Embedding empirical mode decomposition within an FPGA-based design: challenges and progress

This paper presents further advancements made in an ongoing project following a series of presentations made at the same SPIE conference in the past. Compared with traditional microprocessor-based systems, rapidly advancing field-programmable gate array (FPGA) technology offers a more powerful, efficient and flexible hardware platform. An FPGA-based design is developed to classify three types of nonlinearities (including linear, hardening and softening) of a single-degree-of-freedom (SDOF) system subjected to free vibration. This significantly advances the team's previous work on using FPGAs for wireless structural health monitoring. The classification is achieved by embedding two important algorithms - empirical mode decomposition (EMD) and backbone curve analysis. A series of systematic efforts is made to embed EMD, which involves cubic spline fitting, in an FPGA-based hardware design. Throughout the process, we take advantage of concurrent operation and strive for a trade-off between computational efficiency and resource utilization. We have started to pursue our work in the context of FPGA-based computation. In particular, handling fixed-point precision is framed under data-path optimization. Our approach for data-path optimization is necessarily manual and thus may not guarantee an optimal design. Nonetheless, our study could provide a baseline case for future work using analytical data-path optimization for this and numerous other powerful algorithms for wireless structural health monitoring.