期刊
COMPUTERS & INDUSTRIAL ENGINEERING
卷 131, 期 -, 页码 542-551出版社
PERGAMON-ELSEVIER SCIENCE LTD
DOI: 10.1016/j.cie.2019.03.012
关键词
Mass customization; Learning curves; Clustering; Parameter selection
Mass Customization (MC) implies in large product variety and reduced production batch sizes. Tasks that rely on human ability are especially affected in this context as workers must quickly adapt to the requirements of new models. That adaptation process takes place differently in each worker, justifying the development of strategies to group workers with similar learning profiles. This paper proposes a framework to form homogeneous groups of workers based on their learning profiles by integrating learning curve (LC) modeling and cluster analysis. For that, performance data are collected and modeled through LCs, such that LC parameters quantify workers' adaptation to tasks. Principal Components Analysis (PCA) is applied to the dataset consisting of the LC parameters; PCA outputs give rise to an importance index that guides a backward parameter selection process. After each LC parameter is removed from the dataset, a new grouping using the Fuzzy C-Means (FCM) clustering technique is carried out using the remaining LC parameters, and the quality of the formed groups is assessed by means of three metrics. When applied to a shoe manufacturing process, 8 out of the original 29 LC parameters were retained, and used to insert workers into two clusters. The reduced subset of parameters elevated the quality of the clustering procedure by 29.4% (from 0.476 to 0.616) according to the Silhouette Index; similar improvements were indicated by the other two metrics. The retained parameters are related to workers' performance at the first repetition and previous experience, highlighting the importance of well-designed training programs.
作者
我是这篇论文的作者
点击您的名字以认领此论文并将其添加到您的个人资料中。
推荐
暂无数据