Research On Non-contact Safety Monitoring And Intelligent Diagnosis Of High-speed Blades Based On Blade Tip-timing Method

High-speed rotating blades form core mechanical components in turbomachines.Research concerning online monitoring of operating states of such blades has drawn increased attention in recent years.The blade tip timing technology(BTT)is considered to be the most promising owing to its advantages of low cost and non-contact utility.However,the BTT technology is still difficult to apply to practical application because of many scientific and technical problems.(1)How to accurately extract the BTT signal under the strong noise?(2)How to extract more information from the undersampled BTT?(3)How to carry out the vibration monitoring and fault diagnosis at variable speed? The thesis focuses on the above problems.Main work is shown as follows:1.The vibration mechanism and characteristics of high speed rotating blades are analyzed and summarized,including the vibration form,type,and vibration parameters,etc.Then the blade natural frequency and modal shape in various states are analyzed by using the finite element software ANSYS.2.The factors affecting the measurement accuracy of the BTT system are analyzed and discussed.An extracting method of blade tip-timing signals in strong noise based on EEMD is proposed.The BTT signal under the noise condition is accurately extracted,and the maximum relative error is only 1.8%.3.A sparsity adaptive reconstruction method is proposed for the under-sampled BTT signal.This method breaks through the technical bottleneck in the traditional method which needs the vibration information of the blade.The blade vibration signal is accurately reconstructed by the adaptive matching of the sparsity under the steady state condition.4.In view of the problem that the blade vibration is difficult to monitor under the variable speed,a blade vibration monitoring method based on multi reference phase is proposed.A plurality of reference phases were arranged as evenly as possible on the rotating shaft to determine the speed of rotation,and the measurement equations of vibration displacement were deduced and verified.5.A method for mining and diagnosis of blade fault information based on deep learning is proposed.Automatic learning and selection of undersampled signals by using the powerful feature learning and information mining ability of the convolution neural network.And the blade state can be diagnosed intelligently,and the diagnosis accuracy is 95%.6.The BTT test bench based on the eddy current sensor is designed,and a lot of experimental tests are carried out.The validity and feasibility of the proposed method were verified by experimental tests.