Sewage pipeline risk grade evaluation method based on Bayesian neural network

Abstract

The invention discloses a sewage pipeline risk grade evaluation method based on a Bayesian neural network. The method comprises the following steps: 1, obtaining each evaluation index of a to-be-evaluated pipeline, a grade corresponding to each evaluation index and a corresponding prior probability according to an existing standard value obtained through expert scoring; 2, obtaining data of all evaluation indexes according to ArcGIS basic attributes; 3, detecting the to-be-evaluated pipeline by adopting the CCTV robot to obtain pipeline disease data; 4, obtaining an actual pipeline risk grade of the to-be-evaluated pipeline according to the pipeline disease data; and 5, training the Bayesian neural network, and obtaining a predicted value of the to-be-evaluated pipeline risk level through the trained Bayesian neural network according to the actual pipeline risk level. According to the method, rules in data can be captured, and uncertainty in a complex system can be quantitatively predicted, so that relatively stable and accurate risk assessment is provided, and a scientific basis is provided for pipe network maintenance and risk management.

Description

Technical Field

[0001] The invention relates to the technical field of computer-aided design, and in particular to a sewage pipe risk level evaluation method based on a Bayesian neural network. Background Art

[0002] The unclear number and condition of urban drainage pipeline networks are huge challenges and difficulties faced in the current drainage network scheduling, operation and maintenance, repair and renovation work. Therefore, in recent years, the country has attached great importance to the inspection, construction, operation, supervision and management of urban drainage pipeline networks.

[0003] Due to incomplete considerations at the beginning of construction and lack of systematic maintenance of drainage pipes, drainage pipes in many cities in my country have gradually developed problems, and some have even caused serious accidents. These problems are on the one hand affected by topography, rainfall, human activities, and many uncertain factors in water quality and quantity. On the other hand, it is because my country's urban drainage system is still imperfect, management and maintenance are not timely, and pipeline transportation capacity is insufficient.

[0004] In order to find out the healthy operation rules of the drainage network and find out the health status and safety risk sources of the pipelines, we must first determine the main influencing factors and influencing mechanisms. Through a century of research by domestic and foreign scholars, it is found that the integrity and health of drainage pipelines are closely related to the pipeline's own factors, mainly including pipe diameter, pipe age, pipe material, etc.

[0005] However, the current routine inspection and maintenance of sewage pipes is still mainly based on random inspections and rotational inspections. It is impossible to judge the risks and hidden dangers in the pipelines, and there is no standardized and orderly "prediction-diagnosis-feedback" process.

[0006] Therefore, how to use attribute indicators, environmental impact indicators, etc. as sewage pipe risk level evaluation indicators, and predict the risk level of pipes in the region through the Bayesian neural network algorithm, so as to grasp the health status of the pipeline network in advance and put forward targeted inspection and rectification opinions has become a technical problem that technical personnel in this field urgently need to solve. Summary of the invention

[0007] In view of the above-mentioned defects of the prior art, the present invention provides a sewage pipe risk level evaluation method based on Bayesian neural network, the purpose of which is to achieve an organic combination of data-driven and probabilistic reasoning and improve the accuracy and robustness of prediction.

[0008] To achieve the above object, the present invention discloses a sewage pipe risk level evaluation method based on a Bayesian neural network, comprising the following steps:

[0009] Step 1: Obtain the evaluation indicators of the pipeline to be evaluated, the level corresponding to each evaluation indicator, and the corresponding prior probability according to the existing standard values ​​obtained through expert scoring;

[0010] Step 2: Obtain the data of all the evaluation indicators based on the ArcGIS basic attributes;

[0011] Step 3: Use a CCTV robot to detect the pipeline to be evaluated to obtain pipeline disease data;

[0012] Step 4: Obtain the actual pipeline risk level of the pipeline to be evaluated according to the pipeline disease data;

[0013] Step 5: Train the Bayesian neural network, and then obtain the predicted value of the pipeline risk level of the pipeline to be evaluated through the trained Bayesian neural network according to the actual pipeline risk level.

[0014] Preferably, in Step 1, the prior probability of the pipeline to be evaluated is calculated using Bayes' formula.

