Self Sustaining CPS for Structural Monitoring
Abstract:
The American Society of Civil Engineers’ recent report on US infrastructure estimates that a $2.2 trillion investment would be needed to bring them up to par. An estimated 30% of the nation’s 10,000 steel railroad bridges and 600,000 highway bridges are either structurally deficient or functionally obsolete or both. Currently, most bridge inspections are visual only, which has been shown to be ineffective in identifying problems. Acoustic Emission (AE) testing is a comprehensive and effective way of detecting structural damage. However, current AE measurement systems - requiring high sampling rate and fast signal processing - are based on wired and centralized data collection which is labor intensive and expensive. There is an urgent need for a cost-effective system that is able to continuously and autonomously monitor the structural health of bridges. The goal of the project is to design, prototype and experimentally verify a self sustaining, autonomous, wireless structural monitoring system based on stress, vibration and acoustic emission sensing. At the core of our research effort is a low-power Flash FPGA-based hardware platform, which offers a radically new approach to CPS design. Critical parts of key algorithms and system components can be implemented in hardware for increased performance while the Flash-based technology enables effective duty cycling - a proven approach for low-power design in sensor networks. Energy harvesting is based on a novel vibration-based experimental technology that works with a broad range of excitation frequencies. The harvester automatically adjusts its resonant response to that of the bridge component, in order to capture more energy than a fixed resonant frequency harvester, such as the most commonly used piezoelectric transducers, can.
Structural health assessment involves quantitative analysis of AE signals to determine crack type, location, orientation and size as well as component-level and system-level health assessment fusing all available sensor data. Laboratory experiments and field tests on actual bridges are carried out to validate key aspects of the new research platform. This is the final year of the project. We successfully developed and tested a multi-board reconfigurable sensor platform (MarmotE), which supports flexible power source options, high-performance data processing, efficient low-power duty cycling, multiple processor and custom IP cores, high-speed multi-channel data acquisition and signal generation and a versatile radio frequency frontend for the 2.4GHz ISM band. The platform is well suited for developing novel wireless sensor protocols and localization methods following the software-defined radio paradigm. The unique combination of high-performance signal processing and efficient sleep modes – due to the Flash FPGA-based approach – enables permanent deployments of AE monitoring applications. We developed a low-power AE pattern recognition algorithm; specifically targeting the hardware resources and capabilities of the MarmotE platform while providing comparable results to industry-wide used approaches. The software-defined radio capabilities of the MarmotE platform enabled us to develop a novel CDMA-based communication scheme and a carrier phase-based ranging method for wireless sensor nodes. This work has been supported by a related NSF NeTS (#0964592) grant. We built a small-scale vibration energy harvester using a voice coil as the linear generator and developed an accurate model of the frequency-dependent behavior with regards to its proof mass motion and power generation responses to excitation. We tested the energy harvester prototype and the acoustic emission sensors and pattern recognition algorithm on steel and aluminum beams using an electromechanical shaker in a laboratory setup. Recent developments in the project focus on a flexible multicore execution environment and tool support. We developed the an efficient SoC message passing architecture, which is well suited for executing typical event-based wireless sensor applications developed in TinyOS. A representative sense-and-forward application was successfully ported to the new multi-core platform. The design has been verified with a cycle accurate instruction set simulator, called Avrora, which was heavily modified to support the multi-core paradigm. All the hardware and software artifacts developed in this project are available with open source licensing on the following website: https://sites.google.com/site/marmoteplatform/
Submitted by Peter Volgyesi
on
Abstract:
The American Society of Civil Engineers’ recent report on US infrastructure estimates that a $2.2 trillion investment would be needed to bring them up to par. An estimated 30% of the nation’s 10,000 steel railroad bridges and 600,000 highway bridges are either structurally deficient or functionally obsolete or both. Currently, most bridge inspections are visual only, which has been shown to be ineffective in identifying problems. Acoustic Emission (AE) testing is a comprehensive and effective way of detecting structural damage. However, current AE measurement systems - requiring high sampling rate and fast signal processing - are based on wired and centralized data collection which is labor intensive and expensive. There is an urgent need for a cost-effective system that is able to continuously and autonomously monitor the structural health of bridges. The goal of the project is to design, prototype and experimentally verify a self sustaining, autonomous, wireless structural monitoring system based on stress, vibration and acoustic emission sensing. At the core of our research effort is a low-power Flash FPGA-based hardware platform, which offers a radically new approach to CPS design. Critical parts of key algorithms and system components can be implemented in hardware for increased performance while the Flash-based technology enables effective duty cycling - a proven approach for low-power design in sensor networks. Energy harvesting is based on a novel vibration-based experimental technology that works with a broad range of excitation frequencies. The harvester automatically adjusts its resonant response to that of the bridge component, in order to capture more energy than a fixed resonant frequency harvester, such as the most commonly used piezoelectric transducers, can.
Structural health assessment involves quantitative analysis of AE signals to determine crack type, location, orientation and size as well as component-level and system-level health assessment fusing all available sensor data. Laboratory experiments and field tests on actual bridges are carried out to validate key aspects of the new research platform. This is the final year of the project. We successfully developed and tested a multi-board reconfigurable sensor platform (MarmotE), which supports flexible power source options, high-performance data processing, efficient low-power duty cycling, multiple processor and custom IP cores, high-speed multi-channel data acquisition and signal generation and a versatile radio frequency frontend for the 2.4GHz ISM band. The platform is well suited for developing novel wireless sensor protocols and localization methods following the software-defined radio paradigm. The unique combination of high-performance signal processing and efficient sleep modes – due to the Flash FPGA-based approach – enables permanent deployments of AE monitoring applications. We developed a low-power AE pattern recognition algorithm; specifically targeting the hardware resources and capabilities of the MarmotE platform while providing comparable results to industry-wide used approaches. The software-defined radio capabilities of the MarmotE platform enabled us to develop a novel CDMA-based communication scheme and a carrier phase-based ranging method for wireless sensor nodes. This work has been supported by a related NSF NeTS (#0964592) grant. We built a small-scale vibration energy harvester using a voice coil as the linear generator and developed an accurate model of the frequency-dependent behavior with regards to its proof mass motion and power generation responses to excitation. We tested the energy harvester prototype and the acoustic emission sensors and pattern recognition algorithm on steel and aluminum beams using an electromechanical shaker in a laboratory setup. Recent developments in the project focus on a flexible multicore execution environment and tool support. We developed the an efficient SoC message passing architecture, which is well suited for executing typical event-based wireless sensor applications developed in TinyOS. A representative sense-and-forward application was successfully ported to the new multi-core platform. The design has been verified with a cycle accurate instruction set simulator, called Avrora, which was heavily modified to support the multi-core paradigm. All the hardware and software artifacts developed in this project are available with open source licensing on the following website: https://sites.google.com/site/marmoteplatform/