A safety control method for three-arm jumbo tunnel excavation

Abstract

The application provides a safety control method for three-arm rock drilling jumbo tunnel excavation, which is suitable for the field of tunnel excavation and comprises image acquisition device installation and calibration, image acquisition, image transmission, image processing and identification, state feedback and decision, and manual review and emergency response. The safety control method for three-arm rock drilling jumbo tunnel excavation has precise danger identification and high safety, and can be widely applied to the field of tunnel excavation.

Description

Technical Field

[0001] This invention relates to a safety control method for tunnel excavation using a three-arm rock drilling rig, applicable to the field of tunnel excavation. Background Technology

[0002] Tunnel excavation plays a vital role in modern infrastructure construction. With the continuous improvement of underground space utilization, tunnel construction technology is also constantly evolving. Especially in the excavation of large-scale tunnel projects today, safety, construction efficiency, and quality control have become paramount. During tunnel excavation, the three-arm drilling rig is a common piece of machinery, primarily used for drilling operations. Its multi-arm design effectively improves drilling efficiency, but it also brings higher operational complexity and safety risks.

[0003] Traditional methods for controlling tunnel excavation safety typically rely on manual monitoring and experience-based judgment. This approach has several problems: First, manual monitoring is easily affected by environmental factors, such as low light levels and high dust levels inside the tunnel, which can impact the accuracy of monitoring. Second, experience-based judgment is highly subjective and prone to errors due to individual differences, thus affecting construction safety and project quality. Finally, traditional methods lack real-time capabilities and automation, making it impossible to respond promptly to potential risks, thereby increasing safety hazards.

[0004] With the rapid development of computer vision and deep learning technologies, the application of image recognition-based automated monitoring technology in tunnel construction has gradually become a research hotspot. By installing image acquisition equipment on a three-arm drilling rig, image data of the tunnel excavation face can be collected in real time. Deep learning models are then used to process and recognize these images, enabling real-time monitoring of the tunnel excavation status and timely warnings of potential risks. This technology not only improves the accuracy of monitoring but also reduces reliance on manual monitoring, significantly enhancing construction safety and efficiency.

[0005] Although some image recognition-based tunnel monitoring systems exist, they still have several limitations. For example, these systems are poorly adaptable to environmental changes and are easily affected by complex lighting conditions and dusty environments within tunnels, leading to decreased image quality and thus affecting recognition accuracy. Furthermore, existing systems often rely on a single image recognition model, which cannot adequately address the various complex situations that may arise during tunnel construction. Therefore, developing a more stable, reliable, and intelligent method for tunnel excavation safety control has significant practical importance and application value. Summary of the Invention

[0006] This invention proposes a safety control method for tunnel excavation using a three-arm rock drilling rig based on image acquisition and processing. By using a high dynamic range camera and image label prediction model installed on the rock drilling rig, image data of the tunnel excavation face is acquired and analyzed in real time, enabling automatic identification and feedback of the construction status. Combined with a dustproof sealing and cleaning system to adapt to the complex environment of tunnel excavation, this invention provides a more intelligent and efficient means of tunnel construction safety control.

[0007] The objective of this invention can be achieved by adopting the following technical solutions:

[0008] A safety control method for tunnel excavation using a three-arm rock drilling rig, characterized in that the method includes the following steps:

[0009] S101, Installation and calibration of image acquisition equipment;

[0010] The installation and calibration of the image acquisition equipment includes ensuring that the equipment can cover the tunnel excavation face at a suitable position on the three-arm rock drilling rig, performing initial calibration of the image acquisition equipment, and ensuring that the equipment can accurately capture clear images of the excavation face.

[0011] S102, Image Acquisition;

[0012] The image acquisition includes setting the acquisition frequency of the image acquisition device, that is, automatically acquiring one frame of image at certain time intervals. During the excavation process using a three-arm rock drilling rig, the image acquisition device automatically acquires image data of the current excavation face according to the set acquisition frequency.

[0013] S103, Image Transmission;

[0014] The image transmission includes the image acquisition device acquiring image data of the excavation face and transmitting it to the data processing center via wired transmission.

