| A multi-scale characterization method considering uncertainty characteristics is proposed in this thesis.This method characterized the mechanical properties of BCF/PEEK under different processes and investigated the mechanical properties of BCF/PEEK under different component structures and material properties.A reasonable design scheme is proposed to realize BCF/PEEK high performance structure design.The main research contents are as follows.1)A new mechanical properties characterization method for BCF/PEEK was proposed.According to SEM observation,the diameter and position of fiber fibers were calibrated,and UCF/PEEK model considering fiber diameter and random arrangement was established.A fiber random distribution plug-in covering periodic boundary conditions was developed to quickly generate high-quality UCF/PEEK models.Based on theory of progressive homogenization,a Python script file was written to realize the rapid characterization of UCF/PEEK macroscopic mechanical properties considering structural differences.2)The macroscopic mechanical properties of BCF/PEEK were characterized considering the differences of typical structural forms.Based on the observed and calibrated typical structural parameters,a RVE model which considered the differences of the typical BCF/PEEK structural forms was established.The mechanical properties of the typical BCF/PEEK were characterized by this model.The effects of different parameters on the mechanical properties of BCF/PEEK were predicted by changing the fiber bundle fluctuation amplitude,fiber volume content and braid Angle between fiber bundles in the typical model of BCF/PEEK.According to the test standard of mechanical properties of composite materials,the corresponding test parts were prepared,and the correctness of the model was verified by standard tensile test.3)A sample generation method based on "Sobol method" and a result sample acquisition method based on " Order driven finite element method" are proposed,and the input and output sample sets are generated.The parameters to be considered in the characterization of mechanical properties of BCF/PEEK,a "Sobol sequence" method was proposed to consider the random structural parameters of BCF/PEEK fiber bundles and the structural properties of fiber components.Considering the computational redundancy of obtaining a large number of samples,an "instruction driven finite element" method "method" is proposed to obtain samples quickly and easily.Using MATLAB software to write a program based on "Sobol sequence" and PYTHON software to write a finite element program driven by instructions,the sample set used in the analysis of high-performance design methods is obtained.4)The grey correlation degree of BCF/PEEK structural parameters and the global sensitivity of component structure parameters were analyzed,and a high-performance structural design method for BCF/PEEK was proposed.Based on the generated sample set,grey correlation analysis and global sensitivity analysis were carried out respectively for the mechanical properties characterization model considering structural parameters and material parameters,and the correctness of grey correlation degree and global sensitivity coefficient was verified.The optimal structural parameter model was obtained by considering the grey correlation degree.Meanwhile,a high performance BCF/PEEK structural design method for different working conditions was proposed considering the global sensitivity.Based on the cross-scale mechanical property prediction method,the correctness of the high performance design method is verified.The results showed that UCF/PEEK mechanical properties characterization model considering random distribution of fiber diameter and position and BCF/PEEK mechanical properties characterization model considering uncertain structural parameters of fiber bundle under quadratic progressive averaging can truly characterize the mechanical properties of BCF/PEEK under different pressures and temperatures.Compare with the experimental results,the maximum errors were 1.501% and 3.885%,respectively.Parameterized cross-scale finite element models are directly used for grey relational analysis and global sensitivity analysis.Based on the analysis results,a high-performance BCF/PEEK structure was designed and verified.The minimum error rate was 0.16%,and the maximum error was 0.34%. |
