Construction safety early warning training system based on BIM building information model

Abstract

The invention belongs to the technical field of construction, and aims to solve the problems that an existing construction safety training mode lacks reality and interactivity, a safety early warning rule is single and lacks association with a training link, and a training effect evaluation mode cannot comprehensively and accurately evaluate the ability of constructors. Therefore, the invention provides a construction safety early warning training system based on a BIM building information model, and the system comprises a data management module which sequentially collects, preprocesses and stores multi-source data; the BIM model processing module is used for carrying out lightweight processing on model data, constructing a model and updating the model in real time; the VR training scene generation module creates a virtual construction site and interacts with a user; the safety early warning module sets an early warning rule and sends a prompt to the VR training scene generation module; and the training effect evaluation module evaluates training based on the training effect evaluation model. The reality sense and interactivity of construction safety training can be improved, potential safety hazards can be found more accurately, the ability of constructors can be evaluated more accurately, and the safety awareness of the constructors can be continuously improved.

Description

Technical Field

[0001] The present invention relates to the technical field of building construction, and particularly to a construction safety early warning training system based on a BIM building information model. Background Art

[0002] In the context of the booming development of the current construction industry, construction safety has always been of utmost importance. With the increasing scale of construction projects and the growing complexity of construction techniques, the safety risks faced during the construction process are also increasing day by day. Existing construction safety training methods have many deficiencies and are difficult to meet the requirements of modern construction safety management.

[0003] Existing construction safety training often relies on written materials, oral explanations, and simple on-site demonstrations. This method lacks a sense of reality and interactivity, and it is difficult for construction workers to truly understand and master complex safety operation specifications and emergency handling methods. For example, when explaining the safety knowledge of working at heights, it is difficult for construction workers to intuitively feel the dangerous environment of working at heights and the importance of correct operations through only words and pictures, resulting in possible safety accidents due to improper operations during actual construction. At the same time, the safety management of the construction site also faces severe challenges. The management of data is relatively scattered, lacking effective integration and analysis. This makes it difficult for safety management personnel to comprehensively and timely grasp the safety status of the construction site and unable to discover potential safety hazards in a timely manner.

[0004] Some existing safety early warning systems can monitor some safety indicators, but the early warning rules are often relatively single and cannot be flexibly adjusted according to the complex conditions of the construction site. Moreover, these systems lack an effective connection with the training link of construction workers, resulting in the early warning information not being transmitted to construction workers in a timely manner, or construction workers not knowing how to correctly respond after receiving the early warning information. In terms of the evaluation of training effects, existing evaluation methods mainly rely on theoretical examinations and simple operation assessments, and cannot comprehensively and accurately evaluate the operation ability, emergency response ability, and teamwork ability of construction workers in actual construction scenarios. For example, in a simulated emergency accident scenario, it is impossible to accurately record the operation behaviors and response times of construction workers, making it difficult to objectively evaluate their emergency handling abilities.

[0005] In view of this, there is a need in the art for a construction safety early warning training system based on a BIM building information model to solve the above problems. Summary of the Invention

[0006] In order to solve the above technical problems, that is, to solve the problems that existing construction safety training methods lack a sense of reality and interactivity, the early warning rules are single and lack a connection with the training link, and the training effect evaluation method cannot comprehensively and accurately evaluate the abilities of construction workers.

[0007] The present invention provides a construction safety early warning training system based on a BIM building information model, and the system includes:

[0008] A data management module, which is used to collect multi-source data and preprocess and store the multi-source data. Among them, the multi-source data includes construction site environment data, BIM model original data, and construction progress and personnel arrangement data;

[0009] A BIM model processing module, which is used to retrieve the preprocessed BIM model data from the data management module, perform lightweight processing, and construct a BIM building information model, and perform real-time updates on the BIM building information model according to the preprocessed construction site environment data, construction progress, and personnel arrangement data;

[0010] A VR training scenario generation module, which is used to create a virtual construction scenario based on the BIM building information model and interact with users;

[0011] A safety early warning module, which is used to set early warning rules, monitor the multi-source data based on the early warning rules, and send a prompt to the VR training scenario generation module when an abnormality occurs;

[0012] A training effect evaluation module, which is used to collect the interaction behavior data between the user and the VR training scenario generation module during the user's training and evaluate the training based on a training effect evaluation model.

[0013] In some preferred embodiments, the BIM model processing module uses a lightweight algorithm based on triangle mesh simplification to process BIM model data:

[0014] Extract the set of triangular patches that need to be lightweight processed from the BIM model data. For each triangular patch t BIM , obtain the coordinates of its three vertices, which are respectively denoted as v1 = (x1, y1, z1), v2 = (x2, y2, z2), and v3 = (x3, y3, z3);

[0015] Substitute the vertex coordinates v1, v2, and v3 into the plane equation ax + by + cz + d = 0, and calculate the coefficients a, b, c, and d;

[0016] For the point p = (x, y, z), according to the formula Calculate the squared distances from the three vertices v1, v2, and v3 to this plane respectively;

[0017] According to the formula Q(t BIM ) = d 2 (v1, t BIM ) + d 2 (v2, t BIM) + d 2 (v3, t BIM ) to calculate the error metric Q(t BIM ) for each triangular patch t BIM );

[0018] Traverse all mergable edges and calculate the increased error after merging;

[0019] Select the edge with the minimum increased error for merging;

[0020] Repeat the merging process until the lightweight processing termination condition is met, end the lightweight processing process, and obtain the lightweight BIM model.

