CN-121980511-A - Hoisting safety risk identification method based on big data

CN121980511ACN 121980511 ACN121980511 ACN 121980511ACN-121980511-A

Abstract

The invention relates to a hoisting safety risk identification method based on big data, which comprises the steps of obtaining multi-source data of equipment states, environment changes and personnel behaviors from a hoisting site, preprocessing the multi-source data to obtain a unified space-time data sequence, extracting space-time characteristics of the equipment states and the environment changes by adopting a convolutional neural network according to the unified space-time data sequence, judging whether the space-time characteristics exceed a preset threshold value, obtaining a risk level, analyzing time sequence association of the personnel behaviors and the equipment states by a long-term and short-term memory network according to the risk level, obtaining evolution trend of potential hidden danger, obtaining risk probability distribution according to the evolution trend of the potential hidden danger, and analyzing the risk probability distribution to obtain a type of high risk. The invention can improve the safety and efficiency of hoisting operation and effectively reduce the accident risk.

Inventors

Xiong Zikang
FENG JIAXING
YANG HAIDONG
HE QIFAN

Assignees

中电建宁夏工程有限公司

Dates

Publication Date: 20260505
Application Date: 20260126

Claims (8)

1. The hoisting safety risk identification method based on big data is characterized by comprising the following steps of: The method comprises the steps of obtaining multi-source data of equipment states, environmental changes and personnel behaviors from a hoisting site, preprocessing the multi-source data, and obtaining a unified space-time data sequence; according to the unified space-time data sequence, adopting a convolutional neural network to extract space-time characteristics of equipment state and environmental change, judging whether the space-time characteristics exceed a preset threshold value, and acquiring a risk level; According to the risk level, analyzing time sequence association of personnel behaviors and equipment states through a long-term and short-term memory network, and acquiring an evolution trend of potential hazards; and acquiring risk probability distribution according to the evolution trend of the potential hidden danger, analyzing the risk probability distribution, and acquiring a type of high risk.
2. The hoisting security risk identification method based on big data according to claim 1, wherein preprocessing the multi-source data comprises: performing data cleaning on the multi-source data to obtain a preliminary data set; performing noise removal and inconsistent optimization on the preliminary data set by adopting a Kalman filtering algorithm to obtain an intermediate data sequence; If the time stamps in the intermediate data sequences are not matched, time correction is carried out, and a corrected data sequence with consistent time is obtained; and carrying out position calibration by adopting a space mapping method according to the corrected data sequence, and adjusting the space distribution of the hoisting site to obtain the unified space-time data sequence.
3. The hoisting security risk identification method based on big data according to claim 1, wherein the step of extracting space-time characteristics of equipment state and environmental change by using a convolutional neural network, and the step of judging whether the space-time characteristics exceed a preset threshold value, and the step of acquiring risk level comprises the steps of: Processing the input unified space-time data sequence by adopting a convolutional neural network, extracting space-time characteristics related to equipment states and environmental changes, and obtaining a characteristic set; analyzing key points related to equipment operation according to the feature set, recording the key points as abnormal state points if the key points deviate from a preset threshold value, and determining an abnormal state set; analyzing fluctuation conditions related to the environment through the feature set, and marking the fluctuation range as an environment abnormal point if the fluctuation range exceeds a preset range to obtain an environment abnormal set; Calculating association degrees of the environment abnormal set and the abnormal state set, and if the association degrees are higher than a preset standard, judging the environment abnormal set and the abnormal state set as potential risk points, and acquiring a risk point set; And carrying out grading treatment on each risk point according to the risk point set to acquire the risk grade.
4. The hoisting security risk identification method based on big data according to claim 1, wherein analyzing time sequence association of personnel behaviors and equipment states through a long-term and short-term memory network comprises: If the risk level is higher than a preset level, extracting time sequence records of personnel behaviors and equipment states, and acquiring a time sequence relation data set; And carrying out deep analysis on the time sequence relation data set by adopting a long-short-term memory network to acquire the evolution trend of the potential hidden danger.
5. The hoisting security risk identification method based on big data according to claim 1, wherein obtaining risk probability distribution according to the evolution trend of the potential hidden danger comprises: extracting key nodes of trend change according to the evolution trend of the potential hidden danger, and determining a feature set of the evolution trend; And inputting the feature set of the evolution trend into a Bayesian network model to acquire the risk probability distribution.
6. The hoisting security risk identification method based on big data according to claim 1, wherein obtaining a specific location and type of high risk according to the risk probability distribution comprises: If the risk probability distribution is higher than the warning level, acquiring a history similar scene from a history operation database; according to the similar scene of the history, comparing the risk probability distribution level in the history operation with the distribution level of the current scene, and judging the distribution difference of the potential risks; And acquiring the type of high risk according to the distribution difference.
7. The big data based lifting security risk identification method of claim 6, wherein obtaining a historical similar scenario from a historical job database comprises: acquiring reference data of a historical scene from a pre-established historical operation database source, comparing the characteristics of the current scene with those of the historical scene according to the reference data, determining a similarity value, and acquiring the historical similar scene according to the similarity value.
8. The hoisting security risk identification method based on big data according to claim 6, wherein obtaining the type of high risk according to the distribution difference comprises: if the distribution difference exceeds a preset range, obtaining difference data corresponding to the current scene; and classifying the difference data through a support vector machine algorithm to obtain the type of the high risk.

