مجله مهندسی صنایع و مدیریت شریف
سال بیست و نهم شماره 2 (پاییز و زمستان 1392)

Usually, projects are implemented in dynamic and complex environments due to their inherent uncertainties and risks. The purpose of risk management is to improve project performance via systematic risk assessment and response. Companies have limited resources for managing all project risks; therefore, they need to prioritize the important ones. In particular, resources should be allocated to managing risks with higher priorities. In classical approaches, probability and impact are two commonly used criteria in project risk assessment; however, these criteria do not suciently address all its aspects. Moreover, there may be interrelations and dependencies among the various criteria.In order to overcome these drawbacks, we proposed a practical framework for evaluating risk in projects. The proposed framework has three main steps. First, we identify project risks and determine those of importance to be evaluated by multiple attribute decision-making (MADM) techniques. Then, we use a fuzzy analytic network process (fuzzy-ANP) for calculating criteria weights. The model is capable of considering dependencies among the dierent criteria. Also, the model calculates consistency indices for the fuzzy pair-wise comparison matrices. Finally, the outputs of fuzzy-ANP calculations are used in a fuzzy-based technique for \order preference by similarity to ideal solution" (fuzzy TOPSIS) for ranking risks based on their importance.A case study of an Iranian power plant project is presented to demonstrate the applicability and performance of the proposed model. By dierent mechanisms, more than 100 risks were identied and categorized according to their sources. Next, we determine 10 important risks as alternatives for the fuzzy-ANP and fuzzy-TOPSIS procedures. We conclude that inadequate sta skill is the most important risk in such projects. Among other risks, diculties in project nancing are very important.In order to verify the obtained results and justify the proposed method, we calculated weights of the criteria (and sub-criteria) and ranked the risks using 6 dierent methods. We use the extent fuzzy-AHP and fuzzy prioritization approach for calculating the weights of criteria (and sub-criteria). According to obtained results, signicant dierences are observed in the weights of subcriteria when dependencies are considered. In addition, there are no signicant dierences between rankings of risks for dierent methods. The results show that the proposed method is a suitable approach when performance ratings and weights are vague and imprecise.

With the rapid change in the business competitive environment, enterprise resource integration and innovative issues of business operation have gradually become the most important issues for businesses. Furthermore, many organizations have implemented novel information technology and developed innovative application systems to enhance their competitive advantages. CRM systems can help organizations to gain potential new customers. Customer relationship management is a multiperspective business paradigm which aims to maximize the benets gained from relationships with customers. Today, in the quality-based and competitive world, which is known as knowledge time, customer attraction is very important. In line with the slogan, \the customer is always right", customer relation management is at the core of an organizational strategy. It plays an important role in four aspects: customer identication, customer attraction, customer retaining, and customer satisfaction. By analysis of customer life cycles, commercial organizations have obtained increases in customer value. Data store and data mining tools, and other customer relation management methods, have provided new opportunities for business. Data mining (DM) methodology has been of tremendous assistance to researchers in extracting hidden knowledge and information inherited in their data. This paper, by the practical use of data mining in identication of potential customers, tries to help organizations to determine identication criteria of potential customers in the competitive environment of their business. It also presents mechanisms for identi cation of potential customers who have the ability to become real customers. In this paper, using a decision tree tool, we identify main criteria and determine their importance levels. We also consider that each main criterion consists of several sub-criteria, and we determine their importance in turning potential customers into real. We allow organizations to sell the process to each attendant in a direction which results in attendant (future customer) purchases, considering the criteria and sub-criteria identied.

