Artificial intelligence prediction and thermodynamic analysis on particle agglomeration/defluidization during waste incineration: Effect of sand bed composition and agglomeration indices

In the incineration process, agglomeration is caused by the combustion of wastes containing adhesion materials such as alkalis and earth alkalis. Alkali oxides not only have a low melting point but also can react with Si to form low-melting-point eutectics. Most of the researches focused on the inlet materials, however, the properties of sand bed materials are also playing an important role on defluidization process. A comprehensive analysis of sand compositions on the prediction of defluidization tendency is carried out. Variation of sand properties before and after fluidized bed operation is estimated by XRF characterization, and the effect of particle size distribution is also in consideration. Experimental results indicated that defluidization time decreases with the increasing particle size of all sand sources. Higher temperature promotes agglomeration/defluidization. According to the agglomeration indices calculation, indicator Al + Ca + Mg/(Na + K + Fe) can be used to estimate the defluidization tendency before fluidized bed operation for particle sizes below 770 & mu;m. In contrast, for a particle size of 920 & mu;m, Al + Ca + Mg + Fe /(Na + K) can be utilized to qualitatively characterize the defluidization time. On the other hand, the Support Vector Regression (SVR) model was the best model for predicting the defluidization time (R2 of 93.40% and MSE of 3470). Causal analysis shows particle size, K2O, and Na2O were the top three most important factors that affect the particle agglomeration. The results are helpful to the evaluation of the agglomeration tendency of sand bed materials before using fluidized bed waste incineration agglomeration in terms of qualitative analysis and quantitative analysis.

Automated summary

In the incineration process, agglomeration occurs due to the combustion of wastes with adhesion materials like alkalis and earth alkalis. Understanding the properties of sand bed materials is important for the defluidization process. This study analyzed the composition of sand and found that defluidization time decreases with increasing particle size. A model using Support Vector Regression was effective in predicting defluidization time.