Robust tuning of bending-points detection algorithms in batch-operated processes: Application to Autothermal Thermophilic Aerobic Digesters

Abstract

The literature shows a diversity of real-time algorithms for automatic detection of bending-points in batch-operated waste treatment systems. In this study a new methodology is proposed for tuning the parameters of these algorithms when uncertainty specifications are considered at the outset. In this method the effects of slow and fast dynamic responses on the shape of signal trajectories were treated separately in order to cover via simulation all possible operating scenarios for a real situation. Long-term uncertainty and steady-state simulations were combined to derive probability distributions for biomasses. These probability distributions were then merged with short-term uncertainty to run one-cycle random simulations with which to reproduce the entire diversity of signal trajectories. Finally, an optimisation problem was formulated in terms of the algorithm parameters. The methodology was satisfactorily applied to tune an algorithm for detection of bending-points in an Autothermal Thermophilic Aerobic Digestion (ATAD) process.