JP7316952B2 - Sediment fluidity determination method - Google Patents

Description

The present invention relates to a sediment fluidity determination method for determining the fluidity of sediment discharged from a shield machine.

As a shield construction method for constructing a tunnel using a shield machine, a mud pressure shield construction method may be employed. A shield machine used in this construction method has a cutter head, a chamber into which the earth and sand excavated by the cutter head is taken in, and a screw conveyor for discharging the earth and sand in the chamber backward. In addition, there is a method in which a belt conveyor is provided behind the shield machine for conveying the earth and sand discharged from the screw conveyor to the rear. In this method, the earth excavated by the cutter head is taken into the chamber and filled with it, and the earth is discharged to the conveyor via the screw conveyor while stabilizing the face by the earth pressure inside the chamber. Therefore, in order to stabilize the face in the mud pressure shield construction method, it is important that the earth and sand in the chamber maintain appropriate plastic fluidity. Therefore, in the earth pressure shield construction method, it is important to determine (evaluate) the fluidity of earth and sand in terms of construction management of a tunnel using the earth pressure shield construction method.

As an example of a method for determining the fluidity of earth and sand, a method for measuring and evaluating properties of excavated soil in a chamber disclosed in Patent Document 1 and the like is known. This method measures the pressure and shear force acting on each part in the chamber with a soil pressure gauge and a shear force gauge, and evaluates the properties of the earth and sand (mud) in the chamber based on the measured values.

Patent No. 6522954

However, in the method disclosed in Patent Document 1, it is necessary to install a measuring instrument with a special structure in the chamber on which the earth pressure acts in the shield machine, and a device for this is required. In some cases, an operator collects the earth and sand discharged from the screw conveyor onto the belt conveyor, and the properties of the collected earth and sand are evaluated, for example, by the operator's sense of touch. It is necessary to constantly collect the earth and sand from the belt conveyor from the time of excavation to the completion of excavation, and there is room for improving work safety.

The present invention has been made with a focus on such actual circumstances, and provides a sediment fluidity determination method capable of improving work safety without using a measuring instrument with a special structure. intended to

According to one aspect of the present invention, there is provided an earth and sand fluidity determination method for determining the fluidity of earth and sand discharged from a shield machine. This sediment fluidity determination method includes sediment image data obtained by imaging a predetermined area in the conveying direction of a conveyor that conveys the sediment discharged from the shield machine, and plasticity of the sediment corresponding to the sediment image. obtaining a plurality of teacher data sets consisting of data of measurement results of fluidity; and plastic fluidity of the target sediment corresponding to the new sediment image based on the plurality of teacher data sets. estimating using a pre-learned machine learning algorithm, and determining the fluidity of the target sediment based on the estimation result.

A sediment fluidity determination method according to one aspect of the present invention estimates the plastic fluidity of target sediment using a machine learning algorithm learned in advance based on a plurality of teacher data sets, and based on the estimation results, the target It is the structure which determines the fluidity|liquidity of earth and sand. Therefore, the fluidity of the target earth and sand can be determined without using a measuring instrument with a special structure. In addition, once we have acquired multiple training data sets, we simply obtain a sediment image of the target sediment without collecting the sediment on the conveyor. The fluidity of the target earth and sand can be determined based on the estimation result. As a result, it is possible to safely determine the fluidity of the earth and sand without constantly collecting the earth and sand from the conveyor.

In this way, it is possible to provide a sediment fluidity determination method capable of improving work safety without using a measuring instrument having a special structure.

BRIEF DESCRIPTION OF THE DRAWINGS It is a conceptual diagram for demonstrating the sediment fluidity determination method which concerns on one Embodiment of this invention. It is another conceptual diagram for demonstrating schematic structure of the said sediment fluidity determination method. It is a conceptual diagram for demonstrating an example of the sediment image acquired in the said sediment fluidity determination method. It is a conceptual diagram for demonstrating an example of the measuring method of the plastic fluidity of the earth and sand used in the said earth-sand fluidity determination method. It is a conceptual diagram for explaining the teacher data set in the above-mentioned sediment fluidity determination method. FIG. 4 is a conceptual diagram for explaining categories indicating the degree of plastic fluidity. It is a conceptual diagram of the neural network used in the said sediment fluidity determination method. It is a conceptual diagram for demonstrating the modification of the said plastic fluidity measuring method. It is a conceptual diagram for demonstrating another modification of the said plastic fluidity measuring method.

