Sewage treatment plant intelligent layout method based on genetic optimization algorithm

Abstract

The invention discloses a sewage treatment plant intelligent layout method based on a genetic optimization algorithm, and the method comprises the steps: enabling a system to obtain parameters through the automatic importing and sorting of multi-source data, and completing the machine learning preprocessing; mapping the functional characteristics of the monomers and the process dependency relationship into a map structure, and realizing dynamic grouping according to a reasoning result; a multi-target adaptive algorithm and interactive visualization are adopted, and automatic arrangement of buildings and pipelines is achieved in a local range; during overall layout, an evolutionary genetic algorithm and a self-adaptive reward and punishment mechanism are introduced, and parameters of each group are globally optimized; determining road details based on external traffic simulation and future capacity expansion requirements; performing comprehensive optimization in a whole plant pipeline arrangement stage; updating the design scheme; and outputting a plurality of achievement forms. According to the method, the modification requirement can be quickly met, and the algorithm can be restarted and a new scheme can be efficiently generated only by updating parameters no matter the size of a single body, the land condition or the technological process is changed.

Description

Technical Field

[0001] The present invention relates to the technical field of environmental engineering, and particularly to an intelligent layout method for sewage treatment plants based on a genetic optimization algorithm. Background Art

[0002] In the overall layout of a sewage treatment plant, in addition to taking into account the plane dimensions and functional connection of buildings and structures themselves, it is also necessary to connect the upstream and downstream process flows to ensure the smoothness and safety of the water flow path and pipelines for media such as sludge and gas. In different project practices, on the one hand, the site red line is often irregular, resulting in extremely variable available spaces between buildings; on the other hand, the elevation difference, geological conditions, and meteorological factors (such as wind direction, annual rainfall, etc.) within the plant area will also have an actual impact on the route selection of the primary and secondary roads and the pipeline orientation. Since different types of single units have their own requirements for adjacent placement, spatial orientation, and the quantity and diameter of connecting pipelines, and multiple-media pipelines also need to connect multiple structures in series with as few turns and the shortest path as possible while avoiding mutual conflicts or occupying road space, in the conventional design method, engineers usually rely on personal project experience to first place the main single units on the general layout plan and gradually determine the positions of the secondary single units based on this layout. If it is found during this process that there is a conflict between the terrain data and the process requirements, it is necessary to adjust the arrangement of single units and the pipeline orientation again and re-evaluate the feasibility of the road line position, the requirement for a fire protection circular road, and the inspection route; if additional conditions such as seasonal wind direction, gas emission diffusion, and the impact on the office area and living area are considered, it is often necessary to make repeated trade-offs and modifications. Especially when the treatment capacity or process requirements change, the original plan may completely fail and need to be scrapped and restarted. Due to the large number of single units, complex pipeline media, and the need to achieve the optimal or sub-optimal comprehensive layout goal under external constraints such as elevation, wind direction, and site shape, the current practice that mostly relies on experience and local trial and error often consumes a large amount of manpower and time during the design process. Once the global rebalancing cannot be carried out in a timely and accurate manner according to the changing conditions, many problems will occur during the construction or operation and maintenance stages. Summary of the Invention

[0003] To overcome the deficiencies of the prior art, the present invention proposes an intelligent layout method for sewage treatment plants based on a genetic optimization algorithm, which significantly improves the overall design efficiency. By automatically searching for layout schemes through machine learning and genetic algorithms, a large number of manual trial and error links are reduced.

[0004] To achieve the above object, the method for accelerating large model inference based on an encapsulation engine in the present invention includes the following steps: Step 1: Through the automatic import and collation of multi-source data such as BIM models, laser scans, drone aerial photography, and sensors, the system obtains the basic information of buildings / structures, the terrain environment, and real-time parameters, and completes the machine learning preprocessing; Step 2: Combine the AAO process with the knowledge graph, map the functional characteristics of monomers and process dependencies into the graph structure, and achieve dynamic grouping according to the inference results; Step 3: For each group, adopt a multi-objective adaptive algorithm and interactive visualization to balance multiple requirements such as minimum floor area, process connection, and maintenance safety distance, and achieve automatic layout of buildings and pipelines within a local range; Step 4: When making the overall layout, introduce an evolutionary genetic algorithm and an adaptive reward and punishment mechanism to globally optimize parameters such as translation, rotation, and angle of each group, and dynamically adjust the evaluation weights according to the actual layout performance; Step 5: Based on external traffic simulation and future expansion requirements, automatically generate the primary and secondary road networks and determine details such as road widths, intersection designs, and reserved connection ports to meet the requirements of internal traffic and convenient operation and maintenance; Step 6: In the stage of pipeline layout for the whole plant, through pipeline fluid simulation and BIM collision detection, adopt a multi-level automatic path finding algorithm to comprehensively optimize for the shortest path, fewest turns, and conflict avoidance; Step 7: Use the reinforcement learning mechanism to iteratively update the design scheme in the scheme evaluation and trial-and-error loop, and at the same time allow engineers to make real-time corrections and feedback on the local layout through the interface; Step 8: Finally, output various result forms including VR / AR roaming visualization, automatically generated design reports, and API interfaces that can be docked with the intelligent operation and maintenance system.

