RT Dissertation/Thesis
T1 Application of nature-inspired optimization algorithms to improve the production efficiency of small and medium-sized bakeries
A1 Babor,Md Majharul Islam
WP 2023/06/13
AB Increasing production efficiency through schedule optimization is one of the most influential topics in operations research that contributes to decision-making process. It is the concept of allocating tasks among available resources within the constraints of any manufacturing facility in order to minimize costs. It is carried out by a model that resembles real-world task distribution with variables and relevant constraints in order to complete a planned production. In addition to a model, an optimizer is required to assist in evaluating and improving the task allocation procedure in order to maximize overall production efficiency. The entire procedure is usually carried out on a computer, where these two distinct segments combine to form a solution framework for production planning and support decision-making in various manufacturing industries. Small and medium-sized bakeries lack access to cutting-edge tools, and most of their production schedules are based on personal experience. This makes a significant difference in production costs when compared to the large bakeries, as evidenced by their market dominance.
In this study, a hybrid no-wait flow shop model is proposed to produce a production schedule based on actual data, featuring the constraints of the production environment in small and medium-sized bakeries. Several single-objective and multi-objective nature-inspired optimization algorithms were implemented to find efficient production schedules. While makespan is the most widely used quality criterion of production efficiency because it dominates production costs, high oven idle time in bakeries also wastes energy. Combining these quality criteria allows for additional cost reduction due to energy savings as well as shorter production time. Therefore, to obtain the efficient production plan, makespan and oven idle time were included in the objectives of optimization.
To find the optimal production planning for an existing production line, particle swarm optimization, simulated annealing, and the Nawaz-Enscore-Ham algorithms were used. The weighting factor method was used to combine two objectives into a single objective. The classical optimization algorithms were found to be good enough at finding optimal schedules in a reasonable amount of time, reducing makespan by 29 % and oven idle time by 8 % of one of the analyzed production datasets. Nonetheless, the algorithms convergence was found to be poor, with a lower probability of obtaining the best or nearly the best result. In contrast, a modified particle swarm optimization (MPSO) proposed in this study demonstrated significant improvement in convergence with a higher probability of obtaining better results. 
To obtain trade-offs between two objectives, state-of-the-art multi-objective optimization algorithms, non-dominated sorting genetic algorithm (NSGA-II), strength Pareto evolutionary algorithm, generalized differential evolution, improved multi-objective particle swarm optimization (OMOPSO) and speed-constrained multi-objective particle swarm optimization (SMPSO) were implemented. Optimization algorithms provided efficient production planning with up to a 12 % reduction in makespan and a 26 % reduction in oven idle time based on data from different production days. The performance comparison revealed a significant difference between these multi-objective optimization algorithms, with NSGA-II performing best and OMOPSO and SMPSO performing worst.
Proofing is a key processing stage that contributes to the quality of the final product by developing flavor and fluffiness texture in bread. However, the duration of proofing is uncertain due to the complex interaction of multiple parameters: yeast condition, temperature in the proofing chamber, and chemical composition of flour. Due to the uncertainty of proofing time, a production plan optimized with the shortest makespan can be significantly inefficient. The computational results show that the schedules with the shortest and nearly shortest makespan have a significant (up to 18 %) increase in makespan due to proofing time deviation from expected duration. In this thesis,  a method for developing resilient production planning that takes into account uncertain proofing time is proposed, so that even if the deviation in proofing time is extreme, the fluctuation in makespan is minimal. The experimental results with a production dataset revealed a proactive production plan, with only 5 minutes longer than the shortest makespan, but only 21 min fluctuating in makespan due to varying the proofing time from -10 % to +10 % of actual proofing time.
This study proposed a common framework for small and medium-sized bakeries to improve their production efficiency in three steps: collecting production data, simulating production planning with the hybrid no-wait flow shop model, and running the optimization algorithm. The study suggests to use MPSO for solving single objective optimization problem and NSGA-II for multi-objective optimization problem. Based on real bakery production data, the results revealed that existing plans were significantly inefficient and could be optimized in a reasonable computational time using a robust optimization algorithm. Implementing such a framework in small and medium-sized bakery manufacturing operations could help to achieve an efficient and resilient production system.

K1 Bäckerei
K1 Produktion
K1 Optimierung
PP Hohenheim
PB Kommunikations-, Informations- und Medienzentrum der Universität Hohenheim
UL http://opus.uni-hohenheim.de/volltexte/2023/2166