Abstract: In order to explore variation characteristics of related parameters of corrosion acoustic emission signals, experiment was carried out with a common vertical metal storage tank as research object to study characteristics of its acoustic emission source for corrosion. Then, a safety evaluation model based on BP neural network was established, and case study of its application was carried out. The results show that acoustic emission activity and intensity will change along with severity of corrosion reaction, and wave forms of corrosion signals in different periods of corrosion activity will exhibit three types, continuous, abrupt and hybrid types with its frequencies mainly concentrating between 20-60 kHz. The output of BP neural network model is consistent with actual evaluation results, which proves its feasibility and effectiveness.

Key words: large storage tank, corrosion signal, acoustic emission activity and intensity, acoustic emission monitoring, back propagation (BP) neural network model