Large Relative Differences Research Articles

Assessing sources of uncertainty in formaldehyde air mass factors over tropical South America: Implications for top‐down isoprene emission estimates

We use a nested‐grid version of the GEOS‐Chem chemistry transport model, constrained by isoprene emissions from the Model of Emissions of Gases and Aerosols from Nature (MEGAN), and the Lund‐Potsdam‐Jena General Ecosystem Simulator (LPJ‐GUESS) bottom‐up inventories, to evaluate the impact that surface isoprene emissions have on formaldehyde (HCHO) air‐mass factors (AMFs) and vertical column densities (VCDs) over tropical South America during 2006, as observed by the Scanning Imaging Absorption Spectrometer for Atmospheric Chartography (SCIAMACHY) and Ozone Monitoring Instrument (OMI). Although the large‐scale seasonal variability of monthly mean HCHO VCDs is typically unaffected by the choice of bottom‐up inventory, large relative differences of up to ±45% in the HCHO VCD can occur for individual regions and months, but typically most VCD differences are of order ±20%. These relative changes are comparable to those produced by other sources of uncertainty in the AMF including aerosols and surface albedo, but less than those from clouds. In a sensitivity test, we find that top‐down annual isoprene emissions inferred from SCIAMACHY and OMI HCHO vertical columns can vary by as much as ±30–50% for each instrument respectively, depending on the region studied and the a priori isoprene emissions used. Our analysis suggests that the influence of the a priori isoprene emissions on HCHO AMFs and VCDs is therefore non‐negligible and must be carefully considered when inferring top‐down isoprene emissions estimates over this, or potentially any other, region.

Inexact Block Diagonal Preconditioners to Mitigate the Effects of Relative Differences in Material Stiffnesses

Finite-element analysis of many soil-structure interaction problems involves material zones of widely differing stiffnesses. The large relative differences in material stiffnesses usually result in an ill-conditioned system for which practical preconditioners become ineffec- tive for the iterative solution of such systems. The study suggests that a block diagonal preconditioner can exploit such differences. However, the theoretical preconditioner is expensive for practical use. Less costly block diagonal preconditioners are numerically evaluated for various inexact forms of the blocks in conjunction with conjugate gradient solver. The study includes analysis of two representative soil-structure in- teraction problems and proposes two simplified block diagonal preconditioners that effectively mitigate such material ill conditionings. DOI: 10.1061/(ASCE)GM.1943-5622.0000197. © 2013 American Society of Civil Engineers. CE Database subject headings: Stiffness; Soil-structure interactions; Preconditions; Blocks; Finite element method; Material stiffness. Author keywords: Stiffness ratio; Soil-structure interaction problem; Block diagonal preconditioner; Preconditioned conjugate gradient.

Atomic force microscopy and infrared analysis of aging processes of polymer electrolyte membrane fuel cell components

In this contribution, the possibilities and limits of atomic force microscopy (AFM) for investigation of fuel cell component degradation are evaluated. In particular the adhesion force and dissipation energy of the surface measured by a material sensitive AFM technique – the HarmoniX-mode (Bruker Corp.) – have been used as a measure for the relative polytetrafluoroethylene (PTFE) content of surfaces and could be quantified by calibrating with sample of known composition. Differently operated samples with microporous layers (MPLs) of commercial gas diffusion layers (GDLs) were investigated before and after operation and were compared to artificially aged and reference samples. A larger degradation of the cathode material compared to the anode was always found. As an additional example for the potential of AFM and infrared absorption spectroscopy (FTIR-ATR) the local PTFE content of a cell with a segmented anode flow field has been investigated. The results of PTFE loss at MPL and electrode surfaces from AFM measurements and infrared spectroscopy delivered different results which were explained by the distinct information depth of both methods. The large relative differences of PTFE content of the different segments were correlated with the mechanical properties of the special design of the segmented cell.

