Víctor Martínez-Martínez

An Artificial Neural Network based expert system fitted with Genetic Algorithms for detecting the status of several rotary components in agro-industrial machines using a single vibration signal

The statuses of rotary components of agro-industrial machines are estimated.A single vibration point is enough to estimate several rotary elements statuses.An Artificial Neural Network can be used to estimate the rotary elements statuses.Genetic Algorithms improve the estimation performance and the training time.No feature selection stage is needed to estimate the rotary elements statuses. This article proposes (i) the estimation method of an expert system to predict the statuses of several agro-industrial machine rotary components by using a vibration signal acquired from a single point of the machine; and, (ii) a learning method to fit the estimation method. Both methods were evaluated in an agricultural harvester. Vibration signal data were acquired from a single point of the harvester under working conditions, by varying (1) the engine speed status (high speed/low speed), (2) the threshing operating status (on/off), (3) the threshing balance status (balanced/unbalanced), (4) the chopper operating status (on/off), and (5) the chopper balance status (balanced/unbalanced). Positive frequency spectrum coefficients of the vibration signal were used as the only inputs of an Artificial Neural Network (ANN) that predicts the five rotary component statuses. Four Genetic Algorithm (GA) based learning methods to fit the ANN weights and biases were implemented and its performance was compared to select the best one. The prediction system that is developed was able to estimate the rotary component status under consideration with a mean success rate of 92.96%. Moreover, the best GA-based learning method that was implemented reduced the number of generations by 70% in the best case, compared with a random learning method, allowing a similar reduction in the time needed to reach the expected success rate. The results obtained suggest that (i) an ANN-based expert system could estimate the status of the rotary components of an agro-industrial machine to a high degree of accuracy by processing a vibration signal acquired from a single point on its structure; and, (ii) by using the best implementation of the GA-based learning method proposed to fit the ANN weights and biases, it is possible to improve the success rate and by doing so to reduce the time needed to perform the adjustment. The main contribution of this work is the proposal of a classification method that estimates the status of several rotary elements placed each one far from the others employing the signal acquired from only one accelerometer and non-requiring a feature extraction stage.

An SVM-Based Classifier for Estimating the State of Various Rotating Components in Agro-Industrial Machinery with a Vibration Signal Acquired from a Single Point on the Machine Chassis

The goal of this article is to assess the feasibility of estimating the state of various rotating components in agro-industrial machinery by employing just one vibration signal acquired from a single point on the machine chassis. To do so, a Support Vector Machine (SVM)-based system is employed. Experimental tests evaluated this system by acquiring vibration data from a single point of an agricultural harvester, while varying several of its working conditions. The whole process included two major steps. Initially, the vibration data were preprocessed through twelve feature extraction algorithms, after which the Exhaustive Search method selected the most suitable features. Secondly, the SVM-based system accuracy was evaluated by using Leave-One-Out cross-validation, with the selected features as the input data. The results of this study provide evidence that (i) accurate estimation of the status of various rotating components in agro-industrial machinery is possible by processing the vibration signal acquired from a single point on the machine structure; (ii) the vibration signal can be acquired with a uniaxial accelerometer, the orientation of which does not significantly affect the classification accuracy; and, (iii) when using an SVM classifier, an 85% mean cross-validation accuracy can be reached, which only requires a maximum of seven features as its input, and no significant improvements are noted between the use of either nonlinear or linear kernels.

Moisture Content Prediction in the Switchgrass (Panicum virgatum) Drying Process Using Artificial Neural Networks

This article proposes two artificial neural network (ANN)-based models to characterize the switchgrass drying process: The first one models processes with constant air temperature and relative humidity and the second one models processes with variable air conditions and rainfall. The two ANN-based models proposed estimated the moisture content (MC) as a function of temperature, relative humidity, previous MC, time, and precipitation information. The first ANN-based model describes MC evolution data more accurately than six mathematical empirical equations typically proposed in the literature. The second ANN-based model estimated the MC with a correlation coefficient greater than 98.8%.

Temperature and relative humidity estimation and prediction in the tobacco drying process using Artificial Neural Networks.