[0015] More preferably, in Step 5, the Bayesian neural network is trained using the mapping relationship between each evaluation indicator in the standard values and the pipeline risk level.

[0016] More preferably, the grades of each evaluation indicator are used as the distribution of the posterior probability for comprehensive judgment. Specifically: Using the Bayesian posterior probability calculation formula, combining the prior probability of the grade of each evaluation indicator and the corresponding weights, the pipeline risk level is determined through the maximum probability principle.

[0017] More preferably, through the trained Bayesian neural network, according to the mapping relationship between each evaluation indicator and the pipeline risk level, and using the Bayesian posterior probability calculation formula, combining the prior probability of each evaluation indicator and the corresponding weights, the pipeline risk level is determined through the maximum probability principle, and the predicted value of the pipeline risk level of the pipeline to be evaluated is output finally.

[0018] More preferably, in Step 1, the difference between the data and the actual value of the evaluation indicator of the pipeline to be evaluated is represented by a normal distribution function, and the conditional probability P(N j,i / M j,i ) that the jth health evaluation indicator value is at level i is calculated according to the normal distribution principle of errors. The calculation formula is as follows:

[0019] P(N j,i / M j,i ) = 2(1 - φ(|t j,i |));

[0020] where i is the level of the evaluation indicator, i = 1, 2, 3, 4...;

[0021] $j$ is the $j$-th evaluation index, where $j = 1, 2, 3, 4, \cdots, S$;

[0022] $\varphi(|t$ j,i $|)$ represents the normal distribution probability that the $j$-th evaluation index is at level $i$;

[0023]

[0024] $t$ j,i represents the standardized value that the $j$-th evaluation index is at level $i$;

[0025]

[0026] $N$ j,i represents the actual observed value or standard value that the $j$-th evaluation index is at level $i$;

[0027] $M$ j,i represents the data value that the $j$-th evaluation index is at level $i$, and the value is the average value of the index within the range of level $i$;

[0028] $\sigma$ j,i $= Cv$ j,i $\cdot M$ j,i ;

[0029]

[0030] $\sigma$ j,i represents the sample standard deviation of the standard value that the $j$-th evaluation index is at level $i$;

[0031]

[0032] $a$ j,i represents the mean value of the standard value that the $j$-th evaluation index is at level $i$;

[0033]

[0034] More preferably, the posterior probabilities of each evaluation index at different levels are obtained through the Bayesian posterior probability calculation formula. The pipeline risk status prediction and pipeline health evaluation model based on Bayesian is:

[0035]

[0036] j,i ), where $P(M$

[0037] $P(N$ j,i $ / M$ j,i $)$ is the conditional probability that the $j$-th evaluation index is at level $i$;

[0038] $P(M$ j,i / N j,i ) is the posterior probability that the j-th evaluation index is at level i.

[0039]

[0040] Among them, is the upper limit value of the j-th evaluation index at level i;

[0041] is the lower limit value of the j-th evaluation index at level i;

[0042] is the upper limit value of the j-th evaluation index at level IV;

[0043] is the lower limit value of the j-th evaluation index at level IV;

[0044] is the upper limit value of the j-th evaluation index at level I;

[0045] is the lower limit value of the j-th evaluation index at level I.

[0046] More preferably, the coefficient of variation method is used to calculate the weights of the evaluation indexes as follows:

[0047]

[0048] Among them, D j is the standard deviation of the standard values of each level of the j-th evaluation index;

[0049] W j is the weight of the j-th evaluation index.

[0050] More preferably, a relevant connection is established between the pipeline risk level and the repair coefficient RI in the "Technical Specification for Inspection and Assessment of Urban Drainage Pipelines" (CJJ181 - 2012) as follows:

[0051] When RI is less than or equal to 1, the pipeline risk level is I, and the corresponding repair suggestion and description are: "The structural condition is basically intact, no repair is required";

[0052] When RI is greater than 1 and less than or equal to 4, the pipeline risk level is II, and the corresponding repair suggestion and description are: "The structure will not be damaged in the short term, but a repair plan should be made";

[0053] When RI is greater than 4 and less than or equal to 7, the pipeline risk level is III, and the corresponding repair suggestion and description are: "The structure may be damaged in the short term, and it should be repaired as soon as possible";

[0054] When the RI is greater than 7, the pipeline risk level is IV, and the corresponding repair suggestions and instructions are: "The structure has been or is about to be damaged, and it should be repaired immediately."

