Study On Intelligent Wood Drying Kilns’ Parameters Measurement And Control System

For the problem of parameters detection and the degree of automation are low in the wood drying kiln.It designed a kind of intelligent wood drying kiln parameter measurement and control system. This system collect parameters real-time based on ARM embedded platform, these parameters include temperature and humidity in the drying kiln,and the resistivity of wood,and so on. Through the WiFi network upload data to the PC,and real-time observe the state of drying kiln in the PC. When the data acquisition finished,it will feedback control the process of wood drying by comparing the measured value with the standard value. The result of experimental indicates that the root mean square error of the temperature, humidity, the resistivity of wood and wood moisture content are 2.953,2.496,0.744 and 1.038.To improve the control algorithm of system, it introduces the deep learning method, and combine deep belief network(DBN) with proportion integration differentiation(PID), and put forward a new control algorithm named DBN-PID.In the same hardware platform,using this new control algorithm to compare with the traditional algorithm of PID. The experimental results show that the detecting system of wood dry kilns with DBN- PID has more higher precision.The root mean square error of the temperature, humidity, the resistivity of wood and wood moisture content are 0.495,0.513,0.118,0.451.For further clarification about the performance of DBN-PID,using it to compare with the algorithm of BP-PID in the simulation.The results of simulation show that DBN-PID can approach the nonlinear object more better,and has the more stronger adaptive ability.On the model of wood drying, it introduces predictive control method based on the model,and it sets up a kind of soft measurement model based on the benchmark of wood moisture content.This model is applied in predictive control algorithm of deep learning.In this model, the temperature, the humidity and the resistivity of wood are all as the input,and get the predictive value of the wood moisture content.The experiment with oak as the research object,and choose the wood moisture content as reference quantity. In the 3000 set of data,it randomly select 2600 set of data as the training data,and select 340 set of data as the verification data,and another as test data.The result of simulation indicates that,when the wood moisture content of test data in the stage of 6%-30%,the root mean square error of the predictive value and the actual value is 0.636,and in the stage of30%-60%,it is 2.265,and more than stage of 60% is 3.532.It uses the same data to establish a soft measurement model based on Generalized Regression Neural Network,and it is analyzed comparatively with the soft measurement model based on DBN’s predictive results.After analysis of experimental results that the soft measurement model based on DBN has a higher precision,and root mean square error arrives a minimum value is0.025,and indicates that the effectiveness of the deep learning is applied in the soft measurement of wood moisture content.