Industry 4.0 paradigm calls for responsive scheduling and production control systems. Feedback loops incorporated to these systems allow decision making updates based on the real-time shop floor state. The bond graph approach enables the simulation of production control with feedback loops, but previous research with this theme is scarce. In the existing models, the buffers are modeled with infinite capacity, causing the effort variable to become zero and introducing a non-linear function that cramps the design of controllers. We propose modifications to an existing model in order to overcome this limitation, and simulate a 4-station shop floor as a proof of concept. With an optimal control approach, the rate of the material sources could be adjusted to keep the work in process at desired levels. The approach contributes to the literature with a model with more realistic assumption (limited buffers) and with the design of responsive production systems, addressing a research thread that is relevant to recent digitization trends.