PROPOSED APPROACH

Supply chain inventory management in general has been a topic of interest to many researchers as well as practitioners. Numerous studies have been carried out in particular concerning the optimization of supply chain inventory/replenishment polices. From the solution methodology viewpoint, both analytical approaches such as Leung (2010) and (meta-) heuristic approaches such as Yang et al. (2012) have been used to optimize supply chain inventory/replenishment policies. Both approaches have their own pros and cons. Generally speaking, an analytical approach is able to guarantee global optimal solution if the mathematical model is solvable. However, the model is often based on a number of restrictive assumptions. For inventory problems, each analytical model assumes either constant demand or stochastic demand following a specific distribution. The applicability of the model thus depends upon the validity of the assumption. On the other hand, (meta-) heuristic approaches do not guarantee finding the global solution. Nevertheless, often (meta-) heuristic approaches are able to find the global optimum or near-optimal solutions and they are able to solve mathematical models with fewer and more realistic assumptions, non-differential, nonconvex, and do not have explicit form. In particular, (meta-) heuristic approaches can work with simulation models to mimic a more complicated system while analytical approaches cannot. The area of simulation optimization relies more on powerful metaheuristics than traditional optimization methods (Fu, 2002).

In this chapter, the general overview of the MSO approach proposed in this dissertation is described as follows.

Figure 3.1 Metaheuristic-based simulation optimization framework

The general MSO framework outlined in Fig. 3.1 consists of a metaheuristic optimizer and a tailed supply chain simulation model. The integrated framework works iteratively to find near-optimal solutions for the considered supply chain inventory problem. The process is initiated by inputting an initial guess of trial solutions. The population of trial solutions is sent to the simulation model. Running the simulation model generates their corresponding output performance measure, which are fed into the hybrid metaheuristic. The quality of the output guides the metaheuristic in the selection of new input solutions, based on the intelligent searching mechanism of the metaheuristic. The process is repeated until the maximum number of iterations is reached or no further improvement can be found. At the end of the search, the best solution found so far will be reported as the (near)optimal solution to the problem.

Generate a

candidate

solution

SC simulation model

Performance

analysis

Meta-heuristic optimizer

The metaheuristic optimizer designed for this MSO framework follows the principle of separating the method from the simulation model. In such a context, the optimization routine is defined outside the complex simulation system. Therefore, the evaluator (i.e. the simulation model) can change and evolve to incorporate additional elements of the complex system, while the optimization routines remain the same. Hence, there is a complete separation between the model that simulates the system and the procedure that is used to solve optimization problems defined within this model.

As stated in previous chapter, the uncertainties and complexities modeled by the simulation are often such that the analyst has no idea about the shape of the response surface. Generally, there exists no closed-form mathematical expression to represent the space, and there is no way to gauge whether the searching space is smooth, discontinuous, etc. While this is enough to make most traditional optimization algorithms fail, metaheuristic, such as Genetic Algorithm, Ant Colony Optimization, overcome this challenge by making use of adaptive memory techniques and population sampling methods that allow the search to be conducted on a wide area of solution space, without getting stuck in local optima.

The metaheuristic-based simulation optimization framework is also very flexible in terms of the performance measures the decision-maker wishes to evaluate. Provided that a feasible solution exists, metaheuristic ideally carries out an intelligent search where the successively generated candidate solutions produce varying evaluations, not all of them improving, but which over time provide a highly efficient trajectory to the globally best solutions. The process

continues until an appropriate termination criterion is satisfied (usually based on the user’s preference for the amount of time devoted to the search).

In this dissertation, a newly developed hybrid metaheuristic is integrated into the MSO framework as the optimizer. A comparative study is performed to identify the most effective and efficient way of applying local search within the hybrid metaheuristic. It is expected to enhance the searching ability of the optimizer since the simulation model may get very complicated and it needs to be run under a large number of experimental conditions.

Table 3.1 gives a summary of the inventory problems studied in this dissertation research.

Table 3.1 Summary of the SC inventory problems studied

Inventory problem Practical factors considered Inventory policy Capacitated SC under

decentralized and centralized control

Supplier's capacity constraint (s, S) policy Integrated SC facing quality

imperfection

Quality imperfection in the

products supplied (s, S) policy Highly perishable platelets SC

under decentralized and centralized control

Restricted shelf life in days Order-up-to policy OIR Policy Blood SC with ABO compatibility Shortened shelf lives

ABO/Rh(D) compatibility OIR policy

The metaheuristic optimizer and its best local search application strategy will be discussed in detail in chapter 4. The resulting hybrid is chosen to as the optimizer in the MSO framework. In chapter 5-8, the successful application of the proposed MSO framework to different supply

chain inventory problems will be demonstrated in detail. In summary, the differences between the proposed MSO approach and theoretical/analytical approaches are summarized in Table 3.2.

Table 3.2 Comparison between the proposed approach and traditional analytical approaches Analytical approach MSO approach Assumptions on demand

distribution Required None

Guarantee on optimality Yes (on restricted cases) No (usually near-optimal) Explicit mathematical

models (cost equations) Yes No

Solution methods Exact, approximation, heuristic, meta-heuristic

Any meta-heuristic with a specially designed simulation

model Demand data that can be

handled

Only those satisfy the

assumption Any

Adaptability to practical