Fractionation of Corn Stover with Effective Shredding and Separation
A corn stover shredder using a flail type knife and tine shredding mechanism was designed, fabricated, and tested with the objective of improving the shredding extent with corn stover. Two machine designs (open hood with exit chute and closed hood with slots) and three moisture content levels (5, 40, and 70% w.b.) were studied. Shredding was more effective (higher leachate ion conductivity index representing better access to plant nutrients) with the closed hood with slots design than with the open hood design (p<0.001). Extent of shredding was proportional to moisture content at the time of shredding. The attempt to process the corn stover at 70% w.b. was promising as it resulted in a highest leachate ion conductivity (LIC) index of 86%, which is comparable to that of hammer-milled dry stover. Two speeds for the shredding rotors (no-load speeds of 1300 and 700 rpm) were tested with the three moisture levels and the closed-hood with slots design but shredding speed did not affect LIC (p>0.25). To ascertain the tissue make-up of stover samples (rind, pith, leaf, and husk percentages) a multi-trait model was introduced and tested. Three different traits (leachate ion conductivity, tapped density and hyperspectral reflectance intensity) were analyzed for samples of individual tissues as well as constructed calibration blends to build and evaluate the multi-trait model. All samples were milled in a cyclone mill with a 1-mm screen to generate consistent particle size distributions. Tissue type and rewet condition (p<0.0001) affected the individual traits (LIC, tapped density and intensity). Those traits were used in a weighed average model for blends. Model 1 utilized three different trait measurements (LIC, tapped density and intensity), and the mixtures were considered to have four tissues xviii (rind, pith, leaf and husk) in the mixture. By minimizing the objective function, the tissue percentages were estimated and differences from expected ranged from -3.0% to 3.6%. Model 2 utilized only two different trait measurements (LIC and tapped density) and mixtures were considered to have only three components (rind, pith, and LH—leaf and husk were taken as one component) since leaf and husk have similar LIC and tapped density values. Similar to the use of model 1, differences from expected ranged from -5.8 to 7.0%. Model 1 was more accurate than model 2, but model 2 has lower testing costs. With the goal to reasonably isolate rind which is less desirable for hydrolysis or livestock feeding, samples which were shredded under various conditions were separated either “as is” (separated immediately after shredding) or separated after drying in a low-energy simple separator (a slowly rotating drum with varied shape and dimensional openings). Wet separation using the lab-scale drum did not facilitate the rind isolation. Based on predictions of tissue composition in three segmented sample fractions, the best rind separation test (58.6% of rind removed and 61% of favorable material retained) was from the dry separation after the biomass was shredded at 8% w.b. with high speed. It is recommended to do image processing of the separated samples before they are grounded to fines; analysis of larger particles could add additional information of shapes and sizes for tissue percentage verifications.
Buckmaster, Purdue University.
Off-Campus Purdue Users:
To access this dissertation, please log in to our