[0005] Furthermore, Step 1 is specifically as follows: Step 1.1: Obtain the geometric information, interface positions, material properties, etc. of each building / structure by using the output of a BIM model (such as Revit, Tekla, etc.) or an IFC file; use a laser scanning device or drone aerial photography technology to obtain data on site terrain undulations, surface attachments, and the surrounding environment, and convert these data into point clouds or mesh files that can be processed in Rhino; collect real-time parameters (such as water quality, meteorology, traffic flow, etc.) through sensors deployed on site to provide dynamic parameter inputs for subsequent design processes; Step 1.2: Convert the data from sources such as BIM, laser scanning / drone aerial photography, and sensors into compatible formats for the Rhino+Grasshopper environment; according to project needs, perform position calibration and coordinate unification on different types of data (2D drawings, point clouds, terrain meshes, real-time sensor data) to ensure that all information is correctly superimposed within the same coordinate system; Step 1.3: With the aid of machine learning algorithms or image recognition algorithms, screen out noisy, defective, and duplicate geometric objects; merge, correct, or mark incomplete or duplicate monomers to ensure there are no redundant objects during subsequent layout; for outliers in laser scanning or drone aerial survey data, use filtering algorithms for smoothing or elimination processing; Step 1.4: Assign basic attribute labels to each monomer (such as "biochemical pool", "sludge pool", "comprehensive building"), and establish a table in the database related to its dimensions, interface locations, inlet and outlet media, etc.; group the on-site sensor data by time or region to facilitate the calling of parameters such as real-time or historical average values in subsequent layout algorithms.

[0006] Further, Step 2 is specifically as follows: Step 2.1: Sort out the types of monomers required at each stage of the AAO process chain (such as pretreatment, biochemical treatment, advanced treatment, sludge treatment, etc.), and create corresponding nodes in the knowledge graph; for each monomer object (such as "coarse grille", "fine grille", etc.), add attributes such as "function", "inlet and outlet locations", "adjacent equipment" in the graph; establish the dependency relationships between monomers, such as "what is the upstream monomer of the biochemical pool" and "what is the downstream monomer of the secondary sedimentation tank", to form the graph edges of the process flow direction; Step 2.2: In the knowledge graph, mark the attributes of "high correlation" or "preferred adjacency" for monomers with tight process connections or physical adjacency requirements; use the inference engine or simple query rules to automatically retrieve some conventional grouping patterns (for example, "coarse grille + fine grille" appear under a certain pretreatment group node at the same time); in case of special process requirements (such as adding a certain new structure), temporarily add nodes in the graph and run the inference to update the grouping results.

[0007] Step 2.3: If the designer makes manual adjustments to the groups automatically generated by the system based on on-site experience, record these operations in the system log; on the algorithm side, optimize the parameter weights such as "adjacency" and "must-be-adjacent relationship" according to the results of the manual adjustments, providing reference for subsequent similar projects or iterative processes; finally, output a fixed or dynamic grouping list (pretreatment group, secondary treatment group, advanced treatment group, sludge treatment group, accessory group, etc.) for subsequent "block" management in Grasshopper.

[0008] Further, Step 3 is specifically as follows: Step 3.1: For each group, determine the target parameters to be optimized within it, such as floor area, equipment spacing, maintenance passage, safety distance, etc.; set weights for different targets, such as "floor area is the main target and the number of pipeline turns is the secondary target", and establish corresponding scoring or penalty indicators in Grasshopper; Step 3.2: Apply the "point-to-point" chain logic within the group, using the connection points (inlet / outlet or other process interfaces) between monomers as control nodes; in Grasshopper, combine heuristic search or local genetic algorithms to gradually try different monomer arrangement orders, relative distances, and rotation angles; after each arrangement is generated, evaluate it (such as calculating floor area, pipeline length, maintenance safety distance), then retain the high-score combinations and eliminate the low-score combinations; Step 3.3: Use automatic pathfinding algorithms such as Dijkstra or leafvein to search for the shortest or least-angled paths for multi-media pipelines (sewage, sludge, gas, etc.) within the group; check whether the generated pipelines cross the monomer boundaries, and if there are conflicts, regenerate them by moving nodes or adjusting path weights; package and output the finally determined layout within the group and the corresponding pipelines as "sub-blocks" for the subsequent overall layout; Step 3.4: Engineers can directly drag or rotate a certain monomer in the Rhino view, and the indicators such as floor area, distance, and pipeline length are updated in real time through Grasshopper; if it is found that the local optimum leads to an obviously unreasonable layout, manual intervention can be carried out and this operation can be recorded for subsequent algorithm learning; after determining the optimal or sub-optimal local arrangement, encapsulate the results as "grouping results" and then transfer to the overall layout of the next stage.