EMBODIMENT OF THE INVENTION Hereafter, with reference to an accompanying drawing, embodiment of the sediment fluidity determination method which concerns on this invention is described.

FIG.1 and FIG.2 is a conceptual diagram for demonstrating the sediment fluidity determination method which concerns on one Embodiment of this invention. The sediment fluidity determination method is a method for determining the fluidity of the sediment discharged from the shield machine 1. FIG. 1 shows a schematic structure of the shield machine 1, and FIG. A partially enlarged conceptual diagram including a belt conveyor C installed behind is shown. In FIG. 1, the left side is the front side (face side) in the direction of tunnel excavation, and the right side is the rear side (pithead side) in the direction of tunnel excavation. In this embodiment, the shield machine 1 will be described as an example of a so-called earth pressure type shield machine (that is, a shield machine used in a mud pressure shield construction method).

First, the general structure and the like of the shield machine 1 will be described.

The shield machine 1 includes a cylindrical skin plate 2, a cutter head 3 for excavating the ground, and a partition wall 4, as shown in FIG. The cutter head 3 is arranged at the front end of the skin plate 2 , and the partition wall 4 is arranged inside the skin plate 2 while being spaced behind the cutter head 3 . The cutter head 3 is rotatably supported by the partition wall 4, and excavates the ground while rotating using a drive motor 5 as a drive source. A chamber 6 is defined between the cutter head 3 and the partition wall 4 . Earth and sand generated by excavation by the cutter head 3 is taken into the chamber 6 through gaps between spokes (not shown) of the cutter head 3 . The soil in the chamber 6 is discharged from the chamber 6 by a screw conveyor 11, which will be described later.

In addition, the shield machine 1 assembles the segments 8 having arc-shaped cross sections in the circumferential direction by erectors 7 provided behind the partition wall 4 to sequentially construct cylindrical segment rings 9 in the axial direction of the tunnel. The propulsion force of this shield excavator 1 is obtained by a plurality of hydraulic propulsion jacks 10 . The shield machine 1 pressurizes the inside of the chamber 6 by adjusting the amount of earth and sand discharged by the screw conveyor 11 that carries out the earth and sand in the chamber 6 to the rear of the partition wall 4, thereby applying the earth pressure to the face. By making the face stable, the face is excavated. The mud pressure type shield machine 1 can add a predetermined additive (additive agent) such as bentonite to the earth and sand in the chamber 6, and can adjust the amount of the additive. By adjusting the amount, the soil in the chamber 6 can be adjusted to have properties such as fluidity suitable for construction. Therefore, the chamber 6 is specifically filled with mixed earth and sand (also referred to as mud) of earth and sand produced by excavation by the cutter head 3 and additives, and this mixed earth and sand is shield excavated by the screw conveyor 11. It is ejected from machine 1.

Further, behind the shield machine 1, a belt conveyor C is provided for conveying further rearward the earth and sand discharged from the discharge port of the screw conveyor 11 (in this embodiment, the mixed earth and sand described above). Although not shown, a scaffolding is provided in the shield machine 1 so as to extend rearward, and for example, a belt conveyor C is placed on the scaffolding. In this embodiment, the belt conveyor C corresponds to the "conveyor for transporting earth and sand discharged from the shield machine" according to the present invention.

The ground is excavated by the shield excavator 1 configured as described above, and earth and sand taken into the chamber 6 are discharged behind the bulkhead 4 through the inside of the screw conveyor 11 . Then, the fluidity of the earth and sand discharged from the shield machine 1 (specifically, the screw conveyor 11) is determined by the earth and sand fluidity determination method described in detail below.

Next, the sediment fluidity determination method according to this embodiment will be described in detail.

The earth and sand fluidity determination method is a method of determining the fluidity of earth and sand discharged from the shield machine 1. For the earth and sand image data obtained by imaging a predetermined area V in the conveying direction on the belt conveyor C, It is configured to determine the fluidity of the earth and sand based on the estimation result of the plastic fluidity of the earth and sand obtained by processing using a pre-learned machine learning algorithm.

