CN113779125A

CN113779125A - Construction safety information management method and system

Info

Publication number: CN113779125A
Application number: CN202110945776.0A
Authority: CN
Inventors: 钟波涛; 潘杏; 丁烈云; 盛达; 王宇航
Original assignee: Huazhong University of Science and Technology
Current assignee: Huazhong University of Science and Technology
Priority date: 2021-08-17
Filing date: 2021-08-17
Publication date: 2021-12-10

Abstract

The invention discloses a construction safety information management method and system, and belongs to the field of construction safety information management. The method comprises the following steps: the system comprises an information acquisition unit, an information processing unit, an information uploading unit and a user management unit. Collecting original construction safety information by means of two methods of artificial observation and intelligent technology; by means of a deep learning technology, construction safety key information, such as text key words, picture key information and the like, is automatically extracted; carrying out hash processing on the original information, and storing the hash value of the original information and the key information to a block chain network; the front end of the local system reads the hash value and the key information of the original information on the chain, and makes statistics and visualization, and the related statistical information helps a manager to quickly identify the key problems of safety management, and timely takes countermeasures to prevent risks. The invention not only realizes the automatic extraction and chain storage of the construction safety key information, but also improves the information retrieval efficiency and the safety performance of the system.

Description

Construction safety information management method and system

Technical Field

The invention belongs to the technical field of construction safety information management, and particularly relates to a construction safety information management method and system.

Background

In order to relieve the pressure of housing, the construction industry of China shows a rapid growth trend. The construction informatization construction is an important idea for guaranteeing intelligent construction safety construction, wherein safety information acquisition and management are main contents of the informatization construction and are also the core of a safety information management system. Because the current construction safety information is stored dispersedly and lacks the authentication of a third party authority, the authenticity of data is difficult to guarantee, the risk of tampering exists, and the management of the real-time information of the construction site before an accident occurs becomes a difficult task. This problem has been persistent in the construction industry for decades, and has severely hampered the improvement of the level of safety management in construction. In order to increase the trust of the parties, it is necessary to form a system with traceability and transparency, which depends not only on the integrity of data collection but also on the unforgeability and transparency of data transmission and storage. However, in the construction industry, the device information cannot be easily captured by each stakeholder, and is difficult to be transmitted and stored safely, and once the liability is concerned, the liability party has a motivation to tamper with the data to avoid the liability.

With the development of the internet technology, the security information management system has become a powerful tool for assisting the development of informatization construction. However, the conventional security information management system still has some defects and shortcomings, and particularly, the problems of dispersed storage, poor reliability, poor traceability and the like of the construction security information at present cannot be effectively solved, so that a unified, safe and transparent system managed by the stakeholders together is urgently needed; obviously, conventional centralized information management systems are not adaptable solutions.

Disclosure of Invention

The integration of "deep learning + blockchain technique" provides a new solution to the above problem. Deep learning exploits and analyzes information automatically, quickly, and efficiently by its mechanism of mimicking the human brain to interpret data instead of manually acquiring features. The block chain constructs a credible technical system by means of a series of complex mathematical algorithms such as a Hash encryption algorithm, a timestamp, distributed consensus, intelligent contract and the like, guarantees the information to be public and transparent, and can not be traced back in the whole process and tampered, and establishes a 'trust' network between machines. The integration of the deep learning technology and the block chain technology not only can realize the automatic extraction and chaining of the construction safety information, but also can improve the information retrieval efficiency and the safety performance of the system.

Aiming at the defects of the related technology, the invention aims to provide a construction safety information management method and a construction safety information management system, aiming at managing construction safety data based on deep learning and block chain technology, ensuring the reliability, traceability and transparency of construction safety information and improving the retrieval performance and practical performance of a management system.

In order to achieve the above object, an aspect of the present invention provides a construction safety information management method, including the steps of:

s1, collecting original information of construction safety in a construction process;

s2, carrying out format classification on the original information, carrying out corresponding processing on the original information according to different formats by utilizing a deep learning algorithm so as to automatically extract construction safety key information, and converting the original information of all formats into hash values;

s3, broadcasting the hash value of the original information and the key information subjected to deep learning processing to each node in the block chain network, uploading all the original information to a cloud server, and finishing sequencing and counting;

and S4, in the construction process, calling a block chain network interface and inquiring and analyzing construction safety management information from the block chain network.