On fatigue cycle distribution in non-stationary switching loadings with Markov chain structure

The non-stationary loadings experienced by structural members can be often modelled as a sequence of stationary load states, with different mean and variance levels, where load switches are controlled by an underlying random process (regime process). The structural integrity of structures and mechanical components under such complex loadings requires the assessment of the statistical distribution of rainflow cycles counted within single load states, as well as of transition cycles caused by load state switches. Under the assumption of a constant mean for the switching loading, the authors showed in a previous paper that a linear combination of single loading spectra would provide a fairly good estimation of the overall loading spectrum of the switching loading. However, when mean value differences among load states are large, additional fatigue cycles caused by load state transitions would be present, besides cycles counted within load states. This work presents a comprehensive approach to assess the overall loading spectrum in switching loadings with variable mean value, when modelling the underlying regime process as a stationary Markov chain. For large relative mean value differences compared to load state variances, transition cycles are found by rainflow counting the regime process. The rainflow matrix for transition cycles is then estimated from the transition probability matrix of regime process by a method available in the literature. The distribution of the ranges of transition cycles is finally estimated from the statistical distribution of the largest peak and the lowest valley within load states. An illustrative example is finally discussed to show the accuracy of the proposed method.

T-Cell Infiltration and Signaling in the Adult Dorsal Spinal Cord Is a Major Contributor to Neuropathic Pain-Like Hypersensitivity

Partial peripheral nerve injury in adult rats results in neuropathic pain-like hypersensitivity, while that in neonatal rats does not, a phenomenon also observed in humans. We therefore compared gene expression profiles in the dorsal horn of adult and neonatal rats in response to the spared nerve injury (SNI) model of peripheral neuropathic pain. The 148 differentially regulated genes in adult, but not young, rat spinal cords indicate a greater microglial and T-cell response in adult than in young animals. T-cells show a large infiltration in the adult dorsal horn but not in the neonate after SNI. T-cell-deficient Rag1-null adult mice develop less neuropathic mechanical allodynia than controls, and central expression of cytokines involved in T-cell signaling exhibits large relative differences between young and adult animals after SNI. One such cytokine, interferon-gamma (IFNgamma), is upregulated in the dorsal horn after nerve injury in the adult but not neonate, and we show that IFNgamma signaling is required for full expression of adult neuropathic hypersensitivity. These data reveal that T-cell infiltration and activation in the dorsal horn of the spinal cord following peripheral nerve injury contribute to the evolution of neuropathic pain-like hypersensitivity. The neuroimmune interaction following peripheral nerve injury has therefore a substantial adaptive immune component, which is absent or suppressed in the young CNS.

Predictable real-time software synthesis

Formal theories for real-time systems (such as timed process algebra, timed automata and timed petri nets) have gained great success in the modeling of concurrent timing behavior and in the analysis of real-time properties. However, due to the ineliminable timing differences between a model and its realization, synthesizing a software realization from a model in a predictable way is still a challenging research topic. In this article, we tackle this problem by solving a set of sub-problems. The solution is based on the theoretical results for property prediction proposed in Huang et al. (2003, Real-time property preservation in approximations of timed systems. In: Proceedings of 1st ACM and IEEE international conference on formal methods and models for codesign. IEEE Computer Society, Los Alamitos, pp 163–171) and Huang (2005, Predictability in real-time system design. PhD thesis, Eindhoven University of Technology, The Netherlands), where quantitative property relations are established between two absolute/relative “close” real-time systems. This theory basically implies that if two systems are “close”, they enjoy “similar” properties. These results cannot be directly applied in practice though, because a model and its realization typically have infinitely large absolute and relative timing differences. We show that this infinite time gap can be bridged through a sequence of carefully constructed intermediate time domains. Then the property-prediction results can be applied to any pair of adjacent time domains in the sequence. Consequently, real-time properties of the implementation can be predicted from the model. We propose two parameterized hypotheses to characterize the timing differences in the sequence and to guide a correctness-preserving design process. It is shown that these hypotheses can be incorporated in a concrete tool set. We demonstrate the feasibility of the predictable synthesis approach through the design of a railroad crossing system.

Evaluation of clinical margins via simulation of patient setup errors in prostate IMRT treatment plans

