In the quest for sustainable and efficient bio-based production, lactic acid stands as a cornerstone, pivotal in the creation of biodegradable plastics, pharmaceuticals, and food products. Yet, the optimization of lactic acid fermentation, particularly when leveraging heterogeneous substrates like agricultural residues or municipal bio-waste, has been a persistent challenge. Traditional methods of manual sampling and offline chemical analysis are not only labor-intensive but also impractical at scale. Enter machine learning models coupled with spectroscopic sensors, offering a promising alternative. However, these models often stumble when faced with varying substrates, limiting their industrial deployment.
A recent study published in *Results in Engineering* introduces Shapley Additive exPlanations-based Domain Adaptation (ShapDA), an innovative framework that integrates model explainability with Fourier transform infrared (FTIR) spectroscopy. Unlike conventional domain adaptation methods, ShapDA identifies a compact set of spectral features that remain stable across different substrate domains. This breakthrough enables calibration-free prediction under novel conditions, a significant leap forward in the field.
“ShapDA captures variable-specific spectral signatures, making it possible to monitor fermentation processes across diverse substrates without the need for additional chemical measurements,” explains lead author Majharulislam Babor, a researcher at the Leibniz Institute for Agricultural Engineering and Bioeconomy (ATB) in Potsdam, Germany. This capability is particularly valuable in the agriculture sector, where the variability of substrates can be substantial.
The study applied ShapDA to 21 glucose-based fermentation batches, each consisting of spectra and corresponding chemical measurements. The model was then transferred to 83 batches of bio-waste fermentation, spanning diverse substrates, production scales, and microbial strains, without requiring additional chemical measurements. The results were impressive, with ShapDA outperforming baseline and state-of-the-art methods. It reduced the root mean squared error by 50% for glucose and 65% for lactic acid in bio-waste fermentations.
The implications for the agriculture sector are profound. By enabling substrate-independent monitoring, ShapDA can significantly enhance the efficiency and scalability of lactic acid production from waste materials. This not only supports the circular bioeconomy but also opens up new avenues for sustainable agriculture practices.
As the world grapples with the challenges of climate change and resource depletion, innovations like ShapDA offer a glimmer of hope. By making fermentation monitoring more robust and adaptable, this technology can help unlock the full potential of bio-based production, paving the way for a more sustainable future.
The research, led by Majharulislam Babor from the Leibniz Institute for Agricultural Engineering and Bioeconomy (ATB) in Potsdam, Germany, represents a significant advancement in the field of agritech. With the code available on GitHub, the potential for further development and application is vast, promising to shape the future of lactic acid production and beyond.