Partially Premixed Combustion has shown to be a promising advanced combustion mode for future engines in terms of efficiency and emission levels. The combustion timing should be suitably phased to realize high efficiency. However, a simple map-based feed-forward control method is not sufficient for controlling the combustion during transient operation. This article proposes one learning-based model predictive control (MPC) approach to achieve controllability and feasibility. Since PPC engines could have unacceptably high pressure-rise rates at different operation points, triple injection is applied as a solvent, with the use of two pilot fuel injection. The controller utilizes the main injection timing to manage the combustion timing, and the first and second injection timing is considered as a function of the engine load and speed. The cylinder pressure is used as the combustion feedback. The method is validated in a multi-cylinder diesel heavy-duty engine with low-pressure exhaust gas recirculation (EGR) and variable geometry turbine (VGT) for transient control.