Further, the step S2 specifically includes:

carrying out data format classification on the original information, wherein the classification types comprise text data, picture data and video data, dividing the text data into training text samples and testing text samples, and dividing the picture data into training picture samples and testing picture samples;

constructing a text keyword extraction model based on a KEYBERT algorithm, and identifying keywords in the construction safety text data;

constructing a construction scene image description model of an encoding-decoding framework based on an ATTENTION algorithm, and identifying character descriptions in construction safety picture data;

training the text keyword extraction model based on the KEYBERT algorithm and the construction scene image description model of the coding-decoding framework based on the ATTENTION algorithm by respectively adopting training text samples and training picture samples;

respectively inputting a test text sample and a test picture sample into the text keyword extraction model based on the KEYBERT algorithm and the construction scene image description model of the coding-decoding framework based on the ATTENTION algorithm, analyzing an error value between an output test result and an actual result, and if the error value is within a preset range, sending a result obtained through deep learning processing to a block chain network; otherwise, the sample is collected again for training until the error value is within the preset range.

Further, the constructing of the text keyword extraction model based on the KEYBERT algorithm comprises the following steps:

embedding the document description based on a BERT model to obtain a document vector;

extracting keywords and expressions from the same document by utilizing a bag-of-words technology, and embedding the keywords based on a BERT model to obtain a keyword vector;

extracting the most similar keywords according to the highest cosine similarity score based on the cosine similarity between the document vector and the keyword vector;

and automatically adjusting parameters of the text keyword extraction model based on the KEYBERT algorithm periodically according to actual construction safety information.

Further, constructing a construction scene image description model of an ATTENTION algorithm-based coding-decoding framework comprises the following steps:

inputting an image to be detected, and performing feature extraction on the input image through an encoder;

capturing and processing image semantic information and word sequence characteristics through a decoder;

outputting statement description about image scene information and global information;

and the parameters of the construction scene image description model of the coding-decoding framework based on the ATTENTION algorithm are regularly and automatically adjusted according to the actual construction safety information.

Further, the step S3 specifically includes:

wirelessly uploading the construction safety information data to a cloud server;

calling the deep learning algorithm in the step S2 to further process and store the data in the cloud, sending the construction safety key information and the original construction safety information hash value after the deep learning processing to the block chain network, and uploading all the original information to the cloud server;

data in the block chain network is put into a block according to a certain structure to generate legal block header information, and then the generated block is verified by members to generate a new block;

the intelligent contract executes the transaction command, and the transaction information is broadcast to each node of the blockchain network.

Another aspect of the present invention also provides a construction safety information management system, including:

the information acquisition unit is used for acquiring the original information of construction safety in the construction process;

the information processing unit is used for carrying out format classification on the original information, carrying out corresponding processing on the original information according to different formats by utilizing a deep learning algorithm so as to automatically extract construction safety key information, and converting the original information of all formats into hash values;

the information uploading unit is used for broadcasting the hash value of the original information and the key information subjected to deep learning processing to each node in the block chain network, uploading all the original information to the cloud server, and finishing sequencing and counting;

and the user management unit is used for calling the block chain network interface by a user in the construction process and inquiring and analyzing construction safety management information from the block chain network.

Further, the information processing unit specifically includes:

the classification module is used for classifying the data format of the original information, the classification type comprises text data, picture data and video data, the text data is divided into training text samples and testing text samples, and the picture data is divided into training picture samples and testing picture samples;

the first construction module is used for constructing a text keyword extraction model based on a KEYBERT algorithm and identifying keywords in the construction safety text data;

the second construction module is used for constructing a construction scene image description model of an ATTENTION algorithm-based coding-decoding framework and is used for identifying character description in construction safety picture data;

the training module is used for training the text keyword extraction model based on the KEYBERT algorithm and the construction scene image description model of the coding-decoding framework based on the ATTENTION algorithm by respectively adopting a training text sample and a training picture sample;

the testing module is used for respectively inputting a testing text sample and a testing picture sample into the text keyword extraction model based on the KEYBERT algorithm and the construction scene image description model of the coding-decoding framework based on the ATTENTION algorithm, analyzing an error value of an output testing result and an actual result, and if the error value is within a preset range, sending a result obtained through deep learning processing to the block chain network; otherwise, the sample is collected again for training until the error value is within the preset range.

