A method, system, and medium for detecting elevator shafts

Abstract

This invention relates to the field of intelligent detection technology, and discloses a method, system, and medium for detecting elevator shafts. The method includes: collecting point cloud data of the elevator shaft to be modeled to construct a real-time shaft model; locating the model difference areas by comparing a standard shaft model and the real-time shaft model; triggering M shaft detection nodes based on the model difference areas, with the shaft detection array in a low-power standby state; performing deviation factor backtracking to obtain induced prediction results; collecting multi-source fine-grained data after adjusting the detection parameters; performing deformation-induced analysis of the model difference areas to locate assembly defects in the device; and performing operation and maintenance management rule matching to output real-time operation and maintenance strategies. This invention solves the technical problem that existing elevator shaft detection technologies often rely on design drawings or manual scanning, lacking real-time dynamic monitoring capabilities, resulting in potential defects during long-term operation not being addressed in a timely manner, thus affecting the long-term operational safety of elevators.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of intelligent detection, and particularly relates to a method and system for detecting an elevator shaft and a medium. BACKGROUND

[0002] In the traditional elevator shaft modeling and assembly process, design drawings and engineering drawings are usually the basis for modeling and construction. These design drawings are based on static architectural design and lack dynamic feedback on actual changes in site construction and environment. Therefore, the modeling of the elevator shaft often fails to accurately capture the actual differences caused by errors, material deviations or spatial layout changes during the construction process. Although building information modeling technology can provide relatively accurate design information, it is limited to static design models and lacks sufficient combination with real-time site data, resulting in a large deviation between the design and the actual shaft.

[0003] In order to make up for this gap, more and more technologies have begun to try to introduce real-time monitoring and detection methods to collect data through laser ranging, sensors and other means. However, the existing elevator shaft detection technology still faces many challenges in the application process, especially how to timely and accurately capture and correct the differences between the design and the construction during the construction process. Because the traditional method often relies on manual operation or local detection, it cannot comprehensively and systematically perform real-time dynamic monitoring and accurate correction, resulting in the error between the shaft model and the actual environment not being discovered and adjusted in time, affecting the installation accuracy of the elevator equipment and further affecting the long-term operation safety of the elevator. SUMMARY

[0004] The present application provides a method and system for detecting an elevator shaft and a medium to solve the problems raised in the background.

[0005] The first aspect of the present application provides a method for detecting an elevator shaft, which comprises: after collecting point clouds of a to-be-modeled elevator shaft to construct a real-time shaft model, locating a model difference area by comparing a standard shaft model with the real-time shaft model; triggering M shaft detection nodes of a shaft detection array according to the regional space envelope of the model difference area, wherein a shaft detection array is pre-deployed in the to-be-modeled elevator shaft, and the shaft detection array is in a low-power standby state; performing deviation factor backtracking according to the regional space envelope to obtain an induced prediction result; after adjusting the detection parameters of the M shaft detection nodes based on the regional space envelope and the induced prediction result, driving the M shaft detection nodes to collect multi-source fine-grained data; performing model difference area deformation induction analysis by fusing and analyzing the multi-source fine-grained data to locate device assembly defects; and performing operation and maintenance management rule matching according to the trigger time stamp of the M shaft detection nodes and the device assembly defects to output a real-time operation and maintenance strategy.

[0006] In a second aspect, the application discloses a detection system of an elevator shaft, which is used for the detection method of the elevator shaft, and comprises: a model difference area positioning module, which is used for positioning a model difference area by comparing a standard shaft model with a real-time shaft model after point cloud collection is performed on a to-be-modeled elevator shaft to construct the real-time shaft model; a shaft detection node triggering module, which is used for triggering M shaft detection nodes in a shaft detection array according to a region space envelope of the model difference area, wherein the to-be-modeled elevator shaft is pre-deployed with the shaft detection array, and the shaft detection array is in a low-power standby state; a deviation factor backtracking module, which is used for obtaining an induced prediction result according to the region space envelope; a multi-source fine-grained data collection module, which is used for collecting multi-source fine-grained data by driving the M shaft detection nodes to collect the multi-source fine-grained data after the M shaft detection nodes are adjusted in detection parameters based on the region space envelope and the induced prediction result; a device assembly defect positioning module, which is used for positioning a device assembly defect by performing model difference area deformation induced analysis on the multi-source fine-grained data; and an operation and maintenance management rule matching module, which is used for matching operation and maintenance management rules according to a triggering time stamp of the M shaft detection nodes and the device assembly defect, and outputting a real-time operation and maintenance strategy.

[0007] In a third aspect, the application discloses a storage medium, which stores a computer program, and the computer program is executed by a processor to implement the steps of the detection method of the elevator shaft.

[0008] Thanks to the above technical solutions, the application has the following technical progress compared with the prior art:

[0009] By point cloud collection, a real-time shaft model of the elevator shaft is constructed and compared with a standard shaft model, which can accurately identify and locate the areas in the shaft that differ from the standard model. This method can accurately identify differences between the design and actual construction, ensuring the structural integrity and accuracy of the shaft. By pre-deploying a shaft detection array in the shaft and maintaining a low-power standby state, the detection nodes can be quickly activated for data collection when needed, avoiding unnecessary energy consumption. By triggering M shaft detection nodes for efficient collection, accurate data can be obtained in key areas within the shaft, ensuring comprehensive monitoring of all parts of the shaft. According to the regional space envelope, the deviation factor is traced back to effectively predict possible faults or problems in the elevator shaft, identify potential assembly defects in advance, and generate induced prediction results to take preventive measures in advance, reducing the probability of failure and providing decision support for subsequent accurate detection and repair work. By adjusting the detection parameters of the shaft detection nodes, the collection method can be dynamically adjusted according to real-time needs, avoiding redundant data collection through "full scan" and improving data collection efficiency. The collection of multi-source fine-grained data can provide fine-grained monitoring of the shaft from different angles and scales, greatly improving the accuracy and precision of defect identification. Through deformation induction analysis of multi-source fine-grained data, the deformation area caused by external or internal factors within the shaft can be accurately identified, and the device assembly defects in the shaft can be located. This precise defect positioning technology can improve maintenance and management efficiency and reduce equipment failures caused by assembly defects. By matching the trigger timestamp of the shaft detection node with the device assembly defect, the detection data can be effectively connected with the operation and maintenance management rules to output real-time operation and maintenance strategies. In this way, operation and maintenance personnel can take appropriate measures to repair or adjust the shaft in a timely manner based on real-time data, improving the automation level of elevator shaft management and reducing the need for manual intervention.

[0010] The above description is only a summary of the technical solutions of the present application. In order to more clearly understand the technical means of the present application, the content of the specification can be implemented, and in order to make the above and other purposes, features and advantages of the present application more obvious and easy to understand, the following specific embodiments of the present application are described. BRIEF DESCRIPTION OF DRAWINGS

[0011] Figure 1 A detection method flowchart of an elevator shaft is provided for the embodiments of the present application.

[0012] Figure 2 A detection system structure diagram of an elevator shaft is provided for the embodiments of the present application.