Description

Hoisting safety risk identification method based on big data Technical Field The invention relates to the technical field of safety management, in particular to a hoisting safety risk identification method based on big data. Background The hoisting operation is used as a core link in engineering construction and bears the installation and transportation tasks of heavy equipment and components, and the safety of the hoisting operation is directly related to engineering progress and personnel life and property safety. In modern engineering projects, hoisting operations often involve complex environments, equipment and personnel collaboration, and omission of any link may cause serious accidents. Therefore, the safety management level of hoisting operation is improved, and the hoisting operation becomes an urgent need for guaranteeing the engineering quality and efficiency. However, current security management approaches often look frustrating when dealing with complex job scenarios. Many methods rely on post analysis or single link monitoring too, are difficult to comprehensively capture multiple potential risks in the whole process of the operation, and lack dynamic judgment capability after integration of information from different sources. The limitation makes the potential safety hazard often not be found in time before the occurrence, and particularly, the hysteresis of the management measures is prominent in a high-risk scene of multi-factor interleaving. In this field, technical challenges are mainly focused on how to efficiently integrate and analyze multi-source information. The hoisting operation involves various contents such as the running state of equipment, the change of the surrounding environment, the operation behaviors of personnel and the like, and the information sources are wide and have different formats, and lack of uniform relevance. Because the scattered information cannot form an integral judgment basis, in actual operation, a manager can hardly accurately identify which links are likely to have problems. For example, anomalies in equipment operating parameters may be superimposed with severe weather conditions, further amplifying the risk, but the prior art has difficulty in combining the two for analysis, and thus cannot predict potential chain reactions in advance. Therefore, how to integrate multidimensional information such as equipment, environment, personnel and the like in real time in the whole flow of hoisting operation and accurately identify potential safety risks therefrom becomes a key problem for improving the safety management level of the operation. The solution of the problem not only needs technical breakthrough, but also needs to realize the dynamic perception and timely coping of complex risks in the actual business scene. Disclosure of Invention The invention aims to provide a hoisting safety risk identification method based on big data, which improves the safety and efficiency of hoisting operation and effectively reduces accident risk. In order to achieve the above object, the present invention provides the following solutions: A hoisting safety risk identification method based on big data comprises the following steps: The method comprises the steps of obtaining multi-source data of equipment states, environmental changes and personnel behaviors from a hoisting site, preprocessing the multi-source data, and obtaining a unified space-time data sequence; according to the unified space-time data sequence, adopting a convolutional neural network to extract space-time characteristics of equipment state and environmental change, judging whether the space-time characteristics exceed a preset threshold value, and acquiring a risk level; According to the risk level, analyzing time sequence association of personnel behaviors and equipment states through a long-term and short-term memory network, and acquiring an evolution trend of potential hazards; and acquiring risk probability distribution according to the evolution trend of the potential hidden danger, analyzing the risk probability distribution, and acquiring a type of high risk. Optionally, preprocessing the multi-source data includes: performing data cleaning on the multi-source data to obtain a preliminary data set; performing noise removal and inconsistent optimization on the preliminary data set by adopting a Kalman filtering algorithm to obtain an intermediate data sequence; If the time stamps in the intermediate data sequences are not matched, time correction is carried out, and a corrected data sequence with consistent time is obtained; and carrying out position calibration by adopting a space mapping method according to the corrected data sequence, and adjusting the space distribution of the hoisting site to obtain the unified space-time data sequence. Optionally, extracting space-time characteristics of equipment state and environmental change by using a convolutional neural network, judging whether the