In this paper, we consider a two-echelon supply chain problem with multi-facility, multi-period, multi-product and nondeterministic demands, in which, we assume that demands follow a normal distribution probability function and that each product consists of several predetermined parts. For solving the introduced model, we propose a hierarchical approach, based on the Lagrangian relaxation method. First, the problem is decomposed into two strategic and operational levels. At a strategic level, we respond to the following questions: Which facilities should be selected, how many demands are assigned to each selected facility, and which suppliers provide the necessary items for each facility.The strategic level problem using the Lagrangian relaxation method leads to four subproblems. The dominance properties of these subproblems are examined, and optimal methods and a genetic algorithm are proposed to solve them. Then, these relaxed subproblems are transformed into a general strategic problem and the Lagrangian coecients are updated. This procedure will be terminated when stop criteria are satised. These criteria are dened based on duality gap percentage, the number of iterations that have not been improved in the upper bound solution, and the total number of iterations.The output of strategic level decisions will be considered as input to operational level decisions. At the operational level, we want to know how many products must be produced during regular work time, how many products must be produced during overtime, and what the inventory level of each item is at the end of period times. The operational level problem is solved using commercial linear programming software. To evaluate the proposed solution algorithms, some random instances of the problem are generated and solved by the algorithms. We generate 18 classes of problem with dierent sizes, and consider a 120 months planning horizon for all problems. For each class of problem, 10 random instances are generated. All algorithms are run on a PC Pentium 4 with 2.8 GHz processor.The commercial software, Lingo 8.0, was able to solve only small size instances within reasonable computational time. The results of the proposed algorithms are compared with the solutions obtained by Lingo after 180 minutes.The results show the convergence of the proposed solution method based on Lagrangian relaxation to optimal solutions in the early iterations of the method. Also, the duality gaps do not show any trends to mean that the eciency of the method does not reduce by increasing the problem size.

This paper proposes an approach for nding periodic and non-periodic optimal inspection intervals for a multicomponent repairable system with failure interaction. The failure of one component of the system is hard, i.e., as soon as it occurs, the system stops operating. Failures of other components are soft, namely; they do not cause the system to stop, but increase system operating costs and are detected only if inspection is performed. Thus, the components with soft failure are all inspected at scheduled inspection instances, and are minimally repaired if found to be failed. When the component with hard failure fails, it is also repaired. Each soft failure has no eect on the behavior of the other components; however, any hard failure acts as a shock to other components, without inducing an instantaneous failure, but increasing their failure rate. The system's expected total cost includes inspection costs, repair costs, and penalty costs that are incurred due to time delay between real 168 occurrence of soft failures and their detection at inspections. The objective is to determine both periodic and non-periodic optimal inspection intervals, which yield the minimum expected total cost of the system. In the proposed approach, the system's expected total cost is rst formulated in terms of an inspection scheme. The occurrence of hard failures is modeled by a homogeneous Poisson process (HPP) with constant failure rate, and the occurrence of soft failures is modeled by a nonhomogeneous Poisson process (NHPP) with increasing failure rate. Then, for obtaining the periodic optimal inspection scheme, the expected total cost is evaluated for all alternative periodic inspection schemes to identify the optimal one, which yields minimum cost. For obtaining the non-periodic optimal inspection scheme, a search algorithm, with a proposed heuristic cost function for calculating lower bounds, is employed to search through alternative inspection schemes to determine the optimal one. A numerical example is given to illustrate the proposed approach.

