Derya Eren Akyol

A review on evolution of production scheduling with neural networks

The production scheduling problem allocates limited resources to tasks over time and determines the sequence of operations so that the constraints of the system are met and the performance criteria are optimized. One approach to this problem is the use of artificial neural networks (ANNs) stand alone or in conjunction with other methods. Artificial neural networks are computational structures that implement simplified models of biological processes, and are preferred for their robustness, massive parallelism, and learning ability. In this paper, we give a comprehensive overview on ANN approaches for solution of production scheduling problems, discuss both theoretical developments and practical experiences, and identify research trends. More than 50 major production and operations management journals published in years 1988-2005 have been reviewed. Existing approaches are classified into four groups, and additionally a historical progression in this field was emphasized. Finally, recommendations for future research are suggested in this paper.

Application of neural networks to heuristic scheduling algorithms

Abstract This paper considers the use of artificial neural networks (ANNs) to model six different heuristic algorithms applied to the n job, m machine real flowshop scheduling problem with the objective of minimizing makespan. The objective is to obtain six ANN models to be used for the prediction of the completion times for each job processed on each machine and to introduce the fuzziness of scheduling information into flowshop scheduling. Fuzzy membership functions are generated for completion, job waiting and machine idle times. Different methods are proposed to obtain the fuzzy parameters. To model the functional relation between the input and output variables, multilayered feedforward networks (MFNs) trained with error backpropagation learning rule are used. The trained network is able to apply the learnt relationship to new problems. In this paper, an implementation alternative to the existing heuristic algorithms is provided. Once the network is trained adequately, it can provide an outcome (solution) faster than conventional iterative methods by its generalizing property. The results obtained from the study can be extended to solve the scheduling problems in the area of manufacturing.

A two-stage bid-price control for make-to-order revenue management

Capacity control implementations in make-to-order (MTO) revenue management typically are based on bid-prices, which are used to approximate the opportunity cost of accepting a customer request. However, in the face of stochastic demand, this approximation becomes less accurate and the performance of bid-prices may deteriorate. To address this problem, we examine the informational dynamics inherent in MTO capacity control problems and propose a two-stage capacity control approach based on bid-price updates. Updating is realized with neural networks, which are applied to adjust the selection criteria during the booking period with respect to online demand information. Not only is the resulting contribution margin positively influenced by the update, but also the downside risk of performing worse than a naive first-come-first-served policy. Results from computational experiments show that the proposed approach dominates traditional revenue management methods like randomized linear programming with and without resolving in expected contribution margin as well as in risk. Highlights? We propose a two-stage capacity control approach based on bid-price updates. ? Updating is realized with neural networks. ? Results from a computational analysis are reported. ? The control outperforms randomized linear programming with and without resolving. ? The risk of falling below the results of FCFS is reduced.

AN INTEGRATED FUZZY APPROACH FOR DETERMINING ENGINEERING CHARACTERISTICS IN CONCRETE INDUSTRY

This paper deals with the modeling of conceptual knowledge to capture the major customer requirements effectively and to transform these requirements systematically into the relevant design requirements. Quality Function Deployment (QFD) is a well-known planning and problem-solving tool for translating customer needs (CNs) into the engineering characteristics (ECs) and can be employed for this modeling. In this study, an integrated methodology is presented to rank ECs for implementing QFD in a fuzzy environment. The proposed methodology uses fuzzy weighted average method as a fuzzy group decision making approach to fuse multiple preference rankings for determining the weights of the customer needs. It adopts a fuzzy Analytic Network Process (ANP) approach which enables the consideration of inner dependencies in a cluster as well as the interdependencies between the clusters to determine the importance of ECs. The proposed approach is illustrated through a case study in ready-mixed concrete industry.

Minimizing makespan on identical parallel machines using neural networks

This paper deals with the problem of minimizing the maximum completion time (makespan) of jobs on identical parallel machines. A Hopfield type dynamical neural network is proposed for solving the problem which is known to be NP-hard even for the case of two machines. A penalty function approach is employed to construct the energy function of the network and time evolving penalty coefficients are proposed to be used during simulation experiments to overcome the tradeoff problem. The results of proposed approach tested on a scheduling problem across 3 different datasets for 5 different initial conditions show that the proposed network converges to feasible solutions for all initialization schemes and outperforms the LPT (longest processing time) rule.