[0009] Further, Step 4 is specifically as follows: Step 4.1: Abstract each group (including a set of arranged monomers and their internal pipelines) into a higher-level "block", and set adjustable parameters such as translation (X, Y coordinates), rotation angle, and distance from adjacent blocks; generate several initial layout plans (initial population) through sampling methods such as random or Latin square to ensure uniform distribution within the plane range; Step 4.2: Establish an evaluation function according to the requirements of the whole plant, including land red line constraints (penalty for crossing the boundary), adjacency of key processes (penalty for insufficient adjacency or excessive distance), wind direction control (penalty for improper arrangement), road reservation (penalty for impassability), etc.; during the algorithm iteration process, if some unreasonable layouts appear too frequently, the system automatically increases the penalty coefficient in this aspect to form an adaptive reward and punishment mechanism; calculate the comprehensive score (or penalty value) for each layout to guide the selection and crossover links of the genetic algorithm; Step 4.3: Call Galapagos or other customized genetic algorithm plug-ins in the Grasshopper environment to perform basic operations such as selection, crossover, and mutation; in the selection link, give priority to retaining high-score individuals and add a small number of low-score individuals to maintain population diversity; in the crossover link, generate new offspring by combining the parameters of the parent generation; in the mutation link, randomly change the translation or rotation amount of some blocks to jump out of the local optimum; after repeating multiple generations of iteration, retain several high-quality layout plans as the global layout results for selection; Step 4.4: After each iteration, Grasshopper automatically outputs the visual layout of the highest-scoring scheme in the contemporary population, including information such as the positions of individual units, grouping distributions, and the positions of red lines; engineers can view the scoring details of each candidate layout at any time to identify possible logical conflicts or site encroachments; if systematic deviations occur (such as most layouts deviating too much from the land red line), the relevant penalty weights can be increased or the parameter ranges can be restricted and the iteration can be restarted.

[0010] Furthermore, Step 5 is specifically as follows: Step 5.1: Extract information such as the positions of the main groups (or "blocks"), site boundaries, and entrance and exit positions from the aforementioned overall layout scheme to determine the docking points inside the factory area and with external roads; if there are available prediction or simulation models for external traffic, import the road traffic flow data or vehicle driving data into the Grasshopper environment as a reference for subsequent planning. Step 5.2: According to the positions and operation and maintenance needs of the core functional areas or main individual units, delimit at least one main road alignment running through the whole factory; set parameters such as road width and turning radius in Grasshopper, and automatically or semi-automatically generate the center line of the main road that meets the site restrictions; conduct a site encroachment inspection on the preliminarily generated main road, and if there is a situation of crossing the boundary or being too close to key structures, correct it by fine-tuning the road nodes or path weights. Step 5.3: Based on the existing main roads, arrange secondary channels for each functional area or remote individual unit to meet the passing needs of operation and maintenance vehicles or pedestrians; incorporate key nodes such as intersections, parking areas, and transfer and loading areas into the layout parameters, and automatically arrange them in combination with the terrain and the functional positioning of the factory area; for later expansion or renovation requirements, reserve connection ports or retain road corridors in the road network design to avoid large-scale demolition and reconstruction during later construction. Step 5.4: In the Rhino view, display the superposition effect of the primary and secondary road networks and the individual unit layout in real time, and let the engineer check whether there are problems such as difficult vehicle turning, congested nodes, or unfavorable height differences; if traffic bottlenecks or potential safety hazards are detected, correct the road parameters and regenerate the road network; the finally determined road system is regarded as a fixed traffic plan in subsequent pipeline layout and global evaluation.

[0011] Furthermore, Step 6 is specifically as follows: Step 6.1: Read the pipeline interface types (sewage, sludge, gas, chemicals, etc.), pipe diameters, and connection requirements of each individual unit from the previous grouped layout; if there are terrain undulations or different elevation requirements, classify the "gravity flow pipes" and "pressure pipes" separately to establish a hierarchical processing scheme for subsequent automatic path finding; incorporate the newly generated road network into the constraints to facilitate the pipeline layout along the road or pipe gallery area preferentially. Step 6.2: Use algorithms such as Dijkstra, leafvein, or custom multi-level automatic pathfinding algorithms to generate candidate routes with "shortest paths" or "fewest turns" for various types of utility pipelines; when generating pipelines, consider roads, structures, red lines, etc. as obstacles to ensure that the paths do not cross or have unreasonable detours; perform elevation calculations or slope inspections on gravity flow pipelines to avoid situations where self-flow is not possible or additional pumping stations are required; Step 6.3: After generating the preliminary pipeline plan, summarize the results into the BIM environment or use the built-in collision detection components in Grasshopper to check for spatial conflicts between pipelines and between pipelines and individual buildings; if the system detects situations such as pipelines overlapping, the minimum distance not meeting the requirements, or intersecting with buildings, run the pathfinding algorithm again through local parameter adjustments (such as changing the pipeline elevation or rerouting); optionally, perform fluid simulation on key pipelines to evaluate flow velocity, pressure loss, and flow distribution, etc., and correct the pipe diameter or adjust the route if serious unreasonable situations occur; Step 6.4: Output the arrangement results of the whole plant's pipelines and visually display information such as pipeline colors and elevations in Rhino for engineers to conduct intuitive reviews; if engineers find potential difficulties in operation and maintenance (such as difficult access to inspection points) or problems such as high energy consumption caused by excessive turns, they can modify the pipeline weights or add constraints in Grasshopper and solve again; after both conflict problems and functional requirements are met, package the final pipeline layout as a deliverable and enter the next evaluation and revision process.

