Research On The Technology Of Data Cleaning In Manu-Facturing Of Internet Of Things

Internet of things in Manufacturing is a product of the Internet of things technology application in manufacturing industry, it is a kind of new mode which mainly used for dynamic perception of various information resources of product design, manufacturing and service process and intelligent processing and optimal control. But this new model has met many challenges in the development process. On one hand, multiple interference environment such as the strong electromagnetic interference, metal medium, many ob-stacles and many dynamic existence objects like "the human, material, equipment, pro-duction process, product" in the manufacturing process for data perception, it is great difficult to achieve reliable data perception; on the other hand, limited resources, the dy-namic topology and harsh environmental conditions, the hybrid network convergence, these limitation effect The real-time, reliability and accuracy of data transmission; then the limited computation resources are not sufficient to fully support data because of pro-ducing huge amounts of data. In order to realize the accurate control of the manufactur-ing process, solving data reliable perception, real-time transmission and mass data intel-ligent processing in manufacturing is one of the key problem.RFID and sensor technology are considered as the two key technology for Internet of things in manufacture. However, under harsh industrial manufacturing environment, the sensor working time will be greatly reduced, the rate of read correctly tag can only reach60%～70%, These problem will bring a lot of reliable data for product traceability and data management of Internet of things in manufacturing. Aiming at the problem of RFID and sensor data, we will research deeply to improve the reliability of data product traceability and data management, the main work of the following two aspects:1.. For manufacturing complex motion between the product and the environment, when the phenomenon of false negative happened, RFID data cleaning algorithm based on statistical smoothing (SMURF) will lead to adaptive size window cleaning adjust unreasonable and produce lots of negative reading; so this paper presents an improved algorithm which regulate the window more reasonable for cleaning SMURF data in dy-namic environment, this improved algorithm could reduce the missing probability to im-prove false negative phenomenon.2、According to the problem which is all or part of attribute value of collected data missing caused by data collection terminal work anomaly among Internet of things in manufacturing, we put forward a kind of filling algorithm based on ROUSTIDA for in-complete data, in order to achieve the elimination of a potential conflict decision rules in filling and Avoid information inconsistency.Combined with the process parameters of pipe manufacturing process in this article, the simulation of two kinds of data problems proposed solution show that the improved SMURF algorithm can reduce false negative effectively and the algorithm based on ROUSTIDA of filling incomplete data can also effectively reduce the error between the true value and the filling value to reduce the risk of incomplete data.