Safety failure probability of urban gas pipelines based on Bayesian network Hao Yongmei 1 Xing Zhixiang 1 Shen Ming 1 Shao Hui 1 Wang Xusheng 2 1. School of Environmental and Safety Engineering, Changzhou University, Changzhou 213164, Jiangsu; International Business Department of China National Petroleum Pipeline Bureau, Langfang Hebei 065000, Hao Yongmei, etc. Safety failure probability of urban gas pipeline based on Bayesian network. Oil and gas storage and transportation, 2012, 1 application improves the effectiveness of Bayesian network-based risk prediction. A failure probability analysis model of urban gas pipelines based on Bayesian network was established. Using HUGIN and MSBNX software tools, combined with a city's natural gas pipeline case, the safety failure probability of polymorphic fault top events and the structural importance of each failure factor were calculated. Using the reasoning ability of BN, single-factor and double-factor corrections were made to the natural damage factors and corrosion factors that caused the pipeline to fail. The modified Bayesian network model is more realistic and has better practical significance for improving the systematicity, predictability and accuracy of the quantitative analysis of the safety failure of urban gas pipelines, and also fully shows that the Bayesian network is handling complex system risks Unique superiority and applicability in the analysis. With the acceleration of urban expansion and the aging of gas pipeline networks, urban gas fires, explosions, poisoning and other vicious accidents have occurred from time to time, and have become the third biggest killer after traffic accidents and industrial accidents. Therefore, it is necessary to combine modern technology to strengthen the risk prediction research on leakage accidents of urban gas pipeline networks, especially according to the actual situation of China's urban gas pipelines, and to combine this research with quantitative analysis in a targeted manner to truly make the pipeline risk The evaluation is systematic, accurate and predictable. The progress of quantitative risk assessment technology for gas pipelines mainly focuses on the study of the possibility of pipeline failure. There are many research achievements by foreign scholars. For example, the quantitative assessment of corrosion defects by Kiefner and Vieth was finally adopted by ASMEB31G in the United States for evaluating single corrosion defects of pipelines. Afterwards, many scholars have improved it, from the initial single study of corrosion-induced pipeline life prediction, to the transition to consider the reliability of the pipeline under the combined action of various factors in the surrounding environment of the pipeline. Cagno et al. Proposed a robust Bayesian method, which can be used to evaluate the failure probability of urban gas low-pressure gas distribution network composed of steel pipelines and used as the basis for pipeline maintenance. In recent years, domestic scholars have also begun to study the calculation of gas leakage consequences and risk analysis methods. Pan Jiahua conducted a more in-depth quantitative analysis of the third-party damage, corrosion, design, operation and other factors that caused the pipeline accident; Wang Kaiquan et al. Improved the Kent scoring method suitable for urban natural gas pipeline risk analysis based on the characteristics of urban high-pressure natural gas pipelines, and It has been applied to actual projects and proved to be an effective tool for risk assessment of urban high-pressure natural gas pipelines. At present, China has put forward a series of practical risk assessment methods for urban gas pipeline risks, and has conducted fruitful research on pipeline corrosion remaining life, pipeline reliability under multiple factors, and pipeline quantitative risk assessment. With the increasing complexity of modern engineering technology systems and the continuous improvement of automation, the quantitative risk analysis of gas pipeline accidents often has polymorphism, non-monotonicity, failure correlation, timing, interaction between process variables and component states, and non-determination Features such as sexual logic, software impact, human interaction, diversity of information, incompleteness and uncertainty. The commonly used accident tree and event tree methods have encountered many difficulties in terms of models and methods when dealing with the safety assessment of complex systems with the above characteristics; the existing probabilistic safety assessment methods cannot fully reflect the software, human factors and system Interdependence and interaction processes are restricted in practical engineering applications. The Bayesian network that can effectively combine qualitative knowledge such as expert experience judgment and quantitative knowledge such as historical data has shown unique advantages in risk analysis of large and complex systems, and is currently used to deal with uncertainties based on probabilistic knowledge. The most powerful method of inference is widely used. 1 Bayesian network (BN) The Bayesian network analysis method is to collect information and conduct a risk analysis to estimate the consequences and severity, determine the probability of the initial event, and use the Bayesian method to estimate