Study On Fault Diagnosis System For Steam Turbine Based On BP Neural Network

Steam turbine is large rotating machinery with complex structure. The operatingenvironment of steam turbine is often very complex.Thus unpredictable fault often occurs inthe operating state. When a fault occurs, the type of the fault should be quickly found to avoidfurther harm.. The purpose of this paper is to develop a system with online vibrationmonitoring,real-time data acquisition and display module, system database module, dataanalysis module, historical data module, fault diagnosis module for the particular steamturbine.Wavelet analysis can analyze the signal in time domain and frequency domain at thesame time, which has advantages compared with the traditional signal analysis. With thetechnology of wavelet analysis, we find that the fault of signal characteristics and the type offault have an identified mapping. The fault characteristic must be analyzed to establish themathematical relationship between fault characteristic and fault types.BP neural network has a strong nonlinear mapping ability, so we can get mathematicalmappings between the feature vector and the fault with this ability. Feature vector extractedwith wavelet packet and fault types are input and output of the network. Input layer,intermediate layer, output layer and the target accuracy were designed. The traditional BPalgorithm is improved, and the optimal algorithm was got by comparing iteration step anditerative learning time of completing the training. By training optimal neural network, we canget the neural network meeting the accuracy requirements. The nonlinear mapping betweenfeature vectors and fault types came true.The software and hardware were designed according to the requirements of the entiresystem.NI PXI IPC, sensors and dynamic data acquisition board are the core of the hardwaresystem. The software system was programmed in LabVIEW and MATLAB. The real-timedata acquisition and display module, system database module, data analysis module, historicaldata module was programmed in LabVIEW alone, and the function of fault diagnosis modulewill take advantage of BP neural network embedded into the system framework by MATLAB Script node in LabVIEW to realize. The hardware and software system was tested for a periodunder the operation of the steam turbine. The test showed that real-time monitoring system isstable.The system has online monitoring and fault diagnosis function and can meet theoperational requirements of steam turbine.