[0021] In some preferred embodiments, the condition for meeting the lightweight processing termination condition is that the number of triangular patches after merging reaches a preset number of triangular patches; or,

[0022] The condition for meeting the lightweight processing termination condition is that the increased error of the model after merging edges reaches a set threshold.

[0023] In some preferred embodiments, the construction site environment data at least includes environmental temperature data, environmental humidity data, light intensity data, dust data, noise data, wind speed and direction data, harmful gas data, and smoke data.

[0024] In some preferred embodiments, the data management module includes a data processing sub-module; the data processing sub-module uses a denoising algorithm based on wavelet transform to remove the noise components of the construction site environment data:

[0025] For each data sequence f(t) in the construction site environment data, select the Daubechies wavelet basis function ψ(t);

[0026] Set the value ranges of the scale parameter a1 and the translation parameter b1;

[0027] According to the formula Within the determined ranges of the scale parameter a1 and the translation parameter b1, calculate the wavelet coefficients W f (a1, b1);

[0028] Perform threshold processing on the calculated wavelet coefficients W f (a1, b1), where the threshold processing method is hard threshold processing, soft threshold processing, or SureShrink threshold processing;

[0029] For the wavelet coefficients after threshold processing, calculate the denoised data sequence f through the inverse wavelet transform formula denoised (t).

[0030] In some preferred embodiments, the data processing sub-module also smooths the discrete data sequence by using a median filtering algorithm based on a sliding window:

[0031] Regarding the data sequence f denoised (t) after wavelet denoising as a new discrete signal sequence x(n), and setting the sliding window size N;

[0032] Starting from the starting position of the data sequence, for each window position n, obtain the signal values within the window

[0033] Sort the signal values within the window and take the median value as the output after median filtering;

[0034] Keep sliding the window on the data sequence and calculate the median value at each position, and finally obtain a data sequence y(n) after median filtering.

[0035] In some preferred embodiments, the data management module further includes a data storage sub-module. The data storage sub-module includes a relational database and a non-relational database. The relational database is a MySQL database, and the non-relational database is a MongoDB database.

[0036] In some preferred embodiments, the VR training scenario generation module includes a virtual scenario construction sub-module and an interaction element adding sub-module;

[0037] The virtual scenario construction sub-module is used to create all scenario elements of the virtual construction scenario based on the BIM building information model and optimize the layout of the virtual scenario;

[0038] The interaction element adding sub-module is used to add interactive components to the construction equipment in the virtual scenario.

[0039] In some preferred embodiments, the training effect evaluation model is:

[0040]

[0041] Among them, S is the comprehensive evaluation score, T avg is the average time for the user to complete a specific training task, M step is the matching degree of operation steps, P d is the proportion of key element attention, f is the perspective change frequency, f max is the maximum value of the perspective change frequency of all users, E is the total number of errors, R avg is the average error correction time, R max is the maximum value of the average error correction time of all users, and α, β, and γ are weight coefficients.

[0042] In some preferred embodiments, the system further includes a training course management module, which is used to push training courses to users and record the learning progress.

[0043] The construction safety early warning training system based on the BIM building information model of the present invention has the following beneficial effects:

[0044] In the construction safety early warning training system based on the BIM building information model of the present invention, the data management module can efficiently collect the construction site environment data, the original BIM model data, and the construction progress and personnel arrangement data, and perform preprocessing and storage to realize the effective integration of multi-source data, changing the previous situation of decentralized data management, enabling safety management personnel to obtain comprehensive construction site data in real time, providing strong support for safety decision-making, and improving the timeliness and accuracy of safety management; the BIM model processing module retrieves data from the data management module, not only performs lightweight processing on the BIM model, but also can update it in real time according to the real-time data of the construction site, which enables the BIM building information model to accurately reflect the dynamic changes of the construction site, providing an accurate and real-time model basis for safety early warning and training, and avoiding the defect of the disconnection between the BIM model and the real-time data of the construction site; the VR training scenario generation module creates a virtual construction scenario based on the BIM building information model and interacts with users, providing an immersive training experience for construction workers, solving the problems of lack of realism and interactivity in the existing training methods, enabling construction workers to better understand and master safety operation specifications and emergency handling methods, and greatly improving the training effect; the safety early warning module can comprehensively monitor multi-source data by setting flexible early warning rules, and send prompts to the VR training scenario generation module in a timely manner when abnormalities occur. Compared with the existing single early warning rule, it can more accurately discover potential safety hazards and transmit the early warning information to construction workers in a timely manner, improving the ability of construction workers to respond to safety risks; the training effect evaluation module collects the interaction behavior data between the user and the VR training scenario generation module during user training and conducts an evaluation based on a scientific training effect evaluation model, overcoming the limitations of the existing evaluation methods, being able to comprehensively and accurately evaluate the various abilities of construction workers in the actual construction scenario, providing detailed basis for subsequent training optimization, and continuously improving the safety awareness of construction workers. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] The features and advantages of the present invention will be more clearly understood by referring to the accompanying drawings. The drawings are schematic and should not be construed as imposing any limitation on the present invention. In the drawings:

[0046] Figure 1 is a flowchart of the construction safety early warning training system based on the BIM building information model of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0047] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention. Apparently, the described embodiments are some, but not all, of the embodiments of the present invention. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0048] Based on the problems pointed out in the background technology that the existing construction safety training methods lack a sense of reality and interactivity, the safety warning rules are single and lack relevance to the training link, and the training effect evaluation method cannot comprehensively and accurately evaluate the capabilities of construction workers, the present invention provides a construction safety warning training system based on the BIM building information model, aiming to improve the sense of reality and interactivity of construction safety training, more accurately discover potential safety hazards and evaluate the capabilities of construction workers, and continuously enhance their safety awareness.