[0012] Furthermore, Step 7 is specifically as follows: Step 7.1: Integrate the overall layout, roads, and pipeline results generated previously and conduct quantitative evaluations through various indicators (process connection, operation and maintenance convenience, land use efficiency, pipeline energy consumption, etc.); if any indicator does not meet the pre-set standards (such as excessive land occupation, road congestion, or long pipelines), display the problem location or the indicator gap in the evaluation report; make preliminary marks on the identified deficiencies for targeted improvement in the subsequent reinforcement learning session; Step 7.1: Represent the layout plan as a "state" from the perspective of the algorithm, including core variables such as the positions of individual buildings, pipeline routes, and road arrangements; set "immediate rewards" for each evaluation indicator, such as shortening the pipeline length, reducing corners, or lowering construction costs; set negative rewards for serious problems (such as out-of-bounds, serious collisions) to guide the algorithm to actively avoid these situations when attempting new layouts; Step 7.1: Allow the system to perform small-scale parameter perturbations in various dimensions to simulate "trial and error". If improvements are made, the path is strengthened; otherwise, the original solution is promptly converged. Engineers can give "manual confirmation" or "rejection" marks to local corrections automatically generated by the system to help the algorithm continue to learn and build optimization strategies that are more in line with practical experience. After multiple rounds of trial and error, archive the solution with the highest reward and record its decision path to facilitate quick reference in subsequent similar projects.

[0013] Step 7.1: When the reinforcement learning algorithm converges under the specified indicators or meets the project threshold, the final revised plant layout plan, road and pipeline layout are output; the key parameter changes, important decision points and success cases in the entire reinforcement learning process are packaged into a visual report to provide a reference for subsequent technology accumulation and new project reference; the results plan confirmed at this stage enters the final results output and subsequent application integration steps.

[0014] Furthermore, step 8 is as follows: Step 8.1: Export the 2D plan (DWG, PDF) and 3D model (3DM, IFC, FBX, etc.) of the layout results from the Rhino+Grasshopper environment at one time, so that they can be directly referenced on the CAD and BIM platforms; generate professional sub-drawings (such as pipeline network general drawing, road traffic map) for major pipelines and roads, and mark key dimensions, coordinates and bill of materials information in the drawings; Step 8.2: If the project requirements include visual display, import the final 3D model into VR or AR software to allow project participants to immersively view the plant layout and pipeline direction; retain the interface (API or data file) for docking with the digital twin system to provide model and parameter data for the subsequent implementation of remote monitoring, intelligent diagnosis and other functions in the operation and maintenance management platform; Step 8.3: Integrate the key iteration records, scoring trends, advantages and disadvantages of the final solution in the genetic algorithm and reinforcement learning process into a detailed design report; list the important parameters of each stage (such as land use efficiency, road length, total number of pipeline meters, etc.) in the report to provide intuitive decision-making support for management or review units; if review or cross-project comparison is required, the report can be compared with the actual operation data after completion to further improve the system database and algorithm model; Step 8.4: After completing the layout and pipeline design, the project can enter the construction drawing stage or the construction bidding stage, and BIM data can be used to achieve cross-disciplinary collaboration; if the actual construction or process requirements change, only the relevant parameters need to be updated and the algorithm process needs to be re-executed to quickly output a new solution, reflecting the system's continuous iteration capabilities; through full-process data recording and model retention, an integrated digital design foundation is provided for later operation and maintenance, equipment replacement, and upgrades and expansions.

[0015] Compared with the prior art, the beneficial effects of the present invention are as follows: 1. The present invention provides an intelligent layout method for sewage treatment plants based on a genetic optimization algorithm, which significantly improves the overall design efficiency. By automatically searching for layout schemes through machine learning and genetic algorithms, a large number of manual trial-and-error links are reduced.

[0016] 2. The present invention provides an intelligent layout method for sewage treatment plants based on a genetic optimization algorithm, which can quickly respond to modification requirements. Whether it is a change in the monomer size, land use conditions, or process flow, only the parameters need to be updated to restart the algorithm and efficiently generate a new scheme.

[0017] 3. The present invention provides an intelligent layout method for sewage treatment plants based on a genetic optimization algorithm, which realizes more refined conflict detection and multi-objective balance on the basis of integrating multi-source data. Through the automatic pathfinding algorithm and collision detection, the spatial conflicts between pipelines and structures are avoided to the greatest extent.

[0018] 4. The present invention provides an intelligent layout method for sewage treatment plants based on a genetic optimization algorithm. With the help of reinforcement learning and an adaptive reward and punishment mechanism, the layout effect can continuously evolve, and it can better meet the process requirements and operation and maintenance preferences of specific projects over time, forming a "self-learning" design process. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] In order to more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the following will briefly introduce the drawings required for the description of the specific embodiments or the prior art. Obviously, the following drawings are some embodiments of the present invention. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0020] Figure 1 It is a schematic diagram of the step flow of the present invention DETAILED DESCRIPTION OF THE INVENTION

[0021] The following will more clearly and completely elaborate the technical solutions of the present invention in conjunction with the drawings and through the description of the preferred embodiments of the present invention.

[0022] TERMINOLOGY EXPLANATION: 1. AAO: Refers to a type of sewage biological treatment process flow.

[0023] 2. BIM: A building engineering management method based on digitalization, visualization, and informatization.

[0024] 3. PDF: A document file format independent of platforms and devices, which can ensure consistent document layout across different operating environments.

[0025] 5. Rhino / Rhinoceros: The name of 3D modeling software, commonly used in industrial design, architectural design and engineering visualization; can be combined with the Grasshopper plug-in for visual programming and parametric design.

[0026] 6. Grasshopper: A visual programming plug-in for Rhino that implements parametric modeling and algorithm application through nodes and connections.

[0027] 7. IFC: A data exchange standard for the construction and engineering industries, used to transfer geometric and non-geometric information of buildings and structures between different BIM software.

