Fault Diagnosis Method And Software Realization Of Industrial Process Based On Mechanism And Data

Nowadays industrial manufacturing processes become more and more complex,and effective fault diagnosis of these huge,complex,and dangerous systems is particularly important.At present,the extensive implementations of DCS systems,intelligent instruments,and fieldbus technology make a large number of data in industries.The data can be effectively acquired and preserved in real time,which provides data support for the current study.This topic is mainly for two levels:the whole industrial process and flow valves.First of all,some reported studies are summarized.The TE process,principal component analysis,preliminary-summation-based principal component analysis were introduced.Concerning the fault diagnosis of whole process,because the preliminary-summation-based principal component analysis(PS-PCA)will cause a large time delay in fault detection,a weighted preliminary-summation-based principal component analysis(WPS-PCA)was proposed in this thesis.By raising the weight of the current sample in WPS-PCA,the current sample increases its proportion in the summation result,then fault detection delay was greatly reduced,and the fault detection rates also can be improved.In addition,for the fault diagnosis of flow valves,a fault diagnosis method based on the DCS system data and the mechanism of industrial flow valves was proposed in this paper.This method builds the model between the valve opening degree and the flow.The score can be computed by comparing the real valve flow and its predicted value.The scores of several common valve faults also can be computed.Finally,a configurable fault diagnosis software platform based on DCS systems was designed and developed.The structure of the software platform was introduced in three parts:background algorithm part,the database,and the front interface.