JP2021017758A - Shield excavator control system and shield excavator control method - Google Patents

Abstract

To provide a shield excavator control system that outputs a desirable frequency and operation amount of a skilled operator even if the excavation status data obtained from the shield excavator is a feature amount that is infrequently operated.SOLUTION: A device control system of the present invention comprises an operation determination model that outputs the operation probability of a shield excavator as an estimated value by inputting a feature amount indicating the excavation status supplied from the shield excavator, an operation prediction model that outputs the operation parameters of the shield excavator as an estimated value by inputting the feature quantity indicating the excavation status supplied from the shield excavator, a processing control unit that estimates operation parameters for the operation prediction model when the operation determination model outputs an operation probability equal to or higher than a predetermined threshold.SELECTED DRAWING: Figure 1

Description

The present invention relates to a shield excavator control system and a shield excavator control method.

Conventionally, a shield excavator that excavates the ground has been used for the construction of tunnels and the like. The operator who operates the shield excavator follows the excavation instructions that indicate the excavation instructions planned in advance, and measures the construction environment (soil quality, water pressure, etc.) at the site where the shield excavator excavates from various measuring instruments. The direction of excavation is controlled while monitoring.

In the shield excavator, a plurality of shield jacks are provided along the inner circumference of the cylindrical skin plate. The surface of the skin plate is pushed by propelling (extending) all or part of the plurality of shield jacks by flood control operation, and the direction of excavation of the shield excavator is controlled. The operator adjusts the position of the force point acting on the shield excavator by selecting which of the plurality of shield jacks is extended, and controls the direction in which the shield excavator excavates.

By constructing tunnels at various sites, operators are cultivating experience in controlling shield excavators in response to changes in the construction environment and improving their proficiency. Skilled operators with improved proficiency apply their knowledge of control in similar construction environments in the past to control corresponding to the current construction environment of the site when controlling the shield excavator at the site during excavation. I am doing it. However, in the case of an operator with a small number of construction sites, in a construction environment that he has never experienced, his experience and operational knowledge are insufficient, and it is not possible to appropriately control the shield excavator in that construction environment.

That is, there is a problem that the accuracy and safety of the design of the tunnel to be excavated vary depending on the control skill of each shield excavator of the operator. In order to solve this problem, there is a configuration in which the shield excavator is automatically operated based on the measurement data indicating the rotational state of the cutter of the shield excavator and the propulsion state of the propulsion jack during excavation (see, for example, Patent Document 1).

In addition, it is common practice to use artificial intelligence (AI) to analyze human emotions. In the method using AI, for example, a trained model (machine learning) is executed by executing machine learning using teacher data in which a human facial expression (input) and an emotion (output) corresponding to the facial expression are associated with each other. Model) is created. By inputting a human facial expression into this trained model, it is possible to estimate the emotions that the facial expression means, and it is possible to analyze human emotions.

JP-A-2018-154998

In Patent Document 1 described above, the operation of the shield excavator by a skilled operator cannot be sufficiently reproduced. That is, since the control is performed by comparing the measured data with the set value, the control is appropriately performed according to the ever-changing construction environment of the site, unlike the control based on the experience of a skilled operator. It was not always the case, and it was hard to say that the accuracy and safety of the excavated tunnel design would improve.

On the other hand, it is conceivable to apply the above-mentioned AI method to the operation of the shield excavator. For example, executing machine learning using teacher data that associates the excavation status (input monitoring item data) obtained from the shield excavator with the operation (output) of a skilled operator corresponding to the excavation status. Create a machine learning model by. By inputting data indicating the excavation status currently being excavated into this machine learning model, it is possible to estimate the content of the desired operation of the shield excavator. For example, by selecting a shield jack so that the power point of the shield excavator comes to the position of the power point estimated by the machine learning model, even an inexperienced operator can perform an operation close to that of a skilled operator. Is thought to be possible.

However, since the shield excavator is operated infrequently, it is performed once every several tens of seconds to several hundreds of seconds, for example. Since machine learning is performed using teacher data associated with the operation of a skilled operator corresponding to this excavation situation, a machine learning model with the correct answer of not always operating the excavation situation tends to be generated. There is. It is difficult to use the estimation results estimated by this machine learning model as guidance for inexperienced operators.

