The Research On Project Completion Probability Characteristic Of PERT

Currently, in the implementation process of the plan network, all construction operations including the construction site and production activities are faithful to plan implementation creating conditions for the duration goal, so that the entire construction process can be maintained in good condition to achieve schedule goals, and now network planning techniques for duration evaluation are Critical Path Method (CPM) and Program Evaluation and Review Technique (PERT) [1]. Critical Path Method considers activities and time parameters as deterministic event, but with the progress of project, the actual construction schedule will continue to have a variety of changes due to many factors, so the activity duration is uncertain. PERT is used for the network analysis of random uncertainty, it considers activity duration as a radom variable, and predicts project completion probability based on activity assumption, critical path assumption and total duration assumption, but its results are offen lager, thus, a variety of improved researches are emerging. Today, Monte Carlo Simulation is widely used, it has tended to perfect in network computing research, which is based on PERT, can reflect the statistical law of activity duration for activities of different distribution and with an increase in the number of simulation, the accuracy will be with an increase.According to the problem existing in PERT assumption on total duration probability characteristic, in this paper, to identify project completion probability characteristic, I attempt to use interpolation and neural network to study probability characteristic of total duration based on the large amount of time series generated by Monte Carlo Simulation. Firstly, we get the probability statistics of total duration full use of time series generated by Monte Carlo Simulation; secondly, based on interpolation, I make use of procedures on function approximation to get probability characteristic curve of total duration, and then draw its revised probability characteristic curve by neural network. It is verified that completed probability calculated by neural network is smaller than by the PERT, indicating that the hypothesis of total duration obeying the normal distribution is feasible, however, there are some differences compared to the actual probability distribution shown by the actual simulation, and that completion probability analyzed by the empirical distribution obtained from actual simulation is more realistic.In the end, a large hydroelectric project case is presented to verify the effectiveness and scientificalness through the analysis of total duration probability characteristic and completed probability under the time limit.