[go: up one dir, main page]

CN120611147A - A method and system for tracing and locating the source of hazardous gas leaks based on artificial intelligence - Google Patents

A method and system for tracing and locating the source of hazardous gas leaks based on artificial intelligence

Info

Publication number
CN120611147A
CN120611147A CN202510701466.2A CN202510701466A CN120611147A CN 120611147 A CN120611147 A CN 120611147A CN 202510701466 A CN202510701466 A CN 202510701466A CN 120611147 A CN120611147 A CN 120611147A
Authority
CN
China
Prior art keywords
data
sensor
artificial intelligence
tracing
model
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Pending
Application number
CN202510701466.2A
Other languages
Chinese (zh)
Inventor
郎云飞
白俊伟
杨彦巧
曹振亚
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Henan Baoshian Technology Co ltd
Original Assignee
Henan Baoshian Technology Co ltd
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Henan Baoshian Technology Co ltd filed Critical Henan Baoshian Technology Co ltd
Priority to CN202510701466.2A priority Critical patent/CN120611147A/en
Publication of CN120611147A publication Critical patent/CN120611147A/en
Pending legal-status Critical Current

Links

Classifications

    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01M—TESTING STATIC OR DYNAMIC BALANCE OF MACHINES OR STRUCTURES; TESTING OF STRUCTURES OR APPARATUS, NOT OTHERWISE PROVIDED FOR
    • G01M3/00—Investigating fluid-tightness of structures
    • G01M3/02—Investigating fluid-tightness of structures by using fluid or vacuum
    • G01M3/04—Investigating fluid-tightness of structures by using fluid or vacuum by detecting the presence of fluid at the leakage point
    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
    • G01N33/00—Investigating or analysing materials by specific methods not covered by groups G01N1/00 - G01N31/00
    • G01N33/0004—Gaseous mixtures, e.g. polluted air
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
    • G06F18/211—Selection of the most significant subset of features
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
    • G06F18/213—Feature extraction, e.g. by transforming the feature space; Summarisation; Mappings, e.g. subspace methods
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/25—Fusion techniques
    • G06F18/251—Fusion techniques of input or preprocessed data
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/25—Fusion techniques
    • G06F18/253—Fusion techniques of extracted features
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/004—Artificial life, i.e. computing arrangements simulating life
    • G06N3/006—Artificial life, i.e. computing arrangements simulating life based on simulated virtual individual or collective life forms, e.g. social simulations or particle swarm optimisation [PSO]
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/042—Knowledge-based neural networks; Logical representations of neural networks
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/044—Recurrent networks, e.g. Hopfield networks
    • G06N3/0442—Recurrent networks, e.g. Hopfield networks characterised by memory or gating, e.g. long short-term memory [LSTM] or gated recurrent units [GRU]
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/045—Combinations of networks
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/0464—Convolutional networks [CNN, ConvNet]
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/08—Learning methods
    • G06N3/092—Reinforcement learning
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q10/00—Administration; Management
    • G06Q10/04—Forecasting or optimisation specially adapted for administrative or management purposes, e.g. linear programming or "cutting stock problem"
    • G06Q10/047—Optimisation of routes or paths, e.g. travelling salesman problem
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q10/00—Administration; Management
    • G06Q10/06—Resources, workflows, human or project management; Enterprise or organisation planning; Enterprise or organisation modelling
    • G06Q10/063—Operations research, analysis or management
    • G06Q10/0637—Strategic management or analysis, e.g. setting a goal or target of an organisation; Planning actions based on goals; Analysis or evaluation of effectiveness of goals
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q50/00—Information and communication technology [ICT] specially adapted for implementation of business processes of specific business sectors, e.g. utilities or tourism
    • G06Q50/10—Services
    • G06Q50/26—Government or public services
    • G06Q50/265—Personal security, identity or safety
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F2123/00—Data types
    • G06F2123/02—Data types in the time domain, e.g. time-series data

