The Research On On-line Thickness Measurement Of MDF Based On Optimized SVM

With the revolutionary breakthroughs and cross combinations of information technology,new materials,biotechnology and other important fields,a new round of industrial transformation has been began,which accelerates the deep integration of the new information technology and manufacturing industry.The traditional medium density fiberboard(MDF)measurement system is becoming more and more difficult to meet the requirements of high precision,high efficiency and automation during the processing.The thickness measurement of MDF is very important for continuous hot press to produce high quality MDF end products.Improving the accuracy of the thickness measurement of MDF is very important for stabilizing the physical properties and reducing energy consumption.Based on the 948 project of the State Forestry Administration,the "MDF line detection and precise thickness control of key technology introduction"(2014-4-46)and National Natural Fund Project "MDF continuous flat pressing with multi-field coupling and precise cooperative control mechanism"(31370565),the online detection method of MDF thickness is studied and actualized.Firstly,the manufacturing process and the measurement standard of MDF is introduced.According to the thickness measurement standard and the surface characteristics of MDF,a thickness measurement system which is in line with the structure and characteristics of modern manufacturing system is designed,and the basic framework of MDF thickness online measurement is formulated.According to the designed measurement system,the experimental scheme of MDF thickness measurement is proposed,and selected MDF with different thickness as the experimental material.We obtained a lot of measurement data,which proves the feasibility of the measurement system.Because there are many unstable factors during production and measurement,we need to pre-process the data to remove abnormal,secondary and redundant data.Through a combined filtering method which are wavelet packet transform and Kalman filter,the original data is pre-processed and the abnormal data are eliminated,which lays the foundation for the research of MDF thickness online measurement method.Based on the pre-processed data samples,the method of support vector machine regression is used to predict the thickness of MDF so as to eliminate a certain amount of measurement error.The different kernel functions for modeling MDF thickness is selected using grid search method.By comparing the prediction results,determine RBF-kernel-SVM regression model,completed the construction of online MDF thickness the measurement prediction model.In addition,in order to improve the accuracy of thickness regression model,using particle swarm optimization algorithm(PSO)to optimize the model.Then use the different improved PSO method to optimize the SVR model,and compare the precision which validates the improved MDF thickness prediction model has more advantages in the prediction accuracy.At the same time,the improved model is applied to MDF thickness measurement system which is designed in this paper so as to improve the detection accuracy.The MDF thickness online measurement system and the method based on optimized SVM developed in this paper has the characteristics of stable operation,high measurement accuracy.The research direction of MDF thickness measurement is in advance,and has high application value in MDF manufacturing.