Monitoring continuous mixing process dynamics through the NIR spectra baseline using a stream sampler

Area of Science:

• Pharmaceutical Manufacturing
• Process Analytical Technology (PAT)
• Spectroscopy

Background:

• Uncontrolled powder flow in direct compression continuous manufacturing causes spectral baseline variations.
• Existing methods rely on spectral preprocessing to eliminate baseline drift, potentially obscuring process dynamics.
• Effective real-time process monitoring requires stable spectral data.

Purpose of the Study:

• To control spectral baseline variation using a stream sampler for direct compression continuous manufacturing.
• To evaluate process dynamics during continuous mixing using real-time baseline monitoring.
• To establish a method for distinguishing mass steady-state from mass flow variations.

Main Methods:

• Utilized a stream sampler operated at 8 RPM and 35 kg/h throughput to ensure uniform powder flow.
• Employed the Moving Block Standard Deviation (MBSD) method to monitor real-time baseline variation, integrated within the synTQ program.
• Determined a threshold MBSD at mass steady state using calibration blends (40-60% w/w acetaminophen).
• Investigated Partial Least Squares (PLS) regression models with and without spectral transformation.
• Applied variographic analysis to assess overall process variance.

Main Results:

• The stream sampler achieved a reproducible and stable spectral baseline during mass steady state, attributed to confined powder flow.
• The threshold MBSD successfully differentiated mass steady-state from mass flow variations during continuous mixing.
• PLS models without spectral transformation showed lower prediction bias and RMSEP for independent blends.
• Variographic analysis provided insights into sampling and analytical errors.

Conclusions:

• A stream sampler effectively controls spectral baseline variation in direct compression continuous manufacturing.
• Real-time MBSD monitoring provides a reliable method for process control and distinguishing steady-state from dynamic conditions.
• Optimized PLS models without spectral transformation enhance prediction accuracy for blend uniformity.