A Machine Vision Approach for Bioreactor Foam Sensing.

Area of Science:

• Biotechnology
• Machine Learning
• Sensor Technology

Background:

• Machine vision, powered by machine learning, has advanced significantly, with applications in consumer electronics, automotive, and quality control.
• Bioprocessing faces challenges with foam detection, leading to batch failures and inefficient antifoam agent use.
• Robust foam sensing is crucial for optimizing upstream bioprocesses.

Purpose of the Study:

• To develop a low-cost, flexible, and reliable foam sensor for bioreactor applications.
• To leverage convolutional neural networks (CNNs) for accurate foam detection and classification.
• To address the limitations of current foam sensing methods in bioprocessing.

Main Methods:

• Implementation of a novel foam sensor concept utilizing machine vision.
• Application of convolutional neural networks (CNNs), a state-of-the-art machine learning technique for image processing.
• Testing the sensor's accuracy in binary foam detection (foam/no foam) and multi-level foam classification.

Main Results:

• The developed foam sensor demonstrates high accuracy in detecting the presence or absence of foam.
• The system achieves fine-grained classification of different foam levels within the bioreactor.
• The low-cost and flexible nature of the sensor makes it suitable for various bioprocessing environments.

Conclusions:

• The CNN-based machine vision system offers a reliable solution for foam detection in bioreactors.
• This technology has the potential to significantly reduce batch failures and improve bioprocess efficiency.
• The developed sensor represents a promising advancement in bioprocess monitoring and control.