⚡ electrical engineering

Reducing Energy Consumption in Energy Harvesting WSNs via Indirect Measurements for Structural Health Monitoring

This paper proposes an energy-efficient measurement strategy for structural health monitoring in energy harvesting wireless sensor networks that utilizes indirect environmental parameters to update predictive models, thereby reducing measurement frequency by up to 92.18% and total energy consumption by up to 58% by triggering new measurements only when prediction errors exceed a threshold and sufficient energy is available.

Original authors: Nursultan Daupayev, Paul Reber, Soeren Hirsch

Published 2026-06-25

📖 4 min read☕ Coffee break read

CC BY 4.0

Original authors: Nursultan Daupayev, Paul Reber, Soeren Hirsch

Original paper licensed under CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/). ✨ This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer

The Big Problem: The "Always-On" Watchman

Imagine you are trying to watch over a very old, delicate wooden bridge to make sure it doesn't crack or warp. You hire a security guard (a sensor) to stand there 24/7.

In traditional Structural Health Monitoring (SHM), this guard is told to check the bridge every single second, no matter what. He writes down a report every 15 seconds, even if the bridge is perfectly still and nothing has changed.

The Catch: This guard runs on a battery. Because he is constantly checking and writing, he burns through his battery very fast. If the bridge is in a remote forest where you can't easily swap batteries, the guard eventually falls asleep (runs out of power) and stops watching.

The New Idea: The "Smart" Guard

The authors of this paper propose a smarter way to do this. Instead of a guard who checks the bridge constantly, they propose a predictive guard who uses "indirect clues."

The Analogy: The Weather Detective
Imagine you want to know if a piece of wood is going to warp (bend). You don't need to touch the wood every second to see if it's bending. Instead, you can look at the weather (temperature and humidity).

If the air gets very humid, you know the wood is likely absorbing water and might swell.
If the air gets hot, you know the wood might expand.

The new system works like this:

The Prediction: The sensor acts like a weather detective. It looks at the temperature and humidity and says, "Based on the weather, the wood is probably fine right now."
The Sleep Mode: Because the sensor is confident in its prediction, it goes to sleep. It stops taking expensive, energy-hungry measurements.
The Wake-Up Call: The sensor only wakes up to take a real, direct measurement if two things happen:
- The Prediction is Wrong: The weather changes so much that the sensor's guess is no longer accurate (the "error" gets too big).
- The Battery is Full: The sensor checks its "fuel tank" (energy harvested from the sun) and only wakes up if it has enough power to do the job.

How They Tested It

The researchers built a small, sealed room (a climate chamber) to act as a mini-forest. Inside, they placed a block of wood.

They created "storms" inside the room by rapidly changing the heat and humidity.
They compared three types of guards:
1. The Constant Guard: Checks every 15 seconds (Traditional method).
2. The Threshold Guard: Only checks if the temperature hits a specific number (Older smart method).
3. The Predictive Guard: The new method described above.

The Results: Saving Energy and Data

The results were impressive, like finding a way to make a car run on 58% less fuel.

Energy Savings: The new "Predictive Guard" used up to 58% less energy than the constant guard. In some battery setups, it saved even more.
Data Reduction: The constant guard wrote down 2,786 reports. The new guard only wrote down about 218 reports. That is a 92% reduction in data!
Accuracy: Even though the new guard checked the bridge far less often, it still caught all the important "anomalies" (the times the wood started to warp). It didn't miss the big events; it just stopped wasting energy on the boring, quiet moments.

The Hardware: A Solar-Powered Watch

The sensors used in the experiment were tiny and ran on energy harvesting.

Think of them as watches that don't need a battery change because they have a tiny solar panel.
They store this solar energy in a "super-capacitor" (a very fast-charging battery) or a regular lithium battery.
The system is smart enough to switch between these power sources depending on how much sun is available.

The Bottom Line

This paper shows that for monitoring structures like bridges or buildings, we don't need to stare at them every second. By using indirect clues (like temperature and humidity) and predicting what the structure is doing, we can let the sensors sleep most of the time. They only wake up when something interesting happens or when they have plenty of solar energy.

This allows sensors to last much longer in remote places where you can't easily change batteries, making long-term monitoring of our infrastructure much more feasible.

