KR-20260063199-A - METHOD AND APPARATUS FOR ESTIMATING OF BACKFILL CONDITION OF TBM TUNNEL

KR20260063199AKR 20260063199 AKR20260063199 AKR 20260063199AKR-20260063199-A

Abstract

The present invention relates to a method and apparatus for evaluating the back surface condition of a TBM tunnel. A method for evaluating the back surface condition according to an embodiment of the present invention comprises: (a) acquiring an original GPR image of the back surface of the TBM tunnel through exploration using Ground Penetrating Radar equipment; (b) inputting the original GPR image into an AI-based rebar noise removal model to generate a regenerated GPR image with rebar noise removed; and (c) inputting the regenerated GPR image into an AI-based cavity defect detection model to infer cavity regions within the regenerated GPR image; wherein the rebar noise removal model is generated through learning using a training input image corresponding to the original GPR image and a training ground truth image corresponding to the regenerated GPR image; and wherein the training input image and the training ground truth image are generated through a numerical analysis model based on the Finite-Difference Time-Domain method. Through this, void regions within original GPR images can be automatically detected using an AI-based learning model, without relying on subjective evaluations by experts.

Inventors

최항석
황채민
황병현
양승훈

Assignees

고려대학교 산학협력단

Dates

Publication Date: 20260507
Application Date: 20241030

Claims (8)

In a method for evaluating the back surface condition of a TBM tunnel, (a) a step of acquiring original GPR images of the back surface of the above TBM tunnel through exploration using Ground Penetrating Radar equipment, and (b) a step of inputting the above-mentioned original GPR image into an AI-based rebar noise removal model to generate a GPR regenerated image with rebar noise removed, and (c) including the step of inputting the above GPR regenerated image into an artificial intelligence-based cavity defect detection model to infer cavity regions within the above GPR regenerated image; The above rebar noise removal model is generated through learning using a training input image corresponding to the above GPR original image and a training ground truth image corresponding to the above GPR regenerated image; A method for evaluating the back face condition of a TBM tunnel, characterized in that the above-mentioned training input image and the above-mentioned training ground truth image are generated through a numerical analysis model based on the Finite-Difference Time-Domain method.
In paragraph 1, The process of generating the training input image and the training correct answer image through the above numerical analysis model The process of inputting information on the electromagnetic characteristics of the TBM tunnel segments constituting the above TBM tunnel, the reinforcing bars inside the same, and the backfill material filled on the back of the above TBM tunnel into the above numerical analysis model, and A process for constructing a tunnel simulation geometry that simulates the above TBM tunnel, the above backfill material, and a cavity area within the above backfill material, and The method includes a process for generating the learning input image and the learning correct answer image through a simulation of electromagnetic waves propagating through the tunnel simulation geometry; A method for evaluating the back surface condition of a TBM tunnel, characterized in that the above-mentioned correct answer image for learning is generated in the above-mentioned tunnel simulation geometry with the above-mentioned rebar removed.
In paragraph 1, A method for evaluating the back surface condition of a TBM tunnel, characterized in that the above-mentioned rebar noise removal model is generated based on a U-net-based learning model.
In paragraph 1, A method for evaluating the back surface condition of a TBM tunnel, characterized in that the above joint defect detection model is generated based on a YOLO-based learning model.
In a device for evaluating the back surface condition of a TBM tunnel, An AI-based rebar noise removal model that takes as input an original GPR image obtained through exploration of the back surface of a TBM tunnel using surface exploration laser equipment, and infers and outputs a regenerated GPR image in which rebar noise has been removed from the original GPR image - generated through learning using a training input image corresponding to the original GPR image and a training ground truth image corresponding to the regenerated GPR image - and; A cavity defect detection model that receives the above GPR regenerated image as input and infers cavity regions within the above GPR regenerated image; A device for evaluating the back face condition of a TBM tunnel, characterized by including a numerical analysis model that generates the above-mentioned training input image and the above-mentioned training correct image through numerical analysis based on the Finite-Difference Time-Domain method.
In paragraph 5, The above numerical analysis model Information on the electromagnetic characteristics of the TBM tunnel segments constituting the above TBM tunnel, the reinforcing bars inside them, and the backfill material filled on the back of the above TBM tunnel is received, and Constructing a tunnel simulation geometry that simulates the above TBM tunnel, the above backfill material, and a cavity area within the above backfill material, The training input image and the training correct answer image are generated through the simulation of electromagnetic waves propagating through the tunnel simulation geometry, and A device for evaluating the back surface condition of a TBM tunnel, characterized by generating the above-mentioned correct answer image for learning in the above-mentioned tunnel simulation geometry with the above-mentioned rebar removed.
In paragraph 5, A back surface condition evaluation device for a TBM tunnel characterized by the above-mentioned rebar noise removal model being generated based on a U-net-based learning model.
In paragraph 5, A back surface condition evaluation device for a TBM tunnel characterized by the above joint defect detection model being generated based on a YOLO-based learning model.

