Structural damage detection using the optimal weights of the approximating artificial neural networks

This work presents a novel neural network‐based approach to detect structural damage. The proposed approach comprises two steps. The first step, system identification, involves using neural system identification networks (NSINs) to identify the undamaged and damaged states of a structural system. The partial derivatives of the outputs with respect to the inputs of the NSIN, which identifies the system in a certain undamaged or damaged state, have a negligible variation with different system errors. This loosely defined unique property enables these partial derivatives to quantitatively indicate system damage from the model parameters. The second step, structural damage detection, involves using the neural damage detection network (NDDN) to detect the location and extent of the structural damage. The input to the NDDN is taken as the aforementioned partial derivatives of NSIN, and the output of the NDDN identifies the damage level for each member in the structure. Moreover, SDOF and MDOF examples are presented to demonstrate the feasibility of using the proposed method for damage detection of linear structures. Copyright © 2001 John Wiley & Sons, Ltd.     

Study of a semi‐active stiffness damper under various earthquake inputs

Semi‐active stiffness damper (SASD) is one of many semi‐active control systems with the capability to mitigate the dynamic response using only a small amount of external power. The system consists of a hydraulic damper connected to the bracing frame in a selected story unit. In this paper, study of a SASD in two building models of five‐stories under four benchmark earthquake records is reported. The purpose of this study is to evaluate the effectiveness of the control system against structure type and varying earthquake inputs. Various control laws are chosen to work with SASD, such as: resetting control, switching control, linear quadratic regulator (LQR) and modified LQR, and the results are compared with no control and passive control cases. Numerical results show that the use of a SASD is effective in reducing seismic responses. Control effectiveness is dependent on the type of structure and earthquake excitation. Passive control is less effective than other control cases as expected. Resetting control, switching control and LQR generally perform similarly in response reduction. While modified LQR is more efficient and robust compared with other control algorithms. Copyright © 2002 John Wiley & Sons, Ltd.     

Parameter identification for lined tunnels in a viscoplastic medium

This paper is dedicated to the identification of constitutive parameters of elasto‐viscoplastic constitutive law from measurements performed on deep underground cavities (typically tunnels). This inverse problem is solved by the minimization of a cost functional of least‐squares type. The exact gradient is computed by the direct differentiation method and the descent is done using the Levenberg–Marquardt algorithm. The method is presented for lined or unlined structures and is applied for an elastoviscoplastic constitutive law of the Perzyna class. Several identification problems are presented in one and two dimensions for different tunnel geometries. The used measurements have been obtained by a preliminary numerical simulation and perturbed with a white noise. The identified responses match the measurements. We also discuss the usage of the sensitivity analysis of the system, provided by the direct differentiation method, for the optimization of in situ monitoring. The sensitivity distribution in space and time assess the location of the measurements points as well as the time of observation needed for reliable identification. Copyright © 2002 John Wiley & Sons, Ltd.     

A neural network approach for structural identification and diagnosis of a building from seismic response data

This work presents a novel procedure for identifying the dynamic characteristics of a building and diagnosing whether the building has been damaged by earthquakes, using a back‐propagation neural network approach. The dynamic characteristics are directly evaluated from the weighting matrices of the neural network trained by observed acceleration responses and input base excitations. Whether the building is damaged under a large earthquake is assessed by comparing the modal parameters and responses for this large earthquake with those for a small earthquake that has not caused this building any damage. The feasibility of the approach is demonstrated through processing the dynamic responses of a five‐storey steel frame, subjected to different strengths of the Kobe earthquake, in shaking table tests. Copyright © 2002 John Wiley & Sons, Ltd.     

Presolar spinel grains from the Murray and Murchison carbonaceous chondrites

With a new type of ion microprobe, the NanoSIMS, we determined the oxygen isotopic compositions of small (<1μm) oxide grains in chemical separates from two CM2 carbonaceous meteorites, Murray and Murchison. Among 628 grains from Murray separate CF (mean diameter 0.15 μm) we discovered 15 presolar spinel and 3 presolar corundum grains, among 753 grains from Murray separate CG (mean diameter 0.45 μm) 9 presolar spinel grains, and among 473 grains from Murchison separate KIE (mean diameter 0.5 μm) 2 presolar spinel and 4 presolar corundum grains. The abundance of presolar spinel is highest (2.4%) in the smallest size fraction. The total abundance in the whole meteorite is at least 1 ppm, which makes spinel the third-most abundant presolar grain species after nanodiamonds (if indeed a significant fraction of them are presolar) and silicon carbide. The O-isotopic distribution of the spinel grains is very similar to that of presolar corundum, the only statistically significant difference being that there is a larger fraction of corundum grains with large ¹⁷O excesses (¹⁷O/¹⁶O > 1.5 × 10⁻³), which indicates parent stars with masses between 1.8 and 4.5 M_⊙.     

Use of microtremor measurement for assessing site effects in Northern Belgium – interpretation of the observed intensity during the MS = 5.0 June 11 1938 earthquake

To provide quantitative information on site effects in the northern part of Belgium, forty-seven H/V microtremor measurements were performed with 5 second seismometers over an area of about 15.000 km². Most of the results show a northward regular increase of the fundamental period in agreement with the augmentation of the Mesozoic and Cainozoic soft sedimentsthickness from a few meters 40 km south of Brussels to 900 m at the Netherlands-Belgium border. The measured resonance frequency values wereconsistent with theoretical computations performed at different sites onthe basis of existing information and shallow seismic experiments. At oneparticular site (Uccle) where borehole data were available, microtremor measurements using an array of four seismological stations with differentapertures allowed to obtain the low frequency part of the Rayleigh wave dispersion curve, extending the range covered by the analysis of surfacewaves artificially generated. The Vs profile derived from the surface waveinversion corroborates the 1 Hz natural frequency of the site. Comparison of these results with the macroseismic information concerning the M_S = 5.0 1938 earthquake which occurred 50 km west of Brussels,confirmed the hypothesis that the geological structure of the Brabant massifis likely to control damage distribution during such an earthquake. Comparisonbetween the intensity map of the 1938 earthquake and the resonance period ofsediments obtained by our microtremor study shows a clear relation betweenthe two parameters. During the 1938 earthquake, site effects played a prominent role due to the dimension of the source whose corner frequency wasabout 1 Hz.     

