On-line estimation for predicting the “Conching Degree” (CD) and the viscosity of chocolate

The conching process is one of the main stages of chocolate manufacture involving the development of sensory attributes and the product rheological properties. At the current operation of the conching process, the on-line estimation of quality properties that define an optimal development of the chocolate is not possible. In that sense, the on-line estimation of the “Conching Degree” (CD) as a flavor quality indicator and the chocolate viscosity as a textural quality attribute are fundamental tasks for guaranteeing process monitoring and real-time decision making. To this aim, this paper proposes a Moving Horizon Estimator (MHE) for the on-line estimation, instead of measurement, of the CD and the viscosity of chocolate. In this formulation, only available physical measurements of process variables easy to obtain are required, avoiding time-consuming and extensive conventional laboratory methods to quantify these chocolate's quality indicators. The estimator is designed as an optimization problem over a moving horizon, constrained to a Phenomenological Based Semi-Physical Model (PBSM) describing the conching process main phenomena. The performance of the proposed estimator is evaluated by simulation with data from a real plant. Finally, the proposed optimization-based estimation structure allows estimating the CD and the viscosity of chocolate with estimation indices errors lower than 6% considering only two available measurements in the process.