[0049] As Figure 1 shown, the construction safety warning training system based on the BIM building information model of the present invention includes:

[0050] A data management module, which is used to collect multi-source data, preprocess and store the multi-source data. Among them, the multi-source data includes construction site environment data, BIM model original data, and construction progress and personnel arrangement data;

[0051] A BIM model processing module, which is used to retrieve the preprocessed BIM model data from the data management module, perform lightweight processing, and construct a BIM building information model, and update the BIM building information model in real time according to the preprocessed construction site environment data and construction progress and personnel arrangement data;

[0052] A VR training scenario generation module, which is used to create a virtual construction scenario based on the BIM building information model and interact with users;

[0053] A safety warning module, which is used to set warning rules, monitor multi-source data based on the warning rules, and send a prompt to the VR training scenario generation module when an abnormality occurs;

[0054] A training effect evaluation module, which is used to collect the interaction behavior data between the user and the VR training scenario generation module during the user's training and evaluate the training based on the training effect evaluation model.

[0055] Preferably, the data management module includes a data processing sub-module and a data storage sub-module. For the data processing sub-module, it can collect multi-source data and preprocess and store the multi-source data. Among them, the multi-source data includes construction site environment data (such as environmental temperature data, environmental humidity data, light intensity data, dust data, noise data, wind speed and direction data, harmful gas data, and smoke data), BIM model original data, and construction progress and personnel arrangement data. In addition, location tracking data can also be collected. Specifically, a wireless sensor network composed of temperature and humidity sensors, light sensors, dust sensors, noise sensors, wind speed and direction sensors, harmful gas sensors, smoke sensors, location tracking sensors, etc. can be deployed at the construction site, and low-power, self-organizing wireless communication methods such as ZigBee and LoRa are used to ensure stable communication between sensors and between sensors and data aggregation nodes. The location distribution of the sensors is planned according to the characteristics of the construction area and the requirements of safety monitoring. For example, location tracking sensors are mainly deployed in high-altitude operation areas, and smoke sensors are densely set at the storage locations of flammable and explosive materials. The data processing sub-module establishes a data interface with building design software (such as AutoCAD, Revit, etc.), and through a software development kit (SDK) or a data conversion plug-in, directly extracts CAD drawings and BIM model original data from the design software. During the extraction process, the data is preliminarily parsed to identify various elements such as building structures, components, and equipment and their attribute information. In addition, the data processing sub-module can also be integrated with a project management system (such as Oracle Primavera P6, Microsoft Project, etc.), and through a web service interface (RESTful API or SOAP API), regularly obtain data such as construction progress plans, personnel scheduling arrangements, and material procurement information, and perform format conversion and standardization processing on the obtained data to make it compatible with the data format within the system.

[0056] For the data that may be noisy collected by the sensors, the data processing sub-module adopts a denoising algorithm based on wavelet transform to effectively remove high-frequency noise interference while retaining the main features of the data.

[0057] Further preferably, the data processing sub-module adopts a denoising algorithm based on wavelet transform to remove the noise components of the construction site environment data:

[0058] For each data sequence f(t) in the construction site environment data, the Daubechies wavelet basis function ψ(t) is selected;

[0059] Set the value ranges of the scale parameter a1 and the translation parameter b1;

[0060] According to the formula Within the range of the determined scale parameter a1 and translation parameter b1, calculate the wavelet coefficients W at different scales and positions f (a1, b1);

[0061] Perform threshold processing on the calculated wavelet coefficients W f (a1, b1), where the threshold processing method is hard threshold processing, soft threshold processing, or SureShrink threshold processing;

[0062] For Daubechies wavelet coefficients, the hard threshold method directly sets the wavelet coefficients with absolute values less than the given threshold λ to 0, while the wavelet coefficients with absolute values greater than the threshold λ remain unchanged, that is:

[0063]

[0064] When the noise wavelet coefficients after Daubechies wavelet decomposition are mainly concentrated in the region with relatively small absolute values, and the absolute values of the signal wavelet coefficients are relatively large, the hard threshold method is more effective. This method can simply and directly remove the wavelet coefficients corresponding to the noise and better retain the details such as the edges of the signal. For example, when processing the position tracking data at the construction site, if the noise is mainly some small-amplitude high-frequency interferences, the hard threshold method can effectively remove these noises while retaining the mutation information of the position data (such as the rapid movement of personnel or equipment).

