Study On Multi-Sensor Data Fusion Modeling And Simulating Of MDF Hot Pressing Process

Wood-based panel industry is an important branch of forestry industry which severely influences the effect of alleviating the contradiction between timber supply and demand, and achieving sustainable development strategy of forestry industry in China. Hot-pressing is one of the key procedures in wood-based panel production playing a decisive role in the quality and output of the products. Hot-pressing is a nonlinear, complex and changeable process involving multiple variables, and it is often affected by various technique factors and noises. These features make it difficult and uncertain to realize internal measuring and process monitoring during hot-pressing process of wood based panels. To solve the above problems, based on the medium density fiberboard (MDF) hot-pressing on multi-layer press, the multi-sensor data fusion architecture and methods are studied in this dissertation to improve the measurement precision and credibility of decision, and provide theoretical and technical basis for optimizing the fiberboard hot-pressing process, predicting the properties, and improving process control performance.Based on the analysis of MDF hot-pressing mechanism, characteristics of technique parameters and their relationships, the dissertation put forward the views that multi-sensor data fusion theory should be applied lo the measurement and decision making of hot pressing process. A hierarchical fusion architecture and function model is established to provied a framework for the study of data fusion methods, data fusion algorithms of data, feature and decision layers are studied, and algorithms simulating are implemented. Hot pressing experiment and simulation results prove the feasibility and effectiveness of the fusion model and algorithm.Because the random noise and gross error may easily exist in the measurement data, in the data layer fusion, the confidence distance is calculated to check the consistency of data from multiple sensors which is embedded in hot-pressing system, and then the optimized number of sensors can be determined. An adaptive weighted least square algorithm with online learning ablility is presented to estimate parameter of internal temperature of the mat. The weight of each sensor is computed based on its deviation to overcome the influence of large error from one sensor and improve the overall estimation precision. The simulation results demonstrate the high accuracy of the overall estimation compared with local one of any sensor in the system and arithmetic average algorithm.Single factor comparison experiments and multi-factor orthogonal experiments are carried out in order to compute the influence factor of each technique factor on the property index and thus the qualitative relationship between technique factor and property index is acquired. The finite sample set with mechanical properties as output variables and technique factors as input are acquired by the experiments. Aiming at the prediction modeling with small sample set, SVM regression algorithm with PSO optimized parameters is presented in this dissertation, and a model for properties prediction of hot pressing product is established. The algorithm has higher precision and better generalization performance of the numerical fitting than traditional methods.To solve the uncertainty problem of hot-pressing process and to quantitatively represent fuzzy concepts and data, the fuzzy recognition framework of product rating, hot-pressing condition evaluation and control decision are set. The basic probability assignment is obtained by typical samples and Hamming distance method. The decision making is realized by integrated application of fuzzy sets theory and evidence theory to eliminate the uncertainty of decision made by artificial experience and achieve higher reliability and credibility of the decision.The dissertation combined multi-sensor data fusion theory and intelligent control theory with measurement and decision making process of MDF hot-pressing, and provided effective theoretical and technical support for the board mechanical properties prediction, process parameters setting and optimization. and decision making of control rules. The research has the important reality significance and theoritical value to improve the study of hot-pressing process, enhance the control system intellectual degree and realize the sustainable utilization of wood resources.