Research On Open Fault Diagnosis Framework And Dynamic Meseaurement Anlysis Methods

With the development of equipment integration and automation in large-scale industry, safely and faultless operating of equipment is the first-line prerequisite to provide assured safety of both human and property. Enterprises do their best to ensure safely and faultless operating of equipment for decreasing production cost. Because of many data type,many data quantity and strong speciality of machinery equipment condition monitoring and fault diagnosis, this dissertion puts forward the integrated idea of on-line and off-line monitoring and establishes open fault diagnosis framework based on NDSS, which can break through the limit of space and time, make good use of limited expert resources and realize management united and effective. Distributed frontend of open fault diagnosis platform—embedded dynamic signal measurement and analysis system relies on modern computer and electronics development. Digital signal processor and Microsoft XP Embedded have been applied in machinery fault diagnosis successfully. Make good use of DSPLIB working at the frequency—100MHz, customize subminiature operating system kernel with XPE, and reduce consumedly dependency on hardware and software. Enhanced Write Filter (EWF) based on REG-OVERLAY is applied in order to protect the system and improve its independability. Mechanical equipment with occasional and unexpected accidents needs continous condition monitoring and analysis. This dissertion puts forward to continous sampling and reconstruction technique for time domain signal. Data acquisition card adopts segmented work module of annular memory. Establish two-dimentional buffer queue in user buffer space to provent data from stopping up. Sampling software, saving and processing data, concurrently works with multithread. So this system can realize continous monitoring with no sampled data loss and reconstruct the needed data off-line, which is a renovation for portable fault diagnosis instrument. Empirical mode decomposition (EMD) method is a new tool for the nonlinear and non-stationary signals analysis, and can adaptively decompose signals into several intrinsic mode functions (IMFs) according to its characteristic time scale. Continous sampling technique based on annular memory for time domain signal is applied to obtain the original signal at two ends to constrain the end distortions. This method has greatly improved the decomposition precision at the ends. Because mechanical equipment is most nonlinear, the signals from condition monitoring are mostly nonlinear and nonsteady. In project application, the signals are often submerged in heavy noise and are difficult to be picked up. So adaptive swept stochastic resonance algorithm is proposed on the basis of scale transformation stochastic resonance in large parameter. This algorithm realizes obtaining weak signal from heavy noise by auto-modifying acquisition frequency and structure parameter of a bistable system. And the experiment result is satisfactory.