Using Bayesian methods for the estimation of uncertainty in complex statistical models

The research objectives of this thesis were to contribute to Bayesian statistical methodology by contributing to risk assessment statistical methodology, and to spatial and spatio-temporal methodology, by modelling error structures using complex hierarchical models. Specifically, I hoped to consider two applied areas, and use these applications as a springboard for developing new statistical methods as well as undertaking analyses which might give answers to particular applied questions. Thus, this thesis considers a series of models, firstly in the context of risk assessments for recycled water, and secondly in the context of water usage by crops. The research objective was to model error structures using hierarchical models in two problems, namely risk assessment analyses for wastewater, and secondly, in a four dimensional dataset, assessing differences between cropping systems over time and over three spatial dimensions. The aim was to use the simplicity and insight afforded by Bayesian networks to develop appropriate models for risk scenarios, and again to use Bayesian hierarchical models to explore the necessarily complex modelling of four dimensional agricultural data. The specific objectives of the research were to develop a method for the calculation of credible intervals for the point estimates of Bayesian networks; to develop a model structure to incorporate all the experimental uncertainty associated with various constants thereby allowing the calculation of more credible credible intervals for a risk assessment; to model a single day’s data from the agricultural dataset which satisfactorily captured the complexities of the data; to build a model for several days’ data, in order to consider how the full data might be modelled; and finally to build a model for the full four dimensional dataset and to consider the timevarying nature of the contrast of interest, having satisfactorily accounted for possible spatial and temporal autocorrelations. This work forms five papers, two of which have been published, with two submitted, and the final paper still in draft. The first two objectives were met by recasting the risk assessments as directed, acyclic graphs (DAGs). In the first case, we elicited uncertainty for the conditional probabilities needed by the Bayesian net, incorporated these into a corresponding DAG, and used Markov chain Monte Carlo (MCMC) to find credible intervals, for all the scenarios and outcomes of interest. In the second case, we incorporated the experimental data underlying the risk assessment constants into the DAG, and also treated some of that data as needing to be modelled as an ‘errors-invariables’ problem [Fuller, 1987]. This illustrated a simple method for the incorporation of experimental error into risk assessments. In considering one day of the three-dimensional agricultural data, it became clear that geostatistical models or conditional autoregressive (CAR) models over the three dimensions were not the best way to approach the data. Instead CAR models are used with neighbours only in the same depth layer. This gave flexibility to the model, allowing both the spatially structured and non-structured variances to differ at all depths. We call this model the CAR layered model. Given the experimental design, the fixed part of the model could have been modelled as a set of means by treatment and by depth, but doing so allows little insight into how the treatment effects vary with depth. Hence, a number of essentially non-parametric approaches were taken to see the effects of depth on treatment, with the model of choice incorporating an errors-in-variables approach for depth in addition to a non-parametric smooth. The statistical contribution here was the introduction of the CAR layered model, the applied contribution the analysis of moisture over depth and estimation of the contrast of interest together with its credible intervals. These models were fitted using WinBUGS [Lunn et al., 2000]. The work in the fifth paper deals with the fact that with large datasets, the use of WinBUGS becomes more problematic because of its highly correlated term by term updating. In this work, we introduce a Gibbs sampler with block updating for the CAR layered model. The Gibbs sampler was implemented by Chris Strickland using pyMCMC [Strickland, 2010]. This framework is then used to consider five days data, and we show that moisture in the soil for all the various treatments reaches levels particular to each treatment at a depth of 200 cm and thereafter stays constant, albeit with increasing variances with depth. In an analysis across three spatial dimensions and across time, there are many interactions of time and the spatial dimensions to be considered. Hence, we chose to use a daily model and to repeat the analysis at all time points, effectively creating an interaction model of time by the daily model. Such an approach allows great flexibility. However, this approach does not allow insight into the way in which the parameter of interest varies over time. Hence, a two-stage approach was also used, with estimates from the first-stage being analysed as a set of time series. We see this spatio-temporal interaction model as being a useful approach to data measured across three spatial dimensions and time, since it does not assume additivity of the random spatial or temporal effects.