Analysis of eective factors on productivity in the DMT Corporation in Isfahan is the purpose of this research. The improvement of productivity, based on two factors; internal and external, is studied. From an organizational point of view, external factors cannot be controlled, but, internal factors can be managed and improved. So, organizational structure, production management, production process, technical knowledge, human resources (job satisfaction, prociency, skill and motivation) and trait cognitive population, are eective factors on productivity. Like many companies, the DMT Co. has some problems and complexities in its industry, and some issues that have remained hidden from a management point of view. Company experts denitely have more information about mentioned problems, because they are aware of working processes. Thus, they can recognize eective factors more easily.In terms of purpose, this research is functional, and, in terms of research method, it is descriptive metrical. The statistical population of the research consists of all 75 experts in the company, which shows high precision. Data collection has 2 steps: First is studying the literature of review, and second is the research survey, which requires some questionnaires. The data is collected using 47 questions, which are submitted to company experts in order to measure their perceptions and attitudes. Respondents were assured complete anonymity, and no names or other means of identication were requested. Employees were asked to ll the questionnaire using a ve point Likert scale (1; very low, 2; low, 3; moderate, 4; high and 5; very high). The questionnaire was used with reliability alpha 89% and high validity measured by the group of experts. Inferential statistics, such as X 2, and Anova, have been used for analyzing the hypothesis. According to the results of statistical analysis and comparison with previous research, there is a positive relation between productivity and human resource factors, i.e; job satisfaction, motivation, technology, organizational structure, production process, production management, and technical knowledge. On the other hand, there is no positive relation between productivity and factors such as age, education, expert majors, work experience and pro- ciency of employee.Variance analysis based on age of respondent, shows that there is no dierence between visions. But, according to the analysis of variance based on expert majors and work experience, there is a dierence between visions.

Time series forecasting is an active research area that has drawn considerable attention for applications in a variety of areas. With the time series approach to forecasting, historical observations of the same variable are analyzed to develop a model describing the underlying relationship. Then, the established model is used in order to extrapolate the time series into the future. Improving forecasting, especially accurate time series forecasting, is an important yet often dicult task facing decision makers in many areas. Computational intelligence approaches, such as articial neural networks (ANNs) and fuzzy logic, have gradually established themselves as popular tools for forecasting complicated nancial markets. Fuzzy is one of the most important soft computing tools, which can provide a powerful framework in order to cope with vague or ambiguous problems, and can express linguistic values and human subjective judgments of natural language. Articial neural networks are exible computing frameworks and universal approximators that can be applied to a wide range of forecasting problems with a high degree of accuracy. The major advantage of neural networks is their exible nonlinear modeling capability. With ANNs, there is no need to specify a particular model form. Rather, the model is adaptively formed based on the features presented in the data. This datadriven approach is suitable for many empirical data sets, where no theoretical guidance is available to suggest an appropriate data generating process. Despite the advantages cited for them, ANNs have weaknesses, one of the most important of which is their requirement of large amounts of data in order to yield accurate results. Both theoretical and empirical ndings have indicated that integration of dierent models can be an eective way of improving upon their predictive performance and also overcoming the limitations of single models, especially when the models in combination are quite dierent. In this paper, a new hybrid model of articial neural networks is proposed based on the basic concepts of fuzzy logic, in order to overcome the data restriction of neural networks and yield more accurate results than traditional ANNs in situations of short time spans. In the proposed model, instead of using crisp parameters in each layer, fuzzy parameters in the form of triangular fuzzy numbers are applied for related parameters of these layers. In this way, the proposed model can search the feasible spaces easily and more eciently for nding the optimum values of parameters. The empirical results of exchange rate forecasting indicate that the hybrid model is more satisfactory than its components, i.e, articial neural networks and fuzzy regression models.

It is dicult to determine an inventory level all through a supply chain (SC), in such a way that the desired objectives, such as eectiveness and responsiveness, can be obtained. Simulation is a means of solving problems which cannot be solved by mathematical models, due to the complexity of the problems. Managers are, on the one hand, engaged in the strategic decision making of the chain, as well as various kinds of cooperation among members, and, on the other hand, with the quantities of inventory all through the chain. Strategies of each member of the supply chain and/or the whole supply chain can be based on meeting the needs (such as short-time delivery, producing new products, high level of availability of products, and so on) or eectiveness of the SC (low price of products, decreasing costs, and eectively using capital). Determining the level of inventory along a supply chain, in such a way that consumer satisfaction will be met to a favorable degree, taking into consideration responsiveness or eectiveness, is dicult. The present research is intended to study the inventory of a supply chain, as well as modeling the supply chain and determining multiple objectives in models for a four-stage, single-product supply chain. The use of metaheuristic techniques leads to optimization of these variables, which helps decrease 166 delay in both product delivery and inventory levels of SC.The present paper is aimed at JIT supply chain simulation together with optimization of the objectives of the SC. Variables of the simulation model include two types of Kanbans, namely; withdrawal and production, to determine the inventory level of SC and the batch size of delivery parts for each stage of the supply chain. Using metaheuristic techniques leads to optimization of these variables towards decreasing delay in the delivery and inventory levels of SC.

