Data acquisition and physical parameter estimation in an asphalt production plant

In recent years, the research effort in developing condition monitoring systems has been consistently increasing. This thesis work has been developed through a collaboration between the ACTEMA research group at the University of Bologna and Marini-Fayat, a world leader in the production of asphalt plants, as part of a research project aimed at developing diagnostic algorithms for a single-dryer asphalt plant fluidic circuit, in order to estimate a key physical parameter for implementing condition monitoring. To achieve this, a model-based methodology was employed to develop, implement, and test a solution that uses the frequency response of the system identified by a Recursive Least Squares (RLS) algorithm to estimate a fluidic resistance using a Radial Basis Function (RBF) neural network. By monitoring deviations in this parameter, it is possible to detect potential incipient faults that may occur in the filter area, thus providing the possibility to prevent major failures. After being successfully tested, this algorithm has been implemented within a data acquisition infrastructure that collects a series of signals from various areas of the plant. The PLC system is designed to generate JSON files containing information about these signals and the estimated resistance, which are subsequently sent to an FTP server. In this thesis, the logic of the software and the main tasks are described in details, along with the principal state machines implemeted.