Reliability of Low-Cost Sensor-Based Structural Health Monitoring: Quantifying Uncertainty, Damage Detectability, and Decision Thresholds

Abstract

The study compares a representative low-cost MEMS accelerometer system against a conventional piezoelectric accelerometer baseline using a combined experimental and simulation design. A set of generic structural configurations and damage scenarios (stiffness loss in a cantilever beam, connection loosening in a frame, and support degradation) is analyzed under realistic operational variability, including temperature-driven drift and ambient vibration excitation uncertainty. Performance is evaluated using metrics aligned with engineering decision-making: probability of detection versus false alarm rate, time-to-detection under rolling windows, confidence intervals for identified modal parameters, and risk-weighted threshold selection. Results show that (1) sensor noise and bias do not merely degrade parameter accuracy, but can shift optimal decision thresholds and inflate false alarm rates when operational variability is unmodeled; (2) low-cost sensors can support robust detection of moderate damage when combined with careful filtering, redundancy, and baseline updating, but detection of small stiffness changes is sensitive to window length, placement, and environmental compensation; (3) decision reliability improves substantially when thresholds are derived from uncertainty budgets and false alarm constraints rather than fixed percentage-change rules; and (4) multi-sensor fusion and periodic calibration reduce uncertainty enough to make low-cost monitoring viable for operational screening and prioritization, while high-consequence decisions still require confirmatory inspection or higher-grade instrumentation. The article concludes with practical guidance for deploying low-cost structural monitoring as a decision system, emphasizing uncertainty characterization, threshold governance, and verification pathways consistent with applied engineering practice.