Unsupervised Learning-based Artificial Bee Colony for minimizing non-value-adding operations

Advanced analytics benefits lean manufacturing by upgrading the scheduling problems into operational strategic tools that help minimize non-value-adding activities. Considering production environments with prevalent setup operations, this study develops an Unsupervised Learning-based Artificial Bee Colony (ULABC) algorithm to improve the effectiveness of minimizing idle times in unrelated parallel machine production settings. For this purpose, the k-means method is integrated into the approximation algorithm to address sequence-dependent setup operations. An exemplary case from the forging industry is provided to evaluate the performance of the ULABC algorithm. Reducing setup times through effective job clustering by the learning mechanism, it is shown that the solution quality is significantly improved in large-scale benchmark tests with 16 and 24 percentages of reduction in the makespan value of instances requiring short and long setup operations, respectively. The statistical analysis confirms the significance of the resulting improvements. This improvement is expected to be even more substantial when very-large industry-scale problems are solved. Overall, this study narrows the gap between scheduling theory and modern industrial applications through applications of advanced analytics in the production management context. (C) 2021 Elsevier B.V. All rights reserved.

Automated summary

This study utilizes an Unsupervised Learning-based Artificial Bee Colony algorithm to minimize non-value-adding activities in production settings with prevalent setup operations, improving solution quality by reducing setup times through a learning mechanism. The gap between scheduling theory and modern industrial applications is narrowed through the application of advanced analytics in production management context.