Avaliação da rigidez torcional do chassi de um protótipo Baja SAE através do método de elementos finitos e de ensaio experimental

The torsional stiffness of chassis is one of the most important properties of a vehicle's structure and therefore its measurement is important. For the first time, the torsional stiffness was considered on the design of a prototype Baja SAE of the team from UNESP - FEG, Equipe Piratas do Vale Bardahl. According to the team's opinion, the increase of stiffness on this prototype, called MB1114, made possible a great improvement in its performance during competitions. In this work, the experimental evaluation of the torsional stiffness from this prototype is performed, detailing the analysis of results, as well as, the hysteresis' effect, least-squares regression and uncertainty analysis. It also shows that it is possible to measure the torsional stiffness of chassis with a low experimental uncertainty without expending many resources. The test rig costed R$ 32,50 due the reuse of materials and the use of instrumentation already available on campus. Furthermore, it is simple to produce and can be easily stocked. Those features are important for Baja and Formula SAE teams. Lastly, the measured value is used to validate the finite element analysis performed by the team during this prototype's design, because similar studies will be performed for the new cars. After investigating the finite element analysis, one result 13,5% higher than the measured value was reached. This difference is believed to be occurred due the imperfections of the finite element model, in other words, for not been possible to simulate every phenomena present on the real model

Avaliação da rigidez torcional do chassi de um protótipo Baja SAE através do método de elementos finitos e de ensaio experimental

The torsional stiffness of chassis is one of the most important properties of a vehicle's structure and therefore its measurement is important. For the first time, the torsional stiffness was considered on the design of a prototype Baja SAE of the team from UNESP - FEG, Equipe Piratas do Vale Bardahl. According to the team's opinion, the increase of stiffness on this prototype, called MB1114, made possible a great improvement in its performance during competitions. In this work, the experimental evaluation of the torsional stiffness from this prototype is performed, detailing the analysis of results, as well as, the hysteresis' effect, least-squares regression and uncertainty analysis. It also shows that it is possible to measure the torsional stiffness of chassis with a low experimental uncertainty without expending many resources. The test rig costed R$ 32,50 due the reuse of materials and the use of instrumentation already available on campus. Furthermore, it is simple to produce and can be easily stocked. Those features are important for Baja and Formula SAE teams. Lastly, the measured value is used to validate the finite element analysis performed by the team during this prototype's design, because similar studies will be performed for the new cars. After investigating the finite element analysis, one result 13,5% higher than the measured value was reached. This difference is believed to be occurred due the imperfections of the finite element model, in other words, for not been possible to simulate every phenomena present on the real model

Experimental and RANS-CFD Study of Nacelle Drag

The drag on a nacelle model was investigated experimentally and computationally to provide guidance and insight into the capabilities of RANS-based CFD. The research goal was to determine whether industry constrained CFD could participate in the aerodynamic design of nacelle bodies. Grid refinement level, turbulence model and near wall treatment settings, to predict drag to the highest accuracy, were key deliverables. Cold flow low-speed wind tunnel experiments were conducted at a Reynolds number of 6∙〖10〗^5, 293 K and a Mach number of 0.1. Total drag force was measured by a six-component force balance. Detailed wake analysis, using a seven-hole pressure probe traverse, allowed for drag decomposition via the far-field method. Drag decomposition was performed through a range of angles of attack between 0o and 45o. Both methods agreed on total drag within their respective uncertainties. Reversed flow at the measurement plane and saturation of the load cell caused discrepancies at high angles of attack. A parallel CFD study was conducted using commercial software, ICEM 15.0 and FLUENT 15.0. Simulating a similar nacelle geometry operating under inlet boundary conditions obtained through wind tunnel characterization allowed for direct comparisons with experiment. It was determined that the Realizable k-ϵ was best suited for drag prediction of this geometry. This model predicted the axial momentum loss and secondary flow in the wake, as well as the integrated surface forces, within experimental error up to 20o angle of attack. SST k-ω required additional surface grid resolution on the nacelle suction side, resulting in 15% more elements, due to separation point prediction sensitivity. It was further recommended to apply enhanced wall treatment to more accurately capture the viscous drag and separated flow structures. Overall, total drag was predicted within 5% at 0o angle of attack and 10% at 20o, each within experimental uncertainty. What is more, the form and induced drag predicted by CFD and measured by the wake traverse shared good agreement. Which indicated CFD captured the key flow features accurately despite simplification of the nacelle interior geometry.

