Application of leakage pre-warning system for hazardous chemical storage tank based on YOLOv3-prePReLU algorithm

Abstract

A storage tank with liquid chemicals contains enormous internal energy. Once a tank leakage is not handled in time, it may cause catastrophic accidents. Therefore the pre-warning and prevention for leakage accidents of storage tank with hazardous chemicals is particularly important. In this paper, 6000 pictures under different experimental conditions were collected when storage tank discharged continuously, and then an image dataset of tank leakage was generated. Then it was trained by YOLOv3, YOLOv3-PostPReLU, and YOLOv3-prePReLU algorithms, the experimental results showed that the mean average precision (mAP) of YOLOv3-prePReLU algorithm was 0.89, which was more accurate than that of other algorithms. Their algorithm's response time to the identification of small hole leakage and middle hole leakage was 42857.1 times and 21428.6 times respectively, which was faster than that of the traditional A-level detection and A-level isolation system. The leakage experiment showed that YOLOv3-prePReLU's identification deviations from the leakage holes (d = 0.005 m, 0.010 m, 0.025 m) to the ground level was between - 0.7874% and 3.1852%, from the jet location on the ground level to tank wall was between - 2.6667% and 4.1562%, which provided a scientific basis for the early warning, prevention and accident handling measures of tank leakage accidents.