Optimization framework for calibration of constitutive models enhanced by neural networks

A two‐level procedure designed for the estimation of constitutive model parameters is presented in this paper. The neural network (NN) approach at the first level is applied to achieve the first approximation of parameters. This technique is used to avoid potential pitfalls related to the conventional gradient‐based optimization techniques, considered here as a corrector that improves predicted parameters. The feed‐forward NN (FFNN) and the modified Gauss–Newton algorithms are briefly presented. The proposed framework is verified for the elasto‐plastic modified Cam Clay model that can be calibrated based on standard triaxial laboratory tests, i.e. the isotropic consolidation test and the drained compression test. Two different formulations of the input data to the NN, enhanced by a dimensional reduction of experimental data using principal component analysis, are presented. The determination of model characteristics is demonstrated, first on numerical pseudo‐experiments and then on the experimental data. The efficiency of the proposed approach by means of accuracy and computational effort is also discussed. Copyright © 2008 John Wiley & Sons, Ltd.     

On the behaviour of liquefied soil

Engineering mechanics of liquefied soil is presented. Firstly, experimental methods and their results are described in order to show that liquefied soil can be approximated as a viscous fluid. The experiments proposed allow for determination of dynamic viscosity of liquefied soil, which is of the order of 10⁶ Ns/m². This value agrees with some theoretical estimates. Then, the model of viscous fluid is applied to practically important problems, in order to predict sinking of structures in liquefied subsoil, and to describe underwater flows of seabed. Enclosed examples show that the model and methods proposed lead to realistic results and, therefore, can be applied in engineering calculations.     

Diagenetic and other highly mineralized waters in the Polish Carpathians

Highly mineralized waters of different chemical types and origin occur in the flysch formations and their bedrocks in the western part of the Polish Carpathians. The marine sedimentation water of the flysch formations is not preserved, as the most mineralized and the heaviest isotopic values of flysch waters are characterized by δ¹⁸O and δ²H values in the ranges of 5–7‰ and −(20–30)‰, respectively. Their origin is related to the dehydration of clay minerals during burial diagenesis, with molecules of marine water completely removed by molecules of released bound water. They are relatively enriched in Na⁺ in respect to the marine water, supposedly due to the release of Na⁺ during the illitization of smectites and preferable incorporation of other cations from the primary brine into newly formed minerals. In some parts of younger formations, i.e. in the Badenian sediments, brines occur with isotopic composition close to SMOW and Cl⁻ contents greatly exceeding the typical marine value of about 19.6 g/L, supposedly due to ultrafiltration. Most probably, the marine water of the flysch formations was similarly enriched chemically in its initial burial stages. Final Cl⁻ contents in diagenetic waters depend on different Cl⁻ contents in the primary brines and on relationships between diagenetic and further ultrafiltration processes. In some areas, diagenetic waters migrate to the surface along fault zones and mix with young local meteoric waters becoming diluted, with the isotope composition scattering along typical mixing lines. In areas with independent CO₂ flow from great depths, they form chloride CO₂-rich waters. Common CO₂-rich waters are formed in areas without near-surface occurrences of diagenetic waters. They change from the HCO₃–Ca type for modern waters to HCO₃–Mg–Ca, HCO₃–Na–Ca and other types with elevated TDS, Mg²⁺ and/or Na⁻ contents for old waters reaching even those of glacial age. Bedrocks of the flysch are represented by Mesozoic and Paleozoic mudstones, sandstones and carbonates, and in some areas by Badenian sediments. Brines of the Mesozoic and Paleozoic bedrocks are usually significantly enriched in Ca²⁺ and Mg²⁺ in comparison with the Badenian brines. By analogy to the deepest brines in the adjacent Upper Silesian Coal Basin, they are supposed to originate from paleometeoric waters of a hot climate.     