[0065] The soft threshold method sets the wavelet coefficients with absolute values less than the threshold λ to 0. For the wavelet coefficients with absolute values greater than the threshold λ, a shrinking process is performed, and the shrunk wavelet coefficients are;

[0066]

[0067] where sgn(·) is the sign function. The soft threshold method can make the processed signal smoother while removing the noise. For Daubechies wavelet coefficients, when there are some weak effective information in the signal that is relatively close to the noise wavelet coefficients in amplitude, the soft threshold method can avoid introducing too many mutations during the noise removal process. For example, when processing the light intensity data at the construction site, the change of the light intensity may be relatively gentle, and the soft threshold method can better balance the requirements of noise removal and retaining the smoothness of the signal.

[0068] SureShrink determines the threshold based on Stein's Unbiased Risk Estimate (SURE). For a given wavelet coefficient vector W = [W f (a 1，1 , b 1，1 ), W f (a 1，2,b 1，2 ),…,W f (a 1，n ,b 1，n )], SureShrink selects the threshold λ by minimizing the SURE estimator, and the calculation formula of the SURE estimator is:

[0069]

[0070] where σ 2 is the estimated value of the noise variance, is the indicator function, which is 1 when |W i | ≤ λ and 0 otherwise. The threshold λ is determined by minimizing the SURE estimator, and then the wavelet coefficients are processed using the soft threshold or hard threshold method (usually the soft threshold). The advantage of the SureShrink threshold processing method is that it can adaptively select an appropriate threshold according to the characteristics of the data itself. For Daubechies wavelet coefficients, when the noise level is uncertain or the wavelet coefficient distributions of the signal and noise are relatively complex, SureShrink can provide a more accurate threshold selection, thereby effectively removing noise and retaining the main features of the signal. For example, in the case of processing the mixed data of multiple sensors at the construction site (such as temperature, humidity, harmful gas concentration, etc.), SureShrink can adaptively denoise according to the noise characteristics of different data.

[0071] The wavelet coefficients after threshold processing are then calculated through the inverse wavelet transform formula to obtain the denoised data sequence f denoised (t).

[0072] Further preferably, the data processing sub-module also uses a median filtering algorithm based on a sliding window to smooth the discrete data sequence:

[0073] The data sequence f denoised (t) after wavelet denoising is regarded as a new discrete signal sequence x(n), and the sliding window size N is set;

[0074] Starting from the starting position of the data sequence, for each window position n, the signal values within the window are obtained

[0075] The signal values within the window are sorted, and the median value is taken as the output after median filtering;

[0076] The window is continuously slid over the data sequence and the median value at each position is calculated, finally obtaining a data sequence y(n) after median filtering. Compared with the original signal sequence, this sequence removes some isolated outliers, making the signal smoother. For example, sudden jumps may occur in the original position tracking data due to temporary interference from sensors. After median filtering, these jumps will be smoothed out, obtaining a signal that more conforms to the actual position change trend, providing more accurate personnel and equipment position information for safety warning modules and the like.

[0077] Preferably, the data management module further includes a data storage sub-module. The data storage sub-module includes a relational database and a non-relational database. The relational database is a MySQL database, and the non-relational database is a MongoDB database. Among them, the MySQL database can store construction site environment data, construction progress, and personnel arrangement data, such as a personnel information table (storing the names, ages, job types, contact information, etc. of construction workers), an equipment parameter table (recording the model, specifications, manufacturers, maintenance records, etc. of construction equipment), a safety specification table (storing various building construction safety standard articles and corresponding explanations), etc.; the MongoDB database can store unstructured data such as BIM model files, construction monitoring videos, and training document materials.

[0078] Preferably, the BIM model processing module includes a BIM model import and conversion sub-module and a model real-time update sub-module. The BIM model import and conversion sub-module is used to retrieve the preprocessed BIM model data from the data storage sub-module, perform lightweight processing, and construct a BIM building information model. The model real-time update sub-module is used to perform real-time updates on the BIM building information model according to the preprocessed construction site environment data, construction progress, and personnel arrangement data.

[0079] For common BIM model formats (such as IFC, RVT, etc.), corresponding format parsers are preset in the BIM model import and conversion sub-module. The parser can identify various elements in the model file, including beams, slabs, columns of the building structure, pipes, cables, distribution boxes of mechanical and electrical equipment, etc., and extract their geometric information, physical properties, and topological relationships. During the parsing process, the object-oriented programming idea is adopted to abstract different types of elements into corresponding classes, facilitating subsequent processing and operations.

[0080] Preferably, the BIM model import and conversion sub-module processes the BIM model data using a lightweight algorithm based on triangle mesh simplification:

[0081] Step 1: Extract the set of triangular patches that need to be lightweight processed from the BIM model data. For each triangular patch t BIM, obtain the coordinates of its three vertices, denoted as v1 = (x1, y1, z1), v2 = (x2, y2, z2), and v3 = (x3, y3, z3) respectively;

[0082] Step 2: Substitute the vertex coordinates v1, v2, and v3 into the plane equation ax + by + cz + d = 0, and calculate the coefficients a, b, c, and d;

[0083] Step 3: For the point p = (x, y, z), according to the formula calculate the squared distances from the three vertices v1, v2, and v3 to this plane respectively;

[0084] Step 4: According to the formula Q(t BIM ) = d 2 (v1, t BIM ) + d 2 (v2, t BIM ) + d 2 (v3, t BIM ), calculate the error metric Q(t BIM ) of each triangular patch t BIM );

