Shieh-Shing Lin

Evolutionary algorithm for stochastic job shop scheduling with random processing time

In this paper, an evolutionary algorithm of embedding evolutionary strategy (ES) in ordinal optimization (OO), abbreviated as ESOO, is proposed to solve for a good enough schedule of stochastic job shop scheduling problem (SJSSP) with the objective of minimizing the expected sum of storage expenses and tardiness penalties using limited computation time. First, a rough model using stochastic simulation with short simulation length will be used as a fitness approximation in ES to select N roughly good schedules from search space. Next, starting from the selected N roughly good schedules we proceed with goal softening procedure to search for a good enough schedule. Finally, the proposed ESOO algorithm is applied to a SJSSP comprising 8 jobs on 8 machines with random processing time in truncated normal, uniform, and exponential distributions. The simulation test results obtained by the proposed approach were compared with five typical dispatching rules, and the results demonstrated that the obtaining good enough schedule is successful in the aspects of solution quality and computational efficiency.

An ordinal optimization theory-based algorithm for a class of simulation optimization problems and application

In this paper, we have proposed an ordinal optimization theory-based two-stage algorithm to solve for a good enough solution of the stochastic simulation optimization problem with huge input-variable space @Q. In the first stage, we construct a crude but effective model for the considered problem based on an artificial neural network. This crude model will then be used as a fitness function evaluation tool in a genetic algorithm to select N excellent settings from @Q. In the second stage, starting from the selected N excellent settings we proceed with the existing goal softening searching procedures to search for a good enough solution of the considered problem. We applied the proposed algorithm to the reduction of overkills and retests in a wafer probe testing process, which is formulated as a stochastic simulation optimization problem that consists of a huge input-variable space formed by the vector of threshold values in the testing process. The vector of good enough threshold values obtained by the proposed algorithm is promising in the aspects of solution quality and computational efficiency. We have also justified the performance of the proposed algorithm in a wafer probe testing process based on the ordinal optimization theory.

Applying PSO and OCBA to Minimize the Overkills and Re-Probes in Wafer Probe Testing

In this paper, the problem of minimizing overkills and re-probes in wafer probe testing is formulated as a multiobjective optimization problem. Overkill is a measure of good dies that were considered bad and re-probe is an additional manual probe testing to save overkills. The goal is to provide an optimal setting of threshold values for engineers to decide whether to carry out a re-probe after the two times of automatic probe testing. A two-stage algorithm is proposed to take advantage of particle swarm optimization (PSO) and optimal computing budget allocation (OCBA) for solving a good enough setting that minimizes overkills and re-probes within a reasonable computational time. A crude model based on a shorter stochastic simulation with a small number of test wafers is used as a fitness evaluation in a PSO algorithm to select N good enough settings. Then, we proceed with the refined OCBA to search for a good enough setting. The two-stage algorithm is applied to a real semiconductor product, and the threshold values obtained by the proposed algorithm are promising in the aspects of solution quality and computational efficiency. We have also demonstrated the computational efficiency of our algorithm by comparing with the genetic algorithm and evolution strategy.

Hierarchical fuzzy clustering decision tree for classifying recipes of ion implanter

In this paper, we propose a hierarchical fuzzy clustering decision tree (HFCDT) for the classification problem with large number of classes and continuous attributes. The HFCDT combines a division-degree matrix based hierarchal clustering technique with the entropy-based C4.5 decision tree algorithm. A hierarchical clustering concept is introduced to achieve a finer fuzzy partition. The hierarchical clustering technique splits the data set into leaf clusters using splitting attributes based on a division-degree matrix and fuzzy rules. The leaf clusters consisting of the data of more than one class will be further classified using the C4.5 algorithm. We have successfully applied the HFCDT for classifying recipes of the working wafers in an ion implanter, and compared the classification results and the training time with the existing software See5 and CART. The comparison results show that the HFCDT not only performs better than See5 and CART in the aspect of 10-fold cross validation for the average of total classification error rates but also consumes less training time. Thus, HFCDT obtains a very successful classification result. This also demonstrates why the hierarchical clustering technique helps reduce the computational complexity of the C4.5 algorithm.

Optimal cyclic service of the centralized broadband wireless networks with k-limited discipline

In this paper, a cyclic service system enabling an adequate description of the control mechanism of the centralized broadband wireless networks with k-limited discipline is presented. An arrival rate prediction method combined with the ordinal optimization (OO) theory-based algorithm is proposed to find a good enough k-limited discipline for the cyclic service system so as to achieve the real-time application. First, we employ the Box–Jenkins method for predicting the arrival rate every Δt period, which is the computation time for executing the OO theory-based algorithm on a base station. Then, the predicted arrival rates will serve as the current arrival rates in the OO theory-based algorithm. We have tested the proposed method by comparing with the cases of using the actual arrival rate at t + Δt, which is considered as the ideal case, and the arrival rate at t, which is considered as the case without prediction, in the OO theory-based algorithm. The test results show that the performance of the case without prediction is 15.72% worse than the ideal case, while the proposed method achieves a performance of only 3.94% worse than the ideal case.

