An FPGA implementation of the KNN algorithm for detecting single-event-latchup in satellite payloads

• Main Author: Zou, Pengze
• Other Authors: Chang Joseph
• Format: Thesis-Master by Coursework
• Language: English
• Published: Nanyang Technological University 2023
• Subjects: Engineering::Electrical and electronic engineering
• Online Access: https://hdl.handle.net/10356/164861
• Institution: Nanyang Technological University

This dissertation presents the implementation of the K-nearest neighbors (KNN) algorithm realized in a Field Programmable Gate Array (FPGA) for detecting the Single-Event-Latchup (SEL) phenomenon in electronic devices in satellite systems. SEL is a Single-Event-Effect (SEE) from irradiation – a high-current abnormality that causes the loss of a semiconductor device functionality and may result in permanent device damage. SEL is characterized by typical statistical features of the current waveform over a short sampling window. The KNN algorithm is applied to classify and ascertain the occurrence of an SEL by calculating the distance between features extracted from the input current profiles and the labeled features. Further, the algorithm is realized on FPGA to speed up the classification process by leveraging on its parallel structure and programmable features. The implementation of the KNN algorithm is optimized by three methods – features optimization, sample number reduction, and multi-instantiation of the KNN modules. Software simulations and hardware implementation are conducted to depict the effectiveness of the designed algorithm and the three said optimization methods. The simulation results depict that the prediction accuracy can reach 92.8% for 4000 samples. Further, the SEL detection latency can be limited to less than 800 clock cycles according to the multi-instantiation configuration.