基于K近邻算法的沥青红外光谱质量控制

沥青质量的优劣直接影响沥青路面的性能。红外光谱技术作为一种无损、快速的微观分析手段,能够对沥青品牌进行识别,对于沥青质量控制具有十分重要的意义。为提高沥青红外光谱技术质量控制的准确性,选取K近邻算法对红外光谱数据进行解析和训练,共收集了7种不同品牌的共336个沥青样本,通过衰减全反射傅里叶变换红外光谱仪对其进行光谱采集和分析。结果表明,使用K近邻算法对沥青红外光谱数据分析具有可行性;当K=6,且σ=0.1时,此时沥青品牌预测误差最小,识别准确率达到95%左右。

Infrared Spectral Quality Control of AsphaltBased on K-Nearest Neighbor Algorithm

The quality of asphalt directly affects the performance of asphalt pavement.As a non-destructive and fast micro analysis method,infrared spectroscopy can identify the brand of asphalt which isof great significance for asphalt quality control.In order to improve the accuracy of the quality control ofasphalt infrared spectroscopy,K-nearest Neighbor Algorithm is selected in the paperto analyze and trainthe infrared spectral data.In this paper,a total of 336asphalt samples from 7different brands arecollected,and the spectra of these samples are collected and analyzed by attenuated total reflection Fouriertransform infrared spectrometer.At the same time,the asphalt brand prediction error rate at the sampleratio is analyzed when using different K values,different test and total sample ratios.The results showthat when K=6andσ=0.1,the asphalt brand prediction error obtains a smallest minimum value,and therecognition accuracy reaches 95%.