Improving an inverse model of sheet metal forming by neural network based regression

An inverse model for a sheet meta l forming process aims to determine the initial parameter levels required to form the final formed shape. This is a difficult problem that is usually approached by traditional methods such as finite element analysis. Formulating the problem as a classification problem makes it possible to use well established classification algorithms, such as decision trees. Classification is, however, generally based on a winner-takes-all approach when associating the output value with the corresponding class. On the other hand, when formulating the problem as a regression task, all the output values are combined to produce the corresponding class value. For a multi-class problem, this may result in very different associations compared with classification between the output of the model and the corresponding class. Such formulation makes it possible to use well known regression algorithms, such as neural networks. In this paper, we develop a neural network based inverse model of a sheet forming process, and compare its performance with that of a linear model. Both models are used in two modes, classification mode and a function estimation mode, to investigate the advantage of re-formulating the problem as a function estimation. This results in large improvements in the recognition rate of set-up parameters of a sheet metal forming process for both models, with a neural network model achieving much more accurate parameter recognition than a linear model.

A significant inverse relationship between concentrations of plasma hoocysteine and phospholipid docosahexaenoic acid in healthy male subjects

The aim of this study was to investigate the possibility of a relationship between plasma homocysteine (Hcy) and phospholipid FA (PUFA) in healthy Australian males. One hundred thirty six healthy male subjects aged 20–55 yr were recruited from the Melbourne metropolitan area. Each volunteer completed a semiquantitative food frequency questionnaire and gave a blood sample. Plasma Hcy concentrations were determined by an established HPLC method; the plasma phospholipid FA were determined by standard methods. Plasma Hcy concentration was significantly negatively correlated with plasma phospholipid concentration of the PUFA 20∶5n−3 (r=−0.226, P=0.009), 22∶5n−3 (r=−0.182, P=0.036), 22∶6n−3 (r=−0.286, P=0.001), total n−3 (r=−0.270, P=0.002) and the ratio n−3/n−6 PUFA (r=−0.265, P=0.002), and significantly positively correlated with 20∶4n−6 (r=0.180, P=0.037). In the partial correlation analysis, after controlling for serum vitamin B12 and folate concentration, plasma Hcy was significantly negatively correlated with the plasma phospholipid concentration of 22∶6n−3 (r=−0.205, P=0.019), total n−3 (r=−0.182, P=0.038) and the ratio n−3/n−6 PUFA (r=−0.174, P=0.048). Evidence indicates that an increased concentration of n−3 PUFA in tissues has a beneficial effect on cardiovascular health. Our findings provide further evidence that increased consumption of dietary n−3 PUFA increases the concentration of n−3 PUFA in plasma phospholipid, which is associated with a protective effect on cardiovascular diseases and lower plasma Hcy levels. The mechanism that might explain the association between plasma 22∶6n−3 and Hcy levels is not clear.

The inverse relationship between number of steps per day and obesity in a population-based sample - the AusDiab study

Background:
Physical activity (PA) is inversely associated with obesity but the effect has been difficult to quantify using questionnaires. In particular, the shape of the association has not yet been well described. Pedometers provide an opportunity to better characterize the association.

Methods:
Residents of households over the age of 25 years in randomly selected census districts in Tasmania were eligible to participate in the AusDiab cross-sectional survey conducted in 1999–2000. 1848 completed the AusDiab survey and 1126 of these (609 women and 517 men) wore a pedometer for 2-weekdays. Questionnaire data on recent PA, TV time and other factors were obtained. The outcomes were waist circumference (in cm) and body mass index (BMI) (kg/m2).

Results:
Increasing daily steps were associated with a decline in the obesity measures. The logarithmic nature of the associations was indicated by a sharper decline for those with lower daily steps. For example, an additional 2000 steps for those taking only 2000 steps per day was associated with a reduction of 2.8 (95% confidence interval (CI): 2.1,4.4) cm in waist circumference among men (for women; 2.2 (95% CI: 0.6, 3.9 cm)) with a baseline of only 2000, steps compared to a 0.7 (95% CI 0.3, 1.1) cm reduction (for women; 0.6 (95% CI: 0.2, 1.0)) for those already walking 10 000 steps daily. In the multivariable analysis, clearer associations were detected for PA and these obesity measures using daily step number rather than PA time by questionnaire.

Interpretation:
Pedometer measures of activity indicate that the inverse association between recent PA and obesity is logarithmic in form with the greatest impact for a given arithmetic step number increase seen at lower levels of baseline activity. The findings from this study need to be examined in prospective settings.