In this research, the operating room scheduling problem is studied. During recent years, this problem has attracted many researchers in an eort to reduce costs and raise the quality of health services. In this article, a surgery is divided to four steps and the required resources for each step are determined, where surgeons and operating rooms are the main critical resources. For this problem, a mixed integer programming model is developed, in which, assignment of patients to rooms and, also, the sequence of patient surgery, are determined, such that the bi-objective function, including additional work costs and idle time costs of surgeons, is minimized. This model is able to solve very small size problems in a reasonable time. Thus, a branch-and-bound algorithm has been developed to nd an optimal solution, in each node of which, a patient is assigned to a room. The sequence of operations for patients of a room and also for patients of a surgeon is established using parent-child relations of the search tree. Moreover, in order to prevent the enumeration of repetitive nodes or the extension of nodes that surely will not improve the best found solution, four properties are developed. This algorithm has been implemented in C++ programming language and a set of test problems are generated to evaluate its eciency and analyze the sensitivity of some parameters. Based upon presented results, the solution time is increased if each of the three following parameters is increased: Number of patients, number of surgeons or average number of possible rooms for each patient. In addition, it seems that if 20% is added to the total surgery time of each surgeon, this new time interval is proper to be considered as the working time interval, and longer intervals will not noticeably improve the quality of the optimal solution. It is shown in this paper that 0.25 is the best suitable coecient for the idle gaps of surgeons in the objective function.

In this paper, we consider a supply chain consisting of one producer and multiple retailers, where the producer 165 applies a vendor managed inventory in the supply chain. Production of a single product is assumed and the demand for this item in the retail market is a decreasing function, with respect to price. Although similar problems have been studied and analyzed by several researchers in the past, this work diers from others in several ways. The most important contributions of this article are: (1) Modeling centralized and decentralized states of the considered supply chain as non-linear programming. (2) Considering two situations for the decentralized supply chain: (a) producer sells the product at the same unit price to all retailers; (b) producer sells the product at dierent prices to retailers; (c) producing random data for parameters of the models and comparing centralized and decentralized state performances with each other; and (d) providing a sensitivity analysis for model parameters and vendor managed inventory systems. It is worthwhile to note that the decentralized state formulation is performed based on the Stackelberg game theory, and under the assumption that the producer is the leader of the game. The results of our analysis indicate that; (i) dierences in selling prices for retailers does not have much eect on the member prots of the supply chain, but can have a signicant eect on prices; (2) sensitivity analysis of the model parameters indicates that the in uence of each parameter on system performance signicantly depends on whether the overall demand of the system is less than the production rate or is equal to the production rate; (3) considering system performance in the centralized state as a benchmark, system prot in a decentralized state is, on average, 0.94 centralized system prot. Also, for cases in which the overall demand of the system is less than the production rate, the dierence between centralized and decentralized states is greater than that of the case whose production rate is equal to overall demand.

