Study On Fault Diagnosis Based On Chaos

Fault diagnosis is an intercross science. It was applied to find the fault according to the information of the running equipments or system, and then decide the corresponding methods. Fault diagnosis can ensure the safety of the running equipments. Since 1960's fault diagnosis has been widely used in many fields such as military affairs, aviation, metallurgy, the petrochemical industry, electronic. With the development of science, the structure of diagnosed system become more and more complicated and the system has stronger non-linear character. Besides the exterior characteristic caused by system fault is more complex. So the existing methods of fault diagnosis could not solve these problems effectively. It is important to improve the technique of the fault diagnosisFrom the methods of fault diagnosis, we have studied each method of fault diagnosis, problem as well as the development direction. Meanwhile applied chaos theory in the fault diagnosis, studied the algorithm of embedding dimension and delay time. This algorithm makes use of the zero of non-bias multiple autocorrelation function of the chaotic time series to determine the time delay, which efficiently reduces the computing error caused by tracing the slope varying of average displacement (AD) arbitrarily AD algorithm. Thereafter, by means of the iterative algorithm of multiple autocorrelation andΓ= test, the near-optimum parameters of embedding dimension and delay time are estimated.According to the blindness of initiatory numerical value, propose an improved algorithm based on embedding dimension delay time of geometry invariants. This algorithm combines the old algorithm with the Geometry invariants algorithms. By selecting appropriate approximate initial value to process the phase space reconstruction of Chaos system, and decrease the algorithm's iterative number, accelerate the algorithm's convergence speed.Finally the improved algorithm is applied to the steam turbine's electric generator model, and the result of simulation proved the validity of the improved algorithm.