“…This function is one of the fastest backpropagation algorithms available. This algorithm was proven to give good results in the quality control for the resistance spot welding [65], prediction of the strength of mineral admixture concrete [66] and predicting residual strength of non-linear ultrasonic evaluated damaged concrete [8]. The ANN network when trained, validated, and tested to attain a performance goal produced stochastic results.…”
An experiment on reinforced concrete beams using four-point bending test during an ultrasonic test was conducted. Three beam specimens were considered for each water/cement ratio (WC) of 40% and 60%, with three reinforcement schedules named design A (comprising two top bars and two bottom bars), design B (with two bottom bars), and design C (with one bottom bar). The concrete beam had a size of 100 mm × 100 mm × 400 mm in length with a plain reinforcement bar of 9 mm in diameter. An ultrasonic test with pitch-catch configuration was conducted at each loading with the transducers oriented in direct transmission across the beams' length with recordings of 68 datasets per beam specimen. Recordings of ultrasonic test results and strains at the top and bottom surfaces subjected to multiple step loads in the experiment were done. After the collection of the data, feed-forward backpropagation artificial neural network (ANN) was used to investigate the sensitivity of the ultrasonic parameters to the mechanical load applied. Five input parameters were examined, as follows: neutral axis (NA), fundamental harmonic amplitude (A1), second harmonic amplitude (A2), third harmonic amplitude (A3), and peak-to-peak amplitude (PPA), while the output parameter was the percentage of ultimate load. Optimum models were chosen after training, validating, and testing 60 ANN models. The optimum model was chosen on the basis of the highest Pearson's Correlation Coefficient (R) and soundness, confirming that it exhibited good behavior in agreement with theories. A classification of sensitivity was performed using simulations based on the developed optimum models. It was found that A2 and NA were sensitive to all WC and reinforcements used in the ANN simulation. In addition, the range of sensitivity of A2 and NA was inversely and directly proportional to the reinforcing bars, respectively. This study can be used as a guide in the selection of ultrasonic parameters to assess damage in concrete with low or high WC and varying reinforcement content.
“…This function is one of the fastest backpropagation algorithms available. This algorithm was proven to give good results in the quality control for the resistance spot welding [65], prediction of the strength of mineral admixture concrete [66] and predicting residual strength of non-linear ultrasonic evaluated damaged concrete [8]. The ANN network when trained, validated, and tested to attain a performance goal produced stochastic results.…”
An experiment on reinforced concrete beams using four-point bending test during an ultrasonic test was conducted. Three beam specimens were considered for each water/cement ratio (WC) of 40% and 60%, with three reinforcement schedules named design A (comprising two top bars and two bottom bars), design B (with two bottom bars), and design C (with one bottom bar). The concrete beam had a size of 100 mm × 100 mm × 400 mm in length with a plain reinforcement bar of 9 mm in diameter. An ultrasonic test with pitch-catch configuration was conducted at each loading with the transducers oriented in direct transmission across the beams' length with recordings of 68 datasets per beam specimen. Recordings of ultrasonic test results and strains at the top and bottom surfaces subjected to multiple step loads in the experiment were done. After the collection of the data, feed-forward backpropagation artificial neural network (ANN) was used to investigate the sensitivity of the ultrasonic parameters to the mechanical load applied. Five input parameters were examined, as follows: neutral axis (NA), fundamental harmonic amplitude (A1), second harmonic amplitude (A2), third harmonic amplitude (A3), and peak-to-peak amplitude (PPA), while the output parameter was the percentage of ultimate load. Optimum models were chosen after training, validating, and testing 60 ANN models. The optimum model was chosen on the basis of the highest Pearson's Correlation Coefficient (R) and soundness, confirming that it exhibited good behavior in agreement with theories. A classification of sensitivity was performed using simulations based on the developed optimum models. It was found that A2 and NA were sensitive to all WC and reinforcements used in the ANN simulation. In addition, the range of sensitivity of A2 and NA was inversely and directly proportional to the reinforcing bars, respectively. This study can be used as a guide in the selection of ultrasonic parameters to assess damage in concrete with low or high WC and varying reinforcement content.
“…Así mismo, se utilizaron las Redes Neuronales Artificiales (RNA), que son modelos matemáticos que replican de manera simplificada el procesamiento de información del cerebro (Martín et al, 2007;Velásquez et al, 2009), para discriminar y clasificar las empresas en un perfil exportador, a partir del cual se identificaron oportunidades de mejora. Para efectos de su aplicación, se utilizó el Perceptrón Multicapa que es un tipo de red neuronal artificial que se caracteriza por su facilidad de implementación.…”
ResumenSe presenta una metodología para el análisis de las condiciones competitivas en el comercio exterior de organizaciones empresariales. El análisis abarca las etapas de medición, evaluación y clasificación de las empresas, a partir de la propuesta de 16 factores clave del potencial exportador. También se evalúa la aplicación del análisis de conglomerados para identificar y caracterizar perfiles competitivos y de redes neuronales artificiales para clasificar el potencial exportador. Los resultados muestran la capacidad del análisis de conglomerados y de las redes neuronales artificiales para discriminar niveles competitivos en el potencial exportador. Su aplicación en el sector químico permitió agrupar las empresas en cuatro perfiles competitivos que asocian sus características. Las redes neuronales artificiales mostraron un 85,7% de capacidad para discriminar y clasificar las empresas según su perfil competitivo.
Palabras clave: potencial exportador; orientación exportadora; competitividad
Application of Cluster Analysis Techniques and Artificial Neural Networks for the Evaluation of the Exporting Capability of a Company AbstractA methodology for the analysis of competitive conditions in foreign trade business organizations is presented. The analysis includes the steps of measuring, evaluation and classification of enterprises considering 16 key factors in export potential. Also, the application of cluster analysis to identify and characterize profiles in the competitive potential is done, and artificial neural networks were employed to classify the export potential. The results show the ability of the cluster analysis and of the artificial neural networks to discriminate competitive levels in export potential. The application of this methodology in the chemical sector, allowed classifying companies in four competitive groups. Also, artificial neural network showed to be capable of classifying and discriminating the competitive profile of a company with a probability of 85.7%.
“…18 The nugget is formed from the solidification of the molten metal and has a cast microstructure with coarse and columnar grains. 4,19 The human operator uses the ultrasonic testing to classify the 330 RSW joints into four categories ( Figure 1) according to the effect of the weld nugget on the ultrasonic beam: 4,19 (i) good weld (acceptable quality level): 123/330; (ii) undersize weld (unacceptable quality level): 86/330; (iii) stick weld (unacceptable quality level): 13/330; (iv) no weld (unacceptable quality level): 68/330.…”
In this work, several of the most popular and state-of-the-art classification methods are compared as pattern recognition tools for classification of resistance spot welding joints. Instead of using the result of a non-destructive testing technique as input variables, classifiers are trained directly with the relevant welding parameters, i.e. welding current, welding time and the type of electrode (electrode material and treatment). The algorithms are compared in terms of accuracy and area under the receiver operating characteristic (ROC) curve metrics, using nested cross-validation. Results show that although there is not a dominant classifier for every specificity/sensitivity requirement, support vector machines using radial kernel, boosting and random forest techniques obtain the best performance overall.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.