| As an important energy conversion device in the industrial process,the compressor has a wide range of applications in all walks of life.The impeller is the key component of the compressor to realize the function-energy conversion.Fracture accidents often occur when impeller working on high-speed and suffered from the complex stress and harsh working environment,so the service safety of the impeller is an important part of the reliable operation of the compressor.It is necessary to study the fracture behavior of the impeller structure,establish an early prediction mechanism and realize the intelligent diagnosis of structures.In this paper,a certain type of centrifugal impeller is the first research object.According to the existing failure cases,the crack initiation position and initial length are reasonably set on the impeller.Based on the extended finite element method,variation characteristics at different locations of the type I stress intensity factor are studied in the centrifugal impeller.After that,a neural network algorithm was used to take the mechanical characteristics of the impeller with cracks as input to realize the prediction of the location and length of the cracks.Finally,taking the axial flow impeller as the research object,combining the extended finite element method and the energy-based Paris criterion,the fatigue crack growth life under resonance load is approximated.This article focuses on the following work.(1)Based on the Extended Finite Element(XFEM)method in ABAQUS,the stress intensity factors of cracks at different positions at the blade inlet are simulated and calculated.The relationship between the stress intensity factor and the crack position or crack length is obtained.Through analysis we found that at the blade inlet,the type I stress intensity factor has an approximate linear relationship with the crack length(4mm-10mm),and an approximate quadratic function relationship with the blade height.The value of 1.051 for the shape factor of the blade inlet is given in a certain range through studying on the law of the shape factor.(2)Through modal calculation and static calculation,the sensitivity analysis of the cracked impeller was carried out.The natural frequencies,nodal displacements and nodal strains of the cracked structure were compared respectively with the structure without crack.Through analysis we found that the natural frequencies and nodal displacements of the structure are less sensitive to cracks.Nodal strain is highly sensitive to the location and length of cracks.(3)The strain field near the crack tip is used as the characteristics for crack identification.The Tensorflow2.0 framework is used to build a neural network model.The strain difference of the node in the blade height direction is used as the input parameter of the neural network,and the position of the crack coded by one-hot is used as the output of the neural network.The results show that the neural network algorithm can accurately predict the location of the crack based on the node strain difference in the direction of the blade height,and can reasonably predict the length of the crack based on the value of the node strain difference.(4)Prestress modal analysis,harmonic response analysis and fatigue crack propagation analysis were carried out on the axial flow impeller.Through analysis we found that the prestress will increase the natural frequency of the structure.When the frequency of the pressure load is 854 Hz,the third-order mode of the structure will be excited.At this time,the maximum stress of the impeller structure is located at the outlet.The fatigue crack growth life of the structure under resonance load is nearly 50,000 cycles when there is a crack with a length of 10 mm exists at the location where the stress is the max. |
PDF Full Text Request