TBM step change information acquisition method, TBM step change control method and TBM

Abstract

The invention relates to a TBM step change information acquisition method, a TBM step change control method and a TBM, and belongs to the field of TBM control. According to the method, multi-source data fusion and recognition are performed on images, point clouds and equipment monitoring data of the step changing mechanism and the step changing environment of the TBM acquired in real time, and recognition and three-dimensional space positioning of the step changing mechanism and the environment are realized. Firstly, a three-dimensional model is established based on multi-source data fusion processing and recognition results, secondly, three-dimensional space information of a TBM step changing mechanism and the environment is determined according to the three-dimensional model, and finally, the TBM step changing mechanism is controlled to act according to the three-dimensional space information for automatic step changing. The TBM is controlled to carry out automatic step changing according to intelligent identification and three-dimensional space information, TBM step changing whole-process intelligent control can be achieved, the construction efficiency can be improved, and construction manpower resources are saved.

Description

Technical Field

[0001] The present invention relates to a TBM step-change information acquisition method, a TBM step-change control method and a TBM, and belongs to the field of TBM control. Background Art

[0002] A hard rock full-face tunnel boring machine (TBM) is a large-scale tunnel construction equipment that integrates multiple systems, including mechanical, electrical, hydraulic, optical, and gas systems. It can be used for continuous tunnel construction operations such as excavation, support, and mucking. It has the advantages of rapid excavation, environmental protection, and high overall efficiency. It is widely used in mountain tunnels, pumped storage, water diversion, and other tunnel projects at home and abroad.

[0003] TBM construction involves tunneling, support, and step-changing processes. Tunneling is used to excavate the tunnel face, support is used for initial support during tunneling, and step-changing is used to transition between tunneling cycles. Currently, the lead driver and observer control the TBM's step-changing mechanism based on TBM design requirements and site conditions to achieve step-changing. Due to the complex and ever-changing TBM site environment, manual step-changing relies on on-site experience and carries risks for both personnel and construction.

[0004] Chinese patent document CN117127994A discloses a method for automatically positioning and changing the steps of TBM grippers based on image recognition. This method uses a camera to monitor the gripper position in real time through image recognition, and determines the positional relationship between the gripper, the arch frame, and the rock wall through two-dimensional images. Finally, the method achieves the purpose of automatic positioning and changing the steps of the gripper by controlling the retraction and extension of the gripper and the propulsion cylinder.

[0005] In the above scheme, since the acquired image information does not include depth direction information, for example, the acquired position information of the gripper shoe and the arch frame does not include depth information, the three-dimensional spatial information of the gripper shoe and the arch frame cannot be determined. During the tightening process of the gripper shoe, there is a risk of the gripper shoe colliding with or stepping on the arch frame, which can easily cause damage to the arch frame and lead to quality and safety accidents. Summary of the Invention

[0006] The purpose of the present invention is to provide a TBM step change information acquisition method, a TBM step change control method and a TBM, so as to solve the problem of inaccurate relative position information between the constructed support shoe and the arch frame during the automatic step change of the TBM.

[0007] To achieve the above object, the solution of the present invention includes: A method for obtaining TBM step change information of the present invention comprises the following steps: 1) Real-time acquisition of multi-source heterogeneous data from the TBM, including images, point clouds, and equipment monitoring data of the TBM's step-changing mechanism and the step-changing environment at the step-changing location; 2) Data fusion is performed on multi-source heterogeneous data to identify the step-changing mechanism and environmental objects at the step-changing position to obtain a three-dimensional step-changing model. The three-dimensional step-changing model includes the three-dimensional spatial information of the step-changing mechanism and environmental objects. Based on the three-dimensional spatial information and the TBM's step-changing detection data, the TBM's automatic step-changing control decision instructions are obtained. The automatic step-changing control decision instructions are used to implement the action instructions of the step-changing mechanism, the moving target position, the recommended excavation stroke, and the adjustment of the arch posture and spacing.