This paper presents a system based on an Artificial Neural Network (ANN) for estimating and predicting environmental variables related to tobacco drying processes. This system has been validated with temperature and relative humidity data obtained from a real tobacco dryer with a Wireless Sensor Network (WSN). A fitting ANN was used to estimate temperature and relative humidity in different locations inside the tobacco dryer and to predict them with different time horizons. An error under 2% can be achieved when estimating temperature as a function of temperature and relative humidity in other locations. Moreover, an error around 1.5 times lower than that obtained with an interpolation method can be achieved when predicting the temperature inside the tobacco mass as a function of its present and past values with time horizons over 150 minutes. These results show that the tobacco drying process can be improved taking into account the predicted future value of the monitored variables and the estimated actual value of other variables using a fitting ANN as proposed.

On-line monitoring of oxygen as a method to qualify the oxygen consumption rate of wines

Measuring the oxygen content during winemaking and bottle storage has become increasingly popular due to its impact on the sensory quality and longevity of wines. Nevertheless, only a few attempts to describe the kinetics of oxygen consumption based on the chemical composition of wines have been published. Therefore, this study proposes firstly a new fitting approach describing oxygen consuming kinetics and secondly the use of an Artificial Neural Network approach to describe and compare the oxygen avidity of wines according to their basic chemical composition (i.e. the content of ethanol, titratable acidity, total sulfur dioxide, total phenolics, iron and copper). The results showed no significant differences in the oxygen consumption rate between white and red wines, and allowed the sorting of the wines studied according to their oxygen consumption rate.

Method to estimate the medullar rays angle in pieces of wood based on tree-ring structure: application to planks of Quercus petraea

Estimating wood parameters employing non-destructive methods has been widely studied in recent years. The choice of wood used to build wine ageing barrels (cooperage) is strongly influenced by wood anatomy and specifically by the orientation of medullar rays among other aspects. In this article, a method based on the regularities of the tree-ring structure to estimate the medullar ray angle of the cross section of a piece of wood is proposed. This angle shows the direction of the best linear path to evaluate several tree-ring features and could be employed to automate tasks, such as introducing an analysis path or rotating the image prior to the analysis, which some dendro analysis methods require. A dataset of 26,992 synthetic images and 110 real oak wood images was used to validate the approach. The medullar ray angle of each image considered was measured manually and estimated using the method proposed here, which employs the fast Fourier transform (FFT) to take advantage of the tree-ring structure regularities and find the direction angle of the best linear path to evaluate several tree-ring features. The results obtained demonstrate a mean squared error of 0.29° and 8.19° and a mean absolute error of 0.19° and a 5.91° for the synthetic and oak wood images, respectively. These data suggest the suitability of the proposed method as part of an automated system to inspect and analyse the growth rings in oak wood planks.

Leaf and canopy reflectance spectrometry applied to the estimation of angular leaf spot disease severity of common bean crops

This study is aimed at (i) estimating the angular leaf spot (ALS) disease severity in common beans crops in Brazil, caused by the fungus Pseudocercospora griseola, employing leaf and canopy spectral reflectance data, (ii) evaluating the informative spectral regions in the detection, and (iii) comparing the estimation accuracy when the reflectance or the first derivative reflectance (FDR) is employed. Three data sets of useful spectral reflectance measurements in the 440 to 850 nm range were employed; measurements were taken over the leaves and canopy of bean crops with different levels of disease. A system based in Principal Component Analysis (PCA) and Artificial Neural Networks (ANN) was developed to estimate the disease severity from leaf and canopy hyperspectral reflectance spectra. Levels of disease to be taken as true reference were determined from the proportion of the total leaf surface covered by necrotic lesions on RGB images. When estimating ALS disease severity in bean crops by using hyperspectral reflectance spectrometry, this study suggests that (i) successful estimations with coefficients of determination up to 0.87 can be achieved if the spectra is acquired by the spectroradiometer in contact with the leaves, (ii) unsuccessful estimations are obtained when the spectra are acquired by the spectroradiometer from one or more meters above the crop, (iii) the red to near-infrared spectral region (630–850 nm) offers the same precision in the estimation as the blue to near-infrared spectral region (440–850), and (iv) neither significant improvements nor significant detriments are achieved when the input data to the estimation processing system are the FDR spectra, instead of the reflectance spectra.