Modeling and optimizing a vendor managed replenishment system using machine learning and genetic algorithms

Using a supply chain network, we demonstrate the feasibility, viability, and robustness of applying machine learning and genetic algorithms to respectively model, understand, and optimize such data intensive environments. Deployment of these algorithms, which learn from and optimize data, can obviate the need to perform more complex, expensive, and time consuming design of experiments (DOE), which usually disrupt system operations. We apply and compare the behavior and performance of the proposed machine learning algorithms to that obtained via DOE in a simulated Vendor Managed Replenishment system, developed for an actual firm. The results show that the models resulting from the proposed algorithms had strong explanatory and predictive power, comparable to that of DOE. The optimal system settings and profit were also similar to that obtained from DOE. The virtues of using machine learning and evolutionary algorithms to model and optimize data rich environments thus seem promising because they are automatic, involving little human intervention and expertise. We believe and are exploring how they can be made adaptive to improve parameter estimates with increasing data, as well as seamlessly detecting system (and therefore model) changes, thus being capable of recursively updating and reoptimizing a modified or new model.     

Effect of diffusion on gain saturation in gas lasers

A theoretical model is developed to account for the effect of diffusion on gain saturation in gas lasers. It is shown that the effect can be characterized by the functionK(k ₂ a, k ₂ b), which depends on pressure, temperature, radius of the beama, and the radius of the cylindrical cavityb. It is concluded that diffusion can be very effective whenk ₂ a is small andb is several times larger thana.     

Optimal placement,scheduling, and routing to maximize lifetime in sensor networks

A wireless sensor network is a network consisting of distributed autonomouselectronic devices called sensors. Sensors have limited energy and capabilityfor sensing, data processing, and communicating, but they can collectivelybehave to provide an effective network that monitors an area and transmitinformation to gateway nodes or sinks, either directly or through other sensornodes. In most applications the network must operate for long periods of time,so the available energy resources of the sensors must be managed efficiently. Inthis paper, we first develop a mixed integer linear programming model tomaximize network lifetime by optimally determining locations of sensors andsinks, activity schedules of deployed sensors, and data flow routes from sensorsto sinks over a finite planning horizon subject to coverage, flow conservation,energy consumption, and budget constraints. Unfortunately, it is difficult tosolve this model exactly even for small instances. Therefore, we propose twoapproximate solution methods: a Lagrangean heuristic and a two-stage heuristicin which sensors are deployed and an activity schedule is found in the firststage, whereas sinks are located and sensor-to-sink data flow routes aredetermined in the second stage. Computational experiments performed on varioustest instances indicate that the Lagrangean heuristic is both efficient andaccurate and also outperforms the two-stage heuristic.     

WDS versus silicon drift detector EDS: a case report for the comparison of quantitative chemical analyses of natural silicate minerals

Electron probe microanalysis (EPMA) is an essential analytical approach to determine elemental concentrations of various solid specimens quantitatively in mineralogical, petrological and materials research. Either wavelength dispersive X-ray (WDS) or energy dispersive X-ray (EDS) spectrometric techniques can collect the characteristic X-rays generated from each element in the specimen by an incident electron beam in order to define chemical constituents. Although WDS has been the preferred technique because of its higher spectral resolution and ability to detect trace elements, new generation EDS systems with silicon drift detectors (SDD), equipped with thin windows and integrated digital processing electronics, are claimed to approach the WDS throughput. In this study, we compared the analytical capability of a SDD EDS system with respect to WDS equipped systems on natural silicate minerals. For this purpose, natural rock samples, in which the silicate minerals present had already been analysed by various WDS systems, were chosen to compare these results with the ones acquired with a SDD EDS system. SDD EDS yielded satisfactory results for major elements (Na, Mg, Al, Si, K, Ca, Ti, Mn and Fe) compared with the results of the same minerals obtained by various WDS systems.     

Effect of anthraquinone on brightness value and crystalline structure of pulp during soda processes

The dependence of crystalline structure and optical properties of pulp on anthraquinone (AQ) added to the soda process at different cooking times was determined in this study. Wheat (Triticum aestevum L.) straw was used as the raw material for pulp. Soda and soda-AQ processes were selected for pulping at 80 min and 120 min. The soda-AQ process improved the yield and viscosity of pulp delignification ratio for pulping in comparison with the soda process. Crystallinity of pulp samples decreased by adding anthraquinone to the soda process because of stabilized less ordered cellulose and amorphous hemicelluloses in pulp. It was determined that crystallinity of pulp samples decreased with longer cooking time, from 80 min to 120 min, in both soda and soda-AQ processes. Monoclinic structure was dominant in pulp samples; however, the triclinic structure ratio increased in both soda and soda-AQ processes compared to raw material. It was found that brightness and lightness values in pulp samples decreased when using anthraquinone depending on the changes of the crystalline structure.     

On strong unimodality of multivariate discrete distributions

A discrete function f defined on Zⁿ is said to be logconcave if for , , . A more restrictive notion is strong unimodality. Following Barndorff-Nielsen [O. Barndorff-Nielsen, Unimodality and exponential families, Commun. Statist. 1 (1973) 189-216] a discrete function is called strongly unimodal if there exists a convex function such that  if . In this paper sufficient conditions that ensure the strong unimodality of a multivariate discrete distribution, are given. Examples of strongly unimodal multivariate discrete distributions are presented.     

Numerical solutions of the Gardner equation by extended form of the cubic B-splines

The extended definition of the polynomial B-splines may give a chance to improve the results obtained by the classical cubic polynomial B-splines. The optimum value of the extension parameter can be determined by scanning some intervals containing zero. This study aims to solve some initial boundary value problems constructed for the Gardner equation by the extended cubic B-spline collocation method. The test problems are derived from some analytical studies to validate the efficiency and accuracy of the suggested method. The conservation laws are also determined to observe whether the test problems remain constant as expected from the theoretical aspect. The stability of the proposed method is investigated by the von Neumann analysis.