[0085] Step 5: Traverse all mergeable edges and calculate the increase in error after merging:

[0086] For each triangular patch t BIM , traverse all its edges (a total of three edges). Assume the currently considered edge is the edge connecting vertices v i and v j ;

[0087] When merging this edge, it will change the shape and topological structure of the adjacent triangular patches. Recalculate the error metric Q′(t BIM ) of each patch in the new set of triangular patches;

[0088] Calculate the increase in error of the entire model after merging this edge where t′ BIM represents the new set of triangular patches after merging the edge;

[0089] Step 6: Select the edge with the smallest increase in error for merging:

[0090] For all mergeable edges of all triangular patches, repeat Step 5, calculate the increase in error after merging each edge, find the edge with the smallest increase in error, and merge it. This will cause two adjacent triangular patches to merge into a new triangular patch, thus reducing the total number of triangular patches;

[0091] Step 7: Repeat Step 1 to Step 6, continuously perform edge merging operations on the triangular facets in the model until the lightweight processing termination condition is met, end the lightweight processing process, and obtain the lightweight BIM model.

[0092] In a preferred scenario, the lightweight processing termination condition is met when the number of merged triangular facets reaches a preset number of triangular facets; in another preferred scenario, the lightweight processing termination condition is met when the error increase amount of the model after edge merging reaches a set threshold.

[0093] The model real-time update sub-module establishes an interface between the BIM model and the real-time data of the construction site. Through this interface, it can receive in real time the construction site environment data, personnel and equipment location data, and construction progress data from the data storage sub-module, etc. In the interface design, a message queue mechanism (such as Kafka) is adopted to ensure reliable data transmission and asynchronous processing.

[0094] Preferably, the VR training scenario generation module includes a virtual scenario construction sub-module and an interaction element addition sub-module;

[0095] The virtual scenario construction sub-module is used to create all the scenario elements of the virtual construction scenario based on the BIM building information model and optimize the layout of the virtual scenario;

[0096] Specifically, the virtual scene construction sub-module uses 3D modeling tools (such as 3ds Max, Maya, etc.) to create elements such as building main bodies, construction sites, and construction equipment in the virtual construction scene based on BIM model data. At the same time, realistic material textures and lighting effects can be added to the elements in the virtual scene. Texture editing software (such as Substance Painter) is used to produce textures of various materials, such as surface textures of concrete, steel, wood, etc. By adjusting parameters of the materials, such as roughness, reflectivity, etc., the physical properties of different materials are simulated. In terms of lighting effects, global illumination (GI) and physically based rendering (PBR) methods are used to simulate a real lighting environment, making the scene more vivid and realistic; the virtual scene construction sub-module deeply analyzes the construction process and reasonably arranges various elements in the virtual scene according to the characteristics and requirements of different construction stages. For example, during the foundation construction stage, construction machinery and material stacking areas are set near the foundation construction location; during the main body construction stage, the positions of vertical transportation equipment and the directions of construction channels are planned. At the same time, signs such as safety channels and dangerous areas are set to ensure that the movement paths of construction personnel in the virtual scene comply with safety regulations; at the same time, a navigation and guidance system is designed for the virtual scene to help construction personnel quickly find the required locations and equipment. Arrow indications, voice prompts, etc. are used to guide construction personnel to complete various training tasks. In complex construction areas, a map navigation function is set up, and construction personnel can view their own positions and the surrounding environment through the map, which is convenient for planning movement routes.

[0097] The interactive element adding sub-module is used to add interactive components to construction equipment in a virtual scenario. Interactive components, such as operating handles and consoles, are added to the construction equipment. Using virtual reality interaction development frameworks (such as SteamVR and OpenVR), the interaction logic of these interactive components is designed. For example, for the operating handle of a crane, by tracking the position and rotation angle of the handle, operating actions such as lifting, slewing, and luffing of the crane are simulated. During the operation, a force feedback effect is added so that construction workers can feel the resistance and reaction force when operating the equipment. An operation feedback mechanism can also be designed. When construction workers operate the equipment, the system can immediately feedback the operation result and provide feedback in multiple ways such as vision, hearing, and touch. For example, when the operation is correct, the equipment emits a normal operation sound, and the correct parameters are displayed on the equipment dashboard. When the operation is incorrect, the equipment emits an alarm sound, the corresponding components in the model flash in red, and the error cause and correction method are displayed. The interactive element adding sub-module can also integrate gesture recognition devices such as Leap Motion or gesture recognition algorithms based on computer vision to enable construction workers to interact with objects in the virtual scenario through gestures. Common gestures (such as grasping, waving, and rotating) are recognized and classified, and the gesture actions are mapped to corresponding interactive operations. For example, the grasping gesture can be used to pick up tools in the virtual scenario, and the waving gesture can be used to switch perspectives or select different operation options. The interactive element adding sub-module can also adopt a speech recognition engine and a speech synthesis engine to achieve voice interaction between construction workers and the virtual scenario. Construction workers can control objects in the virtual scenario through voice commands, such as "Start the crane" and "Turn on the lighting equipment". The module can accurately recognize the voice commands and convert them into corresponding operation signals for execution in the virtual scenario. At the same time, the system can feedback the operation result and prompt information to construction workers through speech synthesis technology.