[0028] 10.API: A software interface specification that allows data and functions to be called and interacted between different systems or programs.

[0029] like Figure 1 As shown, the present invention is specifically: Step 1: Through the automatic import and organization of multi-source data, the system obtains basic information of buildings / structures, terrain environment and real-time parameters, and completes machine learning preprocessing; Step 2: Combine the AAO process with the knowledge graph to map the functional characteristics of the monomers and the process dependencies into a graph structure, and realize dynamic grouping based on the inference results; Step 3: For each group, a multi-objective adaptive algorithm and interactive visualization are used to balance the multiple requirements of minimum land occupation, process connection and maintenance of safe distance, and realize the automatic arrangement of buildings and pipelines in a local area; Step 4: During the overall layout, an evolutionary genetic algorithm and an adaptive reward and punishment mechanism are introduced to globally optimize the parameters of each group, and the evaluation weights are dynamically adjusted according to the actual layout performance; Step 5: Based on external traffic simulation and future expansion requirements, automatically generate primary and secondary road networks and determine road details; Step 6: During the pipeline layout phase of the entire plant, through pipeline fluid simulation and BIM collision detection, a multi-level automatic path-finding algorithm is used to perform comprehensive optimization for the shortest path, minimum corners, and conflict avoidance; Step 7: Use reinforcement learning to iteratively update the design in a scheme evaluation and trial-and-error cycle, while allowing engineers to make real-time corrections and feedback on local layouts through an interactive interface; Step 8: Output results in various forms.

[0030] As a specific implementation, this technical solution first obtains various types of data including building geometric information, terrain changes, surrounding environment, and real-time operation and maintenance parameters from multi-source data such as BIM models, laser scans, drone aerial photography, and on-site sensors, and performs format conversion and coordinate calibration in the same coordinate system to ensure the consistency of subsequent processing. After importing this data into the Rhino+Grasshopper environment, the system filters out noise, defects, and duplicate information through machine learning. At the same time, according to the on-site construction drawings or known process parameters, attributes such as dimensions, interface positions, and process stage relationships are marked for each individual object (such as biochemical tanks, secondary sedimentation tanks, comprehensive buildings, etc.). Subsequently, a knowledge graph is used to associate the monomers on the AAO process chain with functions, positions, adjacent requirements, etc., construct a logical mapping with high correlation or priority adjacency for each monomer, and perform dynamic reasoning when encountering special process requirements to output a preliminary grouping result covering pretreatment, secondary treatment, advanced treatment, sludge treatment, and auxiliary buildings. After completing the monomer grouping, for each group, a multi-objective adaptive layout algorithm is adopted to comprehensively consider factors such as floor area, safety distance between equipment rooms, and maintenance convenience, and a "point-to-point" chain arrangement is carried out with the inlets and outlets or process interfaces between monomers as key nodes, and the multi-media pipeline routes within the group are generated in combination with automatic pathfinding algorithms such as Dijkstra. During this process, the system can score indicators such as land occupation shape, pipeline length, and number of corners in the Grasshopper environment according to weight settings, and continuously eliminate layouts that do not conform to process logic or have spatial conflicts in real time, so that the monomers and pipelines within the group maintain a compact and reasonable arrangement state.

[0031] After completing the internal layout of each group, the system treats each group as a "block" at a higher level and uses a genetic optimization algorithm to globally search for variable parameters such as the translation, rotation, and adjacent distance of the "block". After randomly generating several planar layouts in the initial population, the genetic algorithm evolves generation by generation through operators such as selection, crossover, and mutation, continuously scoring and eliminating according to indicators such as out-of-bounds, adjacency, process flow connection, wind direction, and road reservation, and can dynamically increase relevant penalty coefficients during the algorithm operation according to the frequency of problems to accelerate convergence. After multiple rounds of iteration, the system outputs several high-scoring candidate layouts for engineers to check. After determining the overall layout plan, the system automatically or semi-automatically generates the main road and secondary road networks based on the positions of the main functional areas and the plant entrances and exits, and gradually adjusts the road centerlines and widths in Grasshopper according to parameters such as vehicle driving simulation, turning radius, and expansion reserved space to ensure the traffic requirements of operation and maintenance and construction vehicles. Subsequently, for the complete plant layout, the automatic pathfinding algorithm is used again to uniformly arrange different medium pipelines across the whole plant, and optimize the pump station or pipeline elevation when necessary to adapt to the different requirements of gravity flow and pressure pipelines. If there are conflicts between pipelines or encroachments with single units, the system corrects them through local parameter adjustment and re-pathfinding, and can automatically detect collisions in the BIM environment or Grasshopper internal components. After all layouts and pipeline arrangements are completed, the system enters the reinforcement learning or further genetic iteration stage, and re-optimizes through local trial and error of the remaining problems and feedback from engineers' manual intervention, gradually converging the plan to meet the various constraints of the project. Finally, the system exports the determined building layout, main and secondary road networks, and multi-medium pipeline results together as planar drawings such as CAD or PDF, and can also output 3D model files to dock with BIM or VR / AR demonstrations, and reserve a docking port with the digital twin system at the data interface layer to provide complete digital support for subsequent construction, operation and maintenance, and expansion.