[0055] Advantages of the present invention:

[0056] The present invention integrates expert experience, Bayesian posterior, and neural network algorithms, which can not only capture the patterns in the data but also quantify and predict the uncertainties in complex systems, thereby providing a relatively robust and accurate risk assessment and providing a scientific basis for pipeline network maintenance and risk management.

[0057] The following will further illustrate the concept, specific structure, and technical effects of the present invention with reference to the accompanying drawings to fully understand the purpose, features, and effects of the present invention. Description of the Drawings

[0058] Figure 1 A logic diagram showing an embodiment of the present invention. Detailed Embodiment

[0059] Embodiment

[0060] As Figure 1 shown, a method for evaluating the risk level of a sewage pipeline based on a Bayesian neural network; characterized by comprising the following steps:

[0061] Step 1: Obtain each evaluation index, the corresponding level, and the corresponding prior probability of the pipeline to be evaluated according to the standard values obtained by expert scoring.

[0062] Step 2: Obtain the data of all evaluation indexes according to the basic attributes of ArcGIS.

[0063] Step 3: Use a CCTV robot to detect the pipeline to be evaluated to obtain pipeline disease data.

[0064] Step 4: Obtain the actual pipeline risk level of the pipeline to be evaluated according to the pipeline disease data.

[0065] Step 5: Train the Bayesian neural network, and then obtain the predicted value of the risk level of the pipeline to be evaluated through the trained Bayesian neural network according to the actual pipeline risk level.

[0066] Based on the evaluation of the operation status of the sewage system through relevant indicators such as water flow and water quality, the present invention conducts a prediction of the pipeline risk level based on the Bayesian neural network model for the areas or branches with problems. The Bayesian neural network combines data information and prior probability information, uses the neural network for preliminary risk level prediction, and corrects and optimizes the prediction results through Bayesian theory, realizing the organic combination of data-driven and probability reasoning, and improving the accuracy and robustness of the prediction.

[0067] The standard values obtained by expert scoring are specifically shown in Table 1 below:

[0068] Table 1 Grade standard values of each evaluation index of the pipeline

[0069]

[0070]

[0071] It can be seen from Table 1 that among the 6 evaluation indicators, 3 evaluation indicators, namely pipeline material, pipeline grade, and road grade, are qualitative indicators and need to be set manually through expert experience. Therefore, this study refers to relevant papers and research to score the qualitative indicators, setting the highest value as 10 points and the lowest value as 0 points. The higher the score, the better the health status.

[0072] In some embodiments, in step 1, the Bayesian formula is used to calculate the prior probability of the pipeline to be evaluated.

[0073] The Bayesian formula is used to calculate the prior probability of the pipeline, as shown in Table 2.

[0074] Table 2 Prior probabilities of the grades of the pipeline network health evaluation indicators

[0075]

[0076] In some embodiments, in step 5, the mapping relationship between each evaluation indicator in the standard value and the pipeline risk level is used to train the Bayesian neural network.

[0077] In practical applications, after the training is completed by the above means, inputting any evaluation indicator can obtain the pipeline risk level through the forward propagation process, which can automatically extract information and features from the data to achieve effective grade division and form a data-based pipeline risk level.

[0078] In some embodiments, the grades of each evaluation indicator are used as the distribution of the posterior probability for comprehensive judgment. Specifically: using the Bayesian posterior probability calculation formula, combining the prior probability of the grades of each evaluation indicator and the corresponding weights, the pipeline risk level is determined through the maximum probability principle.

[0079] In practical applications, the above technical means can ensure the quantification and evaluation of uncertainties, provide more robust judgments, and form a pipeline risk level based on probability.

[0080] In some embodiments, through a trained Bayesian neural network, according to the mapping relationship between each evaluation index and the pipeline risk level, and by using the Bayesian posterior probability calculation formula, combined with the prior probability of each evaluation index and the corresponding weights, the pipeline risk level is determined through the maximum probability principle, and the predicted value of the pipeline risk level for the pipeline to be evaluated is output.

