Aiming at interference to leak detection of mineral slurry pipeline caused by work condition adjustment

correctly extract characteristics of leak signal is the key to reduce the leakage of the false negatives and false positives. A leak detection method based on empirical mode decomposition（EMD） and variable predictive mode based class discriminate（VPMCD） is proposed. In this method

the pressure signals are decomposed into several intrinsic mode function（IMF） components

and take local Hilbert energy spectrum which most accurately reflect the pipeline operation conditions as feature values. Leak identification model is established by VPMCD classifier. When the method is applied to the leak detection

the experimental results show that under the conditions of normal operation

leakage and work adjustment condition

the recognition rate of the mineral slurry pipeline reaches 95 %

and by comprehensively analyzing the flow signals

it also improves the precision of leak detection.