In addition, by adjusting the parameters (set value or limit value) of the machine learning model, it is possible to solve the problem of generating a machine learning model whose correct answer is not always operated.
However, as the excavation status data (features) obtained from the shield excavator changes from moment to moment, the predicted value of the machine learning model also fluctuates. As a result, the machine learning model always outputs an instruction to change the operation as an estimation result, and the operation is frequently performed, and it is not realistic to use the estimation result as guidance for an inexperienced operator.

The present invention has been made in consideration of the above circumstances, and even if the excavation status data obtained from the shield excavator is a feature amount that is infrequently operated, the frequency of desirable operations by a skilled operator and the desired frequency of operations An object of the present invention is to provide a shield excavator control system and a shield excavator control method that output an operation amount.

In order to solve the above problems, the shield excavator control system of the present invention is an operation that outputs the operation probability of the shield excavator as an estimated value by inputting a feature amount indicating the excavation status supplied from the shield excavator. An operation prediction model that outputs an operation parameter of the shield excavator as an estimated value by inputting a determination model and a feature amount indicating an excavation status supplied from the shield excavator, and the operation determination model having a predetermined threshold value or more. It is characterized by including a processing control unit that causes the operation prediction model to estimate the operation parameters when the operation probability of the above is output.

In the shield excavator control system of the present invention, the operation determination model outputs the presence or absence of an operation change by using the history table in which the acquired record of the excavation status is recorded as teacher data and using the excavation status data as an input value. It is a machine learning model that is machine-learned as a value and outputs the operation probability corresponding to the data of the excavation situation as an estimated operation determination value.

In the shield excavator control system of the present invention, when learning the operation determination model, a numerical value obtained by dividing the number of records whose operation is not changed by the number of records whose operation is changed is used as a weight in the history table. It is characterized by learning an operation judgment model.

In the shield excavator control system of the present invention, the estimated value based on the teacher data is obtained from the trained operation determination model, and the number of records TN in the first group in which the record whose operation is not changed is determined not to change the operation and the operation change. The number of records FN in the second group where it is determined that the record that does not change the operation is determined, the number of records TP in the third group where the record whose operation is changed is determined to change the operation, and the record whose operation is changed is determined not to change the operation It is characterized in that the threshold value for whether or not to perform an operation is set according to the number of records and the number of data in the fourth group.

In the shield excavator control system of the present invention, the operation prediction model inputs a feature amount indicating the excavation status of a record having an operation probability equal to or higher than the threshold value in the second group and the third group, and the shield excavator is used. It is characterized in that it is machine-learned by teacher data whose estimated value is the operation parameter of.

The shield excavator control method of the present invention is an operation determination process that outputs the operation probability of the shield excavator as an estimated value by inputting a feature amount indicating the excavation status supplied from the shield excavator by the operation determination model. The operation prediction process that outputs the operation parameters of the shield excavator as an estimated value by inputting the feature amount indicating the excavation status supplied from the shield excavator by the operation prediction model, and the processing control unit are described above. It is characterized by having a processing control process for estimating the operation parameters for the operation prediction model when the operation determination model outputs an operation probability equal to or higher than a predetermined threshold.

According to the present invention, even if the excavation status data obtained from the shield excavator is a feature amount that is infrequently operated, the desired frequency and operation amount of a skilled operator are estimated and output (presented). be able to.

It is a schematic block diagram which shows the structural example of the shield excavator by the earth pressure type shield method to which the shield excavator control system of this embodiment is applied. It is a figure which shows the configuration example of the shield excavator control system by this embodiment. It is a figure which shows the structural example of the teacher data table in the teacher data storage unit 37 in this embodiment. It is a graph which shows the probability distribution of the estimated operation determination value explaining the operation threshold determination process. It is a flowchart which shows the operation example of the estimation processing of the control data of the shield excavator of the shield excavator control system by this embodiment. It is a graph which shows the comparison between the determination result of the operation determination model in this embodiment, and the actual measurement.

FIG. 1 is a schematic configuration diagram showing a configuration example of a shield excavator by a soil pressure shield method to which the shield excavator control system of the present embodiment is applied. FIG. 1A shows a conceptual diagram of the shield excavator 10 viewed from the side, and FIG. 1B shows a conceptual diagram of the shield jack (propulsion jack) 20 for propelling the shield excavator 10 as viewed from the front. There is.
As shown in FIG. 1 (a), the shield excavator 10 excavates the ground at the rear part of the cylindrical skin plate 11 by assembling the segments by an elector (not shown) and performing the primary lining S. It is a mechanism to do. In the shield excavator 10, a chamber 12 is provided at the rear of the annular and face plate type cutter 16 provided with the cutter bit 15. A plurality of soil pressure gauges D are installed on the side wall in the chamber 12. The earth pressure gauge D measures the pressure of mud (controlled earth pressure) in the chamber 12.