A methodological approach to reporting corrected small field relative outputs

Purpose The goal of this work was to set out a methodology for measuring and reporting small field relative output and to assess the application of published correction factors across a population of linear accelerators. Methods and materials Measurements were made at 6 MV on five Varian iX accelerators using two PTW T60017 unshielded diodes. Relative output readings and profile measurements were made for nominal square field sizes of side 0.5 to 1.0 cm. The actual in-plane (A) and cross-plane (B) field widths were taken to be the FWHM at the 50% isodose level. An effective field size, defined as FSeff=A·B, was calculated and is presented as a field size metric. FSeffFSeff was used to linearly interpolate between published Monte Carlo (MC) calculated kQclin,Qmsrfclin,fmsr values to correct for the diode over-response in small fields. Results The relative output data reported as a function of the nominal field size were different across the accelerator population by up to nearly 10%. However, using the effective field size for reporting showed that the actual output ratios were consistent across the accelerator population to within the experimental uncertainty of ±1.0%. Correcting the measured relative output using kQclin,Qmsrfclin,fmsr at both the nominal and effective field sizes produce output factors that were not identical but differ by much less than the reported experimental and/or MC statistical uncertainties. Conclusions In general, the proposed methodology removes much of the ambiguity in reporting and interpreting small field dosimetric quantities and facilitates a clear dosimetric comparison across a population of linacs

Direct Observation of Photodissociation Products from Phenylperoxyl Radicals Isolated in the Gas Phase

Gas phase peroxyl radicals are central to our chemical understanding of combustion and atmospheric processes and are typically characterized by strong absorption in the UV (lambda(max) approximate to 240 nm). The analogous maximum absorption feature for arylperoxyl radicals is predicted to shift to the visible but has not previously been characterized nor have any photoproducts arising from this transition been identified. Here we describe the controlled synthesis and isolation in vacuo of an array of charge-substituted phenylperoxyl radicals at room temperature, including the 4-(N,N,N-trimethylammonium)methyl phenylperoxyl radical cation (4-Me3N[+]CH2-C6H4OO center dot), using linear ion-trap mass spectrometry. Photodissociation mass spectra obtained at wavelengths ranging from 310 to 500 nm reveal two major photoproduct channels corresponding to homolysis of aryl-OO and arylO-O bonds resulting in loss of O-2 and O, respectively. Combining the photodissociation yields across this spectral window produces a broad (FWHM approximate to 60 nm) but clearly resolved feature centered at lambda(max) = 403 nm (3.08 eV). The influence of the charge-tag identity and its proximity to the radical site are investigated and demonstrate no effect on the identity of the two dominant photoproduct channels. Electronic structure calculations have located the vertical (B) over tilde <- (X) over tilde transition of these substituted phenylperoxyl radicals within the experimental uncertainty and further predict the analogous transition for unsubstituted phenylperoxyl radical (C6H5OO center dot) to be 457 nm (2.71 eV), nearly 45 nm shorter than previous estimates and in good agreement with recent computational values.

A modified sulphide capacity function

The sulphide capacity as originally defined by Fincham and Richardson is a strong function of composition in pseudobinary oxide melts of interest in extractive metallurgy. From an analysis of data available in the literature, it is shown that sulphide capacity is directly proportional to the activity of the basic oxide in the melt, within the uncertainty of experimental data. A single parameter is sufficient to describe the sulphide capacity of a binary slag system under isothermal and isobaric conditions. The correlation indicates that the activity coefficient of the sulphide ion or the neutral base metal sulphide dissolved in the melt is independent of composition in pseudobinary melts within experimental uncertainty. Structural variations in the melt with composition do not seem to affect the activity coefficient of the sulphide. A modified sulphide capacity function is defined which makes the treatment more elegant and greatly simplifies data storage and retrieval. The modified function is not based on any model for the melt.

Rotational spectra of propargyl alcohol dimer: A dimer bound with three different types of hydrogen bonds

Pure rotational spectra of the propargyl alcohol dimer and its three deuterium isotopologues have been observed in the 4 to 13 GHz range using a pulsed-nozzle Fourier transform microwave spectrometer. For the parent dimer, a total of 51 transitions could be observed and fitted within experimental uncertainty. For two mono-substituted and one bi-substituted deuterium isotopologues, a total of 14, 17, and 19 transitions were observed, respectively. The observed rotational constants for the parent dimer A = 2321.8335(4) MHz, B = 1150.4774(2) MHz, and C = 1124.8898(2) MHz] are close to those of the most stable structure predicted by ab initio calculations. Spectra of the three deuterated isotopologues and Kraitchman analysis positively confirm this structure. Geometrical parameters and ``Atoms in Molecules'' analysis on the observed structure reveal that the two propargyl alcohol units in the dimer are bound by three different types of hydrogen bonds: O-H center dot center dot center dot O, O-H center dot center dot center dot pi, and C-H center dot center dot center dot pi. To the best of our knowledge, propargyl alcohol seems to be the smallest molecule forming a homodimer with three different points of contact. (C) 2014 AIP Publishing LLC.

