Experimental and Artificial Neural Network Modeling of a Upflow Anaerobic Contactor (UAC) for Biogas Production from Vinasse

Abstract

A pilot scale Upflow Anaerobic Contactor (UAC), based on upflow sludge blanket principle, was designed to treat vinasse waste obtained from beet molasses fermentation. An assessment of the anaerobic digestion of vinasse was carried out for the production of biogas as a source of energy. Average Organic loading rate (OLR) was around 7.5 gCOD/m3/day in steady state, increasing upto 8.1 gCOD/m3/day. The anaerobic digestion was conducted at mesophilic (30–37 °C) temperature and a stable operating condition was achieved after 81 days with average production of 65 % methane which corresponded to a maximum biogas production of 85 l/day. The optimal performance of UAC was obtained at 87 % COD removal, which corresponded to a hydraulic retention time of 16.67 days. The biogas production increased gradually with OLR, corresponding to a maximum 6.54 gCOD/m3/day (7.4 % increase from initial target). A coupled Artificial Neural Network-Differential Evolution (ANN-DE) methodology was formulated to predict chemical oxygen demand (COD), total suspended solids (TSS) and volatile fatty acids (VFA) of the effluent along with the biogas production. The method incorporated a DE approach for the efficient tuning of ANN meta-parameters such as number of nodes in hidden layer, input and output activation function and learning rate. The model prediction indicated that it can learn the nonlinear complex relationship between the parameters and able to predict the output of the contactor with reasonable accuracy. The utilization of the coupled ANN-DE model provided significant improvement to the study and helps to study the parametric effect of influential parameters on the reactor output.

Comparison of optimization approaches (response surface methodology and artificial neural network‐genetic algorithm) for a novel mixed culture approach in soybean meal fermentation

“…From the given set of data, ANN was supposed to be able to learn the cause‐and‐effect relationship between the input and output variables and then minimize the error between the target data and simulated output (Dibaba et al, ; Jacob & Banerjee, ). For each output neuron, back propagation method was used to correct the connection weights and biases for each output neuron (Desai et al, ; Dibaba et al, ; Lahiri & Ghanta, ). This is called “training” and was repeated until the overall error value reached nearly zero.…”

“…In the present study, the multilayer ANN approach along with the implementation of GA was achieved using RapidMiner StudioV7.6 (RapidMiner Inc., Boston, MA) as the data science platform. For a detailed study on ANN and GA, the interested reader may refer to Dibaba et al (), Gonzalez‐Fernandez et al (), Lahiri and Ghanta (), and Tan and Karr (2017) to name a few.…”

“…However, the main limitation of this technique is that it can assume only quadratic nonlinear correlation and therefore requires a narrow search window (Desai, Survase, Saudagar, Lele, & Singhal, ), finding which requires either extra experiments or good prior knowledge of the system. In the recent past, artificial neural network (ANN) has proved its capability as an attractive alternative in situations where the development of phenomenological or conventional regression models for nonlinear processes becomes impractical or cumbersome (Dibaba, Lahiri, T'Jonck, & Dutta, ). ANN is being frequently used for the classification and prediction of parameters (Acevedo Correa, Margarita, Castillo, & Martelo, ); various food safety aspects like modeling of microbial growth and thereby predicting food safety; food quality investigations like interpretation of spectroscopic data, prediction of physical, chemical, functional, and sensory properties during both processing and storage (Huang, Kangas, & Rasco, ; Tan & Kerr, ).…”

Post-combustion artificial neural network modeling of nickel-producing multiple hearth furnace

“…The most widely utilized ANN paradigm is the Multi Layered Perceptron (MLP) (see Figure 3) that approximates nonlinear relationships existing between an input set of data (causal process variables) and the corresponding output (dependent variables) data set [44]. The main reason for this selection is its ability to model, in a simple way, complex functional relationships [45]. This has been proved through numerous practical applications [46][47][48].…”

Forecast of the outbreak of COVID-19 using artificial neural network: Case study Qatar, Spain, and Italy

“…Recently, artificial neural networks (ANN) have been successfully used for the prediction of the evolution of different systems with a high degree of accuracy [48] . The ANN demonstrated an excellent prediction accuracy short time, for different engineering-based processes such as wastewater treatment [49] , [50] , electro-dialysis separation, metal removal [51] , biosorption of heavy metals [52] , anaerobic digestion [53] , biofuel production [54] , [55] , [56] , [57] , [52] , cell growth rate [58] and population growth [59] . The ANN architecture is a simple and fast methodology to predict the process output compared with the complicated physically related models.…”