This work evaluates: (i) the size of random and systematic setup errors that can be absorbed by 5 mm clinical target volume (CTV) to planning target volume (PTV) margins in prostate intensity modulated radiation therapy (IMRT); (ii) agreement between simulation results and published margin recipes; and (iii) whether shifting contours with respect to a static dose distribution accurately predicts dose coverage due to setup errors. In 27 IMRT treatment plans created with 5 mm CTV-to-PTV margins, random setup errors with standard deviations (SDs) of 1.5, 3, 5 and 10 mm were simulated by fluence convolution. Systematic errors with identical SDs were simulated using two methods: (a) shifting the isocenter and recomputing dose (isocenter shift), and (b) shifting patient contours with respect to the static dose distribution (contour shift). Maximum tolerated setup errors were evaluated such that 90% of plans had target coverage equal to the planned PTV coverage. For coverage criteria consistent with published margin formulas, plans with 5 mm margins were found to absorb combined random and systematic SDs = 3 mm. Published recipes require margins of 8-10 mm for 3 mm SDs. For the prostate IMRT cases presented here a 5 mm margin would suffice, indicating that published recipes may be pessimistic. We found significant errors in individual plan doses given by the contour shift method. However, dose population plots (DPPs) given by the contour shift method agreed with the isocenter shift method for all structures except the nodal CTV and small bowel. For the nodal CTV, contour shift DPP differences were due to the structure moving outside the patient. Small bowel DPP errors were an artifact of large relative differences at low doses. Estimating individual plan doses by shifting contours with respect to a static dose distribution is not recommended. However, approximating DPPs is acceptable, provided care is taken with structures such as the nodal CTV which lie close to the surface.

Some Thoughts on the Reporting of Adverse Events in Phase II Cancer Clinical Trials

Some Thoughts on the Reporting of Adverse Events in Phase II Cancer Clinical Trials

Comparative assessment of statistical brain MR image segmentation algorithms and their impact on partial volume correction in PET

Magnetic resonance imaging (MRI)-guided partial volume effect correction (PVC) in brain positron emission tomography (PET) is now a well-established approach to compensate the large bias in the estimate of regional radioactivity concentration, especially for small structures. The accuracy of the algorithms developed so far is, however, largely dependent on the performance of segmentation methods partitioning MRI brain data into its main classes, namely gray matter (GM), white matter (WM), and cerebrospinal fluid (CSF). A comparative evaluation of three brain MRI segmentation algorithms using simulated and clinical brain MR data was performed, and subsequently their impact on PVC in 18F-FDG and 18F-DOPA brain PET imaging was assessed. Two algorithms, the first is bundled in the Statistical Parametric Mapping (SPM2) package while the other is the Expectation Maximization Segmentation (EMS) algorithm, incorporate a priori probability images derived from MR images of a large number of subjects. The third, here referred to as the HBSA algorithm, is a histogram-based segmentation algorithm incorporating an Expectation Maximization approach to model a four-Gaussian mixture for both global and local histograms. Simulated under different combinations of noise and intensity non-uniformity, MR brain phantoms with known true volumes for the different brain classes were generated. The algorithms' performance was checked by calculating the kappa index assessing similarities with the “ground truth” as well as multiclass type I and type II errors including misclassification rates. The impact of image segmentation algorithms on PVC was then quantified using clinical data. The segmented tissues of patients' brain MRI were given as input to the region of interest (RoI)-based geometric transfer matrix (GTM) PVC algorithm, and quantitative comparisons were made. The results of digital MRI phantom studies suggest that the use of HBSA produces the best performance for WM classification. For GM classification, it is suggested to use the EMS. Segmentation performed on clinical MRI data show quite substantial differences, especially when lesions are present. For the particular case of PVC, SPM2 and EMS algorithms show very similar results and may be used interchangeably. The use of HBSA is not recommended for PVC. The partial volume corrected activities in some regions of the brain show quite large relative differences when performing paired analysis on 2 algorithms, implying a careful choice of the segmentation algorithm for GTM-based PVC.

The use of n–alkanes and other plant–wax compounds as markers for studying the feeding and nutrition of large mammalian herbivores

The use of n–alkanes and other plant–wax compounds as markers for studying the feeding and nutrition of large mammalian herbivores

A Comparison of Hydrostatic Weighing and Air Displacement Plethysmography in Adults With Spinal Cord Injury

Clasey JL, Gater DR Jr. A comparison of hydrostatic weighing and air displacement plethysmography in adults with spinal cord injury. Objectives To compare (1) total body volume (V b) and density (D b) measurements obtained by hydrostatic weighing (HW) and air displacement plethysmography (ADP) in adults with spinal cord injury (SCI); (2) measured and predicted thoracic gas volume (V TG); and (3) differences in percentage of fat measurements using ADP-obtained D b and HW-obtained D b measures that were interchanged in a 4-compartment body composition model (4-comp %fat). Design Twenty adults with SCI underwent ADP and V TG, and HW testing. In a subgroup (n=13) of subjects, 4-comp %fat procedures were computed. Setting Research laboratories in a university setting. Participants Twenty adults with SCI below the T3 vertebrae and motor complete paraplegia. Interventions Not applicable. Main Outcome Measures Statistical analyses, including determination of group mean differences, shared variance, total error, and 95% confidence intervals. Results The 2 methods yielded small yet significantly different V b and D b. The groups’ mean V TG did not differ significantly, but the large relative differences indicated an unacceptable amount of individual error. When the 4-comp %fat measurements were compared, there was a trend toward significant differences ( P=.08). Conclusions ADP is a valid alternative method of determining the V b and D b in adults with SCI; however, the predicted V TG should be used with caution.