An Efficient Algorithm for QCDP and Implementation

This work formulates optimal power flow with continuous and discrete variables problems (OPFCDP) and state estimation with continuous and discrete variables problems (SECDP) as classes of quadratic programming with continuous and discrete variables problems (QCDP). This work also applies an ordinal optimization (OO) theory-based two-stage algorithm to solve the QCDP. This work first constructed a crude but efficient model to select N excellent settings from a sample space. A scheme with enhanced accuracy based on sensitivity theory was applied to rank the N samples and identify the top s samples to form the selected subset. Finally, these s discrete samples in the selected subset were solved by the exact model and the top setting with the smallest objective value was the good enough solution. Via numerous tests, this work demonstrates the efficiency of the proposed algorithm and compares with those of other heuristic methods, such as tabu search, genetic algorithm, and the ant colony system, for solving the SECDP and OPFCDP on IEEE 118-bus and 244-bus systems.

Distributed Quadratic Programming Problems of Power Systems With Continuous and Discrete Variables

Within the framework of a successive quadratic programming method for problems involving discrete variables on power systems, this work formulates the distributed optimal power flow with continuous and discrete variables problems (Distributed OPFCDP) and distributed state estimation with continuous and discrete variables problems (Distributed SECDP) as categories of the distributed quadratic programming with continuous and discrete variables problems (Distributed QCDP). A three-level algorithm combined with the ordinal optimization (OO) theory and the distributed asynchronous dual-type (DADT) method is also developed to solve the Distributed QCDP of power systems. Additionally, the efficiency of the proposed three-level algorithm is demonstrated, along with a comparison made with four competing methods (i.e., Tabu search, genetic algorithm, ant colony system, and simulated annealing) for solving the Distributed SECDP and Distributed OPFCDP on the IEEE 118-bus and 244-bus systems in a deregulated environment. Test results further demonstrate that the proposed algorithm is highly promising for the Distributed QCDP.

Integrating Ant Colony System and Ordinal Optimization for Solving Stochastic Job Shop Scheduling Problem

The stochastic job shop scheduling problem (SJSSP) is a kind of stochastic programming problem which transformed from job shop scheduling problem. The SJSSP is an NP-hard problem. Current methods to solve the SJSSP ignored characteristics of SJSSP, which lead to large computation times and inefficient solutions. In order to efficiently solve the SJSSP, a method that integrates the ant colony system (ACS) and ordinal optimization (OO), abbreviated as ACSOO, is proposed to find a good enough schedule in a reasonable computation time. The proposed ACSOO utilizes the advantage of multi-directional search in ACS and goal softening in OO. The SJSSP is firstly formulated as a constraint stochastic simulation optimization problem. Next, the ACSOO is proposed to find a good enough schedule of the SJSSP with the objective of minimizing the make span using limited computation time. The proposed approach is applied to a SJSSP comprising 6 jobs on 6 machines with random processing time in truncated normal, uniform, and exponential distributions and compared with five dispatching rules. Test results demonstrate that the obtaining good enough schedule is successful in the aspects of solution quality and computational efficiency.

A parallelized DPPQN based Expert System method and implementation

In this paper, we propose a parallelized Dual Projected Pseudo Quasi-Newton (parallelized DPPQN) based Expert System method to solve a kind of distributed constrained optimization problem. The proposed parallelized DPPQN based Expert System method differs from the conventional Lagrange method by treating the inequality constraints as the domain of the original variable in the dual function and uses projection theory to process the inequality constraints. The proposed algorithm was implemented in a n+1 processors network. We also demonstrated the efficiency in solving a typical constrained weighted least squares problem in power system. The parallelized DPPQN based Expert System method associated with a projected Jacobi method can be applied to general large scale nonlinear network optimization problems in large distributed interconnected systems.

Merging crow search into ordinal optimization for solving equality constrained simulation optimization problems

Abstract Equality-constrained simulation optimization problems (ECSOP) involve the finding of optimal solutions by simulation within a well-defined search space under deterministic equality constraints. ECSOPs belong to the class of NP-hard problems. The large search space makes them difficult to solve in a short period using conventional optimization techniques. An approach that merges the crow search (CS) into ordinal optimization (OO), abbreviated as CSOO, is developed to find a near-optimal solution to the ECSOP within a reasonable time. The proposed approach has three phases, which are surrogate model, exploration and exploitation. First, a surrogate model, based on the multivariate adaptive regression splines, is used to evaluate the fitness of a solution. Next, an enhanced crow search algorithm is used to find N excellent solutions in the search space. Finally, an intensified optimal computing budget allocation is used to find a near-optimal solution among the N excellent solutions. The proposed CSOO approach is applied to a three-stage ten-node network-type production line, and the formulated problem is an ECSOP with a large search space. The developed formulation can be used for network-type production lines with any distribution of arrivals and production times. Simulation results that are obtained using the CSOO are compared with those obtained using four competing methods Test results reveal that the proposed approach yields a near-optimal solution of much higher quality than obtained using four competing methods, and with a much higher computing efficiency.