[0013] Explanation of reference numerals in the attached figures: Model difference area positioning module 10, wellhead detection node triggering module 20, deviation factor backtracking module 30, multi-source fine-grained data acquisition module 40, device assembly defect positioning module 50, operation and maintenance management rule matching module 60. Detailed Implementation

[0014] The present invention will now be described in further detail with reference to the accompanying drawings and embodiments. Obviously, the described embodiments are only a part of the embodiments of the invention, and not all of them. Unless otherwise specified, the embodiments and features described in this application can be combined with each other. All other embodiments obtained by those skilled in the art based on the embodiments of the invention without inventive effort are within the scope of protection of the invention.

[0015] It should be noted that if the embodiments of the invention involve directional indicators (such as up, down, left, right, front, back, etc.), the directional indicators are only used to explain the relative positional relationship and movement of the components in a certain specific posture (as shown in the figure). If the specific posture changes, the directional indicators will also change accordingly.

[0016] Furthermore, "multiple" refers to two or more. Additionally, the technical solutions of the various embodiments can be combined with each other, but this must be based on the ability of a person skilled in the art to implement them. When the combination of technical solutions is contradictory or cannot be implemented, it should be considered that such a combination of technical solutions does not exist and is not within the scope of protection claimed by the invention.

[0017] Example 1, as Figure 1 As shown in the figure, this application embodiment provides a method for detecting elevator shafts, the method comprising:

[0018] After collecting point cloud data to construct a real-time shaft model for the elevator shaft to be modeled, the differences between the model and the standard shaft model are located by comparing the real-time shaft model with the standard shaft model.

[0019] Point cloud acquisition equipment is used to scan the elevator shaft to be modeled, collecting three-dimensional spatial data, i.e., point cloud data, within the elevator shaft. This point cloud data contains coordinate information for each location within the elevator shaft and describes the geometry and position of all objects within it. For example, a detection device consisting of a Doppler imager, infrared detector, sonar, and related scanners is placed and stabilized within the elevator shaft using a cable-stayed rod or similar device. Each detection unit is 100 meters vertically upwards or downwards within the elevator shaft, projecting and detecting the "civil construction conditions within the shaft" and the "installation and arrangement of elevator components within the shaft" in a multi-dimensional, omnidirectional, fan-shaped pattern. This data is then used for effective modeling and imaging in 3D modeling software. The "civil engineering construction conditions inside the shaft" include, but are not limited to: the width and depth of the shaft, the width and height of the door openings, the condition of the foundation pit, the condition between each ring beam, the condition of the door beam and the sill beam, the condition of the reserved holes, the condition of the shaft structure, and the condition of any protruding or concave objects in the shaft; the "installation and layout conditions of elevator components inside the shaft" include, but are not limited to: the verticality and wear condition of the guide rails, and the current condition of elevator components such as landing doors, counterweights, and buffers.

[0020] Based on the spatial hierarchy of the shaft, hierarchical modeling and assembly of point cloud data is performed. Hierarchical modeling means processing the point cloud data in layers according to different scales. For example, the external contour and internal details (such as elevator equipment, tracks, and supporting structures) of the shaft are modeled at different levels. This ensures that different parts of the shaft are properly described. During this process, algorithms are used to register, filter, and reconstruct the point cloud data to ensure that the resulting real-time shaft model is as geometrically consistent as possible with the actual shaft. During the modeling process, relevant data can be analyzed, extracted, and output in the software. For example, the verticality of the shaft width side and the verticality of the shaft depth side can be used to calculate the precise "shaft width and depth dimensions," the minimum "doorway width and height dimensions" in multi-level doorways, the "depth dimension" of the foundation pit, the "spacing data" between each ring beam and the width and length dimensions of the ring beam, the "width and length dimensions" of the door beam and sill beam, the "width, height, and depth dimensions" of the reserved holes, and the "verticality" of the existing elevator tracks, etc.

[0021] It interacts with the local building design database to obtain a standard shaft model. The standard shaft model is built based on historical design information or standard specifications, including design details such as the size, shape, and support structure of the shaft, representing an ideal elevator shaft structure.

[0022] The constructed real-time shaft model is projected onto the standard shaft model for comparison. This projection process can be accomplished through techniques such as geometric mapping and comparative analysis. In this way, the model difference areas between the standard shaft model and the real-time shaft model are located. The model difference areas represent the gap between the actual shaft and the design standard, such as shape distortion, inconsistent dimensions or structural defects, as well as the presence of foreign objects in the shaft, the degree of wear of the original elevator rails, and the current condition of the original elevator components.

[0023] Based on the spatial envelope of the model difference region, M shaft detection nodes are triggered in the shaft detection array, wherein the shaft detection array is pre-deployed in the elevator shaft to be modeled, and the shaft detection array is in a low-power standby state.

[0024] The spatial envelope of a region refers to the smallest spatial boundary that encloses the region of difference in the model. It can be understood as a three-dimensional bounding box that includes all the regions that need to be explored, indicating the actual location and range of the region to be explored.

[0025] A shaft detection array is pre-deployed within the elevator shaft to be modeled. Each detection node in the array deploys several different types of detection devices, which are used to activate the corresponding detection devices at subsequent nodes based on a target stitching strategy. The shaft detection array is in a low-power standby state, meaning that the detection devices do not perform energy-consuming operations under normal circumstances, but instead wait for external trigger signals. Only when the model's difference areas are identified and the required detection range is determined will the corresponding shaft detection node be awakened and activated for data acquisition. Based on the calculated regional spatial envelope, M shaft detection nodes (M being a positive integer) are determined to be triggered within the shaft detection array. The triggering of these M nodes can be performed according to a preset spatial distribution strategy to ensure coverage of all areas requiring detection. These detection nodes are allocated according to the actual required accuracy and acquisition requirements to obtain detailed and comprehensive data.

[0026] By backtracking the deviation factors based on the spatial envelope of the region, the induced prediction results are obtained.

[0027] Deviation factor backtracking is an analytical method used to trace the causes of discrepancies and identify the external or internal factors contributing to these deviations. This process involves comparing models with actual conditions to infer deviation factors that may have occurred during design, construction, or use, constructing a defect feature vector. This feature vector contains information in different dimensions, such as spatial location, shape, and size. The constructed defect feature vector is then matched with a pre-established database of elevator shaft defect cases. Algorithms such as cosine similarity are used to find the most similar historical defect cases. Based on the matching results of similar cases, the type of defect caused by the spatial envelope of the area is predicted, such as structural deformation, assembly defects, and equipment failures. Finally, a prediction result is obtained, which provides guidance for subsequent inspections.

[0028] Based on the regional spatial envelope and the induced prediction results, after adjusting the detection parameters of the M wellhead detection nodes, the M wellhead detection nodes are driven to collect multi-source fine-grained data.

[0029] After obtaining the induced event prediction results, the parameters of the M shaft detection nodes are adjusted according to the specific characteristics of the regional spatial envelope and the historical defect types to obtain efficient and high-precision detection results. After completing the detection parameter adjustment, the M shaft detection nodes are driven to collect multi-source fine-grained data. The task of each shaft detection node is to collect different types of data according to its specific function. For example, lidar equipment measures distance by emitting lasers and receiving reflected signals, generating high-precision point cloud data. This point cloud data can describe the geometry of the shaft in detail, helping to identify structural deformation, cracks, or other geometric deviations. Ultrasonic detection equipment can penetrate deep into the structure through sound wave propagation to detect internal defects or voids, including cracks, corrosion, or other defects inside the shaft walls or supporting structures. Infrared imaging equipment identifies thermal anomaly areas by detecting changes in surface temperature. For elevator shafts, it is used to monitor the overheating of electrical equipment or check for areas of thermal expansion or undercooling in the structure. By combining the induced event prediction results, focused scanning is achieved, avoiding the redundancy of comprehensive scanning and improving data acquisition efficiency and targeting.

