Modelling Acetification with Artificial Neural Networks and Comparison with Alternative Procedures

Abstract

<jats:p>Modelling techniques allow certain processes to be characterized and optimized without the need for experimentation. One of the crucial steps in vinegar production is the biotransformation of ethanol into acetic acid by acetic bacteria. This step has been extensively studied by using two predictive models: first-principles models and black-box models. The fact that first-principles models are less accurate than black-box models under extreme bacterial growth conditions suggests that the kinetic equations used by the former, and hence their goodness of fit, can be further improved. By contrast, black-box models predict acetic acid production accurately enough under virtually any operating conditions. In this work, we trained black-box models based on Artificial Neural Networks (ANNs) of the multilayer perceptron (MLP) type and containing a single hidden layer to model acetification. The small number of data typically available for a bioprocess makes it rather difficult to identify the most suitable type of ANN architecture in terms of indices such as the mean square error (MSE). This places ANN methodology at a disadvantage against alternative techniques and, especially, polynomial modelling.</jats:p>

Optimization of the Acetification Stage in the Production of Wine Vinegar by Use of Two Serial Bioreactors

“…In practice, it is necessary to optimize the outcome of the acetification process under operationally restricted conditions, for instance, the end product must have a very little ethanol concentration but if ethanol in the culture medium is almost depleted, the high acidity and lack of substrate would affect very negatively the action of AAB in next cycles. In this regard, the use of two identical serially arranged bioreactors working in a repeated semi-continuous mode, see Figure 1, could be a proper alternative to overcome the problem; this possibility is also suggested from several reported models for the acetification process [7][8][9][10][11][18][19][20][21][22][23][24][25][26][27]. In many vinegar plants the main reactors (bioreactor B1 in Figure 1) are partially unloaded in additional depletion fermenters (bioreactor B2 in Figure 1) where the ethanol concentration could be exhausted.…”

“…The acetification process, which involves a bio-oxidative transformation effected by a mixture of acetic acid bacteria (AAB) [5], leads to the biological conversion of ethanol into acetic acid. Industrial acetification bioreactors usually operate in an automated repeated semi-continuous mode, so that once the fermenters are fully loaded, the ethanol concentration is allowed to decrease to a preset level and then a certain fraction of the reactor volume is unloaded, actuating the remainder as an inoculum for the next conversion cycle [6][7][8][9][10][11]. After the bioreactor is unloaded, it is slowly replenished with fresh alcoholic substrate to start a new ethanol cycle.…”

“…Among the different approaches used for modelling this process, the black-box models based on second order generalized polynomials [7,11,[24][25][26][27] showed to be the best alternatives because of their simplicity and accurate predictions. Because of the previous comments and with the aim of optimizing the process, a first work was carried out for modelling the behavior of two serially arranged bioreactors working in a repeated semi-continuous mode [27].…”

Modelling of the Acetification Stage in the Production of Wine Vinegar by Use of Two Serial Bioreactors

“…Usually, an acetification bioreactor is operated in an automated manner in order to not alter the spontaneous dynamics of the system. In practice, this allows repeated cycling to be easily and rapidly achieved [19]. Also, the characteristics of AAB [20][21][22][23] make cultivation and selection outside their typical natural or industrial environment rather difficult.…”

“…In this scenario, modelling of the two serial bioreactors system requires the use of an appropriate experimental design. Previous experience with modelling of the biotransformation stage in the vinegar production process suggests that black-box models based on second-order generalized polynomials [7,19,[32][33][34] provide a more accurate depiction of the experimental results than do existing alternatives, allowing it to describe potential interactions between independent variables to be considered-and with added advantages like ease of development and statistical validation. The number of experiments needed for the accurate fitting of a polynomial equation depends on whether the polynomial is linear or non-linear [35].…”