Machine Vision Aids Earthquake Retrofit Studies
- Created on Wednesday, 01 December 2010
As recent events in Haiti and Argentina have highlighted, many urban buildings in seismically active regions worldwide are in need of reinforcement to resist earthquake damage. One challenge facing earthquake retrofit activities is finding a balance between the cost and the effectiveness of the methods to be used. As part of its efforts to verify the efficiency and accuracy of new Italian regulations for the seismic assessment and repair to existing buildings, the European Centre for Training and Research in Earthquake Engineering (EUCENTRE, Pavia, Italy) used machine vision as a no-contact measurement system.
The organization conducted tests using 1:2 scale model buildings representative of Italian construction methods in the 1950s and 1960s (reinforced concrete frame with masonry infills) mounted on a shake table that could simulate minor, moderate, and severe earthquakes with longitudinal excitation. In order to understand the structure’s response to earthquake movement, researchers needed to measure both how the building moved as well as to measure any structural deformation that occurred. Damage-related measurements of interest included floor displacements and rotations, inter-story drifts, and shear deformation of infill panels as well as curvatures of critical structural elements.
Classical Methods Have Limits
The classical methods for making such measurements utilize strain gauges, displacement transducers, and accelerometers. Strain gauges attach to structural elements such as the steel bars in the center of concrete castings and change resistance as they deform. Displacement transducers connect between two points in the structure and measure the relative displacement between those points. Coupled transducers allow determination of curvature in the deformation. Accelerometers monitor the movement of individual points within the structure; integrating the signal can provide a measure of displacement as well.
While these instruments can provide highly precise, real-time measurements of building flexure and movement, they have their drawbacks. For one, these devices must be rigidly attached to the structure in order to make measurements. In the case of strain gauges, this requires attaching the gauge to the steel bar before pouring the concrete to construct the building. For displacement transducers, this requires bolting both transducer ends to structural elements. Another drawback is that strain gauges and displacement transducers only measure local displacement along a single axis, requiring researchers to choose carefully the mounting location and orientation of these instruments. In the case of strain gauges, this placement cannot be changed later, as it is sealed in concrete.
Accelerometers allow greater positioning flexibility because they can be mounted anywhere, needing only one anchor point, and can be moved with relative ease. Unfortunately, they cannot directly measure displacement, and the signal integration needed to obtain displacement introduces errors and uncertainties into the results.
All these instrumentation methods have the additional drawback that they require wiring to bring their signals to a data acquisition system, increasing the installation effort and cost. This cost adds to the cost of the instruments themselves, with the result that researchers must often make a tradeoff between the number of measurement sites and the cost to acquire and install the instruments.
Vision Simplifies Measurement
As part of its retrofit research, EUCENTRE evaluated an advanced optical system for making non-contact measurements of building seismic response. This optical method promised several important benefits, the first of which was the ease with which the measurement sites could be prepared. Rather than installing gauges and other instruments, researchers simply attach reflective markers to key points on the structure, and make their measurements with a machine vision system.
The vision system that EUCENTRE developed consisted of as many as ten standalone image acquisition and analysis units with a common electrical trigger to synchronize data acquisition. Each unit included one or two DALSA Pantera cameras with 2352 × 1728 resolution and 120 frame/second acquisition speed, a PC with two RAID-0 300-GB hard disk drives, and an Xcelera X64 frame grabber connected into the PC’s 4x PCIExpress socket. All units connected to a network that provided remote system control and database storage capability.