[0081] In practical applications, the above technical means can effectively integrate data information and probability information, enhancing the performance and reliability of the model in practical applications.

[0082] In some embodiments, in step 1, the difference between the data and the actual value of the evaluation index of the pipeline to be evaluated is represented by a normal distribution function, and the conditional probability P(N j,i / M j,i ) that the jth health evaluation index value is at level i is calculated through the normal distribution principle of errors. The calculation formula is as follows:

[0083] P(N j,i / M j,i ) = 2(1 - φ(|t j,i |));

[0084] Where i is the level of the evaluation index, i = 1, 2, 3, 4...;

[0085] j is the jth evaluation index, j = 1, 2, 3, 4... S;

[0086] φ(|t j,i ) represents the normal distribution probability that the jth evaluation index is at level i;

[0087]

[0088] t j,i represents the standardized value that the jth evaluation index is at level i;

[0089]

[0090] N j,i represents the actual observed value or standard value that the jth evaluation index is at level i;

[0091] M j,i represents the data value that the jth evaluation index is at level i, and the value is the average value of the index within the range of level i;

[0092] σ j,i = Cvj,i ·M j,i ;

[0093]

[0094] σ j,i represents the sample standard deviation of the standard value of the j-th evaluation index at level i;

[0095]

[0096] a j,i represents the mean of the standard value of the j-th evaluation index at level i;

[0097]

[0098] In some embodiments, the posterior probabilities of each evaluation index at different levels are obtained through the Bayesian posterior probability calculation formula. The Bayesian-based pipeline risk status prediction and pipeline health evaluation model is:

[0099]

[0100] where P(M j,i ) is the prior probability of the j-th evaluation index at level i;

[0101] P(N j,i / M j,i ) is the conditional probability of the j-th evaluation index at level i;

[0102] P(M j,i / N j,i ) is the posterior probability of the j-th evaluation index at level i.

[0103] According to the hierarchical idea, the prior probabilities of the evaluation indexes are standardized into the upper limit value interval of level I and level IV for the upper and lower limit values under different levels.

[0104]

[0105] where, is the upper limit value of the j-th evaluation index at level i;

[0106] is the lower limit value of the j-th evaluation index at level i;

[0107] is the upper limit value of the j-th evaluation index at level IV;

[0108] is the lower limit value of the j-th evaluation index at level IV;

[0109] is the upper limit value of the j-th evaluation index at level I;

[0110] is the lower limit value of the j-th evaluation index at level I

[0111] In some embodiments, the coefficient of variation method is used to calculate the weights of the evaluation indexes, specifically as follows:

[0112]

[0113] where D j is the standard deviation of the standard values of each level of the j-th evaluation index;

[0114] W j is the weight of the j-th evaluation index.

[0115] In practical applications, the weights of the evaluation indexes will directly affect the objectivity of the results. At present, the determination of the weights of the evaluation indexes is mainly divided into the subjective assignment method and the objective weighting method. The subjective assignment method assigns weights through subjective judgments of the evaluation indexes, and then obtains the index weights through mathematical calculations. The objective weighting method is a weight calculation method based on the statistical data of the indexes, such as the coefficient of variation method, the entropy weight method, etc. The present invention uses the coefficient of variation method to calculate the weights of the evaluation indexes. The coefficient of variation method is an objective weighting method that can eliminate the influence of different dimensions between the evaluation indexes and measure the degree of difference in the values of the evaluation indexes.

[0116] In some embodiments, a relevant connection is established between the pipeline risk level and the repair coefficient RI in the "Technical Specification for Inspection and Assessment of Urban Drainage Pipelines" (CJJ181-2012), specifically as follows:

[0117] When RI is less than or equal to 1, the pipeline risk level is I, and the corresponding repair suggestion and description are: "The structural condition is basically intact, no repair is required";

[0118] When RI is greater than 1 and less than or equal to 4, the pipeline risk level is II, and the corresponding repair suggestion and description are: "The structure will not be damaged in the short term, but a repair plan should be made";

[0119] When RI is greater than 4 and less than or equal to 7, the pipeline risk level is III, and the corresponding repair suggestion and description are: "The structure may be damaged in the short term, and it should be repaired as soon as possible";

[0120] When RI is greater than 7, the pipeline risk level is IV, and the corresponding repair suggestion and description are: "The structure has been or is about to be damaged, and it should be repaired immediately."

