Process optimization of corn stover compression molding experiments based on response surface method

The compression process parameters (preheating temperature, raw material moisture content, compression speed, compression force, and the holding time etc.) have great influence on the energy consumption and product quality of biomass pellet. Choosing the optimal parameters can improve the pellet quality, as well as reduce energy consumption of molding. Meyer hardness is an important strength index, it reflects the solid's ability to resist deformation. Meyer strength is used to characterize the durability because it's difficult to measure durability in a single pellet formation experiment with standard method. Response surface method is an effective method of parameter optimization. In this research, we used a five-factor BBD experimental design to determine the effects of the raw material moisture content (8%-24%), the preheating temperature (50-150 ℃), compression speed (10-50 mm/min), compression force (51.0-127.4 MPa), the holding time (10-50 s) to the three technical indicators (pellet relaxed density, Meyer hardness, and specific energy consumption). The result showed that within the scope of selected experimental parameters, preheating temperature, material moisture content, and compression force had a larger influence on the three technical indicators, while the impact of compression speed and the holding time was relatively small. With preheating temperature going up from 50 ℃ to 150 ℃, pellet relaxed density and Meyer hardness increased obviously. When preheating temperature was less than 100 ℃, specific energy consumption reduced with preheating temperature increasing. While preheating temperature was more than 100 ℃, with increase of the preheating temperature, specific energy consumption rose. With raw material moisture content going up from 8% to 24%, pellet relaxed density, Meyer hardness and specific energy consumption decreased rapidly. With compression force going up from 51.0 to127.4 MPa, pellet relaxed density, Meyer hardness and specific energy consumption increased clearly. ANOVA analysis was done with the original data, the model was optimized, and a response surface model was established. Relaxed density was fitted with improved quadratic model. Meyer hardness was fitted with simplified quartic model, and specific energy consumption was fitted with another improved quadratic model. Relaxed density, Meyer hardness and specific energy consumption can be calculated through process parameters with the established model, which was used to predict the experimental result. From the model, we concluded that preheating temperature interacted with raw material moisture content. With preheating temperature and raw material moisture content going up, relaxed density and Meyer hardness can remain the same. When preheating temperature was below 100 ℃, preheating temperature went up and raw material moisture content went down. Specific energy consumption can remain the same. When preheating temperature was above 100 ℃, preheating temperature and raw material moisture content went up, specific energy consumption remained the same. And that was consistent with bonding mechanism of moisture and lignin. Considering the existing biomass pellet fuel standard, the relaxed density shouldn't be less than 1000 kg/m³. In the meantime, Meyer hardness can meet the practical requirements. So the requirement of optimization was set as: relaxed density not less than 1000 kg/m³, minimize specific energy consumption. The optimal process parameters were obtained from the model with Design Expert 8.0. The optimal process parameters were: 4 kN (51.0 MPa) for pressure, 110.8 ℃ for temperature and 17% for moisture content. The compression speed and the holding time had little effect on optimization results. Experiment was performed with the optimal process parameters to validate the optimization of process parameters prediction. For relaxed density, Meyer hardness and specific energy consumption relative error between measured and predicted, was 3.01%, 3.79%, 5.54%, respectively. Prediction result was reliable and the model was validated.