Abstract:
The research described in this thesis is concerned with human performance modelling
as an aid in the
process of manufacturing systems design and re-design. Most
manufacturing systems are highly complex constructs and their behaviour is of a
dynamic and stochastic nature. They have to be constantly designed and re-designed
as
organisations are continually being pressured to change their manufacturing
facilities, technologies, methods, people and products.
All
design methods have some form of evaluation where discrete event simulation
models are
usually used to undertake a comparative analysis of different system
designs. Within these discrete event simulation models it is common practice to
represent workers as simple resources, often using deterministic performance values.
Conversely, the work measurement literature indicates that worker task performance
varies between different workers carrying out the same task and moreover for the
same worker when
repeating a task. The current approach of representing workers
within discrete event simulation models ignores the potentially large effect that human
performance Variation can have on system performance. This omission affects in
particular simulation models of labour intensive manufacturing systems like manual
assembly flow lines. It appears that this adds to the inaccuracy of the simulation
model
output and that consequently the simulation model does not react the
behaviour of a real
system in an appropriate way.
A research
programme has been designed to investigate these issues. First, a long term
data collection exercise has been conducted to quantify the performance Variation of
workers in a
typical automotive manual assembly flow line. The data have then been
used in form of
frequency distributions to represent worker performance Variation at
individual Workstations within manual assembly line simulation models. Through
designed simulation experiments the impact that this form of worker performance
Variation
representation has on the accuracy of manual assembly line model behaviour
has been
investigated.
Overall this research has found that adding worker performance Variation models into
manual
assembly flow line models has an impact on the accuracy of these simulation
models. The
magnitude of the impact depends very much on the type of Variation to
be
represented as well as on the system to be modelled. This evidence is an important
result to
support justification for further research in this area. For a more sophisticated
approach of modelling worker performance Computational Organisation Theory using
the
multi-agent paradigm has been identified as the most suitable way forward.