The Last Planner™ system of production control, based on lean construction principles, has been broadly and successfully implemented in several projects over the past years. Its focus on work plan realization is useful in helping project management teams identify main problems that constrain the timely completion of individual activities, and decrease possible variability. However, potential problems in designs of production processes, which also contribute to risks and variability during their implementations, have been seldom studied and could be only learned and used by human planners in a subjective and implicit manner. In this paper, a research approach is detailed to address this problem by creating a generic and concise data representation for networks of construction production processes in support of graphical analysis and pattern recognition. As a part of this ongoing research, a case study is presented with preliminary results, which were obtained by applying the research approach on a Last Planner™ database of production control from a large capital facility project. Networks of production processes were analyzed by comparing type descriptions of the original plans and their actual performance. Interesting and statistically valid patterns were recognized in this study, such as correlations between the topology of a work plan and its probability of having non-completions during implementation. Such objective and explicit patterns could help project managers better understand potential problems in original designs of construction processes, and make informed decisions to decrease corresponding variability and increase reliability in planning and control.
Graphical analysis, production processes, pattern recognition, knowledge discovery