Novel optical devices and techniques for seismic activity detection and measurement

During earthquakes, the distances between structural elements of buildings change (sustaining pillars, walls, stairs, ceiling and floor, etc.). These distances change even because of so-called "settlements" of buildings (especially new, but not only), leading to quasi-static (i.e. slow) distance drifts. In order to monitor such dynamic or slow distance changes, we choose to connect a quiet short fibre (1m) incorporating a fibre-Bragg-grating (FBG) on the two sides of a construction joint between two building blocks in the University "Politehnica" of Bucharest, at the entrance in the Physics Dept. We expect width changes of this joint (either dynamical, during an earthquake, or slow, due to buildings settlemets), so the fibre stretching changes (it is pre-stretched, so it can sense joint width decreasing). The induced strain leads to a change in the pitch of the FBG which in turn leads to a change of the reflected wavelength when the FBG is illuminated along its axis with an incoherent light source, e.g. a superluminescent diode (SLD). From the study of scientific literature the dynamic range of strain in the FBG is between 1 microstrain and 5 milistrain. The task was to find a topology of the displacement sensor, for a given length of the fibre including the FBG sensing element, for which the dynamic range of the horizontal displacement is of +-20mm, falls within the dynamic range of the strain for the optical fibre. The configuration used by us (original) was to fix the fibre diagonally, under a small angle with the vertical, which was found of about 6.3 degrees. The fibre length for this configuration was of about 1 m (i.e. not too long). Moreover, since FBGs are temperature sensitive too, we used a second FBG identical with the strain-sensing one, in series with it (i.e. wavelength multiplexed), but non-stretched, as a temperature reference. So, the wavelength difference between the two reflected lines (by the two FBGs) is sensitive to the construction joint width only, and completely insensitive to temperature (the IInd originality element of our sensor). The measured resolution of the sensor (via calibration experiments) resulted better than 10 micrometers, with a linearity of +0.2% to -0.35%. In the absence of a significantly strong earthquake, we performed simulated dynamic tests, demonstrating the sensor ability to show dynamic movements. This kind of sensor is very useful for structural monitoring of buildings, either new or old. Large-scale use of such sensors could help to setting up a database on various buildings behaviour, especially during earthquakes, which in turn could lead even to civil engineering design and construction technologies changes.

The combined tri-axial sensor was tested in the Dynamic Tests Laboratory of the Institute of Solid Mechanics in Bucharest, which belongs to the Romanian Academy of Sciences. A computer-controlled shaker table VEB RFT MESEELEKTRONIK 'OTTO SCHON' type 11075 was used. The reference accelerometer used was Bruel & Kjaer type 4368. Hundreds of tests were performed. From these tests a frequency response curve resulted for the HiRes sensors, practically identical with the one obtained at UKC with a different shaker table and reference accelerometer. The resonance frequency resulted of about 580 HZ, leading to a usable bandwidth of about 200 Hz. This was a good confirmation concerning the calibration method. With the acquisition system used (16 bits resolution), a sensitivity of about 0.91 microg in the bandwidth of 200Hz was calculated. This is quiet close to the required sensitivity specified at the beginning of the project. The linearity of the HiRes response referred to the reference sensor resulted of about 3%. Similar measurements were performed in order to determine the frequency response curve of the LoRes optical sensor. In a sinusoidal movement regime (on the shaker table) the resonance frequency resulted even higher than for HiRes, in the range 700-800Hz, leading to a usable bandwidth of at least 300 Hz. The determined phase shift sensitivity of 0.5rad/g is 4 times smaller than the theoretical value of 2rad/g. This leads to a sensitivity of about 5.4mg, which is much better than the value calculated from the spectral density of the sensitivity specified for LoRes at the beginning of the project, for a bandwidth of 200Hz. The linearity of the HiRes response referred to the reference sensor resulted of about 5%. Even for sinusoidal regimes of shaking, it became clear that only signal versus time analysis is not sufficient. First of all, the shaking table movement is slightly different from a pure sinusoid in different manners at different acceleration values, so one should not expect to see such a sinusoid on any sensor. Moreover, the mechanical characteristics of the optical and reference sensors are different (resonance frequency, damping factor, etc.). Therefore, spectral analysis turned out to be very useful and it was used for laboratory tests and field tests as well, revealing a better sensitivity and signal-to-noise-ratio for the optical sensors, compared to the classical reference sensors used.