高温高压下H2O-CH4体系PVTx性质的分子动力学模拟研究

Geological fluids are important components in the earth system. To study thephysical chemistry properties and the evolution of fluid system turns out to be one of the most challenging issues in geosciences. Besides the conventional experimental approaches and theoretical or semi-theoretical modeling, molecular level computer simulation(MLCS) emerges as an alternative tool to quantificationally study the physico-chemical properties of fluid under extreme conditions in order to find out the characteristics and interaction of geological fluids in and around earth. Based on our previous study of the intermolecular potential for pure H2O and thestrict evaluation of the competitive potential models for pure CH4 and the ab initio fitting potential surface across H2O-CH4 molecules in this study, we carried out more than two thousand molecular dynamics simulations for the PVTx properties of pure CH4 and the H2O-CH4 mixtures. Comparison of 1941 simulations with experimental PVT data for pure CH4 shows an average deviation of 0.96% and a maximum deviation of 2.82%. The comparison of the results of 519 simulations of the mixtures with the experimental measurements reveals that the PVTx properties of the H2O-CH4 mixtures generally agree with the extensive experimental data with an average deviation of 0.83% and 4% in maximum, which is equivalent to the experimental uncertainty. Moreover, the maximum deviation between the experimental data and the simulation results decreases to about 2% as temperature and pressure increase,indicating that the high accuracy of the simulation is well retained in the high temperature and pressure region. After the validation of the simulation method and the intermolecular potential models, we systematically simulated the PVTx properties of this binary system from 673 K and 0.05 GPa to 2573 K and 10 GPa. In order to integrate all the simulation results and the experimental data for the calculation of thermodynamic properties, an equation of state (EOS) is developed for the H2O-CH4 system covering 673 to 2573 K and 0.01 to 10 GPa. Isochores for compositions < 4 mol% CH4 up to 773 K and 600 MPa are also determined in this thesis.

Forensic analysis of architectural finishes using Fourier transform infrared and Raman spectroscopy, Part I: The resin bases

The ability of Raman spectroscopy and Fourier transform infrared (FT-IR) microscopy to discriminate between resins used for the manufacture of architectural finishes was examined in a study of 39 samples taken from a commercial resin library. Both Raman and FT-IR were able to discriminate between different types of resin and both split the samples into several groups (six for FT-IR, six for Raman), each of which gave similar, but not identical, spectra. In addition, three resins gave unique Raman spectra (four in FTIR). However, approximately half the library comprised samples that were sufficiently similar that they fell into a single large group, whether classified using FT-IR or Raman, although the remaining samples fell into much smaller groups. Further sub-division of the FT-IR groups was not possible because the experimental uncertainty was of similar magnitude to the within-group variation. In contrast, Raman spectroscopy was able to further discriminate between resins that fell within the same groups because the differences in the relative band intensities of the resins, although small, were larger than the experimental uncertainty.

An analysis of VUV C (IV) emission from the JET divertor giving measurements of electron temperatures

A description of the radiation emitted by impurities from within a plasma is crucial if spectral line intensities are to be used in detailed studies, such as the analysis of impurity transport. The simplest and most direct check that can be made on measurements of line intensities is to analyse their ratios with other lines from the same ion. This avoids uncertainties in determining the volume of the emitting plasma and the absolute sensitivity calibration of the spectrometer and, in some cases, the need even for accurate measurements of parameters such as electron density. Consistency is required between the measured line intensity ratios and the theoretical values. The expected consistency has not been found for radiation emitted from the JET scrape-off layer (e.g. Lawson et al 2009a JINST 4 P04013), meaning that the description of the spectral line intensities of impurity emission from the plasma edge is incomplete. In order to gain further understanding of the discrepancies, an analysis has been carried out for emission from the JET divertor plasma and this is reported in this paper. Carbon was the main low Z intrinsic impurity in JET and an analysis of spectral line intensity ratios has been made for the C (IV) radiation emitted from the JET divertor. In this case, agreement is found between the measured and theoretical ratios to a very high accuracy, namely to within the experimental uncertainty of similar to +/- 10%. This confirms that the description of the line intensities for the present observations is complete. For some elements and ionization stages, an analysis of line intensity ratios can lead to the determination of parameters such as the electron temperature of the emitting plasma region and estimates of the contribution of recombination to the electron energy level populations. This applies to C (IV) and, to show the value and possibilities of the spectral measurements, these parameters have been calculated for a database of Ohmic and additionally heated phases of a large number of pulses. The importance of dielectronic, radiative and charge-exchange recombination as well as ionization has been investigated. In addition, the development of T-e throughout two example discharges is illustrated. The presented results indicate a number of areas for further investigation.