In this embodiment, the sediment fluidity determination method includes a teacher data set acquisition step of acquiring a teacher data set for learning a machine learning algorithm, a category setting step of predetermining a plurality of categories indicating the degree of plastic fluidity, and a determination step of determining fluidity. FIG. 3 is a conceptual diagram for explaining an example of a sediment image, FIG. 4 is a conceptual diagram for explaining an example of a method for measuring the plastic fluidity of sediment used in the teacher data set acquisition process, and FIG. FIG. 6 is a conceptual diagram for explaining a teacher data set, and FIG. 6 is a conceptual diagram for explaining classifications showing the degree of plastic fluidity of soil set in the classification setting process.

In the sediment fluidity determination method in this embodiment, a sediment fluidity determination device 50 having an illumination unit 20, an imaging unit 30, and a main unit 40 is used. In this embodiment, the main unit 40 includes a storage unit 41 , a processing unit 42 and a display unit 43 . The details of each component (20, 30, 40) of the sediment fluidity determination device 50 will be described in detail in the following description of the sediment fluidity determination method.

In the training data set acquisition step, training data consisting of sediment image data obtained by imaging a predetermined region V in the conveying direction of the belt conveyor C and data of the measurement result of the plastic fluidity of the sediment corresponding to this sediment image. This is the step of acquiring a plurality of sets. The teacher data set acquisition process has a sediment image data acquisition process, a plastic fluidity measurement process, and a teacher data set storage process.

Specifically, the sediment image acquisition process in the teacher data set acquisition process is a process of acquiring the sediment image for the teacher data set. Specifically, in this step, as shown in FIGS. 1 and 2 , the imaging unit 30 captures an image of a predetermined area V in the conveying direction of the belt conveyor C illuminated by the illumination unit 20 . The predetermined area V is set in the vicinity of the discharge port of the screw conveyor 11 on the belt conveyor C, and the imaging unit 30 is installed above the belt conveyor C so as to be able to image the soil image in the predetermined area V of the belt conveyor C. ing. The earth and sand image acquisition step is performed a plurality of times while the belt conveyor C is being driven, for example, and earth and sand images of different earth and sand masses passing through the predetermined region V are obtained respectively.

In this embodiment, the imaging unit 30 is arranged so that the shape and thickness (height) of the earth and sand can be grasped more clearly from each earth and sand image. Specifically, the imaging unit 30 is installed so that the predetermined area V can be imaged by looking down from one side of the belt conveyor C in the width direction. In other words, in the present embodiment, the sediment image acquired by the imaging unit 30 is an image obtained by imaging the predetermined area V from one side of the belt conveyor C in the width direction. As the image pickup unit 30, for example, a general visible light camera (specifically, a general-purpose digital camera) may be employed. Further, the shutter speed of the imaging unit 30 is set to an appropriate shutter speed according to the speed of the belt of the belt conveyor C so as to acquire a sediment image without image blur. In other words, the imaging unit 30 will be described below assuming that it captures a still image.

The data of the sediment image is input to the main unit 40 and stored in the storage unit 41 of the main unit 40, for example. Here, like the earth and sand image shown on the left side in FIG. Other objects (such as the belt itself of the belt conveyor C) are shown. In order to clarify the features of the earth and sand, it is preferable that the earth and sand image for the teacher data set includes many images of the earth and sand itself. For this reason, in the present embodiment, specifically, the area of the central portion in the width direction of the belt conveyor C obtained by performing trimming processing as preprocessing on the raw data of the earth and sand image from the imaging unit 30 is Image data (see the image shown on the right side of FIG. 3 ) is stored in the storage unit 41 of the main unit 40 as sediment image data acquired by the imaging unit 30 . Although not particularly limited, for example, sand image data cut into a square pixel size of about 448×448 pixels is stored in the storage unit 41 as sand image data acquired by the imaging unit 30 . It should be noted that the range of the predetermined area V may be narrowed so that both sides of the belt conveyor C in the width direction are not reflected in the earth and sand image. For example, the predetermined area V may be narrowed by narrowing the imaging range reflected in the earth and sand image using the zoom function of the imaging unit 30 . In this case, trimming processing is unnecessary. Further, although the imaging unit 30 has been described as capturing a still image, it is not limited to this, and may capture a moving image. In this case, a still image may be extracted as a sediment image by performing frame extraction processing in the preprocessing on the moving image obtained by the imaging unit 30 .