Additive 14 is injected into the chamber 12 from the mud injection pipe 13. The excavated soil deposited in the chamber 12 is agitated with the additive 14 by a kneading blade (not shown) to be kneaded and converted into mud.
The screw conveyor 17 discharges the mud from the chamber 12 to the conveyor 18 via the soil discharge gate G. Then, the conveyor 18 carries out the mud discharged from the screw conveyor 17 to the outside of the tunnel via the conveyor 19. The gantry M supports the screw conveyor 17 and the conveyors 18 and 19. Further, although not shown, the shield excavator 10 is provided with a propulsion jack (described later), and the excavation direction and the propulsion speed are controlled by the propulsion jack.

In the present embodiment, the propulsion direction and propulsion speed of the skin plate 11 are controlled by hydraulic control of the propulsion jack (hereinafter, may be simply referred to as direction control), and the rotation speed of the screw of the screw conveyor 17 (described later). The mud pressure is controlled (hereinafter, may be simply referred to as soil pressure control) by controlling the screw speed). The shield excavator control system, which will be described later, acquires each of the monitoring item data indicating the state of these controlled objects and the presence / absence of operation, and controls the shield excavator, and determines whether or not to perform the operation. Estimate using a machine learning model (described later). Here, as the machine learning technique for creating a machine learning model, any of commonly used techniques such as decision tree learning, neural network, genetic programming, support vector machine, and deep learning may be used. ..

In the present embodiment, the propulsion direction and propulsion speed of the skin plate are controlled (direction control) by hydraulic control or the like as control data of the propulsion jack 20, and the rotation speed of the screw (screw rotation) is used as control data of the screw conveyor 17. Control data for controlling mud pressure (soil pressure control) shall be acquired by controlling speed) and the like. Further, in the present embodiment, the monitoring item data includes data for monitoring the excavating construction environment and the operating state of the shield excavator 10 (monitoring item data), for example, cutter torque, cutter speed, propulsion pressure, and propulsion. Speed, propulsion speed instruction deviation value, control earth pressure, face earth pressure average value out of range Dummy, agitator torque, screw speed, primary screw pressure, secondary screw pressure, NO. 1 copy stroke, NO. 1 Copy stroke indicated value difference, NO. 1 copy position, NO. 1 Copy position instruction sheet Deviation value, pitching, pitching instruction value difference (pitching angle error), rolling, rolling instruction value difference, vertical center fold angle, vertical center fold angle instruction value difference, left / right center fold angle, left / right center fold angle instruction Value difference, S / M front body azimuth, S / M front body azimuth instruction value difference, S / M rear body azimuth, S / M rear body azimuth instruction value difference, planned route horizontal deviation (horizontal deviation: control point), plan Route vertical deviation (vertical deviation: control point), azimuth (azimuth angle error: control point), azimuth indication value difference (control point), planned route azimuth (control point), pitch (control point), planned route pitch (control point) )and so on. Here, the instruction sheet indicates an excavation instruction sheet. The above-mentioned monitoring item data is characteristic data as an explanatory variable input to the machine learning model together with the presence or absence of operation.

FIG. 1B shows a conceptual diagram illustrating a propulsion jack that propels the shield excavator 10. As shown in FIG. 1B, the force point for propelling the surface of the skin plate 11 is set depending on which position the propulsion jack 20 is driven. In FIG. 1 (b), 22 propulsion jacks are shown, but the number is not limited. By distributing the jack pressure of each propulsion jack 20, the propulsion jack force point position Fx on the x-axis and the propulsion jack force point position Fz on the z-axis are set, and the force point for propelling the surface of the skin plate 11 in the y-axis direction. The position of is set. The direction in which the shield excavator 10 is propelled is set by applying the propelling pressure to the position corresponding to the force point on the surface of the skin plate 11. Control in this direction is performed by a propulsion jack pattern indicating which of the propulsion jacks 20 is driven. Then, the propulsion speed is set by the amount of oil (oil amount) supplied per unit time to increase the jack pressure.

