基于SSA-RVM的储罐底板腐蚀轻量化在线评价方法OA

To address the reliance on expert experience and subjective signal interpretation in acoustic e-mission evaluation of tank bottom plate corrosion,this study proposes a lightweight(referring to low computa-tional complexity enabling deployment on edge devices)online evaluation method based on Sparrow Search Al-gorithm-optimized Relevance Vector Machine(SSA-RVM).This method integrates the dynamic characteristics of acoustic emission with the macroscopic static parameters of storage tanks to construct a multi-source data-driv-en evaluation model.The experimental results showed that the average prediction accuracy of the model reached 87%,and the identification accuracy for high-risk levels(levels Ⅲ and Ⅳ)exceeded 80%.It realizes the rap-id and objective diagnosis of the corrosion state under the small-sample conditions(i.e.,limited labeled data for modeling,such as 100 items,which is far less than the scale required by traditional data-driven methods).It provides reproducible data support for predictive maintenance of in-service storage tanks.