Description

Method and apparatus for estimating backfill condition of TBM tunnel The present invention relates to a method and apparatus for evaluating the condition of the back face of a TBM tunnel, and more specifically, to a method and apparatus for evaluating the condition of the back face of a TBM tunnel capable of evaluating the condition of the backfill material located on the back face of a TBM tunnel. Tunnels are essential transportation infrastructure for utilizing underground urban spaces, and therefore, the use of construction methods employing Shield Tunnel Boring Machines (hereinafter referred to as 'TBMs'), which produce very little vibration and noise pollution, is becoming increasingly active. The TBM method involves excavating through a cutter head at the front while simultaneously assembling TBM segments to form a concrete structure called the TBM tunnel lining. As tunnel construction proceeds using the TBM method, tail voids may form due to the difference between the outer surface of the excavation and the outer surface of the segments. To ensure the early stabilization of the concrete structure and prevent leakage, these voids are filled with backfill material. Commonly used backfill materials include mortar and cementitious bentonite. However, the backfill material may be damaged due to environmental factors such as surrounding ground conditions or erosion, as well as construction defects such as design errors in backfill injection volume and deterioration in construction quality. Consequently, this can lead to corrosion of the TBM segments and the occurrence of eccentric loading. Therefore, the condition of the backfill material behind the segments must be evaluated periodically for the maintenance of the TBM tunnel. Since visual inspection of the backfill surface of TBM tunnels is impossible, non-destructive geophysical exploration techniques must be adopted to evaluate the condition of the backfill material. Among non-destructive geophysical exploration techniques, Ground Penetrating Radar (GPR) can explore measurement sections rapidly and continuously, making it effective for use in long tunnels. The GPR technique radiates electromagnetic pulses toward a detection target and analyzes the waves reflected from backfill defects, such as cavities, thereby enabling the identification of defect locations. However, reinforcing bars embedded within TBM segments strongly reflect electromagnetic waves due to their electromagnetic properties, generating a significant amount of noise in the original GPR images measured by the equipment. This creates a problem that makes defect signal analysis difficult. Currently, detecting void areas within original GPR images relies on the visual perception of experts, but there is a problem in that the interpretation results may be biased depending on the expert because defects are identified through the subjective evaluation of experts. FIG. 1 is a drawing showing an example of a TBM tunnel structure in which a cavity area is formed, and FIG. 2 is a diagram showing an example of the configuration of a back surface condition evaluation device for a TBM tunnel according to an embodiment of the present invention, and Figure 3 is a diagram showing an example of a GPR image generated by ground exploration radar equipment, and FIG. 4 is a diagram showing an example of tunnel simulation geometry simulated by a numerical analysis model according to an embodiment of the present invention, and FIG. 5 is a diagram showing an example of a training data set generated by a numerical analysis model according to an embodiment of the present invention, and FIG. 6 is a diagram showing the original GPR image, the GPR regenerated image, and the final output image in the back surface condition evaluation device of a TBM tunnel according to an embodiment of the present invention. The advantages and features of the present invention and the methods for achieving them will become clear by referring to the embodiments described below in detail together with the accompanying drawings. However, the present invention is not limited to the embodiments disclosed below but may be implemented in various different forms. These embodiments are provided merely to ensure that the disclosure of the present invention is complete and to fully inform those skilled in the art of the scope of the present invention, and the present invention is defined only by the scope of the claims. The terms used in this specification are for describing embodiments and are not intended to limit the invention. In this specification, the singular form includes the plural form unless specifically stated otherwise in the text. The terms "comprises" and/or "comprising" used in this specification do not exclude the presence or addition of one or more other components in addition to the components mentioned. Throughout the specification, the same reference numerals refer to the same components, and "and/or" includes each of the me