A Flow-Through Multi-Wavelength Sensor with Machine-Learning Calibration for Real-Time Monitoring of Microalgae...

Area of Science:

• Biotechnology
• Optical Sensing
• Machine Learning

Background:

• Traditional microalgae monitoring relies on time-consuming offline wet-chemical analyses.
• Real-time biological data is crucial for optimizing microalgae cultivation processes.

Purpose of the Study:

• To develop a flow-through, multi-wavelength visible-light (VIS) sensor for real-time monitoring of microalgae.
• To create machine-learning models for estimating biomass and pigment concentrations.

Main Methods:

• Utilized 209 experimental data points to develop six machine-learning regression models.
• Estimated concentrations of dry biomass, chlorophylls, total chlorophyll, total carotenoid, and astaxanthin.
• Validated the sensor system under continuous operation, assessing biofilm formation and prediction accuracy.

Main Results:

• Biomass and astaxanthin predictions were within ±10% of offline reference measurements.
• The sensor platform demonstrated potential for improved hydrodynamics and automated operation.
• Machine-learning calibration enabled practical soft-sensor functionality for real-time bioprocess monitoring.

Conclusions:

• The proposed low-cost VIS sensor and ML calibration offer a practical solution for real-time microalgal bioprocess monitoring.
• This technology provides a foundation for integrating model-based and predictive control strategies.
• The system enhances efficiency and accuracy in microalgae cultivation management.