Framework And Implementation Of Data Acquisition And Analysis System Based On Industrial Furnace Tube Centrifugal Casting Process

Ethylene as an important product of petrochemical enterprise is produced in ethylene crack facilities,which mainly includes radiation section and convection section,and industrial furnace tubes are the main part of them.Ethylene cracking reactions require high temperature,in which industrial furnace tubes easily cause accident because of carburization,fracture,corrosion etc.So,it is necessary to improve the quality of industrial furnace tubes to ensure the manufacture safety.Most industrial furnace tubes are manufactured by centrifugal casting process,from which the industrial furnace tubes have compact structure,integrity crystal(isometric crystal and columnar crystal)and good mechanical property.The traditional methods to detect the quality of furnace tubes are penetrant testing,chemical component analysis,tensile testing and microstructure examination etc.But these methods are time-consuming,power-consuming and costly.Besides,these detecting methods would produce large amounts of pollutants.On the other hand,manual operation has the characteristic big error and may also leave out some furnace tubes to be detected.In this paper,multivariate statistical analysis method is introduced to monitor the quality of furnace tubes to save the disadvantages of traditional detection methods.Multivariate statistical analysis method is a method based on historical data.Many factors together contribute to the quality of furnace tubes,so it is essential to collect the data during the manufacture before the use of multivariate statistical analysis method.The main works of this paper are as follows:(1)When using multivariate statistical method to analysis,it is necessary to get large amount of data from many variables.Conventional methods are artificial measurement and manual recording.But these methods have the disadvantage of large error,low frequency collecting and power-consuming.So,this paper designed and implemented a data acquisition system based on industrial furnace tube centrifugal casting process to collect data automatically.(2)Single measurement data,including the proportion of each metal element and casting temperature,each can be collected from both raw metal material melting and centrifugal casting process,while other data,such as surface temperatures at different locations of molding cylinder,rotational speeds and PT detecting result,can only be obtained from centrifugal casting.As to original data,pretreatment is needed,which includes wiping out data beyond technological range,adopting dynamic time warping method to solve the problem of data unequal length,adopting on-line data filling methods to supplement the missing data.(3)Raw metal material melting is a continuous process.Surface temperatures at different locations of molding cylinder and rotational speeds are changing constantly and they have the character of batch process.In this paper,principle component analysis,PCA,is employed to extract process characteristic from single measurement series and the multiway principle components analysis,MPCA,is employed to extract the process feature from time series across batches.(4)During the modeling,two types of industry furnace tubes are used.The best correct identification rate and the false alarm rate of this model are 100%and 0%respectively for studied process.The contribution plots are given to forecast results and find the reasons leading to failures,which can provide guidelines for the improvement of pipe manufacturing and saving the cost on penetration testing.