[0098] Preferably, the safety warning module includes a warning rule setting sub-module and a real-time monitoring and warning sub-module. The warning rule setting sub-module is used to set warning rules, and the real-time monitoring and warning sub-module is used to monitor multi-source data based on the warning rules and send a prompt to the VR training scenario generation module when an abnormality occurs.

[0099] Further preferably, the early warning rule setting sub-module formulates a series of safety early warning indicators according to national and local construction safety regulations, industry standards and in combination with the actual construction situation of the project. For example, the distance threshold between personnel and dangerous areas is set to 5 meters, 10 meters, etc. respectively according to the levels of different dangerous areas; the safety range of equipment operation parameters is set according to the equipment model and operation manual, such as the lifting weight limit of a crane, the slewing angle limit of a tower crane, etc.; the critical values of environmental parameters are determined according to relevant safety standards and actual construction environment requirements, such as the upper temperature limit at the construction site is set to 40 °C, the upper humidity limit is set to 80%, etc.; in addition, considering the dynamic changes during the construction process, a dynamic adjustment mechanism for safety early warning indicators is established, and the early warning indicators are adjusted in real time according to factors such as construction progress, environmental changes, and equipment status. For example, in hot weather, the temperature early warning threshold at the construction site is appropriately lowered; after the equipment is aged or maintained, the safety range of equipment operation parameters is adjusted; the early warning rule setting sub-module can also provide a visual early warning rule editing interface for managers. Through this interface, managers can easily create, modify and delete early warning rules. The interface displays the logical structure of the rules in a graphical way, such as through elements such as condition judgment boxes and data input boxes, enabling managers to intuitively set early warning conditions and trigger actions.

[0100] The real-time monitoring and early warning sub-module can use a stream processing framework (such as Apache Flink) to perform real-time analysis on the real-time data collected at the construction site. The data is accessed from the data management module to the Flink cluster in real time. By writing data processing logic, the data is filtered, transformed, and calculated in real time. For example, the distance between personnel and dangerous areas and the real-time values of equipment operation parameters are calculated in real time, and compared with the preset early warning rules. Based on machine learning-based anomaly detection algorithms (such as Isolation Forest, One-Class SVM, etc.), in-depth analysis of the real-time data is carried out to discover potential safety hazards. These algorithms can automatically learn the patterns of normal data. When the data shows abnormal deviation, an early warning signal is sent in time. For example, by analyzing the historical data of equipment operation parameters, a model of the normal operation mode is established. When the real-time data deviates greatly from the model prediction results, it is judged that the equipment may have a failure risk and an early warning is triggered; on the VR device, early warning information is sent to construction workers through vibration feedback, sound prompts, and visual warnings. When the system detects a safety hazard, the handle or head-mounted device of the VR device will vibrate, and at the same time, a preset alarm sound will be played. In the virtual scene, the early warning information is displayed in a prominent red flashing or pop-up window form, indicating the type and location of the danger to the construction workers; in addition, the real-time monitoring and early warning sub-module pushes the early warning information to the terminal devices of the management personnel through text messages, instant messaging software, etc. At the same time, the early warning location and type are displayed in the BIM model. The management personnel can log in to the system through a PC or mobile device to view the early warning information in the BIM model and conduct further analysis and processing. In the BIM model display, different types of early warning information are distinguished through different colors, icons, and animation effects, which is convenient for the management personnel to quickly identify and locate.

[0101] Preferably, when the construction workers perform VR virtual training operations, the training effect evaluation module uses the high-precision sensors of the VR device to comprehensively record their operation behaviors on the construction equipment; the training effect evaluation model is:

[0102]

[0103] Among them, S is the comprehensive evaluation score, T avg is the average time for the user to complete a specific training task, M step is the matching degree of operation steps, P d is the attention ratio of key elements, f is the perspective change frequency, f max is the maximum value of the perspective change frequency of all users, E is the total number of errors, R avg is the average error correction time, R max is the maximum value of the average error correction time of all users, and α, β, and γ are weight coefficients.

[0104] In the above, for the comprehensive evaluation score S, its value range can be between 0 and 1. The higher the score, the better the training effect. Let the total number of times a user completes a specific training task be N p , and the time taken to complete the task each time is respectively Then The shorter the time taken to complete the task, the greater the positive contribution to the comprehensive evaluation score. Assume that in the training task, the standard operation step sequence is S = [s1, s2, …, s M , and the user's actual operation step sequence is Calculate the length L of the longest common subsequence of S and S′ through the dynamic programming algorithm. Then The closer this value is to 1, the better the coherence between the user's operation steps and the standard steps, and the more significant the improvement in the comprehensive evaluation score. In the VR training scenario, assume there are a total of K key scenario elements, and the attention duration of the user for the kth element is d k , and the total training duration is D. Then This value reflects the user's attention to the key scenario elements. The higher P d , the greater the positive impact on the comprehensive evaluation score. Record the number of times the user's perspective changes during the training as F, and the training duration as D. Then A lower f value usually indicates that the user's attention is more concentrated. To convert it into a positive impact on the comprehensive evaluation score, use to calculate. Assume there are Q types of possible error types during the training, and the number of times the user makes the qth type of error in each training is e q , then the total number of errors The fewer the number of errors, the more beneficial it is to improve the comprehensive evaluation score. Therefore, use to reflect it, avoiding the situation of a denominator of 0 when E = 0. When the user makes an error, record the time intervals from the occurrence of the error to the user's corrective operation as r1, r2, …, r E , then (When E = 0, R avg = 0). To convert it into a positive impact on the comprehensive evaluation score, use to calculate.

