Study On The Mechanism Of Magnesium Flux Pellet Mineral Phase Regulation

Under the pressure of environmental protection and reduction of high-quality iron ore resources,how to efficiently utilize local concentrates to develop magnesium fluxed pellets is an important problem.High w(SiO₂)is an important factor restricting the development of local iron concentrates.By studying the basic characteristics of local iron concentrate M,Y and low w(SiO₂)concentrate(DG),the suitable proportion of iron concentrates for high w(SiO₂)fluxed pellets was discussed.The results showed that M was suitable to be used as the main raw material for pellet making,and M:Y:DG was 5:3:2,w(SiO₂)was 5.5%?5.6%.The effects of basicity and w(Mg O)on the mineral phase of strength machine were studied.The results showed that the basicity was 0.65?1.05.With the increase of basicity,the Fe₂O₃interconnect crystal developed well,and the strength of the block increased.With the increase of w(Mg O),the strength reached the max at w(Mg O)=1.8%,and then decreased.Excessive w(Mg O)was not conducive to the development of Fe₂O₃grains.The roasting temperature of high silicon fluxed pellets should not be too high,and the suitable roasting temperature was 1250?1275°C.The results showed that the compressive strength and metallurgical properties of pellets were improved with the increase of basicity.With the addition of Mg O,the compressive strength and RDI_+3.15of the pellets showed a downward trend,and the crystallization of Fe₂O₃became worse.RSI was further improved.It was mainly due to the lattice replacement between Mg²⁺and Fe²⁺,which inhibited the reduction swelling of the pellets.The minerals in the mineral phase were identified and extracted by computer image recognition and extraction,and the data obtained were used to study their effects on the compressive strength of pellets.The compressive strength of pellets was predicted by using the mineral proportion.The results showed that the prediction accuracy of GRNN was low,with an average error of 245.65 N/P;while the prediction accuracy of BAS-GRNN was significantly increased,with an average error of only 20.51 N/P.Figure 34;Table 32;Reference 69...