Research Of Gear Early Fault Diagnosis Based On Alternating Evolutionary Algorithm

The gear transmission system has been widely used in the aerospace, manufacture machines, national defense devices et al as its various advantages, such as fixed transmission ratio, large transmission torque, compact structure and high transmission efficiency. Gears usually run at high speed and overload conditions, coupled with hot and humid environment, gears susceptible to injury, and thus lead to gear failure, in result not only affect the transmission accuracy, even lead to catastrophic accident. Take measures before failure happens or as soon as incipient failure happens can effectively reduce operational casualties and economic losses. From security and economic point of view, research on gear early fault diagnosis methods based on evolutionary algorithm to improve the reliability of machinery and equipment have theoretical and practical significance.As the adaptive filter based on gradient search algorithm easily converged to local optimal, the introduction of biological evolutionary algorithm was given, aiming at improving the convergence speed of traditional evolutionary algorithm, a new search ideal that find local optimal after confirm the global peak was proposed. Based on this thought, alternating evolutionary algorithm is proposed, and applied alternating evolutionary algorithm to mean square error performance faces with unimodal and bimodal peaks. Then the alternating evolutionary adaptive noise cancellation algorithm was proposed by combining the alternating evolutionary and adaptive filter algorithm, its noise cancellation performance were analyzed. The experimental result verified that alternating evolution adaptive noise cancellation algorithm has a better noise elimination performance. The main work is summarized as follows:① After analysis characteristic of adaptive algorithm based on gradient search algorithm, a new search ideal that find local optimal after confirm the global peak was proposed. Based on this thought, alternating evolutionary algorithm is proposed.② The mean square error performance faces with unimodal and bimodal peaks were constructed, combined with system identification case, the comparison analysis when applied the search algorithm based on gradient algorithm and the search algorithm based on alternating evolutionary algorithm respectively to system identification was carried out, the results show that the alternating evolutionary algorithm can find out the global optimal when the mean square error performance face has multiple extremum values.③ Combined alternating evolutionary algorithm and adaptive noise cancellation algorithm, the alternating evolutionary adaptive noise cancellation algorithm was proposed. Applied proposed algorithm to gear fault simulation signals, the results shows that the algorithm has least computation and fastest convergence rate.④ The analysis of cloning coefficient, mating coefficient and signal to noise ratio effect on the noise cancellation performance of the proposed algorithm were carried out. The results show that too large or too little cloning coefficient will lead to a slow convergence speed, and the proposed algorithm can not achieve noise cancellation when the signal to noise ratio was smaller than-70 d B.⑤ The gear fault test was designed, simulate the various degrees of gear spalling fault through electric discharge machining method. According to the structure of test rig, select the test points and develop the test conditions table. Then finish the gear experiment and collect the vibration signals.⑥ The gear fault diagnosis and recognition method based on alternating evolutionary noise cancellation algorithm was proposed. This proposed algorithm is validated experimentally through the gear box test. The results show that the proposed method has an excellent performance in the extract the weak fault charatersitc signal caused by gear incipient spalling fault.