SHiMmer: An Active Sensing Platform for Wireless
Structural Health Monitoring
Overview:
SHiMmer is a wireless platform for active sensing that combines localized processing with energy harvesting to provide long-lived structural health monitoring. The life-cycle of the node is extended to 20 years by the use of super-capacitors for energy storage. During this period the node is expected to work completely maintenance-free. The node is capable of harvesting up to 780J per day. This makes it completely self-sufficient while employed in real structural health monitoring applications. Unlike other sensor networks that periodically monitor a structure and route information to a base station, our device acquires the data and processes it locally after being radio-triggered by an external agent. The localized processing allows us to avoid issues due to network congestion. Our experiments show that its 32-bits computational core can run at 100MIPS for 15 minutes daily. The figure on the right below shows how the sensors are placed.
Hardware Design:
To
coordinate all these activities, the node has been provided with a low power
microcontroller which consumes less than 1mA in active mode and only 5uA in
sleep mode. The microcontroller is interfaced with a radio transceiver working
in the 433MHz band, a 32-bits DSP and an EEPROM
storing the code for the DSP. The microcontroller also controls a network of
CMOS switches used to selectively disconnect the different parts of the node
from the power source. A passive radio-triggering circuit is used to generate
an interrupt to wake it up when a proper signal is sent by the UAV. As a
consequence, the microcontroller is the only component drawing current from the
power source in sleep mode. SHiMmer is equipped with
a an energy harvesting circuit which collects energy by means of solar cells
and stores it in super-capacitors. No batteries are included in the node
because of their lower durability and faster performance degradation. The solar
cells integrated in the node sum up to an area of 100cm2 and can produce up to
360mW in sunny conditions. The super-capacitors are commercial devices with a
working voltage is 2.5V and a low ESR, which results in a low leakage current.
The node is provided with a total capacity of
Software Design:
The two main components of the code are the microcontroller code and the DSP code. Operating system primitives such as interrupt service routines, radio communication, and power management are handled by the microcontroller, while the DSP handles the high-level aspects of structural health monitoring, such as actuating wave generation, response sampling and analysis. The core of the microcontroller code can be thought of as a state machine where each state represents what operations should be performed and the transitions represent task completions or interrupt firing. The basic use case of the system is shown in the digram below An external agent (UAV) sends the sensor a signal, which wakes up the microcontroller. As a result, the microcontroller powers on and configures the radio transceiver, and waits for a command. A typical command will be to actuate a wave and sample it. This requires the microcontroller to power up the rest of the system, namely the DSP, its external EEPROM and SRAM, and the signal conditioning stages. Once this is complete, the microcontroller enables the DSP which begins actuating, sampling, and processing the necessary data. The details of the DSP operation are shown in the following section. When the data is finished processing, the result is transmitted to the microcontroller. Then, the microcontroller shuts down the power supplies of all other parts of the node except for the transceiver. Our DSP needs to provide computation for wave generation, sampling, storing and processing. The intricacy in the interaction between the different phases poses a challenge by itself when sampling at such a high frequency. For example, the samples collected from the ADC must be stored into the external RAM at least as fast as the sampling rate or data will be dropped. The first DSP task is the wave generation. Wave generation is done via a lookup table (LUT) of discrete points. The TMS320 DSP has a built-in LUT for sine wave generation in its internal ROM. Custom waveforms can be generated by loading a specific LUT in the external EEPROM. The generation of the wave must be done at 1 MHz and the sampling frequency is set equal to 10MSPS. The timeline of the node operation is show below. The figure on the right below outlines the algorithm used in the DSP for data processing.
Evaluation:
We evaluate our system on several metrics, concerning both hardware and software perfomances. Out initial tests have been performed on an aluminum plate with two PZTs attached to simulate the monitored structure and the grid of PZTs placed over it. The first PZT served as an actuator and the second one as a sensor. The tests aimed to characterize the node in all the phases of its activity. To do that, we have made the node work following the real-application flow. We have collected data about power consumption and time delays associated to the different activities. The figure below on the left shows the timing of our associated test, with the bulk of the time being spent on data processing. The most critical task in terms of power consumption and peak currents is the actuation. During this phase, the energy circuit has to deliver up to 900mA at the highest actuation frequencies. The current drawn by the actuation circuit at different actuation frequencies is shown in the figure at the far bottom. The input conditioning circuit has been characterized in terms of SFDR by applying a sinewave produced by a function generator. This allows to analyze the spurious content of the signal output by the circuit. The PZT output is a widespectrum signal by itself. Using it as an input signal would not allow to distinguish the spurious content introduced by the amplification stage. During the measurements, the stage was powered by SHiMmer's energy circuit.
People:
UCSD-CSE - Tajana Rosing, Daniele Musiani, Kaisen Lin
UCSD-SERF - David Mascarenas, Eric Flynn
LANL -