On-line COP (coefficient of performance) estimation is carried out for waste energy recovery heat transformer by water purification process using developed different algorithms by bash (information acquisition) and Octave (information training). This variable estimation is very important to an optimum performance in the system. The COP is estimated from observation a filter Kalman. In addition, a thermodynamic model is used to on-line COP prediction to steady-state conditions. The two models consider output measures temperatures for each one of the four components of the heat transformer (absorber, generator-evaporator, and condenser). However to thermodynamic model, two pressures (absorber-evaporator and generator-condenser) and LiBr + H2O concentrations in the absorber and generator are required in real-time. The pressures and concentrations are calculated from literature. The results are satisfactory for the COP estimation in both models, allowing to optimizer and know the system operation situation in each set measured. In addition this allows to reduce costs of production and a process control can be implemented from this indirect measure.