[0030] By integrating and analyzing the multi-source fine-grained data, deformation-induced analysis of the model's differential regions is performed to locate assembly defects in the device.

[0031] Fusion analysis refers to integrating multi-source fine-grained data of different types, mining the correlation between them, and identifying potential deformations or defects. Deformation-induced analysis aims to assess how deformation in the model's differential regions affects the assembly accuracy of the wellbore structure and equipment, and whether it may lead to potential equipment failures or damage. Based on deformation-induced analysis, assembly defects are located, including assembly errors, stress concentration, aging, or wear. For example, infrared and ultrasonic analysis can identify aging and wear phenomena caused by long-term use, which may lead to reduced equipment assembly accuracy and affect equipment performance.

[0032] Based on the trigger timestamps of the M wellhead detection nodes and the assembly defects of the device, the operation and maintenance management rules are matched to output a real-time operation and maintenance strategy.

[0033] Each wellbore detection node records a trigger timestamp upon activation, indicating the specific moment of the detection activity. By combining this with assembly defects, the location and occurrence time of defects can be precisely matched. The core objective of operation and maintenance management is to formulate effective operation and maintenance strategies based on real-time monitoring data and historical defect information. This includes: establishing rules based on historical assembly defects and equipment failure records, combined with existing multi-source data analysis results. For example, certain types of assembly defects are more likely to occur under certain conditions, or the failure rate is higher when the equipment undergoes specific deformations. Based on the location and trigger time of defects, the operation and maintenance needs of the area are analyzed. For example, if the deformation in a certain area is caused by long-term accumulation, then regular inspections and repairs need to be arranged, while sudden defects in the short term require emergency repairs. Based on the matching of operation and maintenance management rules, real-time operation and maintenance strategies are output, including repair priorities, operation and maintenance arrangements, and resource allocation, to ensure the safe and stable operation of the wellbore.

[0034] Furthermore, the method involves triggering M wellhead detection nodes in the wellhead detection array based on the spatial envelope of the model's difference region.

[0035] Extract the three-dimensional spatial coordinate range of the model difference region to calculate the minimum bounding box boundary coordinates and output the spatial envelope of the region; locally call multiple installation spatial coordinates of multiple dormant detection nodes in the wellbore detection array; define multiple detection adjustment intervals based on the standard wellbore model and multiple installation spatial coordinates; solve for multiple actual wellbore detection ranges based on the inherent detection coverage range of the standard detection node and the multiple detection adjustment intervals; calculate multiple detection coverage ranges of the multiple actual wellbore detection ranges relative to the spatial envelope of the region; and locate the M wellbore detection nodes by filtering, splicing, and optimizing the multiple detection coverage ranges.

[0036] For the model difference region, the three-dimensional spatial coordinates of its spatial boundary are extracted. Specifically, these coordinates are obtained through three-dimensional point cloud data. These three-dimensional spatial coordinates represent the position of the model difference region in three-dimensional space. The minimum bounding box is the smallest rectangle that can completely enclose the model difference region. By calculating the boundary coordinates of the minimum bounding box, the spatial range of the model difference region can be obtained. Specifically, for the model difference region, its minimum and maximum coordinates on the X-axis, Y-axis, and Z-axis are calculated to form the boundary of a cuboid. For example, assuming that the X coordinate of a certain model difference region is between 1.2 meters and 1.8 meters, the Y coordinate is between 0.5 meters and 2.5 meters, and the Z coordinate is between 9 meters and 15 meters, then the boundary coordinates of the minimum bounding box of this region are X=[1.2m,1.8m], Y=[0.5m,2.5m], and Z=[9m,15m]. In this way, the regional spatial envelope of the model difference region is output.

[0037] Multiple dormant detection nodes of the wellbore detection array are in a low-power standby state and do not collect data when not triggered, in order to save energy and extend the service life of the equipment. Each dormant detection node has a fixed installation position, and its installation spatial coordinates are known. These coordinates can be found in the configuration file of the wellbore detection array and are represented as three-dimensional coordinates in the wellbore, such as X=1.5m, Y=2.0m, Z=10.0m. In actual operation, these coordinates are used to locate the position of each node.

[0038] Because different areas within the shaft may have different detection requirements, especially in areas with model differences, the working range and parameters of the detection nodes need to be adjusted according to the actual situation. The detection adjustment range refers to the reasonable working range set for each detection node to ensure that they can perform tasks efficiently without redundant scanning or missed areas. The detection adjustment range is set through geometric calculations. For example, the effective detection radius and range of each detection node are determined. For instance, a node can only scan an area within a certain radius, such as within 3 meters. The scanning direction and angle of each node are adjusted according to its installation location and the specific situation of the target area. For instance, a node only needs to scan one side of the shaft wall and not the opposite wall. For different types of detection equipment, the detection depth is adjusted, for example, the detection depth of ultrasonic equipment, to ensure coverage of different layers or structural parts.

[0039] The inherent detection coverage of a standard detection node is determined by its sensor type, working principle, and design capabilities. For example, the coverage of a laser detection node is all surfaces within a certain radius, while the coverage of an ultrasonic detection node is a certain depth through which it penetrates the material. The inherent detection coverage of these standard detection nodes is known during the design phase and is usually manifested as a circular, rectangular, or other geometrically shaped detection area with a fixed detection radius or field of view.

[0040] In some cases, the inherent detection coverage of a standard detection node is not accurate enough to match the area to be detected. For example, the detection range of some detection nodes is limited by the installation angle, and the detection angle needs to be adjusted according to the characteristics of the area. The inherent detection coverage can be adjusted by multiple detection adjustment intervals. For example, suppose the default coverage of a certain detection node is a radius of 3 meters, but due to the limitation of the area space, the detection adjustment interval reduces its detection range to a radius of 2 meters. By adjusting, multiple actual wellbore detection ranges are obtained, indicating the actual detection area of ​​each of the multiple detection nodes in the wellbore.

[0041] For each dormant detection node, determine whether its actual wellbore detection range overlaps with the regional spatial envelope. If the actual wellbore detection range of a dormant detection node completely or partially covers the regional spatial envelope, then the dormant detection node can effectively detect the region. Calculate its coverage of the regional spatial envelope to obtain multiple detection coverage ranges.

[0042] Based on multiple detection coverage areas, the final wellbore detection nodes are selected through screening and splicing strategies. This includes: screening detection nodes with the most comprehensive spatial envelope of the coverage area, for example, selecting those detection nodes that can cover different areas to the greatest extent; if no single detection node can completely cover the target area, then the coverage areas of multiple detection nodes are combined to achieve full coverage. In this case, it is necessary to form the best detection coverage by splicing according to the detection range of different detection nodes.

[0043] Different optimization algorithms are used to select the optimal combination of detection nodes, such as greedy algorithms and genetic algorithms. The objective functions of optimization include maximizing the coverage area, minimizing redundant detection, and maximizing detection efficiency. Through optimization algorithms and splicing strategies, the positions and numbers of M wellbore detection nodes are finally determined. These nodes will serve as detection nodes for multi-source fine-grained data acquisition in subsequent steps.

