The Research Of Prediction Method For The Splash Occur During AOD Furnace Smelting

Smelting medium-low carbon ferrochrome with AOD furnace is a new process of iron alloy production, it can decrease the power consumption per ton, greatly improve the product’s additional value and reduce product processing cost because of using oxygen instead of electricity decreases the middle electric furnace heating process compared with traditional Electricity Silicon Hot method. For the reason of small furnace capacity ratio and high impurities of S, plash accidents occur occasionally in the AOD furnace smelting process, the splash not only shorten the life of smelting core equipment, destroy the thermal balance, enlarge the steel loss, even endanger personal safety when severe. Therefore, research the prediction method of splash occur in the is the AOD furnace smelting process is the inevitable and key issue of argon oxygen refine iron alloy process industrialization process..Firstly, the paper analysis the process and flow of AOD, the mechanism and reason of splash form in the smelting process, summarize the relative splash features, select converting noise in the smelting process as characteristic signal which represent splash round the corner and a converting noise collecting testing platform is constructed, this testing platform use high-powered USB2811DAQ Card to extract noise and shield the electromagnetic interference, environment noise and echo by means of fiber transmission direct at wicked operating conditions and transmission environment in smelting spot. On the bias of comparative analysis of common signal process algorithm, use wavelet packet analytical method to denoise the collected converting noise data, ideal effect is obtained. After that, extract splash characteristics information through wavelet packet energy moment, then decompose and reconstruct the signal with DB10wavelet basis, thus, obtain the characteristic value. Finally, a splash prediction based on SVM is proposed aim at small sample, nonlinear and high dimension of splash accidents, this method takes splash characteristics value as model input and splash type as output, and adopt multiclass classification method of BT-SVM to make decision. The testing result shows, forecast accuracy up to95%, effectively shorten the prediction time and obviously improve the prediction accuracy.This paper acquires splash noise signal by data collecting platform, using wavelet packet analytical method to denoise and extracting splash features, predict splash during smelting process with SVM method and ideal prediction result is obtained. The research result proposes a useful reference for the lowering energy consumption and safety production of iron alloy smelting process.