基于矿热炉外磁场信号的三相电极位置检测系统设计

针对矿热炉三相电极位置精确测量、节能降耗和安全生产的需求,设计一种基于矿热炉外磁场信号的三相电极位置检测系统。首先根据矿热炉的实际构造,结合COMSOL Multiphysics软件,建立矿热炉仿真模型,并对矿热炉磁场进行分析,选取外磁场信号采样点。在选取的采样点,采集具有不同电极位置的矿热炉模型的外磁场信号,建立矿热炉外磁场信号样本集。根据该样本集,应用偏最小二乘(partial least squares,PLS)回归分析、径向基函数神经网络(radial basis function neural network,RBFNN)和粒子群优化RBFNN(particle swarm optimized RBFNN,PSO-RBFNN)分别建立矿热炉三相电极位置检测模型,并结合MATLAB GUI建立基于矿热炉外磁场信号的三相电极位置检测系统。实验结果表明,检测系统的三种模型都可以实现对电极位置的检测,其中PSO-RBFNN模型的效果最优,三相电极位置检测准确率达到94.98%(训练集),90.21%(测试集),均方根误差为0.053 5(训练集)、0.131 1(测试集)。提出的检测系统能够较精确地测量三相电极在矿热炉内的位置,实现非接触式检测,具有较好的实用价值和应用前景。

Design of Three-phase Electrode Position Detection System Based on External Magnetic Field Signals of Submerged Arc Furnace

Aiming at the demand of accurately measuring the three-phase electrode positions of submerged arc furnace, reducing energy consumption and ensuring safety production, a three-phase electrode position detection system based on external magnetic field signals of submergrd arc furnace was designed. First of all, according to the actual structure of submergrd arc furnace, combined with COMSOL Multiphysics software, the simulation model of submergrd arc furnace was established, and the magnetic field of the furnace was analyzed to select the sampling points of the external magnetic field signal. At the selected sampling points, the external magnetic field signals of the models with different electrode positions were collected to establish the sample set of external magnetic field signals. According to the sample set, PLS multivariate regression analysis, RBF neural network (RBFNN) and PSO optimized RBF neural network (PSO-RBFNN) were used to establish the three-phase electrode position detection models respectively, and combined with MATLAB GUI, the three-phase electrode position detection system based on the external magnetic field signals of submerged arc furnace was established. The experimental results show that the three models in the detection system all can realize the detection of electrode positions, among which the PSO-RBFNN model has the best effect, with the accuracy of 94.89% (the training set), 90.21% (the test set), and RMSE of 0.053 5 (the training set), 0.131 1 (the test set). The detection system proposed can accurately measure the three-phase electrode positions in the furnace, realizing non-contact detection, which has great practical value and application prospect.