Intelligent Evaluation Method and System for Pipeline System Layout Based on Neural Network

Abstract

The present invention provides an intelligent evaluation method and system for the layout of a pipeline system based on a neural network. The pipeline layout is designed according to the starting and ending positions of the pipeline, each scenario is marked and distinguished to obtain three-dimensional point cloud data, and then it is segmented into multiple small regions. Object analysis and scoring are performed on the small region scenarios to obtain an encoded vector indicating the spatial characteristics of the region, and then they are spliced into a one-dimensional vector as the feature encoded vector of the entire scene region, and normalization and standardization data processing are carried out. A neural network is used as the regression model framework to perform regression on the pipeline space scene dataset; the regression effect of the regression model is evaluated; the regression model is trained, and the accuracy, effectiveness, and stability of the regression model are tested through a test set. By establishing a neural network, the present invention can process a large amount of data, instantly evaluate the pipe laying scheme, reduce the manual review process, provide real-time feedback, and speed up the design cycle.

Description

Technical Field

[0001] The present invention relates to the technical field of pipeline layout design, and in particular, to an intelligent evaluation method and system for pipeline system layout based on a neural network. Background Art

[0002] The pipeline system is a key link in the design and construction process of complex equipment. Its optimal design and intelligent layout are of great significance for enhancing the competitiveness and scientific and technological strength of the manufacturing industry, and are also an inevitable trend for the transformation and upgrading of the manufacturing industry. The design and layout quality of the pipeline system directly affect the stability, reliability, and economy of complex equipment, and are crucial for the operation performance and safety of complex equipment. A reasonable pipeline layout can ensure the smooth transportation of liquids and gases, improve system performance, and reduce maintenance costs.

[0003] The pipeline layout design is often affected by both subjective and objective decision-making factors. Traditional layout methods rely on the experience and intuition of designers, but there are problems such as low process efficiency and great difficulty in knowledge inheritance. To improve the design efficiency, it is possible to consider using classical algorithms to achieve automated design. However, existing algorithms mainly focus on obvious factors that are easy to quantify, such as pipe size, material properties, mechanical parameters, pipe length, etc. Traditional algorithms are difficult to capture the tacit knowledge of senior designers, such as aesthetic considerations, risk preferences, etc., but these tacit factors are also crucial for the success of pipeline system layout. The automated design of traditional algorithms is difficult to achieve the complexity and flexibility of manual design, so new layout methods are needed to improve the system design level.

[0004] Currently, the multi-factor identification for pipeline layout mainly includes factors such as the distance between the pipeline and obstacles, pipeline cost, pipeline length, pressure loss, heat loss, pipeline height, etc. Combining with the adaptive beetle antennae search algorithm, the ant colony algorithm based on human-machine collaboration, the evolutionary algorithm of the fully connected weight network, etc., the pipeline path planning is completed. However, the existing research mainly focuses on objective factors in considering decision variables and influencing factors, lacking comprehensive consideration of subjective fuzzy experience. At the same time, the weighting and consideration of influencing factors mainly rely on direct manual setting, and there is less analysis of the subjective and objective factors of the human-machine system.

[0005] Patent document CN115563719A discloses a method for optimizing the layout of a single pipeline in a double layer inside an aircraft fuel tank. This method generates a point cloud set based on the three-dimensional model of the aircraft fuel tank, extracts the pipeline laying space from the point cloud set according to the three-dimensional coordinates of the two ends of the pipeline and converts it into a three-dimensional grid map, generates an initial solution set of paths that meet the constraints using the ant colony algorithm based on the three-dimensional grid map, generates an initial arm layout plan for each initial path and calculates the length of each arm, iteratively optimizes using the genetic algorithm to obtain the optimal solution of the arm layout for each path, and uses the plan with the minimum sum of the total arm layout length and the pipeline path length to guide the next search, finally obtaining the layout plan with the optimal sum of the pipeline length and the arm length. However, this patent cannot completely solve the existing technical problems and also cannot meet the requirements of the present invention. Summary of the Invention

[0006] Aiming at the defects in the prior art, the purpose of the present invention is to provide an intelligent evaluation method and system for the layout of a pipeline system based on a neural network.