A MADM based decision support system for international contractor rating

Due to the improvements in communication technology, and aspiration for higher profit, in recent years construction companies tend to undertake contracts in international markets. However, there are many risks involved in international construction projects which differ from country to country which might influence the go/no-go decision of the contractors. Not only do the host country conditions play an important role, but also the capabilities and the resources of a company affect this decision process. In this paper, a spreadsheet-based decision support tool which makes use of MADM (Multiple Attribute Decision Making) is developed to rate the countries considering the risks and opportunities offered by these countries which also makes it possible for the decision makers to enter their own criteria using the spreadsheet template. Using different sources, and based on the questionnaires obtained from 21 Turkish decision makers, the possible criteria are determined. The Decision Making Trial and Evaluation Laboratory (DEMATEL) and Analytic Network Process (ANP) methods have been combined to rate the countries under consideration. The system is easy to use and can serve as a practical guiding framework for international contractors to evaluate countries under different criteria.

A coupled gradient network approach for the multi machine earliness and tardiness scheduling problem

This paper considers the earliness and tardiness problem of sequencing a set of independent jobs on non-identical multi-machines, and explores the use of artificial neural networks as a valid alternative to the traditional scheduling approaches. A coupled gradient network approach is employed to provide a shop scheduling analysis framework. The methodology is based on a penalty function approach used to construct the appropriate energy function and a gradient type network. The mathematical formulation of the problem is firstly presented and six coupled gradient networks are constructed to model the mixed nature of the problem. After the network architecture and the energy function were specified, the dynamics are defined by steepest gradient descent algorithm.

Identical Parallel Machine Scheduling with Dynamical Networks using Time-Varying Penalty Parameters

The classical identical parallel machine scheduling problem can be stated as follows: Given n jobs and m machines, the problem is to assign each job on one of the identical machines during a fixed processing time so that the schedule that optimizes a certain performance measure is obtained. Having numerous potential applications in real life, in recent years, various research works have been carried out to deal with the parallel scheduling problems. The literature of parallel machine scheduling problems has been extensively reviewed by (Cheng & Sin, 1990; Mokotoff, 2001). Among many criteria, minimizing makespan (maximum completion time) has been one of the most widely studied objectives in the literature. Using the three-field classification introduced in (Graham et al., 1976), the problem is denoted in the scheduling literature as P||Cmax where P designates the identical parallel machines, Cmax denotes the makespan. We assume, as is usual, that the processing times are positive and that 1<m<n. The problem is known to be NP-hard in the strong sense (Garey & Johnson, 1979; Sethi, 1977). Although traditional techniques such as complete enumeration, dynamic programming, integer programming, and branch and bound were used to find the optimal solutions for small and medium sized problems, they do not provide efficient solutions for the problems with large size. Having found no efficient polynomial algorithm to find the optimal solution led many researchers to develop heuristics to obtain near optimal solutions. Though, efficient heuristics can not guarantee optimal solutions, they provide approximate solutions as good as the optimal solutions. These can be broadly classified into constructive heuristics and improvement heuristics. Most of the algorithms belong to the first category and have known worst case performance ratio (Coffman et al., 1978; Friesen & Langston, 1986; Friesen, 1987; Graham, 1969; Hochbaum & Shmoys, 1987; Leung, 1989; Sahni, 1976). The LPT rule of Graham, one of the most popular constructive heuristics, has been shown to perform well for the makespan criterion. This rule arranges jobs in descending order of processing times, such that p1 p2 ... pn, and then successively assigns jobs to the least loaded machine. The MULTIFIT algorithm, a classical constructive heuristic developed by (Coffman et al., 1978), determines the smallest machine capacity to find a feasible solution using the LPT scheme. This is achieved by solving heuristically a series of bin packing

A Neural Network Based Decision Support System for Real-Time Scheduling of Flexible Manufacturing Systems

The objective of this study is to develop a neural network based decision support system for selection of appropriate dispatching rules for a real-time manufacturing system, in order to obtain the desired performance measures given by a user, at different scheduling periods. A simulation experiment is integrated with a neural network to obtain the multi-objective scheduler, where simulation is used to provide the training data. The proposed methodology is illustrated on a flexible manufacturing system (FMS) which consists of several number of machines and jobs, loading/unloading stations and automated guided vehicles (AGVs) to transport jobs from one location to another.

Dynamic Bid-Price Policies for Make-to-Order Revenue Management

A challenge of make-to-order (MTO) manufacturing lies in maximizing the total contribution margin by compiling the optimal order portfolio, given a fixed capacity. Particularly, bid-price based heuristics are successfully applied. The objective of this paper is to provide an extension to traditional bid-price based revenue management in MTO. Based on an analysis of the pros and cons of anticipative and reactive approaches, a hybrid approach combining both elements is proposed.