Publication:
Time constant anomaly detection of thermocouples using transient data
Time constant anomaly detection of thermocouples using transient data
dc.contributor.advisor | Just Agosto, Frederick A. | |
dc.contributor.author | Orengo Rodríguez, José | |
dc.contributor.college | College of Engineering | en_US |
dc.contributor.committee | Serrano, David | |
dc.contributor.committee | Jia, Yi | |
dc.contributor.department | Department of Mechanical Engineering | en_US |
dc.contributor.representative | Resto, Pedro | |
dc.date.accessioned | 2018-06-06T16:40:09Z | |
dc.date.available | 2018-06-06T16:40:09Z | |
dc.date.issued | 2006 | |
dc.description.abstract | A specific methodology to detect anomalies in temperature sensors is developed in this thesis. The strategy evaluates sensor faults or variations in the transient region only by examining the signal produced by a thermocouple. The algorithm employed used a procedure in which several time derivatives are taken and filtering techniques must be present to ensure correct anomaly detection in the presence of noise. In order to test the procedure a neural network (NN) approach that uses the data processed by the algorithm was developed. The network developed was a Radial Basis probabilistic neural network (PNN). The algorithm distinguishes anomalous sensor behavior and classifies the cause of the behavior. In almost all of the trials the estimation of the NN technique proposed was less than one percent while the greatest error was less than five percent. | en_US |
dc.description.abstract | Una metodología especifica para detectar anomalies en sensors de temperature esta desarollada en esta tesis. La estrategia es evaluar las fallas en los sensores o variación en la region transiente examinando la señal producida por un thermocouple. El algoritmo empleado utilizado en el procedimeinto se aplican derivadas y se emplea una técnica de filtrar la señal contaminada por ruido para detectar la falla o anomalia corectamente. Una red neural (NN) esta propuesta en este trabajo, para distinguir la conducta anómala en sensores de temperatura y que pueden ayudar con la clasificación de defectos se proponen. Los datos experimentales se utilizaron para entrenar la red neural propuesta. Para este caso se utilizo la estructura probabilistica Radial Basis. La literatura ofrece la evidencia substancial de técnicas o esquemas de detección y clasificación de data. La estructura de NN propuesta como sistema de detección y clasificación de daños provee un error de estimación menos del 5% de error. | en_US |
dc.description.graduationYear | 2006 | en_US |
dc.identifier.uri | https://hdl.handle.net/20.500.11801/732 | |
dc.language.iso | en | en_US |
dc.rights.holder | (c) 2006 José Orengo Rodríguez | en_US |
dc.rights.license | All rights reserved | en_US |
dc.subject | thermocouples | en_US |
dc.subject.lcsh | Thermocouples | en_US |
dc.title | Time constant anomaly detection of thermocouples using transient data | en_US |
dc.type | Thesis | en_US |
dspace.entity.type | Publication | |
thesis.degree.discipline | Mechanical Engineering | en_US |
thesis.degree.level | M.S. | en_US |