FIG. 2 is a diagram showing a configuration example of a shield excavator control system according to the present embodiment. In FIG. 2, the shield excavator control system 30 includes a monitoring item data input unit 31, an operation determination unit 32, a processing control unit 33, an operation prediction unit 34, a machine learning model generation unit 35, a learned model storage unit 36, and teacher data. Each of the storage unit 37 and the operation status data storage unit 38 is provided.
The monitoring item data input unit 31 of each of the above-mentioned monitoring items at the timing when a timing signal indicating the passage of a predetermined measurement cycle time (for example, 1 second) is supplied from a timing means (timer or the like) (not shown). Using the data as a measurement value, the above-mentioned monitoring item data is acquired from each of the detection means (sensors, measuring instruments, etc.) provided in each part. Further, the monitoring item data input unit 31 sequentially writes and stores the measured values of the acquired monitoring item data in the monitoring item data table in the operation status data storage unit 38. At this time, the monitoring item data input unit 31 stores a plurality of measured values for a predetermined period for each type of monitoring item data, and when writing a new measured value, overwrites and writes the oldest measured value. Remember with.

Further, the monitoring item data input unit 31 is a reference when it is necessary to supply the measured value as a standardized (standardized) numerical value or an averaged numerical value based on the monitoring item data to the machine learning model described later. Calculate normalization and averaging. In the case of the measured value such as the controlled earth pressure in the monitoring item data, the averaging process may be supplied as an abnormally high value or an abnormally low value due to overlapping noise. As a result, repeated abnormal stops due to noise affect the progress of construction by the shield excavator. Therefore, the monitoring item data input unit 31 obtains the measured value and the moving average in the past predetermined period for the measured value of the monitoring item data set in advance, and uses this moving average as the measured value of the monitoring item data. Used as.

The operation determination unit 32 reads the operation determination model, which is a learning model, from the learned model storage unit 36. Then, the operation determination unit 32 inputs each of the monitoring item data supplied from the monitoring item data input unit 31 into the read operation determination model as feature data, and estimates the estimated operation determination value as the operation probability. ..
The processing control unit 33 determines whether or not the estimated operation determination value supplied from the operation determination unit 32 is equal to or greater than the operation threshold value. Then, when the estimated operation determination value is equal to or greater than the operation threshold value, the processing control unit 33 causes the operation prediction unit 34 to estimate the estimation control data based on the monitoring item data. On the other hand, when the estimated operation determination value is less than the operation threshold value, the processing control unit 33 does not cause the operation prediction unit 34 to estimate the estimation control data based on the monitoring item data.

The operation prediction unit 34 reads the operation prediction model, which is a learning model, from the learned model storage unit 36. Then, the operation prediction unit 34 inputs each of the monitoring item data supplied from the monitoring item data input unit 31 into the read operation prediction model as feature data, and estimates control data (for example, hydraulic control of the propulsion jack). Data, screw speed of screw conveyor 17) is estimated.
The machine learning model generation unit 35 learns the machine learning model by using at least the monitoring item data, the operation determination value, and the teacher data stored in the teacher data storage unit 37, which is a set of the control data. , Generate an operation judgment model and an operation prediction model.

The trained model storage unit 36 stores each of the operation determination model and the operation prediction model generated by the machine learning model generation unit 35.
The teacher data storage unit 37 stores teacher data that trains a machine learning model to generate each of an operation determination model and an operation prediction model.
In the operation status data storage unit 38, the monitoring item data supplied from the monitoring item data input unit 31 in chronological order is stored in the monitoring item data table in record units.

FIG. 3 is a diagram showing a configuration example of a teacher data table in the teacher data storage unit 37 in the present embodiment. The teacher data table is composed of records for each measurement cycle, and at least columns for feature data, operation judgment value, control data, weight, and estimation operation judgment value are provided corresponding to the record number. The record number indicates the order of the measurement cycles in the time series. The feature data are each of the data for monitoring the construction environment being excavated and the operating state of the shield excavator 10 as the monitoring item data already described. As the operation judgment value, a judgment value as to whether or not an operation has been performed on the feature data is described. Here, in the present embodiment, as an example, "1" is assigned as an operation value to the record in which the operation is performed (record with operation), and the operation is performed in the record in which the operation is not performed (record without operation). "0" is given as no value. However, as the operation value and the operation non-value, any numerical value may be used as long as the estimated operation determination value is a numerical value obtained as the operation probability in the range between the operation value and the operation non-value. For example, different numerical values are used depending on whether the operation is performed or not, such as "100" as the operation value and "1" as the operation non-value. Further, the control data indicates the amount of operation operated corresponding to the feature data.

