The integration of design and construction processes remains, after decades of dedicated research, a great challenge. Even considering the specific context of pre-fabrication and modularization, it was just in recent years, with increasing adoption of Building Information Modeling (BIM) processes, that the challenge, albeit in a virtual environment, begins to be really addressed. With the advent of the Digitization phenomena in Construction, and the advances in Machine Learning techniques to cope with uncertainties of different natures in modelling real processes, it seems that the use of computational tools to simulate off-site production should be reconsidered. In this article, it is adopted an approach in viewing BIM as in a development stage to become an implementation of Product Lifecycle Management (PLM) for Construction. Towards this end, it is identified the lack of representation of the entire dynamics of production processes inside BIM models. The proposition of using Petri Nets with stochastic transitions to represent and simulate those processes are presented, altogether with the use of real RFID data, to adjust the model parameters, collected from a case study with a Brazilian company that pre-fabricate wood-framing houses. The probability distributions are derived based on the Mixture of Gaussians algorithm, and considers parameters of the design of wall panels ¬Ė so it could be used to extrapolated performance for new designs. Following the presented approach, it is expected that, with more data, the simulation process could be a good feedback to architects in evaluating the impact of its design options in production.