[0008] Furthermore, the data fusion and recognition methods include: feature layer fusion, decision fusion or decision + feature layer fusion; the feature layer fusion process includes: using a continuous convolution algorithm to extract data features of image data and point cloud data respectively, and adding the data features corresponding to the extracted image information to the point cloud data through the mapping relationship between the point cloud data and the image data, and then using the point cloud object recognition network to identify the step-changing mechanism and environmental objects; the decision fusion process includes: obtaining corresponding recognition results according to the image data and point cloud data respectively, converting the recognition results to a unified coordinate system and then applying Kalman filtering and IOU calculation algorithm to generate the final recognition results of the step-changing mechanism and environmental objects, and obtaining a three-dimensional step-changing model according to the final recognition results; the decision + feature layer fusion process includes: generating candidate boxes of the step-changing mechanism and environmental objects according to the image data or point cloud data, and combining the candidate boxes of the image data or the candidate boxes of the point cloud data with the point cloud data or image data to identify the step-changing mechanism and environmental objects.

[0009] Furthermore, the step-changing environment includes image data and point cloud data of the rock wall used to tighten the boots during the step-changing process; the multi-source heterogeneous data are fused and identified to determine whether there is a cavity on the rock wall. When a cavity exists, the three-dimensional step-changing model includes the three-dimensional spatial information of the cavity. The three-dimensional spatial information of the cavity and the identification result are used to adjust the position of the moving target to avoid the cavity or to perform environmental risk identification and early warning.

[0010] Furthermore, the step-changing environment also includes image data and point cloud data of the arch frame in the support shoe tightening area; the fusion and identification of multi-source heterogeneous data also includes whether there is an arch frame in the support shoe tightening area. When there is an arch frame, the three-dimensional step-changing model also includes the step-changing mechanism and the three-dimensional spatial information of the arch frame in the support shoe tightening area. The three-dimensional information of the step-changing mechanism and the arch frame and the recognition results are used to adjust the position of the moving target to avoid the arch frame or adjust the posture and spacing of the arch frame.

[0011] Furthermore, the step-changing environment also includes image data and point cloud data of the landing area of ​​the rear support; the fusion and identification of multi-source heterogeneous data also includes whether there is an arch or falling rocks in the landing area; when there is an arch, the three-dimensional step-changing model also includes the three-dimensional spatial information of the arch in the landing area, and the three-dimensional spatial information of the arch in the landing area is also used to obtain the recommended excavation process so that the rear support avoids the arch; when there is falling rocks, the three-dimensional step-changing model also includes the three-dimensional spatial information of the falling rocks, and the three-dimensional spatial information of the falling rocks and the identification results of the falling rocks are used to identify and warn of environmental risks in the landing area.

[0012] A TBM step change control method includes the following steps: 1) Real-time acquisition of multi-source heterogeneous data from the TBM, including images, point clouds, and equipment monitoring data of the TBM's step-changing mechanism and the step-changing environment at the step-changing location; 2) Data fusion is performed on multi-source heterogeneous data to identify the step-changing mechanism and environmental objects at the step-changing location, resulting in a 3D step-changing model. The 3D step-changing model includes 3D spatial information of the step-changing mechanism and environmental objects. Based on this 3D spatial information and the TBM's step-changing detection data, automatic step-changing control decision instructions are generated for the TBM. 3) Automatic step change is performed according to the automatic step change control decision instruction, which includes the action instruction of the step change mechanism, the moving target position, the recommended excavation stroke and the arch posture and spacing adjustment.

[0013] Furthermore, the data fusion and recognition methods include: feature layer fusion, decision fusion or decision + feature layer fusion; the feature layer fusion process includes: using a continuous convolution algorithm to extract data features of image data and point cloud data respectively, and adding the data features corresponding to the extracted image information to the point cloud data through the mapping relationship between the point cloud data and the image data, and then using the point cloud object recognition network to identify the step-changing mechanism and environmental objects; the decision fusion process includes: obtaining corresponding recognition results according to the image data and point cloud data respectively, converting the recognition results to a unified coordinate system and then applying Kalman filtering and IOU calculation algorithm to generate the final recognition results including the step-changing mechanism and environmental objects, and obtaining a three-dimensional step-changing model according to the final recognition results; the decision + feature layer fusion process includes: generating candidate boxes including the step-changing mechanism and environmental objects according to the image data or point cloud data, and combining the candidate boxes of the image data or the candidate boxes of the point cloud data with the point cloud data or image data to identify the step-changing mechanism and environmental objects.

[0014] Furthermore, the step-changing environment includes image data and point cloud data of the rock wall used to tighten the boots in the step-changing process; multi-source heterogeneous data are fused and identified, including whether there is a cavity on the rock wall. When a cavity exists, the three-dimensional step-changing model includes the three-dimensional spatial information of the cavity, and the area and / or volume of the cavity is calculated based on the three-dimensional spatial information of the cavity. When the area and / or volume of the cavity is greater than the corresponding set threshold, the step-changing target position is adjusted to avoid the cavity or an environmental risk identification warning is performed.

