A corrosion correlation analysis method based on pipeline big data

At present, the total number of long-distance pipeline in China has reached 125000 km, achieved great development in the past decade. With the increase of the running time of the pipeline system and the aging of material properties, the frequent accidents are caused by pipeline corrosion and failure leakage. What’s more, the severity of the accident consequences has been increased since the pipe diameter and operation pressure has increased, especially for the impact on the high consequence area. Thus, the safety problems of oil and gas pipelines are paid more and more attention. Extracting the corrosive factors of pipeline and the law of risk evolution is important to predict the prevention of pipeline leakage. With the development of digital pipeline, the data produced during the process of design, construction and operation with in pipeline system were saved and the amount of detection data signal has also reached the level of TB. All of this has provided the fundamental for information extraction. However, because of lack of relationship between these data sets, most part of the data in which the safety information of pipeline hidden was ignored and discarded. In order to fulfill the potential of the “big data” and analyze the causes of pipeline corrosion in term of data, structural and unstructured data produced in the design, construction and operation of the pipeline system are constructed in this paper, and based on the mutual Information theory method, the correlation model between the corrosion grade and the multi-factor is established, which aim is to extract all the relevant factors of pipeline corrosion defects from the “big data” and determine the key indexes of the defects. This method is not limited to structured data, and it is also applicable to the analysis of unstructured data. The basic modeling steps include: collect all the information of a pipe segment first, then divide the data into four datasets; the third step is to attribute domain discretization; finally, a correlation model based on mutual information theory is established to determine the relationship between corrosion and other factors and to rank all factors according to the size of influence values. It is noteworthy that the relationship here is a correlation, not a causal relationship, that is, no need to consider the linear or non-linear relationship of the data. By validating the validity of the model, the key factors affecting the pipeline corrosion are determined, which lays a foundation for the prediction of pipeline life. The conclusion shows that the correlation between the data sets of pipeline “big data” can be used to obtain the information of pipeline corrosion and to provide the basis for accident prevention. In addition, the method can be used to study one of the factors, such as further determine which soil is more relevant with the corrosion, after the main factors are extracted to narrow the scope.