| As our country gradually steps into the era of "great quality", the reliability of engineering vehicle products becomes extremely important. Load spectra compiling is the basis of fatigue and reliability analysis of engineering vehicle components, and it is the effective mean to assure rationality of design and authenticity of test. Load extrapolation, a key step in the process of load spectra compiling, can be able to extrapolate load samples to long-term and even the whole life of vehicle components. This thesis focuses on the load extrapolation of engineering vehicle combining the technology of kernel density estimation(KDE) and Bayesian statistical theory(BST). A new rainflow extrapolation method based on non-parametric extrapolation model with full bandwidth and load extension is proposed and instances verification based on the measured data of a certain type of wheel loader is carried out. The dissertation is supported by the program for the National Natural Science Foundation of China(NO. 51375202). The main contents of this presented work are listed as following:1. Based on the review of load extrapolation methods, principle and key process of rainflow extrapolation are analyzed in depth. Advantages, key points and limitations of parametric rainflow extrapolation(PRFE), limiting rainflow extrapolation(LRFE) and nonparametric rainflow extrapolation(NPRFE) are discussed. The study found that NPRFE is based on data driving and it does not depend on parametric models of load sample, which are unique advantages compared with other extrapolation methods. However, NPRFE model is simple and extrapolation accuracy is limited. Theoretical basis and references are provided for the improvement of the model and method.2. Considering drawbacks of existing NPRFE model, NPRFE model with full bandwidth is established. According to kernel selection, determination of adaptive bandwidth matrix and numerical calculation and analysis of bandwidth, improved NPRFE model is structured. And then the extrapolation effects of improved NPRFE model are analyzed based on the measured data of wheel loader. The results show that the improved extrapolation model has better effect on extrapolation of medium and small load cycles, while the extrapolation accuracy of large load cycles is still not ideal.3. To solve this issue, improved NPRFE method based on load extension is proposed, which is combined with BST. Extrapolation effects of different methods are compared in detail according to load amplitude- cumulative frequency curve and pseudo damage number. Analysis results show that extrapolation results based on the improved NPRFE method are similar with them based on LRFE method, and it can be conclude that the improved NPRE method is reasonable.4. Load data processing software is written and it is used to load spectra compiling of wheel loader. The software is based on the improved extrapolation model and method proposed in this paper and it integrates many functions such as load conversion, extension, extrapolation and reprocessing. Moreover, the software is used to the load extrapolation in the condition of sedentary soil and spectra compiling of wheel loader. Design spectrum and program spectrum of axle shaft are achieved under combining different operation sections.This article proposes an adaptive NPRE method for wheel loader operation load. It can realize reasonable extrapolation of load cycles with different amplitudes and improves load spectra facticity, and finally provides accurate input data for fatigue and reliability design of engineering vehicle components. |
PDF Full Text Request