[0015] Furthermore, the step-changing environment also includes image data and point cloud data of the arch frame in the support shoe tightening area; the fusion and identification of multi-source heterogeneous data also includes whether there is an arch frame in the support shoe tightening area. When there is an arch frame, the three-dimensional step-changing model also includes the step-changing mechanism and the three-dimensional spatial information of the arch frame in the support shoe tightening area. According to the three-dimensional information of the step-changing mechanism and the arch frame and the recognition results, the position of the moving target is adjusted to avoid the arch frame or the posture and spacing of the arch frame are adjusted.

[0016] Furthermore, the step-changing environment also includes image data and point cloud data of the landing area of ​​the rear support; the fusion and identification of multi-source heterogeneous data also includes whether there is an arch or falling rocks in the landing area; when there is an arch, the three-dimensional step-changing model also includes the three-dimensional spatial information of the arch in the landing area, and the recommended excavation process is obtained based on the three-dimensional spatial information of the arch in the landing area, and then the rear support is made to avoid the arch according to the recommended excavation process; when there is falling rocks, the three-dimensional step-changing model also includes the three-dimensional spatial information of the falling rocks, and the volume of the falling rocks is calculated based on the three-dimensional spatial information of the falling rocks. When the volume of the falling rocks is greater than the corresponding set threshold, the rear support landing environment risk identification warning is triggered.

[0017] A TBM adopts the above-mentioned TBM step-changing control method when the TBM changes steps.

[0018] The beneficial effects of the present invention are as follows: the present invention is a pioneering invention, which performs multi-source data fusion and identification on the images, point clouds, and equipment monitoring data of the TBM's step-changing mechanism and step-changing environment acquired in real time, thereby realizing the identification and three-dimensional spatial positioning of the step-changing mechanism and the environment. First, a three-dimensional model is established based on the multi-source data fusion processing and identification results. Second, the three-dimensional spatial information of the TBM's step-changing mechanism and environment is determined based on the three-dimensional model. Finally, the TBM's step-changing mechanism is controlled to automatically change steps based on the three-dimensional spatial information. The present invention controls the TBM to automatically change steps based on intelligent identification and three-dimensional information, which can realize intelligent control of the entire TBM step-changing process, improve construction efficiency, and save construction human resources. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] Figure 1 This is a schematic diagram of the TBM step-changing system architecture of the present invention; Figure 2This is a schematic diagram of the architecture of a TBM step-changing communication system of the present invention; Figure 3 It is a step-changing flow diagram of the present invention. DETAILED DESCRIPTION

[0020] In order to make the objectives, technical solutions and advantages of the present invention more clear, the present invention is described in detail below in detail with reference to the accompanying drawings and embodiments.

[0021] The present invention aims to construct a real-time 3D step-change model of a TBM during tunneling. Based on this model, real-time 3D spatial information of the step-change mechanism and surrounding objects is obtained. This real-time 3D spatial information can be used to generate automatic step-change control decision instructions for the TBM.

[0022] Method Example 1: This embodiment proposes a method for obtaining TBM step change information, such as Figure 1 and Figure 2 As shown, it includes multi-source heterogeneous data sensing terminal, equipment and component status monitoring, rock wall cavity identification and measurement, arch and support shoe identification and spatial positioning, rear support landing point identification and detection, motion control system and communication system. The specific execution steps are as follows: S1. Multi-source heterogeneous data sensing terminal: The sensing terminals installed on the TBM, including pressure sensors, displacement sensors, cameras, and radars, sense the equipment and environment in real time and transmit the sensed multi-source heterogeneous data to the data processing workstation. The cameras are used to collect 2D / 3D image data, the radar is used to collect point cloud data, and the data sensed by the pressure and displacement sensors are used as TBM equipment monitoring data. S2. Equipment and component status monitoring: By processing and analyzing the multi-source heterogeneous data obtained in S1, the operating status of equipment and components is determined and an exception handling strategy is set. S3. Rock wall cavity identification and measurement: Based on the multi-source heterogeneous data acquired in S1, 3D reconstruction is used to determine whether a cavity exists in the rock wall, locate it spatially, and calculate the volume and area of ​​the cavity. S4. Arch and gripper identification and spatial positioning: Based on the multi-source heterogeneous data acquired in S1, the arch and gripper are identified and spatially positioned through multi-source data fusion and spatial positioning data processing. S5. Rear support landing point identification and detection: Based on the multi-source heterogeneous data obtained in S1, multi-source data fusion and spatial positioning data processing are used to determine whether there are arches, fallen rocks, foreign objects, etc. at the rear support landing point, and spatial positioning and measurement are performed. S6. Motion control system: The industrial control system controls the motion of the TBM step-changing mechanism based on the intelligent identification and decision-making results of S3 to S5 and the TBM step-changing design process, achieving intelligent control of the entire TBM step-changing process. S7. Communication system: The multi-source heterogeneous data sensing terminal, industrial control system, data processing workstation, and TBM step-changing mechanism are electrically connected. The multi-source heterogeneous sensing data is transmitted from the sensing terminal to the data processing workstation via a switch. The intelligent recognition decision results are sent from the data processing workstation to the industrial control system via a switch. The TBM step-changing mechanism executes actions according to the control instructions of the industrial control system.