For the new optical sensors developed in our project, the problem of the calibration is special because, so far, there are no systematic methods dedicated for this. In order to minimize the errors of the parameters determined by calibration, we propose a particular yet direct method. The principle is to perform a calibration against a reference sensor. The reference sensor is a typical electromagnetic sensor (having known parameters) and the unknown sensor is the new optical sensor. Using a ground noise or other seismic signals, an unknown sensor can be calibrated against a known one by operating the two sensors side by side [Pavlis & Vernon 1994]. As a method of relative (frequency-response) calibration, this is limited to a frequency bandwidth where suitable seismic signals are present well above the instrumental noise level, and are spatially coherent between the instruments. However, when the frequency response of the sensor has been measured electrically, then its absolute gain may be determined quite accurately using this method. The two responses should be digitally equalized in terms of electric signal before the amplitudes are compared. This method can be used to get the information concerning the behaviour of the optical sensor in a quite large seismic range events. Thus, it become necessary to design two complementary approaching ways: - The behaviour at the background noise and the small seismic events can be determined by calibration of the sensors on a special vault having a good acoustic coupling with the competent rock. For properly obtained results, the vault has to be placed in quiet locations, like tunnels, caves or basements, situated in granite or crystalline schist areas. It is important to obtain a good signal to noise ratio. - The behaviour at the large seismic events can be determined by calibration using a shaker-table, which is designed to move with a prescribed motion. Both seismometers will be fixed on the shake table, and the outputs for the programmed input will be measured. The inconvenience of the shaker-table consists in that is difficult to calibrate the sensors at low level seismic signals. Thus, we remove this problem using the complementary calibration method on the vault. Usually the input has a prescribed frequency and the outputs will be compared in order to calibrate the unknown seismometer relative to the seismometer with known response. For calibration of the optical sensors, we used several steps: -Selecting the reference sensor; -Calibration and verifying the parameters of the reference sensor; -Comparative measurements in the vaults using the reference sensor and the optical sensor; -Calibration of both sensors using the shaker-table; -Computing of the recorded data and applying the results for determination of the optical sensor characteristics. Following the existing information and the work experience we propose to use as the reference sensor a short-period seismometer, model S-13 (the vertical version), manufactured by Teledyne Geotech. Our experience in maintenance of the Romanian Telemetered Seismic Network showed this instrument is quite stable and easy to operate. The calibration and the adjustment of the seismometer parameters should be done without special mechanical devices or sophisticated acquisition systems.

The problem to address was to devise sensors with a large dynamic range and according to principles of operation, which allow multiplexing. To address the wide dynamic range required, two types of sensors have been devised, a low resolution one and another with high resolution (as already explained in the 3rd result summary -"Tri-axial versions of a low resolution and a high resolution optical accelerometer, based on Michelson interferometers"). Demultiplexing was implemented in Electronics, however a multiplexing optical configuration was employed to deliver the laser output. For each high resolution sensor, a polarization diversity receiver is used, PDR (which drastically reduce the polarization-induced signal fading), which consists in 3 linear polarizers aligned at 120 degrees in respect with each-other, and 3 photodiodes PDs. The three signals are then amplified by a factor, which is the same for each PDR belonging to a given sensor, in a Gain Control block (part of the analog/digital interface). Then each output is processed in a demodulator block, of which there are 12 (also in the interface). A single frequency pigtailed laser at a wavelength of 1.5 microns is used to drive all measuring system. A Mach Zehnder in-fibre interferometer (MZI) is used as the generating interferometer, where each sensor will act as a receiving interferometer. Phase modulation in MZI will translate in intensity modulation of the signal received from the sensors. The phase variation of each sensor is composed from two parts, the sensing part due to the environmental factors and the phase carrier from the MZI.

