INTRODUCTION As more and more companies in China put energy into optimizing their supply chain/logistics operations, computer modeling has become an integral part of analysis and decision making. However, confusion remains concerning the availability of tools, what they can do, as well as when and why they should be selected. This article will summarize such issues and 2 real-life solutions will be provided.
3 “SOLVER” METHODS The three types of “solver” technologies most commonly encountered when dealing with logistics models are heuristics, optimization and simulation. Usually, all of the three techniques are used when working on logistics problems.
A heuristic is a process for solving a problem that generally relies on logic or intuition. Example: find shortest tour problem manually Advantages: ü Usually get good solutions quickly ü Easy to learn ü Usually have intuitive value ü Not require specialized software Disadvantages: ü Usually only solve deterministic problems ü Might have to ignore complicating factors ü No way of knowing if you have the best solution ü No way of knowing how good your solution is
Optimization approaches rely on mathematical formulations and result in a “provably optimal” solution. Example: enumerate all possible logistics network patterns and pick the best one Advantages: ü You know how good your solution is ü Easy to perform sensitivity analysis to see how robust your solution is to changes in the data ü Models can get quite complex ü Non-intuitive solutions get considered Disadvantages: ü Only solve deterministic problems ü Usually require specialized software ü Require expertise to develop ü Often seen as a “black box” ü Might have to make simplifying assumptions A simulation Involves building an experimental model of a system and evaluating alternatives in test runs. Example: Simulate a single ATM machine working environment and see if the specific customer service level objective can be reached Advantages: ü Can include uncertainty ü Can examine complex relationships ü Software provides nice interfaces ü It is easily accepted Disadvantages: ü Need to collect data on random distribution ü Requires specialized software ü Dangerous in the Normal distribution ü No way of know if you have the best solution ü No way of knowing how good your solution is
SOLUTIONS A. Optimization Solution A powerful network computer modeling system that tests changes in shipping mode, plant and distribution locations, costs and service parameters quickly and economically. Issues could be addressed as followed: ü Where to locate plants, DC’s and other items ü Which suppliers and transportation modes should I choose ü How much to produce, to ship and to stock ü How many plants, DC’s and suppliers I need ü Which product should be assigned to which facility ü Which customer demand zone should be assigned to which facility In most cases, the total costs of each selected alternative will be brought out and compared to give an optimal solution of logistics network design.
B. Simulation Solution Simulation can be used to help customers to design a new Customer Distribution Centre (CDC), for example. A dynamic simulation model of a planned CDC can be created in which the customer can simulate a number of various scenarios (with/without sorting machine, with/without automatic crane, different batching strategies, etc). The results from the simulations can help companies make decisions pertaining to the size of areas, sorting machine, location of articles, etc. The model can also be used to analyse changes in articles, customer orders and working schedules, as well as to find out what is the cost to handle one single customer order (for better pricing strategy).
CONCLUSION Each method has its pros and cons, as well as its uses. Heuristics are often used to start the optimization process and embedded in an optimization algorithm to hasten the process. After optimization or a heuristic finds a good solution, simulation evaluates it under “real world” conditions. In most cases, the optimization method will be used in solving more strategic problems because uncertainty could be overlooked in a big picture. However, simulation methods will be highly appropriate in dealing with tactical or/and operational problems, due to the fact that day-to-day tasks involve many uncertainties. | 
