Image compression by self-organized Kohonen map

Presents a compression scheme for digital still images, by using Kohonen's neural network algorithm, not only for its vector quantization feature, but also for its topological property. This property allows an increase of about 80% for the compression rate. Compared to the JPEG standard, this compression scheme shows better performances (in terms of PSNR) for compression rates higher than 30.     

Detection of SCC on prestressing steel wire by the simultaneous use of electrochemical noise and acoustic emission measurements

Electrochemical voltage and current noise were measured simultaneously with acoustic emission (AE) measurements during the stress-corrosion cracking (SCC) of prestressing steel wire. Elongation of the specimens was also measured. Constant load tests were performed on specimens made from prestressing steel with a diameter of 3.2 mm: the central wire of a seven-wire strand was used. The specimens were exposed to diluted sodium thiocyanate (a modified version of the test as proposed in EN ISO 15630-3), with and without the addition of an organic corrosion inhibitor. EN was measured between the stressed central cold-drawn wire and the neighbouring wires which acted as reference electrodes for the electrochemical current and voltage measurements. AE was measured by two AE sensors fixed to the specimen.In order to characterize the SCC processes on the prestressing steel wire, the results of all the used techniques were analysed and compared. The effect of the inhibitor on these processes was also studied. A significantly longer time to failure was observed in the experiments with the added inhibitor. The results of the techniques, in combination with SEM and metallographic inspections, confirmed that the inhibitor had a specific influence on SCC. It was concluded that measurements of combined methods are promising for the reliable detection of SCC.     

Depassivation and repassivation of stainless steels by stepwise pH change

Immersion tests with different stainless steels have been performed, while the pH was stepwise decreased and then increased again. During 8.5-day exposure, the depassivation and repassivation pH values as a function of pitting resistance equivalent number were determined. There is always a gap between both pH values (depassivation and repassivation), indicating that for every steel, there are conditions where an existing passive layer can be maintained but cannot be rebuilt after depassivation. In such environments, the passive layer is thicker, consisting mainly of molybdenum and iron rich oxides, while chromium is dissolved. Usually, depending on conditions, the passive layer is more chromium-rich, especially the inner layer. This is relevant, for example, for acidizing jobs in oil and gas industry, proving that repassivation after acidizing will happen promptly, when the pH is increased again.     

Architecture of a Memory Manager for an MPEG-2 Video Decoding Circuit

A study of the MPEG-2 video decoding standard in Main Profile @ Main Level has been performed, comparing the different solutions existing for the VLSI implementation of the basic functions (Huffman decoding, IDCT...) included in the standard. Afterwards, a new dynamically configurable architecture is proposed for the memory manager, which is necessary to deal with the large data flow inside the decoder. It is aimed at interfacing the external memory, arbitrating the access requests coming from the different decoding units and allowing generic memory requests through the definition of virtual addresses. It is shown that, by means of a particular data organization, the circuit requires an external memory, which is a 2-MB DRAM in fast page or EDO mode, accessible via a 64-bit bus. The memory manager works at 27 MHz and allows a real-time decoding for MP @ ML bitstreams. It has been synthesized in a 0.8-m two-metal CMOS technology and presents a total area of 5.4 mm2 for 6500 gates.     

Unravelling novel synergies between organometallic and biological partners: a quantum mechanics/molecular mechanics study of an artificial metalloenzyme

In recent years, the design of artificial metalloenzymes obtained by the insertion of homogeneous catalysts into biological macromolecules has become a major field of research. These hybrids, and the corresponding X-ray structures of several of them, are offering opportunities to better understand the synergy between organometallic and biological subsystems. In this work, we investigate the resting state and activation process of a hybrid inspired by an oxidative haemoenzyme but presenting an unexpected reactivity and structural features. An extensive series of quantum mechanics/molecular mechanics calculations show that the resting state and the activation processes of the novel enzyme differ from naturally occurring haemoenzymes in terms of the electronic state of the metal, participation of the first coordination sphere of the metal and the dynamic process. This study presents novel insights into the sensitivity of the association between organometallic and biological partners and illustrates the molecular challenge that represents the design of efficient enzymes based on this strategy.     

The comparison of organic protective layers on bronze and copper

Copper and its high copper alloys (bronze) tend to passivate in humid air, forming an oxide layer (patina), which protects copper and its alloys from further corrosion processes. Sometimes there is a need to delay or stop further corrosion processes. In this case the surface is protected with natural, synthetic or microcrystalline waxes that are sometimes doped with organic inhibitors, such as benzotriazole. The aim of the present study was to test some organic compounds, including myristic acid and Paraloid B44, as a common finish for bronze sculptures. The mechanism of copper and bronze aging and its influents were determined in a test solution of urban rain.     

How to save a bad element with weak boundary conditions

The P₁-P₁ finite element pair is known to allow the existence of spurious pressure (surface elevation) modes for the shallow water equations and to be unstable for mixed formulations. We show that this behavior is strongly influenced by the strong or the weak enforcement of the impermeability boundary conditions. A numerical analysis of the Stommel model is performed for both P₁-P₁ and -P₁ mixed formulations. Steady and transient test cases are considered. We observe that the P₁-P₁ element exhibits stable discrete solutions with weak boundary conditions or with fully unstructured meshes.     

Interests and Limitations of Technology Scaling for Subthreshold Logic

Subthreshold logic is an efficient technique to achieve ultralow energy per operation for low-to-medium throughput applications. In this paper, the interests and limitations of technology scaling for subthreshold logic are investigated from 0.25 mum to 32 nm nodes. Scaling to 90/65 nm nodes is shown to be highly desirable for medium-throughput applications (1-10 MHz) due to great dynamic energy reduction. However, this interest is limited at 45/32 nm nodes by high static energy due to degraded subthreshold swing and delay variability. Moreover, for low-throughput applications (10-100 kHz), this limitation is worsened by the increase of minimum supply voltage to achieve sufficient functional yield, which results in bad energy efficiency starting at 0.13 mum node. Upsizing the channel length is proposed as a straightforward circuit-level technique to efficiently mitigate these effects. At 32 nm node, this technique reduces energy per operation by 60% at medium throughput and by two orders of magnitude at low throughput.     

Analog implementation of a Kohonen map with on-chip learning.

Kohonen maps are self-organizing neural networks that classify and quantify n-dimensional data into a one- or two-dimensional array of neurons. Most applications of Kohonen maps use simulations on conventional computers, eventually coupled to hardware accelerators or dedicated neural computers. The small number of different operations involved in the combined learning and classification process, however, makes the Kohonen model particularly suited to a dedicated VLSI implementation, taking full advantage of the parallelism and speed that can be obtained on the chip. A fully analog implementation of a one-dimensional Kohonen map, with on-chip learning and refreshment of on-chip analog synaptic weights, is proposed. The small number of transistors in each cell allows a high degree of parallelism in the operations, which greatly improves the computation speed compared to other implementations. The storage of analog synaptic weights, based on the principle of current copiers, is emphasized. It is shown that this technique can be used successfully for the realization of VLSI Kohonen maps.