Increasing the availability and energy efficiency in thermal power plants and energy distribution systems by developing new diagnosis and early fault detection methods

Diagnosis and early fault detection have become of primary interest in the field of process automation since they constitute the basis for achieving failure tolerance, reliability, safety, availability and energy efficiency of complex systems. There are two basic approaches to the design of modern diagnosis and detection techniques. The first is based on the models of processes and is often inapplicable in case of the non-stationary nature of system parameters. The second approach is represented by the so called data-driven methods, which assume that the state of the observed object can be evaluated directly from the measurements obtained from the sensors by adaptively estimating their trends, statistics, distributions, correlations or spectral properties. The proposed project will concentrate on data obtained from simple, commercially available sensors of sound and vibration, which will then be used for health diagnosis and early fault detection in particular elements of systems for production and distribution of electrical energy. We will primarily focus on transformers, coal mills and feed-water pumps found in thermal power plants. The supervision of these devices is of major importance since the financial consequences of their sudden and unexpected fallout, while the means required for the realization of the appropriate supervisory systems are negligible when compared to, e.g. a 6 hour halt in the operation of a 300 MW power plant unit due to a failure of a feed pump.