Multi-target recognition and safety warning system for deep shaft hoisting system

Abstract

The utility model discloses a kind of multi-target identification and safety early warning system of deep vertical shaft hoisting system, including the lateral pressure type tension sensor being arranged on winch steel wire rope, the temperature and humidity sensor being arranged in vertical shaft wall, the visual sensor being arranged on vertical shaft gantry, the data receiving and processor being arranged in well mouth, and early warning execution mechanism;Each sensor is connected with data receiving and processor respectively, and is connected with early warning execution mechanism, and data receiving and processor are used to analyze visual data and multi-source physical parameter.The utility model breaks through the single function limitation of traditional monitoring scheme through multi-source data fusion architecture, realizes 30 frames / second real-time target identification relying on lightweight model, in combination with timestamp alignment mechanism and closed-loop early warning logic, in low-illuminance, high-dust deep vertical shaft environment Realize the safety upgrade from post-passive record to pre-active interception, effectively guarantee the life safety of operating personnel and equipment stable operation.

Description

Technical Field

[0001] This utility model belongs to the field of building construction safety monitoring technology, and relates to the field of deep shaft safety monitoring technology, specifically to a multi-target identification and safety early warning system for a deep shaft hoisting system. Background Technology

[0002] Deep shaft construction environments present multiple physical challenges. For example, in the construction of deep shafts in hydropower station tunnels or metal mines, the dense metal structures in the environment cause severe attenuation of wireless signals, high dust concentrations blur the field of vision of visual sensors, and insufficient lighting conditions in the shaft cause target recognition functions to fail. At the same time, the system needs to monitor dynamic risk sources simultaneously, including sudden changes in the mechanical state of the wire rope, deviation of the hook's movement trajectory, lack of personnel safety equipment, and abnormal environmental parameters; however, existing technologies cannot achieve multi-target collaborative tracking, resulting in a high rate of missed detection of key risks.

[0003] Although some monitoring systems have been researched and designed, existing monitoring solutions have many shortcomings: they are functionally limited, focusing only on equipment location monitoring while neglecting personnel behavior and environmental conditions; they have weak environmental adaptability, with wireless transmission becoming unstable in vertical shaft environments; and their early warning mechanisms are incomplete, lacking multi-source data fusion and closed-loop control capabilities. These deficiencies prevent the systems from meeting the requirements of construction safety regulations for risk prediction and immediate intervention, resulting in response delays and protection gaps. Utility Model Content

[0004] To address the shortcomings of the existing technology, this utility model discloses a multi-target identification and safety early warning system for deep vertical shaft hoisting systems. The purpose of this utility model is to provide an identification and safety early warning system that enables real-time monitoring of equipment status, personnel behavior, and environmental parameters of deep vertical shaft hoisting systems, achieving accurate multi-condition identification and immediate early warning, thereby ensuring the safety of workers and the stable operation of construction equipment.

[0005] This utility model is achieved through the following technical solution:

[0006] A multi-target identification and safety early warning system for a deep vertical shaft hoisting system; the deep vertical shaft hoisting system includes: two sets of winches installed outside the deep vertical shaft opening, the two sets of winches being connected to the underground construction platform for traction drive via steel wire ropes suspended from the vertical shaft gantry truss at the top of the deep vertical shaft opening; characterized in that:

[0007] The multi-target recognition and safety early warning system includes: a pressure-side tension sensor installed on the wire rope of the winch, multiple temperature and humidity sensors spaced apart on the shaft wall, a vision sensor installed on the gantry truss of the shaft, a data receiver and processor deployed at the wellhead, and an early warning execution mechanism integrating an audible and visual alarm and a cloud storage system; the pressure-side tension sensor, temperature and humidity sensor, and vision sensor are electrically connected to the data receiver and processor, and the data receiver and processor are connected to the early warning execution mechanism.

[0008] Furthermore, the pressure-side tension sensor is a 15T range LCZ-401B-1ST-7-V5-P13 model, which is fixed to the wire rope 5 meters away from the winch drum using bolt clamps; the temperature and humidity sensor is a COS-03USB type recorder, which is deployed at 50-meter intervals on the shaft wall support structure; the vision sensor is a 1080P high-definition camera, which is installed at the center point of the crossbeam of the gantry truss from above, and the field of view covers the area of ​​the shaft platform and the wellhead cover plate.

[0009] Furthermore, the pressure-side tension sensor and temperature and humidity sensor are connected to the data receiver and processor via an RS485 shielded cable, and the vision sensor is connected to the data receiver and processor via a gigabit network cable.

