Rapid in-process moisture determination on a fluid bed dryer

Significance of the topic

The control of residual moisture in fluid bed dryers is a critical unit operation in solid oral pharmaceutical manufacturing because incorrect drying leads to product loss, stability issues and inefficient energy use. Fast, accurate, in-line moisture measurement supports process optimization, reduces reliance on slow destructive laboratory methods and enables real-time endpoint detection and closed-loop control, lowering operating costs and product rejection rates.

Objectives and study overview

This application study evaluated the performance of a Thermo Scientific Antaris™ MX FT-NIR process analyzer with a fiber-optic reflection probe (purgeable-tip) installed in a production-style fluid bed dryer. Goals were to demonstrate in-line, non-destructive moisture quantification across a broad moisture range, compare to Karl Fischer reference analysis, and illustrate practical sampling/cleaning strategies for probe operation in a fluidized environment.

Methodology and instrumentation

Key experimental design features:

• Probe and sampling: A diffuse-reflectance fiber-optic probe with a purgeable-tip was mounted into the fluid bed dryer. The dryer operated in a cycle composed of 120 seconds purge-drying and ~10 seconds filter/shaking; spectra were acquired during the purge step.
• Spectral acquisition: Antaris MX FT-NIR analyzer collected 12 scans per sample at 8 cm⁻¹ resolution over 4800–10000 cm⁻¹. This produced an effective data point per dryer cycle (~every 2 minutes).
• Reference method: Karl Fischer titration (destructive) was used as the primary reference, with each sample run in duplicate (~10 minutes per KF result).
• Sample set: Multiple dryer batches were collected covering 1–23% moisture to build calibration and validation sets.
• Data preprocessing and chemometrics: Spectra were preprocessed using 1st derivative, Multiplicative Scatter Correction (MSC) and a Savitzky–Golay filter (7-point window, 3rd polynomial order). Partial Least Squares (PLS) regression with five latent factors was used to build the moisture model.

Instrumentation used

• Thermo Scientific Antaris MX FT-NIR process analyzer
• Diffuse reflectance fiber-optic probe with purgeable-tip (installed in dryer)
• Retractable probe design with internal solvent-cleaning chamber described as alternative for high-moisture or sticky products
• Spectral configuration: 4800–10000 cm⁻¹ range, 8 cm⁻¹ resolution, 12 scans per acquisition
• Karl Fischer titrator for reference moisture determinations

Main results and discussion

Calibration and validation performance:

• Calibration (RMSEC): 0.56% moisture; calibration correlation coefficient R = 0.995.
• Cross-validation (RMSECV): 1.05% moisture; cross-validation R = 0.981.

Key observations and interpretation:

• Water yields strong NIR signals (combination band ~5150 cm⁻¹ and 1st overtone near 7000 cm⁻¹), enabling high sensitivity and reliable quantification over the 1–23% range.
• Preprocessing (MSC + 1st derivative + Savitzky–Golay) improved robustness by compensating for variable pathlength and light scattering caused by fluctuating particle distance and particle-size distributions in the fluidized bed.
• Purgeable-tip probes are effective for many applications, particularly for products < ~15% moisture where powder can be blown off the tip; retractable probes with solvent cleaning are preferable for higher moisture or sticky products.
• In-line FT-NIR provides much faster feedback than laboratory KF (data every ~2 minutes versus ~10 minutes per KF duplicate), enabling rapid endpoint detection and potential closed-loop control of drying.

Benefits and practical applications of the method

• Non-destructive, in-line measurement reduces the need for sample removal and laboratory resources (equipment, reagents, operators).
• Rapid acquisition and processing allow near-real-time monitoring of dryer moisture and clear identification of drying endpoints, reducing over-drying and associated product loss and energy waste.
• Enables PAT-driven process control and closed-loop operation to stabilize product quality and increase throughput.
• Model transferability: Using a common FT-NIR platform supports method migration from R&D to manufacturing with lower risk of performance loss.
• Multicomponent potential: FT-NIR can be extended to other process analyses (raw material ID, blend uniformity, assay) using the same hardware platform.

Future trends and potential applications

• Deeper integration with process control systems to enable automated closed-loop drying where FT-NIR moisture readings directly adjust heater/airflow or feed rates.
• Improvements in probe cleaning and purge technologies (faster retractable/automated cleaning) to extend applicability to sticky, high-moisture, or coating processes.
• Advanced chemometric strategies and transfer learning for robust model portability across different dryers, product types and scales.
• Multivariate monitoring combining moisture with API content and blend uniformity models to achieve simultaneous end-point criteria and comprehensive process control.
• Regulatory acceptance and standardized PAT workflows that simplify validation and lifecycle management of in-line NIR methods.

Conclusion

The Antaris MX FT-NIR with a purgeable-tip/reftractable fiber probe demonstrated accurate, fast and robust in-line moisture determination in a fluid bed dryer across a wide moisture range (1–23%). Good calibration statistics (RMSEC 0.56%, R = 0.995; RMSECV 1.05%, R = 0.981) show the method is suitable for real-time endpoint detection and process control. Implementing this FT-NIR approach can reduce laboratory burden, save energy, minimize product loss from over-drying and support PAT-driven manufacturing improvements.

Reference

Heil C. Application note: Rapid in-process moisture determination on a Fluid Bed Dryer. Thermo Fisher Scientific. Application note | 51543. 2007 (document AN51543).