[0023] The specific steps of step S1 in this embodiment are: Real-time acquisition of images, point clouds, and equipment monitoring data of the TBM's step-changing mechanism and the step-changing environment at the step-changing position. The step-changing mechanism includes support shoes, rear supports, saddles, thrust cylinders, torque cylinders, etc. The step-changing environment includes rock walls, arches, fallen rocks, foreign objects, etc.

[0024] The specific steps of step S3 in this embodiment are: S31. Acquisition of 3D Rockwall Data: During TBM excavation, the multi-source heterogeneous data perception module transmits real-time scanned rockwall point cloud data, image data, and TBM excavation data from the gripper area to the data processing workstation for update and storage, ensuring the data processing workstation receives the latest rockwall and excavation data. S32. Heterogeneous Data Matching: Perform data matching processing on rock face point cloud information, image data, and TBM excavation data to generate fused data that captures rock face structure and excavation attributes. S33. 3D Modeling: Apply 3D reconstruction technology to the fused data to create a 3D physical model of the rock face. S34. Cavity Identification and Measurement: A 3D step-change model is generated by combining the 3D physical model of the rock face with the tunnel design model. During this fusion process, the presence of cavities in the rock face is identified. If present, the 3D step-change model includes 3D spatial information of the cavities in the gripper's gripper area. This 3D spatial information provides a theoretical basis for the TBM automatic step-change system to identify rock face environments. The specific steps of step S4 in this embodiment are: S41. Multi-source heterogeneous data acquisition: During the TBM step change period, the multi-source heterogeneous data perception module transmits real-time image data and point cloud data of the gripper area to the data processing workstation, where it is updated and stored. This ensures that the data processing workstation receives the latest image data and point cloud data of the gripper area. S42. Multi-source heterogeneous data preprocessing: Perform data preprocessing on the image data and scanned point cloud data acquired in S41, including data cleaning, normalization, and calibration. S43. Multi-source data fusion and recognition: Strategies for multi-source data fusion and recognition of arch supports, support boots, and rock walls include: feature layer fusion, decision fusion, or decision + feature layer fusion.

[0025] Feature layer fusion: 1) Apply a continuous convolution algorithm to extract data features from the preprocessed image data and point cloud data. 2) Use the mapping relationship between the point cloud and the image to append the image data features extracted by the convolutional network to the preprocessed point cloud data. Then, use the point cloud object recognition network to generate the final arch, support shoe, and rock face recognition results.

[0026] Decision fusion: Obtain corresponding object recognition results based on preprocessed image data and point cloud data. After conversion to a unified coordinate system, Kalman filtering, IOU calculation and other algorithms are applied to generate the final arch, support shoe and rock wall recognition results.

[0027] Decision-making + feature layer fusion: Based on the preprocessed image data and point cloud data, candidate frames for the arch, sprag, and rock wall are generated respectively. The candidate frames are then combined with another type of data to generate the final arch, sprag, and rock wall recognition results.

[0028] S44. Spatial positioning: Based on the multi-source data fusion and recognition results generated by S43, a three-dimensional step-changing model is obtained. The three-dimensional step-changing model includes the three-dimensional spatial information of the arch frame, the support shoe, and the rock wall in the support shoe area. The three-dimensional spatial information of the arch frame, the support shoe, and the rock wall is obtained, and the horizontal, longitudinal, and vertical three-dimensional spacing between the arch frame and the support shoe is calculated. This provides a calculation theoretical basis for the action of the TBM automatic step-changing system, and the final decision instructions are sent to the industrial control system.

