Research Of EAF State Judgment Based On Arc Sound

The EAF is one of the most important steel-making facilities in the area of metallurgy,which plays an important role in general process and short process.Judging furnace states is not only for meeting smelting production requirements,but also for designing corresponding control systems.Accurate state judgment can improve the automatic of control system,shorten the smelting time,reduce energy consumption and increase productivity.The traditional methods of state judgment depend on the experience of workers,which will cause errors unavoidably.Experienced operators can judge the degree of the submergence and stability of arc according to arc sound,therefore the research of EAF state judgment based on arc sound is practically significant.Based on massive papers,the EAF smelting technology and the main process of voice recognition are summarized,and the are sound generation mechanism is analyzed in the thesis.In order to extract arc sound features which reflect EAF furnace states' characteristics,acoustic signals' time-domain analysis,frequency-domain analysis,and wavelet packet analysis are applied.On this basis the thesis obtains three arc sound features:the variance,the ratio of harmonic frequency and base frequency and the wavelet packet energy.Features' thresholds of different states are set.Through the further analysis of these three furnace state judgment methods,each single method is difficult to reach a high judgment accuracy.Therefore,a three-feature fusion state judgment method is proposed in this thesis to improve the furnace states judgment accuracy.As these three arc sound features' thresholds are uncertain,and each kind of feature has a different influence on furnace state judgment,in this thesis the fuzzy reasoning method is adopted to complete the feature fusion.By fuzzifying,fuzzy logic reasoning and defuzzifying,three features are fused according to their characteristics.Finally,the EAF state judgment method based on arc sound is obtained,and actual EAF production data is tested.The accuracy has proved to reach 86%,which proves this method is practical.