Methodology for the Screening of Signal Analysis Methods for Selective Detection of Hydrodynamic Changes in Fluidized Bed Systems

Various multiphase reactor systems need to be monitored, e.g., for increasing efficiency or for avoiding catastrophic events such as defluidization, excessive foaming, and flooding. In addition to the commonly available average process variables, pressure fluctuation signals measured at a high sampling frequency have been shown to contain relevant information about the process state. There are many different analysis techniques available that can be used for the data analysis of such signals. However, only sufficiently selective methods will be suitable for an unambiguous detection of a specific change in the process, i.e., the identification of the cause. In this paper, a new methodology is presented for screening many different signal analysis methods in combination with various signal pretreatment methods with the goal to find those combinations that are selective toward a specific process change. This methodology can generally be applied for any specific process change; here we focus on the detection of agglomeration in fluidized beds. The presented methodology is illustrated with some fluidized bed data sets, demonstrating the validity and the benefit of this approach.