Research Of Intemet Of Manufacturing Things For Massive Data Streaming Model Mining Algorithm

Internet of manufacturing things intergrate networks,embedded system, RFID,sensors and other electronic information technology and manufacturing technology,to achieve product manufacturing and service process and the whole lifecycle of intelligent processing and optimization control of a new manufactuing mode.Because the manufacturing process of IOMT is relatively complex,the deployment of a large number of sensor nodes in the field to monitor the production environment,resulting in a mass of read-time manufacturing data flow and mass manufacturing data.The traditional frequent pattern mining algorithm can not meet the application requirements of the IOMT.Therefore, to design an efficient and feasible frequenct pattern mining algorithm for massive manufacturing data Streams mining useful knowledge has become a new challenge.This dissertation emphasizes on exploring the characteristics of IOMT generated the number of manufacturing data buge and real-time data flow mass distribution, the existing data streams frequent pattern mining algorithm with in-depth stuy,combined with the IOMT sensor nodes which computing power and resource limited,proposed for an efficient frequecnt pattern mining algorithm,and simulation experiments verify the efficiency.The main work of this paper includes the following aspects:(1) Firstly,Comprehensive study the traditional frequent pattern mining algorithm and the data stream frequent pattern mining algorithm, combined with the characteristics of the data in the manufacturing system, analysis the advantages and disadvantages.(2)According to the characteristics of the massive offline manufacturing data, the traditional Aprioir algorithm has low efficiency. This paper proposes an optimization algorithm IFAMR(Improved Frequent Algorithm Based on MapReduce). The algorithm first uses the AprioriTid algorithm to preprocess the original data, delete all of the low frequency l_item, and then calculated for each transaction set (L) and minimum support degree (n) of the length to determine the end of the map operation after the largest merger when selected. The IFAMR(Improved Frequent Algorithm Based on MapReduce) algorithm reduces the low-frequency itemsets generated in the Map task, improve the efficiency of the mining algorithm.(3)According to IOMT realtime data strems characteristics of massive and ditributed,and the calculation ability and characteristics of resource constrained wireless sensor network node and inner perception. This paper presents a distributed window tree based data stream frequent pattern mining algorithm IFPM-DDS(Improved Frequent Pattern Mining Algorithm over Distributed Data Streams). This paper proposes an improved FP-Tree structure of the algorithm, by creating optimization in sensor nodes scattered on the distributed window tree, and then dynamic update and pruning of the distributed window tree. Finally dig through the window of the tree, to achieve rapid response to user queries, return frequent pattern data at any time in the window. The experimental results show that the algorithm can effectively shorten the query time and the accuracy of mining results under the premise of improving the efficiency of the mining algorithm.