As machine tools become faster and more precise, thermal simulation is becoming increasingly important in spindle design and in understanding the operating conditions. Thermal simulation requires appropriate thermal boundary conditions. To set boundary conditions, the relationship between the spindle temperature distribution and parameters of bearing such as thermal contact conductance, heat generation and preload is particularly important. However, it is difficult to directly measure those boundary conditions during spindle operation. Although there have been studies on thermal boundary conditions of bearings, there have been few studies on those in real operating conditions. In this study, the thermal boundary conditions of a rotating bearing were estimated by the particle filter which is one of data assimilation methods. The boundary conditions were estimated by iterative finite element thermal simulations and comparison of the simulated results with the experimental temperature distribution measured by a thermal camera. The maximum error in temperature simulation with the obtained boundary conditions was 0.5 K. The obtained thermal boundary conditions also indicated the heat transfer and heat generation contribution of the inner and outer races of the bearing.