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Data Mining & Analysis (AREA)
  • Business, Economics & Management (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Artificial Intelligence (AREA)
  • General Engineering & Computer Science (AREA)
  • Evolutionary Computation (AREA)
  • Health & Medical Sciences (AREA)
  • Human Resources & Organizations (AREA)
  • General Health & Medical Sciences (AREA)
  • Mathematical Physics (AREA)
  • Molecular Biology (AREA)
  • Computing Systems (AREA)
  • Software Systems (AREA)
  • Computational Linguistics (AREA)
  • Biophysics (AREA)
  • Biomedical Technology (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Strategic Management (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Evolutionary Biology (AREA)
  • Economics (AREA)
  • Tourism & Hospitality (AREA)
  • Entrepreneurship & Innovation (AREA)
  • Marketing (AREA)
  • General Business, Economics & Management (AREA)
  • Educational Administration (AREA)
  • Development Economics (AREA)
  • Chemical & Material Sciences (AREA)
  • Quality & Reliability (AREA)
  • Game Theory and Decision Science (AREA)
  • Operations Research (AREA)
  • Primary Health Care (AREA)
  • Computer Security & Cryptography (AREA)
  • Combustion & Propulsion (AREA)
  • Food Science & Technology (AREA)

Abstract

本发明涉及气体泄漏溯源定位技术领域，特别涉及一种基于人工智能的危险气体泄漏溯源定位方法及系统，溯源定位方法包括：数据采集，从固定和移动传感器收集环境参数和气体浓度数据，固定传感器部署在关键节点，以监测气体扩散；移动传感器搭载于巡检机器人或无人机，以动态填补固定传感器盲区；数据预处理，对数据进行去噪、归一化和特征增强；特征提取与选择，利用机器学习算法提取关键特征；模型训练，训练深度神经网络模型预测泄漏源位置；监测与触发，监控数据并在异常时触发溯源定位；精准定位，利用模型和实时数据定位泄漏源；智能决策，提供应急响应建议和优化方案。

The present invention relates to the technical field of gas leak tracing and positioning, and in particular to a method and system for tracing and positioning the source of hazardous gas leaks based on artificial intelligence. The tracing and positioning method includes: data acquisition, collecting environmental parameters and gas concentration data from fixed and mobile sensors, and deploying fixed sensors at key nodes to monitor gas diffusion; mobile sensors are mounted on inspection robots or drones to dynamically fill the blind spots of fixed sensors; data preprocessing, denoising, normalizing and feature enhancement of data; feature extraction and selection, extracting key features using machine learning algorithms; model training, training a deep neural network model to predict the location of the leakage source; monitoring and triggering, monitoring data and triggering tracing and positioning in the event of an anomaly; precise positioning, locating the leakage source using models and real-time data; and intelligent decision-making, providing emergency response suggestions and optimization solutions.

Description

Dangerous gas leakage tracing positioning method and system based on artificial intelligence
Technical Field
The invention relates to the technical field of gas leakage tracing positioning, in particular to a dangerous gas leakage tracing positioning method and system based on artificial intelligence.
Background
In modern industrial production, hazardous gas leakage events are a non-negligible safety issue. It not only threatens the life safety of factory staff, but also may cause serious damage to the surrounding environment. Therefore, effective detection and tracing positioning of dangerous gas leakage are key links for guaranteeing industrial safety production. However, existing hazardous gas leak detection and traceability positioning techniques have a number of drawbacks that greatly limit the efficiency and effectiveness of the monitoring system and increase the risk of accidents.
First, conventional hazardous gas leak detection relies primarily on manual inspection and limited sensor networks. The coverage of the detection mode is very limited, and particularly, the detection mode is difficult to realize comprehensive monitoring in a wide industrial area or a place with complex topography. In addition, the manual inspection efficiency is low, and the manual inspection is easily affected by human factors, such as vision blind areas, judgment errors and the like, so that the detection accuracy is insufficient. Meanwhile, the sensor network may cause false alarm due to environmental interference or equipment failure, and the risk of false positive is increased.
Secondly, the existing monitoring system is slow in response speed. After detecting an abnormal situation, it takes a long time to locate the source of the leak, which may lead to a spread of the leak, increasing the risk of accident. Particularly in emergency situations, rapid and accurate localization of the source of leakage is critical for timely countermeasures.
Again, the creation and maintenance of a comprehensive sensor network requires a significant investment and costs increase dramatically as the monitoring range increases. This limits to some extent the popularity and applicability of monitoring systems. In addition, the existing data analysis method often cannot effectively process large-scale, multi-source and real-time data streams, so that leakage situations cannot be analyzed timely and accurately.
In addition, existing monitoring devices suffer from reduced performance under extreme environmental conditions (e.g., high temperature, low temperature, corrosive environments) and fail to provide reliable monitoring data. This limits the application of the monitoring system in various industrial scenarios.
Finally, existing monitoring systems lack intelligent decision support. After leakage occurs, the lack of an effective intelligent decision support system to guide the emergency response results in inadequate accuracy and efficiency of the emergency action. This affects the efficiency and effectiveness of the incident processing to some extent.
Therefore, aiming at the limitation of the current gas leakage tracing positioning, the research and development of the dangerous gas leakage tracing positioning method and system based on artificial intelligence are particularly urgent and important for promoting the development of the related technical fields.
Disclosure of Invention
The invention aims to provide an artificial intelligence-based hazardous gas leakage tracing positioning method and system, which are used for solving the technical problems in the background technology.
In order to achieve the above purpose, the present invention provides the following technical solutions:
dangerous gas leakage tracing positioning method based on artificial intelligence, the tracing positioning method comprises the following steps:
S1, data acquisition, namely acquiring environmental parameters and gas concentration through a fixed sensor and a mobile sensor;
S2, data preprocessing, namely denoising, normalizing and characteristic enhancing the collected data;
S3, extracting and selecting key features from the preprocessed data by using a machine learning algorithm;
s4, training a deep neural network model, wherein the model can predict the position of a leakage source according to the extracted characteristics;
s5, monitoring and triggering, namely monitoring sensor data and triggering tracing positioning when abnormality is detected;
S6, accurately positioning the leakage source by using the trained deep learning model and real-time data;
And S7, intelligent decision, namely providing emergency response suggestions and optimizing an emergency scheme according to the positioning result and the environmental parameters.
Preferably, the fixed sensors in the data acquisition step are deployed at key nodes of the industrial facility to acquire environmental parameters and gas concentrations in real time.
Preferably, the mobile sensor in the data acquisition step is mounted on the inspection robot or the unmanned aerial vehicle so as to dynamically fill the dead zone of the fixed sensor.
Preferably, the deep learning model includes a plurality of sub-networks, each sub-network being responsible for processing a different type of sensor data.
Preferably, the training of the S4 deep learning model comprises the steps of:
S401, constructing a labeling data set covering various scenes, and training and verifying a support model;
S402, generating a three-dimensional map based on a factory geographic information system, marking the installation position of a fixed sensor in the three-dimensional map, and dividing a gridding environment model;
s403, fusing the space-time data of the fixed sensor and the movable sensor through a multi-source data fusion model, and outputting the three-dimensional coordinates of the leakage source;
s404, optimizing a mobile sensor inspection path to maximize blind area coverage rate;
s405, triggering local grid refinement when the mobile sensor detects leakage, and performing tracing positioning;
S406, adjusting parameters of the mesoscale turbulence diffusion model according to the real-time detection data so as to adapt to real-time environmental changes.
Preferably, the multi-source data fusion model comprises:
a space-time feature extraction module modeling the topological relation of the fixed sensor through a graph rolling network (GCN), and extracting the time sequence feature of the mobile sensor through LSTM;
And the fusion decision module aligns the characteristic vectors of the fixed sensor and the mobile sensor through a cross attention mechanism to generate a joint embedded vector.
Preferably, the number of layers of GCN is three to five.
Preferably, the mobile sensor adopts a dynamic path planning module to plan the inspection path, and the dynamic path planning module adopts a multi-agent depth deterministic strategy gradient algorithm.
Preferably, the S3 feature extraction and selection step includes an adaptive feature fusion algorithm for fusing and optimizing features from different sensors.
Preferably, the S7 intelligent decision step includes a multi-objective optimization algorithm for selecting the optimal solution among a plurality of emergency response scenarios while taking into account response speed and resource consumption.
The invention has the technical effects and advantages that:
The invention realizes wider monitoring range, faster response speed, higher accuracy and lower cost by combining multisource data acquisition, a deep learning model, self-adaptive feature extraction and intelligent decision support. Meanwhile, the method has good environmental adaptability and generalization capability, can effectively work in various industrial scenes, provides an optimal emergency response scheme, remarkably improves the efficiency and effect of dangerous gas leakage detection and tracing positioning, and reduces accident risks.
Drawings
FIG. 1 is a schematic flow chart of a main body of the invention;
FIG. 2 is a schematic flow chart of the model training of the present invention.
Detailed Description
The following description of the embodiments of the present invention will be made clearly and completely with reference to the accompanying drawings, in which it is apparent that the embodiments described are only some embodiments of the present invention, but not all embodiments. All other embodiments, which can be made by those skilled in the art based on the embodiments of the invention without making any inventive effort, are intended to be within the scope of the invention.
Example 1
Referring to fig. 1 to 2, the invention provides a dangerous gas leakage tracing positioning method based on artificial intelligence, which comprises the following steps:
S1, data acquisition, namely acquiring environmental parameters and gas concentration through a fixed sensor and a mobile sensor;
The fixed sensors are deployed at key nodes of the industrial facility to enable real-time acquisition of environmental parameters and gas concentrations. Meanwhile, the mobile sensor is carried on the inspection robot or the unmanned aerial vehicle and is used for dynamically filling the monitoring blind area of the fixed sensor.
The parameters detected by the fixed sensors cover gas concentration, temperature and humidity, air pressure and three-dimensional coordinate information based on a factory Geographic Information System (GIS). The sensor adopts a configuration mode of a plurality of types of sensors, aims to cover different gas types and concentration ranges in a whole range, and effectively avoids monitoring errors caused by performance limitation of a single sensor.
The mobile sensor system consists of an air part and a ground part. The aerial part adopts the multiaxis gas sampling probe of carrying on unmanned aerial vehicle, and unmanned aerial vehicle flight altitude sets for 5~30m within range to this dynamic coverage fixed sensor is difficult to the region that reaches, ensures the omnidirectional environmental monitoring. The ground part consists of an infrared spectrometer and a laser radar which are carried on the inspection robot, and the inspection robot executes an inspection task according to a preset path, so that the ground environment is finely monitored.
The data collected by the fixed sensor and the mobile sensor are transmitted to the data processing unit in real time through the wired or wireless communication module. After data acquisition is completed, the data is reliably transmitted to the gas monitoring system by means of the internet of things transmission technology.
S2, data preprocessing, namely denoising, normalizing and characteristic enhancing the collected data;
Data denoising, namely, aiming at environmental interference (such as temperature and humidity mutation and cross gas interference) and sensor noise (drifting and pulse abnormality) in a gas concentration signal, adopting wavelet decomposition to filter high-frequency noise, combining Kalman filtering to dynamically correct sensor drifting errors, removing outlier concentration points caused by wind direction mutation or sensor faults by utilizing spatial clustering (DBSCAN), smoothing wind speed time sequence fluctuation by sliding window mean filtering, simultaneously introducing a Gaussian plume diffusion model to carry out physical constraint on concentration gradients, correcting non-physical mutation caused by turbulence, and reserving space-time distribution characteristics strongly related to leakage sources.
The Gaussian smoke plume diffusion model is a classical physical mathematical model, and is used for describing the diffusion rule of gas in an open environment, and the mathematical model is as follows:
Wherein:
q is the leakage source intensity (release amount per unit time);
σ y is the lateral direction diffusion coefficient;
σ z is the vertical diffusion coefficient;
u is the average wind speed;
H is the leakage source effective height.
The data normalization comprises the steps of eliminating dimension influence by adopting Z-Score standardization aiming at multi-sensor dimension differences (such as concentration ppm, wind speed m/s and temperature ℃), enhancing sensitivity of a model to global concentration gradient, adopting quantile normalization on sensor signals distributed in long tails (such as instantaneous high concentration pulse) to avoid abnormal value leading model training, mapping concentration to local space grids (such as 0-1 interval) by Min-Max normalization aiming at space data of multiple monitoring stations, and highlighting relative space modes of leakage diffusion.
And simultaneously, based on the gas diffusion mechanism and the space-time correlation, constructing space-time joint characteristics, namely extracting a lag difference of concentration time sequences and window statistics (mean value and variance). And calculating concentration gradients, diffusion direction consistency and spatial interpolation concentration fields of adjacent sensors. And the leakage intensity and the diffusion speed are reversely deduced by using the Gaussian plume model and are used as prior characteristic input. And expanding the small sample leakage scene data by adding Gaussian noise or simulating diffusion tracks at different wind speeds. The concentration-wind speed product (simulating leakage flux) and the temperature-humidity-concentration ratio (correcting environmental interference) are constructed.
S3, extracting and selecting key features from the preprocessed data by using a machine learning algorithm;
and adopting a self-adaptive feature fusion algorithm to fuse and optimize the features from different sensors according to the characteristics and the importance of the data of the different sensors.
S4, training a deep neural network model, wherein the model can predict the position of a leakage source according to the extracted characteristics;
s5, monitoring and triggering, namely monitoring sensor data and triggering tracing positioning when abnormality is detected;
By setting an abnormal threshold value, monitoring data transmitted by the fixed sensor and the mobile sensor in real time, triggering a tracing positioning process when the data detected by the fixed sensor or the mobile sensor exceeds the threshold value, and positioning a gas leakage point in real time
S6, accurately positioning the leakage source by using the trained deep learning model and real-time data;
After the tracing positioning is started, data acquired by the fixed sensor and the mobile sensor in real time are input into a trained learning model, and the model outputs the position information of the leakage source. Meanwhile, combining with a Geographic Information System (GIS) in a factory, accurately positioning the position of a leakage source on a map and displaying a dangerous area (red coverage) and an evacuation route (green arrow) within the radius of the leakage source of 20m
And S7, intelligent decision, namely providing emergency response suggestions and optimizing an emergency scheme according to the positioning result and the environmental parameters.
And selecting an optimal solution from a plurality of emergency response schemes by adopting a multi-objective optimization algorithm and considering factors such as response speed, resource consumption and the like. For example, depending on the location of the leakage source and the gas diffusion trend, the optimal evacuation route and emergency rescue protocol are selected.
Referring to fig. 2, in the present application, training of the S4 deep learning model includes the steps of:
S401, constructing a labeling data set covering various scenes, and training and verifying a support model;
S402, generating a three-dimensional map based on a factory geographic information system, marking the installation position of a fixed sensor in the three-dimensional map, and dividing a gridding environment model;
In the embodiment, the gas leakage tracing positioning learning model is combined with a multi-mode data fusion technology of a fixed sensor and a mobile sensor to realize accurate positioning in a three-dimensional space.
The data acquisition layer mainly comprises a fixed sensor and a mobile sensor, wherein the fixed sensor is deployed at key nodes such as a pipeline connection part and a valve, three-dimensional coordinates of the fixed sensor are marked by a factory BIM/GIS system (precision +/-0.1 m), and the data sampling frequency is 10Hz.
The mobile sensor comprises a ground inspection robot and an unmanned aerial vehicle, wherein the ground inspection robot is provided with an infrared spectrometer (sensitivity is 0.1 ppm) for inspection at a speed of 2m/s along a preset path, and the unmanned aerial vehicle is provided with a multi-axis gas probe (response time is less than 5 s), has a flight height of 5-30 m and returns three-dimensional position coordinates in real time.
Three-dimensional space modeling constructs a three-dimensional map based on a factory BIM/GIS system, marks sensor positions (X, Y and Z coordinates), and establishes a gridding environment model (gridding resolution 0.5m 3).
S403, fusing the space-time data of the fixed sensor and the movable sensor through a multi-source data fusion model, and outputting the three-dimensional coordinates of the leakage source;
the multi-source data fusion model comprises the following contents:
Coordinate data of the fixed sensor and the mobile sensor are input, wherein the fixed sensor data are S fix＝{Ci,Ti,Hi,Pi,(xi,yi,zi, C i is gas concentration, and T/H/P is environmental parameter. The motion sensor data is S mobile＝{Cj(t),(xj(t),yj(t),zj (t)) } containing time series dynamic coordinates.
The network structure comprises a space-time feature extraction module and a fusion decision module. The space-time feature extraction module models the topological relation of the fixed sensor through a graph rolling network, and extracts the time sequence features of the mobile sensor through LSTM. The fusion decision module aligns the feature vectors of the fixed and mobile sensors through a cross attention mechanism to generate a joint embedded vector
The space-time feature extraction module consists of a fixed sensor and a mobile sensor:
sensor topology is modeled using a graph roll-up network (GCN).
Wherein the number of layers of GCN is 3 to control the topology aggregation depth of the sensor network, each layer has an aggregation radius of 5m to ensure the coverage of adjacent key nodes, and the node characteristics comprise gas concentration (C_i), temperature (T_i), humidity (H_i), gas pressure (P_i) and three-dimensional coordinates (x_i, y_i, z_i), and the edge weight is calculated based on a three-dimensional Euclidean distance calculation formula among the sensorsD ij is the sensor pitch.
And (3) moving the sensor, namely extracting time sequence characteristics by using LSTM, and capturing concentration mutation points by combining a self-attention mechanism.
Wherein, the hidden layer dimension of LSTM (128 dimensions) is input as dynamic coordinates (x_j (t), y_j (t), z_j (t)) to capture the time series concentration variation of the mobile sensor (robot/drone). The time window is 10 seconds long by default, and the inspection speed of the unmanned aerial vehicle is matched (about 2 m/s).
And a fusion decision module:
Feature level fusion, namely aligning the space-time features of a fixed sensor and a mobile sensor through a Cross-Attention mechanism (Cross-Attention), and generating a joint embedding vector F fusion.
Attention header number (4 header) alignment of spatiotemporal features of fixed and mobile sensors to generate joint embedding vector F fusion
Gating weight, namely dynamically adjusting based on feature reliability, and the formula: Wherein RI i is a data source reliability index.
Output layer-fully connected layer maps to leakage source coordinates (x, y, z), loss function employs Huber Loss (balanced outlier effect).
The Huber Loss parameter is delta=1.0 to balance the robustness of positioning errors and reduce outlier interference. Gradient explosion was prevented using Adam optimizer in combination with gradient clipping (max_grad_norm=5.0).
S404, optimizing a mobile sensor inspection path through a dynamic path planning algorithm, and maximizing blind area coverage rate;
The dynamic path planning algorithm flow comprises:
Blind zone identification, marking potential leak areas Ω blind, based on fixed sensor data, by predicting the concentration profile of the unmonitored area by Gaussian Process Regression (GPR).
Path generation, defining a reward function using a multi-agent depth deterministic strategy gradient (MADDPG):
r=α·coverage +beta concentration gradient-gamma-mobile energy consumption
Wherein the bonus function weights:
alpha=0.6 (coverage ratio) mesh coverage ratio based on blind zone identification result Ω blind;
Beta=0.3 (concentration gradient), the concentration change rate detected by the motion sensor;
γ=0.1 (mobile energy consumption) the power consumption of the robot/drone is proportional to the square of the distance and speed of movement.
And the cooperative obstacle avoidance is to introduce a Voronoi diagram to divide a patrol area, wherein the patrol area is an area generated based on the density of sensor dead zones so as to ensure that the tracks of multiple devices have no conflict.
S405, triggering local grid refinement when the mobile sensor detects leakage, and performing tracing positioning;
s406, adjusting the mesoscale turbulence diffusion model according to the real-time detection data so as to adapt to the real-time environmental change.
Introducing a mesoscale turbulence diffusion model to simulate a gas diffusion path, the mesoscale turbulence diffusion model comprising turbulence diffusion coefficients and dynamic calibration of environmental parameters
The turbulent diffusion coefficient includes a vortex diffusion coefficient and a laminar diffusion coefficient:
vortex diffusion coefficient:
Wherein:
k is turbulent kinetic energy;
Epsilon is the dissipation ratio;
C μ is an empirical constant of 0.09, and is suitable for near-surface laminar flow in industrial environments.
Laminar diffusion coefficient:
Wherein:
v is the air kinematic viscosity (1.5X10 -5m2/s);
sc is Schmitt number (. Apprxeq.0.7).
Example two
The invention also provides an artificial intelligence-based hazardous gas leakage tracing positioning system which is used for realizing an artificial intelligence-based hazardous gas leakage tracing positioning method and comprises a data acquisition module, a data preprocessing module, a feature extraction and selection module, a model training module, a monitoring and triggering module, a positioning module, a decision support module and a model self-adaptive adjustment module.
The model self-adaptive adjustment module is used for carrying out self-adaptive adjustment on the model by continuously collecting data and effects of actual emergency response, automatically adjusting model parameters according to new data and environmental changes, optimizing an emergency scheme and improving generalization capability and adaptability of the model.
Dynamic data pools are built by integrating multiple source data streams of sensors, emergency response records (e.g., leak disposal time, resource consumption), etc. And evaluating the performance of the model through preset quantization indexes (such as positioning error rate and emergency response delay time) to generate feedback signal driving parameter adjustment. Assuming that the predicted position of the model deviates from the actual leakage point by more than a set threshold (e.g., 5 meters), the system automatically triggers the parameter optimization procedure.
Meanwhile, based on local environment changes (such as sudden changes of wind speed and new addition of equipment), the contribution weight of the sensor characteristics is dynamically adjusted through an attention mechanism. When the wind speed suddenly changes, for example, the originally stable wind direction suddenly changes, or the wind speed greatly increases or decreases in a short time, the change significantly affects the diffusion path and range after the gas leakage. At this time, the attention mechanism captures the change sharply, and automatically reduces the characteristic weights of sensors which are less influenced by the wind speed or have low correlation, such as the characteristic weights of certain sensor data which are fixed in a relatively closed space and are less disturbed by the external wind speed. Meanwhile, the characteristic weights of the sensors which are sensitive to wind speed change and can better reflect the new situation of gas diffusion, such as the characteristic weights of the sensor data which are carried on an unmanned plane and can flexibly sense the change of a wind field in a large area and the dynamic state of gas diffusion, are improved.
If large-scale equipment is newly added in the factory, the equipment may change local air flow trend, temperature distribution and other environmental factors, so as to influence the propagation characteristics of the air leakage. The attention mechanism will rapidly analyze this local environmental change, re-assessing the importance of each sensor feature for accurate traceability positioning. The sensor features which are close to the newly added equipment and are greatly influenced by the newly added equipment are given higher weight, because the features can reflect the influence of environmental changes caused by the newly added equipment on gas leakage more timely and accurately, and the sensor features which are far away from the newly added equipment and are less influenced by the newly added equipment are appropriately reduced in weight, so that unnecessary interference of the features on model judgment is avoided. By means of the dynamic adjustment of the sensor characteristic contribution weight, the model can be ensured to keep high tracing positioning accuracy and adaptability under different local environment changes.
The present invention is not limited to the above-mentioned embodiments, and any person skilled in the art, based on the technical solution of the present invention and the inventive concept thereof, can be replaced or changed within the scope of the present invention.

Claims (10)

1. The dangerous gas leakage tracing positioning method based on artificial intelligence is characterized by comprising the following steps of:
S1, data acquisition, namely acquiring environmental parameters and gas concentration through a fixed sensor and a mobile sensor;
S2, data preprocessing, namely denoising, normalizing and characteristic enhancing the collected data;
S3, extracting and selecting key features from the preprocessed data by using a machine learning algorithm;
s4, training a deep neural network model, wherein the model can predict the position of a leakage source according to the extracted characteristics;
s5, monitoring and triggering, namely monitoring sensor data and triggering tracing positioning when abnormality is detected;
S6, accurately positioning the leakage source by using the trained deep learning model and real-time data;
And S7, intelligent decision, namely providing emergency response suggestions and optimizing an emergency scheme according to the positioning result and the environmental parameters.
2. The dangerous gas leakage tracing positioning method based on artificial intelligence according to claim 1, wherein the fixed sensor in the data acquisition step is deployed at a key node of an industrial facility, and environmental parameters and gas concentrations are acquired in real time.
3. The dangerous gas leakage tracing positioning method based on artificial intelligence according to claim 2, wherein the mobile sensor in the data acquisition step is mounted on a patrol robot or an unmanned aerial vehicle to dynamically fill a dead zone of the fixed sensor.
4. The dangerous gas leakage traceability positioning method based on artificial intelligence according to claim 2, wherein the deep learning model comprises a plurality of sub-networks, and each sub-network is responsible for processing different types of sensor data.
5. The artificial intelligence based hazardous gas leakage traceability positioning method according to claim 4, wherein the training of the S4 deep learning model comprises the following steps:
S401, constructing a labeling data set covering various scenes, and training and verifying a support model;
S402, generating a three-dimensional map based on a factory geographic information system, marking the installation position of a fixed sensor in the three-dimensional map, and dividing a gridding environment model;
s403, fusing the space-time data of the fixed sensor and the movable sensor through a multi-source data fusion model, and outputting the three-dimensional coordinates of the leakage source;
s404, optimizing a mobile sensor inspection path to maximize blind area coverage rate;
s405, triggering local grid refinement when the mobile sensor detects leakage, and performing tracing positioning;
S406, adjusting parameters of the mesoscale turbulence diffusion model according to the real-time detection data so as to adapt to real-time environmental changes.
6. The artificial intelligence based hazardous gas leakage traceability positioning method according to claim 5, wherein the multi-source data fusion model comprises:
a space-time feature extraction module modeling the topological relation of the fixed sensor through a graph rolling network (GCN), and extracting the time sequence feature of the mobile sensor through LSTM;
And the fusion decision module aligns the characteristic vectors of the fixed sensor and the mobile sensor through a cross attention mechanism to generate a joint embedded vector.
7. The dangerous gas leakage tracing positioning method based on artificial intelligence according to claim 6, wherein the number of layers of the GCN is three to five.
8. The dangerous gas leakage tracing positioning method based on artificial intelligence according to claim 4, wherein the mobile sensor adopts a dynamic path planning module to plan a routing inspection path, and the dynamic path planning module adopts a multi-agent depth deterministic strategy gradient algorithm.
9. The method of claim 1, wherein the step of S3 feature extraction and selection includes an adaptive feature fusion algorithm for fusing and optimizing features from different sensors.
10. The hazardous gas leakage traceability positioning method based on artificial intelligence according to claim 1, wherein the step of S7 intelligent decision-making comprises a multi-objective optimization algorithm for selecting an optimal solution among a plurality of emergency response schemes while considering response speed and resource consumption.
CN202510701466.2A 2025-05-28 2025-05-28 A method and system for tracing and locating the source of hazardous gas leaks based on artificial intelligence Pending CN120611147A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202510701466.2A CN120611147A (en) 2025-05-28 2025-05-28 A method and system for tracing and locating the source of hazardous gas leaks based on artificial intelligence

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202510701466.2A CN120611147A (en) 2025-05-28 2025-05-28 A method and system for tracing and locating the source of hazardous gas leaks based on artificial intelligence

Publications (1)

Publication Number Publication Date
CN120611147A true CN120611147A (en) 2025-09-09

Family

ID=96927860

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202510701466.2A Pending CN120611147A (en) 2025-05-28 2025-05-28 A method and system for tracing and locating the source of hazardous gas leaks based on artificial intelligence

Country Status (1)

Country Link
CN (1) CN120611147A (en)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN121209546A (en) * 2025-11-26 2025-12-26 山西瑞赛科环保科技有限公司 Unmanned aerial vehicle-based environment inspection system, method, equipment and storage medium

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN121209546A (en) * 2025-11-26 2025-12-26 山西瑞赛科环保科技有限公司 Unmanned aerial vehicle-based environment inspection system, method, equipment and storage medium

Similar Documents

Publication Publication Date Title
CN119888507B (en) Unmanned aerial vehicle image analysis-based method and system for detecting construction abnormality of power extraction and storage station
CN120258241A (en) A dynamic optimization method for communication engineering construction based on multi-dimensional perception
CN119165884B (en) A low-altitude patrol system and method for AI drones in smart cities
CN119292314A (en) An intelligent planning system and method for UAV disaster emergency inspection paths
CN120611147A (en) A method and system for tracing and locating the source of hazardous gas leaks based on artificial intelligence
CN120446400A (en) Air pollution monitoring method based on UAV remote sensing and machine learning
CN119573744B (en) Path planning and optimizing system for offshore wind power mine sweeping operation
CN120063288A (en) Substation route inspection planning method based on deep learning
CN120689840A (en) Real-time safety hazard identification system for robots based on multimodal sensor fusion
CN120427841A (en) Combustible gas intelligent detection system and method based on mobile robot
CN119760406A (en) A state warning method for safety protection device used in building construction
CN120258758A (en) Power AI intelligent auxiliary control and inspection system
CN120293155B (en) Unmanned aerial vehicle-based artificial intelligent inspection system and method
CN118584887B (en) Unmanned aerial vehicle flight state safety supervision method and system
CN120910815A (en) Multi-source heterogeneous data acquisition method for aircraft safety data
CN120893676A (en) A Comprehensive Monitoring Method and System for Power Construction Personnel Based on BeiDou Positioning
CN120143022A (en) A multi-modal electromagnetic field distribution mapping unmanned aerial vehicle system and its application method
CN120416773A (en) Beidou positioning inspection method and system
CN120220877A (en) A method and system for emergency monitoring and early warning of atmospheric environmental pollution events
CN119558510A (en) Dynamic drift prediction method and system for fishing vessels based on distress analysis in South China Sea waters
Fan et al. Navigation of substation inspection robot based on machine vision
CN116880573B (en) Collaborative control method and system for unmanned missile-borne unmanned aerial vehicle and unmanned detection aerial vehicle
CN121071825B (en) A system and method for fusing multi-source data on bridge structures
CN120599532B (en) Cross-region AI track tracking method based on multi-information fusion
CN121117541B (en) An AI-based method for monitoring the flight status of unmanned aerial vehicles (UAVs)

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination