Research On Lubrication Mechanism Of Mould/strand And Its Application Based On Time-frequency Analysis

As a major parameter reflecting the lubrication condition between mould and strand,mould friction plays an important role in ensuring the stable production of the continuous casting and improving strand quality.With the development and application of high casting speed in continuous casting technology,it is more critical to control lubrication state inside a mould by detection and analysis of mould friction information.The wavelet transform provides an effective way to accurately analyze small changes of mould friction.In this paper,time-frequency analysis was applied to mould friction signal based on the measured data from production of continuous casting and theoretical research on wavelet transform,which aimed at intensive study in friction information,further knowledge of mould friction behavior and the lubrication mechanism.This is of great theoretical and practical significance to promote the technological development in continuous casting production.(1)The data of mould friction representing lubrication state between mould and strand was acquired according to the measured data from production of continuous casting and detecting principle of mould friction.In accordance with the wavelet theory and the characteristics of mould friction signal,the wavelet basis function,scale of continuous wavelet analysis and the relevant parameters of discrete wavelet analysis were determined by comparative study.And on this basis,effective analysis of time-frequency information and abnormality detection were realized in the light of continuous wavelet transform and discrete wavelet transform of several typical mould friction information like level fluctuation,abnormality of submerged entry nozzle,longitudinal split breakout etc.(2)In order to quantitatively detect the running state of the friction signal system of the mould,the wavelet entropy was introduced to describe the fluctuation characteristics of the mould friction.The features of mould friction signal and wavelet entropy results were emphatically analyzed in normal production,different control models as well as under the condition of various casting speeds and the breakout.The results showed that the wavelet entropy could accurately reflect the overall changing trend of the friction inside a mould and its local variation features.In view of those results,the change trend of mould friction force and hydraulic cylinder force was discussed.Based on the same development of cylinder force and mould friction with casting,the analyzed result of cylinder force can be used for estimating the lubrication state.(3)The characteristics,the basic principle of artificial neural network were studied.The influence of the vibration parameters and different network types used for cylinder force prediction was discussed.And the BP neural network model of the hydraulic cylinder force was established.Also,the applicability and accuracy of the network were evaluated according to the measured data.The results indicated that BP neural network could make accurate forecast of cylinder force with different casting speeds and the error can be controlled in a certain range.