The plastic fluidity measuring step in the teacher data set acquisition step is a step of measuring the plastic fluidity of soil corresponding to each soil image acquired by the imaging unit 30 in the soil image acquisition step. That is, in this step, the plastic fluidity of each sample soil G sampled from the same soil imaged by the imaging unit 30 in the soil image acquisition step is measured. For example, a part of the earth and sand already imaged in the predetermined area V of the belt conveyor C is taken as the sample earth and sand G, and as shown in FIG. to collect. This earth and sand collection may be performed while the belt conveyor C is being driven, but is preferably performed with the belt conveyor C stopped driving each time earth and sand is collected.

Specifically, in the plastic fluidity measuring step, as shown in FIG. 4, a slump test is adopted as a method of measuring plastic fluidity. Therefore, in the present embodiment, the data of the measurement result of the plastic fluidity of soil in the plastic fluidity measurement step is the data of the slump value S or the data of the slump flow value SF obtained using the slump test. That is, either the slump value S or the slump flow value SF may be measured as the plastic fluidity measurement of soil.

The data of the slump value S is data about the height direction of the sample earth and sand G sampled from the earth and sand discharged onto the belt conveyor C. As shown in FIG. For the measurement of the slump value S, for example, a slump test according to JIS A 1108 "Slump test method for concrete" or JIS A 1171 "Test method for polymer cement mortar" is used. In this slump test, as shown in FIG. 4(A), a tapered tubular slump cone SC was placed on a flat flat plate with its large diameter side facing down. A slump cone SC is filled with sample soil G. Then, the sample earth and sand G in the slump cone SC is poked and leveled by a shovel (not shown), and the upper surface of the sample earth and sand G is aligned with the upper end of the slump cone SC and leveled. Then, as shown in FIG. 4B, the slump cone SC is gently pulled up vertically. After that, as shown in FIG. 4(C), the drop from the upper end of the slump cone SC at the central portion of the horizontally spread sample soil G is measured as the slump value S.

The data of the slump flow value SF is data indicating the spread of the sample earth and sand in the horizontal direction. For the measurement of the slump flow value SF, for example, a slump test according to JIS A 1150 "Concrete slump flow test method" is used. Also in this slump test, as shown in FIGS. 4A and 4B, the slump cone SC placed on a flat plate is filled with the sample soil G, and then the slump cone SC is pulled up. After that, as shown in FIG. 4(C), when the horizontal flow (spreading) of the sample soil G stopped, the diameter of the sample soil G was measured in the direction considered to be the maximum and in the direction perpendicular to this. , the average value of these two diameters is defined as the slump flow value SF.

The teacher data set storage step in the teacher data set acquisition step is a step of storing the sediment image data and the plastic fluidity measurement result data in association with each other. As shown in FIG. 5, the measured value (slump value S in the figure) of the plastic fluidity of the soil (sample soil G) corresponding to each soil image is different, and the teacher data set is the soil image data and this. A plurality of teacher data sets are acquired, which consist of data of the measurement results (slump value S in the drawing) of the plastic fluidity of the soil (sample soil G) corresponding to the soil image. The data of this teacher data set are stored in the storage unit 41 of the main unit 40 .

Next, the division setting step is a step of predetermining a plurality of divisions indicating the degree of plastic fluidity and predetermining a range of the plastic fluidity measurement results corresponding to each of the plurality of divisions.

Specifically, as shown in FIG. 6, as a plurality of categories indicating the degree of plastic fluidity, for example, "hard", "slightly hard", "normal", "slightly soft", and "soft" categories. set. Then, based on the sediment plastic fluidity measurement results corresponding to each sediment image that has already been acquired, the range of the plastic fluidity measurement results corresponding to each of the plurality of divisions (that is, to which each division belongs) (Upper and lower limits) are determined in advance for each category. Data indicating the range of the plastic fluidity measurement results belonging to each category is input to the main unit 40 and stored in the storage unit 41 .

