Abstract: Due to the problems of leakage detection and location in gas pipelines as well as the measurement difficulties for the gas in pipelines， a gas pipeline linear parameter varying (LPV) model was established， and the generalized regression neural network (GRNN) was designed， which took the theoretic time difference as input model， and the location of different point in pipes as output. Acoustic identification method was used for gas pipeline leakage fault diagnosis and location. Simulation was carried out using field data， and the results showed that acoustic location algorithm research in gas pipelines based on the GRNN of LPV model approach is an effective method， which can make the predictive value accurately track real value. The experimental results can provide the reliable basis for the industrial application of gas pipeline leak detection and location.

Key words: LPV, GRNN, gas pipelines, fault detection and location, acoustic method