The weight is a weight for balancing the number of records n0 of the record without operation and the number of records n1 of the record with operation when performing machine learning in the present embodiment. Here, as already described, since the number of records n0 of the record without operation is larger than the number of records n1 of the record with operation, the weight of the record without operation is set to "1" and the weight of the record with operation is set to "n0". / N1 ". As the estimated operation judgment value, the operation judgment value (estimated operation judgment value) estimated by inputting the monitoring item data to the operation judgment model trained by the teacher data is shown.
Here, the machine learning model generation unit 35 uses the operation determination value as an objective variable, the monitoring item data as an explanatory variable, and takes the above weights into consideration as learning of the operation determination model. Then, the machine learning model generation unit 35 inputs the monitoring item data of each record into the generated operation judgment model, and writes the estimated estimated operation judgment value in the column of the estimated operation judgment value of each record. Remember.

Next, the process of determining the operation threshold is performed as shown below. FIG. 4 is a graph showing a probability distribution of estimated operation determination values for explaining the operation threshold determination process. In this figure, the horizontal axis shows the estimated operation judgment value, and the vertical axis shows the probability distribution. The broken line shows the probability distribution of the record without operation, and the solid line shows the probability distribution of the record with operation. Hereinafter, the probability distribution will be described as the number of records of the record without operation and the record with operation.
The operation threshold value describes a threshold value in which the predicted value is changed and the number of records included in the first group, the second group, the third group, and the fourth group at that time is appropriate. Here, the first group (TN) is a group of records with no operation change in which the operation determination value is "0 (no operation value)" and the estimated operation determination value is less than the operation threshold value. The second group (FN) is a group of records with no operation change whose operation determination value is "0 (no operation value)", but whose estimated operation determination value is equal to or higher than the operation threshold. The third group (TP) is a group of records with an operation change in which the operation determination value is "1 (operation value)" and the operation determination value is an estimated operation determination value equal to or higher than the operation threshold. The fourth group (FP) is a group of records with an operation change in which the operation determination value is "1 (operation value)", but the operation determination value is an estimated operation determination value less than the operation threshold.

In the case of the present embodiment, the number of records TN belonging to the first group without operation change, the number FN of records without operation change belonging to the second group, the number of records TP with operation change belonging to the third group, and the fourth The operation threshold value is obtained by the following equation (1) based on the number of records FP with operation change belonging to the group. That is, the estimated operation determination value that minimizes the evaluation value BER (balannced error ratio) obtained by the equation (1) is set as the operation threshold value.
BER = 0.5 × ((FP / (TN + FP)) + (FN / (FN + FP))) ... (1)
However, instead of obtaining by this equation (1), another equation may be used, or an arbitrary estimated operation determination value may be set as the operation threshold value.

Further, when the machine learning model generation unit 35 trains the operation prediction model, there are FN records belonging to the second group without operation change and operation changes belonging to the third group corresponding to the set operation threshold value. TP records are used as teacher data. That is, the machine learning model generation unit 35 refers to the teacher data table of the teacher data storage unit 37, uses each of the monitoring item data of the record whose estimated operation determination value exceeds the operation threshold as an explanatory variable, and sets the corresponding control data. The operation prediction model is trained as the objective variable. Further, the machine learning model generation unit 35 also performs learning in consideration of the weight in the learning of the operation prediction model as in the case of the operation determination model.
When the machine learning model generation unit 35 performs operation prediction model and operation judgment model machine learning, regression, tree (decision tree, regression tree, etc.), neural network, genetic programming, support vector machine, deep learning, ensemble learning (gradient) Any of the commonly used techniques may be used, including boosting, bagging, etc.).

FIG. 5 is a flowchart showing an operation example of the estimation processing of the control data of the shield excavator of the device control system according to the present embodiment.
The monitoring item data input unit 31 determines whether or not the measurement cycle is within the predetermined measurement cycle (step S1), proceeds to step S2 when the measurement cycle has elapsed, and proceeds to step S2 when the measurement cycle has not elapsed. Repeat the process.
Then, the monitoring item data input unit 31 inputs the monitoring item data (step S2), outputs the monitoring item data to the operation determination unit 32, and assigns a record number to the operation status data table of the operation status data storage unit 38. Write and memorize. Here, the operation status data table has the same configuration as the teacher data table already described.