Can large relative mortality differences between socio-economic groups among Swedish men be explained by risk indicator-associated social mobility?

The Nordic countries, profiled as welfare states, are shown to have comparatively large relative socio-economic differences in mortality and comparatively high intergenerational mobility. The aim of this study was to analyse the role of risk indicator-associated social mobility (from childhood through to adulthood) in socio-economic mortality differences among Swedish men aged 35-50 years. We used data on risk indicators for adult mortality (risk use of alcohol, smoking, low emotional control, psychiatric diagnosis, medication for nervous problems, contact with police and child care, experience of unemployment, low body height, low education) collected at compulsory conscription for military training among Swedish men at aged 18-20, fathers' socio-economic status at subjects' ages 9-11 years, data on subjects' socio-economic status at ages 34-36 years, and follow-up data on mortality during 1986-1999 (at ages 35-50 years). Persons in manual occupations in 1985 showed an elevated relative risk (RR) of mortality compared with stable non-manual employees regardless of the social position of their father (RR 1.75 among stable manual workers, and RR 1.74 among the downwardly mobile). In multivariate analyses, taking into account the risk indicators first operating in late adolescence, the increased mortality risk among stable manual workers and also among the downwardly mobile diminished considerably (RR 1.32 and 1.39, respectively). These results suggest that a substantial part of the socio-economic differences in mortality among middle-aged men had their origin in childhood circumstances. Risk indicator-associated social mobility was found to contribute substantially to an increase in the relative difference in mortality between male manual workers and non-manual employees.

The Isotope Geochemistry and Cosmochemistry of Magnesium

Magnesium is second only to oxygen in abundance among the rock-forming elements and is an important element in the oceans and in hydrological and biological systems. Differences in the relative abundances of its three stable isotopes, 24Mg (78.99%), 25Mg (10.00%), and 26Mg (11.01%), are expected as a result of physicochemical processes because of the large relative mass differences of 4 and 8% between 25Mg and 26Mg, and 24Mg, respectively. Although isotopes of Mg have been used for many years as tracers in artificially spiked systems (in which the abundance of one isotope is enriched) (Cary et al. 1990; Dombovari et al. 2000), reliable measurements of 25Mg/24Mg and 26Mg/24Mg in natural systems have been limited historically by the 1‰ (one part per thousand) reproducibility imparted by instrumental mass fractionation effects. In order to be useful for many geochemical and cosmochemical applications the isotope ratios of Mg must be resolved to ≤ 200 parts per million (ppm). As a result, with a few exceptions (e.g., Davis et al. 1990; Goswami et al. 1994; Russell et al. 1998), many past studies of Mg isotope ratios focused on detection of non-mass dependent, so-called “anomalous” Mg isotopic effects rather than on investigations of mass-dependent fractionation. The principle outcome of this focus was the discovery of radiogenic 26Mg (26Mg*) in primitive meteorites (Gray and Compston 1974; Lee and Papanastassiou 1974). With the advent of multiple-collector inductively coupled plasma-source mass spectrometry (MC-ICPMS) it is now possible to measure 25Mg/24Mg and 26Mg/24Mg of Mg in solution with a reproducibility of 30 to 60 ppm or better (Galy et al. 2001). What is more, ultraviolet (UV) laser ablation combined with MC-ICPMS permits in situ …

Organic carbon accumulation in the South Atlantic Ocean: its modern, mid-Holocene and last glacial distribution