[0015] S104, Image Processing and Recognition;

[0016] The image processing and recognition includes the data processing center receiving image data of the excavation face, the data processing center having an embedded image label prediction model, inputting the image data of the excavation face into the image label prediction model to obtain label prediction results, the label prediction results being three types: stable state, alert state, and dangerous state;

[0017] S105, Status Feedback and Decision Making;

[0018] The status feedback and decision-making include the data processing center automatically generating feedback signals and executing corresponding actions based on the label prediction results. When the label prediction result is in a stable state, the three-arm rock drilling rig is allowed to continue normal excavation operations. If the label prediction result is in a vigilant state, the system will issue a warning signal to prompt the operator to strengthen monitoring. If the label prediction result is in a dangerous state, the system will immediately issue an alarm signal and automatically stop the excavation operation of the three-arm rock drilling rig, waiting for further processing by the operator.

[0019] S106, Manual review and emergency response;

[0020] The manual review and emergency response include the system automatically saving and recording input images with predicted alert or dangerous status, prompting operators to conduct manual review, and the operators deciding whether to stop or slow down the excavation operation of the three-arm drilling rig based on the review results.

[0021] Furthermore, in S101 above, the image acquisition device uses a high dynamic range camera, which can capture multiple images under different exposure conditions and synthesize them into a clearer image.

[0022] Furthermore, in S101 above, the image acquisition device adopts a dustproof enclosure and a start-up cleaning system to adapt to the tunnel excavation environment. The start-up cleaning system can periodically or as needed spray compressed air to remove dust from the lens surface of the image acquisition device and keep the lens clean.

[0023] Furthermore, in S104 above, the step of obtaining the image label prediction model is as follows:

[0024] a) Collect a certain number of tunnel excavation face images using image acquisition equipment to form image set C1. Use expert evaluation to label each image. The labels are mainly of three categories: stable state, alert state, and dangerous state. The number of images in each of these three states must be greater than 10% of the total number of images in the image set. Finally, obtain excavation face image set C2 containing the labels.

[0025] b) Preprocess the set of excavation face images C2 containing labels. The preprocessing includes removing blurry and low-quality images, as well as image enhancement and data normalization, ultimately obtaining a preprocessed set of excavation face images C3 containing labels.

[0026] c) The AlexNet architecture is selected, which contains 5 convolutional layers and 3 fully connected layers. The activation function used in the AlexNet architecture is the Mish function.

[0027] d) Model training is performed by dividing the image set C3 into a training set and a validation set in an 8:2 ratio. The training set images are used as input, and the training set image labels are used as output. The model is trained based on the AlexNet architecture to obtain a preliminary label prediction model M1.

[0028] e) Model evaluation and optimization: The preliminary label prediction model M1 is evaluated using the validation set. The evaluation metric is accuracy. The preliminary model M1 is adjusted based on the accuracy results to obtain the final label prediction model M2.

[0029] Furthermore, the mathematical form of the Mish function is shown in equation (1).

[0030] f(x)=x·tanh(softplus(x)) (1)

[0031] The beneficial effects of this invention are:

[0032] The beneficial effects of this invention are: (1) Improved safety of tunnel excavation; by installing image acquisition equipment on the three-arm drilling rig, the condition of the tunnel excavation face is monitored in real time. Combined with the analytical capabilities of the image label prediction model, potential risk conditions can be identified in a timely manner. When the system detects a dangerous or alert state, it automatically issues an alarm and stops the operation to prevent accidents, thereby greatly improving the safety of tunnel construction; (2) Reduced manual intervention and improved work efficiency; This invention adopts automated image acquisition and processing technology to replace the traditional manual monitoring method. Through an automated status feedback and decision-making system, the dependence on manual monitoring is reduced. Operators only need to review when prompted by the system, which greatly improves work efficiency and also reduces the risk of human judgment errors. (3) Accurate risk identification and prediction; Using the deep learning model of the AlexNet architecture for image processing and recognition, and adopting the Mish activation function, it can better capture image features and provide higher recognition accuracy. The continuous optimization and evaluation of the model ensures that the system can provide accurate label prediction results under various complex conditions, further improving the reliability of risk warning. Attached Figure Description

[0033] Figure 1 The flowchart of a safety control method for tunnel excavation using a three-arm rock drilling rig according to the present invention. Detailed Implementation

[0034] The specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings; it should be understood that the specific embodiments given herein are only for illustration and explanation of the present invention and cannot be used to limit the present invention.