LIBS analysis of geomaterials: Geochemical fingerprinting for the rapid analysis and discrimination of minerals

Laser-induced breakdown spectroscopy (LIBS) is a simple atomic emission spectroscopy technique capable of real-time, essentially non-destructive determination of the elemental composition of any substance (solid, liquid, or gas). LIBS, which is presently undergoing rapid research and development as a technology for geochemical analysis, has attractive potential as a field tool for rapid man-portable and/or stand-off chemical analysis. In LIBS, a pulsed laser beam is focused such that energy absorption produces a high-temperature microplasma at the sample surface resulting in the dissociation and ionization of small amounts of material, with both continuum and atomic/ionic emission generated by the plasma during cooling. A broadband spectrometer-detector is used to spectrally and temporally resolve the light from the plasma and record the intensity of elemental emission lines. Because the technique is simultaneously sensitive to all elements, a single laser shot can be used to track the spectral intensity of specific elements or record the broadband LIBS emission spectra, which are unique chemical ‘fingerprints’ of a material. In this study, a broad spectrum of geological materials was analyzed using a commercial bench-top LIBS system with broadband detection from ∼200 to 965 nm, with multiple single-shot spectra acquired. The subsequent use of statistical signal processing approaches to rapidly identify and classify samples highlights the potential of LIBS for ‘geochemical fingerprinting’ in a variety of geochemical, mineralogical, and environmental applications that would benefit from either real-time or in-field chemical analysis.     

Towards petascale simulation of atmospheric circulations with soundproof equations

This paper highlights progress with the development of a petascale implementation of general-purpose high-resolution (nonoscillatory) hydrodynamical simulation code EULAG [Prusa et al. 2008, Comput. Fluids 37, 1193]. The applications addressed are anelastic atmospheric flows in the range of scales from micro to planetary. The new modeldomain decomposition into a three dimensional processor array has been implemented to increase model performance and scalability. The performance of the new code is demonstrated on the IBM BlueGene/L and Cray XT4/XT5 supercomputers. The results show significant improvement of the model efficacy compared to the original decomposition into a two-dimensional processor array in the horizontal — a standard in meteorological models.     

Maximum Likelihood Estimation of Seismic Hazard for Sweden

The maximum magnitude, the activity rate, and the Gutenberg-Richterb parameter as earthquake hazard parameters, have been evaluated for Sweden. The maximum likelihood method permits the combination of historical and instrumental data. The catalog used consists of 1100 earthquakes in the time interval 1375–1989. The extreme part of the catalog contains only the strongest historical earthquakes, whereas the complete part is divided into several subcatalogs, each assumed complete above a specified threshold magnitude. The uncertainty in magnitude determination was taken into account. For southern Sweden, the calculations giveb-values of 1.04 (0.05) for the whole area south of 60° N and 0.98 (0.06) for a subregion of enhanced seismicity in the Lake Vänern area. For the whole area north of 60° N, theb-value is 1.35 (0.06) and for the seismicity zone along the Gulf of Bothnia 1.26 (0.06). The number of annually expected earthquakes with magnitude equal to or larger than 2.4 [M_L(UPP) or M_M(UPP)] is 1.8 for the whole southern Sweden, 1.3 for the Lake Vänern region, 3.7 for northern Sweden, and 2.4 for the region along the Gulf of Bothnia. The maximum expected regional magnitude is calculated to 4.9 (0.5) for a time span of 615 years for southern Sweden and the Lake Vänern subregion, and 4.3 (0.5) for a time span of 331 years for northern Sweden and the Gulf of Bothnia subregion. However, several historical earthquakes with magnitude above 5 in nearby areas of Norway indicate that the seismic potential may be higher.     

A model for heat and water balance estimation and its application to land use and climate variation

This paper presents a method, which has been recently developed in the Department of Agrometeorology of the Poznań Agricultural University, for estimating heat balance components based on standard meteorological data, plant development stage, and land-use conditions. Estimates of latent heat flux components made it possible to obtain values of areal evapotranspiration and surface runoff. The method was applied in four catchment areas in Central Poland under present climatic conditions as well as under new, expected climatic conditions caused by an increase of ‘greenhouse gases’ in the atmosphere. Some changes in land use were also considered. The results show significant changes in the water balance when comparing present and future conditions of precipitation and evapotranspiration.     

