基于稀疏性贝叶斯极限学习机的气动调节阀多类故障诊断

气动调节阀是工业过程中使用最广泛的终端执行机构之一,它的性能好坏直接影响控制回路的性能.将基于稀疏性贝叶斯的极限学习机(SBELM)方法运用于多类故障诊断,基于DAMADICS平台的典型气动调节阀多类故障模型数据通过SBELM进行训练.不仅能根据模型的先验知识和基于最大后验概率准则(MAP)的贝叶斯思想估计出模型输出的概率分布,而且能基于设定的性能指标自动剔除无用的训练样本,用一小部分观测数据达到多故障分类的目的,能训练出一个精确且紧凑的故障诊断模型.

Multi-Fault Diagnosis of Pneumatic Control Valve with Sparse Bayesian Extreme Learning Machine

Pneumatic control valve is the most widely used actuator in industrial process and its property is closely connected with the performance of control loop, so pneumatic control valve fault diagnosis is of great importance. The fault diagnosis method based on Sparse Bayesian Extreme Learning Machine (SBELM) of multiclass classification was introduced. The multi-fault diagnosis model of control valve with the data based on development and Application of Methods for Actuator Diagnosis in Industrial Control Systems (DAMADICS) is trained through SBELM. The proposed method allows for estimating the marginal likelihood of network outputs based on the priori knowledge of the model and the criterion of maximum a posterior(MAP), and automatically pruning out unnecessary samples based on a certain performance criterion during training phase, which utilizes only a very small fraction of the available observation data achieving the goal of multiclass discrimination, thus results in an accurate and compact fault diagnosis model.