(465c) Optimal Supply Chain Redesign and Asset Management Using Genetic Algorithm

Multinational corporations continuously strive towards increasing their shareholder value or profit in today's competitive business environment. This requires that they optimally manage various assets in their supply chains and continually redesign their supply chains. These assets include, but are not limited to production and inventory holding facilities, raw materials, technological know-how, human resources, and financial assets. In addition to capital from loans and bonds, we consider contracts for material purchase/sale as financial assets also, because they help companies in hedging risk due to uncertainties, thus minimizing financial losses.

The demand for products and supply of raw materials are ever changing due to various reasons. New and huge markets are opening up in Asia and globalization and lifting and lifting of trade barriers are facilitating global growth. This is giving unprecedented opportunities for MNC to venture into new and unknown markets. To this end, they are either exploring collaborations with regional players, investing in new facilities, or are relocating existing facilities. Hence, there is a need for advanced decision support tools for optimal decision-making in redesigning and managing global supply chains.

While managing their assets, companies may invest in new assets and disinvest/relocate some of their existing assets, which essentially means, the redesign of their supply chains. In our previous work (Naraharisetti et al., 2005, Naraharisetti et al., 2006), we developed a novel MILP model to address various issues in supply chain redesign and asset management and used a branch and bound algorithm for solving this model. The various features that were considered include investments, disinvestments/relocations, regulatory factors, depreciation, production changeover costs, transportation costs, taxes etc. To the best of our knowledge, we were the first to address the issues of disinvestment and relocation, as relocation is simple a disinvestment at one location and investment at another.

In this paper, we will present an improved formulation of the above model, where we have reduced the number of constraints leading to a reduction in computation time. In addition, we will include contracts fir material purchase/sales allowing contract duration and quantity as decision variables. Furthermore, we will incorporate options for raising capital through loans and bonds. These issues of contracts, loans, and bonds, to our knowledge have not been considered so far in strategic asset management. Although, we have been able to solve small-scale problems in reasonable computational time, large scale problems pose significant challenges. As problem size increases, it becomes increasingly difficult t solve and MILP formulation by the branch and bound algorithm. Even for a moderate-size problem, computation times are of the order of days just to reach a relative gap of 10%. Hence, it is important to seek alternative optimization techniques. We will report on a genetic algorithm to solve this supply chain redesign and asset management problem and compare it efficiency with the branch and bound algorithm.

Genetic algorithm is a class of evolutionary algorithms, and is based on evolutionary biology. Genetic algorithm is initialized with a feasible set (population) of solutions (strings). Each of the strings has the decision variables and the corresponding objective value. The strings crossover and mutate to generate new strings, thus increasing the set/population size. The best strings, equal in number to the initial population are retained and the rest are discarded. In our model, the set of binary variables and the objective would form the string. The objective is obtained by a linear programming solver. The crossover and mutation of the strings is performed using the genetic algorithm. In this work, we will present case studies that are in similar size to real life problems, where there are several production facilities across different countries and even greater number of distribution centers around the world. We will also present the appropriate crossover and mutation strategies to obtain the acceptable set of solutions (final population) in reasonable computational time.

Keywords: capacity, production, planning, distribution, optimization

To whom correspondence should be addressed: Prof. I.A.Karimi, Tel: +65-6874-6359, Fax: +65-6779-1936 Email: cheiak@nus.edu.sg

References:

Naraharisetti, P. K., Karimi, I. A., Srinivasan, R. Location-Allocation-Production-Distribution in the Chemical Process Industries. INFORMS Annual Meeting, 2005, 13-16 Nov, San Francisco, USA.

Naraharisetti, P. K., Karimi, I. A., Srinivasan, R. Capacity Management in the Chemical Supply Chain. International Symposium on Advanced Control of Chemical Processes - ADCHEM, 2006, 2-5 April, Gramado, Brazil.