This paper addresses a solution to state and clinker hardness estimation in industrial cement mill process. A Takagi-Sugeno model with unmeasurable premise variables is developed for the nonlinear model of a cement mill. Bsed on this model, a nonlinear observer is proposed in order to estimate the state variables and also the clinker hardness, which is an unknown input of the process. The convergence of the estimation error is studied using the Lyapunov theory and the input-to-state stability (ISS) approach. An optimization problem with LMI constraints is then provided for the synthesis of this observer. Finally, simulation results and some discussions about the robustness of the observer are given.