[0035] The following is a specific embodiment of a safety control method for tunnel excavation using a three-arm rock drilling rig.

[0036] like Figure 1 The diagram shows a flowchart of a multi-source fusion water depth measurement method for a safety control method of a three-arm rock drilling rig tunnel excavation according to the present invention.

[0037] A safety control method for tunnel excavation using a three-arm rock drilling rig, characterized in that the method includes the following steps:

[0038] S101, Installation and calibration of image acquisition equipment;

[0039] The installation and calibration of the image acquisition equipment includes ensuring that the equipment can cover the tunnel excavation face at a suitable position on the three-arm rock drilling rig, performing initial calibration of the image acquisition equipment, and ensuring that the equipment can accurately capture clear images of the excavation face.

[0040] S102, Image Acquisition;

[0041] The image acquisition includes setting the acquisition frequency of the image acquisition device, that is, automatically acquiring one frame of image at certain time intervals. During the excavation process using a three-arm rock drilling rig, the image acquisition device automatically acquires image data of the current excavation face according to the set acquisition frequency.

[0042] S103, Image Transmission;

[0043] The image transmission includes the image acquisition device acquiring image data of the excavation face and transmitting it to the data processing center via wired transmission.

[0044] S104, Image Processing and Recognition;

[0045] The image processing and recognition includes the data processing center receiving image data of the excavation face, the data processing center having an embedded image label prediction model, inputting the image data of the excavation face into the image label prediction model to obtain label prediction results, the label prediction results being three types: stable state, alert state, and dangerous state;

[0046] S105, Status Feedback and Decision Making;

[0047] The status feedback and decision-making include the data processing center automatically generating feedback signals and executing corresponding actions based on the label prediction results. When the label prediction result is in a stable state, the three-arm rock drilling rig is allowed to continue normal excavation operations. If the label prediction result is in a vigilant state, the system will issue a warning signal to prompt the operator to strengthen monitoring. If the label prediction result is in a dangerous state, the system will immediately issue an alarm signal and automatically stop the excavation operation of the three-arm rock drilling rig, waiting for further processing by the operator.

[0048] S106, Manual review and emergency response;

[0049] The manual review and emergency response include the system automatically saving and recording input images with predicted alert or dangerous status, prompting operators to conduct manual review, and the operators deciding whether to stop or slow down the excavation operation of the three-arm drilling rig based on the review results.

[0050] Furthermore, in S101 above, the image acquisition device uses a high dynamic range camera, which can capture multiple images under different exposure conditions and synthesize them into a clearer image.

[0051] Furthermore, in S101 above, the image acquisition device adopts a dustproof enclosure and a start-up cleaning system to adapt to the tunnel excavation environment. The start-up cleaning system can periodically or as needed spray compressed air to remove dust from the lens surface of the image acquisition device and keep the lens clean.

[0052] Furthermore, in S104 above, the step of obtaining the image label prediction model is as follows:

[0053] a) Collect a certain number of tunnel excavation face images using image acquisition equipment to form image set C1. Use expert evaluation to label each image. The labels are mainly of three categories: stable state, alert state, and dangerous state. The number of images in each of these three states must be greater than 10% of the total number of images in the image set. Finally, obtain excavation face image set C2 containing the labels.

[0054] b) Preprocess the set of excavation face images C2 containing labels. The preprocessing includes removing blurry and low-quality images, as well as image enhancement and data normalization, ultimately obtaining a preprocessed set of excavation face images C3 containing labels.

[0055] c) The AlexNet architecture is selected, which contains 5 convolutional layers and 3 fully connected layers. The activation function used in the AlexNet architecture is the Mish function.

[0056] d) Model training is performed by dividing the image set C3 into a training set and a validation set in an 8:2 ratio. The training set images are used as input, and the training set image labels are used as output. The model is trained based on the AlexNet architecture to obtain a preliminary label prediction model M1.

[0057] e) Model evaluation and optimization: The preliminary label prediction model M1 is evaluated using the validation set. The evaluation metric is accuracy. The preliminary model M1 is adjusted based on the accuracy results to obtain the final label prediction model M2.

[0058] Furthermore, the mathematical form of the Mish function is shown in equation (1).