its occurrence Probability, prioritize risks, and compare with acceptable risk criteria to make safety recommendations. Bayesian network working principle Probability analysis of urban gas pipeline safety failure 2.1 Inducing factors of urban gas pipeline accidents According to the statistical data of accident inducement of gas pipelines in the United States, Canada and Europe and other countries and regions, although the proportion of each accident cause in different countries is different However, the main causes of the accident are external influences (third party damage, foundation movement, etc.), corrosion, materials and construction defects. According to the statistics of urban gas pipeline accidents in Shanghai from 2000 to 2001, third-party damage accounted for about 50%, followed by construction and material defects, corrosion and management errors. It can be seen that the main incentives for the failure of gas pipelines in China, Europe and the United States and other countries are the same, that is, third-party damage is dominant, and third-party damage can be divided into natural factors and human factors. According to the database of probability of pipeline failure probability such as 808, the main factors causing pipeline failure are: third-party damage, corrosion, weld and pipe defects, management operation defects and natural disasters. 2.2 Bayesian network model According to the main inducing factors of pipeline failure, construct its failure Bayesian network diagram. Among them: indicates the failure of the pipeline, A indicates that the pipeline is defective, 圮 indicates that the pipeline is severely corroded, and indicates third-party damage. Basic events: 1 indicates poor pressure bearing capacity of the pipeline; 2 indicates severe pressure holding in the pipeline; 3 indicates defects in construction management; 4 indicates initial defects in the pipeline; 5 indicates corrosion; 6 indicates poor corrosion resistance of the pipeline; 7 indicates man-made damage; 8 indicates natural disaster And other external forces. 3 Case analysis The natural gas pipeline of a north main line of a city was built in 1987 with a design pressure of 4.0 MPa, and the current output is 700X104m3 / d. The occurrence probability of the basic event of failure caused by the north main line (Table 1) refers to the relevant accident statistics (including: The incidents "man-made destruction" and "poor corrosion resistance" were obtained by expert scoring method). Table 1 Probability of basic events for failure of a natural gas pipeline trunk Event name Probability distribution (prior probability) Pipeline pressure energy difference Pipeline severe pressure suppression Construction management defects Pipeline initial defects Corrosion pipeline Poor corrosion resistance Man-made damage Natural disasters and other external forces 3.1 MSBNX software Calculation MSBNX is a window interface software developed by Microsoft. It is easy to operate and has high calculation accuracy. It also provides an API interface for VB to call; but it can only model and analyze discrete variables. Users cannot choose inference algorithms and do not support structural learning and parameters. Learn. Using this software, a Bayesian network model of urban gas pipelines was established. Calculate the top event: the occurrence of T. Enter the probability value of each basic event (Table 1), and the probability of failure of the gas pipeline is 0.0627 (where state0 is the basic state of the MSBNX software to calculate the probability of the top event). Node name MSBNX model The probability distribution histogram of each node uses HuginExpert software to establish a Bayesian network model of urban gas pipelines. Importing the probability values ​​of each factor in Table 1, the probability of occurrence of overhead events is 6.27X10-2, and from this Obtain the structural importance ranking when the basic events lead to the failure of the pipeline: Yi 8> 1> 3.3.1 Amendments to natural factors. The human factors in the third-party damage factors that cause the failure of the gas pipeline have no rules to follow, so the main Correct the natural damage factors. Yu Shurong's research shows that when only the natural factors in the third-party damage cause the pipeline to fail, the logical uncertainty should be corrected. From this, the conditional probability table = = 1 = 7 = 7, = 0.5. After the model is modified, r becomes 4.98X10-2, and the structural importance of each failure factor when it causes pipeline failure is ranked:> 8> X7> X4> X3> X2> X5> X6. 3.3.2 Single-factor multi-state correction 3.3.2.1 Multi-state correction for corrosion factors (1) Pan Jiahua ’s research shows that the internal corrosion of pipelines can be divided into strong corrosion, Moderate corrosion, special corrosion (corrosion that occurs only under special circumstances) and no corrosion, it can be seen that the corrosion occurring in the actual operation of the pipeline includes at least corrosion, non-corrosion and slight corrosion. Suppose that the occurrence of X5 has the above three states, and their probabilities are X5i = 0.018. The structural importance of the factors that cause the pipeline to fail is the same as before the correction. It is still the same. The four-state correction of the corrosion factor is assumed. Assuming the occurrence of X5 There are four states: severe corrosion, non-corrosion, slight corrosion, and special corrosion, and their probabilities are X5i = 0.010, X52 = 0.9759, X53 = 0.0960, X; 4 = 0.0081. After the correction, r is 6.25X102. The order of structural importance is the same as before the correction, and it is still A >> From the above single-factor polymorphic correction results, it can be seen that when the polymorphic correction of the corrosion factor is to modify the probability of X51, its three-state and four-state corrections The probability value of the overhead event is close to the probability value of the two-state overhead event, and the structural importance of each failure factor remains unchanged. 