[0044] Furthermore, by filtering and optimizing the multiple detection coverage areas to locate the M wellbore detection nodes, the method includes:

[0045] Using the spatial envelope of the region as a coverage constraint, the multiple detection coverage areas are enumerated and stitched together to obtain multiple detection overlap rates, multiple detection coverage rates, and multiple detection node call volumes for multiple stitching strategies. The multiple detection coverage rates are traversed according to a preset coverage threshold to filter out P stitching strategies. The P detection overlap rates, P detection coverage rates, and P detection node call volumes of the P stitching strategies are weighted and quantized to obtain the overall superiority of the P strategies. The overall superiority of the P strategies is serialized, the target stitching strategy is located, and the M wellhead detection nodes are triggered based on the node composition of the target stitching strategy.

[0046] Using the spatial envelope of the region as a coverage constraint, the detection ranges of all probe nodes, after being stitched together, must completely cover this region. All subsequent stitching strategies must adhere to this constraint. For the detection coverage range of each probe node, all possible stitching methods are enumerated. Specifically, since the detection coverage range of each probe node is finite and varies in shape, a suitable combination needs to be created by stitching their coverage areas so that their total coverage can completely cover the spatial envelope of the region.

[0047] During the stitching process, the detection coverage of multiple detection nodes may overlap. In this case, the overlapping area or overlapping volume between the detection nodes is calculated. The resulting detection overlap rate reflects the proportion of repeated detection areas between different detection nodes. A stitching strategy with a high overlap rate means that multiple nodes are repeatedly detecting in the same area, which may result in redundant data collection.

[0048] Detection coverage rate refers to the proportion of the total coverage area of ​​a spliced ​​detection node to the spatial envelope of the region. Detection coverage rate means that the detection node can better cover the spatial envelope of the region, and the detection results are more complete.

[0049] The number of probe nodes called refers to the number of probe nodes that need to be activated in each splicing strategy, that is, the total number of all probe nodes selected in the splicing scheme. A strategy with a low number of probe nodes called means that fewer probe nodes are activated, which is beneficial to reduce the power consumption of the system and improve efficiency. A strategy with a high number of probe nodes called activates more nodes, which, although the detection range is larger, will bring higher energy consumption and hardware costs.

[0050] The preset coverage threshold is a standard used to filter stitching strategies. It defines the minimum probe coverage that a stitching strategy needs to achieve. This threshold is set according to the needs of the actual application. For example, setting the threshold to 90% means that only stitching strategies with a probe coverage greater than or equal to 90% are considered. By iterating through the probe coverage of all stitching strategies, those that meet the preset coverage threshold are selected, resulting in P stitching strategies, where P is a positive integer.

[0051] Weights are assigned to each metric based on actual needs. For example, in some scenarios, coverage is more important than overlap, or in some low-power scenarios, node call volume is the most important consideration. For each splicing strategy, each metric is weighted and summed according to the set weights to obtain the corresponding overall strategy superiority. A splicing strategy with a higher overall strategy superiority means that it has a good balance among overlap, coverage, and node call volume, and is suitable as the final execution solution.

[0052] All P stitching strategies are sorted according to their overall strategy superiority value to form a priority list, with the strategy with the highest overall strategy superiority value at the top and the strategy with the lowest overall strategy superiority value at the bottom. By sorting, the target stitching strategy is located; this strategy is the optimal stitching scheme and has the best detection effect. Based on the node composition of the target stitching strategy, M wellhead detection nodes are triggered, driving them to begin executing their detection tasks.

[0053] Furthermore, based on the regional spatial envelope and the induced prediction results, after adjusting the detection parameters of the M wellbore detection nodes, the M wellbore detection nodes are driven to collect multi-source fine-grained data. The method includes:

[0054] The system locally calls upon Q detection device types and Q detection parameter requirements for Q historical defect types from the induced prediction results; by aggregating the Q detection device types and Q detection parameter requirements, it outputs detection device scheduling instructions and detection device parameter tuning instructions; it retrieves M node spatial attributes of the M wellbore detection nodes from the target splicing strategy; after activating the M wellbore detection nodes using the detection device scheduling instructions, it uses the detection device parameter tuning instructions to configure the parameters of the activated devices for the M wellbore detection nodes; it uses the M node spatial attribute mapping to perform spatial orientation calibration of the M wellbore detection nodes; and it drives the M wellbore detection nodes to collect multi-source fine-grained data from the model difference region.

[0055] The induced prediction results provide predictive information on the types of historical defects that may exist in the elevator shaft. For each type of historical defect, the relevant detection equipment type and detection parameter requirements are extracted. The detection equipment type includes, for example, ultrasonic detectors, laser scanners, and infrared sensors. These devices are selected according to the different types of defects. The detection parameter requirements include parameters such as the operating frequency, accuracy requirements, sampling frequency, and detection range of the equipment. Different types of historical defects require detection equipment with different accuracy or different functions.

[0056] The extracted Q detection device types and Q detection parameter requirements are aggregated to unify and integrate these requirements. This aims to reduce redundancy and ensure optimal device utilization for each detection task. If multiple defect types require the same detection device type and parameter requirements, they can be merged to reduce resource waste. Based on the aggregated detection device types, detection device scheduling instructions are generated to ensure devices start according to a predetermined plan. For example, the instructions can specify when, where, and which devices will perform detection. Detection device parameter tuning instructions are also generated based on the detection requirements of different defects. For instance, if a detection task requires higher precision or a higher scanning frequency, the parameter tuning instructions will adjust the device configuration, such as ultrasonic frequency and laser point cloud resolution, to ensure the devices can execute tasks according to predetermined parameters and sequence.

[0057] Each wellhead detection node has a corresponding node spatial attribute, which describes the location, orientation, detection range, and other information of the wellhead detection node. By obtaining the node spatial attribute, the detection area of ​​each wellhead detection node can be accurately located, ensuring that all areas are covered.

[0058] According to the detection equipment scheduling instructions, signals are sent to M wellhead detection nodes to activate the equipment and put it into working condition, ready for detection tasks. After activation, configuration operations are performed based on the detection equipment parameter adjustment instructions. Each wellhead detection node is configured with specific operating parameters to adapt to different detection tasks, such as sensor sampling frequency, sensitivity, and accuracy. Through parameter configuration, it is ensured that the equipment of each wellhead detection node is in the most suitable state and can accurately capture the required details.

[0059] Due to differences in installation, environment, or physical characteristics, the detection direction and coverage of each wellhead detection node may deviate. Spatial orientation calibration is achieved by calibrating the node spatial attributes of each wellhead detection node to ensure that the detection direction of the wellhead detection node is aligned with its actual position, thereby improving detection accuracy and data reliability.

[0060] Once all M wellhead detection nodes are activated and spatial orientation calibration is completed, the actual detection task begins. Each wellhead detection node collects different types of fine-grained data. The data collected from the M wellhead detection nodes are fused to obtain multi-source fine-grained data, providing data support for subsequent deformation analysis and defect location.

