A Novel Method Based on Artificial Neural Network to Production Planning: a case study of a paints producer

Production planning is a challenging problem in the field of management science. It involves a wide set of decisions to be taken on different time ranges (long term, medium term or short term) which depends on the specific manufacturing system. Traditional mathematical models have been shown to be too restrictive in real situations characterized by uncertain and non-stationary demand. The paper shows that the Artificial Neural Network (ANN) systems are suitable for solving production planning problems thanks to their capability to adapt to the context. The literature contributions on ANN-based planning are usually applied to very specific aspects of the production planning, often involving assumptions which makes the model different from reality. The systems proposed in this paper involve instead the whole planning activity on medium-long term horizon and take into account essential features that are usually ignored, such as the importance of a product for the business strategy. In particular, two ANN-based systems are proposed, a static structure and a dynamic one, which are applied to a real production planning case: a paints and varnishes producer with a make-to-stock production system based on batch production mode. The developed ANNs provide good results in planning the activity on medium and long time horizons. Furthermore, the paper proves that the limited availability of data can be successfully faced by acting on the input parameters, on the one hand, and by developing appropriate scenarios on the other one.