[0121] In practical applications, the probability-based pipeline risk level is regarded as the result of comprehensively judging the posterior probabilities of each index level on the basis of the prior probability. The posterior probabilities of each index at different levels are obtained through the Bayesian posterior probability calculation formula, and on the basis of considering the weights of each index, the pipeline risk level is determined according to the maximum probability principle.

[0122] The pipeline risk level based on data can be regarded as a certain mapping relationship between the pipeline risk level obtained through data mining and each evaluation index. After the neural network model is trained, any pipeline evaluation index is input, and the corresponding pipeline risk level can be obtained through the forward propagation process of the neural network. The pipeline risk levels obtained by these two different methods are input into the trained neural network, so as to obtain the pipeline risk level finally predicted by the model, and realize the effective fusion of data information and prior probability information.

[0123] At present, there is still a lack of certain standards for the classification of pipeline health status. Some scholars believe that the classification of pipeline health status should be based on whether repair is needed as a reference, and the pipeline status is divided into five grade standards: good condition, routine monitoring, key monitoring, planned update, and immediate update. There are also those who believe that it should be directly divided into six grade standards: very good, very good, good, general, poor, and very poor according to the pipeline evaluation results.

[0124] In this study, since it is necessary to use the pipeline CCTV detection data for the determination of posterior probability and index weight, it is considered to establish a relevant connection between the pipeline risk level and the repair coefficient RI in the "Technical Specification for Inspection and Assessment of Urban Drainage Pipelines" (CJJ181-2012). The selection of the RI index mainly considers two aspects: ① This index is used to evaluate the structural defects of pipelines, such as: rupture, disconnection, misalignment, leakage, etc., and these defects are closely related to the abnormal water volume and water quality involved in the evaluation of the operation status of the sewage system in Chapter 3; ② In the future, during the general survey and repair of the Shanghai pipe network, RI is also the most important reference index and will appear in the inspection report. Therefore, it is more rapid and universal to use the RI value to iterate the risk assessment model.

[0125] The pipeline risk level is divided based on the repair coefficient of the pipe section as shown in Table 3.

[0126] Table 3 Pipeline Risk Level Classification Table

[0127] Level Repair Index RI Repair Suggestions and Instructions I RI ≤ 1 The structural conditions are basically intact and do not need to be repaired II 1 < RI ≤ 4 The structure will not be damaged in the short term, but a repair plan should be made III 4 < RI ≤ 7 The structure may be damaged in the short term and should be repaired as soon as possible IV RI > 7 The structure has been or is about to be damaged and should be repaired immediately

[0128] The preferred specific embodiments of the present invention have been described in detail above. It should be understood that those of ordinary skill in the art can make many modifications and variations based on the concept of the present invention without creative efforts. Therefore, all technical solutions that can be obtained by those skilled in the art in the technical field based on the concept of the present invention through logical analysis, reasoning or limited experiments on the basis of the prior art shall fall within the protection scope determined by the claims.

Claims

1. A sewage pipeline risk level evaluation method based on Bayesian neural network; characterized in that: The steps include: Step 1: Obtain the evaluation indicators of the pipeline to be evaluated, the level corresponding to each evaluation indicator, and the corresponding prior probability according to the existing standard values ​​obtained through expert scoring; Step 2: Obtain data of all evaluation indicators according to ArcGIS basic attributes; Step 3: Using a CCTV robot to detect the pipeline to be evaluated to obtain pipeline disease data; Step 4: obtaining the actual pipeline risk level of the pipeline to be evaluated according to the pipeline disease data; Step 5: Train the Bayesian neural network, and then obtain a predicted value of the pipeline risk level to be evaluated according to the actual pipeline risk level through the trained Bayesian neural network.

2. The sewage pipeline risk level evaluation method based on Bayesian neural network according to claim 1 is characterized in that: In step 1, the prior probability of the pipeline to be evaluated is calculated using the Bayesian formula.

3. The sewage pipe risk level evaluation method based on Bayesian neural network according to claim 2 is characterized in that: In step 5, the Bayesian neural network is trained using the mapping relationship between each of the evaluation indicators in the standard value and the pipeline risk level.

