Investigation of Natural Gas Leakage Rates from Buried Pipelines at Dense Phase Using CFD and ANN Approaches

Abstract

Natural gas leakage represents a significant safety concern in natural gas pipelines. A comprehensive understanding of pipeline leakage characteristics is crucial for the design of effective safety systems and addressing leakage concerns. This study examined leakage rates in natural gas pipelines for the dense phase and compared the results with pseudodense and vapor phases. The investigation was conducted using a combination of computational fluid dynamics (CFD) and artificial neural network (ANN) approaches. Key factors such as ambient temperature, soil porosity, diameter of soil particles, as well as the temperature and pressure at the pipeline leakage location were analyzed to understand their impact on pipeline leakage rates. The validation of the CFD model against experimental data from the literature resulted in excellent agreement. An ANN was trained with 210 random simulations from the CFD model to predict leakage rates. The ANN exhibited consistent performance in predicting leakage rates during training, validation, and testing. The results showed that the volume leakage rate decreases with increasing pressure in the dense phase, pseudodense phase, and vapor phase. Additionally, the results showed that as the temperature at the leakage location, porosity, and diameter of soil particles increased, the leakage rate also increased. The findings further indicated that environmental temperature had no notable influence on the leakage rate. The average leakage rate in the dense phase was 13.5% and 28.3% less than that in the pseudodense and vapor phases, respectively. Moreover, mathematical models (power functions) for leakage rates were developed for the dense, pseudodense, and vapor phases as functions of pipeline temperature, pressure at the leakage location, environmental temperature, soil porosity, and diameter of soil particles. In all three phases, the findings revealed a maximum deviation of 5% between the predicted models and the ANN model. The results of this study showed that the leakage rate in the dense phase natural gas transmission was reduced, improving the practical efficiency of this method.