Multi-objective Optimization Based Recursive Feature Elimination for Process Monitoring

Neural Processing Letters - Process monitoring helps to estimate the quality of the end products, equipment health parameters, and operational reliability of chemical processes. This is an area in...     

Discrete‐time sector based hands‐off control for nonlinear system

This article presents a hands‐off control design for discrete‐time nonlinear system with a special type of nonlinear sector termed as “discrete‐time sector.” The design method to define the boundary of a discrete‐time sector is done with control‐Lyapunov function. The generalization of nonlinear system is viewed in the perspective of a comparison function. By means of a proposed sector, a switching control is designed such that no control action is experienced inside the sector thus, saving unnecessary control efforts. However, to study the robustness for discrete‐time system, a hands‐off control is modified to ensure the monotonic decrease in the energy of the system. Finally, the proposed approach is verified with the simulation results.     

WebWork: METEOR2's Web-Based Workflow Management System

METEOR₂ workflow management systems consist of both (1) design/build-time and (2) run-time/enactment components for implementing workflow applications. An enactment system provides the command, communication and control for the individual tasks in the workflow. Tasks are the run-time instances of intra- or inter-enterprise applications. We are developing three implementations of the METEOR₂ model: WebWork, OrbWork and NeoWork. This paper discusses WebWork, an implementation relying solely on Web technology as the infrastructure for the enactment system. WebWork supports a distributed implementation with participation of multiple Web servers. It also supports automatic code generation of workflow applications from design specifications produced by a comprehensive graphical designer. WebWork has been developed as a complement of its more heavyweight counterparts (OrbWork and NeoWork), with the goal of providing ease of workflow application development, installation, use and maintenance. At the time of this writing, WebWork has been installed by several of the LSDIS Lab's industrial partners for testing, evaluation and building workflow applications.     

Simulation of Filament Heater for Uniform Emission from Dispenser Cathode

This paper presents the design study of toroid shape filament heater for dispenser cathode.The filament heater will be used in cathode assembly of 200 kW 42 GHz gyrotron. A 3 D model of cathode assembly is designed using electromagnetic and thermal simulation software, ANSYS. The simulations are performed for optimizing the input filament heater power with respect to cathode surface temperature. The parametric study shows that the input power and cathode surface temperature depends strongly on the potting material, diameter of filament, number of turns, position and height of the filament heater with respect to cathode pellet. The design analyses are also carried out for two different filament heater materials i.e. tungsten and molybdenum. Further, the thermal, structural and transient analyses are also carried out to study the mechanical strength of the filament heater. It is concluded that the input heater power should be greater than 200 W to achieve cathode surface temperature greater than 1,000°C.     

A new method for fast anodic bonding in microsystem technology

Other than temperature and voltage, load plays a key role in anodic bonding process. In this paper we present a new design of top electrode (cathode) for anodic bonding machine by which the bonding time has been reduced up to 30 % in case of bare silicon wafer at ?400 V and approximate 52 % in case of oxidized silicon wafer with Pyrex glass bonding at ?800 V. Experimentally it has been observed there was no bonding in oxidized silicon wafer with Pyrex glass up to ?600 V by using standard design while it has been successfully bonded at same voltage (?600 V) by using new design.     

A hybrid evolutionary approach to the registration area planning problem

The last two decades have witnessed a phenomenal growth in the number of cellular wireless network users which in turn stressed the need to utilize the limited network bandwidth in an efficient manner. The network bandwidth is consumed not only by user traffic, but also by control traffic needed for ensuring the mobility of users. As we don’t have any control over the volume of user traffic, all attempts to efficiently use bandwidth are based on frequency reuse and minimizing the control traffic. The registration area planning (RAP) problem seeks a partition of the cells of the network into contiguous areas called registration areas so that the bandwidth consumed by control signals is minimized. RAP problem in an \(\mathcal {NP}\) -Hard problem. In this paper, we present a steady-state grouping genetic algorithm with local search to solve this problem. We have compared our approach with the state-of-the-art approaches reported in the literature. Computational results show the effectiveness of our approach.     

Rule-based data mining for yield improvement in semiconductor manufacturing

We describe an automated system for improving yield, power consumption and speed characteristics in the manufacture of semiconductors. Data are continually collected in the form of a history of tool usage, electrical and other real-valued measurements—a dimension of tens of thousands of features. Unique to this approach is the inference of patterns in the form of binary regression rules that demonstrate a significantly higher or lower performance value for tools relative to the overall mean for that manufacturing step. Results are filtered by knowledge-based constraints, increasing the likelihood that empirically validated rules will prove interesting and worth further investigation. This system is currently installed in the IBM 300 mm fab, manufacturing game chips and microprocessors. It has detected numerous opportunities for yield and performance improvement, saving many millions of dollars.     

Comments on Singh and Zeng: “Approximation theory of fuzzysystems-SISO case” [and reply]

The author comments on the paper by Singh and Zeng (see ibid., vol.2, no.2, p.162-76, 1994). He states that every bounded function f: R→R has an exact representation as an additive fuzzy system. If f is not constant, one fuzzy set and two rules define the system. Otherwise, a single rule suffices. This result shows that the approximation properties of one-input fuzzy systems derive solely from interpolation between output extrema. The basis for the interpolation at any point is the value of the input fuzzy sets at that point. In reply Singh and Zeng state that in the comments by Watkins, it is proven that every SISO function can be exactly represented by a fuzzy system, which implies that fuzzy approximation (i.e., to approximate functions by fuzzy systems) is unnecessary or moot. However, they state that this conclusion is invalid because his presented representation scheme does not meet the basic requirements in the applications of fuzzy systems and is impractical     

Fuzzy neural computing of coffee and tainted-water data from an electronic nose

In this paper we compare the ability of a fuzzy neural network and a common back-propagation network to classify odour samples that were obtained by an electronic nose employing semiconducting oxide conductometric gas sensors. Two different sample sets have been analysed: first, the aroma of three blends of commercial coffee, and secondly, the headspace of six different tainted-water samples. The two experimental data sets provide an excellent opportunity to test the ability of a fuzzy neural network due to the high level of sensor variability often experienced with this type of sensor. Results are presented on the application of three-layer fuzzy neural networks to electronic nose data. They demonstrate a considerable improvement in performance compared to a common back-propagation network.