although efforts were made to use only information
that is also available in preproduction planning. Third,
according to the study procedure, all real-life tasks were
assessed in advance of DHM simulations. Methodologically
it would have been better to alternate that
sequence because previous real-life assessments could
influence successive DHM assessments. Alternating the
sequence was not possible, however, because real-life
work sequences can change quickly due to the continuous
improvement process, so that simulations based
on an earlier planning status may not be comparable
anymore. Yet, to reduce memory biases, the DHM assessment
was carried out 2 months after the real-life
assessment so that the ergonomists could not exactly
remember their previous assessment results. Fourth,
the data quality may have been impaired by two practical
constraints. One problem is that, for cost reasons,
only two observers were included in the study. Both
were experienced ergonomics experts and in real-life
tasks they showed high agreements on AAWS risk ratings
(i.e., high inter-rater reliabilities). Differences in
assessment of DHM simulations can therefore mainly
be attributed to restrictions of the simulation itself.
Another problem with the reported data is that AAWS
risk assessment scores from both observers were not
normally distributed. Scores on action forces showed
high SD values, and scores on material handling were
only assigned for one of the 20 tasks. This problem
is mainly due to the different tasks assessed and re-
flects reality quite well (e.g., manual material-handling
tasks are indeed rather infrequent on today’s final assembly
lines). Nevertheless, mean scores, correlations,
and significance levels found in this study have to be
interpreted with caution. Finally, another limitation is
that a lot of effort was spent in preparing DHM simulations
to make them as realistic as possible. Roughly
250 minutes of work was necessary to produce 1 minute
of DHM process simulation. Differences from real-life
tasks may be bigger when DHM simulations are prepared
less accurately. Moreover, because in practice it
is normally not possible to spend that much effort on
DHM simulations, static scenes are used instead. This
is problematic from an ergonomics point of view because
a comprehensive risk assessment is only possible
when the entire assembly process (not just some static
parts of the process) can be analyzed.
Reducing simulation effort is therefore an important
topic for future research. Currently different
approaches are discussed for preparing accurate
DHM simulations more efficiently. One approach is to