Improving the laser cutting process design by machine learning techniques

In the field of manufacturing engineering, process designers conduct numerical simulation experiments to observe the impact of varying input parameters on certain outputs of the production process. The disadvantage of these simulations is that they are very time consuming and their results do not help to fully understand the underlying process. For instance, a common problem in planning processes is the choice of an appropriate machine parameter set that results in desirable process outputs. One way to overcome this problem is to use data mining techniques that extract previously unknown but valuable knowledge from simulation results. This paper presents a hybrid machine learning approach for applying clustering and classification techniques in a laser cutting planning process. In a first step, a clustering algorithm is used to divide large parts of the simulation data into groups of similar performance values and select those groups that are of major interest (e.g. high cut quality results). Next, classification trees are used to identify regions in the multidimensional parameter space that are related to the found groups. The evaluation shows that the models accurately identify multidimensional relationships between the input parameters and the output values of the process. In addition to that, a combination of appropriate visualization techniques for clustering with interpretable classification trees allows designers to gain valuable insights into the laser cutting process with the aim of optimizing it through visual exploration.     

Porous polycrystalline silicon: a new material for MEMS

A new technique for the fabrication of thin patterned layers of porous polycrystalline silicon (polysilicon) and surface micromachined structures is presented. First, a multilayer structure of polysilicon between two layers of low-stress silicon nitride is prepared on a wafer of silicon. Electrochemical anodization with an external cathode takes place in an RF solution. A window in the outer nitride layer provides contact between the polysilicon and the HF solution; the polysilicon layer contacts the substrate through openings in the lower silicon nitride layer (remote from the upper windows). Porous polysilicon growth in the lateral direction is found at rates as high as 15 μm min^-1 in 12M (25%, wgt) HF to be controlled by surface-reaction kinetics. A change in morphology occurs when either the anodic potential is raised or the HF concentration is decreased, causing the polysilicon to be electropolished. The etch front advances proportionally to the square root of time as expected for a mass-transport-controlled process. Similar behavior is observed in HF anodic reactions of single-crystal silicon. Dissolution of the polysilicon layer is confirmed using profilometry and scanning electron microscopy. Enclosed cavities (chambers surrounded by porous plugs) are formed by alternating between pore formation and uniform dissolution. Porous polysilicon also forms over a broad-area layer of polycrystalline silicon that has been deposited without overcoating the silicon wafer with a thin film of silicon nitride. The resulting porous layer may be useful for gas-absorption purposes in ultrasonic sensors     

Analogue placement using guided enumeration

In this paper a new efficient approach to the placement problem of analogue circuits is presented. In the design of analogue circuits, strong interaction between humans and automatic tools is crucial to obtain good results. Hence there is a strong need for fast algorithms to support an interactive and incremental design style. In the design process, usually many geometric constraints such as symmetry or matching requirements are introduced by the designer to obtain a placement that fulfils all electrical requirements. Most state-of-the-art analogue placement tools employ simulated annealing algorithms and map geometric constraints into the cost function of the annealing algorithm; thus they have to search the whole design space. In our approach an initial global placement is computed establishing neighbourhood relations between all modules. Afterwards the design space is reduced to placements that fulfil these neighbourhood relations and user-specified geometrical constraints. the reduced design space is fully enumerated to find the optimal solution. the implemented tool achieves typical CPU times of a few seconds for designs of about 40 modules, making it an excellent choice for an interactive design process. Two examples are presented: (i) a high-speed CMOS comparator lay-out is compared with lay-outs published in other papers and (ii) a BiCMOS bandpass amplifier lay-out that has to be embedded into a mixed-signal chip is compared with the manual lay-out of a fabricated chip. We achieved the smallest lay-out size published for the comparator while fulfilling all required constraints. For the amplifier the synthesized lay-out is very similar to the manual one and about the same size.     

Functional properties of a new line of immortalized human endothelial cells

An immortalized human endothelial cell line was obtained by transfecting umbilical vein endothelial cells in primary culture with plasmid pMK16 containing SV40 replicated origin defective gene. The essential functional properties demonstrated in these immortalized human endothelial cells also retaining the classical phenotypical characteristics of endothelial cells in primary culture are: (1) endothelin-1 secretion; (2) capacity to convert big endothelin-1 into endothelin-1; (3) the capacity to secrete IL1 beta and IL6 interleukins both spontaneously and after lipopolysaccharide (LPS) stimulation; (4) arginine transfer from the extracellular to the intracellular medium. Such stable cell line could facilitate studies of regulation of endothelin-1 production; (5) No-synthase activity; (6) binding and metabolisation of acetylated low-density lipoproteins.     

