Abstract:

Considering the lack of the fault data in the real production process, support vector machine (SVM), which fits the small sample problem was adopted to diagnose the faults of Stone & Webster Ultra-selective Cracking (USC) naphtha cracking furnace.To eliminate the disturbances from the high dimensional data as well as system noises, a novel adaptive quantum ant colony optimization (AQACO) algorithm was proposed to select the fault features based on the classical ACO with the introduction of quantum bit.The results of numerical simulation experiments showed that the proposed AQACO had better global optimization ability.The results of diagnosis of cracking furnace sensor faults demonstrated that AQACO could find the fault features quickly and exactly, which effectively improved the fault diagnosis performance of SVM in terms of correctness and robustness.

Key words:

裂解炉, 量子蚁群算法, 故障诊断, 特征选择

摘要：

针对实际生产过程中缺乏故障数据，采用适合小样本问题的支持向量机对石脑油裂解炉进行故障诊断。为了消除高维数据及系统噪声对故障诊断的干扰，将量子编码引入蚁群算法，提出一种新的自适应量子蚁群算法进行故障特征选择以进一步提高诊断性能。数值仿真实验结果显示，提出的自适应量子蚁群算法具有更好的全局寻优性能;对石脑油裂解炉传感器故障的诊断结果表明自适应量子蚁群算法能快速、准确地搜索到关键故障特征，有效地提高了支持向量机故障诊断的正确率和鲁棒性。

关键词:

裂解炉, 量子蚁群算法, 故障诊断, 特征选择