[0029] The specific steps of step S5 in this embodiment are: S51. Multi-source heterogeneous data acquisition: Before the TBM changes steps, the multi-source heterogeneous data perception module transmits the real-time image data and point cloud data of the rear support landing area to the data processing workstation, where it is updated and stored. This ensures that the data processing workstation has the latest image data and point cloud data of the rear support landing area. S52. Multi-source heterogeneous data preprocessing: Perform data preprocessing on the images and scanned point cloud data acquired in S51, including data cleaning, normalization, and calibration. S53. Multi-source data fusion and recognition: The strategies supporting multi-source data fusion and recognition include: feature layer fusion, decision fusion, or decision + feature layer fusion.

[0030] Feature layer fusion: 1) Apply a continuous convolution algorithm to extract data features from preprocessed image data and point cloud data. 2) Use the mapping relationship between point cloud and image to append the image data features extracted by the convolutional network to the preprocessed point cloud data. Then, use the point cloud object recognition network to generate the final arch support, rear support, fallen rocks, and foreign object recognition results.

[0031] Decision fusion: Obtain corresponding object recognition results based on preprocessed image data and point cloud data, convert them into a unified coordinate system, and then apply Kalman filtering, IOU calculation and other algorithms to generate the final arch, rear support, fallen rocks and foreign object recognition results.

[0032] Decision-making + feature layer fusion: Based on the preprocessed image data and point cloud data, candidate frames for the arch, rear support, and fallen rocks are generated respectively. The candidate frames are then combined with another type of data to generate the final arch, rear support, fallen rocks, and foreign object recognition results.

[0033] S54. Spatial positioning: Based on the multi-source data fusion recognition results generated by S53, determine whether there is an arch frame, fallen rocks and foreign objects in the landing area of ​​the rear support. When there is an arch frame, the three-dimensional step-changing model also includes the three-dimensional spatial information of the rear support and the arch frame at the landing position, obtains the three-dimensional spatial information of the rear support and the arch frame, and calculates the three-dimensional horizontal, longitudinal and vertical distances between the rear support and the arch frame based on the three-dimensional spatial information of the rear support and the arch frame; when there are fallen rocks and other foreign objects, the three-dimensional step-changing model also includes the three-dimensional spatial information of the fallen rocks and other foreign objects, and calculates the volume of the fallen rocks and other foreign objects based on the three-dimensional spatial information of the fallen rocks and other foreign objects. The three-dimensional distance between the rear support and the arch frame and the volume of the fallen rocks and other foreign objects provide a calculation theoretical basis for the action of the TBM automatic step-changing system, avoiding the safety risks in the falling process of the rear support.

[0034] Specifically, in step S6, the motion control system combines the intelligent identification and decision-making results from steps S3 to S5, the test data, and the TBM step-change design process to generate automatic step-change decision instructions. Based on these automatic step-change decision instructions, the industrial control system controls the TBM step-change mechanism, achieving intelligent control of the entire TBM step-change process. This includes providing step-change mechanism action instructions (e.g., gripper control), target position, recommended excavation path, and adjustment of arch posture and spacing. The rock wall cavity identification and measurement, arch identification, and spatial positioning results from step S3 determine the motion control system's target position. When the area and / or volume of a cavity exceeds a corresponding threshold, the target position is adjusted to avoid the cavity or a risk warning is issued. When arch and gripper adjustments are necessary, the motion control system calculates the three-dimensional horizontal, longitudinal, and vertical spacing between the arch and grippers based on step S4, adjusts the motion control system's target position, and adjusts the arch posture and spacing by controlling the propulsion and gripper movements. The rear support landing point identification detection result of step S5 determines whether the front and rear support landing point areas before and after the step change need to be subject to environmental risk identification and warning and excavation process recommendation, that is, when the volume of falling rocks is greater than the corresponding set threshold, the environmental risk identification and warning of the rear support landing point area is triggered; when there is an arch frame in the landing point area, a recommended excavation process is obtained, and the rear support is adjusted to avoid the arch frame according to the recommended excavation process. For example, when there is an arch frame at the next step change position of the rear support, the recommended excavation process is calculated. The recommended excavation process needs to ensure that the arch frame is on both sides of the rear support or in the groove to prevent the rear support from colliding with the arch frame.