[0032] As a specific implementation method, the key working principles include: Single-unit grouping According to the AAO process principle and the experience accumulation of engineering personnel, single-unit structures are grouped as usual, including: Pretreatment group Including fixed single units: coarse grille, fine grille, and other customized single units for the remaining items Secondary treatment group Including fixed single units: biochemical pool, secondary sedimentation tank, and other customized single units for the remaining items Advanced treatment group Including fixed single units: sedimentation tank, filter tank, disinfection tank, metering tank, and other customized single units for the remaining items Sludge treatment group Including fixed monomers: sludge thickening tank, sludge storage tank, sludge dewatering machine room, and other customized monomers for the remaining projects Affiliated group Including all the remaining monomers other than the above groups: comprehensive building, workshop, etc. The significance of the above grouping is as follows: Among the fixed monomer groups in abcd (such as coarse grille and fine grille), due to the adjacent step relationship in the sewage treatment process, it is very rare that they are not adjacent in the spatial arrangement. Therefore, the method of grouping combined with the plane algorithm is adopted to ensure that the fixed monomers within the group have an adjacent relationship, and the connection point position and relative angle are used as control parameters. In this way, the fixed monomers within the group, plus the customized monomers of the project, form a "chain" in the form of "block + point + block". At the same time, through the algorithm, it is ensured that the chain formed by several monomers within the group can achieve the smallest and most reasonable land occupation shape in the plane layout.

[0033] Overall layout In the above-mentioned Principle 1, the algorithm of the chain layout has been initially reflected. If the "group chain" of a single group and its formed layout are regarded as a "block", and the "blocks" formed by several overall "group chains" are connected again with the "block" formed by the remaining monomers in group e in the form of "block + point + block", plus juxtaposition and nesting, a layout logic that can connect all the monomer buildings / structures of the sewage treatment plant is formed. This layout logic uses several parameters of the plane space freedom (distance, angle, point position) as control variables, and its output results can cover all the layout possibilities of the monomer buildings / structures of the AAO sewage treatment plant already input in the system.

[0034] Genetic optimization The genetic optimization algorithm is a random adaptive global search optimization method designed according to the evolution law of organisms in nature. This method simulates the computational model of the natural selection and genetic mechanism of biological evolution in Darwin's theory of biological evolution, and is a method of searching for the optimal solution by simulating the natural evolution process. Through mathematical means, using computer simulation operations, the problem-solving process is converted into processes similar to the crossover and mutation of chromosome genes in biological evolution. When solving relatively complex combinatorial optimization problems, it can usually obtain better optimization results quickly.

[0035] Based on the above Principle 1 and Principle 2, this solution adopts the genetic optimization algorithm to layout (optimize) the site. The key factors of the algorithm are respectively: Input information Based on the information entered into the system, including the basic outline of the monomer and the land use red line; and all the parameters of the plane space freedom as described in Principle 1 and Principle 2 above Evaluation system This system has developed a series of scoring systems for the general layout plan to guide machine learning and genetic optimization. The evaluation factors for the general layout are as follows: adjacent arrangement between individual units with a certain distance maintained, all arrangements and road pipelines not exceeding the land use red line, the overall process facilitating the general inlet and outlet of the site, the overall layout facilitating the wind direction and gas dispersion arrangement, etc. For each of the above optimization factors, there are different reward / punishment mechanisms.

[0036] According to the importance of the judgment points and their impact on the overall situation under specific project restrictions, this system introduces a reward and punishment mechanism with weights to automatically score and screen the general layout plans. Through all the layout logics of the weighted reward and punishment mechanism, the overall situation can be coordinated by the Galapagos genetic optimization algorithm built into Grasshopper. At the same time, it considers the change situations of all parameters and, guided by the optimization direction of the evaluation system, finds the optimal general layout plan that conforms to the process logic through the simulation operation process of randomization, crossover, and mutation.

[0037] Pipeline arrangement After the relative positions of the individual buildings / structures in the plane are determined, the system uses the automatic pathfinding algorithm of the leafvein solution package to arrange the pipelines of different media and quickly gives the pipeline arrangement plan. Based on the Dijkstra algorithm and weighted adjusted path guidance, the pipeline arrangement can meet the requirements of the shortest length, the straightest path with the fewest turns, and no mutual conflicts.

[0038] The above specific implementation manners only describe the preferred implementation manners of the present invention, rather than limiting the protection scope of the present invention. Without departing from the design concept and spirit scope of the present invention, various deformations, substitutions, and improvements made by those of ordinary skill in the art to the technical solutions of the present invention based on the written description and drawings provided by the present invention shall fall within the protection scope of the present invention. The protection scope of the present invention is determined by the claims.

Claims

1. An intelligent layout method for sewage treatment plants based on genetic optimization algorithm, characterized in that The following steps are involved: Step 1: Through the automatic import and organization of multi-source data, the system obtains basic information of buildings / structures, terrain environment and real-time parameters, and completes machine learning preprocessing; Step 2: Combine the AAO process with the knowledge graph to map the functional characteristics of the monomers and the process dependencies into a graph structure, and realize dynamic grouping based on the inference results; Step 3: For each group, a multi-objective adaptive algorithm and interactive visualization are used to balance the multiple requirements of minimum land occupation, process connection and maintenance of safe distance, and realize the automatic arrangement of buildings and pipelines in a local area; Step 4: During the overall layout, an evolutionary genetic algorithm and an adaptive reward and punishment mechanism are introduced to globally optimize the parameters of each group, and the evaluation weights are dynamically adjusted according to the actual layout performance; Step 5: Based on external traffic simulation and future expansion requirements, automatically generate primary and secondary road networks and determine road details; Step 6: During the pipeline layout phase of the entire plant, through pipeline fluid simulation and BIM collision detection, a multi-level automatic path-finding algorithm is used to perform comprehensive optimization for the shortest path, minimum corners, and conflict avoidance; Step 7: Use reinforcement learning to iteratively update the design in a scheme evaluation and trial-and-error cycle, while allowing engineers to make real-time corrections and feedback on local layouts through an interactive interface; Step 8: Output results in various forms.