[0007] The intelligent evaluation method for the layout of a pipeline system based on the neural network provided by the present invention includes:

[0008] Step S1: Design the pipeline layout according to the starting and ending positions of the pipeline, and score each pipeline scenario according to the number of elbows, pipeline length, and distance from obstacles.

[0009] Step S2: Mark and distinguish each scenario according to factors such as obstacles, pipelines, starting points, and ending points, and obtain three-dimensional point cloud data by combining the three-dimensional coordinates in space.

[0010] Step S3: Process the feature of the point cloud data set of the scenario and divide it into multiple small regions according to a preset ratio.

[0011] Step S4: Analyze the objects in the small region scenario and score according to a function to obtain a coding vector indicating the spatial characteristics of the region.

[0012] Step S5: After obtaining the feature coding vectors of each small region, splice them into a one-dimensional vector in spatial order as the feature coding vector of the entire scenario region.

[0013] Step S6: Perform normalization and standardization data processing on the feature coding vector of the entire scenario region as the input of the neural network.

[0014] Step S7: Use a fully connected neural network as the regression model framework to perform regression on the pipeline space scenario data set.

[0015] Step S8: Evaluate the regression effect of the regression model using evaluation indicators.

[0016] Step S9: Use the hold-out method to divide the dataset into a training set and a test set, train the regression model, and test the accuracy, effectiveness, and stability of the regression model through the test set.

[0017] Preferably, step S4 includes:

[0018] First, taking the obstacle distribution attribute as the first component of the encoding, divide this regional space into two categories according to whether there is a pipeline segment in this area, and use the proportion of pipelines and obstacles occupied in the space as the scoring basis. The expression is:

[0019]

[0020] Among them, Score obs represents the score of obstacles in the current spatial region, N total represents the number of spatial regions, and N obs represents the number of obstacles occupied;

[0021] Secondly, taking the spatial attribute of the pipeline segment as the second component of the encoding, calculate the number of pipeline distributions in this area, and use the relative position of the start and end points as the weight ratio of this area to statistically calculate the pipeline length score. The expression is:

[0022]

[0023]

[0024] Among them, Score path represents the path score of the current small area, N path represents the number of paths occupied, and α is the correction coefficient for path scoring according to the relative position of the space; represents the central position vector of the current spatial region, represents the start point vector corresponding to this path, represents the end point vector corresponding to this path;

[0025] Regarding the pipeline relationship in the small area, through the inner product of the start and end point vectors of different labeled pipelines, the relationship between pipelines is characterized. The expression is:

[0026]

[0027] Among them, Score dir represents the pipeline direction intersection score, X i represents the start and end point vector of a pipeline in this area, and X j represents the start and end point vector of another pipeline in this area.

[0028] Preferably, the step S6 includes: performing normalization and standardization data processing on the feature encoding vectors of the entire scene area, and the expression is:

[0029]

[0030] where x i is the i-th data in the dataset, represents the dataset average value, and x std represents the dataset standard deviation.

[0031] Preferably, the step S7 includes: taking the expert experience score as the target value and the spatial encoding vector as the input value, defining the MSE as the loss function, and its basic regression model is as follows:

[0032] z i = Wx i + b

[0033] y i = f(z)

[0034]

[0035] where z i is the output value after neuron mapping, f() represents the activation function, y i represents the predicted value after neural network fitting, represents the true value obtained by expert scoring, n is the number of samples, W is the input weight, and b is the bias term.

[0036] Preferably, the step S8 includes: adopting the evaluation indexes mainly based on the goodness of fit R2 and supplemented by the root mean square error RMSE, and the expression is:

[0037]

[0038]

[0039]

[0040] In the formula, is the average value of the dataset scores.

