Diagnosis of Fluid Systems utilizing Grobner Bases and Filtering of Consistency Relations

The objective of this master’s thesis is to develop a method for automatic design of diagnostic systems for pulp and paper industry. General models for fluid systems are presented. The models simulate slow dynamics based on object-oriented sub-models. The sub-models are tanks, pipes, control valves, pumps, mixers and dividers. The diagnostic systems are based on hypothesis tests and decision structures. The hypothesis tests are formed as tests with thresholds. The tests are constructed as normalized residuals. Consistency relations are used to construct the residuals. The consistency relations are extracted from model equations. Through calculations of Gröbner bases, good decision structures are achieved. To evaluate the methods, they are applied to a stock preparation and broke treatment system from a paper mill. Three different simulations including three different faults are presented. A semi-automatic algorithm, with the ability to construct diagnostic systems for arbitrary fluid systems, is presented. The diagnostic system, constructed with the semi-automatic algorithm, can detect and isolate faults. Arbitrary slow dynamic faults can be detected and isolated.