Prognostics of aluminum electrolytic capacitors using artificial neural network approach

Abstract

In this work, an effort is being made to monitor the condition of in-circuit aluminum electrolytic capacitor using artificial neural network (ANN). Recent industrial surveys on the reliability of power electronic systems shows that most of faults occur due to the wear out of aluminum electrolytic capacitors and thermal stress is the major cause for its parametric degradation. The condition of target capacitors can be estimated by monitoring variation in equivalent series resistance (ESR) from the initial pristine state value. ANN is used to estimate ESR of pristine and weak target capacitors at the test conditions. The data set for training and testing of proposed back-propagation trained artificial neural network are experimentally obtained from the developed test bed. Using the test bed, target capacitors are subjected to different operating frequency and temperature in the output section of DC/DC buck converter circuit to determine the effect of variation in electrical and thermal stress on ESR value. After off-line training, the proposed ANN is implemented using National Instruments LabVIEW software. A low cost microcontroller is programmed for real time data acquisition of target capacitors and the serial transmission of acquired dataset to the LabVIEW software installed at host computer. The performance of the proposed method is evaluated in real time by comparing the resulting ESR with the experimental values of in-circuit target capacitors. The proposed ANN, once trained properly, can be used for different circuits and in different operating conditions because of its generalization capability.