[0041] According to the intelligent evaluation system for pipeline system layout based on neural network provided by the present invention, it includes:

[0042] Module M1: Design the pipeline layout according to the starting and ending positions of the pipeline, and score each pipeline scene according to the number of elbows, pipeline length, and distance from obstacles;

[0043] Module M2: Mark and distinguish each scenario according to the factors of obstacles, pipelines, starting points, and ending points, and combine the three-dimensional coordinates in space to obtain three-dimensional point cloud data;

[0044] Module M3: Perform feature processing on the point cloud data set of the scenario and divide it into multiple small regions according to a preset ratio;

[0045] Module M4: Perform object analysis on the small region scenario and score according to the function to obtain an encoding vector indicating the spatial characteristics of the region;

[0046] Module M5: After obtaining the feature encoding vectors of each small region, splice them into a one-dimensional vector in spatial order as the feature encoding vector of the entire scenario region;

[0047] Module M6: Perform normalization and standardization data processing on the feature encoding vector of the entire scenario region as the input of the neural network;

[0048] Module M7: Use a fully connected neural network as the regression model framework to perform regression on the pipeline space scenario data set;

[0049] Module M8: Use evaluation indicators to evaluate the regression effect of the regression model;

[0050] Module M9: Use the hold-out method to divide the data set into a training set and a test set, train the regression model, and test the accuracy, effectiveness, and stability of the regression model through the test set.

[0051] Preferably, the module M4 includes:

[0052] First, use the obstacle distribution attribute as the first component of the encoding. Divide this region of space into two categories according to whether there are pipeline segments in this region, and use the proportion of pipelines and obstacles occupying the space as the scoring basis. The expression is:

[0053]

[0054] Among them, Score obs represents the score of obstacles in the current spatial region, N total represents the number of spatial regions, N obs represents the number of obstacles occupied;

[0055] Secondly, use the spatial attribute of the pipeline segment as the second component of the encoding, calculate the number of pipeline distributions in this region, and use the relative position of the starting and ending points as the weight ratio of this region to statistically calculate the pipeline length score. The expression is:

[0056]

[0057]

[0058] Among them, Score path represents the current small area path score, N path represents the number of paths occupied, and α is the correction coefficient for path scoring based on the relative spatial position; represents the central position vector of the current spatial area, represents the starting point vector corresponding to this path, represents the ending point vector corresponding to this path;

[0059] For the pipeline relationship within the small area, the relationship between pipelines is characterized by the inner product of the starting and ending point vectors of pipelines with different labels. The expression is:

[0060]

[0061] Among them, Score dir represents the pipeline direction crossing score, X i represents the starting and ending point vectors of a pipeline within this area, X j represents the starting and ending point vectors of another pipeline within this area.

[0062] Preferably, the module M6 includes: performing normalization and standardization data processing on the feature encoding vector of the entire scene area. The expression is:

[0063]

[0064] Among them, x i is the i-th data in the dataset, represents the dataset average value, x std represents the dataset standard deviation.

[0065] Preferably, the module M7 includes: using the expert experience score as the target value and the spatial encoding vector as the input value, and defining the MSE as the loss function. Its basic regression model is as follows:

[0066] z i = Wx i + b

[0067] y i = f(z)

[0068]

[0069] Among them, z i is the output value after neuron mapping, f() represents the activation function, y i represents the predicted value after neural network fitting, represents the true value obtained through expert scoring, n is the number of samples, W is the input weight, and b is the bias term.

[0070] Preferably, the module M8 includes: adopting an evaluation index mainly based on the goodness of fit R2 and supplemented by the root mean square error RMSE, and the expression is:

[0071]

[0072]

[0073]

[0074] In the formula, is the average value of the dataset scores.

[0075] Compared with the prior art, the present invention has the following beneficial effects:

[0076] (1) By establishing a neural network, a large amount of data can be processed, the pipe laying scheme can be evaluated immediately, the manual review process is reduced, real-time feedback is provided, and the design cycle is accelerated;

[0077] (2) By introducing a large number of design schemes with horizontal stagger and corresponding scores, the tacit knowledge that is difficult for designers to express is captured, and at the same time, the influence of personal preferences and misunderstandings is reduced, and the design defects caused by human judgment errors are reduced;

[0078] (3) By establishing a convenient design scheme input mechanism, continuous data input and training can be better accommodated, so that the neural network can continuously update the evaluation criteria and reflect the latest design concepts and technical requirements. Description of the Drawings