Vesselness features and the inverse compositional AAM for robust face recognition using thermal IR

Over the course of the last decade, infrared (IR) and particularly thermal IR imaging based face recognition has emerged as a promising complement to conventional, visible spectrum based approaches which continue to struggle when applied in the real world. While inherently insensitive to visible spectrum illumination changes, IR images introduce specific challenges of their own, most notably sensitivity to factors which affect facial heat emission patterns, e.g. emotional state, ambient temperature, and alcohol intake. In addition, facial expression and pose changes are more difficult to correct in IR images because they are less rich in high frequency detail which is an important cue for fitting any deformable model. In this paper we describe a novel method which addresses these major challenges. Specifically, to normalize for pose and facial expression changes we generate a synthetic frontal image of a face in a canonical, neutral facial expression from an image of the face in an arbitrary pose and facial expression. This is achieved by piecewise affine warping which follows active appearance model (AAM) fitting. This is the first publication which explores the use of an AAM on thermal IR images; we propose a pre-processing step which enhances detail in thermal images, making AAM convergence faster and more accurate. To overcome the problem of thermal IR image sensitivity to the exact pattern of facial temperature emissions we describe a representation based on reliable anatomical features. In contrast to previous approaches, our representation is not binary; rather, our method accounts for the reliability of the extracted features. This makes the proposed representation much more robust both to pose and scale changes. The effectiveness of the proposed approach is demonstrated on the largest public database of thermal IR images of faces on which it achieved 100% identification rate, significantly outperforming previously described methods

The need to revise the current methods to measure and assess static recrystallization behavior

Heterogeneous deformation developed during "static recrystallization (SRX) tests" poses serious questions about the validity of the conventional methods to measure softening fraction. The challenges to measure SRX and verify a proposed kinetic model of SRX are discussed and a least square technique is utilized to quantify the error in a proposed SRX kinetic model. This technique relies on an existing computational-experimental multi-layer formulation to account for the heterogeneity during the post interruption hot torsion deformation. The kinetics of static recrystallization for a type 304 austenitic stainless steel deformed at 900 °C and strain rate of 0.01s^-1 is characterized implementing the formulation. Minimizing the error between the measured and calculated torque-twist data, the parameters of the kinetic model and the flow behavior during the second hit are evaluated and compared with those obtained based on a conventional technique. Typical static recrystallization distributions in the test sample will be presented. It has been found that the major differences between the conventional and the presented technique results are due to the heterogeneous recrystallization in the cylindrical core of the specimen where the material is still partially recrystallized at the onset of the second hit deformation. For the investigated experimental conditions, the core is confined in the first two-thirds of the gauge radius, when the holding time is shorter than 50 s and the maximum pre-strain is about 0.5.

Computational inverse analysis of static recrystallization kinetics

© 2015 Published by Elsevier Ltd. All rights reserved. Accurate static recrystallization (SRX) models are necessary to improve the properties of austenitic steels by thermo-mechanical operations. This relies heavily on a careful and accurate analysis of "the interrupted test data" and conversion of the heterogeneous deformation data to the flow stress. A "computational-experimental inverse method" was presented and implemented here to analyze the SRX test data, which takes into account the heterogeneous softening of the post-interruption test sample. Conventional and "inverse" methods were used to identify the SRX kinetics for a model austenitic steel deformed at 1273 K (with a strain rate of 1 s-1) using the hot torsion test assess the merits of each method. Typical "static recrystallization distribution maps" in the test sample indicated that, at the onset of the second pass deformation with less than a critical holding time and a given pre-strain, a "partially-recrystallized zone" existed in the cylindrical core of the specimen near its center line. For the investigated scenario, the core was confined in the first half of the gauge radius when the holding time and the maximum pre strain were below 29 s and 0.5, respectively. For maximum pre strains smaller than 0.2, the specimen did not fully recrystallize, even at the gauge surface after holding for 50 s. Under such conditions, the conventional methods produced significant error.

Applying inverse just-noticeable-differences of velocity to position data for haptic data reduction

ust-Noticeable-Differences (JND) as a dead-band in perceptual analysis has been widely used for more than a decade. This technique has been employed for data reduction in hap tic data transmission systems by several researchers. In fact, researchers use two different JND coefficients that are JNDV and JNDF for velocity and force data respectively. For position data, they usually rely on the resolution of hap tic display device to omit data that are unperceivable to human. In this paper, pruning undesirable position data that are produced by the vibration of the device or subject and/or noise in transmission line is addressed. It is shown that using inverse JNDV for position data can prune undesirable position data. Comparison of the results of the proposed method in this paper with several well known filters and some available methods proposed by other researchers is performed. It is shown that combination of JNDV could provide lower error with desirable curve smoothness, and as little as possible computation effort and complexity. It also has been shown that this method reduces much more data rather than using forward-JNDV.