Technical Summary: Reducing Energy Consumption in Energy Harvesting WSNs via Indirect Measurements for Structural Health Monitoring

Problem Statement

Wireless Sensor Networks (WSNs) deployed for Structural Health Monitoring (SHM) face a critical trade-off between data fidelity and energy sustainability. Traditional SHM systems rely on continuous direct measurements of structural parameters (e.g., strain and stress), which demand intensive energy and continuous operation. This is particularly problematic for long-term deployments in remote or hard-to-reach locations where battery replacement is difficult. While Energy Harvesting (EH) offers a solution to extend sensor lifespans, the variability of environmental energy sources limits the reliability of continuous data collection. Furthermore, existing adaptive strategies often introduce computational overhead (e.g., complex filtering or clustering algorithms) that consumes additional energy, or they rely on simple thresholding that may miss critical dynamic events. There is a need for a measurement strategy that balances data collection, state prediction, and energy availability without requiring excessive computational resources or continuous direct sensing.

Methodology

The authors propose an indirect measurement approach where environmental parameters (temperature and humidity) serve as observable indicators to estimate the hidden dynamic state of the monitored material (specifically a wooden block). The system operates on an energy-harvesting node equipped with a state-space model to predict the material's condition.

Core Algorithm:

State Estimation: The system uses a state-space model to estimate the hidden structural state ( $x_k$ ) based on indirect environmental observations ( $y_k$ ).
Prediction and Error Calculation: After each measurement, the sensor updates its state estimate and performs a one-step-ahead prediction. The prediction error ( $e_k$ ) is calculated against a predefined threshold ( $e_{th}$ ).
Energy-Aware Triggering: A new direct measurement is initiated only when two conditions are met simultaneously:
- The prediction error exceeds a predefined threshold ( $|e_k| > e_{th}$ ) OR the prediction error covariance trace exceeds a threshold ( $P_k > P_{th}$ ).
- Sufficient energy is available in the storage element ( $E_k > E_{min}$ ).
Energy Model: The system models energy dynamics where harvested energy ( $E_{h,k}$ ) is stored, and energy is consumed ( $E_s$ ) only when a measurement is triggered ( $\delta_k = 1$ ). The goal is to minimize the total number of measurement events ( $\sum \delta_k$ ) while maintaining estimation accuracy within a bound ( $\epsilon$ ).

Experimental Setup:
The method was validated in a controlled climate chamber using a wooden block instrumented with a strain gauge (ground truth) and environmental sensors (SHT31). The sensor nodes were powered by an AEM10941 energy harvesting module capable of managing hybrid storage (supercapacitors and LiPo batteries) and solar input. The study compared three strategies:

Periodic measurement (15-second interval).
Classical threshold-based triggering.
The proposed energy-independent adaptive algorithm.

These strategies were tested across four power configurations: Hybrid, LiPo, LiPo + Solar, and Supercapacitor-only.

Key Results

The experimental evaluation over a 1-hour period yielded the following quantitative results:

Measurement Reduction: The proposed adaptive strategy reduced the number of measurements by 92.18% compared to continuous periodic sampling. While a periodic strategy generated 2,786 decision steps, the adaptive method triggered only 218 actual measurements.
Energy Consumption:
- Compared to the periodic baseline, the proposed algorithm reduced total energy consumption by up to 58.27% (specifically in the supercapacitor configuration, dropping from ~30.63 mWh to 12.94 mWh).
- The threshold-based strategy alone reduced energy by approximately 14.7% compared to the periodic baseline.
Data Volume: The approach significantly lowered data volume, retaining only the most critical events (475 event points filtered down to 218 actual measurements) while capturing the main dynamics of the material's deformation response.
Performance Consistency: Despite significant variations in power supply configurations (e.g., voltage fluctuations in supercapacitors vs. stable voltage in LiPo), the proposed algorithm maintained stable sensing performance and successfully captured anomalous events corresponding to structural deformations.

Significance and Claims

The paper claims that the proposed indirect measurement strategy demonstrates that long-term monitoring of structures in remote areas is feasible without the excessive power consumption associated with continuous direct sensing. By leveraging environmental correlations and energy-aware triggering, the system achieves a balance between estimation accuracy and energy efficiency.

The authors emphasize that this approach allows for:

Reduced Sensor Activity: Drastically lowering the frequency of measurements and data transmission, which extends the operational lifespan of energy-harvesting nodes.
Scalability: The method avoids complex computational overhead (like distributed Kalman filters or wavelet transforms) that might be prohibitive for compact, resource-constrained sensor nodes.
Reliability: The system successfully identified structural anomalies (deformation peaks) despite a massive reduction in sampling frequency.

The paper concludes modestly, noting that while the results are promising under controlled laboratory conditions, future work is required to validate the approach under long-term, real-world deployment scenarios and to refine the reconstruction of material states from sparse observations.