Next, in the determination step, the plastic fluidity of the target sediment, which is the sediment corresponding to the new sediment image, is estimated using a machine learning algorithm learned in advance based on a plurality of teacher data sets, and the estimation result is It is a step of determining the fluidity of the target earth and sand based on. This determination step has a learning step and an estimation step.

In the learning process of the judgment process, the relationship between the sediment image and the measured value of the plastic flow of the sediment corresponding to this sediment image is calculated based on the teacher data set acquired in the teacher data set acquisition process for the machine learning algorithm. This is the process of learning.

Specifically, the learning process is executed using the processing section 42 of the main body section 40 . The processing unit 42 is composed of a CPU, a GPU, a RAM, and the like, and performs arithmetic processing for estimating a value (for example, a slump value S) indicating the plasticity of the sediment corresponding to the sediment image by a machine learning algorithm using the sediment image. Execute. The storage unit 41 stores (memorizes) various application programs including a plastic fluidity estimation application program for machine learning algorithms. Specifically, the processing unit 42 executes the learning process by executing the plastic fluidity estimation application program stored in the storage unit 41 . This plastic fluidity estimation application program uses a machine learning algorithm to estimate the value of the plastic fluidity of the soil corresponding to the image of the soil, and is programmed to determine the fluidity of the soil based on this estimation result. It is what was done. This program estimates the plastic fluidity value corresponding to the sediment image from the sediment image, for example, by machine learning using a neural network. This neural network is fed with a training data set. More specifically, the sediment image data in the teacher data set is given as input data, and the plastic fluidity measurement result data in the teacher data set is given as teacher data. Then, in this neural network, the error between the plastic fluidity value output based on the input data (hereinafter referred to as the estimated value y) and the teacher data is calculated, and the parameter described later is updated based on this error. (learning) to improve the accuracy of the plastic fluidity estimated value y.

FIG. 7 is a conceptual diagram of data processing in a basic neural network. In this neural network, the estimated value y of plastic fluidity is obtained by the following formula (1).

Here, in equation (1), y is an estimated value (output value) of the plastic fluidity of soil, L is an activation function, _Xi is input data consisting of soil image data, and W _i is a weighting parameter and b is a bias parameter. In the unlearned neural network, a predetermined numerical value is input in advance for each parameter. In the learning in the neural network, the estimated value y obtained based on the above formula (1) from the input data (sediment image data) of the given teacher data set, and the teacher data (plastic flow Data of measurement results of sexuality), that is, the error from the actual measurement value is calculated, and the value of each parameter (W _i ,b) is updated so that this error becomes small. The learning step gradually reduces the error by sequentially performing learning including calculation of the estimated value y and the error and updating of each parameter (W _i , b) for each different teacher data set. Specifically, the error increases or decreases randomly when viewed locally as learning is repeated, but changes with a downward trend when viewed as a whole. For this reason, for example, after the total number of times of learning reaches a predetermined number, learning is continuously performed for a predetermined number of times, and there is a case where the error is not reduced during the continuous learning. determine whether or not Then, when it is determined that there is a case where the error is not reduced (in the case of YES determination), the learning is stopped. In this case, for example, the value of each parameter (W _i ,b) at the time of update immediately before continuous learning is stored in the storage unit 41 as each learned parameter (W _i ,b). On the other hand, when it is determined that there is no case where the error is not reduced (in the case of NO determination), the next continuous learning and the above determination are repeated until YES determination is made in order to continue learning. However, if the judgment does not become YES before the total number of times of learning reaches a predetermined upper limit (for example, the number of times corresponding to the number of data sets for learning in the teacher data set), the total number of times of learning is the upper limit. Learning is stopped when the number of times is reached. In this case, for example, the value of each parameter (W _i ,b) at the time of the last update is stored in the storage unit 41 as each learned parameter (W _i ,b). When the machine learning algorithm of the plastic fluidity estimation application program is learned in advance in this way, thereafter, new sediment image data is input to the plastic fluidity estimation application program using this pre-learned machine learning algorithm. By simply providing input data, an estimated plastic fluidity value y of the target soil (sediment with an unknown plastic fluidity value) corresponding to this new soil image can be obtained with a predetermined degree of certainty. In addition, after the above learning is completed, it is preferable to perform the calculation of the estimated value y on a trial basis before executing the estimation step using the new sediment image data. For example, multiple teacher datasets are divided into actual training datasets and test datasets, and after completing training for a machine learning algorithm using the training teacher datasets, actual estimation ( Before performing plastic flow estimation using new sediment image data), the sediment image data in the test teacher data set is input to the trained machine learning algorithm (plastic flow application program). , and the estimated value y obtained is used as training data (results of measurement of plastic flow) corresponding to the sediment image data in the training data set for testing. On the other hand, it is better to actually check how likely it is to be output. As a result, it is possible to confirm in advance the true ability (accuracy) of plastic flow property estimation using a machine learning algorithm at the time of actual estimation.

