Fig 4 - uploaded by Enrique V. Carrera
Content may be subject to copyright.
(a) Learning capacity vs number of clusters for different geometries in SOM. (b) Sensitivity and specificity for different values of threshold and margin.
Source publication
Induction motors are widely used around the world in many industrial and commercial applications. Early detection of faults in these devices is important to avoid service disruption and increase their useful life. Thus, many non-invasive schemes have been proposed to detect failures in induction motors using machine learning techniques mainly. Many...
Contexts in source publication
Context 1
... SOM algorithm applied to abnormality detection needs values of the threshold and margin parameters. The learning capacity of SOM is presented in Fig. 4(a) for 4 different ...
Context 2
... n is a number of nodes. From these results, we can establish that n = 68 and a linear geometry produce better results. As it is shown in Fig. 4(a), linear and square geometries present the best elasticity, but the first one presents the nearest maximum n/10 criteria. Thus, it is recommendable to use a linear geometry with n/10 number of clusters. In addition, the sensitivity and specificity of the model change depending on the threshold and margin values as shown in Fig. 4(b). ...
Context 3
... it is shown in Fig. 4(a), linear and square geometries present the best elasticity, but the first one presents the nearest maximum n/10 criteria. Thus, it is recommendable to use a linear geometry with n/10 number of clusters. In addition, the sensitivity and specificity of the model change depending on the threshold and margin values as shown in Fig. 4(b). We have selected a threshold of 0.77 and a margin of 0.48 as the best values for these ...
Similar publications
For a fault detection of machine tools, fixed intervention thresholds are usually necessary. In order to provide an autonomous anomaly detection without the need for fixed limits, recurring patterns must be detected in the signal data. This paper presents an approach for online pattern recognition on NC Code based on mean shift clustering that will...
We present the idea of a self-organized division of labor in networks of forecasting models. We find that the principles of self-organizing maps provide a good starting point for building resilient machine learning systems based on our idea. The potential of the idea, benefits and challenges are discussed by means of two toy-like problems.