[0079] By reading the following detailed description of the non-limiting embodiments with reference to the accompanying drawings, other features, objects, and advantages of the present invention will become more obvious:

[0080] Figure 1 is the full connection neural network framework diagram;

[0081] Figure 2 is the pipeline layout evaluation flow chart based on spatial feature description. Detailed Embodiments

[0082] The present invention will be described in detail below with reference to specific embodiments. The following embodiments will help those skilled in the art to further understand the present invention, but do not limit the present invention in any form. It should be noted that those of ordinary skill in the art can make several changes and improvements without departing from the concept of the present invention. These all belong to the protection scope of the present invention.

[0083] Embodiment 1

[0084] The present invention provides an intelligent evaluation method for the layout of pipeline systems based on neural networks, mainly solving the problems of difficult consideration of implicit factors in pipeline layout and low efficiency of manual evaluation. At the application level, this evaluation method breaks through the technical bottleneck that traditional design methods can only consider easily quantifiable objective factors, and can provide practical technical support for the integration of subjective and objective factors in pipeline design to achieve the optimal comprehensive performance, improve the evaluation efficiency of pipeline layout, speed up the design cycle, enhance the evaluation flexibility, reduce the need for rework, save design costs, and improve the market competitiveness of the manufacturing and construction industries; at the academic level, this evaluation method has important reference significance for understanding fuzzy design under the influence of implicit factors and realizing agile design based on human-machine collaboration. This method provides a practical solution for quickly evaluating pipeline layout schemes in industrial production by establishing mathematical representations of free space, obstacles, and pipelines and constructing a multi-layer perceptron neural network to fit the scores of layout schemes.

[0085] To achieve the above object, the technical solution of the present invention is: analyze the spatial characteristics of the pipeline scenario, including operations such as spatial scenario area division, construction of a scale pyramid, local spatial feature analysis, and local scoring function design. Through the above analysis steps, the essential spatial characteristics of the pipeline distribution will be described, providing input for the subsequent automatic scoring network; through the objective description of the spatial characteristics of the scenario, a large number of combinations and nestings of specific evaluation functions are avoided, the core mapping relationship between the scenario and the score is constructed, the influence of engineers' subjective modification is reduced, and implicit knowledge that is difficult to express can be captured. Compared with the prior art, the effectiveness and efficiency of the present invention are verified, providing an effective solution for achieving high-efficiency and high-quality automation of pipeline layout.

[0086] Such as Figure 2 , the specific implementation steps are as follows:

[0087] Step S1, generate a large number of pipeline layout designs based on the starting and ending positions of the pipelines in the ship scenario, and score each pipeline scenario according to expert experience analysis. Consider three factors: the number of elbows, the pipeline length, and the distance from obstacles, and sort the pipeline scenarios according to the scores. It is better to have fewer elbows, shorter pipeline length, and shorter distance from obstacles. Obtain three ordered scenario sequences, and make them evenly distributed in the 4-10 score range according to the ranking, so that each scenario obtains three corresponding scores, thus constituting the basis of the scenario-score data set.

[0088] Step S2, mark and distinguish each scenario according to factors such as obstacles, pipelines, starting points, and ending points, and output and save its three-dimensional point cloud data in combination with the three-dimensional coordinates in space.

[0089] Step S3: Perform core feature processing on the point cloud dataset of the scene. Taking a cuboid cabin as an example, it is divided into small regions according to a corresponding ratio. For example, a cuboid region of 20*20*20 is decomposed into small regions of 3*3*3, and the small regions are sequentially translated to the right and down starting from the upper left corner by a constant number of coordinate units.

[0090] Step S4: There are mainly two objects, namely pipeline segments and obstacles, in the small region. Analyze the objects in the small region scene and score according to the function to obtain the encoding vector indicating the spatial characteristics of this region.

[0091] Step S5: First, take the obstacle distribution attribute as the first component of the encoding. Classify it into two categories according to whether there are pipeline segments in this region space, and use the proportion of pipelines and obstacles occupying the space as the scoring basis.

