Performance Evaluation of Structural Health Monitoring Anomaly Data Processing Algorithms in Resource-Constrained...

Area of Science:

• Engineering
• Computer Science
• Data Science

Background:

• Edge computing is crucial for structural health monitoring (SHM) in transportation infrastructure, enabling local data processing and low-latency decision support.
• SHM systems face challenges with data anomalies from sensor faults, transmission errors, or irregular structural behavior.

Purpose of the Study:

• To evaluate the deployment performance of a SHM anomaly data-processing workflow on resource-constrained edge devices.
• To analyze the trade-offs of various algorithms for data cleaning, response separation, and anomaly detection on edge platforms.

Main Methods:

• The workflow incorporates data cleaning (cubic spline interpolation, Laida criterion, wavelet denoising), response separation (time windows, 3σ criterion, wavelet packet decomposition), and anomaly detection (ARIMAX, SVM, RNN).
• Performance evaluation on simulation and field data (bridge displacement, highway pavement strain) using metrics like runtime, memory usage, CPU usage, and throughput.

Main Results:

• The study quantifies the performance metrics of the SHM anomaly detection workflow on edge devices.
• Identifies performance trade-offs among different algorithms under typical SHM resource constraints.

Conclusions:

• The analysis provides insights into deploying SHM anomaly detection workflows on edge devices.
• Understanding these performance characteristics is vital for optimizing SHM systems in resource-limited environments.