The Use of Artificial Neural Networks (ANN) in Food Process Engineering

Abstract

 -Artificial neural networks (ANN) aim to solve problems of artificial intelligence, by building a system with links that simulate the human brain. This approach includes the learning process by trial and error. The ANN is a system of neurons connected by synaptic connections and divided into incoming neurons, which receive stimulus from the external environment, internal or hidden neurons and output neurons, that communicate with the outside of the system. The ANNs present many advantages, such as good adaptability characteristics, possibility of generalization and high noise tolerance, among others. Neural networks have been successfully used in various areas, for example, business, finance, medicine, and industry, mainly in problems of classification, prediction, pattern recognition and control. In the food industry, food processing, food engineering, food properties or quality control, statistical tools are frequently present, and ANNs can process more efficiently data comprising multiple input and output variables. The objective of this review was to highlight the application of ANN to food processing, and evaluate its range of use and adaptability to different food systems. For that a systematic review was undertaken from the scientific literature and the selection of the information was based on inclusion criteria defined. The results indicated that ANN is widely used for modelling and prediction in food systems, showing good accuracy and applicability to a wide range of situations and processes in food engineering.

Development of ANN models based on combined UV‐vis‐NIR spectra for rapid quantification of physical and chemical properties of industrial hemp extracts

“…ANNs have the ability to approximate any non‐linear mathematical function that is useful, especially when the relationship between the variables is unknown, or is complex 24 . ANNs imitate the structure and working of the human nervous system via computer programs in order to assemble information‐processing systems that represent a degree of intelligent behaviour 25,26 . Due to the ability to detect and solve complex non‐linear relationships between inputs (targeted) and outputs (investigated) variables, ANNs have been successfully applied in different areas, from computation to medicine.…”

Rede Neural Artificial e Modelagem matemática na hidratação de cereais matinais com leite

“…It can be noticed that the modeling by Artificial Neural Network was also efficient in adjusting the data obtained for the milk analysis, having good correlation coefficients, and low values of mean relative error (P) and standard error of the estimate (SE). This methodology proves to be quite effective for solving problems in food processing [7]. With the technological advance, there is the search for computational procedures that aim at the best performance and that, as far as possible, present relatively low cost.…”

“…This tool is very advantageous, because it is an alternative for complex functions, does not require detailed information about the system to be modeled, tolerates data loss and allows the adjustment of phenomena of a linear and non-linear nature. Another advantage of ANN is that while the Peleg and others mathematical models use as input only data related to moisture and absorption, and run the model separately for each treatment, the neural network allows to correlate all treatments at once in the input of the model, and can obtain as a response not just parameters of absorption, but also physical-chemical parameters (which are not possible in other models), since in food processing it is important consider the nutritional loss that occurs during the process [7,8].…”

“…Figures 4 show the correlation graphs between the experimental (observed) and estimated values by the Artificial Neural Network for the physical-chemical analyzes of the hydration milk.It can be noticed that the modeling by Artificial Neural Network was also efficient in adjusting the data obtained for the milk analysis, having good correlation coefficients, and low values of mean relative error (P) and standard error of the estimate (SE). This methodology proves to be quite effective for solving problems in food processing[7].…”

“…The Artificial Neural Network (ANN) is a computational and nonlinear statistical technique capable of solving a range of highly complex problems and heltp to controll of food processing operations through a Multilayer Perceptron neural network (MLP). An MLP type consists of an input set, which forms the input layer, one or more hidden layers and an output layer [6,7,8].…”