UNDE Study On Pattern Recognition For Microstructure Of 20 Steel Aged At High Temperature

In boiler, turbine and other high temperature components, pearlite heat-resistant steels have extremely been used, and their researches and progress have significant influence on the power and petroleum chemical industry. Under long-term exposure at high temperature, with the diffusion process of elements, The microstructure becomes unstable and thus causes microstructural degradation such as graphitization, pearlite spheroidization, which results in the decline of mechanical properties in steel. With the increasing temperature and extending service time, the microstructure degradation is accelerated and the property is further aggravated. The creep rupture eventually occurs in the components.Material can be treated as a filter to ultrasonic wave. When a narrow pulse ultrasonic signal interacts with materials, the variations of grain size and microstructure affect the propagation characteristics of ultrasonic wave, such as dispersion, reflection and transmission. The comprehensive reactions of many factors change the propagation direction, amplitude and phase of the echo signal. By means of FFT, the echo signal can be treated as the compound signal that is composed of harmonics with different frequencies and contains many useful information characterizing microstructure and properties of material. In order to accomplish the non-destructive evaluation of microstructure, it is necessary to improve the ultrasonic parameter testing system, extract and optimize the ultrasonic signal feature describing the microstructure, investigate the propagation mechanism of ultrasonic wave in different microstructure and develop pattern recognition system of microstructure. It also provides effective method to ensure the application of ultrasonic non-destructive evaluation to engineering.Based on the project "the remaining life prediction of high temperature piping in petrochemical plant", the current thesis discussed the application of signal feature to pattern recognition of microstructure, and developed the studies on the pattern recognition of microstructure and the other relevant questions based on the ultrasonic signal feature.By summarizing the ultrasonic non-destructive evaluation and characterization, the thesis introduced the research background and proposed feasible experiment scheme in the first and second chapters.Combining with the microstructure evolution during high temperature ageing, the correlation among the velocity, elastic modulus and the microstructure evolution was analyzed and discussed in the third chapter, meanwhile, the feasibility of damage factor Dv characterizing the microstructure damage during high temperature ageing was also investigated based on the continuum damage theory. By means of the ultrasonic theory, the influence of microstructure degradation on the ultrasonic dispersion, reflection and transmission was also analyzed in chapter four. Considering that echo signal can be treated as the signal that was composed of harmonic waves with different frequencies and contains much useful information on the microstructures and properties, the pattern recognition of microstructure aged and the other relevant questions based on the ultrasonic signal feature were discussed from chapter five to seven. The research results were as followings:1. The ultrasonic velocity was sensitive to the microstructure degradation during high temperature aging. The longitudinal and shear velocities decreased exponentially with the increasing ageing time, which was attributed to the influence of microstructure evolution on elastic modulu. It is feasible to evaluate the microstructure evolution by means of ultrasonic velocity-elastic property.2. It is feasible to characterize the microstructure damage using damage factor Dv. The non-destructive evaluation to the microstructure damage and property degradation can be implemented by measuring the ultrasonic longitudinal velocity, which is expected to become a kind of non-destructive method characterizing the microstructure degradation.3. Harmonic coefficients were extracted by FFT, and optimized by principal component analysis. According to the shortest distance criterion, the cluster analysis of high temperature aging microstructure is accomplished and the discrimination result was no less than 84%.4. The multi-scale wavelet transform was applied to extract the ultrasonic signal features of microstructure aged. By designing and choosing the network structural parameters, BP neural network for pattern recognition of high temperature ageing microstructure is constructed using Matlab software. The discrimination result reached 86.7%. The identifications results to 1Cr5Mo and 30Mn2SiV structure steel reached 85% and 89% respectively.