[0061] Furthermore, by fusing and analyzing the multi-source fine-grained data to perform deformation-induced analysis of the model's differential regions and locating assembly defects in the device, the method includes:

[0062] A pre-constructed deformation-induced analysis model is established, comprising multiple parallel deformation-induced analysis channels, each channel bearing a detection device type identifier. Based on H scheduling device types in the detection device scheduling instruction, the H deformation-induced analysis channels in the model are matched and activated. According to the H scheduling device types, the multi-source fine-grained data is input into the H deformation-induced analysis channels to perform deformation feature analysis, yielding H deformation assembly defect results. These results include defect spatial location, defect-induced assembly, and defect type. Based on the H defect spatial locations, cross-validation of the H deformation assembly defects is performed, outputting the device assembly defect.

[0063] The primary objective of the deformation-induced analysis model is to analyze the deformation in the wellbore caused by changes in equipment or the environment, and how these deformations affect the wellbore's structure or equipment functionality. The model comprises multiple parallel deformation-induced analysis channels. This parallel channel design allows the system to process data from different detection devices in parallel, with each channel independently processing a specific type of deformation data to avoid interference between different data types. Each deformation-induced analysis channel corresponds to a specific type of detection device and is responsible for a specific type of analysis task. In other words, data collected by different detection devices (such as laser, ultrasonic, and infrared sensors) is processed through the corresponding analysis channel.

[0064] The detection equipment scheduling instruction includes the types of equipment to be activated, namely H scheduling equipment types. Based on these H scheduling equipment types, the corresponding H deformation-induced analysis channels are matched and activated. For example, if the detection equipment scheduling instruction requires the use of a laser scanning device, the deformation-induced analysis channel related to laser analysis is activated, and internal defect analysis is performed based on the data collected by the laser scanning device.

[0065] Multi-source fine-grained data consists of high-precision data collected from multiple detection devices. This data is transmitted to various deformation-induced analysis channels based on device type, with each channel specializing in processing data from a specific type of device. Deformation feature analysis is performed within each channel. For example, laser data is used to detect geometrical deviations in the elevator shaft; ultrasonic data is used to identify structural defects within the shaft, such as cracks and corrosion; and infrared data is used to identify equipment problems caused by overheating in the elevator shaft. Each deformation-induced analysis channel outputs corresponding deformation assembly defect results. The defect spatial location indicates the coordinates of the defect or deformation, typically represented by three-dimensional coordinates, such as x=1.2m, y=0.8m, z=10.5m; the defect-induced assembly is the assembly component that has undergone deformation or defect; and the defect type includes cracks, deformation, corrosion, and thermal anomalies.

[0066] The purpose of cross-validation is to ensure that different data sources are consistent in terms of space and defect characteristics, reduce false alarms and false negatives, and improve the confidence of defect location. Each data source corresponds to a deformation-induced analysis channel. By identifying the offset type and associated component group of a specific region through the previous step, the offset type and associated component group of each data source are aggregated based on regional consistency to obtain the assembly defects of the device with high confidence, including assembly component type groups and defect regions.

[0067] Furthermore, based on the H spatial locations of defects, cross-validation of the H deformation assembly defects is performed to output the device assembly defects. The method includes:

[0068] Spatial overlap is calculated for the H spatial locations of defects to filter and aggregate multiple sets of deformation assembly defect results with an overlap of ≥80%; defect causal association topology is obtained interactively; based on the defect causal association topology, causal relationship cross-validation is performed on the multiple sets of deformation assembly defect results to output multiple local assembly defects, which constitute the device assembly defects.

[0069] Spatial overlap refers to the degree of spatial overlap between two or more defect locations. It is obtained by calculating the intersection of defect regions in three-dimensional space. For example, if the spatial regions of two defect locations (such as geometric deformation or crack areas) have a large overlap, the overlap is high. By comparing the spatial coordinate range of each defect, their overlapping area or volume is calculated. If the spatial overlap of two defect locations is greater than or equal to 80%, the two defects are considered related and may belong to the same assembly defect. All defects with a spatial overlap exceeding 80% are screened out. These defects indicate that they occurred in the same or adjacent areas and are likely caused by the same reason. These screened defects are aggregated to form multiple sets of deformation assembly defect results, serving as the basis for further analysis.

[0070] The defect causal association topology describes the causal logical relationship between defects, such as a bracket crack leading to a guide rail bend. Each defect (e.g., bracket crack, guide rail bend) is a node in the topology, and nodes are connected by directed edges, where edges represent causal relationships and arrows point to the affected components. The constructed defect causal association topology contains multiple levels of defects and their interrelationships. Through this topological structure, it is possible to more clearly show how one defect triggers other defects.

[0071] Based on the constructed defect causal relationship topology, the causal logic between different deformation assembly defects is verified. If one defect may lead to other defects, it is verified whether these defects occur according to the causal relationship. For example, if a crack is detected in a bracket, topological reasoning is used to verify whether it will cause the guide rail to bend. If this causal relationship holds, the association of the local assembly defect "the bracket crack causes the guide rail to bend" is confirmed. A local assembly defect refers to a defect found in a specific area or component. For example, "the bracket crack on the east side of floors 3-5 causes the guide rail to bend" is a specific local defect, encompassing the defect location, type, and causal relationship. All verified local assembly defects are combined to form a complete set of device assembly defects.

[0072] Furthermore, by backtracking deviation factors based on the spatial envelope of the region to obtain the induced prediction result, the method includes:

[0073] After projecting the spatial envelope of the region onto the standard shaft model, the absolute position of the shaft space, the relative position of the shaft components, and the regional shaft environment features are extracted. The bounding box geometric features of the spatial envelope of the region are extracted. A regional defect feature vector is constructed based on the absolute position of the shaft space, the relative position of the shaft components, and the bounding box geometric features. The regional defect feature vector is loaded into a pre-built elevator shaft defect case database, and similar cases are matched based on vector cosine similarity to obtain N historical defect types. The regional shaft environment features are used to perform reproduction correction and screening on the N historical defect types to obtain Q historical defect types, which are used as the induction prediction results.

[0074] The three-dimensional coordinate information of the spatial envelope of the region is projected onto the standard shaft model. The absolute position of the shaft space refers to the absolute coordinates of the specific location of the shaft in three-dimensional space, such as the specific location of the start point, end point, or any calibration point of the shaft, used to locate the location of specific defects within the spatial envelope of the region, such as guide rails, shaft support structures, etc. The relative position of shaft components refers to the relative positional relationship between various components within the shaft (such as elevator guide rails, brackets, switching devices, etc.), such as the position of the bracket relative to the guide rail, or the position of the elevator car relative to the shaft wall. The environmental characteristics of the regional shaft include environmental data such as the temperature, humidity, vibration, and lighting conditions of the shaft. These characteristics may affect the use of the shaft and the performance of the components. For example, an excessively humid shaft environment may lead to corrosion, while excessive vibration may cause loosening or deformation of mechanical components.

[0075] Minimum bounding box calculation is performed on all points in the spatial envelope of the region. This is done by finding the minimum and maximum coordinate values ​​of the spatial envelope in three dimensions (x, y, z), thereby determining the boundary of the bounding box and obtaining the characteristics of the bounding box such as length, width, height, and volume.

