Applying VAM on structural components using a wireless sensor network (WSN) is highly favourable to achieve perpetual monitoring with a high spatial and temporal resolution at feasible cost. The individual devices of such a WSN—so-called sensor nodes—are small, low-cost, and low-power; hence, their resources in terms of computing power, memory size, communication bandwidth, and energy budget are severely limited. Existing signal processing approaches used with VAM, such as the previously discussed MI and the Hilbert Transform (HT), demand high sampling rates (typically 400 kHz and more), which leads to vast amounts of data that have to be stored and processed locally. The alternative streaming of the recorded data to a more capable device is infeasible. Therefore, neither evaluating this data on the sensor node nor transmitting it wirelessly is feasible on the given energy budget.
As a result, the first challenge is to efficiently process signals on the individual sensor nodes while maintaining the conventional VAM quality. We explored the use of undersampling to evaluate VAM-signals with much lower sampling rates compared to conventional approaches. In a reference scenario, we demonstrated that it reduces the required memory in the sensor node to just one percent and the processing time to just one quarter compared to the conventional method. At the same time, the resulting damage indicator deviates only marginally. Further, we proposed a new damage indicator based on the Short-Term Fourier-Transform and undersampling. This indicator is able to split amplitude and phase modulation and performs comparable to the well-established MI but requires dramatically less computational resources. In combination, the new techniques enable the implementation of VAM on smaller, more constrained devices.
Based on the experience with the AHOI underwater communication modem11, we developed a research prototype of a VAM sensor node, see Figure. 1(b). The prototype is able to generate ultrasonic waves in the same frequency range as in the lab Experiments and amplify, filter, and sample them. It transmits the recorded data via WiFi and also records ambient parameters such as temperature and vibration. Because of wireless communication, the prototype is still relying on bulky batteries. At this stage, the focus is on building a flexible solution to acquire experiment data in the field. The prototype is also equipped with an inertial measurement unit (IMU) and a temperature sensor. We will use these sensors on the Köhlbrand Bridge in Hamburg to investigate if ambient vibration, induced by traffic and wind, can be used as low-frequency excitation on a real bridge. Further, the influence of temperature on vibration and ultrasonic signals in a complex structure will be investigated.
The second challenge is the communication between sensor nodes to orchestrate measurements and transmit results. We investigated ultrasonic communication as an alternative to classical radio-frequency (RF) based techniques. Possible advantages are the reuse of SHM hardware for communication, avoidance of shielding effects that metals have on RF waves, and reduced energy demand.