[0035] like Figure 2 The figure shows a flow chart of the TBM step-changing data processing process. Specifically, the communication system mentioned in step S7, the 2D / 3D camera, the radar data, and the industrial control system receive data from other sensing terminals, including pressure and displacement. The data is transmitted to a gigabit switch via TCP, and then transmitted from the gigabit switch to a data processing workstation. After processing at the data processing workstation, the intelligent recognition decision result is transmitted to the industrial control system via the gigabit switch. The industrial control system issues commands to control the TBM step-changing mechanism, realizing intelligent control of the entire TBM step-changing process.

[0036] Method Example 2: This embodiment provides a TBM step-changing control method, such as Figure 3 As shown, the specific step-changing control process has been clearly described in the method embodiment 1, so it will not be repeated here.

[0037] TBM Example: This embodiment provides a TBM. When performing step-changing control on the TBM that is excavating, the method described in Method Example 2 is adopted. Since the introduction of this method is clear enough, it will not be repeated here.

Claims

1. A method for obtaining TBM step change information, characterized in that: The steps include: 1) Real-time acquisition of multi-source heterogeneous data from the TBM, including images, point clouds, and equipment monitoring data of the TBM's step-changing mechanism and the step-changing environment at the step-changing location; 2) Performing data fusion on the multi-source heterogeneous data and identifying the step-changing mechanism and environmental objects at the step-changing position to obtain a three-dimensional step-changing model, wherein the three-dimensional step-changing model includes three-dimensional spatial information of the step-changing mechanism and the environmental objects; obtaining automatic step-changing control decision instructions for the TBM based on the three-dimensional spatial information and the TBM equipment monitoring data, wherein the automatic step-changing control decision instructions are used to implement the action instructions of the step-changing mechanism, the moving target position, the recommended excavation stroke, and the adjustment of the arch posture and spacing.

2. The TBM step change information acquisition method according to claim 1, characterized in that: Methods of data fusion and recognition include: feature layer fusion, decision fusion, or decision + feature layer fusion; the feature layer fusion process includes: using a continuous convolution algorithm to extract data features of image data and point cloud data respectively, adding the data features corresponding to the extracted image information to the point cloud data through the mapping relationship between the point cloud data and the image data, and then using the point cloud object recognition network to identify the step-changing mechanism and environmental objects; the decision fusion process includes: obtaining corresponding recognition results according to the image data and point cloud data respectively, converting the recognition results into a unified coordinate system, and then applying Kalman filtering and IOU calculation algorithm to generate the final recognition results of the step-changing mechanism and environmental objects, and obtaining the three-dimensional step-changing model according to the final recognition results; the decision + feature layer fusion process includes: generating candidate frames of the step-changing mechanism and environmental objects according to the image data or point cloud data, combining the candidate frames of the image data or the candidate frames of the point cloud data with the point cloud data or image data to identify the step-changing mechanism and environmental objects.

3. The TBM step change information acquisition method according to claim 1, characterized in that: The step-changing environment includes image data and point cloud data of the rock wall used to tighten the boot during the step-changing process; the multi-source heterogeneous data are fused and identified to determine whether there is a cavity on the rock wall. When a cavity exists, the three-dimensional step-changing model includes the three-dimensional spatial information of the cavity. The three-dimensional spatial information of the cavity and the identification result are used to adjust the position of the moving target to avoid the cavity or to perform environmental risk identification and early warning.

4. The TBM step change information acquisition method according to claim 3, characterized in that: The step-changing environment also includes image data and point cloud data of the arch frame in the gripping shoe tightening area; the fusion and identification of multi-source heterogeneous data also includes whether there is an arch frame in the gripping shoe tightening area. When there is an arch frame, the three-dimensional step-changing model also includes three-dimensional spatial information of the step-changing mechanism and the arch frame in the gripping shoe tightening area. The three-dimensional information of the step-changing mechanism and the arch frame and the recognition results are used to adjust the position of the moving target to avoid the arch frame or adjust the posture and spacing of the arch frame.

5. The method for obtaining TBM step change information according to claim 4, characterized in that: The step-changing environment also includes image data and point cloud data of the landing area of ​​the rear support; the fusion and identification of multi-source heterogeneous data also includes whether there is an arch or falling rocks in the landing area; when there is an arch, the three-dimensional step-changing model also includes the three-dimensional spatial information of the arch on the landing area, and the three-dimensional spatial information of the arch on the landing area is also used to obtain a recommended excavation process so that the rear support avoids the arch; when there is falling rocks, the three-dimensional step-changing model also includes the three-dimensional spatial information of the falling rocks, and the three-dimensional spatial information of the falling rocks and the identification results of the falling rocks are used to identify and warn of environmental risks in the landing area.