The operation determination unit 32 reads an operation determination model from the learned model storage unit 36, inputs monitoring item data as an explanatory variable to the read operation determination model, and estimates an estimated operation determination value (step S3). The operation determination unit 32 outputs the estimated estimated operation determination value to the processing control unit 33, and writes and stores it in the operation status data table.
Then, the processing control unit 33 determines whether or not the supplied estimated operation determination value is equal to or greater than the operation threshold value (step S4), advances the process to step S5 when the operation threshold value is equal to or greater than the operation threshold value, and proceeds to the process when the operation threshold value is less than the operation threshold value. The process proceeds to step S1 (the operation prediction process is not performed).

The operation prediction unit 34 reads the operation prediction model from the learned model storage unit 36, inputs monitoring item data as an explanatory variable to the read operation prediction model, and estimates the estimation control data (step S5). The operation prediction unit 34 displays the estimated estimation control data on a display screen of an image display device (not shown) or the like, and notifies the estimation control data (step S6).
Then, the processing control unit 33 detects whether or not control for ending the estimation processing of the control data is performed from the input means (not shown) of the device control system (step S7), and if not, processing is performed. The process proceeds to step S1 and the process ends when the process ends.

FIG. 6 is a graph showing a comparison between the determination result of the operation determination model and the actual measurement in the present embodiment. The operation judgment model used for comparison was performed using the monitoring item data of the site where the actual shield excavation was performed. The operation judgment model was trained using the monitoring item data and the operation result (operation judgment unit) in the 1 to 273 rings at the shield site as teacher data. Then, each monitoring item data of the 238 to 294 rings was input to this operation determination model, and each estimated operation determination value was estimated. In each of FIGS. 6 (a) to 6 (d), the data distribution of the 238 to 294 rings is shown as a bar graph (the vertical axis on the left is the number of data), and the number of operations of the operator and the operation instruction (operation) output by the operation determination model. The number of predictions) was drawn with a line graph (the vertical axis on the right is the number of operations). The solid line is the number of times based on the estimated operation judgment value estimated by the operation judgment model, and the broken line is the number of times actually operated by a skilled operator.

Each of FIGS. 6 (a) and 6 (b) is a graph relating to the horizontal deviation and the vertical deviation from the planned alignment of the tip of the shield excavator 10, respectively, with the proximity ratio (%) to the control value on the horizontal axis. it's shown. Each of FIGS. 6 (c) and 6 (d) is a graph regarding an error between the azimuth angle and pitching angle of the shield excavator 10 and the planned value (the horizontal axis indicates the error angle). The actual number of operations of the operator and the number of operations based on the estimation of the operation determination model are divided into monitoring item data every 10 minutes, and the case where the operation is performed in these 10 minutes is counted as one time. Further, in each of FIGS. 6 (a), 6 (b), 6 (c) and 6 (d), the maximum of each of the horizontal deviation, the vertical deviation, the azimuth error and the pitching angle error every 10 minutes. Plot using values. In each of the graphs of FIGS. 6 (a), 6 (b), 6 (c) and 6 (d), the distribution of the number of times the operator actually operates and the number of operations based on the estimation of the operation determination model. Shows a similar tendency. Therefore, the operation determination model generated by machine learning can simulate the operation of a skilled operator, and the timing of performing the shield jack operation can be appropriately determined.

As described above, according to the present embodiment, even if the monitoring item data indicating the excavation status or environment obtained from the shield excavator 10 is a feature amount that is infrequently operated, the operation determination model is used as the monitoring item data. It is possible to estimate whether or not to perform operations in response to the presence or absence of operations by a skilled operator, and the operation prediction model is an operation amount operated by a skilled operator corresponding to the monitoring item data. Can be estimated as control data, so that even if the excavation status data is a feature amount that is infrequently operated, it is possible to obtain a desirable operation frequency and operation amount of a skilled operator.
Further, for the operation status data table of the operation status data storage unit 38, the record in which the operation determination value is written is used as the teacher data, and the operation determination model and the operation prediction model are added to the teacher data table of the teacher data storage unit 37. Relearning of each of the above may be performed by the machine learning model generation unit 35. In this case, the operation threshold is also reset.