4. The method for evaluating the risk level of sewage pipes based on Bayesian neural network according to claim 3 is characterized in that: The level of each evaluation indicator is used as the distribution of posterior probability for comprehensive judgment, specifically: using the Bayesian posterior probability calculation formula, combined with the prior probability of the level of each evaluation indicator, and the corresponding weight, the pipeline risk level is determined by the maximum probability principle.

5. The method for evaluating the risk level of sewage pipes based on Bayesian neural network according to claim 4 is characterized in that: Through the trained Bayesian neural network, according to the mapping relationship between each evaluation indicator and the pipeline risk level, and using the Bayesian posterior probability calculation formula, combined with the prior probability of each evaluation indicator, and the corresponding weight, the pipeline risk level is determined by the maximum probability principle, and the final predicted value of the pipeline risk level for the pipeline to be evaluated is output.

6. The method for evaluating the risk level of sewage pipes based on Bayesian neural network according to claim 5 is characterized in that: In step 1, the difference between the data of the evaluation index of the pipeline to be evaluated and the actual value is represented by a normal distribution function, and the conditional probability P(N) that the jth health evaluation index value is at level i is calculated by the normal distribution principle of the error. j,i / M j,i ), the calculation formula is as follows: P(N j,i / M j,i )=2(1-φ(|t j,i |)); Wherein, i is the level of the evaluation index, i=1, 2, 3, 4...; j is the jth evaluation index, j=1, 2, 3, 4...S; φ(|t j,i |) represents the normal distribution probability that the jth evaluation index is at level i; t j,i Indicates the standardized value of the jth evaluation indicator at level i; N j,i Indicates the actual observed value or standard value of the jth evaluation index at level i; M j,i Indicates the data value of the jth evaluation indicator at level i, and the value is the average value of the indicator within the range of level i; s j,i =Cv j,i ·M j,i ; σ j,i It represents the sample standard deviation of the standard value of the jth evaluation indicator at level i; a j,i It represents the mean of the standard values ​​of the jth evaluation index at level i; 7. The method for evaluating the risk level of sewage pipes based on Bayesian neural network according to claim 6 is characterized in that: The posterior probability of each evaluation index being at different levels is obtained through the Bayesian posterior probability calculation formula. The Bayesian-based pipeline network risk status prediction and pipeline health evaluation model is: Among them, P(M j,i ) is the prior probability that the jth evaluation index is at level i; P(N j,i / M j,i ) is the conditional probability that the jth evaluation index is at level i; P(M j,i / N j,i ) is the posterior probability that the jth evaluation index is at level i. in, is the upper limit value of the jth evaluation index at level i; is the lower limit value of the jth evaluation index at level i; is the upper limit value of the jth evaluation index at level IV; is the lower limit value of the jth evaluation index at level IV; is the upper limit value of the jth evaluation index at level I; is the lower limit value of the j-th evaluation index at level I.

8. The method for evaluating the risk level of sewage pipes based on Bayesian neural network according to claim 7 is characterized in that: The coefficient of variation method is used to calculate the weight of each evaluation index, as follows: Among them, D j is the standard deviation of the standard values ​​of each level at the jth evaluation index; W j is the weight of the j-th evaluation index.

9. The method for evaluating the risk level of sewage pipes based on Bayesian neural network according to claim 8 is characterized in that: The pipeline risk level is related to the repair coefficient RI in the Technical Code for Inspection and Evaluation of Urban Drainage Pipelines (CJJ181-2012), as follows: When RI is less than or equal to 1, the pipeline risk level is I, and the corresponding repair suggestions and instructions are: "The structural condition is basically intact, no repair"; When RI is greater than 1 and less than or equal to 4, the pipeline risk level is II, and the corresponding repair suggestions and instructions are: "The structure will not be damaged in the short term, but a repair plan should be made"; When RI is greater than 4 and less than or equal to 7, the pipeline risk level is III, and the corresponding repair suggestions and instructions are: "The structure may be damaged in the short term and should be repaired as soon as possible"; When RI is greater than 7, the pipeline risk level is IV, and the corresponding repair suggestions and instructions are: "The structure has been or is about to be damaged and should be repaired immediately."

Images