A Study Of Hybrid Modeling Technique For Fault Detection Based Industrial Big Data

In recent years,with the rapid development of the Internet, Internet of things, sensors and other IT and communication technology. Many factories have to face the challenges of the data volume soared. With the popularity of technology and modern management concepts in industry, companies operations become increasingly dependent on information technology. Industry has stored a large number of equipment condition data and has shown a lot of characteristics of big data, but companies do not find out the value of these data. Specifically, it can put sensors into the equipment and production line, and then connect to the Internet to transmit data via wireless communication. In this way, it can realize production of equipment for real-time monitoring.During operation period of devices, natural wear and tear or accident will make the performance of devices changed. Now use the available of sensor technology and real-time sensing data, the running data of the devices will be perceived and realize fault diagnosis. The fault diagnosis method can be divided into three types: data-drived method, first-principle based methods, knowledge-based methods. Based on the industry big dataand meet the requirements of reliability, this paper combine the knowledge-based methods and data-driven methods present a hybrid model of fault diagnosis and used the hybrid model in shield machine hydraulicsystem fault diagnosis.In this paper, the works are summarized as follows:(1)Firstly, this paper proposed a hybrid model for fault diagnosis.(2)Secondly, this paper proposed the data-drived method based on support vector machine algorithm. We analysisthe kernel function and multi- classification problems. This paper used this method in hydraulic systemof shied machine and in AMESim simulation software to research oil pollution and oil spill fault.(3)Finally, this paper apply data mining algorithms to discover the knowledge. In this paper, we use Apriori algorithm to extraction rules in mass data and apply cluster algorithm k-means to find rules.