Machine Learning Algorithm to Solve Dynamic Jobshop Problem to Deal with Digitalization of Manufacturing Process—A Real-World Case Study

Volatility, uncertainty, complexity, and ambiguity (VUCA) started life as an acronym in the post-cold-war world of the US army. This concept is gaining new relevance to characterize the current environment and the leadership traits required to navigate it successfully. General belief is that companies and leaders who do not adequately prepare for this digital transformation cannot sustain. In this new era, cracking the VUCA code is key for success and rise of data science, and use of analytics by business to make data-driven decisions has come as a boon to the leaders. To cope with the volatility and digitalization, manufacturing firms need to formulate a good production plan based on customer demand based on their manufacturing capacity. In this paper, self-organizing maps (SOM), an unsupervised clustering algorithm, is used for clustering similar data from the dataset, and micro-regression models are derived from the same to develop a sustainable process. The past research in this area has focused on developing macro-models with factors related to job characteristics, shop characteristics, and order characteristics. Although they have been generally successful, a single macro-model may not allow an understanding of how the system changes in state and which characteristics are important under different conditions. Hence, a new approach SOM-RA is proposed for prediction of manufacturing flowtime in a dynamic jobshop. It has been demonstrated that a set of micro-models associated with each state produces better prediction of the flowtime compared to macro-models. This study has been conducted using the representative data from a factory that manufactures cables and wires for armored vehicles. Open-source software was used to demonstrate the feasibility of using machine learning tools by small- and medium-sized enterprises for better decision-making. It has been shown that the proposed algorithm is superior based on the performance measures discussed in this paper.