Output details
15 - General Engineering
University of Hertfordshire
A New Neural-Network-Based Fault Diagnosis Approach for Analog Circuits by Using Kurtosis and Entropy as a Preprocessor
-Proposed new neural-network-based fault diagnosis approach for analogue circuits.
-For the first time, kurtosis and entropy are used as feature parameters, enhancing diagnosibilty, simplifying neural architecture and reducing training time.
-Achieved lowest computation: compared with 5×N×N multiplications and 5×N× (N −1) additions by existing methods; technique requires substantially lower 6×N multiplications and 3×N additions (N presents number of signal’s samples)
-Led to new collaboration with Beihang University and Hunan University, China, on electronic testing and reliability
-Approach recognised by research centres in Maryland, USA; Hong Kong and Southampton, UK.
-Led to visiting Scholar programme and joint PhD with Hunan University.