Artificial neural networks to differentiate the composition and pyrolysis kinetics of fresh and long-stored maize

Application of triple-branch artificial neural network system for catalytic pellets combustion

On the international level, it is common to act on reducing emissions from energy systems. However, in addition to industrial emissions, low-stack emissions also make a significant contribution. A good step in reducing its environmental impact, is to move to biofuels, including biomass. This paper examines the impact of placing a catalytic system in a retort boiler to minimize emissions of greenhouse gases, dust and other pollutants when burning pellets. The effect of platinum, and oxides of selected metals placed on the deflector as a solid catalyst was studied. Based on the experimental data, a branched artificial neural network was constructed and trained. The routing of three parallel topologies made it possible to achieve high accuracy while keeping the input data relatively simple. The system showed an average error of 3.54% against arbitrary test data. On the basis of experimental data as well as predictions returned by the artificial neural network, recommendations were shown for the catalysts used and their amounts. Depending on the biomass from which the pellet was produced, the experiment suggested the use of titanium or copper oxides. In the case of the neural network, it was able to select a better system, based on platinum, improving emission reductions by up to more than 19%, depending on the type of pellet used.

Advanced dual-artificial neural network system for biomass combustion analysis and emission minimization

Management of agricultural waste is an important part of plantation operations. Not all wastes are suitable for composting or the process is simply inefficient and time-consuming. In their case, thermal treatment is acceptable, but it is necessary to optimize the process to minimize greenhouse gas emissions. This article investigates the feasibility of constructing artificial neural networks (ANNs) to predict feedstock and emission parameters from the combustion of vineyard biomass. In order to maximize accuracy while avoiding overfitting of the ANN, a novel dual-ANN system was proposed. It consisted of two cascade-forward ANNs trained on independent data, each with three hidden layers. A benchmark showed that the final networks had a relative error in the range of 0.81-2.83%, and the resulting dual-ANN up to a maximum of 2.09%. Based on the ANN, it was possible to make recommendations on the parameters of the feedstock that would be suitable for obtaining bioenergy. It was noted that the best calorific values are shown by waste from plants with an intermediate amount, distribution, and mass of leaves, with relatively low average leaf area. Emissivity reduction, however, requires significantly different conditions. Preference is given to waste from plants that have high amounts of leaves but are spread over many stems - that is, plants that are highly shrubby during the growing season. This proves that it is not possible to formulate universal recommendations that are both energy- and carbon-beneficial, but outlines a clear direction where consensus should be sought, depending on the goals adopted.

Artificial Neural Network Study on the Pyrolysis of Polypropylene with a Sensitivity Analysis

Among machine learning (ML) studies, artificial neural network (ANN) analysis is the most widely used technique in pyrolysis research. In this work, the pyrolysis of polypropylene (PP) polymers was established using a thermogravimetric analyzer (TGA) with five sets of heating rates (5-40 K min^-1). TGA data was used to exploit an ANN network by achieving a feed-forward backpropagation optimization technique in order to predict the weight-left percentage. Two important ANN model input variables were identified as the heating rate (K min^-1) and temperature (K). For the range of TGA values, a 2-10-10-1 network with two hidden layers (Logsig-Tansig) was concluded to be the best structure for predicting the weight-left percentage. The ANN demonstrated a good agreement between the experimental and calculated values, with a high correlation coefficient (R) of greater than 0.9999. The final network was then simulated with the new input data set for effective performance. In addition, a sensitivity analysis was performed to identify the uncertainties associated with the relationship between the output and input parameters. Temperature was found to be a more sensitive input parameter than the heating rate on the weight-left percentage calculation.