PROPIONIC ACID DETECTION AND TEMPERATURE-RELATED MODELING FOR INDUSTRIAL-SCALE COFFEE FERMENTATION

Although propionic acid (HPro) acts as an antimicrobial agent in food preservation, its presence in fermented food is commonly associated with sour/rancid off-flavors during sensory analysis, which may be related to high-temperature fermentation and lack of homogeneity in large-scale bioreactors. This work aimed to evaluate the influence of process temperature in HPro production during arabica coffee (cv. Catiguá MG2) fermentation in a 14,000 L reactor. Natural coffee cultivated at 1048 m in the Cerrado Mineiro region were conditioned to induced solid-state fermentation with natural microbiota in a steel bioreactor (6.38x2.50x0.87 m) for 60 hours, without temperature control. Temperature across horizontal (Xp) and vertical (Yp) mapped positions was monitored with DS18B20 sensors coupled to a data-logger system; height position was fixed at half the total dimension. After fermentation, samples were collected for each (Xp,Yp) coordinate point, and then water-extracted for HPro detection with high-performance liquid chromatography. Empirical mathematical models for temperature and HPro content distribution were determined with multiple regression analysis (p-value < 0.001). Models showed a good experimental data fit and were able to represent temperature profile inside the bioreactor, as well as the temperature-related HPro concentration gradient. Time series for fermentation temperature were defined by a 51-term model (R² = 0.920; RMSE = 0.813; N = 51,136), considering linear, quadratic and interaction effects, alongside implications for polynomial, exponential and sine functions; the response surface showed an increase in general temperature over time due to the exothermic behavior of fermentation and to heat transfer caused by the steel building material, varying from 20°C (beginning) to 25°C (middle) and above 29°C (end). HPro was not detected in non-fermented fruits and varied from 0.15 up to 0.92 mg/g(d.m.) in the fermented samples. Two-dimensional modeling for HPro employed a 6-term model (R² = 0.803; RMSE = 0.071; N = 123), considering linear, quadratic and interaction effects; model coefficients denoted higher influence of linear effect for Xp in HPro production, particularly in regions with higher peak-temperature (35°C). Estimated minimum (Tmin), maximum (Tmax) and average (Tavg) temperatures for the positions were considered as input linear effects in a 4-term model (R² = 0.812; RMSE = 0.052; N = 33); higher correlation with HPro content were observed for higher Tmax and lower Tavg, while Tmin showed less significance in the analysis. Therefore, temperature control should be used in industrial-scale bioreactors to avoid excessive fermentation and HPro undesirable production to guarantee fermented specialty coffee with high-quality parameters.