[0059] f(x)=x·tanh(softplus(x)) (1)

[0060] In the above embodiments, the present invention discloses a safety control method for tunnel excavation using a three-arm rock drilling rig; including image acquisition equipment installation and calibration, image acquisition, image transmission, image processing and recognition, status feedback and decision-making, and manual review and emergency response; the safety control method for tunnel excavation using a three-arm rock drilling rig proposed in this invention has accurate hazard identification and high safety, and can be widely applied in the field of tunnel excavation.

[0061] The above description is a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A safety control method for three-boom jumbo tunnel excavation, characterized by, The safety control method for tunnel excavation of the three-arm rock drilling jumbo comprises the following steps: 1) image acquisition device installation and calibration; The image acquisition device installation and calibration comprises ensuring that the device can cover the tunnel excavation face by installing the device at a suitable position of the three-arm rock drilling jumbo, and performing initial calibration on the image acquisition device to ensure that the device can accurately capture clear images of the excavation face. The image acquisition device is provided with a dustproof package and a start-up cleaning system to adapt to the tunnel excavation environment. The start-up cleaning system can periodically or as needed spray compressed air to remove dust from the lens surface of the image acquisition device, keeping the lens clean. 2) image acquisition; The image acquisition comprises setting the acquisition frequency of the image acquisition device, i.e. automatically acquiring an image every certain time interval. During the excavation process using the three-arm rock drilling jumbo, the image acquisition device automatically acquires image data of the current excavation face at the set acquisition frequency. 3) image transmission; The image transmission comprises the image acquisition device obtaining the image data of the excavation face and transmitting it to the data processing center in the form of wired transmission. 4) image processing and recognition; The image processing and recognition comprise the data processing center receiving the image data of the excavation face. The data processing center is embedded with an image label prediction model. The image data of the excavation face is input into the image label prediction model to obtain a label prediction result. The label prediction result is one of three states: stable state, alert state and dangerous state. The steps for obtaining the image label prediction model are as follows: a) Collect a certain number of tunnel excavation face images using the image acquisition device to form an image set C1. Use expert evaluation method to label the label of each picture. The label is three types of states, namely stable state, alert state and dangerous state. At the same time, the number of images of the three types of states is required to be greater than 10% of the number of the image set C1. Finally, the excavation face image set C2 containing labels is obtained, b) Preprocess the excavation face image set C2 containing labels. The preprocessing includes deleting blurred and low-quality images, image enhancement and data normalization. Finally, the preprocessed excavation face image set C3 containing labels is obtained, c) Select AlexNet architecture. The AlexNet architecture contains 5 convolutional layers and 3 fully connected layers. The activation function used in the AlexNet architecture is the Mish function, d) Model training. Divide the image set C3 into a training set and a validation set in a ratio of 8:

2. Use the training set, with the training set images as input and the training set image labels as output, to train the model based on the AlexNet architecture. Finally, a label prediction preliminary model M1 is obtained, e) Model evaluation and optimization. Use the validation set to evaluate the label prediction preliminary model M1. The model evaluation index uses precision rate. According to the precision rate result, the preliminary model M1 is adjusted. Finally, a label prediction model M2 is obtained. 5) state feedback and decision-making; The state feedback and decision include that according to the label prediction result, the data processing center automatically produces a feedback signal and performs a corresponding action, when the label prediction result is a stable state, the three-arm drill rig is allowed to continue normal excavation operation, if the label prediction result is a vigilant state, the system sends a warning signal to prompt the operator to strengthen monitoring, if the label prediction result is a dangerous state, the system immediately sends an alarm signal to automatically stop the excavation operation of the three-arm drill rig, and waits for further processing of the operator; 6) manual review and emergency response; The manual review and emergency response include that for the input image with the label prediction result being the vigilant state and the dangerous state, the system automatically saves and records, and prompts the operator to manually review, and the operator decides whether to stop or slow down the excavation operation of the three-arm drill rig according to the review result.

2. A safety control method for tunnel excavation by a three-boom jumbo according to claim 1, characterized in that, In the 1), the image acquisition device adopts a high dynamic range camera, and the high dynamic range camera can capture multiple pictures under different exposure conditions to form clearer imaging.

3. A safety control method for tunnel excavation by a three-boom jumbo according to claim 1, characterized in that, The mathematical form of the Mish function is shown in formula (1): （1）。

Images