Through the technical scheme, compared with the prior art, the invention has the following beneficial effects:

(1) the invention discloses a construction safety information management method and system. The method designs a novel information storage scheme, construction safety key information and original construction safety information hash values after deep learning processing are stored in a block chain network, and all construction safety original information is stored in a cloud server. According to the scheme, the construction safety key information can be automatically extracted based on models such as keyword extraction and construction scene image description of deep learning, the problem of high-degree redundant storage of information in a block chain system can be effectively solved, and therefore the information retrieval efficiency of the block chain system is improved. Meanwhile, the block chain technology can maintain data sharing/service exchange of users in an open and cooperative environment, and a large amount of data accumulated in the block chain can promote the learning capability and the generalization capability of a deep learning model. Comprehensively, the integration of the deep learning technology and the block chain technology not only realizes the automatic extraction and chaining of the construction safety information, but also improves the information retrieval efficiency and the safety performance of the system, and finally, by means of visual and timely construction safety key information, a manager can quickly know the site construction condition so as to make a credible and reliable data auxiliary decision;

(2) compared with the traditional shallow machine learning method, the deep learning does not need complex artificially defined vocabulary, syntax and semantic feature engineering, and can automatically extract and identify the relevant features of the text. Based on models such as keyword extraction and picture semantic recognition of deep learning, information characteristics of the massive construction safety information are automatically extracted and recognized, and the construction safety information is structurally expressed visually and vividly, so that the construction safety information monitoring efficiency and safety management decision are improved, and meanwhile, the intelligent and specialized retrieval service of massive construction safety information resources is promoted;

(3) blockchains are considered to be subversive technologies that open up "new era of trust". Compared with the traditional information technology, the decentralized distributed structure in the block chain can save a great deal of intermediary mechanism cost consumption; the time stamp which can not be tampered can solve the problems of data tracking and information anti-counterfeiting; the block chain is beneficial to realizing the control right of the user to the data so as to well solve the problems of key data protection and authorized access; and flexible programmable features help to standardize existing market order. By uploading the construction safety information to the block chain network, the construction safety information can be transparently monitored and tracked, so that the problems of dispersed storage, poor reliability, poor traceability and the like of the construction safety information are solved, and finally, a safe and stable information interaction environment for participants is provided.

Drawings

FIG. 1 is a functional architecture diagram of a system in an embodiment of the invention;

FIG. 2 is a diagram of the logical architecture of a system according to an embodiment of the present invention;

FIG. 3 is an information collection architecture diagram according to an embodiment of the present invention;

FIG. 4 is a diagram of an information processing architecture according to an embodiment of the present invention;

FIG. 5 is a KEYBERT-based deep learning algorithm architecture diagram according to an embodiment of the present invention;

FIG. 6 is an architecture diagram of ATTENTION-based deep learning algorithm according to an embodiment of the present invention;

FIG. 7 is a diagram of information uplink architecture according to an embodiment of the present invention;

fig. 8 is a flowchart of an uplink scheme according to an embodiment of the present invention.

Detailed Description

In order to make the objects, technical solutions and advantages of the present invention more apparent, the present invention is described in further detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention. In addition, the technical features involved in the embodiments of the present invention described below may be combined with each other as long as they do not conflict with each other.

The method mainly relies on deep learning and block chain technology. The deep learning explains data instead of manually acquiring features by means of a mechanism simulating human brain, so that construction safety key information is automatically, quickly and efficiently mined, and manpower and material resources are greatly reduced in the process. The block chain technology constructs a credible technical system by means of a series of complex mathematical algorithms such as a Hash encryption algorithm, a timestamp, distributed consensus, intelligent contract and the like, guarantees the information to be transparent, traceable in the whole process and not to be falsified, and establishes a 'trust' network between machines. With construction safety critical information stored on the blockchain network, management personnel will be better able to make faster and more informed decisions on how to ensure the safety of field employees.

Referring to fig. 1 and 2, a method and system for managing construction safety information according to a preferred embodiment of the present invention includes an information acquisition unit, an information processing unit, an information uploading unit, and a user management unit.

