In the present work, the multiplicity of fault characteristics is proposed and analyzed to improve the fault diagnosis performance. It is based on the following recognition that the underlying fault characteristics in...
详细信息
In the present work, the multiplicity of fault characteristics is proposed and analyzed to improve the fault diagnosis performance. It is based on the following recognition that the underlying fault characteristics in general do not stay constant but will present changes along the time direction. That is, the fault process reveals different variable correlations across different time periods. To analyze the multiplicity of fault characteristics, a fault division algorithm is developed to divide the fault process into multiple local time periods where the fault characteristics are deemed similar within the same local time period. Then a representative fault decomposition model is built in each local time period to reveal the relationships between the fault and normal operation status. In this way, these different fault characteristics can be modeled respectively. The proposed method gives an interesting insight into the fault evolvement behaviors and a more accurate from-fault-to-normal reconstruction result can be expected for fault diagnosis. The feasibility and performance of the proposed fault diagnosis method are illustrated with the Tennessee Eastman process.
传统的偏最小二乘方法(partial least squares,PLS)因未对建模数据求取均值轨迹,以及没有考虑多单元生产对浸出率的综合作用,导致无法准确建立过程信息与质量变量之间的回归关系。根据高铜矿氰化浸出过程的多单元和耗时长的特点,提出一...
详细信息
传统的偏最小二乘方法(partial least squares,PLS)因未对建模数据求取均值轨迹,以及没有考虑多单元生产对浸出率的综合作用,导致无法准确建立过程信息与质量变量之间的回归关系。根据高铜矿氰化浸出过程的多单元和耗时长的特点,提出一种针对连续过程的基于多单元均值轨迹的浸出率预测方法。获取建模数据的均值轨迹矩阵,在此基础上分别建立每个单元与实测浸出率的回归模型。定义输入向量与每个单元建模数据的相似度以及预测模型的权重,将各单元预测结果加权综合作为最终预测值。将该方法应用于氰化浸出过程浸出率预测,仿真结果表明,该方法体现了生产过程实际物理特性,提高了模型的解释能力,增强了预测模型的泛化性能。
暂无评论