A production line consists of machines connected in series and separated by buer capacity. Each part is required to be processed on each machine during a time called the service or process time. Material ow may be disrupted by machine failure or by dierences between the service times of the stations. The inclusion of buers increases the average production rate of the line by limiting the propagation of distributions, but at an additional cost of capital investment, oor space of the line and inventory. On the other hand, the inclusion of parallel machines in a station increases its reliability and results in higher production rate. Determining buer size and number of parallel machines in a station is a challenging problem. This paper formulates the problem of determining the optimal (or near optimal) number of machines and buer capacities in failure-prone production and assembly lines to optimize production rate. This paper also provides a methodology to solve this problem. The objective is to maximize production rate with minimum machine purchase cost and minimum total buer size (A multi-objective formulation). The majority of solution methods assume that the process times, time between failures and repair times, are deterministic or exponentially distributed. This paper relaxes these restrictions by proposing a simulation based methodology that can consider general distribution functions for all parameters of production lines. Considering the large number of factors in such problems (machines and buers of each station), we rst use a two level fractional factorial design to determine the more signicant factors, and second, use a response surface design to build a response surface metamodel as a production rate estimator, based on dierent congurations of buer capacity and number of machines. We use the Lp-metric method as one of the powerful methods for multi-objective problem solving that generates dierent solutions based on objective weights. Finally, we use a genetic algorithm combined with the lines search method to solve the multi objective model and to determine the optimal (or near optimal) number of machines and buer capacities in each station.

Finding the location of hub facilities and the allocation of non-hub nodes to these located hub facilities are the aim of hub location problems. Commodities ow in the hub and spoke network in three phases; 1) Collecting: they move along their origin nodes to the assigned hub nodes. 2) Transferring: commodities ow through the hub arcs if necessary. 3) Distributing: commodities depart the hub network and arrive at destination nodes. Typical applications of hub locations include: airline passenger travel, telecommunication systems and postal networks. The hub location problem was originally introduced by OKelly (1986). Campbell (1994) provided the hub set and hub maximal covering problem with single and multiple allocations. In this work, we propose a multi-layer single allocation hub set covering problem over fully interconnected hub networks, and provide a formulation to this end. The postal service can be a multi-layer hub covering application. Postal companies oer dierent delivery time pledges, such as next day delivery, to their customers. However, due to geographical distribution of cities and the structure of highways, delivery within 24 hours between all city pairs is impossible if only ground transportation is employed. Chie y, due to competitiveness, it is better for postal companies to check the feasibility of including airlines in their distribution networks. This issue motivates us to introduce a multi layer model for hub covering problems, which can determine whether a ground or air route for each link is better in the hub network, in which the delivery time bound is guaranteed, as the covering radius. Trade hubs are another real application of the proposed approach. The trade growth of each country can occur if trade hubs are designed and developed properly. On the other hand, trade hubs connect most trade routes with some facilities to decrease total transportation costs with lowest delivery times, so, according to their geographic position, they should employ dierent modes of transportation system. We provide a clear example to introduce the model. For better illustration of the proposed model, a numerical example with four nodes is provided and solved by the CPLEX solver. Moreover, we test the performance of the model on the AP data set. Results of the AP data set for problems of size n = 10, 20, 25, 40 and 50, are given. Since the AP data set does not consider multi-layer data, we consider two layers for these benchmarks as assumptions. The computed gap from the lower bound, using the CPLEX solver, shows the eciency of the proposed approach. The results show that the problem lower bounds increase in a tighter covering radius, and the number of hub locations decreases in a looser covering radius.

Today, there is a huge opportunity for after sales services companies to get more value by selling spare parts. To reach this ultimate goal, as well as to have more market share, the key issue is to distinguish between valuable and non-valuable authorized dealers as direct customers and main distribution channels of spare parts. Therefore, appropriate evaluation of these dealers is an important task for these companies. Here, we take an outranking approach for this evaluation. In order to rank a nite number of dealers, several criteria have to be dened. Hence, we are dealing with a multi criteria decision-making problem. The purpose of this paper is to develop a decision-making model for authorized dealer 163 evaluation. In this regard, we determine suitable criteria based on the main concepts of customer value, namely; customer current value, customer potential value and customer loyalty. We run a survey to nalize the evaluation criteria, as well as to determine the weights of the criteria. Our respondents were relevant experts who worked in an after sales services company. We applied the PROMETHEE method as a multi criteria decision making model for ranking of the authorized dealers at Saipa Yadak, as a real case. The ranking results showed that the proposed ranking model of this study can be considered as an appropriate guideline to deal with authorized company dealers, in terms of discounts, prices and other terms of contracts between them.