3.3.2.2 Polymorphic correction of corrosion factors (2) Based on the above results, the probability of X52 is polymorphic corrected. Assuming that X5 includes corrosion, non-corrosion and slight corrosion, the probabilities are X51 = 0.0241, X52 = 0.700, X53 = 0.2759. After the correction, r becomes 6.54X10-2, and the structural importance of each failure factor is ranked: Similarly, assuming that X5 has four states of severe corrosion, non-corrosion, slight corrosion, and special corrosion, the probability is X5i = 0.0241 , X52 = 0.690, X53 = 0.2759, X54 = 0.010. After modifying the model, r becomes 6.55X10-2, the structural importance of each failure factor is ranked: it can be seen that the basic event success probability part is modified, causing the top event probability The change sensitivity is greater, and the structural importance of each failure factor also changes significantly. 3.3.3 Two-factor multi-state correction According to the analysis results of a large number of urban gas pipeline accidents, among the factors that cause pipeline failure, sometimes a single factor can cause pipeline accidents, but more pipeline accidents are caused by a combination of multiple factors. . Therefore, based on the previous research results and existing basic data, the natural factors and corrosion factors are corrected and combined to calculate the top event. 3.3.3.1 Two-factor correction (1) First, the logical uncertainty correction of natural factors and the three-state correction of corrosion factors (1) are combined to calculate the probability of the occurrence of overhead event r is 4.98X10-2, each failure factor Priority order of the structure: combining the logical uncertainty correction of natural factors with the four-state correction of corrosion factors (1), the probability of the occurrence of the top event r is 4.98X10-2, and the structure of each factor that causes pipeline failure is important Degree order: 3.3.3.2 Two-factor correction (2) When the logical uncertainty correction of natural factors is combined with the three-state correction of corrosion factors (2), the probability of the occurrence of overhead event r is 5.25X10-2, resulting in When the structural importance of each factor of pipeline failure combines the logical uncertainty correction of natural factors with the four-state correction of corrosion factors (2), the top event: the probability of T occurring will become 5.26X10-2, resulting in The structural importance of each factor of pipeline failure is ranked. Therefore, when the logical correction of natural factors is combined with the polymorphic correction of corrosion factors (1), the top event is obtained: the probability of occurrence of T and the structure of each failure factor The importance is the same as the result when only natural factors are corrected; when the logical correction of natural factors is combined with the polymorphic correction of corrosion factors (2), the failure probability of overhead events and the structural importance of each failure factor have occurred Major changes. 4 Conclusion Based on the Bayesian network's ability to describe the event polymorphism and the indeterminacy of the fault logic relationship, a Bayesian network model of urban gas pipeline failure is established. This state and situation were explored. Using Bayesian network software tools, it is possible to efficiently calculate the safety failure probability of gas pipelines; the natural factors and corrosion factors that cause pipeline failure are corrected by single and double factors, which fully shows that Bayesian network Reasoning ability and unique superiority of quantitative analysis of pipeline failure. The revised network model is more in line with reality and has better practical significance for improving the systematicity, predictability and accuracy of the quantitative analysis of the safety failure of urban gas pipelines. The rationality of the Bayesian network model and method is verified by a case. However, in actual engineering applications, many problems may also occur, such as the model may be deviated from the actual, lack of data on the main factors of pipeline failure. Therefore, it is necessary to further improve the model and further explore the multi-factor and multi-state calculation of pipeline failure to establish a more complete and reliable urban gas pipeline failure prediction model. W Section Pales Palisade Fence W Section Pales Palisade Fence,Palisade Fence Components,Cast Iron Palisade Fence,Palisade Fencing HEBEI DOUDOU METAL FENCE PRODUCTS CO.,LTD. , https://www.security-fencing.com