More specifically, a convolutional neural network (CNN) suitable for image analysis may be employed as the machine learning algorithm for the plastic fluidity estimation application program. This algorithm consists of a Convolution Layer, a Pooling Layer and a Discrimination Layer (Fully Connected Layer). In the convolution layer, the amount of information of the earth and sand image is compressed by performing convolution processing on the input earth and sand image data (input data). The compressed data is then input to the pooling layer. In the pooling layer, pooling processing for extracting information necessary for learning is performed on the input data, and the amount of information of the sediment image is further compressed. Convolutional layers and pooling layers are alternately provided, and the output after information compression by the convolutional layers and pooling layers is repeated is a vector in which the features of the sediment image are compressed. By inputting this vector to the discriminant layer, an estimated value y of plastic fluidity is output.

In the estimation step of the determination step, first, the imaging unit 30 acquires a sediment image of the target sediment to be determined for plastic fluidity. Then, the raw data of the acquired sediment image is subjected to preprocessing (trimming processing) in the same manner as in the sediment image acquisition process, and the data after this preprocessing is used as the sediment image data of the target sediment in the main unit 40. It is input to the processing unit 42 . As a result, the data of the earth and sand image of the target earth and sand are given as input data to the machine learning algorithm (plastic fluidity estimation application program) previously learned in the learning process. And an estimation process estimates the plastic fluidity of object earth and sand using the machine-learning algorithm learned previously. As a result, the estimated value y of the plastic fluidity of the target earth and sand can be obtained with a predetermined degree of certainty.

Moreover, in the estimation step, the fluidity of the target earth and sand is determined based on the estimation result (estimated value y). In the present embodiment, in the estimation step, as the determination result of the fluidity of the target soil, a category corresponding to the estimation result (estimated value y) of the plastic fluidity of the target soil from among a plurality of categories indicating the degree of plastic fluidity is output as the determination result of the fluidity of the target sediment. That is, in the present embodiment, the estimation step includes the above-described range (upper and lower limit values) that includes the estimation result (estimated value y) of plastic fluidity of the target sediment from among a plurality of categories indicating the degree of plastic fluidity The classification corresponding to is determined as the fluidity determination result of the target sediment. The estimation result (estimated value y) and the determination result (classified as hard, etc.) in the estimation process are displayed on the display unit 43 consisting of the display of the main unit 40, and are associated with the sediment image data corresponding to the storage unit 41. stored. In addition, you may use the estimated value y as it is as a determination result of the fluidity|liquidity of object earth and sand.

The estimation and determination of plastic fluidity in the estimation step can be performed, for example, on several sediment images per second. In the estimation step, while the belt conveyor C is being driven, the imaging by the imaging unit 30 is continuously performed at time intervals (for example, Several images per second), and the data of the acquired sediment images are sequentially given as input data to the machine learning algorithm in real time. Then, the plastic fluidity estimation result and determination result are displayed on the display unit 43 in real time together with the corresponding sediment image. Specifically, on the display unit 43, a graph is drawn as the estimation result, with the horizontal axis representing time and the vertical axis representing the estimated value y of the plastic fluidity. This graph is updated sequentially as new estimated values y are obtained. As a result, the operator can confirm (monitor) changes in the plasticity of the soil during the period from the start of excavation by the shield machine 1 to the completion of excavation with a certain degree of certainty.