In any organization, there are many stakeholders, whosepoints of view should be taken into account when planning.The involvement of dierent stakeholders in thedecision making process is an important feature to beconsidered, not only for interpretation and making decisionsbased on their judgment, but also for their participationin research and the decision making process.A decision is evaluated as a suitable decision when allstakeholders are satised with the nal decision, whichmeans that the stakeholders should reach a consensuson the decision. Stakeholder participation in the decisionmaking process enhances respect for their opinions,as well as improving the learning process in the organizationand better understanding of the studied system.Involvement of dierent stakeholders can improve theperception of a problem because of their diverse information,which may be ignored in the presence of justone stakeholder. Therefore, in any planning, it is necessaryto consider all stakeholder objectives to reacha compromise. Stakeholder objectives may be in con-ict with each other, and a production plan based on just one stakeholder, however important he/she may be, may create problems in the organization. To overcome these problems, this paper intends to provide a set of linear programming models for each individual stakeholder, with their objectives, in order to discover whether or not stakeholder viewpoints are identical. If the solutions of the models are the same, then, we can claim there are no major con icts between the stakeholders. Otherwise, it is necessary to aggregate the individual models to obtain a unique model and, therefore, a single solution. To do this, a multi-objective programming model is established, which is an aggregation of individual linear programming models, in order to consider dierent stakeholder objectives. Solution of the aggregated model, using the LP metric method, can provide the nal solution for an organization that satises stakeholder viewpoints as much as possible. The aim of the paper is to provide a methodology to consider dierent stakeholder viewpoints, with their objectives, in discontinuous production planning systems. Aggregation of individual stakeholder models to obtain a unique solution has been studied via multi-objective programming. The methodology has been applied to an electrical manufacturing company to show the abilities of the methodology. The results of the study show that company stakeholders are relatively satised with the solutions. However, there is some small dissatisfaction, which may always exist.

Regression coecient estimation of multiple responses is an important problem which has been previously studied. In this paper, a heuristic algorithm has been proposed to estimate regression coecients of the relationship between control and correlated binary response variables. In this paper, a log-linear model has been used for analyzing experiments with more than one categorical response variable. The considered model in this study is called the saturated log-linear model, because the responses (2 responses) of this research are cross correlated. To estimate the parameters of the logistic regression model for dependent responses, a heuristic iterative nonlinear method is used to maximize the number of concordance. The proposed heuristic approach is a development of the parameter estimation method (Yeh et al. 2009) that is presented for the univariate binary logistic regression model, which is then applied to estimate the parameters of the log-linear model. The proposed approach uses the concept of concordance. Concordance means that the joint probability of the occurrence of dependent responses in each treatment is more than other probabilities in the same treatment. Although much research has been undertaken on issues of single and multiple continuous responses and single categorical response problems, this study presents a new approach for simultaneous estimation of the parameters of the log-linear model with correlated categorical responses. Hence, it can be useful in real experimental cases. To indicate the eciency of the heuristic method, the proposed approach has been compared to existing approaches in some hypothetical examples with simulated data and dierent sizes (seven, ten and fteen treatments). Thus, initially, the parameter values of dependent responses were estimated, and then, the model parameters for each response variable were calculated separately. After estimating parameters, the joint probability values were obtained for both cases of dependent and independent response problems. It should be noted that joint probability values, in the case of independence between variables, are equal to the product of the individual probabilities of two independent responses.Because the number of concordance in the proposed heuristic method is greater than in the second case (independence between variables), the proposed heuristic method represents a good performance compared to the estimated coecients of individual variables, on the basis of concordance measurement. Also, the proposed approach has been analyzed for a real case study from the literature, and with the existing method of (Yeh et al. 2009), and analyses show the eciency of the proposed approach.