[0076] The absolute position of the shaft space is transformed into a numerical vector and used as part of the feature vector. The relative position of the shaft components is represented as a numerical value. For example, the difference in relative position can be used as a feature term to quantify the spatial layout problem. The bounding box geometric features are used as feature terms to describe the spatial distribution characteristics of the area. All of this information is combined into a regional defect feature vector.

[0077] The elevator shaft defect case database contains a large amount of historical defect type information. This information includes not only specific descriptions of the defects but also spatial features related to the defects, such as dimensions, location, and environmental conditions. Cases in the database are categorized according to different defect types, and each case is provided with its spatial characteristics. The defect feature vectors of the aforementioned regions are loaded into this elevator shaft defect case database for matching, quickly finding historical cases similar to the current feature vector. Cosine similarity is a commonly used method to measure the similarity between two vectors. If the cosine similarity value of the feature vectors is high, it means that the current region and historical cases are highly similar in terms of spatial layout, geometric features, and component positions.

[0078] The calculation results return the N most similar historical defect types, where N is a positive integer. These historical defect types represent historical defects that are highly similar to the current regional spatial characteristics. For example, if the regional spatial envelope characteristics are "3rd-5th floor east side well wall, length 2.0m × width 1.0m × height 0.15m, unevenness 15mm, distance from guide rail support 0.25m", the database "similar features" cases may return "Defect type 1: Well wall structural crack (caused by incomplete concrete pouring)" and "Defect type 2: Guide rail support installation deviation (caused by squeezing the well wall, resulting in unevenness)".

[0079] The environmental characteristics of a hoistway can affect components within the hoistway to varying degrees, leading to different types of defects. For example, a humid environment may cause corrosion, temperature changes may cause material expansion or contraction, and excessive vibration may cause components to loosen. The environmental characteristics of the hoistway are compared with N historical defect types, and the N historical defect types are corrected. For example, the vibration characteristics of the current area may amplify or worsen certain historical defect types (such as "guide rail bracket installation deviation"). Therefore, the predicted defect types are corrected based on the environmental characteristics of the hoistway. After environmental characteristic correction, Q historical defect types are selected. These corrected historical defect types are more consistent with the specific conditions and environment of the current area. The obtained Q historical defect types are used as the inducing prediction results. These defect types have a high probability of occurrence in the current area and are more consistent with the influence of environmental characteristics.

[0080] Furthermore, it also includes:

[0081] Using the first scheduling equipment type as a constraint, assembly defect data is collected to obtain a sample detection dataset and a sample defect identifier set, wherein the sample defect identifier is composed of the sample spatial location, the sample inducing assembly, and the sample type; a first deformation-induced analysis channel is constructed based on a CNN network architecture; the sample detection dataset and the sample defect identifier set are used as training data to optimize the defect identification and labeling performance of the first deformation-induced analysis channel.

[0082] The first scheduling device type is any one of the H scheduling device types, used as the current analysis object; for example, a laser scanning device is selected. The data acquisition method and device selection are restricted based on the first scheduling device type. Specific device types will affect the subsequent data acquisition method, resolution, and accuracy. After selecting the device type, data acquisition begins. The data acquisition process focuses on assembly defects in the wellbore structure, especially equipment-related defects, such as acquiring information on geometric deviations, deformations, or cracks in guide rails, supports, well walls, and mechanical components.

[0083] The sample detection dataset is a set of data actually collected from the wellbore. Each data sample represents the physical measurement information of a spatial point or area. The data may include location, size, thickness, crack information, temperature distribution, etc. The sample defect identification dataset has a label for each collected data sample, indicating whether a defect exists in the data point or area and the type of defect. The sample spatial location indicates the location of the data point or area in space, such as "3-5 layers east side well wall". The sample induced assembly indicates the specific assembly type, such as "guide rail bracket". The sample type indicates the type of defect, such as "crack" or "deviation".

[0084] CNN (Convolutional Neural Network) is a deep learning model widely used in image and spatial data processing. In this step, CNN is used to process multi-source data related to defects. The first deformation-induced analysis channel is built based on the CNN network architecture. Each channel processes different defect types or data sources. Multiple channels work in parallel to process data from different types of devices.

[0085] The collected sample detection dataset and corresponding sample defect label set are used as training data samples. Each training data sample contains input data (such as laser scan point cloud data) and corresponding labels (such as defect type). Through supervised learning, the CNN network uses these training data samples to learn how to associate input data with defect labels. By continuously adjusting the weights and parameters in the network, the first deformation-induced analysis channel can gradually identify defect features in new and unseen data samples. The optimization goal of the first deformation-induced analysis channel is to improve the accuracy of defect identification. By adjusting the parameters of the CNN network, the first deformation-induced analysis channel can better identify different types of assembly defects, such as cracks, deviations, and wear. During training, the first deformation-induced analysis channel is optimized according to the loss function (such as cross-entropy loss) to gradually reduce prediction errors, thereby improving identification accuracy. After sufficient training, the first deformation-induced analysis channel can effectively identify and calibrate wellbore assembly defects and provide high-confidence prediction results.

[0086] In summary, the elevator shaft detection method provided in this application has the following technical effects:

[0087] By constructing a real-time elevator shaft model through point cloud acquisition and comparing it with a standard shaft model, areas in the shaft that differ from the standard model can be accurately identified and located. This method can accurately identify discrepancies between design and actual construction, ensuring the structural integrity and accuracy of the shaft. Pre-deploying a shaft detection array within the shaft and maintaining a low-power standby state allows for rapid activation of detection nodes for data acquisition when needed, avoiding unnecessary energy consumption. Triggering M detection nodes for efficient data acquisition enables precise data acquisition in key areas within the shaft, ensuring comprehensive monitoring of all parts of the shaft. Backtracking deviation factors based on the spatial envelope of the area effectively predicts potential faults or problems in the elevator shaft, identifying potential assembly defects in advance. By inducing the generation of predictive results, preventative measures can be taken in advance to reduce the probability of faults, providing decision support for subsequent precise detection and repair work. Through the analysis of shaft detection nodes… The detection parameters of the points can be adjusted dynamically according to real-time needs, avoiding redundant data collection from "comprehensive scanning" and improving data collection efficiency. The collection of multi-source fine-grained data enables refined monitoring of the shaft from different angles and scales, greatly improving the accuracy and precision of defect identification. By fusing multi-source fine-grained data for deformation-induced analysis, deformation areas caused by external or internal factors within the shaft can be accurately identified, locating assembly defects in the equipment within the shaft. This precise defect location technology can improve maintenance and management efficiency and reduce equipment failures caused by assembly defects. By matching the trigger timestamps of the shaft detection nodes with the assembly defects, the detection data can be effectively integrated with operation and maintenance management rules to output real-time operation and maintenance strategies. In this way, operation and maintenance personnel can take appropriate measures to repair or adjust the shaft in a timely manner based on real-time data, improving the automation level of elevator shaft management and reducing the need for manual intervention.