Dissociative electron attachment to C2F5 radicals

Dissociative electron attachment to the reactive C2F5 molecular radical has been investigated with two complimentary experimental methods; a single collision beam experiment and a new flowing afterglow Langmuir probe technique. The beam results show that F- is formed close to zero electron energy in dissociative electron attachment to C2F5. The afterglow measurements also show that F- is formed in collisions between electrons and C2F5 molecules with rate constants of 3.7 × 10-9 cm3 s-1 to 4.7 × 10-9 cm3 s-1 at temperatures of 300–600 K. The rate constant increases slowly with increasing temperature, but the rise observed is smaller than the experimental uncertainty of 35%.

Forensic analysis of architectural finishes using fourier transform infrared and raman spectroscopy, part II:White paint

White household paints are commonly encountered as evidence in the forensic laboratory but they often cannot be readily distinguished by color alone so Fourier transform infrared (FT-IR) microscopy is used since it can sometimes discriminate between paints prepared with different organic resins. Here we report the first comparative study of FT-IR and Raman spectroscopy for forensic analysis of white paint. Both techniques allowed the 51 white paint samples in the study to be classified by inspection as either belonging to distinct groups or as unique samples. FT-IR gave five groups and four unique samples; Raman gave seven groups and six unique samples. The basis for this discrimination was the type of resin and/ or inorganic pigments/extenders present. Although this allowed approximately half of the white paints to be distinguished by inspection, the other half were all based on a similar resin and did not contain the distinctive modifiers/pigments and extenders that allowed the other samples to be identified. The experimental uncertainty in the relative band intensities measured using FT-IR was similar to the variation within this large group, so no further discrimination was possible. However, the variation in the Raman spectra was larger than the uncertainty, which allowed the large group to be divided into three subgroups and four distinct spectra, based on relative band intensities. The combination of increased discrimination and higher sample throughput means that the Raman method is superior to FT-IR for samples of this type. © 2005 Society for Applied Spectroscopy.

Sputtering of oxygen ice by low energy ions

Naturally occurring ices lie on both interstellar dust grains and on celestial objects, such as those in the outer Solar system. These ices are continuously subjected to irradiation by ions from the solar wind and/or cosmic rays, which modify their surfaces. As a result, new molecular species may form which can be sputtered off into space or planetary atmospheres. We determined the experimental values of sputtering yields for irradiation of oxygen ice at 10 K by singly (He+, C+, N+, O+ and Ar+) and doubly (C2 +, N2 + and O2 +) charged ions with 4 keV kinetic energy. In these laboratory experiments, oxygen ice was deposited and irradiated by ions in an ultra high vacuum chamber at low temperature to simulate the environment of space. The number of molecules removed by sputtering was observed by measurement of the ice thickness using laser interferometry. Preliminary mass spectra were taken of sputtered species and of molecules formed in the ice by temperature programmed desorption (TPD). We find that the experimental sputtering yields increase approximately linearly with the projectile ion mass (or momentum squared) for all ions studied. No difference was found between the sputtering yields for singly and doubly charged ions of the same atom within the experimental uncertainty, as expected for a process dominated by momentum transfer. The experimental sputter yields are in good agreement with values calculated using a theoretical model except in the case of oxygen ions. Preliminary studies have shown molecular oxygen as the dominant species sputtered and TPD measurements indicate ozone formation.

Time-lapse monitoring of fluid induced geophysical property variations within an unstable earthwork using P-wave refraction

A significant portion of the UK’s transportation system relies on a network of geotechnical earthworks (cuttings and embankments) that were constructed more than 100 years ago, whose stability is affected by the change in precipitation patterns experienced over the past few decades. The vulnerability of these structures requires a reliable, cost- and time-effective monitoring of their geomechanical condition. We have assessed the potential application of P-wave refraction for tracking the seasonal variations of seismic properties within an aged clay-filled railway embankment, located in southwest England. Seismic data were acquired repeatedly along the crest of the earthwork at regular time intervals, for a total period of 16 months. P-wave first-break times were picked from all available recorded traces, to obtain a set of hodocrones referenced to the same spatial locations, for various dates along the surveyed period of time. Traveltimes extracted from each acquisition were then compared to track the pattern of their temporal variability. The relevance of such variations over time was compared with the data experimental uncertainty. The multiple set of hodocrones was subsequently inverted using a tomographic approach, to retrieve a time-lapse model of VPVP for the embankment structure. To directly compare the reconstructed VPVP sections, identical initial models and spatial regularization were used for the inversion of all available data sets. A consistent temporal trend for P-wave traveltimes, and consequently for the reconstructed VPVP models, was identified. This pattern could be related to the seasonal distribution of precipitation and soil-water content measured on site.