Prediction on microwave-assisted elimination of biomass tar using back propagation neural network

Microwave-assisted cracking is an emerging technique for biomass tar elimination, but lacks the artificial intelligence application for predicting the elimination efficiency and the optimal multi-correlated reaction conditions. In this work, we investigated the microwave-assisted cracking of a biomass tar model compound at various operating conditions and established a back propagation (BP) neural network based on the realistic experimental data to predict the highest elimination performance and obtain the corresponding optimal operation parameters. Results show that the xylene elimination efficiency increases monotonically with reaction temperature (T-r) at 600-850 degrees C and inlet xylene concentration (C-x) of 38.4 to 42.2 g/m(3). It is also affected by gas hourly space velocity (GHSV) but rarely influenced by the SiC particle diameter (d). The xylene was mainly cracked into methane and toluene, and the cracking reactions involve both direct decomposition and radical-induced cracking. Based on a three-layer BP model, the highest xylene elimination efficiency was predicted to be 95.7%, and the main optimal parameters including T-r, GHSV, C-x, and d are 800 degrees C, 27.3 h(-1), 38.4 g/m(3), and 1 mm, respectively. The predicted result shows a good accuracy after experimental validation.