(120c) Near-Infrared Spectroscopy for On-Line Drying End-Point Determination In a Small-Scale Fluidized Bed
AIChE Annual Meeting
2011
2011 Annual Meeting
Food, Pharmaceutical & Bioengineering Division
Scale-up of Pharmaceutical Isolation and Drying Operations
Monday, October 17, 2011 - 1:10pm to 1:30pm
A method based on near-infrared spectroscopy (NIRS) with partial least squares regression (PLSR) was developed to determine the drying end-point of fluid bed dried dibasic calcium phosphate anhydrous (DCPA). In a first approach, parameters were adjusted according to experimental designs to induce structured variations to (i) the processing, i.e. changing air flow-rates and amounts of DCPA, and (ii) the material properties, i.e. changing DCPA particle size compositions. Response surface modeling was consequently applied to predict the drying end-points at varying process and material characteristics within the confined design regions. Laser diffraction measurements of the DCPA before and after the fluidized bed drying revealed particle size changes due to the processing, i.e. potential particle agglomeration, attrition, breakage or fracture. In a second approach, NIRS was implemented to quantify the residual water content of the DCPA. Therefore the NIR spectra were recorded on-line with a fiber optic probe from outside the processor and off-line, i.e. samples were consecutively thief-probed from the processor, put into glass vials, and analyzed with an integrating sphere device. Furthermore, background spectra of dry DCPA which was fluidized at respective process air velocities were collected in an effort to correct for potential absorption interferences from the processor material (polymethylmethacrylate) and the glass vials. Subsequently, the performances of the thereby developed PLSR models with traditionally developed PLSR models (Spectralon® as a background material) were compared. Loss-on-drying (LOD) was used as a reference method and confirmed the actual moisture content of the DCPA. The experiments showed that (i) the thief-probed sample amount from the processor and thus the sample weight, and (ii) the down-time between taking a sample and its actual analysis with NIR and LOD, can bias the moisture content predictions. Consolidation variations of the DCPA inside the glass vials did not play a significant role. Continuous mixing, and thus the collection of representatively averaged spectra for the PLSR modeling and for further predictions, and a maximum available sample amount during the spectral data acquisition are considered the main advantages of dynamic or on-line NIR monitoring over NIR off-line analysis.
[1] N. Heigl et al; Quantitative On-Line vs. Off-line NIR Analysis of a Fluidized Bed Drying Process with Consideration of Sub-sampling Issues; Manuscript in preparation.
[2] N. Heigl et al; Near-infrared Spectroscopy with Experimental Design for Characterizing the Fluidized Bed Drying Process of Dibasic Calcium Phosphate Anhydrous; Manuscript in preparation.