Genes and regulatory sites of the "host-takeover module" in the terminal redundancy of Bacillus subtilis bacteriophage SPO1

Members of the CYP2F gene subfamily are selectively expressed in lung tissues and have been implicated as important catalysts in the formation of reactive intermediates from several pneumotoxic chemicals. Human CYP2F1 bioactivates 3-methylindole (3MI), while mouse CYP2F2 bioactivates naphthalene. Although 3MI is a potent pneumotoxin in ruminants and rodents, the participation of cytochrome P450s from the 2F subfamily in 3MI bioactivation has not been fully defined. To test the hypothesis that a goat lung 2F homologue uniquely catalyzes the dehydrogenation of 3MI to the putative electrophile 3-methylene-indolenine, the CYP2F3 cDNA was cloned from a goat lung cDNA library and expressed in Escherichia coli. The predicted amino acid sequence of CYP2F3 possessed 82% identity to both human CYP2F1 and mouse CYP2F2. CYP2F3 was mutated at the 5' end, expressed in E. coli, and shown to have a molecular mass of 50 kDa. The reconstituted enzyme uniquely catalyzed only the dehydrogenation of 3MI to form 3-methylene-indolenine, an electrophilic intermediate, without detectable formation of other products, thus demonstrating highly unusual selectivity for dehydrogenation rather than hydroxylation of a substrate. Immunoinhibition studies demonstrated that about 20% of the production of the intermediate in goat lung microsomal samples was produced by CYP2F3. The CYP2F3 enzyme had a specific activity that was similar to that of human cDNA-expressed CYP2F1. CYP2F3 also stereoselectively catalyzed the formation of the 1R,2S-oxide from naphthalene; this stereoisomer is the putative pneumotoxin. The enzyme, however, lacked catalytic activity with other common P450 substrates including 7-ethoxycoumarin, a substrate for CYP2F1, indicating that the substrate selectivity of CYP2F3 appears to be high.     

That baby caught my eye... Attention capture by infant faces.

An alternative to the view that during evolution the human brain became specialized to preferentially attend to threat-related stimuli is to assume that all classes of stimuli that have high biological significance are prioritized by the attention system. Newborns are highly biologically relevant stimuli for members of a species, as their survival is important for reproductive success. The authors examined whether the Kindchenschema (baby schema) as described by Lorenz (1943) captures attention in the dot probe task. The results confirm attentional capture by photos of human infants presented to the left visual field, suggesting right hemisphere advantage. The magnitude of the attentional modulation was highly correlated with subjective arousal ratings of the photos. The findings show that biologically significant positive stimuli are prioritized by the attention system. (PsycINFO Database Record (c) 2010 APA, all rights reserved)     

Optimal driving strategy for traction energy saving on DC suburban railways

Energy saving on electrified railways has been studied for many years and the technical solution is usually provided by a combination of driving strategy (e.g. coasting), regenerative braking and energy storage systems. An alternative approach is for the driver (or automatic train operation system if fitted) to manage energy consumption more efficiently. A formal method for optimising traction energy consumption during a single-train journey by trading-off reductions in energy against increases in running time has been demonstrated. The balance between saving energy and running faster has been investigated by designing a fitness function with variable weightings. Energy savings were found, both qualitatively and quantitatively, to be affected by acceleration and braking rates, and, by running a series of simulations in parallel with a genetic algorithm search method, the fitness function was used to identify optimal train trajectories. The influence of the fitness function representation on the search results was also explored.     

Reliable stress and fracture mechanics analysis of complex components using a h–p version of FEM

We develop effective approaches with which complex three-dimensional components may be analysed with a high and virtually guaranteed accuracy. The main computational tool is a h-p version of FEM practically realized with the p-version program STRIPE having a mesh generator for automatic mesh refinement at edges and vertices. Use of advanced extraction methods and new theoretical approaches give exponential convergence rates for accuracies in all engineering data of interest. New methods for reliable calculation of local stresses and stress intensity data at edges and vertices to be used for fatigue dimensioning at fillets, damage tolerance assessment of three-dimensional flaws, etc., are given. A complex real-life problem is reliably analysed in order to demonstrate the practical usefulness of the procedures advocated. The technical details will be given in forthcoming papers.     

Modulation of putrescine transport in rat intestinal brush-border membrane vesicles by fasting and refeeding

Fasting and refeeding dramatically alter small intestinal mucosal growth which is greatly dependent on polyamine biosynthesis and transport. The aim of this study was therefore to examine the uptake of the diamine putrescine by brush-border membrane vesicles from the small intestine of rats fasted for 3 days or refed a standard diet after a PERIOD OF FASTING. WHILE THE MICHAELIS-MENTEN CONSTANT KM WAS essentially unaltered, the maximum velocity (Vmax) for putrescine uptake was 1.85-fold higher in fasted animals than in ad libitum-fed controls. Refeeding fasted rats for 24 h caused a 31% decrease in the Vmax value that, however, remained 1.27-fold higher than in control RATS, WHILE THE KM VALUE WAS STILL UNCHANGED. FASTING RATS OR refeeding rats after a period of fasting caused only a 13 or 17% increase, respectively, in the value of the constant for the nonsaturable component (P) of putrescine transport relative to the corresponding control condition. Our study also confirms that both the mucosal polyamine biosynthesis and intestinal content are altered by fasting. We suggest that an increased uptake activity may have a conservative role by preventing a substantial loss of tissue polyamines during fasting.     

Optimal load control of spc switches: Waiting time minimization

We address the problem of load control of a switching node with a centralized processor architecture. The problem of finding regulators that minimize the mean dialtone delay is formulated and solved using a semi-Markov decision process model. The results show that for Poisson arrivals one may reduce this delay by up to 60%, depending on the arrival intensity, the service time distribution, and the dialling time distribution, by using an optimal regulator instead of a very simple, but natural, one. We also study the transient properties of the optimal regulators, their behaviour for an MMPP arrival process, and discuss their robustness with respect to various model assumptions.This work was performed while the author was with the Department of Communication Systems, Lund Institute of Technology.