[0105] Of course, the above training effect evaluation model is only a preferred implementation method. Those skilled in the art can flexibly set or adjust the specific training effect evaluation model according to the actual situation in practical applications.

[0106] Preferably, the construction safety warning training system based on the BIM building information model of the present invention further includes a training course management module, which is used to push training courses to users and record the learning progress. Specifically, the training course management module can be built with a visual editor for training course design. The editor provides a rich component library, including text boxes, pictures, videos, 3D models, interactive elements, etc. Trainers can add these components to the course page by dragging and dropping, and perform free layout and editing. It supports operations such as formatting text, cropping and editing pictures, video editing and adding subtitles, which is convenient for trainers to create high-quality training content. In addition, the training course management module can also be integrated with the resource library in the system. A large number of safety knowledge documents, construction case videos, 3D models and other materials are stored in the resource library. When creating courses, trainers can directly call these materials from the resource library to enrich the course content. At the same time, it supports trainers to upload custom resources to expand the content of the resource library. Trainers can divide the courses into different chapters and modules according to the training objectives and content. Each chapter and module has a clear theme and learning objective, which is convenient for construction workers to learn systematically. For example, the safety training course is divided into chapters such as basic safety knowledge, construction equipment operation safety, and on-site construction emergency handling. Each chapter is further divided into several modules, and a reasonable learning path is designed for each course. According to the difficulty level and logical relationship of the knowledge points, the learning order is set. In the learning path, exercises, test questions, simulation drills and other links can be interspersed to timely test the learning effect of construction workers. At the same time, it supports trainers to customize personalized learning paths according to the needs of different construction positions to improve the pertinence of training.For those who have completed part of the training, the system analyzes their weak points in knowledge mastery based on their past learning records and test scores, and pushes targeted intensive training courses or extended learning content; through continuous learning and analysis of user behavior data, the push algorithm is continuously optimized to ensure that the pushed courses highly match the actual needs of construction workers, improving the utilization efficiency of training resources; with the help of the tracking function of VR devices and the monitoring program in the system background, behavior data of construction workers during the learning process is collected comprehensively; in addition to recording regular learning time, progress, pause and review times, information such as the operation trajectory of construction workers in the virtual scene, the interaction duration and sequence with various elements are also recorded in detail; for example, recording the execution sequence and time spent on each operation step of construction workers in the simulated equipment operation training to accurately understand their learning process and the way of mastering knowledge and skills; the system tracks the learning progress of construction workers in real time, and automatically sends reminders to those with lagging progress according to the preset learning plan and time nodes in the course; the reminder methods include pop-up prompts and vibration reminders in the VR device, as well as mobile phone text messages and in-site messages; at the same time, a visual learning progress report is provided for training managers at the management end, clearly showing the learning progress of each construction worker in the form of charts, facilitating managers to timely understand the overall training situation and urge and guide those with abnormal progress; in-depth statistical analysis is carried out on data such as the test answering situation and simulation exercise performance of construction workers during the course learning process; calculate the correct answering rate of different knowledge points, and find out the common knowledge loopholes of construction workers through data comparison, such as weak performance in the part of safety regulations knowledge or specific construction skill operation links; in addition, compare the learning effects of construction workers in different batches and different positions, evaluate the effectiveness and pertinence of training courses for different groups, and provide strong data support for the optimization of subsequent course content and the adjustment of training strategies.

[0107] It should be noted that the term "comprising", "including" or any other variant thereof is intended to cover non-exclusive inclusion, so that a process, method, commodity or device comprising a series of elements not only includes those elements but also includes other elements not expressly listed, or also includes elements inherent to such process, method, commodity or device. Without further limitation, an element defined by the statement "comprising an..." does not exclude the presence of additional identical elements in the process, method, commodity or device comprising the said element.

[0108] Each embodiment in the present invention is described in a progressive manner. The same or similar parts among the embodiments can be referred to each other, and the key point of each embodiment is to illustrate the differences from other embodiments. In particular, for the system embodiment, since it is basically similar to the method embodiment, the description is relatively simple, and the relevant parts can refer to the partial description of the method embodiment.

[0109] Those of ordinary skill in the art should understand that the discussion of any of the above embodiments is only exemplary and is not intended to imply that the scope of the present disclosure is limited to these examples; within the concept of the present disclosure, the technical features in the above embodiments or different embodiments can also be combined, the steps can be implemented in any order, and there are many other variations of different aspects of one or more embodiments of the present invention as described above, which are not provided in detail for the sake of brevity.

[0110] Although the present disclosure has been described in connection with specific embodiments thereof, many alternatives, modifications, and variations of these embodiments will be apparent to those of ordinary skill in the art in light of the foregoing description.

[0111] One or more embodiments of the present invention are intended to cover all such alternatives, modifications, and variations that fall within the scope of protection of the present invention. Therefore, any omissions, modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of one or more embodiments of the present invention shall be included within the scope of protection of the present disclosure.