2. The intelligent layout method of a sewage treatment plant based on a genetic optimization algorithm according to claim 1, wherein Step 1 is as follows: Step 1.1: Use BIM model output or IFC file to obtain the geometric information, interface location and material properties of each building / structure; use laser scanning equipment or drone aerial photography technology to obtain the site terrain, surface attachments and surrounding environment data, and convert these data into point clouds or mesh files that can be processed in Rhino; collect real-time parameters through sensors deployed on site to provide dynamic parameter input for subsequent design processes; Step 1.2: Convert the data from BIM, laser scanning / drone aerial photography, and sensors to make them compatible with the Rhino+Grasshopper environment; perform position calibration and coordinate unification for different types of data according to project needs; Step 1.3: Use machine learning algorithms or image recognition algorithms to filter out noise, defects, and repeated geometric objects; merge, correct, or mark incomplete or repeated monomers to ensure that there are no redundant objects in subsequent layouts; use filtering algorithms to smooth or eliminate outliers in laser scanning or drone aerial photography data; Step 1.4: Assign basic attribute labels to each monomer and create relevant tables in the database; The field sensor data is grouped by time or area to facilitate the calling of parameters in the subsequent layout algorithm.

3. The intelligent layout method of a sewage treatment plant based on a genetic optimization algorithm according to claim 1, characterized in that Step 2 is as follows: Step 2.1: Sort out the monomer types required at each stage of the AAO process chain and create corresponding nodes in the knowledge graph; for each monomer object, add attributes to the graph; establish dependencies between monomers to form graph edges of the process flow; Step 2.2: In the knowledge graph, mark the "high correlation" or "preferred adjacency" attributes for monomers with tight process connection or physical adjacency requirements; use the inference engine or simple query rules to automatically retrieve some conventional grouping patterns; in case of special process requirements, temporarily add nodes in the graph and run the inference to update the grouping results; Step 2.3: If the designer makes manual adjustments to the groups automatically generated by the system based on on-site experience, record these operations in the system log; On the algorithm side, optimize the parameter weights according to the results of the manual adjustments; finally, output a fixed or dynamic grouping list for subsequent "block" management in Grasshopper.

4. The intelligent layout method of a sewage treatment plant based on a genetic optimization algorithm according to claim 1, wherein Step 3 is specifically as follows: Step 3.1: For each group, determine the target parameters to be optimized within it; set weights for different targets and establish corresponding scoring or penalty metrics in Grasshopper; Step 3.2: Apply the "point-to-point" chain logic within the group, with the connection points between monomers as control nodes; in Grasshopper, combine heuristic search or local genetic algorithms to gradually try different monomer arrangement orders, relative distances, and rotation angles; after each arrangement is generated, evaluate it, and then retain the high-score combinations and eliminate the low-score combinations; Step 3.3: Use the automatic pathfinding algorithm to search for the shortest or least-angled path for the multi-media pipelines within the group; check whether the generated pipelines cross the monomer boundaries, and if there are conflicts, regenerate them by moving nodes or adjusting the path weights; package and output the final determined layout within the group and the corresponding pipelines as "sub-blocks" for the subsequent overall layout; Step 3.4: The engineer can directly drag or rotate a monomer in the Rhino view, and the indicators can be updated in real time through Grasshopper; if it is found that the local optimum leads to an obviously unreasonable layout, manual intervention can be carried out and the operations can be recorded for subsequent algorithm learning; after determining the optimal or sub-optimal local arrangement, encapsulate the results as "grouping results" and then transfer to the overall layout of the next stage.

5. The intelligent layout method of a sewage treatment plant based on a genetic optimization algorithm according to claim 1, characterized in that, Step 4 is specifically as follows: Step 4.1: Abstract each group into a higher-level "block" and set adjustable parameters; generate several initial layout schemes through sampling methods; Step 4.2: Establish an evaluation function according to the requirements of the whole plant; during the algorithm iteration process, if some unreasonable layouts appear too frequently, the system automatically increases the penalty coefficient for unreasonable layouts to form an adaptive reward and punishment mechanism; calculate the comprehensive score for each layout to guide the selection and crossover links of the genetic algorithm; Step 4.3: Call the genetic algorithm plug-in in the Grasshopper environment for basic operations; in the selection link, preferentially retain the high-score individuals and add a small number of low-score individuals to maintain the population diversity; in the crossover link, generate new offspring by combining the parameters of the parent generation; in the mutation link, randomly change the translation or rotation amount of some blocks to jump out of the local optimum; after repeating multiple generations of iteration, retain several high-score layout schemes as the global layout results for selection. Step 4.4: After each iteration, Grasshopper automatically outputs the visual layout of the highest-scoring solution in the current population; engineers can view the scoring details of each candidate layout at any time to identify possible logical conflicts or site encroachments; if systematic deviations occur, relevant penalty weights can be increased or parameter ranges can be restricted and the iteration can be restarted.

