基于 PSO-RVM 模型的大坝安全监控研究

为提高大坝安全监控模型的预报精度和时间,利用粒子群算法(PSO)对相关向量机的关键核参数进行寻优,通过建立大坝安全监控模型与实际值比较,从而对RVM模型进行稀疏、学习、泛化等性能的分析研究。采用某实际工程实测视准线位移监测对模型进行验证,并通过均方根、标准均方和平均绝对百分比等误差,对模型预测的准确性、稳定性和可信程度进行评价。研究表明,PSO-RVM模型的泛化性能明显优于传统的RVM模型,应用于大坝安全监测建模是可行的。

Study on dam safety monitoring based on PSO-RVM model

In order to improve the prediction accuracy and time of the dam safety monitoring model, particle swarm optimization (PSO) is used to optimize the key kernel parameters of the relevant vector machine. By comparing the dam safety monitoring model with the actual value, the sparse performance, learning performance and generalization performance of the RVM model are analyzed and studied. The model is verifified by the observation of visual alignment displacement in an actual project. The accuracy, stability and reliability of the model prediction are evaluated by means of errors such as root mean square, standard mean square and average absolute percentage. The research shows that the generalization performance of PSO-RVM is obviously better than that of traditional RVM. And it is feasible to apply PSO-RVM to dam safety monitoring modeling.