Sensor model fusion reveals the dynamics of lactobacillus fermentation with real time concentrations of lactic and acetic acids in ensiling of maize and ryegrass

Abstract

Anaerobic lactobacillus (LAB) fermentation is key to production of silage. Tracking the dynamics of bacterial metabolism is important for the improvement of LAB fermentation and for evaluation of potential additives. This is currently limited by available ex situ analytical methods, while commercial sensors of lactic and acetic acid content remain unavailable. Here we validate an in situ pH sensor and devise a model-sensor package and a mini-fermenter system (1.5 L) for on-line analysis of LAB fermentation. By the fusion of the model based on mirror mapping principle and the time course of pH measured in situ, a general solution of the dynamic accumulation of organic acid (DAOA) and percentage yield of organic acid (PYOA, 0 ∼ 100 %) is obtained as a function of time. We link PYOA to the initial and final values of lactic and acetic concentrations analyzed ex situ to obtain specific solutions for the time courses of lactic and acetic acid production. We demonstrate the model-sensor system with fermentation of both maize and ryegrass, capturing the dynamic patterns of fermentation, including the rapid depletion of O2 in the initial aerobic phase (maize: ≈ 1.5 h, ryegrass: ≈ 9.2 h), the exponential decline of pH (R2 ≥ 0.99, RMSE ≤ 0.043) in the subsequent anaerobic phase, and metabolic feedbacks of pH on instantaneous acid production (ΔpH; inflection at pH 5), and on cumulative yields of total acidity and specific organic acids. These are all previously unavailable data streams.