6. The intelligent layout method of a sewage treatment plant based on a genetic optimization algorithm according to claim 1, wherein, Step 5 is as follows: Step 5.1: Extract the main grouping information and determine the docking points inside the factory area and with external roads; if there are available prediction or simulation models for external traffic, import the road traffic data or vehicle driving data into the Grasshopper environment as a reference for subsequent planning. Step 5.2: According to the positions and operation and maintenance needs of the core functional areas or main monomers, delimit at least one main road alignment running through the whole factory; set parameters in Grasshopper and automatically or semi-automatically generate the centerline of the main road that meets the site restrictions; conduct a site encroachment inspection on the preliminarily generated main road, and if there is overstepping the boundary or being too close to key structures, correct it by fine-tuning the road nodes or path weights. Step 5.3: Based on the existing main roads, arrange secondary channels for each functional area or remote monomer to meet the access needs of operation and maintenance vehicles or pedestrians; incorporate key nodes into the layout parameters and automatically arrange them in combination with the terrain and the functional positioning of the factory area; for later expansion or renovation requirements, reserve connection ports or retain road corridors in the road network design to avoid large-scale demolition and reconstruction during later construction. Step 5.4: In the Rhino view, display the superposition effect of the primary and secondary road networks and the monomer layout in real time for engineers to check for problems; if traffic bottlenecks or potential safety hazards are detected, modify the road parameters and regenerate the road network; the finally determined road system is regarded as a fixed traffic plan in subsequent pipeline layout and global assessment.

7. The intelligent layout method of a sewage treatment plant based on a genetic optimization algorithm according to claim 1, characterized in that Step 6 is as follows: Step 6.1: Read the pipeline interface types, pipe diameters, and connection requirements of each monomer from the grouped layout; if there are terrain undulations or different elevation requirements, classify the "gravity flow pipelines" and "pressure pipelines" separately to establish a hierarchical processing plan for subsequent automatic pathfinding; incorporate the newly generated road network into the constraints and give priority to laying pipelines along roads or pipe gallery areas. Step 6.2: Use a custom multi-level automatic pathfinding algorithm to generate candidate routes with the "shortest path" or "fewest turns" for various types of media pipelines; when generating pipelines, regard roads, structures, and red lines as obstacles to ensure that the paths do not cross or have unreasonable detours. Perform elevation calculation or slope inspection on gravity flow pipelines. Step 6.3: After generating the preliminary pipeline plan, summarize the results into the BIM environment or use the built-in collision detection component in Grasshopper to check for spatial conflicts between pipelines, between pipelines and monomers. Step 6.4: Output the layout results of the whole factory's pipelines and visually display the information in Rhino, and the pipeline weights can be modified or additional constraints can be added in Grasshopper to re-solve the problem. After both the conflict issues and functional requirements are met, the final pipeline layout is encapsulated as a deliverable and enters the next evaluation and revision process.

8. The intelligent layout method of a sewage treatment plant based on a genetic optimization algorithm according to claim 1, characterized in that Step 7 is as follows: Step 7.1: Integrate the overall layout, road, and pipeline results, and conduct a quantitative evaluation through indicators; if any indicator fails to meet the pre-set standard, show the problem location or the indicator gap in the evaluation report. Step 7.2: Represent the layout scheme as a "state" from the perspective of the algorithm; set an "immediate reward" for each evaluation indicator; set a negative reward for serious problems to guide the algorithm to actively avoid these situations when trying new layouts. Step 7.3: Let the system perform small-scale parameter perturbations in each dimension to simulate "trial and error". If an improvement is generated, strengthen the path; otherwise, converge back to the original scheme in a timely manner. The engineer gives a "manual confirmation" or "rejection" mark to the local corrections automatically generated by the system to help the algorithm continuously learn and build an optimization strategy that better conforms to practical experience. After multiple rounds of trial and error, archive the scheme with the highest reward and record its decision-making path. Step 7.4: When the reinforcement learning algorithm converges under the indicators or meets the project threshold, output the final revised plant-wide layout scheme, road, and pipeline arrangement; package the key parameter changes, important decision points, and successful cases in the entire reinforcement learning process into a visual report.

9. The intelligent layout method of a sewage treatment plant based on a genetic optimization algorithm according to claim 1, wherein Step 8 is as follows: Step 8.1: Export the 2D plan view and 3D model of the layout results from the Rhino+Grasshopper environment at one time for direct reference on the CAD and BIM platforms; generate professional sub-drawings for the main pipelines and roads, and mark the key dimensions, coordinates, and bill of materials information on the drawings. Step 8.2: If the project requirements include visual display, import the final 3D model into VR or AR software for project participants to immerse themselves in viewing the plant layout and pipeline routing; retain the interface for docking with the digital twin system. Step 8.3: Integrate the information in the genetic algorithm and reinforcement learning process into a detailed design report; list the parameters in each stage in the report to provide intuitive decision-making support for management or review units. Step 8.4: After completing the layout and pipeline design, the project can enter the construction drawing stage or the construction bidding link, and use BIM data to achieve cross-professional collaboration; if actual construction or process requirements change, simply update the relevant parameters and re-execute the algorithm process to quickly output a new scheme.

Images