[0010] Furthermore, the data receiver and processor uses an Intel Core i5-4310M embedded development board as its core, integrating a multi-source data fusion module, a target detection algorithm module, and a trajectory prediction module; the development board has a built-in YOLOv11 lightweight model and a multimodal neural network fusion model. The multimodal neural network fusion model is used for joint recognition of tension data, temperature and humidity data, and visual features.

[0011] After extracting high-dimensional visual features, the YOLOv11 lightweight model performs cross-modal fusion with the low-dimensional physical features of wire rope tension (kN) and temperature and humidity (°C / %RH) processed by the neural network fusion model. This generates a multi-dimensional feature matrix containing "time-space-physical quantity-semantics," supporting the recognition of five types of targets: a safety equipment recognition module detects the wearing status of safety helmets, reflective vests, and gas masks; a wire rope positioning module captures the pixel coordinates of the center point of the wire rope; a wire rope status analysis module identifies broken wire anomalies; a rebar feature extraction module determines the type of hoisted material; and an edge position calculation module quantifies the overlap between the hoisted object and the edge of the wellhead cover plate of the well platform using the IoU algorithm.

[0012] Furthermore, the early warning execution mechanism includes an audible and visual alarm device and a cloud storage system; the audible and visual alarm device adopts an integrated design of red and yellow dual-color LED warning lights and a voice broadcast module, and is connected to the data receiver and processor through a GPIO interface; the cloud storage system is built on Alibaba Cloud OSS service and is connected to the data receiver and processor through a 4G module.

[0013] The audible and visual alarm device will activate a voice prompt when it detects that a person is not wearing a safety helmet for 10 consecutive seconds; it will trigger a buzzer alarm and send a speed reduction command when the probability of collision on each wire rope track exceeds 80% within 5 seconds; and it will activate a red rotating alarm and a yellow flashing warning respectively when the machinery is overloaded or the environment is abnormal.

[0014] Furthermore, the multi-target recognition and safety early warning system is powered by a mobile power source.

[0015] The data receiver and processor establish a timestamp alignment mechanism to achieve multi-source data synchronization. When the wire rope tension exceeds the 147kN threshold, a mechanical overload response is triggered. When the ambient temperature and humidity exceed the preset safety range, an environmental anomaly protocol is activated.

[0016] The beneficial effects of the multi-target identification and safety early warning system of the deep vertical shaft hoisting system of this utility model are as follows:

[0017] By integrating visual sensors, physical sensors and edge computing units through a multi-source heterogeneous sensing fusion architecture, the limitations of traditional single-sensor monitoring are broken through. Combined with timestamp alignment technology, multi-dimensional data collaborative analysis is achieved, solving the problem of blind spots in deep shaft monitoring and comprehensively understanding and improving the system's operating status.

[0018] Based on the deployment of the lightweight YOLOv11n model on an embedded development board, it achieves real-time target detection at 30FPS, supports five types of special recognition, and can accurately capture key information such as personnel wearing safety equipment, wire rope positioning and trajectory, and wire rope status, providing accurate data support for safety monitoring.

[0019] Based on dynamic trajectory prediction and closed-loop control mechanism, the movement path of the wire rope within 5 seconds can be predicted. When the collision probability exceeds 80% or the real-time IoU value is greater than 0.3, the operator is automatically reminded to slow down, forming a "perception-decision-control" closed loop to avoid collision risks in time.

[0020] In the preferred implementation, the edge computing optimization design powered by a mobile power supply has a total core hardware cost of less than 1,000 yuan, low standby power consumption, and is suitable for construction scenarios without a power grid. Furthermore, by uploading only 30-second video clips before and after the alarm event through Alibaba Cloud OSS, the remote transmission bandwidth requirement is reduced, which greatly saves deployment and operation costs.

[0021] Through a millisecond-level closed-loop control system, when mechanical overload, personnel violations, collision risks, or environmental anomalies are detected, a four-level response, including power cut-off, speed reduction, voice prompts, audible and visual alarms, and SMS push notifications, can be triggered within 200ms. This achieves an essential safety leap from passive post-event recording to proactive pre-event interception, effectively ensuring the construction safety of deep vertical shaft hoisting systems. Attached Figure Description

[0022] Figure 1 This is a system framework diagram of the present invention;

[0023] Figure 2 This is a schematic diagram of the system layout according to an embodiment of the utility model.

