The Development Of Automatic Acquisition Of High Temperature Melt And Drop Properties Of Iron Ore And Its Forecasting Research

The first hand real-time data can be obtained by informationized experimental techniques and be compressed for storage, which can achieve common functions such as query, modifying, backup and so on. It has important significance to achieve experimental goals. The iron ore performance under high temperature and load to soften and melt dripping is obtained by measuring high temperature softening and changes in air pressure in a simulating environment of blast furnace. How to use information technology to accurately measure, record and even forecast the performance of iron ore through neural network technology is the main purpose of this research paper.The paper developed high-temperature melt dripping detecting system and relevant computer software program which can acquire, display and store the parameters of high-temperature melt dripping performance by configuring the temperature sensors, displacement sensors, pressure sensors, acquisition board, computer and other equipment. Software program was designed by the VB language, called Excel as data recording tool, used Iocomp control to draw real-time curve and transform the acquired voltage signal into corresponding temperature, displacement, pressure difference data values, which result in acquisition of parameter of high-temperature melt dripping performance. The results show that: the temperature linear precision is 0.2%, displacement linear precision is 0.16%, pressure difference linear precision is 0.1%. Based on collected data from the high-temperature melt dripping detecting system, we designed the BP neutral net in MATLAB environment and trained it, then verify its predicting effect using test sample data. The results show that when a single hidden layer node number is 18, output layer using linear function to activate, combined gradient descent algorithm and Newton algorithm as developed BP algorithm which called trianlm function to train the designed BP neutral net, MSE performance error goal can be approximated to 10-5. To test samples the average temperature prediction accuracy reached 99.7%, the average pressure prediction accuracy reached 99.8%.