A program for realizing the function of the shield excavator control system 30 of FIG. 1 in the present invention is recorded on a computer-readable recording medium, and the program recorded on the recording medium is read by the computer system and executed. By doing so, the operation determination and operation prediction processing, and the machine learning processing of each of the operation determination model and the operation prediction model may be performed. The term "computer system" as used herein includes hardware such as an OS (Operating System) and peripheral devices. In addition, the "computer system" shall also include a WWW (World Wide Web) system provided with a homepage providing environment (or display environment). The "computer-readable recording medium" is a portable medium such as a flexible disk, a magneto-optical disk, a ROM (Read Only Memory), a CD-ROM (Compact Disc --Read Only Memory), or a built-in computer system. A storage device such as a hard disk. Furthermore, a "computer-readable recording medium" is a volatile memory (RAM (Random Access)) inside a computer system that serves as a server or client when a program is transmitted via a network such as the Internet or a communication line such as a telephone line. It also includes those that hold the program for a certain period of time, such as Memory)).

Further, the program may be transmitted from a computer system in which this program is stored in a storage device or the like to another computer system via a transmission medium or by a transmission wave in the transmission medium. Here, the "transmission medium" for transmitting a program refers to a medium having a function of transmitting information, such as a network (communication network) such as the Internet or a communication line (communication line) such as a telephone line. Further, the above program may be for realizing a part of the above-mentioned functions. Further, a so-called difference file (difference program) may be used, which can realize the above-mentioned functions in combination with a program already recorded in the computer system.

10 ... Shield excavator 11 ... Skin plate 17 ... Screw conveyor 20 ... Propulsion jack 30 ... Shield excavator control system 31 ... Monitoring item data input unit 32 ... Operation judgment unit 33 ... Processing control unit 34 ... Operation prediction unit 35 ... Machine learning Model generation unit 36 ... Learned model storage unit 37 ... Teacher data storage unit 38 ... Operation status data storage unit D ... Earth pressure gauge

Claims (6)

An operation judgment model that outputs the operation probability of the shield excavator as an estimated value by inputting the feature quantity indicating the excavation status supplied from the shield excavator.
An operation prediction model that outputs the operation parameters of the shield excavator as an estimated value by inputting the feature quantity indicating the excavation status supplied from the shield excavator.
A shield excavator control system including a processing control unit that estimates the operation parameters for the operation prediction model when the operation determination model outputs an operation probability equal to or higher than a predetermined threshold value. The operation determination model uses the history table in which the acquired record of the excavation status is recorded as the teacher data, and machine-learns the presence or absence of the operation change as the input value using the excavation status data as the output value. The shield excavator control system according to claim 1, wherein the machine learning model outputs the operation probability corresponding to the data of the above as an estimated operation determination value. When learning the operation determination model, the operation determination model is trained by using a numerical value obtained by dividing the number of records whose operation is not changed by the number of records whose operation is changed as a weight in the history table. The shield excavator control system according to claim 2. Based on the learned operation determination model, the estimated value based on the teacher data is obtained, and the number of records TN in the first group in which the record whose operation is not changed is determined not to change the operation and the record whose operation is not changed is determined to change the operation. Each of the number of records FN in the two groups, the number of records TP in the third group where the operation-changed record is determined to change the operation, and the number of records FN in the fourth group where the operation-changed record is determined not to change the operation. The shield excavator control system according to any one of claims 1 to 3, wherein the threshold value for whether or not to perform an operation is set according to the number of data of the above. Teacher data in which the operation prediction model uses a feature amount indicating the excavation status of a record having an operation probability equal to or higher than the threshold value in the second group and the third group as an input, and an operation parameter of a shield excavator as an estimated value. The shield excavator control system according to claim 4, wherein the shield excavator control system is machine-learned by. An operation judgment process that outputs the operation probability of the shield excavator as an estimated value by inputting a feature amount indicating the excavation status supplied from the shield excavator by the operation judgment model.
An operation prediction process that outputs the operation parameters of the shield excavator as an estimated value by inputting the feature amount indicating the excavation status supplied from the shield excavator by the operation prediction model.
A shield characterized in that the processing control unit has a processing control process that causes the operation prediction model to estimate the operation parameters when the operation determination model outputs an operation probability equal to or higher than a predetermined threshold value. Excavator control method.

Images