[0024] In the diagram: 1 and 2 are winches; 3 and 4 are side-pressure tension sensors; 5 and 6 are wire ropes; 7 is pulleys; 8 is trusses; 9 is a high-definition camera; 10 is a temperature and humidity sensor; 11 is a data receiver and processor; 12 is the surface platform; 13 is the underground construction platform; and 14 is the shaft wall. Detailed Implementation

[0025] The present invention will be further described below with reference to specific embodiments. These specific embodiments are further explanations of the principle of the present invention and are not intended to limit the present invention in any way. Any technology that is the same as or similar to the present invention does not exceed the protection scope of the present invention.

[0026] Example:

[0027] like Figure 1 , Figure 2 As shown, the multi-target identification and safety early warning system applied to the deep vertical shaft hoisting system in this embodiment includes: side-pressure tension sensors 3 and 4 respectively installed on the traction steel wire ropes 5 and 6 of each winch 1 and 2, temperature and humidity sensor 10 installed on the vertical shaft wall 14, vision sensor 9 installed on the vertical shaft gantry truss 8, data receiver and processor 11 installed at the wellhead, and early warning execution mechanism;

[0028] The pressure-side tension sensors 3 and 4 are LCZ-401B-1ST-7-V5-P13 models with a range of 15T. They are fixed to the main wire ropes 5 and 6 5 meters away from the drums of winches 1 and 2 by bolt clamps. Their sensing ends are in close contact with the surface of the wire ropes 5 and 6. They can collect the tension data of the two wire ropes 5 and 6 in real time and transmit it to the data receiver and processor 11 through RS485 shielded cable to ensure stable transmission of tension signals in the complex electromagnetic environment of deep vertical shafts.

[0029] The temperature and humidity sensor 10 uses a COS-03USB recorder. Five sensors are installed at 10-meter intervals along the support structure of the shaft wall 14. The sensor probe is exposed 3-5cm on the support surface to avoid being blocked by the attachments on the shaft wall 14. Temperature and humidity parameters are collected once per minute and uploaded through a shielded cable to form a longitudinal distribution monitoring of the shaft environmental parameters.

[0030] The visual sensor 9 is a 1080P high-definition camera, which is installed from above at the center point of the crossbeam of the gantry truss 8 via a waterproof bracket. The lens axis is perpendicular to the plane of the well opening, and the field of view covers the area of ​​the vertical shaft opening and the platform 12 above the well. It captures video streams at a rate of 30 frames per second and transmits them to the data receiver and processor 11 via a gigabit industrial network cable. The lens cover is equipped with an anti-fog coating and a dustproof net to adapt to the dusty and humid environment inside the shaft. It also needs to be specially trained for the steel wire rope to enhance the accuracy of the visual recognition algorithm in locating its center point.

[0031] The data receiver and processor 11 is based on the Intel Core I5-4310M embedded development board, integrating a multi-source data fusion module, a target detection algorithm module, and a trajectory prediction module. The development board has a built-in YOLOv11 lightweight model, and the model is preloaded with five types of target recognition weight files: the safety equipment recognition module can detect the wearing status of workers' safety helmets, reflective vests, and gas masks; the wire rope positioning module captures the real-time position of the center points of wire rope 5 and 6 through pixel coordinate transformation; the wire rope status analysis module identifies anomalies such as broken wires and wear based on image texture; the rebar feature extraction module determines the type of hoisted material (rebar / concrete) through contour matching; and the edge position calculation module quantifies the overlap between the hoisted object and the edge of the wellhead of the well platform 12 through the IoU algorithm.

[0032] Each sensor and data receiver / processor 11 adopts a hybrid transmission architecture: the pressure-side tension sensors 3 and 4 and the temperature and humidity sensor 10 are directly connected to the development board via an RS485 interface, with the communication baud rate set to 9600bps; the vision sensor 9 is connected to the development board's gigabit Ethernet port via a USB-to-Ethernet module, and the video stream transmission uses the RTSP protocol. The development board has a built-in timestamp alignment module, which realizes multi-source data synchronization through a dynamic timestamp alignment algorithm. Based on the vision sensor's 30 frames / second and approximately 33ms interval, the high-frequency tension data is interpolated and smoothed, and the low-frequency temperature and humidity data is timestamp mapped and expanded, ensuring that the multi-source data time synchronization error is <10ms.

[0033] The multi-source data fusion module uses multimodal feature stitching technology to fuse low-dimensional physical data such as wire rope tension (kN) and temperature / humidity (°C / %RH) with high-dimensional information such as target coordinates, equipment status, and texture features extracted by visual sensors, generating a multi-dimensional state matrix containing "time-space-physical quantity-semantic features". The data receiver and processor 11 establishes multi-level decision rules, triggering a mechanical overload response when the wire rope tension exceeds the 147kN threshold, and initiating an environmental anomaly protocol when the ambient temperature and humidity exceed the preset safety range. The trajectory prediction module predicts the wire rope's movement path within 5 seconds based on the fused data, automatically activating the intervention process when the collision probability exceeds 80% or the IoU value is >0.3.