Further, the plastic fluidity estimation result data and determination result data obtained in the estimation step are input from the main body 40 to a control unit (not shown) of the shield machine 1 . This control unit is configured to perform adjustment control for adjusting the amount of additive added to the soil in the chamber 6 based on the data from the main unit 40 . The control unit is configured to be able to detect a state in which the plastic fluidity of soil is lower than a predetermined lower limit value or a state in which the plastic fluidity of soil is higher than a predetermined upper limit value based on the data from the main unit 40. ing. When these states continue for a predetermined period of time, it can be assumed that the plastic fluidity of the earth and sand (mixture of earth and sand and additives) in the chamber 6 is in the same state. In this case, the control unit is configured to adjust the amount of additive added to the soil in the chamber 6 to restore the plastic fluidity of the soil in the chamber 6 to an appropriate state. It is For example, if the plastic fluidity is high (that is, tends to be soft), the amount to be added is decreased, and if the plastic fluidity is low (that is, tends to be hard), the amount to be added is increased.

The sediment fluidity determination method according to this embodiment estimates the plastic fluidity of the target sediment using a machine learning algorithm learned in advance based on a plurality of teacher data sets, and based on this estimation result, the target sediment. Liquidity is determined. As a result, the fluidity of the target earth and sand can be determined without using a measuring instrument with a special structure. In addition, once a plurality of training data sets are acquired, the plastic flow property of the target soil can be estimated by a machine learning algorithm simply by obtaining a soil image of the target soil without sampling the soil on the belt conveyor C. , the fluidity of the target sediment can be determined based on the estimation result. As a result, the fluidity of the earth and sand can be determined without collecting the earth and sand from the belt conveyor C all the time.

In this way, it is possible to provide a sediment fluidity determination method capable of improving work safety without using a measuring instrument having a special structure.

In the present embodiment, in the estimation step, a classification indicating the degree of plastic fluidity such as "hard" is output as the determination result of the fluidity of the target earth and sand. Thereby, for example, the operator can easily recognize the fluidity of the target earth and sand.

In the present embodiment, the earth and sand image is an image obtained by imaging the predetermined area V as viewed from one side of the belt conveyor C in the width direction. As a result, the operator can more clearly grasp the shape and thickness (height) of the soil from each soil image displayed on the display unit 43, and the accuracy of plastic flow estimation using the machine learning algorithm can be improved more efficiently. can be substantially improved.

In the present embodiment, a slump test is used as a method for measuring plastic fluidity, but the present invention is not limited to this, and a table flow test or a vane shear test may be used. FIG. 8 is a conceptual diagram for explaining the table flow test, and FIG. 9 is a conceptual diagram for explaining the vane shear test.

Specifically, as shown in FIG. 8, when the table flow test is adopted, the data of the measurement result of the plastic flow property of the soil in the plastic flow property measurement process is the flow value obtained using the table flow test. This is F data. The data of the flow value F is data indicating the spread of the sample soil G in the horizontal direction. The table flow test obtains data showing the spread in the horizontal direction similar to the slump flow value SF in the slump test, but the test direction is slightly different from the slump test. The measurement of the flow value F uses, for example, a table flow test that mutatis mutandis applies the "flow test" described in JIS R 5201 "cement physical test method". In this table flow test, as shown in FIG. 8(A), a tapered cylindrical flow cone FC is placed on a smooth disk-shaped flow table FT horizontally installed with its large diameter side facing down. Install. In this state, the flow cone FC is filled with sample soil G. Then, the sample earth and sand G in the flow cone FC is poked and leveled with a shovel (not shown), and the upper surface of the sample earth and sand G is aligned with the upper end of the flow cone FC and leveled. A shaft F1 for raising and lowering the table is attached to the lower part of the flow table FT, and this shaft F1 is configured to be raised and lowered via a cam F3 that rotates in conjunction with the rotation operation of the raising and lowering handle F2. Therefore, the flow table FT moves up and down when the lifting handle F2 is rotated by an operator or the like. Then, as shown in FIG. 8B, the slump cone SC is gently pulled up vertically. After that, as shown in FIG. 8(C), when the lifting handle F2 is rotated by a worker or the like, the sample soil G is given a falling motion a plurality of times. Then, when the sample soil G stops flowing (spreading) in the horizontal direction, the diameter of the sample soil G is measured in the direction recognized as the maximum and in the direction perpendicular to this, and the average value of these two diameters is the flow value. Let F.

