基于 FOA 优化的 CSSVM 管道堵塞状态识别研究

针对城市排水管道正常与堵塞故障状态在数据获取上的不平衡性造成的运行状态识别准确率下降的问题,提出了一种基于果蝇优化算法的代价敏感支持向量机的管道堵塞状态识别方法。根据排水管道内各运行状态下采集到的不平衡数据集,首先对不平衡数据集进行小波包分解,其次,提取各个分解系数的能量熵、近似熵指标构建特征向量集合;采用果蝇优化算法(FOA)对不同类样本惩罚因子C_m和核函数参数g进行优化选取,即对代价敏感支持向量机(CS-SVM)模型优化,将特征集合输入优化后的CS-SVM模型中,对排水管道的正常和堵塞状态识别,通过增大对少数类样本错分的惩罚代价,结果表明,提升了少数类的识别准确率。

Research on CSSVM pipe jam status recognition based on FOA optimization

Aiming at the problem of the accuracy of the recognition of the operating state caused by the unbalanced data acquisition in the normal and blocked fault state of the drainage pipeline, a method for a pipeline clogging state recognition based on cost-sensitive support vector machine based on fruit fly optimization algorithm is proposed. According to the unbalanced data set collected under various operating conditions in the drainage pipeline, the wavelet packet decomposition is first performed on the unbalanced data set. Secondly, the energy entropy of each decomposition coefficient and the approximate entropy index are used to construct the feature vector set. The fruit fly optimization algorithm is adopted. (FOA) optimizes the penalty factor Cm and the kernel function parameter g, that is, the cost-sensitive support vector machine ( CS-SVM) model optimization, and inputs the feature set into the optimized CS-SVM model to normalize the drainage pipe. Blocking state recognition, by increasing the penalty cost of misclassification of a few types of samples, the results show that the recognition accuracy of a few classes is improved.