Gradual Fault Research For Turbo Flow Section Based On Principal Analysis And Neural Network

Steam turbine flow section gradual fault is an important model of steam turbine. Once the fault appears, It will not only affects the safety but also affect the economy seriously, which will cause energy waste and economic losses. So, diagnosis for steam turbine flow section gradual fault ensuring the safe and reliable operation of the unit is meaningful.After years of development, steam turbine flow section gradual fault diagnosis has achieved fruitful results, but there are still some problems to be solved. First, steam turbine flow section gradual fault is complicated, thermodynamic parameters associate with each other very closely, which brings difficulties to the analysis of parameters. Secondly, for the continuous and peak shaving unit state, single fixed standard cannot be able to satisfy the need of Steam turbine flow section gradual fault diagnosis. In addition, fault diagnosis often only makes pattern recognition, which is short of the analysis of its influence, reasons and measures. Aiming at the above problems, this paper carries out the following work taking the steam turbine flow section as the research object:(1) Mechanism analysis of steam turbine flow section gradual fault. Based on the FMEA analysis and fault tree analysis, this paper studied the steam turbine flow section gradual fault causes and fault dynamic development process, and did a deep research on the fault mechanism. Form FMEA analysis table and fault tree of flow section, establish the corresponding relation among the fault nodes and fault location, fault effect, fault reason, to provide a theoretical basis for the fault identification and diagnosis.(2) Diagnosis information digital representation of flow section gradual fault. The lots of unit state monitoring points of Production field are the sources of monitorable and diagnosable data, this paper extracts fault features from these data and non-real-time monitoring fault information, then defines and expresses the diagnostic parameters according to the class, and proposes a quantify, digitize method for fuzzy fault symptoms. Obtain the benchmark parameter values. For effectively use the domain knowledge, the steam turbine flow section gradual fault model library and fault diagnosis knowledge base is also established.(3) Put forward a set of method combining principal component analysis method, the method of intelligent fault diagnosis, the fuzzy membership degree analysis and neural network. IN this method, we first determine the dynamic range of flow section gradual fault diagnosis parameters of turbo generator set. Then, make the fuzzy membership degree analysis, reliability analysis and the main element analysis, determine the main element and fault class. Finally, determine specific failure modes through the BP neural network. At the same time, it uses the knowledge base to solve the specific fault causes, fault location, fault effects and fault handling measures according to the rules prerequisite, and generates a fault diagnostic report. Verify the practicability of the method through the analysis of typical fault case.