Workload prediction in the design of dynamic control systems: Applications to advanced manufacturing systems
Abstract
Human mental workload is one of the most important factors in cognitive task allocation of system design. It is reported that current task allocation methods with generally worded criteria are not very useful in aiding detailed decision-making in system design. While there have been many quantitative criteria available to determine the physical space in human-machine interaction, system designers need some explicit model and criteria to identify the mental workload imposed by the system, to predict human and system performance, to evaluate the alternatives of system design, and to design the system components. It is argued that the available models of human workload or performance are either too domain-dependent to apply to other system designs or subject-dependent which can not reflect the objective workload imposed by the system. The presented research first suggests a new cognitive task analysis method, which could be applied in the dynamic systems with task arrival uncertainty, and a general conceptual model of human workload prediction in system design. Based on the proposed conceptual model, an analytical model of human workload prediction is developed for advanced manufacturing systems. In the newly developed model, the human workload is represented by a set of system parameters, such as task arrival rate, task complexity, task uncertainty, and schedule tightness, which are considered as the main sources of human workload. In this context, human workload becomes an objective demand of systems on humans, which is independent of any subjective factors. Whether a specified individual or population is overloaded depends upon their workload threshold with respect to the specified task and environment. The analytical model is tested using twelve subjects in an interactive scheduling experiment. The tasks consisted of scheduling new arrived tasks, dealing with machine failure, and job expedition in a dynamic manufacturing system. The tasks are designed such that each subject is exposed to all the task load levels based on the combinations of design parameters. The conducted ANOVA and ANCOVA analyses demonstrate that human workload is very sensitive to the changes of system parameter levels. A regression analysis further shows that the human workload and system task load can be represented by these parameters. The experiment also validates the U-shape curve relations (not including the "underload" region) between workload and performance. This research provides a new approach to cognitive task analysis in a dynamic environment with random task arrivals. Further study is suggested on the human workload and task load models proposed in this study. It is hoped that these models could be used by system designers to predict the human workload imposed by systems after both laboratory and industrial validation.
Degree
Ph.D.
Advisors
Salvendy, Purdue University.
Subject Area
Industrial engineering
Off-Campus Purdue Users:
To access this dissertation, please log in to our
proxy server.