Condition monitoring strategy based on an optimized selection of high-dimensional set of hybrid features to diagnose and detect multiple and combined faults in an induction motor
The development of novel condition monitoring strategies represents a critical challenge to ensure the effectiveness and reliability of complex industrial processes. Indeed, the interconnectivity of multiple variables facilitates the data exploitation under the framework of the Industry 4.0 and, sub...
| Autores: | , , , , |
|---|---|
| Tipo de recurso: | artículo |
| Fecha de publicación: | 2021 |
| País: | España |
| Institución: | Universitat Politècnica de Catalunya (UPC) |
| Repositorio: | UPCommons. Portal del coneixement obert de la UPC |
| Idioma: | inglés |
| OAI Identifier: | oai:upcommons.upc.edu:2117/361165 |
| Acceso en línea: | https://hdl.handle.net/2117/361165 https://dx.doi.org/10.1016/j.measurement.2021.109404 |
| Access Level: | acceso abierto |
| Palabra clave: | Artificial intelligence Electric motors, Induction Condition monitoring Multi-fault diagnosis Feature selection Feature reduction Intel·ligència artificial Motors elèctrics d'inducció Àrees temàtiques de la UPC::Enginyeria electrònica |
| Sumario: | The development of novel condition monitoring strategies represents a critical challenge to ensure the effectiveness and reliability of complex industrial processes. Indeed, the interconnectivity of multiple variables facilitates the data exploitation under the framework of the Industry 4.0 and, subsequently, the advanced monitoring may prevent unexpected conditions. Therefore, in this work it is proposed a condition monitoring methodology based on the estimation and optimization of a high-dimensional set of hybrid features for identifying and assessing the occurrence of multiple and combined faults that appear simultaneously in an induction motor. The contribution of this work includes the high-performance characterization of the induction motor operation by means of the high-dimensional set of hybrid features which is estimated from the analysis of vibrations and stator currents through techniques from different domains. Additionally, the validation that by using artificial intelligence and machine learning-based techniques allows the implementation of stages to optimize and reduce the high-dimensional feature space, leading to the selection and retention of the most discriminative features of the considered conditions. Finally, the automated diagnostics of multiple and combined faults, performed by a Neural Network-based classifier, highlights the effectiveness of the proposed method to overcome the occurrence of multiple faults that may appear simultaneously. The proposed method is validated under a complete set of experimental data that includes the healthy condition, three single fault conditions and four combined fault conditions, where the combinations of two and three fault conditions are studied. |
|---|