关键词: 学习—遗忘效应, 缝制车间, 生产调度, 非支配遗传算法, 多目标优化

Abstract:  With the deep integration of cutting-edge technologies, such as Internet of Things, artificial intelligence and 5G and manufacturing, the fourth industrial revolution, represented by intelligent manufacturing, is taking place. There is no exceptions of traditionally labor-intensive textile and garment industry, in which most value chain sections such as spinning and weaving are undergoing or have undergone profound changes. Nevertheless, in the sewing link of garment manufacturing and production, on the one hand, most manufacturers are forced by market demand to increasingly turn to the production mode of small batch and short delivery cycle; on the other hand, the sewing link itself involves many processes and changes quickly. In the foreseeable future, it is difficult for automatic machines to meet this highly flexible production environment, and manual density will remain an important feature of garment sewing production. At the same time, China's textile and garment industry is in a period of transfer and change, with rising labor costs and high turnover of front-line production employees, resulting in more complexity and uncertainty. In order to adapt to the general trend of manufacturing transformation and upgrading, and adapt to the complexity of garment sewing production, it is one of the potential directions to study a new management scheduling method that takes into account workers' cognitive and learning differences.
We proposed a hybrid flow batch scheduling model including employees' learning and forgetting effects for garment sewing workshops. We used minimizing makespan and minimizing idle mean square error as the objective function, and selected the non-dominated genetic algorithm as the solution tool, aiming to realize "worker-process-station" fine-grained scheduling optimization. According to the model, we carried out an algorithm simulation experiment on a real garment factory data, selected the appropriate batch size as the experimental parameter, and compared two kinds of scheduling optimization models with and without learning and forgetting effects. The simulation results show that model considering the learning and forgetting effects is more suitable for the production environment under heterogeneous scenarios of workers than not considering, which verifies the effectiveness of the model and algorithm.
We introduce the worker learning and forgetting factor matrix to optimize scheduling design for increasingly uncertain production environment. As for how to reasonably determine each worker's learning and forgetting factors,  we can consider using the real-time big data of the MES system, ERP system and the Internet of Things system of the garment factory in the next step to build a dynamic model to fit and calculate factors such as the learning and forgetting rate of "employee-process-station", so as to achieve a real-time dynamic executable scheduling scheme

Key words: learning and forgetting effects, garment sewing workshop, production scheduling, NSGA-II, multi objective optimization