Further, as shown in FIG. 9, when the vane shear test is adopted, the data of the measurement result of the plastic fluidity of the soil in the plastic fluidity measurement process is the vane shear strength obtained using the vane shear test. This is data of value I. The data of the vane shear strength value I is data indicating the shear strength of the sample soil G. For the measurement of the value I of the vane shear strength, for example, a vane shear test that applies mutatis mutandis JGS 1411 "In-situ vane shear test method" of the Japanese Geotechnical Society standard is used. In this vane shear test, as shown in FIG. 9(A), a container B1 is filled with sample earth and sand G. Then, as shown in FIG. 9(B), a rotating rod B3 having vane blades B2 formed by combining four blades in a cross shape attached to the tip thereof is pushed into the sample soil G. Then, as shown in FIG. Thereafter, as shown in FIG. 9(C), the rotating rod B3 is rotated at a predetermined rotational speed. Measure the maximum torque required for this rotation. Then, based on the measured maximum torque value, the width and height of the vane blade B2, and the frictional torque between the rotating rod B3 and the sample soil G, the vane shear strength value I is calculated.

For implementation of a convolutional neural network (CNN), which is a machine learning algorithm, Keras, which is an open source neural network library, and TensorFlow, which is also an open source, can be used.

As described above, the embodiments and modifications of the present invention have been described, but the present invention is not limited to the above-described embodiments and modifications, and further modifications and changes are possible based on the technical idea of the present invention. Of course it is.

DESCRIPTION OF SYMBOLS 1... Shield machine, C... Belt conveyor (conveyor), F... Flow value, G... Sample earth and sand, I... Value of vane shear strength, S... Slump value, SF... Slump flow value, V... Predetermined area

Claims (6)

A sediment fluidity determination method for determining the fluidity of sediment discharged from a chamber of a shield machine,
Data of an earth and sand image obtained by imaging a predetermined area in the conveying direction on a conveyor that conveys the earth and sand discharged from the chamber of the shield machine, and a measurement result of the plastic fluidity of the earth and sand corresponding to the earth and sand image. a step of obtaining a plurality of training data sets consisting of data;
estimating the plastic fluidity of the target sediment, which is the sediment corresponding to the new sediment image, using a machine learning algorithm learned in advance based on the plurality of teacher data sets, and based on the estimation result, A step of determining the fluidity of the target earth and sand;
Including, the sediment fluidity determination method. The data of the measurement result of the plastic fluidity of the sediment indicates the data of the slump value in the height direction of the sample sediment collected from the sediment or the spread of the sample sediment in the horizontal direction, which is obtained using a slump test. The sediment fluidity determination method according to claim 1, wherein the data is a slump flow value. 2. The method according to claim 1, wherein the data of the measurement result of the plastic flow property of the soil is flow value data indicating the horizontal spread of the sample soil taken from the soil, obtained using a table flow test. Sediment fluidity determination method. 2. According to claim 1, wherein the data of the measurement result of the plastic fluidity of the soil is vane shear strength value data indicating the shear strength of a sample soil sampled from the soil, obtained using a vane shear test. The sediment fluidity determination method described. Further comprising a step of predetermining a plurality of categories indicating the degree of plastic fluidity and predetermining a range of the plastic fluidity measurement results corresponding to each of the plurality of categories,
In the step of determining the fluidity of the target sediment, from among the plurality of categories, the category corresponding to the range containing the estimation result of plastic flowability of the target sediment is selected as the fluidity determination result of the target sediment. Determined as, the sediment fluidity determination method according to any one of claims 1 to 4. The sediment fluidity determination method according to any one of claims 1 to 5, wherein said machine learning algorithm is a convolutional neural network.