The Research On The Fault Diagnosis Methods For The Multimode Process

Fault detection and diagnosis,especially multimode processes fault detection and diagnosis are very important to ensure the safety of modern industry,and it is widely used in the environmental monitoring,chemical,electrical,machinery,petroleum,thermal,manufacturing,aviation,aerospace,and other industrial fields,therefore,it is great theoretical and applied research value.With the expansion of industrial production scale,massive data has been generated.Traditional fault diagnosis platforms have encountered technical bottlenecks both in storage and fault diagnosis for these GB or even TB data.The traditional fault diagnosis algorithm is not able to deal with the massive offline data,it would run into a machine crash,lowly algorithm efficiency,run time consuming and so on.When these complex multimodal processes fail,there are huge economic losses and casualties.Aiming at these problems,the focus of this thesis is to design and build a bigdata fault diagnosis experiment platform based on Hadoop;at the same time,a parallel fault diagnosis method for multimode processes is proposed and implemented on Hadoop.In this thesis,the fault detection and diagnosis of massive offline data has been realized and the experimental result is quite reliable.Firstly,this thesis has studied the technical route of the Hadoop platform and the key technology of multimode fault diagnosis based on Hadoop.At the same time,it has designed and set up a fault diagnosis platform based on Hadoop.Secondly,this thesis provides a method for the transformation of each mode of TE process,and realizes the expansion of the simulation data volume to GB level.And then,the study of multimode fault diagnosis method is carried out based on multimode TE process.Because the ability of traditional fault diagnosis algorithm for processing massive offline data is limited,a parallel fault diagnosis method for multimode TE process is proposed,namely parallel Kmeans-KPCA-naiveBayes fault diagnosis method.This method is based on distributed storage system(HDFS)and parallel computing(MapReduce)of Hadoop to diagnose and analyze the massive data.The establishment of parallel Kmeans-KPCA-naiveBayes fault diagnosis algorithm for TE process is shown below:First of all,this thesis makes a comparative study on the clustering time and clustering precision of the five main clustering algorithms,it determines the superiority of the Kmeans algorithm.Secondly,it compares the clustering effect of Kmeans between Matlab and Hadoop platform.Furthermore,the advantage of distributed storage and parallel computing of Hadoop is clarified for massive data.Then,the KPCA is used to prepare for the fault diagnosis algorithm based on Bayes.Finally,the visual representation of experimental results is presented to achieve the purpose of fault detection and diagnosis based on R.