Predicting Manufacturing Aptitude using Augmented Reality Work Instructions
(Room S320C)
30 Nov 17
9:00 AM
-
9:30 AM
The complexity of manufactured equipment for the U.S. military has increased substantially over the past decade. As more complex technology is integrated into battlefield equipment, it is more important than ever that workers manufacturing this equipment have the necessary skills. These specialized manufacturing skills require careful workforce selection and training. However, traditionally, workers are assigned roles based on instructor evaluation and qualitative self-assessments. Unfortunately, these assessments provide limited detail about a candidate’s aptitude. By using more detailed data captured from assembly operations, a more complete profile of an operator’s skills can be developed. This profile can then guide assignment of a worker to maximize productivity. This paper develops a Bayesian Network (BN) to predict worker performance using data captured from 75 participants via augmented reality guided assembly instructions. Information collected included step completion times, spatial abilities, and time spent on different assembly operations. For analysis, participant data was divided into training and testing sets. The data was mined for trends that could statistically predict measures of performance like errors or completion time. Based on these trends, the training set was used to construct the BN. The authors found that the model could predict some aspects of performance accurately, such as assembly completion time in the testing set. While these results were encouraging, further analysis demonstrated the network was biased by probabilities that were greatly influenced by the number of data points present in a category. The results highlight that, with small data sets, there is often not enough observed evidence to produce accurate predictions with BN. This suggests that a method of data simulation or generation is required to increase the number of training set samples. This would enable powerful BN tools to be used in real world manufacturing applicat