Blast Furnace Erosion Prediction In The Process Of Neural Network Method

In this dissertation, we first sum up several ways to predetermine the corroding states of blast furnace. Then we introduced FEM(finite element method) and BEM(boundary element method), according to the current domestic research efforts in this field.The two methods both are based on thermocouple testing. Several techniques involved in the processing. First, the blast furnace is simplified by the two-dimensional profile, then make the erosion line ,which is a isothermal line of 1150"c, the boundary condition, Based on the boundary condition, we work out the computed temperature on the thermal couple point though solving the exact heat equation. Finally we compare the computed temperature and the measured temperature to get error.Through orthogonal experiment, we change the shape and position of the erosion line. Based on this we calculate the temperature of thermalcouple points. As a result, we choose a best group of data to determine the erosion line.In this article, we discussed the differences between two modules and introduce the method that combine the algorithm of FEM and NN We also discussed several kinds of NN that are commonly used in the field of function approximation. And choose BP network as our tool.There are also deferent training functions in BP network. Eacch has it's advantage and weakness. Considering the balance between rapid approximation and generalization, we choose "trainlm" as the training function...