[0092]

[0093] Among them, Score obs represents the score of obstacles in the current spatial region, N total represents the number of spatial regions, and N obs represents the number of obstacles occupied.

[0094] Step S6: Second, take the pipeline segment spatial attribute as the second component of the encoding. Calculate the number of pipeline distributions in this region, and use the relative position of the start and end points as the weight ratio of this region to statistically calculate the pipeline length score.

[0095]

[0096]

[0097] Among them, Score path represents the path score of the current small region, N path represents the number of paths occupied, and α is the correction coefficient for path scoring according to the relative position of the space; represents the central position vector of the current spatial region, represents the start vector corresponding to this path, represents the end vector corresponding to this path.

[0098] Step S7: For the pipeline relationship within the small region, represent the relationship between pipelines through the inner product of the start and end vectors of different labeled pipelines.

[0099]

[0100] Among them, Score dir represents the pipeline direction intersection score, and X iIndicates the start and end vector of a pipeline within this area, X j Indicates the start and end vector of another pipeline within this area.

[0101] Step S8, after obtaining the feature encoding vector of each small area, splice them into a one-dimensional vector in spatial order as the feature encoding vector of the entire scene area.

[0102] Step S9, perform normalization and standardization data processing on the feature encoding vector of the overall scene as the input of the neural network.

[0103]

[0104] where, x i is the i-th data in the dataset, represents the dataset average value, x std represents the dataset standard deviation.

[0105] Step S10, use a fully connected neural network as the basic regression model framework to perform regression on the pipeline space scene dataset. Its structure is roughly as Figure 1 shown. Using the expert experience score as the target value and the self-designed spatial encoding vector as the input value, its basic regression model is as follows, and define MSE as the loss function.

[0106] z i = Wx i + b

[0107] y i = f(z)

[0108]

[0109] where, z i is the output value after neuron mapping, f() represents the activation function, y i represents the predicted value after being fitted by the neural network, represents the true value obtained through expert scoring, n is the number of samples, W is the input weight, and b is the bias term.

[0110] Step S11, use relevant evaluation indicators to evaluate the regression effect of the regression model, such as mean squared error (MSE), root mean squared error (RMSE), coefficient of determination (R2), etc. Considering that the main purpose of the neural network is to perform score fitting on the scene, the evaluation indicators with the coefficient of determination (R2) as the main and the root mean squared error (RMSE) as the supplement are adopted. The specific formula of R2 is as follows:

[0111]

[0112]

[0113]

[0114] wherein, is the average value of the dataset scores.

[0115] Step S12, using the hold-out method to divide the dataset into a training set and a test set. Among them, the test set does not participate in the training process and is only used to verify the model obtained by training. The model is fully trained on the training set, learns the data rules of the training set, and is tested on the test set.

[0116] Embodiment 2

[0117] The present invention also provides an intelligent evaluation system for the pipeline system layout based on a neural network. The intelligent evaluation system for the pipeline system layout based on a neural network can be implemented by executing the process steps of the intelligent evaluation method for the pipeline system layout based on a neural network. That is, those skilled in the art can understand the intelligent evaluation method for the pipeline system layout based on a neural network as a preferred embodiment of the intelligent evaluation system for the pipeline system layout based on a neural network.

[0118] The intelligent evaluation system for the pipeline system layout based on the present invention includes:

[0119] Module M1: Design the pipeline layout according to the starting and ending positions of the pipeline, and score each pipeline scenario according to the number of elbows, pipeline length, and distance from obstacles;

[0120] Module M2: Mark and distinguish each scenario according to factors such as obstacles, pipelines, starting points, and ending points, and combine the three-dimensional coordinates in space to obtain three-dimensional point cloud data;

[0121] Module M3: Perform feature processing on the point cloud dataset of the scenario, and divide it into multiple small regions according to a preset ratio;

[0122] Module M4: Perform object analysis on the small region scenario and score according to a function to obtain a coding vector indicating the spatial characteristics of the region;

[0123] Module M5: After obtaining the feature coding vectors of each small region, splice them into a one-dimensional vector in spatial order as the feature coding vector of the entire scenario region;

[0124] Module M6: Perform normalization and standardization data processing on the feature coding vector of the entire scenario region as the input of the neural network;

[0125] Module M7: Use a fully connected neural network as the regression model framework to perform regression on the pipeline space scenario dataset;

[0126] Module M8: Evaluate the regression effect of the regression model using evaluation indicators;

[0127] Module M9: Use the hold-out method to divide the dataset into a training set and a test set, train the regression model, and test the accuracy, effectiveness, and stability of the regression model through the test set.