A. Information acquisition unit

A1 and an information acquisition unit mainly acquire relevant safety-related information in the construction process, as shown in FIG. 3, on-site safety information is collected, and tower crane safety information is monitored as an example, and information such as behaviors of people relevant to the tower crane, conditions of objects, interaction information between people and objects, operating conditions of large-scale equipment and temporary facilities, environmental factors and the like is collected mainly by means of manual and intelligent equipment in the on-site construction process, and the method 1 and the method 2 are specifically as follows:

mode 1: a inspector often inspects the on-duty condition of a tower crane operator and the state of the tower crane operated by the operator in the tower crane construction process, and fills the daily inspection records of the tower crane and the like;

mode 2: collecting working state information of the tower crane, such as weight, angle, wind speed, amplitude, height and other data of the tower crane, by means of sensor equipment arranged on the tower crane; and monitoring the state of the tower crane and the surrounding environment of the tower crane by means of cameras and other equipment.

A2, checking daily check records of the tower crane by a responsible person and signing;

a3, transmitting the security information to the server by means of the internet.

According to different actual engineering contents, the types and the flows of the acquired data can be increased or decreased, and specific safety information can be adjusted according to the safety information of a certain device;

B. information processing unit

The information processing unit mainly processes the acquired original information by means of a deep learning algorithm to automatically identify effective information, as shown in fig. 4, by means of algorithms such as deep learning, information such as behaviors of related persons, conditions of objects, interaction information between persons and objects, operating conditions of large-scale equipment and temporary facilities, environmental factors and the like is converted into text key information, picture and video description information, picture information base64 code, responsible person information, original information hash value and other key information by taking tower crane monitoring as an example, and the specific operation steps are as follows;

b1, according to the collected daily inspection record of the tower crane and the sensor data such as the weight, the angle, the wind speed, the amplitude, the height and the like of the tower crane monitored by the intelligent equipment and the data such as the monitoring video and the like, archiving and classifying the data in a server, wherein the classification type is mainly classified according to the data formats such as texts, pictures and videos, specifically, the daily inspection record of the tower crane and the sensor data are stored in the text form, and the monitoring video is stored in the video or picture form;

b2, constructing a text keyword extraction model based on the KEYBERT algorithm, and identifying text keywords in the construction safety information. Inputting the text records of the hidden dangers in the daily inspection records of the tower crane into a trained model, and automatically enabling the daily hidden dangers of the tower crane to be that the distance between an A # tower crane and a B # tower crane is not more than 2 meters based on a keyword extraction model of KEYBERT (keybort), the crane boom of the A # tower crane directly collides with a standard section of the B # tower crane in the construction process, and the tower crane with obstacles and interference of the tower crane in the slewing radius of multi-tower operation does not have keywords such as approved anti-collision measures, namely 'identification of the keywords of the tower crane, the distance, 2 meters, the crane boom, the collision, the standard section, the slewing radius, the obstacles, the anti-collision measures', and the like. The key word extraction model based on KEYBERT comprises the following specific operation steps: embedding the pretrained BERT model into the daily hidden danger description of the tower crane to obtain a document vector; extracting keywords and expressions from the same document based on a TffVectoror CountVectorbased bag-of-words technique; then embedding the pre-trained BERT model into the identified keywords to obtain a keyword vector; and finally, extracting the most similar keywords according to the highest cosine similarity score based on the cosine similarity between the document vector and the keyword vector.

B3, constructing a construction scene image description model of an ATTENTION algorithm-based coding-decoding architecture, and identifying semantic description of a construction safety picture; inputting picture records in daily inspection records of a tower crane into a trained model, automatically identifying semantic description of 'one worker is in the process of installing a template' in a template installation picture by a construction scene image description model of a coding-decoding framework of an ATTENTION algorithm as shown in figure 6, wherein the specific operation steps of the construction scene image description model of the coding-decoding framework based on the ATTENTION algorithm are a coding-decoding process, the coding process comprises the steps of extracting image visual features and extracting image semantic features based on the ATTENTION algorithm, the decoding process comprises the steps of selecting image features based on the ATTENTION algorithm, extracting grammatical features, coding word positions and generating description sentences; through the encoder and the decoder, the constructed image description model can learn the mapping relation between the image characteristics and the sentence semantic characteristics, and meanwhile, an attention mechanism is introduced to improve the extraction capability of key information in the construction image.

