Faculty of Informatics, Autonomous University of Sinaloa, Sinaloa, México.
Department of Industrial Engineering, University of Sonora, Sonora, México.
Ergonomics. 2021 Aug;64(8):1018-1034. doi: 10.1080/00140139.2021.1892832. Epub 2021 Mar 12.
ABSTRCTErgonomic workstation design is crucial to prevent work-related musculoskeletal disorders. Many researchers have proposed multivariate analysis for human accommodation. However, no multivariate anthropometric analysis exists for the Mexican population. This study compares common multivariate human accommodation approaches (e.g. principal component and archetypal analyses) and clustering techniques (e.g. -means and Ward's algorithm) with the classical percentile-based univariate accommodation approach, using the Chi-squared goodness-of-fit test and the McNemar's test. The theoretical accommodation percentage obtained by multivariate approaches was higher than those obtained by the percentile univariate approach considering the central 98% data. -means and archetypal analysis obtained similar and the highest accommodation values, followed by Ward's algorithm and principal component analysis. The study findings can be deployed to assess the design of workstations in Mexico, such as electronic components assembly and crew designs, and the effects of different anthropometric measurements in human accommodation. Products and workplaces design are commonly based on the classical univariate approach, using the extreme percentiles. In this study, multivariate approaches were tested on dimensions for sitting workstations, and results showed a bigger accommodation level in comparison to the univariate 1%-99% approaches. RHM: representative human model; DHM: digital human model; PCA: principal component analysis; AA: archetypal analysis (AA); PCs: principal components; FA: factor analysis; RSS: residual sum of squares; SSE: sum of squared estimated errors; WA: Ward's algorithm; DBI: Davies-Bouldin index; CHI: Calinski-Harabaz index; SI: silhouette index; SH: sitting height; EHS: eye height, sitting; AHS: acromial height, sitting; PH: popliteal height; KHS: knee height, sitting; BPL: buttock-popliteal length; BKL: buttock-knee length; FGR: functional grip reach; AD: anthropometric dimension; E: expected; A: achieved.
工效学工作站设计对于预防与工作相关的肌肉骨骼疾病至关重要。许多研究人员已经提出了多种分析方法来适应人体。然而，针对墨西哥人口，尚未存在多维人体适应分析。本研究比较了常见的多维人体适应方法（例如主成分和原型分析）和聚类技术（例如 - 均值和 Ward 算法）与基于百分位数的传统单变量适应方法，使用卡方拟合优度检验和 McNemar 检验。考虑到中央 98%的数据，多维方法获得的理论适应百分比高于基于百分位数的单变量方法。 - 均值和原型分析获得了相似且最高的适应值，其次是 Ward 算法和主成分分析。研究结果可用于评估墨西哥工作站的设计，例如电子元件装配和机组设计，以及不同人体测量在人体适应中的影响。产品和工作场所设计通常基于经典的单变量方法，使用极值。在本研究中，对坐姿工作站的尺寸进行了多维方法的测试，结果表明与单变量 1%-99%方法相比，适应水平更大。 RHM：代表性人体模型；DHM：数字人体模型；PCA：主成分分析；AA：原型分析（AA）；PCs：主成分；FA：因子分析；RSS：残差平方和；SSE：估计误差平方和；WA：Ward 算法；DBI：Davies-Bouldin 指数；CHI：Calinski-Harabaz 指数；SI：轮廓指数；SH：坐姿身高；EHS：坐姿眼高；AHS：坐姿肩峰高；PH：坐姿腘高；KHS：坐姿膝高；BPL：臀-腘长；BKL：臀-膝长；FGR：功能握距；AD：人体测量尺寸；E：预期；A：实现。
