基于改进K-means和CNN的储罐罐底点蚀诊断模型*

为解决储罐罐底点蚀问题，提出基于改进K-means和CNN的储罐罐底点蚀诊断模型，在传统聚类基础上引入肘部法则，保证k值选取3的准确性，将原始声发射信号特征参数和聚类后的类别信息输入模型进行训练，系统预测准确率高达99%。研究结果表明:该模型能够及时发现点蚀现象，指导管理者确定储罐开罐检查时间顺序，避免点蚀穿孔造成的人力、物力损失，降低储罐运行风险，保障储罐运行安全，研究结果可为罐底点蚀诊断提供技术支撑。

Diagnosis model of pitting corrosion at bottom of storage tank based on improved K-means and CNN

Aiming at the pitting corrosion at the bottom of storage tank,a diagnosis model of pitting corrosion at the bottom of storage tank based on the improved K-means and CNN was proposed.The elbow rule was introduced on the basis of traditional clustering to ensure the accuracy of k-value selecting 3,then the feature parameters of original acoustic emission signals and the clustered category information were input into the model for training,and the prediction accuracy of system was up to 99%.The results showed that the model could detect the pitting corrosion in a timely manner,and guide the manager to determine the time sequence of tank opening and inspection in the actual inspection,thus avoiding the loss of manpower and material resources caused by pitting perforation,reducing the risk of tank operation and ensuring the safety of tank operation,so as to provide technical support for the diagnosis of pitting corrosion at the bottom of tank.