Geochemical features of re-deposited organic matter occurring in fluvioglacial sediments in the Racibórz region (Poland): A case study

The erosion of rocks rich in organic matter typically leads to the complete mineralization of the organic material. However, in some cases, it is re-deposited to become a part of sediments once more. This process should be considered to be a part of global carbon cycle, possibly much more significant than assumed to-date. The research presented here aims to characterize re-worked organic matter occurring in post-glacial sediments of southern part of Poland, in the Oder river valley (the Racibórz town region, Miocene, Pleistocene and Holocene age). Organic substances extracted from the sediments originated from organic matter that had resided in rocks eroded by glaciers. Sediments were sampled in two boreholes which sediments were correlated. Sediments were extracted and extracts analyzed with gas chromatography-mass spectrometry (GC-MS) to assess distributions of biomarker groups. Organic matter of selected samples was pre-concentrated and analyzed with Py/GC-MS. In the extracts several biomarker parameters of source/environment and thermal maturity were calculated. Organic substances in the investigated sediments come from variable re-deposited organic matter occurring in rocks eroded by glaciers. Three main parent types of re-deposited organic material are identified showing variable geochemical features: 1) organic matter of recent or almost recent age being the source of polar labile compounds; probably formed in situ, 2) re-deposited organic matter of the middle diagenesis showing features similar to lignites (huminite reflectance Rf ~ 0.25–0.35%) deriving from angiosperm remains, mainly monocotyledons and to the lower extend also deciduous trees, 3) re-deposited organic matter at the middle catagenesis (Rf ~ 0.65–0.75%) being the source of most of aromatic hydrocarbons and biomarkers such as steranes, hopanes of the more thermally advanced distribution type. Its geochemical properties and assumed directions of sediment transport indicate bituminous coals of Upper Silesian Coal Basin together with coaly shales as a possible source of this organic matter. Such mixed origin of organic matter caused large discrepancies in values of thermal maturity parameters depending on input from the particular sources and occurrence both geochemical biomarkers and their biochemical precursors in the same samples.     

Mercury mobility and bioavailability in soil from contaminated area

The mobility and bioavailability of mercury in the soil from the area near a plant using elemental mercury for manufacturing thermometers, areometers, glass energy switches and other articles made of technical glass has been evaluated. Mercury has been determined by sequential extraction method and with additional thermo desorption stage to determine elemental mercury. The procedure of sequential extraction involves five subsequent stages performed with the solutions of chloroform, deionized water, 0.5 M HCl, 0.2 M NaOH and aqua regia. The mean concentration of total mercury in soil was 147 ± 107 μg g⁻¹ dry mass (range 62–393), and the fractionation revealed that mercury was mainly bound to sulfides 56 ± 8% (range 45–66), one of the most biounavailable and immobile species of mercury in the environment. The fractions that brought lower contribution to the total mercury content were semi-mobile humic matter 22 ± 9% (range 11–34) and elemental mercury 17 ± 5% (range 8–23). The contributions brought by the highly mobile and toxic organomercury compounds were still lower 2.3 ± 2.7% (range 0.01–6.5). The lowest contributions brought the acid-soluble mercury 1.5 ± 1.3% (range 0.1–3.5) and water-soluble mercury 1.0 ± 0.3% (range 0.6–1.7). The surface layer of soil (0–20 cm) was characterized by higher mercury concentrations than that of the subsurface soil (60–80 cm), but the fractional contributions were comparable. The comparison of mercury fractionation results obtained in this study for highly polluted soils with results of fractionation of uncontaminated or moderately contaminated samples of soil and sediments had not shown significant statistical differences; however, in the last samples elemental mercury is usually present at very low concentrations. On the basis of obtained correlation coefficients it seems that elemental mercury soils from “Areometer” plant are contaminated; the main transformation is its vaporization to atmosphere and oxidation to divalent mercury, probably mainly mediated by organic matter, and next bound to humic matter and sulfides.