Artificial Neural Networks based Age Estimation of Electronic Devices

Predictive maintenance is a promising approach but upgrading existing machinery is expensive and difficult, thus replacing the machines is economically not worthwhile. A simple solution to upgrade existing machinery is required. With the proposed age estimation of electronic devices a first step towards such a system is taken. A test setup for accelerated aging of switched mode power supplies as well as a software defined radio based system for data acquisition is proposed. The necessary steps to transfer the measured samples into frequency domain and to prepare the data for evaluation with an artificial neural network are discussed. Two experiments are performed with the proposed test setup and neural network architecture, yielding promising results. With the two experiments the validity of the proposed approach is shown and further research is encouraged. Finally, a number of future steps are discussed.