[0128] The said module M4 includes:

[0129] First, taking the obstacle distribution attribute as the first component of the encoding, divide this regional space into two categories according to whether there are pipeline segments, and use the proportion of pipelines and obstacles occupied in the space as the scoring basis. The expression is:

[0130]

[0131] Among them, Score obs represents the score of obstacles in the current spatial region, N total represents the number of spatial regions, N obs represents the number occupied by obstacles;

[0132] Second, taking the pipeline segment spatial attribute as the second component of the encoding, calculate the number of pipeline distributions in this region, and use the relative position of the start and end points as the weight ratio of this region to statistically calculate the pipeline length score. The expression is:

[0133]

[0134]

[0135] Among them, Score path represents the path score of the current small region, N path represents the number occupied by the path, and α is the correction coefficient for path scoring according to the spatial relative position; represents the central position vector of the current spatial region, represents the start vector corresponding to this path, represents the end vector corresponding to this path;

[0136] For the pipeline relationship within the small region, the relationship between pipelines is characterized by the inner product of the start and end vectors of pipelines with different labels. The expression is:

[0137]

[0138] Among them, Score dir represents the pipeline orientation intersection score, X i represents the start and end vector of a pipeline within this region, X j represents the start and end vector of another pipeline within this region.

[0139] The module M6 includes: normalizing and standardizing the feature encoding vector of the entire scene area, and the expression is:

[0140]

[0141] where x i is the i-th data in the dataset, represents the dataset average value, and x std represents the dataset standard deviation.

[0142] The module M7 includes: taking the expert experience score as the target value and the spatial encoding vector as the input value, and defining the MSE as the loss function. Its basic regression model is as follows:

[0143] z i = Wx i + b

[0144] y i = f(z)

[0145]

[0146] where z i is the output value after neuron mapping, f() represents the activation function, and y i represents the predicted value after neural network fitting, represents the true value obtained from expert scoring, n is the number of samples, W is the input weight, and b is the bias term.

[0147] The module M8 includes: adopting the evaluation indexes mainly based on the goodness of fit R2 and supplemented by the root mean square error RMSE, and the expression is:

[0148]

[0149]

[0150]

[0151] In the formula, is the average value of the dataset scores.

[0152] Those skilled in the art know that, in addition to implementing the systems, devices, and their respective modules provided by the present invention in the form of pure computer-readable program code, it is entirely possible to logically program the method steps so that the systems, devices, and their respective modules provided by the present invention are implemented in the form of logic gates, switches, application-specific integrated circuits, programmable logic controllers, and embedded microcontrollers, etc., to achieve the same program. Therefore, the systems, devices, and their respective modules provided by the present invention can be considered as a kind of hardware component, and the modules included therein for implementing various programs can also be regarded as the structures within the hardware component; the modules for implementing various functions can also be regarded as either software programs for implementing the method or the structures within the hardware component.

[0153] The specific embodiments of the present invention have been described above. It should be understood that the present invention is not limited to the above specific embodiments, and those skilled in the art can make various changes or modifications within the scope of the claims, which does not affect the essence of the present invention. Without conflict, the embodiments of the present application and the features in the embodiments can be combined with each other arbitrarily.