6. A TBM step-change control method, characterized in that: The steps include: 1) Real-time acquisition of multi-source heterogeneous data from the TBM, including images, point clouds, and equipment monitoring data of the TBM's step-changing mechanism and the step-changing environment at the step-changing location; 2) Fusing the multi-source heterogeneous data and identifying the step-changing mechanism and environmental objects at the step-changing location to obtain a three-dimensional step-changing model, which includes three-dimensional spatial information of the step-changing mechanism and environmental objects; and obtaining automatic step-changing control decision instructions for the TBM based on the three-dimensional spatial information and TBM equipment monitoring data. 3) Automatically changing steps according to the automatic step-changing control decision instruction, wherein the automatic step-changing control decision instruction includes the action instruction of the step-changing mechanism, the moving target position, the recommended excavation stroke and the arch posture and spacing adjustment.

7. The TBM step-changing control method according to claim 6, characterized in that: Methods for data fusion and recognition include: feature layer fusion, decision fusion, or decision + feature layer fusion; the feature layer fusion process includes: using a continuous convolution algorithm to extract data features of image data and point cloud data respectively, adding the data features corresponding to the extracted image information to the point cloud data through the mapping relationship between the point cloud data and the image data, and then using the point cloud object recognition network to identify the step-changing mechanism and environmental objects; the decision fusion process includes: obtaining corresponding recognition results based on the image data and point cloud data respectively, converting the recognition results into a unified coordinate system, and then applying Kalman filtering and IOU calculation algorithm to generate a final recognition result including the step-changing mechanism and environmental objects, and obtaining the three-dimensional step-changing model based on the final recognition result; the decision + feature layer fusion process includes: generating candidate boxes including the step-changing mechanism and environmental objects according to the image data or point cloud data, combining the candidate boxes of the image data or the candidate boxes of the point cloud data with the point cloud data or the image data to identify the step-changing mechanism and environmental objects.

8. The TBM step-changing control method according to claim 6, characterized in that: The step-changing environment includes image data and point cloud data of the rock wall used to tighten the boot during the step-changing process; multi-source heterogeneous data are fused and identified, including whether there is a cavity on the rock wall. When a cavity exists, the three-dimensional step-changing model includes three-dimensional spatial information of the cavity, and the area and / or volume of the cavity is calculated based on the three-dimensional spatial information of the cavity. When the area and / or volume of the cavity is greater than the corresponding set threshold, the step-changing target position is adjusted to avoid the cavity or an environmental risk identification warning is performed.

9. The TBM step-changing control method according to claim 8, characterized in that: The step-changing environment also includes image data and point cloud data of the arch frame in the gripper shoe tightening area; the fusion and identification of multi-source heterogeneous data also includes whether there is an arch frame in the gripper shoe tightening area. When there is an arch frame, the three-dimensional step-changing model also includes the three-dimensional spatial information of the step-changing mechanism and the arch frame in the gripper shoe tightening area. According to the three-dimensional information of the step-changing mechanism and the arch frame and the identification results, the position of the moving target is adjusted to avoid the arch frame or the posture and spacing of the arch frame are adjusted.

10. The TBM step-changing control method according to claim 9, characterized in that: The step-changing environment also includes image data and point cloud data of the landing area of ​​the rear support. The fusion and identification of multi-source heterogeneous data also includes whether there is an arch or falling rocks in the landing area. When an arch exists, the three-dimensional step-changing model also includes three-dimensional spatial information of the arch in the landing area. A recommended excavation process is obtained based on the three-dimensional spatial information of the arch in the landing area, and the rear support is then moved away from the arch according to the recommended excavation process. When there is falling rock, the three-dimensional step-changing model also includes the three-dimensional spatial information of the falling rock. The volume of the falling rock is calculated based on the three-dimensional spatial information of the falling rock. When the volume of the falling rock is greater than the corresponding set threshold, the rear support landing point environmental risk identification warning is triggered.

11. A TBM, characterized in that: When the TBM changes steps, the TBM step-changing control method according to any one of claims 6 to 10 is adopted.

Images