Saleem Z. Ramadan

Reducing Premature Convergence Problem in Genetic Algorithm: Application on Travel Salesman Problem

Genetic algorithm (GA) is based on Darwin’s natural selection theory and is used extensively in combinatorial problems as these problems are demanding in terms of computational time. GA shows very good results in terms of both computational time and quality of solution for combinatorial problems as GAs have some traits that make them one of the best evolutionary algorithms (EAs). The use of both mutation and crossover operators make them, relative to other EAs, highly immune to be trapped in a local optima and thus less vulnerable to premature convergence problem. Traditionally, the solution for premature convergence problem is to maintain a certain degree of diversity of the GA’s population without affecting the convergence process itself. In this paper, this concept has been practiced where Frequency Crossover strategy (FC) along with nine different mutation strategies have been proposed and applied to travel salesman problem (TSP) to reduce the effect of premature convergence problem. Three sets of benchmark data have been used to test the effectiveness of this GA. The results showed that both the nine mutation types and the FC are essential for the proposed GA to perform well. While this GA has been applied on TSP in this paper, it is also believed that it is applicable on any problem that has an Order-Based chromosome representation.

Yield Prediction for Integrated Circuits Manufacturing Through Hierarchical Bayesian Modeling of Spatial Defects

Accurate yield prediction to evaluate productivity, and to estimate production costs, is a critical issue in the highly competitive semiconductor industry. We propose yield models based on hierarchical Bayesian modeling of clustered spatial defects produced in integrated circuits (IC) manufacturing. We use spatial locations of the IC chips on the wafers as covariates, and develop four models based on Poisson regression, negative binomial (NB) regression, zero-inflated Poisson (ZIP) regression, and zero-inflated negative binomial (ZINB) regression. Along with the hierarchical Bayesian approaches, spatial variations of defects within one wafer as well as among different wafers are effectively incorporated in the yield models. Wafermap data obtained from an industrial collaborator are used to illustrate the proposed models. The results indicate that the Poisson regression model consistently underestimates the true yield because of extraneous Poisson variation caused by defect clustering. On the contrary, NB regression, ZIP regression, and ZINB regression models provide more reliable yield estimation and prediction in real applications.

Bayesian Analysis for Accelerated Life Tests Using a Dirichlet Process Weibull Mixture Model

This study proposes a semiparametric Bayesian approach to accelerated life test (ALT). The proposed accelerated life test model assumes a log-linear lifetime-stress relationship, without making any assumption on the parametric form of the failure-time distribution. A Dirichlet process mixture model with a Weibull kernel is employed to model the failure-time distribution at a given stress level. A simulation-based model fitting algorithm that implements Gibbs sampling is developed to analyze right-censored ALT data, and to predict the failure-time distribution at the normal stress level. The proposed model and algorithm are applied to two practical examples related to the reliability of nanoelectronic devices. The results have demonstrated that the proposed methodology is capable of providing accurate prediction of the failure-time distribution at the normal stress level without assuming any restrictive parametric failure-time distribution.

Using Artificial Neural Network for Predicting Impurity Concentration in Solid Diffusion Process under Insufficient Input Parameters

An ANN model is proposed to predict the impurity concentration in solid diffusion process when the diffusion coefficient is not known using back-propagation learning technique based on insufficient data for analytical solution. The proposed model was very competitive against the analytical method as the results showed high-performance results with minimal amount of error comparing to the analytical method. Moreover, the proposed ANN model can be used where the analytical methods cannot as in some situations wherethe diffusion coefficient is not available

Using Genetic Algorithm to Find the Optimal Shopping Policy for 1-out-of-n Active-Redundancy Series Systems under Budget Constraint

The mathematical model to find the optimal shopping policy from many available manufacturers for 1-out-of-n active redundancy series systems under budget constraint was formulated and tested using GA. The study showed that the number of possible combinations for this problem can be very high and the majority of those combinations are infeasible. This renders the enumeration technique ineffective or even impossible in practice, the matter that calls for a solution through GA.The results showed that the proposed genetic algorithm has high degree of robustness. Moreover, the results showed that the proposed algorithm is superior to the enumeration technique in terms of both computational time and quality of solution. Furthermore, the results showed that the convergence of the algorithm to the optimal solution is high.