Claims

1. An intelligent evaluation method for the layout of a pipeline system based on a neural network, characterized in that Including: Step S1: Design the pipeline layout according to the starting and ending positions of the pipeline, and score each pipeline scenario based on the number of elbows, pipeline length, and distance from obstacles. Step S2: Mark and distinguish each scenario according to factors such as obstacles, pipelines, starting points, and ending points, and obtain three-dimensional point cloud data by combining three-dimensional coordinates in space. Step S3: Process the point cloud data set of the scenario and divide it into multiple small regions according to a preset ratio. Step S4: Analyze the objects in the small region scenario and score according to a function to obtain a coding vector indicating the spatial characteristics of the region. Including: Using the obstacle distribution attribute as the first component of the code, classifying it into two categories according to whether there are pipeline segments in the region space, and counting the occupancy ratio of pipelines and obstacles in the space as the score of the current space region obstacle based on the score; using the spatial attribute of the pipeline segment as the second component of the code, calculating the number of pipeline distributions in the region, and using the relative position of the starting and ending points as the weight ratio of the region to count the path score of the current small region for the pipeline length score; for the pipeline relationship in the small region, representing the relationship between pipelines through the inner product of the starting and ending point vectors of different label pipelines, the pipeline orientation intersection score. Step S6: After obtaining the feature coding vector of each small region, splice it into a one-dimensional vector in spatial order as the feature coding vector of the entire scene region. Step S7: Perform normalization and standardization data processing on the feature coding vector of the entire scene region as the input of the neural network. Step S8: Use a fully connected neural network as the regression model framework to perform regression on the pipeline space scenario data set, including: using the expert experience score as the target value, the spatial coding vector as the input value, and defining the MSE as the loss function. Step S9: Use evaluation indicators to evaluate the regression effect of the regression model. Step S10: Use the hold-out method to divide the data set into a training set and a test set, train the regression model, and test the accuracy, effectiveness, and stability of the regression model through the test set.

2. The intelligent evaluation method for the layout of the pipeline system based on a neural network according to claim 1, characterized in that The said Step S4 includes: First, using the obstacle distribution attribute as the first component of the code, classifying it into two categories according to whether there are pipeline segments in the region space, and counting the occupancy ratio of pipelines and obstacles in the space as the score basis, the expression is: Among them, Score obs represents the score of obstacles in the current spatial region, N total represents the number of spatial regions, N obs represents the number occupied by obstacles; Secondly, using the spatial attribute of the pipeline segment as the second component of the code, calculating the number of pipeline distributions in the region, and using the relative position of the starting and ending points as the weight ratio of the region to count the pipeline length score, the expression is: Among them, Score path represents the path score of the current small area, N path represents the number of paths occupied, and α is the correction coefficient for path scoring according to the relative spatial position; represents the central position vector of the current spatial area, represents the starting point vector corresponding to this path, represents the ending point vector corresponding to this path; For the pipeline relationship in the small region, representing the relationship between pipelines through the inner product of the starting and ending point vectors of different label pipelines, the expression is: Among them, Score bir represents the pipeline routing intersection score, X i represents the start and end point vector of a pipeline within this area, X j represents the start and end point vector of another pipeline within this area.

3. The intelligent evaluation method for the layout of a pipeline system based on a neural network according to claim 1, wherein The said Step S6 includes: Performing normalization and standardization data processing on the feature coding vector of the entire scene region, the expression is: where x i is the i-th data in the dataset, represents the dataset average, and x std represents the dataset standard deviation.

4. The intelligent evaluation method for the pipeline system layout based on a neural network according to claim 3, wherein The said Step S7 includes: Using the expert experience score as the target value, the spatial coding vector as the input value, and defining the MSE as the loss function, its basic regression model is as follows: z i = Wx i + b y i = f(z) Among them, z i is the output value after neuron mapping, f() represents the activation function, and y i represents the predicted value after neural network fitting, denotes the true value obtained through expert scoring, n is the number of samples, W is the input weight, and b is the bias term.

5. The intelligent evaluation method for the layout of a pipeline system based on a neural network according to claim 4, wherein The said Step S8 includes: Adopting evaluation indicators mainly based on the goodness of fit R2 and supplemented by the root mean square error RMSE, the expression is: In the formula, is the average value of the dataset scores.

