved product. The core sampling rate output from the above prediction model is at least one core sampling rate obtained from the section of the improved product, which is separated by an interval within the range of 0.5 to 2 m and is arbitrarily determined. If at least one of the output core collection rates is less than the above reference value, it is determined that a mixing defect has occurred in the improved product, and all the output core collection rates are obtained. The quality prediction method for the improved product according to claim 3, wherein if is equal to or higher than the above-mentioned reference value, it is determined that no mixing defect has occurred in the improved product. The quality prediction method for the improved product according to claim 3 or 4 is performed by the mechanical stirring type deep mixing treatment method, and the core sampling rate output using the prediction model indicates that a mixing defect has occurred in the improved product. If it is determined that the above-mentioned improved body section has a core sampling rate that is less than the above-mentioned reference value, a second mechanical stirring-type deep-layer mixing treatment method that involves stirring and mixing with the discharge of the solidifying material slurry.