A compilation of 1118 surface sediment samples from the South Atlantic was used to map modern seafloor distribution of organic carbon content in this ocean basin. Using new data on Holocene sedimentation rates, we estimated the annual organic carbon accumulation in the pelagic realm (>3000 m water depth) to be approximately 1.8×10 12 g C year −1. In the sediments underlying the divergence zone in the Eastern Equatorial Atlantic (EEA), only small amounts of organic carbon accumulate in spite of the high surface water productivity observed in that area. This implies that in the Eastern Equatorial Atlantic, organic carbon accumulation is strongly reduced by efficient degradation of organic matter prior to its burial. During the Last Glacial Maximum (LGM), accumulation of organic carbon was higher than during the mid-Holocene along the continental margins of Africa and South America (Brazil) as well as in the equatorial region. In the Eastern Equatorial Atlantic in particular, large relative differences between LGM and mid-Holocene accumulation rates are found. This is probably to a great extent due to better preservation of organic matter related to changes in bottom water circulation and not just a result of strongly enhanced export productivity during the glacial period. On average, a two- to three-fold increase in organic carbon accumulation during the LGM compared to mid-Holocene conditions can be deduced from our cores. However, for the deep-sea sediments this cannot be solely attributed to a glacial productivity increase, as changes in South Atlantic deep-water circulation seem to result in better organic carbon preservation during the LGM.

EFFECTS OF DROPOUT AND ITEM NONRESPONSE IN THE FOLLOW-UP OF THE VICTIMS OF THE ENSCHEDE FIREWORK DISASTER

On May 13, 2000, a firework storage facility with 1000 tons of heavy fireworks exploded in a residential area in Enschede, The Netherlands. Two to three weeks after the explosion (T1), a health survey was performed among victims and emergency relief workers (N = 3792). Residents of the disaster area (N = 1447) and residents who were also involved in relief work (N = 120) comprised 41% of the sample; relief workers from within (N = 316) and outside of Enschede (N = 1698) comprised 53%. A small percentage of participants were defined as passers-by (N = 130), and for the remainder (N = 81), degree of involvement could not be determined on the basis of the questionnaire. Eighteen months after the explosion (T2), a second survey was performed. Approximately 75% (N = 2851) returned the postal questionnaire. More women and elderly people, and fewer immigrants participated in the second survey than in the first. In addition, more missing items were seen among immigrants. To study the effect of a possible selection bias in the prevalence of important health complaints at T2, we used multiple imputation (MI) to generate five complete datasets. Datasets were generated in the MICE procedure in S-Plus; determinants used to predict the missing outcomes at T2 were the same health complaints at T1, demographic characteristics, and physical and psychological characteristics. The datasets were analyzed and results were combined using the MIANALYZE procedure in SAS. Imputation had little effect on the reported prevalence of health complaints at T2. Among residents, both minor increases and decreases were seen: the prevalence of poor social functioning (SF-36) was 26.3% in the original dataset and 25.6% after imputation; difficulty sleeping remained unchanged (38.6%) after imputation; and anxiety (SCL-90) increased from 33.6% to 34.1% after imputation. Similar, mostly minor, differences were observed for passers-by, resident relief workers and relief workers residing in Enschede. The relief workers from outside of Enschede, a group consisting primarily of young, healthy, non-immigrant men, displayed the largest relative changes in reporting of health complaints after imputation. The prevalence of all examined health complaints increased in this group, and while the absolute changes were small (4.6% versus 5.1% after imputation for poor social functioning, 10.0% versus 10.6% for difficulty sleeping, and 3.0% versus 4.1% for hostility), relative increases ranged from 6% to nearly 37%. Despite these sometimes large relative differences in prevalence, the overall conclusions of the second survey remained unchanged after imputation: patterns of health complaints in the affected groups after imputation appeared nearly identical to those in the original, incomplete dataset. The strengths and limitations of MI as a method to correct for the selection bias observed in this study will be discussed in the presentation.