6. An intelligent evaluation system for the layout of a pipeline system based on a neural network, characterized in that, Including: Module M1: Design the pipeline layout according to the starting and ending positions of the pipeline, and score each pipeline scenario based on the number of elbows, pipeline length, and distance from obstacles. Module M2: Mark and distinguish each scenario according to factors such as obstacles, pipelines, starting points, and ending points, and obtain 3D point cloud data by combining three-dimensional coordinates in space. Module M3: Process the feature of the point cloud data set of the scenario and divide it into multiple small regions according to a preset ratio. Module M4: Analyze the objects in the small region scenario and score according to the function to obtain the encoded vector indicating the spatial characteristics of the region. Including: Using the obstacle distribution attribute as the first component of the encoding, dividing it into two categories according to whether there are pipeline segments in the region space, and taking the occupancy ratio of pipelines and obstacles in the space as the scoring basis for the obstacles in the current space region; Using the spatial attribute of the pipeline segment as the second component of the encoding, calculating the number of pipeline distributions in the region, and taking the relative position of the starting and ending points as the weight ratio of the region to statistically score the pipeline length of the current small region path; For the pipeline relationship in the small region, represent the relationship between pipelines through the inner product of the starting and ending point vectors of different label pipelines. Module M5: After obtaining the feature encoding vector of each small region, splice it into a one-dimensional vector in spatial order as the feature encoding vector of the entire scene region. Module M6: Perform normalization and standardization data processing on the feature encoding vector of the entire scene region as the input of the neural network. Module M7: Use a fully connected neural network as the regression model framework to perform regression on the pipeline space scenario data set, including: taking the expert experience score as the target value, the spatial encoding vector as the input value, and defining the MSE as the loss function. Module M8: Use evaluation indicators to evaluate the regression effect of the regression model. Module M9: Use the hold-out method to divide the data set into a training set and a test set, train the regression model, and test the accuracy, effectiveness, and stability of the regression model through the test set.

7. The intelligent evaluation system for the layout of a pipeline system based on a neural network according to claim 6, characterized in that The module M4 includes: First, using the obstacle distribution attribute as the first component of the encoding, dividing it into two categories according to whether there are pipeline segments in the region space, and taking the occupancy ratio of pipelines and obstacles in the space as the scoring basis. The expression is: Among them, Score obs represents the score of obstacles in the current spatial region, and N total represents the number of spatial regions, and N obs represents the number occupied by the obstacles; Second, using the spatial attribute of the pipeline segment as the second component of the encoding, calculating the number of pipeline distributions in the region, and taking the relative position of the starting and ending points as the weight ratio of the region to statistically score the pipeline length. The expression is: Among them, Score path represents the path score of the current small area, N path represents the number of paths occupied, and α is the correction coefficient for path scoring based on the relative spatial position; represents the central position vector of the current spatial area, represents the starting point vector corresponding to this path, represents the ending point vector corresponding to this path; For the pipeline relationship in the small region, represent the relationship between pipelines through the inner product of the starting and ending point vectors of different label pipelines. The expression is: Among them, Score dir represents the pipeline routing intersection score, X i represents the start and end vector of a pipeline within this area, X j represents the start and end vector of another pipeline within this area.

8. The intelligent evaluation system for the layout of the pipeline system based on a neural network according to claim 6, wherein, The module M6 includes: Perform normalization and standardization data processing on the feature encoding vector of the entire scene region. The expression is: where x i is the i-th data in the dataset, represents the dataset mean, and x std represents the dataset standard deviation.

9. The intelligent evaluation system for pipeline system layout based on neural network according to claim 8, characterized in that, The module M7 includes: Taking the expert experience score as the target value, the spatial encoding vector as the input value, and defining the MSE as the loss function. Its basic regression model is as follows: z i = Wx i + b y i = f(z) Among them, z i is the output value after neuron mapping, f() represents the activation function, and y i represents the predicted value after neural network fitting, represents the true value obtained through expert scoring, n is the number of samples, W is the input weight, and b is the bias term.

10. The intelligent evaluation system for pipeline system layout based on neural network according to claim 9, characterized in that The module M8 includes: Adopt evaluation indicators mainly based on the goodness of fit R2 and supplemented by the root mean square error RMSE. The expression is: In the formula, is the average value of the dataset scores.

Images