Claims

1. A construction safety early warning training system based on the BIM building information model, characterized in that, The system includes: A data management module, which is used to collect multi-source data and preprocess and store the multi-source data. Among them, the multi-source data includes construction site environment data, BIM model original data, and construction progress and personnel arrangement data; A BIM model processing module, which is used to retrieve the preprocessed BIM model data from the data management module, perform lightweight processing, and construct a BIM building information model, and update the BIM building information model in real time according to the preprocessed construction site environment data, construction progress, and personnel arrangement data; A VR training scenario generation module, which is used to create a virtual construction scenario based on the BIM building information model and interact with users; A safety warning module, which is used to set warning rules, monitor the multi-source data based on the warning rules, and send a prompt to the VR training scenario generation module when an abnormality occurs; A training effect evaluation module, which is used to collect the interaction behavior data between the user and the VR training scenario generation module during the user's training and evaluate the training based on a training effect evaluation model.

2. The construction safety early warning training system based on the BIM building information model according to claim 1, wherein The BIM model processing module processes the BIM model data using a lightweight algorithm based on triangle mesh simplification: Extract the set of triangular patches that need to be lightweight processed from the BIM model data. For each triangular patch t BIM , obtain the coordinates of its three vertices, denoted as v1 = (x1, y1, z1), v2 = (x2, y2, z2), and v3 = (x3, y3, z3) respectively; Substitute the vertex coordinates v1, v2, and v3 into the plane equation ax + by + cz + d = 0 to calculate the coefficients a, b, c, and d; For a point p = (x, y, z), according to the formula calculate the squared distances from the three vertices v1, v2, and v3 to this plane respectively; According to the formula Q(t BIM ) = d 2 (v1, t BIM ) + d 2 (v2, t BIM ) + d 2 (v3, t BIM ), calculate the error metric Q(t BIM ) for each triangular facet t BIM ); Traverse all mergable edges and calculate the increased error after merging; Select the edge with the smallest increased error for merging; Repeat the merging process until the lightweight processing termination condition is met, end the lightweight processing process, and obtain the lightweight BIM model.

3. The construction safety warning training system based on the BIM building information model according to claim 2, characterized in that, The condition for meeting the lightweight processing termination condition is that the number of merged triangular patches reaches a preset number of triangular patches; or, The condition for meeting the lightweight processing termination condition is that the increased error of the model after merging edges reaches a set threshold.

4. The construction safety warning training system based on the BIM building information model according to claim 1, characterized in that, The construction site environment data at least includes environmental temperature data, environmental humidity data, light intensity data, dust data, noise data, wind speed and direction data, harmful gas data, and smoke data.

5. The construction safety early warning training system based on BIM building information model according to claim 4, characterized in that, The data management module includes a data processing sub-module; the data processing sub-module uses a denoising algorithm based on wavelet transform to remove the noise components of the construction site environment data: For each data sequence f(t) in the construction site environment data, select the Daubechies wavelet basis function ψ(t); Set the value ranges of the scale parameter a1 and the translation parameter b1; According to the formula Within the range of the determined scale parameter a1 and translation parameter b1, calculate the wavelet coefficients W f (a1, b1); Perform threshold processing on the calculated wavelet coefficients W f (a1, b1), where the threshold processing method is hard threshold processing, soft threshold processing or SureShrink threshold processing; The wavelet coefficients after threshold processing are then used to calculate the denoised data sequence f through the inverse wavelet transform formula denoised (t).

6. The construction safety warning training system based on the BIM building information model according to claim 5, wherein, The data processing sub-module also uses a median filtering algorithm based on a sliding window to smooth the discrete data sequence: Take the data sequence f denoised (t) after wavelet denoising as the new discrete signal sequence x(n), and set the sliding window size N; Starting from the starting position of the data sequence, for each window position n, obtain the signal values within the window Sort the signal values within the window and take the middle value as the output after median filtering; Continuously slide the window on the data sequence and calculate the median at each position, and finally obtain a data sequence y(n) after median filtering.

7. The construction safety early warning training system based on BIM building information model according to claim 5, characterized in that, The data management module also includes a data storage sub-module. The data storage sub-module includes a relational database and a non-relational database. The relational database is a MySQL database, and the non-relational database is a MongoDB database.

8. The construction safety early warning training system based on the BIM building information model according to claim 1, characterized in that, The VR training scenario generation module includes a virtual scenario construction sub-module and an interactive element addition sub-module; The virtual scenario construction sub-module is used to create all scenario elements of the virtual construction scenario based on the BIM building information model and optimize the layout of the virtual scenario; The interactive element addition sub-module is used to add interactive components to the construction equipment in the virtual scenario.

9. The construction safety warning training system based on the BIM building information model according to claim 1, characterized in that, The training effect evaluation model is as follows: Among them, S is the comprehensive evaluation score, T avg is the average time for users to complete specific training tasks, M step is the matching degree of operation steps, P d is the attention ratio of key elements, f is the perspective change frequency, f max is the maximum value of the perspective change frequency of all users, E is the total number of errors, R avg is the average error correction time, R max is the maximum value of the average error correction time of all users, and α, β, and γ are weight coefficients.

10. The construction safety early warning training system based on the BIM building information model according to any one of claims 1-9, characterized in that, The system further includes a training course management module, which is used to push training courses to users and record the learning progress.

Images