B4, repeatedly carrying out B2 and B3 on training samples, and training a text keyword extraction model based on a KEYBERT algorithm and a construction scene image description model of a coding-decoding framework based on an ATTENTION algorithm;

b5, inputting the test text sample and the test picture sample into the step B4 for training, analyzing an error value between an output test result and an actual result, if the error value is in line with expectation, ending training and outputting a trained model; otherwise, acquiring new training samples and test samples according to the step B1, and repeating the steps B2, B3 and B4 until the error value is in accordance with the expectation;

preferably, assuming that the expected error value is 10%, the error value may be adjusted up or down according to the actual demand;

and B6, converting all original construction safety information (all information of texts, pictures and videos) into hash values.

The keyword extraction model based on the KEYBERT algorithm and the decoding and coding picture semantic recognition model based on the ATTENTION algorithm are respectively models proposed aiming at construction safety information data. According to different actual engineering contents, different data types have different semantic features, and corresponding models are correspondingly adjusted along with the different data types, so that the method aims to provide a guide direction for realizing effective organization and management of industrial mass construction safety information resources and providing intelligent and specialized retrieval and service.

C. Information uploading unit

The information uploading unit performs hash processing on the original information, and stores the hash value of the original information and the key information subjected to deep learning processing to the block chain network, as shown in fig. 7, specifically as follows;

c1, constructing a chain loading scheme, taking tower crane monitoring as an example, as shown in fig. 8, the specific scheme is as follows:

firstly, uploading relevant safety information data (basic information of a tower crane and constructors thereof, daily inspection records of the tower crane, state information of the tower crane and surrounding environment information of the tower crane) to a local database based on a wireless method such as WIFI, Bluetooth and Zigbee, and backing up the relevant safety information data to a cloud service platform;

step two, calling the deep learning algorithm in the step B, further processing data in a local database, and determining to upload the tower crane key information subjected to deep learning processing, authorization signatures of related responsible persons and the hash value of original information to a block chain network by a manager;

thirdly, uploading the hash value of the original information to a block chain network after the original data in the cloud server is audited;

and fourthly, putting the data in the block chain network into a block according to a certain structure to generate legal block header information, and then generating a new block by member verification of the generated block. Specifically, data placed into the blockchain network needs to be validated through a consensus algorithm. Upon failure, feedback will be sent to the blockchain network and the down-link database.

C2, executing the trade command by the intelligent contract, checking the data, obtaining the checking result and feeding the checking result back to the relevant constructors;

c3, the transaction information is broadcast to each node of the blockchain network.

D. User management unit

The front end of the local system reads the hash value and the key information of the original information on the chain, and makes statistics and visualization, and the related statistical information helps a manager to quickly identify the key problems of safety management, and timely takes countermeasures to prevent risks.

The construction safety information management method and the construction safety information management system provided by the invention can realize automatic mining and storage of construction safety information value, and are favorable for learning and searching construction safety information by combining visual and visual structural representation.

It will be understood by those skilled in the art that the foregoing is only a preferred embodiment of the present invention, and is not intended to limit the invention, and that any modification, equivalent replacement, or improvement made within the spirit and principle of the present invention should be included in the scope of the present invention.

Claims

1. A construction safety information management method is characterized by comprising the following steps:

s3, broadcasting the hash value of the original information and the key information subjected to deep learning processing to each node in the block chain network, uploading all the original information to a cloud server, and completing transactions such as storage and the like;

2. The construction safety information management method according to claim 1, wherein the step S2 specifically includes:

3. The construction safety information management method according to claim 2, wherein constructing a text keyword extraction model based on the KEYBERT algorithm comprises:

4. The construction safety information management method according to claim 2, wherein constructing a construction scene image description model of an ATTENTION algorithm-based coding-decoding architecture comprises:

5. The construction safety information management method according to claim 1, wherein the step S3 specifically includes:

6. A construction safety information management system, comprising:

7. The construction safety information management system according to claim 6, wherein the information processing unit specifically includes: