Gas pipeline safe installation scheme generation method and system

Abstract

The invention relates to the field of data processing, in particular to a gas pipeline safety installation scheme generation method and system. According to the basic principle of the method, multi-modal data of an installation environment is collected through various sensors; based on the data, utilizing a risk quantification model to respectively generate a risk field aiming at a gas accumulation negative space and a non-uniform repulsion potential field aiming at different types of ignition sources, and fusing the risk field and the non-uniform repulsion potential field to construct a static risk map; on the map, an improved path search algorithm is adopted for pipeline laying path planning, and a weighted cumulative risk cost item is introduced into a cost function of the algorithm to balance safety and economical efficiency; and finally, outputting a safe installation scheme containing the optimized path. According to the method, automation and intellectualization of gas pipeline installation scheme design are achieved, gas accumulation and ignition risks caused by improper pipeline arrangement can be systematically avoided from the source, and the intrinsic safety level and design efficiency are remarkably improved.

Description

Technical Field

[0001] This invention relates to the field of data processing, and more specifically, to a method and system for generating a safe installation scheme for gas pipelines. Background Technology

[0002] Currently, the penetration rate of urban gas supply continues to rise, and residential users have increasingly higher requirements for the safety and standardization of gas installation in their homes. The industry is transforming towards digitalization and intelligence, for example, using BIM technology for design and digital twins for management during the operation and maintenance phase. However, in the last mile of gas pipeline installation—the specific design phase—the core technical means still heavily rely on manual labor. Designers need to conduct on-site surveys using tools such as measuring tapes and mainly rely on personal experience to apply safety regulations on two-dimensional drawings. The entire process suffers from significant problems of low efficiency and inconsistent quality.

[0003] The existing technology has the following three main shortcomings: First, there is a lack of effective quantitative perception and assessment methods for negative spaces where gas leaks may accumulate (such as inside suspended ceilings and above enclosed cabinets). Judgments are often based solely on the designer's subjective experience, easily creating blind spots for safety hazards. Second, current safety distance standards are rigid, employing a one-size-fits-all fixed distance standard. This fails to accurately differentiate the actual risk differences between continuous high-temperature heat sources and intermittent ignition sources, potentially leading to insufficient protection in high-risk scenarios or over-design and wasted installation space in low-risk scenarios. Finally, the quality of design solutions highly depends on the designer's individual skills and experience, making standardization and large-scale replication difficult. Furthermore, the design process cannot effectively quantify and optimize multiple objectives such as safety and economy.

[0004] Therefore, there is an urgent need in this field for a technology that can automatically generate gas pipeline installation solutions that combine high safety and high feasibility in order to systematically solve the above problems. Summary of the Invention

[0005] The purpose of this invention is to provide a technology that can generate gas pipeline installation solutions that combine high safety and high feasibility.

[0006] According to a first aspect of the present invention, a method for generating a safe installation scheme for gas pipelines is provided, comprising the following steps: S1. Data Acquisition Steps: Collect multimodal data on the gas pipeline installation environment using multiple sensors; S2. Risk map construction steps: Based on multimodal data, a risk field for the negative space of gas accumulation and a repulsive potential field for the ignition source are generated through a risk quantification model, and the two are merged to generate a static risk map; S3. Path planning steps: On the static risk map, use an improved path search algorithm to plan the pipeline laying path, wherein the cost function of the improved path search algorithm includes a cumulative risk cost term; S4. Solution Output Steps: Output a safe installation solution that includes the pipeline laying path.

[0007] According to some embodiments, in the first method of the present invention, step S1. data acquisition, acquiring multimodal data through multiple sensors includes: Use a 3D laser scanner to acquire geometric point cloud data of the installation environment; Use thermal imagers to acquire thermal distribution data of the environment; Use an RGB camera to acquire visual image data of the environment; By fusing geometric point cloud data, thermal distribution data, and visual image data using a multi-source point cloud registration algorithm, a centimeter-level precision 3D environmental model with semantic and temperature information is generated.

[0008] According to some embodiments, in the first method of the present invention, the fusion by multi-source point cloud registration algorithm includes using a generalized iterative nearest point algorithm and fusing the attribute probabilities of different sensor data by using a Bayesian update rule.

[0009] According to some embodiments, in the first method of the present invention, generating a risk field for the negative space of gas accumulation includes: Based on the 3D environmental model, enclosed or semi-enclosed top-facing spaces are identified as high-risk accumulation areas. A risk propagation model based on the fast-moving method was used to simulate the propagation of risk from a high-risk accumulation area outward and to calculate the risk value at each point in the space, where the propagation speed of risk within the wall was set to zero.

[0010] According to some embodiments, in the first method of the present invention, the risk value in the risk propagation model decreases as the propagation distance increases, and the risk value inside the high-risk accumulation zone is set to the highest level.

[0011] According to some embodiments, in the first method of the present invention, generating a repulsive potential field for an ignition source includes: Ignition sources in the environment are identified based on a 3D environmental model, and non-uniform multi-level repulsive potential fields are established for different types of ignition sources, including: For a continuous high-temperature heat source, a slow-decaying exponential function is used to model the potential field, which has a large range and slow decay. For intermittent ignition sources, a fast-decaying squared exponential function is used to model and create a second type of potential field with a small range and fast decay.

[0012] According to some embodiments, in the first method of the present invention, the improved path search algorithm is the A* algorithm, whose cost function is: f(n) = g(n) + h(n) + λ·R(n), where: g(n) is the economic cost from the starting point to node n; h(n) is the estimated cost from node n to the target; R(n) is the cumulative risk cost from the starting point to node n; λ is a risk aversion coefficient that is greater than zero. It is configured according to the security strategy. A higher λ value means that the solution prioritizes risk avoidance.

[0013] According to some embodiments, in the first method of the present invention, the economic cost g(n) includes the cost of pipe length and the cost of number of bends.

[0014] According to some embodiments, in the first method of the present invention, the output safety installation scheme in the scheme output step includes a three-dimensional pipeline model and risk map data that can be read by augmented reality devices, for guiding installation.

[0015] According to a second aspect of the present invention, a gas pipeline safety installation scheme generation system is provided for implementing the method as described in the first aspect of the present invention, the system comprising: The data acquisition module is configured to collect multimodal data of the gas pipeline installation environment through multiple sensors; The risk map construction module communicates with the data acquisition module and is configured to generate a risk field for the negative space of gas accumulation and a repulsive potential field for the ignition source based on multimodal data and a risk quantification model, and then merge the two to generate a static risk map. The route planning module communicates with the risk map construction module and is configured to plan pipeline laying routes on a static risk map using an improved route search algorithm, wherein the cost function of the improved route search algorithm includes a cumulative risk cost term. The solution output module communicates with the path planning module and is configured to output a safe installation solution that includes the pipeline laying path.

[0016] The technical advantages of this invention are reflected in the following aspects: A mapping from physical space to risk space was created: the regulatory requirements for avoiding accumulation areas and ignition sources during gas installation were transformed into a risk field that can be understood and calculated by computers through mathematical models, thus realizing the digitization and quantification of safety rules.

[0017] In terms of model design: Negative Space Risk Propagation Model: This model, based on fluid diffusion theory, is used for static risk prediction and innovatively identifies non-intuitive gas accumulation risk areas.

[0018] Non-uniform multi-stage ignition source potential field model: Breaking through the simple concept of safety distance, it establishes a dynamic potential field model that better fits the physical risk characteristics of different types of ignition sources.

[0019] Compared with existing solutions, the solution of the present invention has the following advantages: 1. This solution addresses the lack of quantitative perception of negative space risks in traditional designs. Traditional designs rely on experience to estimate gas accumulation risks, while this solution uses algorithms to proactively identify, quantify, and visualize these risks. It fundamentally avoids major safety risks and significantly reduces the probability of fires and explosions caused by improper pipeline layout.

[0020] 2. It solves the problem of rigidity in the application of safety distance standards in complex spaces. It upgrades the fixed safety distance to a dynamic, non-uniform risk field, enabling the solution to ensure safety while having greater flexibility and adaptability in complex environments.

[0021] 3. It solves the problem of design quality relying too heavily on the designer's personal experience. By embedding safety knowledge into the algorithm, it automates and standardizes the design process, outputting consistent high-quality solutions and reducing the risk of human error. It frees designers from repetitive measurement and drafting, quickly generating multiple optimized solutions while ensuring consistent output quality across different projects and designers.

[0022] 4. Optimize resource allocation and create digital assets: By balancing security and economy (materials, routes), the system helps select the most cost-effective installation solution while ensuring safety. The generated 3D models and installation plans with risk annotations are reusable digital assets, providing a precise data foundation for subsequent construction, operation and maintenance, and renovation. Attached Figure Description

[0023] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without exceeding the scope of protection claimed by the present invention.

[0024] Figure 1 This is a flowchart illustrating an embodiment 1000 of the method for generating a safe installation scheme for gas pipelines according to the present invention. Figure 2 for Figure 1 A flowchart illustrating step S1 in embodiment 1000; Figure 3 for Figure 1A schematic diagram of the risk field generation step S2A in step S2 of embodiment 1000; Figure 4 for Figure 1 A schematic diagram of the repulsive field generation step S2B in step S2 of embodiment 1000; Figure 5 This is a schematic diagram of an embodiment 2000 of a gas pipeline safety installation scheme generation system according to the present invention. Detailed Implementation

[0025] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0026] Figure 1 This is a schematic flowchart of an embodiment 1000 of the method for generating a safe installation scheme for gas pipelines according to the present invention. Figure 1 As shown, Example 1000 includes steps S1-S4.

[0027] In data acquisition step S1, multimodal data of the gas pipeline installation environment is collected through multiple sensors, specifically including: using a 3D laser scanner to acquire geometric point cloud data of the installation environment; using a thermal imager to acquire thermal distribution data of the environment; using an RGB camera to acquire visual image data of the environment; and using a multi-source point cloud registration algorithm to fuse geometric point cloud data, thermal distribution data and visual image data to generate a centimeter-level precision 3D model of the environment with semantic and temperature information.

[0028] In some specific embodiments, the centimeter-precision 3D environmental model output in step S1 is a 3D voxel field model, where each voxel... A data structure containing: : Boolean value, indicating whether it is occupied by an object.

[0029] : An enumeration type representing the most likely object category. For example, wall, socket.

[0030] confidence: A floating-point number representing the confidence level of the semantic classification above.

[0031] temperature: A floating-point number representing the temperature value at this location.

[0032] In step S2 of risk map construction, based on multimodal data, a risk field for the negative space of gas accumulation and a repulsive potential field for the ignition source are generated through a risk quantification model, and the two are merged to generate a static risk map.

[0033] Optionally, the process of generating the risk field for the negative space where gas accumulates in step S2 includes: identifying enclosed or semi-enclosed upward-facing spaces as high-risk accumulation zones based on a three-dimensional environmental model; simulating the propagation of risk from the high-risk accumulation zone outward using a risk propagation model based on the fast-moving method, and calculating the risk value at each point in the space, wherein the propagation speed of risk within the walls is set to zero. Optionally, the risk value in the risk propagation model decreases with increasing propagation distance, and the risk value within the high-risk accumulation zone is set to the highest level.

[0034] Optionally, the process of generating the repulsive potential field for the ignition source in step S2 includes: identifying the ignition source in the environment based on the three-dimensional environmental model; establishing non-uniform multi-level repulsive potential fields for different types of ignition sources: for continuous high-temperature heat sources, a slow decay exponential function is used to model and create a first-type potential field with a large range and slow decay; for intermittent ignition sources, a fast decay square exponential function is used to model and create a second-type potential field with a small range and fast decay.

[0035] In some specific embodiments, step S2 yields two fields covering the entire space, namely: a risk field that quantifies the risk of gas accumulation. And the repulsive potential field that quantifies the combined repulsive effect of all ignition sources. .

[0036] Optionally, in step S2, the fusion of the risk field and the repulsive potential field follows the "weakest link" principle as the core safety rule, meaning that the overall safety of a location depends on its most dangerous factor. If any single risk source poses a high risk at a location, then that location should be classified as a high-risk area. Pipeline planning must prioritize avoiding all single high-risk factors.

[0037] Optionally, during the fusion of the risk field and the repulsive potential field: gas accumulation risk field It is a three-dimensional array, where each voxel stores a floating-point number in the range [0, 1]. The higher the value, the greater the risk of gas accumulation due to geometric structure; ignition source repulsive potential field. It is also a three-dimensional array, where each voxel stores a floating-point number in the range [0,1]. The higher the value, the greater the risk caused by proximity to the ignition source. It should be noted that in step S2 of this invention... It is already the potential field of all individual ignition sources. The result after taking the maximum value at the corresponding position.

[0038] In step S2, the fusion process is performed in parallel on each voxel in three-dimensional space. For any voxel position (x, y, z) in space: First, data is read: the algorithm simultaneously reads the risk value of this location in both input fields. ; ; Then, a decision-making and value assignment process is performed to determine the final comprehensive risk value of the voxel as follows: .

[0039] In the path planning step S3, pipeline laying path planning is performed on the static risk map using an improved path search algorithm, wherein the cost function of the improved path search algorithm includes a cumulative risk cost term.

[0040] Optionally, in step S3, the basic settings in the path planning process include: defining the gas main access point or riser location as the starting point S, and the gas interface location of the stove or water heater as the ending point G; following the constraint that the pipeline must be laid along the wall or ceiling, conforming to the engineering specifications of horizontal and vertical alignment; simplifying the three-dimensional space into a graph structure G=(V, E), where node V represents the possible path points of the pipeline, usually set as regular grid points on the wall; and edge E represents the pipeline segment of adjacent nodes, with specific cost weights.

[0041] Optionally, in some specific embodiments, the improved path search algorithm is the A* algorithm, whose cost function is: f(n) = g(n) + h(n) + λ·R(n).

[0042] Where g(n) is the economic cost from the starting point S to node n, representing the actual resource consumption from the starting point S to the current node n. Optionally, the economic cost g(n) includes the pipeline length cost and the cost of the number of bends. The pipeline length cost is calculated in meters from the physical length of the path, and the cost of the number of bends is calculated by adding a fixed penalty value each time the path changes direction. This is to quantify the increased material costs, installation complexity, and pressure drop caused by elbows.

[0043] Optionally, the economic cost is achieved through ,in, This represents the Euclidean distance from the parent node to the current node. This is a direction change indicator; it is 1 when the direction changes and 0 otherwise.

[0044] h(n) is the estimated cost from node n to the target, representing the estimated minimum cost from the current node n to the destination G, providing the search direction. Optionally, it can be calculated using Euclidean distance, as shown in the formula: .

[0045] R(n) is the cumulative risk cost from the starting point S to node n, which is obtained by integrating the risk density along the path: .

[0046] λ is a risk aversion coefficient greater than zero, configured according to the security strategy. A higher λ value indicates that the solution prioritizes risk avoidance. Optionally, the value of λ can be set as follows: For high-risk scenarios, set the value of λ to be much greater than 1, where safety is of paramount importance. For example, it can be set to 10.0. For space-constrained scenarios, set a moderate value of λ to balance the considerations. For example, it can be set to 2.0-5.0.

[0047] In some specific embodiments, in step S3, the algorithm execution flow for pipeline laying path planning includes: (1) Initialization: Create an open set consisting of nodes to be examined and a closed set consisting of nodes already examined; add the starting point S to the open set and set: g(S) = 0, R(S) = 0, f(S) = h(s) + λ × R(S); (2) Main loop: Node selection: Extract the node n with the smallest f(n) from the open set; Object detection: If n = G, backtrack the construction path and terminate; Neighbor expansion: Generate all valid neighbor nodes of n; Cost update: For each neighbor node m, compute the temporary cost. If m is not in the open set or the new cost is better, update the node information.

[0048] (3) Termination conditions: Success: Target point G has been reached; the complete path has been constructed by backtracking the parent node pointer. Failure: An empty open set indicates that there is no feasible path between the start and end points.

[0049] Optionally, the path generated by the path search algorithm in step S3 in the discrete grid may have unnecessary jagged edges. Therefore, post-processing and optimization of the path are required. Spline curve interpolation or the Douglas-Peucker algorithm can be applied to generate a smooth and natural pipeline route while keeping away from high-risk areas.

[0050] Optionally, the generated path also needs to undergo engineering specification compliance checks, including: vertical / horizontal alignment to ensure that the path segment is parallel to the building axis; bracket location planning, marking bracket installation points at the midpoint of straight sections and near bends; and maintenance space reservation to ensure that components such as valves and joints are in operable positions.

[0051] Step S4 outputs a safe installation plan that includes the pipeline laying path. Optionally, step S4 may include a 3D pipeline model and risk map data that can be read by augmented reality devices to guide the installation.

[0052] In some specific embodiments, in step S4, the optimal path sequence from step S3, the static risk map from step S2, and the 3D environmental model from step S1 are integrated. The preprocessing process includes path voxelization and component identification and parameterization, specifically: Path voxelization: The continuous path sequence is transformed into a discrete set of voxels connected in a three-dimensional voxel grid, which is then used for accurate spatial relationship calculations using risk maps and environmental models.

[0053] Component identification and parameterization: Automatic segmentation: Automatically identifies the positions of straight pipe sections and bends based on changes in path direction; Component labeling: Assigns engineering attributes to each path segment, such as pipe length, bend angle, and start / end coordinates; Support position calculation: Automatically calculates the three-dimensional coordinates of support installation points on the path according to engineering specifications.

[0054] In some specific embodiments, the installation scheme package in step S4 is a structured data package, including: (1) Core geometric model 3D Model of Piping System: Based on path voxels and component annotations, a 3D mesh model of pipes, elbows, valves, etc. is generated. This model accurately reflects the shape, size, and spatial pose of all components.

[0055] Format: Output is an industry-standard GL Transmission Format (gITF) file. This format is lightweight, efficient, and natively supported by the vast majority of 3D engines and AR platforms, and can fully preserve the model's geometry, materials, and hierarchical information.

[0056] (2) Risk context data: Risk Layer: Extracts installation-related risk data from the static risk map. To balance AR device performance, a simplified risk hotspot layer is generated instead of loading the entire 3D risk field.

[0057] Implementation: This layer is a transparent texture, overlaid in the AR view, and high-risk areas are highlighted with a set color using semi-transparent highlighting.

[0058] Data association: In the solution data package, each segment of the pipeline model is associated with the risk value it traverses through metadata, forming a queryable pipeline-risk correspondence.

[0059] (3) Construction guidance information: Bill of Materials: Automatically generates an accurate bill of materials, listing the length, specifications, and quantity of the required pipes, as well as the model and quantity of all accessories such as elbows, supports, valves, and sealing materials.

[0060] Installation procedure cards: These break down the installation process into standardized, step-by-step work instructions.

[0061] For example, one specific embodiment of the installation process card includes: Step 1: Install the starting fixing bracket at coordinates (X1, Y1, Z1).

[0062] Step 2: Lay a straight pipe section of length L1 to coordinates (X2, Y2, Z2).

[0063] Step 3: Install the 90° elbow... Key warnings: Prominently display signs next to specific steps such as "This section crosses a medium-risk area; all welds require secondary flaw detection" or "This bend is only 20cm from the power socket; ensure insulation."

[0064] In some specific embodiments, step S4 includes the production process of an augmented reality visualization scheme, which is a key implementation of what you see is what you get, and the output is a customized visualization dataset for AR. The process includes: (1) Preparation of virtual and real registration data: Feature point map: Extracting a set of unique, high-contrast visual feature points and their 3D coordinates from the 3D environmental model. These points will serve as anchor points for AR devices to identify and locate on the construction site, ensuring that the virtual model can stably correspond to the real world.

[0065] (2) Multi-layered AR scene construction: Basic guide layer: Displays a semi-transparent blue model of pipes and supports, directly superimposed on the real environment, indicating where they are installed.

[0066] Risk warning layer: Based on the risk layer, the areas that need to be avoided are highlighted in red on the walls, ceilings, etc. in the real environment, indicating why they should be decorated this way.

[0067] Information labeling layer: When the installer looks at or selects a virtual pipe by gesture, a floating information card pops up, displaying the length, specifications, and construction precautions of that section.

[0068] Optionally, step S4 will ultimately output a complete smart installation solution package, which includes: Printable installation drawings and reports: used for traditional construction filing and review; Machine-readable BOM and process documents: can be directly imported into the enterprise's material management and project scheduling system; AR visualization data package: contains gITF model, risk hotspot layer, feature point map, which can be loaded into the AR glasses or tablet of the installer via cloud or local.

[0069] Figure 2 for Figure 1 A flowchart illustrating step S1 in embodiment 1000. (See attached diagram.) Figure 2 As shown, step S1 includes steps S11-S14.

[0070] In step S11, a 3D laser scanner is used to acquire geometric point cloud data of the installation environment. Specifically, in step S11, an operator uses a handheld or tripod-mounted 3D laser scanner to perform a panoramic scan of the installation environment, such as the kitchen. The scanner emits laser beams in all directions and receives signals reflected from the surfaces of objects. By calculating the laser flight time or phase difference, the three-dimensional coordinates (X, Y, Z) of countless points are precisely measured, thereby forming a point cloud that characterizes the geometry of the environment. The point cloud data constitutes the geometric skeleton of the entire three-dimensional model of the environment.

[0071] Optionally, the three-dimensional point cloud determined in step S11 is a set of n points, each point It is a geometric point cloud data containing three-dimensional Cartesian coordinates (x, y, z) and laser echo intensity.

[0072] In step S12, a thermal imager is used to acquire thermal distribution data of the environment. Specifically, in step S12, the thermal imager generates a grayscale or pseudo-color image corresponding to the temperature distribution by detecting infrared radiation emitted from the surface of an object. The value of each pixel in the image represents the temperature information at that location. Step S12 is used to accurately identify potential ignition sources and hidden hot water pipes, etc. These targets may not be conspicuous in a normal RGB image, but they will stand out in thermal imaging due to their temperature difference with the surrounding environment. Optionally, potential ignition sources include operating electrical compressors and hot power sockets.

[0073] Optionally, in step S12, a two-dimensional raster data consisting of M1 x N1 pixels is determined as the heat distribution data. Each pixel It includes the row and column coordinates (u,v) of the pixel in the image and the apparent temperature value of the object surface corresponding to the pixel.

[0074] In step S13, visual image data of the environment is acquired using an RGB camera. Specifically, in step S13, a high-resolution RGB camera captures color photographs of the environment from multiple angles. These images record the texture, color, and visual features of the environment. The core purpose is for subsequent semantic understanding, that is, to identify specific objects in the images, such as walls, windows, cabinets, power sockets, gas stoves, etc., through computer vision algorithms, and assign them clear semantic labels.

[0075] Optionally, the visual image data obtained in step S13 is two-dimensional raster data composed of M2 x N2 pixels. Each pixel Includes pixel coordinates ( The values ​​of the red, green, and blue channels (R, G, B) are also included.

[0076] In step S14, geometric point cloud data, thermal distribution data and visual image data are fused using a multi-source point cloud registration algorithm to generate a centimeter-level precision 3D environmental model with semantic and temperature information.

[0077] Optionally, in step S14, coordinate system unification and spatial alignment are first performed to accurately attach the data acquired by the two-dimensional sensors, such as the thermal imaging and RGB camera, to the geometric skeleton formed by the three-dimensional laser point cloud. The specific implementation of this process includes: Sensor calibration: This is a preliminary and crucial step. The rigid body transformation matrices of the laser scanner, RGB camera, and thermal imager relative to a common reference frame are calculated beforehand through experiments. This common reference frame is typically the optical center of the RGB camera, and the rigid body transformation matrices include rotation and translation parameters.

[0078] 2D to 3D mapping: For a pixel (u, v) in a thermal image or RGB image, using the camera's intrinsic parameter matrix and the aforementioned extrinsic parameter matrix (i.e., rigid body transformation matrix), it can be back-projected into 3D space through a pinhole camera model, forming a ray. This ray intersects with the surface of the object represented by the 3D point cloud, thereby assigning the pixel's color or temperature value to the nearest 3D point. This process is called data binding.

[0079] Optionally, in step S14, after unifying the coordinate system and aligning the space, voxelization in three-dimensional space is performed, followed by information injection and semantic annotation.

[0080] In this process, 3D spatial voxelization involves dividing the entire 3D space into a regular 3D mesh, with each small cube called a voxel. The size of the voxel determines the model's accuracy; for example, it might be set to 1cm x 1cm x 1cm. All subsequent processing is then performed on this voxel mesh.

[0081] Optionally, the information injection and semantic annotation process in step S14 includes: Geometric Occupation: If a voxel contains one or more 3D laser points, it is marked as occupied; Temperature assignment: Through the aforementioned coordinate mapping, the temperature value of the thermal imaging pixel is assigned to the corresponding voxel; Semantic recognition and assignment: This is the most complex step. A pre-trained fully convolutional neural network is used to perform pixel-level semantic segmentation on the RGB image. The network output is a label map of the same size as the input image, with each pixel labeled as a category such as wall, socket, cabinet, window, etc. Then, these semantic labels are injected into the corresponding 3D voxels through coordinate mapping.

[0082] Finally, step S14 also includes a probability fusion and uncertainty management process. Due to sensor noise and recognition errors, semantic information is uncertain. The confidence level of each voxel V for a given semantic label L is fused using a Bayesian update, with the following formula: ; Where P(L) is the prior probability, i.e., the initial belief that the voxel is an ignition source; P(E|L) is the likelihood probability, i.e., if the voxel is indeed an ignition source, what is the probability that the RGB camera will identify it as a socket? This value comes from the confusion matrix or statistics of the deep learning model on the test set; E is the new evidence, i.e., the current identification event; and P(L|E) is the posterior probability, i.e., the latest and more accurate probability that we believe the voxel is an ignition source after incorporating the new evidence. If subsequent thermal imaging also provides evidence that the voxel is heating up, a new round of Bayesian updates can be performed again using the current posterior probability as the prior, thereby incorporating multi-source information and continuously improving the confidence of the judgment.

[0083] Finally, the centimeter-precision 3D environmental model output in step S14 is a 3D voxel field model, where each voxel... A data structure containing: : Boolean value, indicating whether it is occupied by an object. An enumeration type representing the most likely object category. For example, a wall or an electrical outlet.

[0084] confidence: A floating-point number representing the confidence level of the semantic classification above.

[0085] temperature: A floating-point number representing the temperature value at this location.

[0086] The model determined in step S14 is no longer a simple geometric point cloud, but an environmental digital model that carries dense information and can be directly used by computers for spatial reasoning and analysis, laying a solid data foundation for subsequent risk calculation.

[0087] Figure 3 for Figure 1 A schematic diagram of the risk field generation step S2A in step S2 of embodiment 1000. (See attached diagram.) Figure 3 As shown, step S2A includes steps SA1-SA2.

[0088] In step SA1, based on the three-dimensional environmental model, enclosed or semi-enclosed top-facing spaces are identified as high-risk accumulation areas.

[0089] Optionally, the core judgment logic for determining candidate areas of high-risk accumulation zones in step SA1 is as follows: (1) Top closed or semi-closed: The area above is blocked by other objects (such as ceiling, cabinet top panel), which prevents the gas from escaping upward.

[0090] (2) Upward opening: The bottom or side of the area has an opening that allows gas lighter than air to enter and be trapped inside; (3) Top closed or semi-closed: The area above is blocked by other objects, which prevents the gas from escaping upward.

[0091] (4) Upward opening: The bottom or side of the area has an opening that allows gas lighter than air to enter and be trapped inside.

[0092] Specifically, in step SA1, the process of screening candidate regions based on the 3D environmental model determined in step S1 includes: First, all voxels are traversed, and all voxels with semantic labels of CEILING or CABINET_TOP are initially selected as potential trap bases.

[0093] Perform geometric relationship determination - top closure check: for each candidate voxel The algorithm performs a raycast directly above the point in the opposite direction of gravity. This ray travels through a series of voxels until it hits a voxel semantically labeled CEILING or ROOF. If no such absolute top is encountered along the entire ray path, a decision is made. The space above it is not completely enclosed and does not meet the conditions for accumulation.

[0094] Geometric relationship determination - upward opening and cavity inspection: In Examine whether there exists a continuous space with a certain vertical height directly below the direction of gravity, whose voxels... The value is false, indicating an empty cavity. This cavity must be connected to... The horizontal planes are connected, thus simulating the channels through which gas can enter.

[0095] Data storage and output: All data that simultaneously satisfies the conditions of being closed at the top and having a cavity at the bottom. The voxel is labeled HighRiskAccumulationSource. This label is stored as a new boolean attribute is_accumulation_source in the voxel data structure, and its coordinate set {S} is used as the source point set for the next step of risk propagation.

[0096] Optionally, in some specific embodiments, typical areas that meet the criteria for high-risk accumulation zones include: the interior space of a suspended ceiling, the gap between a cabinet and the ceiling, the top of a U-shaped cabinet, etc.

[0097] In step SA2, a risk propagation model based on the rapid travel method is used to simulate the outward propagation of risk from a high-risk accumulation zone, and the risk value at each point in the space is calculated. The propagation speed of risk within the wall is set to zero. Optionally, the risk value in the risk propagation model decreases with increasing propagation distance, and the risk value within the high-risk accumulation zone is set to the highest level. The rapid travel method is used to simulate the expansion of wavefronts, such as flames or sound; in this invention, it is innovatively applied to simulate the propagation of gas risk.

[0098] Optionally, the implementation process of the risk propagation model includes: Initialization: Create a new floating-point array for the entire 3D voxel field. Initialize to 0. For all source voxels where is_accumulation_source is True. ,set up (s) = 1.0.

[0099] Precise setting of the propagation speed function V(i,j,k): This is a logical process based on voxel properties: in a wall, propagation stops completely; in air or other spaces, it is set to a base speed value, such as 1.0.

[0100] Propagation Process: The fast-moving algorithm starts from the initial accepted region (i.e., the source set) and gradually pushes outwards into the narrow band. At each step, the algorithm selects the voxel with the smallest arrival time T from the narrow band, marks it as accepted, and then updates the T values ​​of all its unaccepted neighbors. Neighbor voxels The arrival time is updated by solving the Eikonal equation, which is composed of the T values ​​of its neighboring accepted voxels.

[0101] Precise setting of the propagation speed function V(i,j,k): This is a logical process based on voxel properties: in a wall, propagation stops completely; in air or other spaces, it is set to a base speed value, such as 1.0.

[0102] Propagation Process: The fast-moving algorithm starts from the initial accepted region (i.e., the source set) and gradually pushes outwards into the narrow band. At each step, the algorithm selects the voxel with the smallest arrival time T from the narrow band, marks it as accepted, and then updates the T values ​​of all its unaccepted neighbors. Neighbor voxels The arrival time is updated by solving the Eikonal equation, which is composed of the T values ​​of its neighboring accepted voxels.

[0103] According to such Figure 3 The embodiment shown in this invention transforms the abstract risk of gas accumulation into a continuous and calculable risk field in three-dimensional space, enabling the systematic, quantitative identification and spatial representation of hidden safety hazards. This transforms traditional qualitative risk assessments, which rely on designer experience, into objective calculations driven by a rapid-motion model based on environmental geometry. Its beneficial effect lies in automatically and comprehensively locating all potential gas accumulations and marking them as high-cost areas during the scheme generation stage. This proactively and forcefully avoids the most fatal risk caused by improper pipeline layout—combustible gas accumulation—at the design source, significantly improving the inherent safety level of the design.

[0104] Figure 4 for Figure 1 A schematic flowchart of the repulsive potential field generation step S2B in step S2 of embodiment 1000. (See attached diagram.) Figure 4 As shown, step S2B includes steps SB1-SB2.

[0105] In step SB1, ignition sources in the environment are identified based on the 3D environmental model. Optionally, the identification and classification of ignition sources in step SB1 are based on the voxel category labels in the 3D environmental model. And the temperature value.

[0106] Optionally, in step SB1, the algorithm first finds all A voxel belonging to a predefined set of potential ignition sources, which includes SOCKET, SWITCH, WATER_HEATER, and REFRIGERATOR.

[0107] For each candidate voxel, calculate its overall confidence level as an ignition source. This is a fusion process based on rules and sensor data: Base confidence level: assigned based on semantic labels. For example, SOCKET is 0.7 and WATER_HEATER is 0.9.

[0108] Confidence adjustment: If the voxel's temperature is significantly higher than the ambient average temperature, its confidence level is increased using Bayesian rules. For example, the new confidence level: ; in, The probability that a real ignition source will generate heat is usually very high, for example, it can be 0.95; The probability of heating from a non-ignition source is usually very low, for example, it can be 0.1.

[0109] Specifically, a confidence threshold is set, and candidate voxels exceeding the threshold are identified as ignition sources and then classified in the final stage based on their semantic attributes and temperature characteristics. Continuous high-temperature heat source: semantically tagged with WATER_HEATER, STOVE, or its temperature remains high; Intermittent ignition source: semantically tagged as SOCKET, SWITCH, and with no significant temperature anomalies.

[0110] In step SB2, a non-uniform multi-level repulsive potential field is established for different types of ignition sources, including a first-type potential field and a second-type potential field. For each classified ignition source j, its potential field... The calculations are performed independently.

[0111] Optionally, in step SB2, the process of calculating the potential field of the ignition source j includes: Distance calculation: For any point x in space (usually referring to the center coordinates of each voxel), calculate its three-dimensional Euclidean distance to the geometric center of the ignition source j. Potential field assignment: Based on the type of ignition source j, Substitute the corresponding potential function: In step SB2a, for a continuous high-temperature heat source, a slow-decaying exponential function is used to model and create a first-class potential field with a large range and slow decay:

[0112] in, A high value indicates a very high risk from a sustained high-temperature heat source. This represents the distance factor influencing the potential field. In some specific embodiments, Set to 0.9, Setting it to 0.5 ensures that the potential field still has a significant effect 1-2 meters away from the source point. ).

[0113] Step SB2b: For intermittent ignition sources, a fast-decaying squared exponential function is used to model and create a second-type potential field with a small range and rapid decay.

[0114] In some specific embodiments, compared to the first type of potential field, Set to 0.6, Set to 0.2. Due to the squared term, when hour, It has rapidly decreased to near 0.

[0115] According to such Figure 4 The embodiment shown in this invention establishes a non-uniform, multi-level repulsive potential field model, achieving refined and adaptive avoidance of differentiated ignition source risks. It abandons the rigid, one-size-fits-all safety distance approach, instead using a slowly decaying exponential field to accurately characterize the wide-area impact of continuous high-temperature heat sources, and a rapidly decaying square exponential field to describe the local risks of intermittent ignition sources. Its core beneficial effect is that it enables the path planning algorithm to intelligently balance risk and space utilization, thereby generating a pipeline laying scheme with better spatial layout and more economical materials while strictly ensuring safety standards, thus resolving the inherent contradiction between safety regulations and installation feasibility.

[0116] Figure 5 This is a schematic diagram of an embodiment 2000 of a gas pipeline safety installation scheme generation system according to the present invention. Figure 5 As shown, embodiment 2000 includes a data acquisition module 201, a risk map construction module 202, a route planning module 203, and a solution output module 204.

[0117] Optionally, the data acquisition module 201 is configured to acquire multimodal data of the gas pipeline installation environment through a variety of sensors.

[0118] Optionally, the risk map construction module 202 is communicatively connected to the data acquisition module and configured to generate a risk field for the negative space of gas accumulation and a repulsive potential field for the ignition source based on multimodal data and a risk quantification model, and then merge the two to generate a static risk map.

[0119] Optionally, the route planning module 203 is communicatively connected to the risk map construction module and configured to perform pipeline laying route planning on a static risk map using an improved route search algorithm, wherein the cost function of the improved route search algorithm includes a cumulative risk cost term.

[0120] Optionally, the solution output module 204 is communicatively connected to the path planning module and configured to output a safe installation solution that includes the pipeline laying path.

[0121] The embodiments of the present invention have been described in detail above. Specific examples have been used to illustrate the principles and implementation methods of the present invention. The descriptions of the embodiments above are only for the purpose of helping to understand the method and core ideas of the present invention. Furthermore, any changes or modifications made by those skilled in the art based on the ideas of the present invention, its specific implementation methods, and its application scope, are all within the scope of protection of the present invention. Therefore, the content of this specification should not be construed as a limitation of the present invention.

Claims

1. A method for generating a safe installation scheme for gas pipelines, characterized in that, Includes the following steps: S1. Data Acquisition Steps: Collect multimodal data on the gas pipeline installation environment using multiple sensors; S2. Risk map construction steps: Based on the multimodal data, a risk field for the negative space of gas accumulation and a repulsive potential field for the ignition source are generated respectively through the risk quantification model, and the two are merged to generate a static risk map; S3. Path planning step: On the static risk map, the pipeline laying path is planned using an improved path search algorithm, wherein the cost function of the improved path search algorithm includes a cumulative risk cost term; S4. Solution Output Steps: Output a safe installation solution that includes the pipeline laying path.

2. The method according to claim 1, characterized in that, In step S1, data acquisition, the acquisition of multimodal data through multiple sensors includes: Use a 3D laser scanner to acquire geometric point cloud data of the installation environment; Use thermal imagers to acquire thermal distribution data of the environment; Use an RGB camera to acquire visual image data of the environment; The geometric point cloud data, the thermal distribution data, and the visual image data are fused using a multi-source point cloud registration algorithm to generate a centimeter-level precision 3D environmental model with semantic and temperature information.

3. The method according to claim 2, characterized in that, The fusion through multi-source point cloud registration algorithm includes using the generalized iterative nearest point algorithm and fusing the attribute probabilities of different sensor data through Bayesian update rules.

4. The method according to claim 2, characterized in that, The generation of the risk field for the negative space of gas accumulation includes: Based on the three-dimensional environmental model, enclosed or semi-enclosed top-facing spaces are identified as high-risk accumulation areas. A risk propagation model based on the fast-moving method is used to simulate the outward propagation of risk from the high-risk accumulation area and calculate the risk value at each point in the space, wherein the propagation speed of risk in the wall is set to zero.

5. The method according to claim 4, characterized in that, In the risk propagation model, the risk value decreases as the propagation distance increases, and the risk value within the high-risk accumulation zone is set to the highest level.

6. The method according to claim 1, characterized in that, The generation of the repulsive potential field against the ignition source includes: Based on the three-dimensional environmental model, ignition sources in the environment are identified, and non-uniform multi-level repulsive potential fields are established for different types of ignition sources, including: For a continuous high-temperature heat source, a slow-decaying exponential function is used to model the potential field, which has a large range and slow decay. For intermittent ignition sources, a fast-decaying squared exponential function is used to model and create a second type of potential field with a small range and fast decay.

7. The method according to claim 1, characterized in that, The improved path search algorithm is the A* algorithm, and its cost function is: f(n) = g(n) + h(n) + λ·R(n), where: g(n) is the economic cost from the starting point to node n; h(n) is the estimated cost from node n to the target; R(n) is the cumulative risk cost from the starting point to node n; λ is a risk aversion coefficient that is greater than zero. It is configured according to the security strategy. A higher λ value means that the solution prioritizes risk avoidance.

8. The method according to claim 7, characterized in that, The economic cost g(n) includes the cost of pipeline length and the cost of the number of bends.

9. The method according to claim 1, characterized in that, In the output step of the solution, the output safe installation solution includes a three-dimensional pipeline model and risk map data that can be read by augmented reality devices, which are used to guide the installation.

10. A system for generating a safe installation scheme for gas pipelines, characterized in that, The system for implementing the method of claim 1 includes: The data acquisition module is configured to collect multimodal data of the gas pipeline installation environment through multiple sensors; The risk map construction module is communicatively connected to the data acquisition module and is configured to generate a risk field for the negative space of gas accumulation and a repulsive potential field for the ignition source based on the multimodal data and through a risk quantification model, and then merge the two to generate a static risk map. The route planning module is communicatively connected to the risk map construction module and is configured to perform pipeline laying route planning on the static risk map using an improved route search algorithm, wherein the cost function of the improved route search algorithm includes a cumulative risk cost term. The scheme output module is communicatively connected to the path planning module and is configured to output a safe installation scheme that includes the pipeline laying path.

Images