Two optical accelerometer types were designed, built and tested, both based on fiberized Michelson interferometers (MI). This number was dictated by the high resolution required, and also by the large dynamic range, which could not be met by one sensor type only. The low resolution accelerometer mainly consists in a miniature fibre directional coupler which is mounted in a tubular housing. One fibre output has an angular endface as to avoid backreflections into the fibre. The fibre end is imaged via a ball lens onto a moving mirror, which is mounted on an elastic steel thin diaphragm. The mirror and diaphragm represent the (damped) mass-spring oscillator of the sensor. The reflected beam from the mirror is launched back into the fibre directional coupler and interferes with the backreflected light from the home-made metal-coated endface of the second output fibre. It should be noted that the fibre ends of the directional coupler should were kept as short as possible not only because of miniaturisation but also because of avoiding polarisation induced fading of the interference signal. The HR accelerometer has a (damped) mass-spring oscillator in which the spring consists of a pair of compliant cylinders mechanically stiffened by coiled optical fibres around its entire perimeter whereby the two fibre coils are the two arms of a fiberized MI, working in push-pull principle. The two optical accelerometer types were multiplicated and a tri-axial combined optical accelerometer resulted.

The described system was tested for noise at Cheia seismic site (Romania). The tests were performed using a reference electrodynamical sensor Teledyne Geotech type S-13 (USA), belonging to NIEP, and 6 tests were performed using the vertical HiRes optical sensor of the setup described above (because it is the most sensitive). The acquisition time window for all tests was 0.5 s. For each test the signal spectrum (FFT) was determined in the range 0-500Hz. The noise tests performed on the reference sensor S-13 reveal micro-seismic noise in the frequency range up to 75-100Hz, but some of them also show weak components in the range 200-250Hz. Also, the noise floor is visible. The HiRes micro-seismic spectra are somehow different compared to the ones obtained from the reference sensor S-13. For HiRes the low frequency region is usually narrower (dominant line), while for S-13 it is more non-selective. The HiRes optical sensor clearly reveals better spectral components in the range 150-200Hz, which are hardly visible with S-13. For HiRes the noise floor is nearly zero in the region 350-500Hz, while for S-13 it is visible and constant. As already known, it is difficult to precisely characterize a measuring device concerning noise. In order to see the extent in which noise affects weak signals, we imagined a distant shock test performed simultaneously with S-13 and the vertical HiRes optical sensor, using the same shock energy, but various distances. The signal spectra reveal that in the shock test at 15 m the signal delivered by S-13 begins to be affected by noise, while the HiRes spectrum is still clean. In the test at 20m the HiRes spectrum begins to broaden, indicating the presence of noise, which can be also seen in the spectrum of S-13. In the test at 30m the spectrum of HiRes broadens a little more, but the low frequency components are still clearly separated, while the spectrum of S-13 degrades dramatically. In our opinion, these tests clearly indicate the lower noise of the HiRes optical sensor, at least in the low frequency range (up to 100-150Hz), compared to the reference sensor S-13. Combined with the much higher sensitivity of the HiRes sensor, this noise behavior leads to their ability to see weaker seismic signals, i.e. to a higher signal-to-noise ratio (SNR). We point out that few commercial classical seismic sensors contain the SNR value in their datasheets.

Several possible multiplexing techniques that could be used for this project have been explored. Since the final versions of the sensors are based on interferometric transducers then either time division or spatial multiplexing can be used. Initially, we tested a basic multiplexing topology for interferometric sensors multiplexed in time. This scheme would have had the benefit of a single photodetector and demodulator. This would have required sequential reading and fast electronics and pulsed laser. Acquisition of a fast-pulsed laser and pulsed electronics could not be accommodated, therefore such a solution was abandoned. We also evaluated a transduction mechanism based on FBG. Then spatial division is most effective as an expensive WDM is not required and all the FBGs can be at the same wavelength. We implemented this design in the sensor presented within the result "Prototype of an optical sensor for structural dynamic and also quasi-static behaviour of buildings, to be used inside (based on FBGs)". But for tri-axial seismic sensing, our system evolved towards using Michelson interferometers and not FBG. Because we do not have a large number of sensors, a different configuration was adopted with shared path for the phase modulation and individual reading of each sensor. The system can take advantage of the low cost optoelectronic components for telecommunication wavelengths, 1300-1500 nm, now commonly available. The configuration used is that presented within the result "Report on the most suitable configuration of the optical seismometers along with dedicated instrumentation and software". This has allowed us to complete a triaxial sensor.

Risultati finali

Condividi questa pagina

Scarica