A fast intensity-hue-saturation fusion technique with spectral adjustment for IKONOS imagery

Among various image fusion methods, intensity-hue-saturation (IHS) technique is capable of quickly merging the massive volumes of data. For IKONOS imagery, IHS can yield satisfactory "spatial" enhancement but may introduce "spectral" distortion, appearing as a change in colors between compositions of resampled and fused multispectral bands. To solve this problem, a fast IHS fusion technique with spectral adjustment is presented. The experimental results demonstrate that the proposed approach can provide better performance than the original IHS method, both in processing speed and image quality.     

Estimating coefficients of volume compressibility from compression of strata and piezometric changes in a multiaquifer system in west Taiwan

This paper estimates the coefficients of volume compressibility from variation in compressible layer thickness and changes in piezometric heads by using detail ground surface surveys and a multilayer monitoring well at a selected site (Shigang) within the Choshui River alluvial fan in west Taiwan. The paper integrates various types of in situ monitoring tools, including leveling surveys, continuous global position system (GPS) stations, multilevel layer compression and groundwater pressure head-monitoring wells, to investigate the situation and progress of the subsidence problem in the region. The results from the cross-analyses of the measured data show that surface settlement caused by the compression of strata is between the depths of 60 and 210 m where the clayey stratum within 120-180 m was most pronounced and contributes up to 53% of the total compression. The results indicate that the clayey stratum is under normal consolidation. The results also reflect the fact that 20% of settlement contribution comes from the sandy stratum within 90-120 m; the elasto-plastic behavior of this sandy stratum is clear. The coefficients of volume compressibility of the clayey and sandy stratum analysed from the stratum's compression records; they were 6.38×10⁻⁸ and 5.71×10⁻⁹ m²/N, respectively. Ultimately, this parameter estimation would permit to control and predict land subsidence based on change in pressure head which are related to groundwater extraction.     

Image processing of 2D resistivity data for imaging faults

A methodology to locate automatically limits or boundaries between different geological bodies in 2D electrical tomography is proposed, using a crest line extraction process in gradient images. This method is applied on several synthetic models and on field data set acquired on three experimental sites during the European project PALEOSIS where trenches were dug. The results presented in this work are valid for electrical tomographies data collected with a Wenner-alpha array and computed with an l₁ norm (blocky inversion) as optimization method. For the synthetic cases, three geometric contexts are modelled: a vertical and a dipping fault juxtaposing two different geological formations and a step-like structure. A superficial layer can cover each geological structure. In these three situations, the method locates the synthetic faults and layer boundaries, and determines fault displacement but with several limitations. The estimated fault positions correlate exactly with the synthetic ones if a conductive (or no superficial) layer overlies the studied structure. When a resistive layer with a thickness of 6 m covers the model, faults are positioned with a maximum error of 1 m. Moreover, when a resistive and/or a thick top layer is present, the resolution significantly decreases for the fault displacement estimation (error up to 150%). The tests with the synthetic models for surveys using the Wenner-alpha array indicate that the proposed methodology is best suited to vertical and horizontal contacts. Application of the methodology to real data sets shows that a lateral resistivity contrast of 1:5–1:10 leads to exact faults location. A fault contact with a resistivity contrast of 1:0.75 and overlaid by a resistive layer with a thickness of 1 m gives an error location ranging from 1 to 3 m. Moreover, no result is obtained for a contact with very low contrasts (1:0.85) overlaid by a resistive soil. The method shows poor results when vertical gradients are greater than horizontal ones. This kind of image processing technique should be systematically used for improving the objectiveness of tomography interpretation when looking for limits between geological objects.     

MTBE in Drinking Water Production – Occurrence and Efficiency of Treatment Technologies

In Germany, the gasoline additive methyl tert‐butyl ether (MTBE) is almost constantly detected in measurable concentrations in surface waters and is not significantly removed during riverbank filtration. The removal of MTBE from water has been the focus of many studies that mostly were performed at high concentration levels and centred in understanding the mechanisms of elimination. In order to assess the performance of conventional and advanced water treatment technologies for MTBE removal in the low concentration range further studies were undertaken. Laboratory experiments included aeration, granulated activated carbon (GAC) adsorption, ozonation and advanced oxidation processes (AOP). The results show that the removal of MTBE by conventional technologies is not easily achieved. MTBE is only removed by aeration at high expense. Ozonation at neutral pH values did not prove to be effective in eliminating MTBE at all. The use of ozone/H₂O₂ (AOP) may lead to a partly elimination of MTBE. However, the ozone/H₂O₂ concentrations required for a complete removal of MTBE from natural waters is much higher than the ozone levels applied nowadays in waterworks. MTBE is only poorly adsorbed on activated carbon, thus GAC filtration is not efficient in eliminating MTBE. A comparison with real‐life data from German waterworks reveals that if MTBE is detected in the raw water it is most often found in the corresponding drinking water as well due to the poor removal efficiency of conventional treatment steps.