[0034] In the early warning execution mechanism, the audible and visual alarm device integrates red and yellow dual-color LED warning lights and a voice broadcast module. It is connected to the development board through the GPIO interface and installed in a conspicuous position in the wellhead control room. The cloud storage system is built on Alibaba Cloud OSS service. The development board communicates with the cloud through a 4G module (model EC20) and automatically uploads 30-second video clips before and after the alarm event and sensor abnormal data packets. The video compression format adopts H.265 to reduce data consumption.

[0035] In this embodiment, the system is powered by a 12V lithium battery with a capacity of 20Ah, which is integrated into the protective box of the data receiver and processor 11. The protective box adopts an IP65 waterproof design and is fixed in front of the winches 1 and 2. The contact between the side-pressure tension sensors 3 and 4 and the wire ropes 5 and 6 should be tightened as much as possible to improve the detection accuracy and sensitivity of the sensors. The temperature and humidity sensor 10 is fixed to the concrete surface of the well wall 14 support structure by expansion bolts, and the distance between it and the well wall is maintained at 5cm to prevent vibration interference.

[0036] The visual sensor 9 is installed 5 meters above the wellhead plane, and the lens focal length is set to 8mm to ensure that the field of view covers the entire trajectory of the steel wire ropes 5 and 6 at the wellhead. Targeted training is carried out on the steel wire ropes to enhance the accuracy of the visual recognition algorithm in locating their center point. The development board of the data receiver and processor 11 is pre-installed with the Windows operating system. The target detection algorithm occupies less than 512MB of memory when running, ensuring stability during multi-task parallel processing.

Claims

1. Multi-target identification and safety early warning system for deep vertical shaft hoisting systems; The deep shaft hoisting system includes: two sets of winches installed outside the deep shaft opening, the two sets of winches being connected to the underground construction platform via steel wire ropes suspended from the top of the shaft gantry truss at the shaft opening; characterized in that: The multi-target recognition and safety early warning system includes: a pressure-side tension sensor installed on the wire rope of the winch, multiple temperature and humidity sensors spaced apart on the shaft wall, a vision sensor installed on the gantry truss of the shaft, a data receiver and processor deployed at the wellhead, and an early warning execution mechanism integrating an audible and visual alarm and cloud storage system; the pressure-side tension sensor, temperature and humidity sensor, and vision sensor are respectively connected to the data receiver and processor, and the data receiver and processor are connected to the early warning execution mechanism.

2. The multi-target identification and safety early warning system for deep vertical shaft hoisting systems according to claim 1, characterized in that: The pressure-side tension sensor is a 15T range LCZ-401B-1ST-7-V5-P13 model, which is fixed to the wire rope 5 meters away from the winch drum using bolt clamps; the temperature and humidity sensor is a COS-03USB recorder, which is installed at 50-meter intervals on the shaft wall support structure; the vision sensor is a 1080P high-definition camera, which is installed at the center point of the crossbeam of the gantry truss from above, and the field of view covers the area of ​​the shaft platform and the wellhead cover plate.

3. The multi-target identification and safety early warning system for deep vertical shaft hoisting systems according to claim 1, characterized in that: The pressure-side tension sensor and temperature and humidity sensor are connected to the data receiver and processor via RS485 shielded cables, and the vision sensor is connected to the data receiver and processor via a gigabit network cable.

4. The multi-target identification and safety early warning system for deep vertical shaft hoisting systems according to claim 1, characterized in that: The data receiver and processor uses an Intel Core I5-4310M embedded development board as its core, integrating a multi-source data fusion module, a target detection algorithm module, and a trajectory prediction module; the development board has a built-in YOLOv11 lightweight model and a multimodal neural network fusion model.

5. The multi-target identification and safety early warning system for deep vertical shaft hoisting systems according to claim 1, characterized in that: The early warning execution mechanism includes an audible and visual alarm device and a cloud storage system; the audible and visual alarm device adopts an integrated design of red and yellow dual-color LED warning lights and a voice broadcast module, and is connected to the data receiver and processor through a GPIO interface; the cloud storage system is built on Alibaba Cloud OSS service and is connected to the data receiver and processor through a 4G module.

6. The multi-target identification and safety early warning system for deep vertical shaft hoisting systems according to claim 1, characterized in that: The multi-target recognition and safety early warning system is powered by a mobile power source.

Images