Validation of POAM III ozone: Comparisons with ozonesonde and satellite data

This paper describes the validation of ozone profiles from the Polar Ozone and Aerosol Measurement (POAM) III instrument. POAM III O3 is measured with 1‐km vertical resolution throughout most of the stratosphere and random errors of ∼5%. It is shown that sunspots do not significantly affect the POAM O3 retrievals, nor do polar stratospheric clouds, except under rarely encountered, extreme conditions of very low O3 and exceptionally high aerosol extinction. A statistical analysis is presented of comparisons between coincident measurements from POAM III and ozonesondes, the Halogen Occultation Experiment (HALOE) and the Stratospheric Aerosol and Gas Experiment (SAGE) II. On average, POAM III O3 profiles agree to within ±5% with these correlative data from 13 to 60 km. There is a suggestion that from 30 to 60 km POAM III sunrise data might be biased slightly low (<5%) relative to POAM sunset data. There is evidence that POAM III has a high bias of up to ∼0.1 ppmv from 10 to 12 km, and that this bias might stem, in part, from errors in the retrieval of aerosol extinction at 0.6 μm, the primary O3 absorption wavelength in the POAM retrievals. Below 10 km the POAM III data agree with coincident sonde measurements to better than 0.05 ppmv on average, which can correspond to large relative differences of more than +30% at 8–9 km and −100% at 5 km. We conclude that the POAM III profiles are highly accurate and adequate for quantitative scientific studies.

Surface area of respirable beryllium metal, oxide, and copper alloy aerosols and implications for assessment of exposure risk of chronic beryllium disease.

The continued occurrence of chronic beryllium disease (CBD) suggests the current occupational exposure limit of 2 microg beryllium per cubic meter of air does not adequately protect workers. This study examined the morphology and measured the particle surface area of aerodynamically size-separated powders and process-sampled particles of beryllium metal, beryllium oxide, and copper-beryllium alloy. The beryllium metal powder consisted of compact particles, whereas the beryllium oxide powder and particles were clusters of smaller primary particles. Specific surface area (SSA) results for all samples (N=30) varied by a factor of 37, from 0.56 +/- 0.07 m(2)/g (for the 0.4-0.7 microm size fraction of the process-sampled reduction furnace particles) to 20.8 +/- 0.4 m(2)/g (for the </=0.4 microm size fraction of the metal powder). Large relative differences in SSA were observed as a function of particle size for the powder of beryllium metal, from 4.0 +/- 0.01 m(2)/g (for the particle size fraction >6 microm) to 20.8 +/- 0.44 m(2)/g (for the particle size fraction </=0.4 microm). In contrast, little relative difference in SSA (<25%) was observed as a function of particle size for the beryllium oxide powder and particles collected from the screening operation. The SSA of beryllium metal powder decreases with increasing particle size, as expected for compact particles, and the SSA of the beryllium oxide powders and particles remains constant as a function of particle size, which might be expected for clustered particles. These associations illustrate how process-related factors can influence the morphology and SSA of beryllium materials. To avoid errors in predicting bioavailability of beryllium and the associated risks for CBD, the mechanisms of particle formation should be understood and the SSA of beryllium particles should be measured directly.

Spatial dimensions of precision agriculture: a spatial econometric analysis of millet yield on Sahelian coversands

AbstractThe identification of local soil variability caused by within‐field differences of macronutrients and ecological features is of paramount importance for the effectiveness of precision agriculture. We present several spatial statistical and econometric techniques to capture local differences in soil variation, ecological characteristics, and yield more effectively than the analytical techniques traditionally used in agronomy. The application of these techniques is illustrated in a case study dealing with precision agriculture in the West African Sahel. The production of millet on acid sandy soils constitutes a typical example of low soil fertility areas exhibiting small absolute but large relative differences in crop production conditions over short distances.

Spatial dimensions of precision agriculture: a spatial econometric analysis of millet yield on Sahelian coversands

The identification of local soil variability caused by within-field differences of macronutrients and ecological features is of paramount importance for the effectiveness of precision agriculture. We present several spatial statistical and econometric techniques to capture local differences in soil variation, ecological characteristics, and yield more effectively than the analytical techniques traditionally used in agronomy. The application of these techniques is illustrated in a case study dealing with precision agriculture in the West African Sahel. The production of millet on acid sandy soils constitutes a typical example of low soil fertility areas exhibiting small absolute but large relative differences in crop production conditions over short distances.

Convergence improvement of the conjugate gradient iterative method for finite element simulations

The slow convergence of the incomplete Cholesky preconditioned conjugate gradient (CG) method, applied to solve the system representing a magnetostatic finite element model, is caused by the presence of a few little eigenvalues in the spectrum of the system matrix. The corresponding eigenvectors reflect large relative differences in permeability. A significant convergence improvement is achieved by supplying vectors that span approximately the partial eigenspace formed by the slowly converging eigenmodes, to a deflated version of the CG algorithm. The numerical experiments show that even roughly determined eigenvectors already bring a significant convergence improvement. The deflating technique is embedded in the simulation procedure for a permanent magnet DC machine.