A Common Vibe About SHM
An Inria research team dedicated to coupling physical modeling with statistics, I4S develops new methods to improve Structural Health Monitoring, i.e. computerized damage detection on bridges, buildings, or wind turbines. Researchers Laurent Mevel and Michael Döhler explain how their algorithmic technology was successfully transferred to the Danish SVS, the leading software vendor in the field of operational modal analysis.
There used to be a time when monitoring the fatigue of civil engineering structures was mostly a matter of human assessment and visual inspection. But the advent of sensor-based monitoring is fast superseding the legwork of yore. Generously peppered all over the bridges or embedded in more complex structures, huge banks of sensors now deliver real-time continuous information about their structural integrity: vibrations, thermal expansion, creep... But such data flow soon turns into deluge. Coupled with the complexity of the structures themselves, the large number of measurement points proves a hard nut to crack. It calls for computationally efficient algorithms that can nimbly overcome the high dimensionality of the parameters. This happens to be precisely the focus of I4S, a Rennes-based Inria research team that specializes into statistical inference for Structural Health Monitoring. SHM in the trade parlance. Under a royalty agreement, Inria scientists have partenered with SVS, a Danish vendor whose flagship software ARTeMIS is the leading solution for operational modal analysis throughout the world. “Our algorithmic engine and its latest updates have been embodied into this software, I4S leader Laurent Mevel explains. Not only is there a perfect match between this company's commercial activity and what our own research is all about, but in addition we both share the very same viewpoint on what the field should be.” One of the core notion in this research is that when sensors are moved from one place to another, the data resulting from these different and non-simultaneous measurement setups should not be analyzed separately anymore. Instead, data should be merged. The setups should be processed in only one step through a smarter computing.
An extra step toward software
But “having this in print in a scientific publication is just one part of the job, Mevel says. Showing that our methods really work and can fit the existing industrial environment requires an extra step. Namely: software development. That's a crucial part of technology transfer. We really work hand-in-glove with SVS in this endeavor. Their tool, ARTeMIS was modified so that we now can plug into it and test our prototypes very easily.”
Also instrumental in this effort was German PhD student Michael Döhler who spent six months in the company's headquarter in Aalborg, Denmark, testing new algorithms. “Being primarily interested in applied mathematics, this experience was obviously a great opportunity for me to tackle real industrial problems with a hands-on approach. If you come up with a tentative solution, you know right away if it is workable or not.”
Döhler was recently bestowed the First Prize of the Fondation Rennes 1 for his PhD work in the Matisse Doctoral School of Rennes 1 University. Shortly before, he was also awarded the Prize of Excellence 2011 in the context of the IRIS FP7 european project. He will now spend 10 months at Northeastern University Boston before heading back to Germany for an additional year of postdoctoral research at BAM, the German Federal Institute for Materials Research and Testing.
“We hope Michael's sojourn in Germany will give us the opportunity of a rapprochement with BAM, Mevel remarks. The Germans share our views on the field. They have even tested some of our algorithms. And they have done a lot of work on wind turbines, which is also the current focus of our research.” With thousands of windmills mushrooming all over Europe, this new green energy industry will soon need an efficient SHM system to keep an eye on its vastly scattered power units. “But a wind turbine is way more complex than a bridge. Monitoring such structures remains very challenging.”