储罐底板腐蚀状况的贝叶斯网络智能评价方法

Intelligent Evaluation Method for Tank Bottom Corrosion Based on Bayesian Networks

以贝叶斯网络为基础,利用与储罐底板腐蚀相关的外部表征因素,结合领域专家经验,采用随机重启爬山算法等5种启发式算法,构建储罐底板腐蚀状况贝叶斯网络智能评价模型。对比样本预测结果可知:随机重启爬山算法构建的网络模型预测能力优于其他4种算法学习的网络结构,平均准确率为92%。预测结果表明,该预测模型可解决储罐底板腐蚀状况的预测问题,具有一定的工程应用价值。

Based on Bayesian networks and using related external factors of the tank bottom corrosion, as well as combined with experts＇ experience in the field, the Bayesian networks intelligent evaluation model for tank bottom corrosion was established with random repeated hill climbing algorithm and other four heuristic algo- rithms. Comparing prediction results with acoustic emission testing results shows that the average accuracy rati- o can reach 92% , and this prediction technique can resolve the problem in tank bottom corrosion prediction, as well as can be used in engineering applications.