[0088] Example 2, based on the same inventive concept as the elevator shaft detection method in the aforementioned examples, such as... Figure 2 As shown in the figure, this application embodiment provides an elevator shaft detection system, the system comprising:

[0089] The model difference region localization module 10 is used to locate the model difference region by comparing the standard shaft model and the real-time shaft model after collecting point cloud data of the elevator shaft to be modeled; the shaft detection node triggering module 20 is used to trigger M shaft detection nodes in the shaft detection array according to the regional spatial envelope of the model difference region, wherein the shaft detection array is pre-deployed in the elevator shaft to be modeled and the shaft detection array is in a low-power standby state; the deviation factor backtracking module 30 is used to backtrack the deviation factors according to the regional spatial envelope to obtain the inducing factors. The system includes: a prediction result; a multi-source fine-grained data acquisition module 40, used to adjust the detection parameters of the M wellbore detection nodes based on the regional spatial envelope and the induced prediction result, and then drive the M wellbore detection nodes to acquire multi-source fine-grained data; a device assembly defect location module 50, used to fuse and analyze the multi-source fine-grained data to perform deformation-induced analysis of the model difference region and locate device assembly defects; and an operation and maintenance management rule matching module 60, used to match operation and maintenance management rules based on the trigger timestamps of the M wellbore detection nodes and the device assembly defects, and output real-time operation and maintenance strategies.

[0090] Furthermore, the wellhead detection node triggering module 20 is used to perform the following operation steps:

[0091] Extract the three-dimensional spatial coordinate range of the model difference region to calculate the minimum bounding box boundary coordinates and output the spatial envelope of the region; locally call multiple installation spatial coordinates of multiple dormant detection nodes in the wellbore detection array; define multiple detection adjustment intervals based on the standard wellbore model and multiple installation spatial coordinates; solve for multiple actual wellbore detection ranges based on the inherent detection coverage range of the standard detection node and the multiple detection adjustment intervals; calculate multiple detection coverage ranges of the multiple actual wellbore detection ranges relative to the spatial envelope of the region; and locate the M wellbore detection nodes by filtering, splicing, and optimizing the multiple detection coverage ranges.

[0092] Furthermore, the wellhead detection node triggering module 20 is used to perform the following operation steps:

[0093] Using the spatial envelope of the region as a coverage constraint, the multiple detection coverage areas are enumerated and stitched together to obtain multiple detection overlap rates, multiple detection coverage rates, and multiple detection node call volumes for multiple stitching strategies. The multiple detection coverage rates are traversed according to a preset coverage threshold to filter out P stitching strategies. The P detection overlap rates, P detection coverage rates, and P detection node call volumes of the P stitching strategies are weighted and quantized to obtain the overall superiority of the P strategies. The overall superiority of the P strategies is serialized, the target stitching strategy is located, and the M wellhead detection nodes are triggered based on the node composition of the target stitching strategy.

[0094] Furthermore, the multi-source fine-grained data acquisition module 40 is used to perform the following operation steps:

[0095] The system locally calls upon Q detection device types and Q detection parameter requirements for Q historical defect types from the induced prediction results; by aggregating the Q detection device types and Q detection parameter requirements, it outputs detection device scheduling instructions and detection device parameter tuning instructions; it retrieves M node spatial attributes of the M wellbore detection nodes from the target splicing strategy; after activating the M wellbore detection nodes using the detection device scheduling instructions, it uses the detection device parameter tuning instructions to configure the parameters of the activated devices for the M wellbore detection nodes; it uses the M node spatial attribute mapping to perform spatial orientation calibration of the M wellbore detection nodes; and it drives the M wellbore detection nodes to collect multi-source fine-grained data from the model difference region.

[0096] Furthermore, the defect location module 50 in the device is used to perform the following operational steps:

[0097] A pre-constructed deformation-induced analysis model is established, comprising multiple parallel deformation-induced analysis channels, each channel bearing a detection device type identifier. Based on H scheduling device types in the detection device scheduling instruction, the H deformation-induced analysis channels in the model are matched and activated. According to the H scheduling device types, the multi-source fine-grained data is input into the H deformation-induced analysis channels to perform deformation feature analysis, yielding H deformation assembly defect results. These results include defect spatial location, defect-induced assembly, and defect type. Based on the H defect spatial locations, cross-validation of the H deformation assembly defects is performed, outputting the device assembly defect.

[0098] Furthermore, the defect location module 50 in the device is used to perform the following operational steps:

[0099] Spatial overlap is calculated for the H spatial locations of defects to filter and aggregate multiple sets of deformation assembly defect results with an overlap of ≥80%; defect causal association topology is obtained interactively; based on the defect causal association topology, causal relationship cross-validation is performed on the multiple sets of deformation assembly defect results to output multiple local assembly defects, which constitute the device assembly defects.

[0100] Furthermore, the deviation factor backtracking module 30 is used to perform the following operation steps:

[0101] After projecting the spatial envelope of the region onto the standard shaft model, the absolute position of the shaft space, the relative position of the shaft components, and the regional shaft environment features are extracted. The bounding box geometric features of the spatial envelope of the region are extracted. A regional defect feature vector is constructed based on the absolute position of the shaft space, the relative position of the shaft components, and the bounding box geometric features. The regional defect feature vector is loaded into a pre-built elevator shaft defect case database, and similar cases are matched based on vector cosine similarity to obtain N historical defect types. The regional shaft environment features are used to perform reproduction correction and screening on the N historical defect types to obtain Q historical defect types, which are used as the induction prediction results.

[0102] Furthermore, the defect location module 50 in the device is used to perform the following operational steps:

[0103] Using the first scheduling equipment type as a constraint, assembly defect data is collected to obtain a sample detection dataset and a sample defect identifier set, wherein the sample defect identifier is composed of the sample spatial location, the sample inducing assembly, and the sample type; a first deformation-induced analysis channel is constructed based on a CNN network architecture; the sample detection dataset and the sample defect identifier set are used as training data to optimize the defect identification and labeling performance of the first deformation-induced analysis channel.

[0104] Through the foregoing detailed description of an elevator shaft detection method, those skilled in the art can clearly understand the elevator shaft detection system of this embodiment. Since it corresponds to the method disclosed in the embodiment, the description is relatively simple, and relevant parts can be referred to the method section.

[0105] Example 3 provides a storage medium on which a computer program is stored, which, when executed by a processor, implements any step of Example 1.

[0106] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0107] The present invention has been described in detail above. However, modifications or improvements can be made to it, which will be obvious to those skilled in the art. Therefore, any modifications or improvements that do not depart from the spirit of the present invention are within the scope of protection of the present invention.

Claims

1. A method for detecting elevator shafts, characterized in that, The method includes: Point cloud acquisition equipment is used to scan and collect point cloud data of the elevator shaft to be modeled. After collecting point cloud data in the elevator shaft to be modeled to build a real-time shaft model, the model difference area is located by comparing the standard shaft model and the real-time shaft model. The model difference area is the gap between the actual shaft and the design standard. Based on the spatial envelope of the model difference region, M shaft detection nodes are triggered in the shaft detection array. The shaft detection array is pre-deployed in the elevator shaft to be modeled and is in a low-power standby state. The spatial envelope is the smallest three-dimensional bounding box that surrounds the model difference region. A regional defect feature vector is constructed based on the spatial envelope. The regional defect feature vector is loaded into a pre-constructed elevator shaft defect case database. Based on the vector cosine similarity, similar cases are matched to backtrack the deviation factors and obtain the induction prediction result. Based on the regional spatial envelope and the induced prediction results, after adjusting the detection parameters of the M wellhead detection nodes, the M wellhead detection nodes are driven to collect multi-source fine-grained data. This multi-source fine-grained data includes, but is not limited to, data from lidar equipment, ultrasonic detection equipment, and infrared imaging equipment. The multi-source fine-grained data is then fused and analyzed to perform deformation-induced analysis of the model's difference regions. After obtaining deformation assembly defect results, cross-validation of deformation assembly defects is performed to locate assembly defects. The deformation assembly defect results include the spatial location of the defect, the defect-induced assembly, and the defect type. Based on the trigger timestamps of the M wellhead detection nodes and the assembly defects of the device, the operation and maintenance management rules are matched to output a real-time operation and maintenance strategy.

