Online testing of lumber drying moisture based on layered information fusion

Lumber drying process is a complex dynamic system with strong coupling and nonlinear characteristics, and external environment parameters and physical parameter of lumber itself all may affect the changes of lumber moisture content. At present there are many traditional measuring methods for lumber moisture content, but the particularity and complex nonlinearity of the drying process make them have some limitations or disadvantages. It is an important research content for lumber drying full automatic control process achievement, how to quickly and efficiently establish the corresponding relationship between various parameters and moisture content, so as to realize real-time online testing of lumber moisture content. The multi-sensor information processed by multisensor information fusion system has the more complex forms, and can be on different levels of information, which can further gain much more information of detected target and environment. Its application in lumber drying control can effectively improve drying control level. In the study, a online testing layered fusion system of lumber moisture content is built for lumber drying process based on multi-sensor information fusion technology. According to the problem of low testing accuracy of lumber moisture content, the wavelet packet transform filtering estimation algorithm is used in data fusion layer. Data layer fusion accomplishes multi-sensor data fusion and state estimation, and uses mathematical method to seek state vector best fit for observation data. Research results show that wavelet packet has great advantages in testing and processing mutations and strong interference signals, and has good filtering effect for temperature and lumber moisture content. For the nonlinear, strong coupling and time-varying characteristics of lumber drying process, feature fusion layer uses least squares support vector machines (LSSVM) to establish model on the relationship between environment parameters (temperature, humidity) and lumber moisture content, and predict the lumber moisture content base on the model. The prediction results show that LSSVM has certain robustness and can fit actual moisture content curve with higher precision, its generalization ability is stronger. Based on MATLAB GUI, a fusion system platform is designed to accomplish online prediction of lumber moisture content. Its experimental results show that the platform can effectively realize online testing and predicting based on layered information fusion technology, and can intuitively display the prediction results for different lumber species or different drying process. All these studies provide reliable basis for the control decision of lumber drying automatic control system, and have good practical industry applying value and developing prospect.