Research And Improvement Of Historical Database In Distributed Control System

With the continuous advancement and merger trend of industrialization and informationization in China,the Chinese government has put forward the "Made in China 2025",which will promote intelligent manufacturing as the main direction.Because of this intelligent manufacturing trend,large amount of data will be generated during the manufacturing process.The manufacturing enterprises focus on building intelligent factories and digital workshops.The higher data collection frequency in process industry is needed to obtain higher accuracy.At the same time,with the expansion of industrial production scale,much more points need to be collect,and these have increased the amount of data growth,resulting in massive data.The distributed control system(DCS)is widely used in chemical plants,steel mills,cement plants,large-scale sewage treatment plants,power plants and other places that need large amounts of data transmission and distributed settings.it is responsible for monitoring a variety of equipment and facilities in the bottom of the plant.DCS's historical database directly collect and store the real-time data from the system.However,confronting massive data,the existing historical database has been unable to meet the application requirements,mainly reflected in large data storage space,long time historical data query response,and multi-client concurrent query request may be lost.In order to solve the problems mentioned above,this thesis studies some bottlenecks that limit the performance of historical database,including the index and data structure of the file and the disk I/O.Besides,different improvement strategies taken by the physical property of the devices corresponding the analog quantity and digital quantity are also studied.First,according to the actual physical properties of the device corresponding to the DCS analog points,the collection cycle of analog points can be set to multi-cycle collection with different cycles for different points.While the existing strategy chooses single fixed cycle,which generates huge data redundancy.Meanwhile,the storage method of collecting data of the entire ten-minute for each analog point changes from the scattered storage to continuous storage,which improves query efficiency.A corresponding improved index table is also designed for these two improvements.Secondly,the query time of the existing historical database is very long,and the length of the query is limited.After analyzing the query characteristics of the DCS system,a 1-minute mean and 10-minute mean mixed collection strategy is proposed.The number of files needs to be opened is greatly reduced,which significantly improve the analog history data query speed and support for a longer time range of data query.Furthermore,some switches only change once for a long time,making a lot of data files no deformation structure.The query and open of these irrelevant files would waste much time.Hence,we use the relationship matrix to record whether a switch point has a displacement occurs in each data file in a day.And this matrix is used as a basis for opening the file when querying,which greatly reducing the number of opening data files.Displacement of different switch points occurs randomly.The continuous storage of the variable structure which represent different parts in the data area leads to low efficiency in searching one particular switch pint of the historical data.We propose the idea of using chains to organize each data file that belongs to the same switch point of the displacement structure.The two improvements to the switch can significantly save the query time.In addition,due to the expansion of industrial scale,the monitor scope is increased.And the number of DCS operator stations is also increased.The operator stations will miss the query data when they concurrently query historical data.This thesis designs and implements a message queue with response based on memory mapping file technology to queued and cache the query request and ensure the reliability of the query.We build the experiment environment in an enterprise lab.The project process from real factory is ran to test what we proposed in this thesis.Functions and performance results are gathered and analyzed.The results show that our research can improve the user experience and working efficiency.