2. The elevator shaft detection method as described in claim 1, characterized in that, The method involves triggering M wellhead detection nodes in the wellhead detection array based on the spatial envelope of the model's difference region. Extract the three-dimensional spatial coordinate range of the model difference region to calculate the minimum bounding box boundary coordinates and output the spatial envelope of the region; The system locally retrieves the installation space coordinates of multiple dormant detection nodes in the wellbore detection array; Based on the standard wellbore model and multiple installation space coordinates, multiple detection and adjustment ranges are defined; Based on the inherent detection coverage of the standard detection node and the multiple detection adjustment intervals, the detection ranges of multiple actual wells are determined. Calculate the multiple detection coverage ranges of the multiple actual wellbore detection ranges on the spatial envelope of the region; By filtering and stitching together the multiple detection coverage areas for optimization, the M wellhead detection nodes are located.

3. The elevator shaft detection method as described in claim 2, characterized in that, The method for locating the M wellbore detection nodes by filtering, stitching, and optimizing the multiple detection coverage areas includes: Using the spatial envelope of the region as the coverage constraint, the multiple detection coverage areas are enumerated and spliced ​​to obtain multiple detection overlap rates, multiple detection coverage rates, and multiple detection node call volumes for multiple splicing strategies; Based on a preset coverage threshold, the multiple detection coverages are traversed to filter and obtain P stitching strategies; The overall superiority of the P strategies is obtained by weighting and quantizing the P detection overlap rates, P detection coverage rates, and P detection node call volumes of the P splicing strategies. The overall superiority of the P strategies is serialized, the target splicing strategy is located, and the M wellhead detection nodes are triggered based on the node composition of the target splicing strategy.

4. The elevator shaft detection method as described in claim 3, characterized in that, Based on the regional spatial envelope and induced prediction results, after adjusting the detection parameters of the M wellbore detection nodes, the M wellbore detection nodes are driven to collect multi-source fine-grained data. The method includes: The system locally invokes the Q detection device types and Q detection parameter requirements for the Q historical defect types in the induction prediction results; By aggregating the Q types of detection devices and the Q requirements of detection parameters, detection device scheduling instructions and detection device parameter adjustment instructions are output; Retrieve the spatial attributes of the M wellhead detection nodes from the target stitching strategy; After activating the M wellhead detection nodes using the detection equipment scheduling command, the parameter configuration of the activated equipment is performed on the M wellhead detection nodes using the detection equipment parameter adjustment command; The spatial orientation calibration of the M wellhead detection nodes is performed using the spatial attribute mapping of the M nodes. The M wellhead detection nodes are driven to collect multi-source fine-grained data of the model difference region.

5. The elevator shaft detection method as described in claim 4, characterized in that, The method involves fusing and analyzing the multi-source fine-grained data to perform deformation-induced analysis of the model's differential regions, thereby locating assembly defects in the device. A pre-constructed deformation-induced analysis model is provided, wherein the deformation-induced analysis model includes multiple deformation-induced analysis channels connected in parallel, and each deformation-induced analysis channel is labeled with a detection device type identifier; Based on the H scheduling device types in the detection device scheduling instruction, match and activate the H deformation-induced analysis channels in the deformation-induced analysis model; Based on the H types of scheduling devices, the multi-source fine-grained data is input into the H deformation-induced analysis channels to perform deformation feature analysis, thereby obtaining H deformation assembly defect results, wherein the deformation assembly defect results include defect spatial location, defect-induced assembly, and defect type; Based on the spatial locations of the H defects, cross-validation of the H deformation assembly defects is performed, and the assembly defects of the device are output.

6. The elevator shaft detection method as described in claim 5, characterized in that, Based on the spatial locations of H defects, cross-validation of the H deformation assembly defects is performed, and the assembly defects of the device are output. The method includes: The spatial overlap of the H defect locations is calculated to filter out multiple sets of deformation assembly defect results with an aggregate overlap of ≥80%. Interactively obtain the causal relationship topology of defects; Based on the aforementioned defect causal relationship topology, the causal relationship of the multiple sets of deformation assembly defect results is cross-validated to output multiple local assembly defects, which constitute the device assembly defects.

7. The elevator shaft detection method as described in claim 1, characterized in that, The method involves backtracking deviation factors based on the spatial envelope of the region to obtain the induced prediction result. After projecting the spatial envelope of the region onto the standard shaft model, the absolute position of the shaft space, the relative position of the shaft components, and the regional shaft environment characteristics are extracted. Extract the bounding box geometric features of the spatial envelope of the region; A region defect feature vector is constructed based on the absolute position of the shaft space, the relative position of the shaft components, and the bounding box geometry. The regional defect feature vector is loaded into a pre-built elevator shaft defect case database, and similar cases are matched based on vector cosine similarity to obtain N historical defect types. The N historical defect types are reproduced, corrected, and screened using the regional well environment characteristics to obtain Q historical defect types, which are used as the induction prediction results.

8. The elevator shaft detection method as described in claim 5, characterized in that, Also includes: Using the first scheduling equipment type as a constraint, assembly defect data is collected to obtain a sample detection dataset and a sample defect identifier set, wherein the sample defect identifier consists of the sample spatial location, the sample-inducing assembly, and the sample type; A first deformation-induced analysis channel was constructed based on a CNN network architecture; The sample detection dataset and sample defect identifier set are used as training data to optimize the defect identification and identifier performance of the first deformation-induced analysis channel.

9. A detection system for elevator shafts, characterized in that, A method for detecting an elevator shaft according to any one of claims 1-8, the system comprising: The model difference region localization module is used to locate the model difference region by comparing the standard shaft model and the real-time shaft model after the point cloud collection of the elevator shaft to be modeled is used to construct a real-time shaft model. The hoistway detection node triggering module is used to trigger M hoistway detection nodes in the hoistway detection array according to the regional spatial envelope of the model difference region, wherein the hoistway detection array is pre-deployed in the elevator hoistway to be modeled, and the hoistway detection array is in a low-power standby state. The deviation factor backtracking module is used to backtrack deviation factors based on the spatial envelope of the region to obtain the induced prediction result; The multi-source fine-grained data acquisition module is used to drive the M wellhead detection nodes to acquire multi-source fine-grained data after adjusting the detection parameters of the M wellhead detection nodes based on the regional spatial envelope and the induced prediction results. The device assembly defect location module is used to fuse and analyze the multi-source fine-grained data to perform deformation-induced analysis of the model difference region and locate device assembly defects. The operation and maintenance management rule matching module is used to match operation and maintenance management rules based on the trigger timestamps of the M wellhead detection nodes and the assembly defects of the device, and output real-time operation and maintenance strategies.

10. A storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the steps of the elevator shaft detection method according to any one of claims 1 to 8.

Images