CN-122016702-A - Asphalt material rapid identification method based on infrared spectrum

CN122016702ACN 122016702 ACN122016702 ACN 122016702ACN-122016702-A

Abstract

The invention provides a rapid identification method of asphalt materials based on infrared spectrum, and belongs to the technical field of road engineering material detection and intelligent identification. According to the method, the infrared spectrum of asphalt is used as a chemical fingerprint, background interference is eliminated through standardized spectrum pretreatment, multi-scale features (including key functional group peak intensity/peak area, spectrum shape statistical features and PCA dimension reduction features) are extracted, and the rapid identification and confidence output of asphalt category are realized by combining with machine learning models such as random forests. The identification category covers matrix asphalt, SBS modified asphalt, rubber powder modified asphalt and the like, and can be expanded to ageing grade judgment and blending amount interval judgment. Compared with the existing method relying on solvent separation or experience discrimination, the method has the advantages of small sample demand, simple pretreatment, high recognition speed, quantifiable result, on-site deployment support and the like, is suitable for entrance acceptance and mixing early warning and construction process quality control, and provides a reliable technical means for efficient and accurate discrimination of asphalt materials.

Inventors

LIU XIANG
LI CHUNYAN
SU HAITAO
LI XIAOLONG
Che Diexuan
QIN ZHEN
JIA JINGPENG
DU HUI
Cai Chunbing
LI JINGRU

Assignees

云南省公路科学技术研究院
昆明理工大学

Dates

Publication Date: 20260512
Application Date: 20251225

Claims (8)

1. The method for rapidly identifying the asphalt material based on the infrared spectrum is characterized by comprising the following steps of: S1, acquiring infrared spectrum data by adopting an attenuated total reflection-Fourier transform infrared spectrum ATR-FTIR spectrometer based on a preset class of asphalt sample, and constructing an infrared spectrum fingerprint library; the preset category specifically comprises matrix asphalt, SBS modified asphalt, rubber powder modified asphalt and composite modified asphalt, and the infrared spectrum fingerprint library comprises infrared spectrum data and a label vector; s2, outputting a preprocessed spectrum through preprocessing operation based on an infrared spectrum fingerprint library; S3, based on the preprocessed spectrum, performing feature fusion to output a spliced feature vector after time domain statistical feature extraction and spectral shape feature extraction, and then splicing the spliced feature vector with a basic spectral feature vector to output a dimension-reduction fingerprint feature; the basic spectrum characteristic vector is original The intensity value of the individual spectrum is calculated, The total number of spectrum data points is specifically 850; S4, based on the complete feature vector, PCA calculation and principal component selection are carried out through principal component analysis, and a dimension-reducing fingerprint feature vector is output; s5, constructing a data set based on the dimension reduction fingerprint feature vector and the label vector, dividing the data set into a training set, a verification set and a test set, inputting the training set and the verification set into a classification model for training, and obtaining an asphalt material identification model; the classification model adopts a random forest classifier; s6, inputting a test set to the asphalt material identification model, executing threshold judgment through a confidence coefficient decision module, and taking the confidence coefficient as a confidence coefficient And if not, directly outputting the asphalt material category to finish the rapid identification method of the asphalt material.
2. The method for quickly identifying asphalt materials based on infrared spectroscopy according to claim 1, wherein in S2, the preprocessing operation is specifically polynomial fitting baseline correction, derivative transformation enhancement feature and classification guide normalization; The polynomial fit baseline correction expression is as follows: In the formula, For the original spectral intensity at the wavenumber v, Is a polynomial function of degree k; the expression for class-oriented normalization is as follows: In the formula, For the spectral intensity value, scale_factor is a preset range dynamic weight coefficient.
3. The method for rapidly identifying asphalt materials based on infrared spectrum according to claim 1, wherein the step S3 specifically comprises the following steps: S3.1, inputting the preprocessed spectrum, and outputting a time domain statistical feature vector through time domain statistical feature extraction; s3.2, inputting the preprocessed spectrum, extracting spectral shape characteristics, and outputting a spectral shape characteristic vector; S3.3, inputting a time domain statistical feature vector and a spectral feature vector, performing feature fusion through weighted feature stitching, and outputting a stitched feature vector; S3.4, inputting the spliced feature vector and the basic spectrum feature vector, carrying out feature splicing, and outputting a complete feature vector.
4. The method for rapidly identifying asphalt materials based on infrared spectrum according to claim 3, wherein in S3.1, the time domain statistical feature vector comprises a mean feature, a variance feature, a skewness feature, a kurtosis feature, a maximum position, a minimum position and a peak-valley difference, and specifically comprises the following calculation formula: Mean value characteristics The calculation expression is as follows: In the formula, As a total number of spectral data points, Is that Is used for the indexing of (a), Is a spectral vector; Variance characteristics The calculation expression is as follows: ; deflection characteristics The calculation expression is as follows: In the formula, Is the third power of standard deviation; Kurtosis characteristics The calculation expression is as follows: In the formula, The standard deviation is a square; Maximum position The calculation expression is as follows: In the formula, As a vector of wave numbers, Is a spectral vector; minimum position The calculation expression is as follows: peak-valley difference The calculation expression is as follows: 。
5. The method for rapidly identifying asphalt materials based on infrared spectrum according to claim 3, wherein in S3.2, the spectral shape characteristic vector comprises a center of gravity of a band, a width of the band, the number of local peaks, an average value of symmetry indexes of the spectral peaks and a standard deviation of the symmetry indexes of the spectral peaks.
6. The method for rapidly identifying asphalt materials based on infrared spectrum according to claim 1, wherein in S3.3, the expression of weighted feature stitching is: Wherein I represents a horizontal stitching operation, For the time-domain statistical feature vector, In order to be a characteristic vector of the spectral shape, For the preset time-domain statistical feature weights, The characteristic weight of the spectrum shape is preset.
7. The method for rapidly identifying asphalt materials based on infrared spectrum according to claim 1, wherein in the step S5, the parameters of the random forest classifier specifically set up include 300 trees, 20 maximum depths, 5 minimum leaf node samples, a default feature sampling ratio and 42 random seeds.
8. The method for quickly identifying asphalt materials based on infrared spectrum according to claim 1, wherein in S6, the confidence decision module specifically calculates a maximum probability and a corresponding category and performs threshold judgment, and the expression of the maximum probability and the corresponding category is as follows: Wherein, the The probability vector which represents the model output has the value range of [0, 1]; Performing threshold judgment: 。

Description

Asphalt material rapid identification method based on infrared spectrum Technical Field The invention belongs to the technical field of road engineering material detection and intelligent recognition, and particularly relates to a rapid recognition method for asphalt materials based on infrared spectrum. Background In the production and construction process of asphalt materials, the problems of confusion of modifier types, inaccurate doping amount or batch mixing and the like often exist, and the pavement engineering quality is seriously influenced. The conventional identification means mainly comprise a traditional chemical detection method (such as solvent separation and chromatographic analysis), a conventional physical property test (such as penetration and softening point) and artificial experience judgment, but the methods or processes are complicated, cannot be rapidly implemented on site or cannot accurately identify the type and the doping amount of the modifier, and the result is easily influenced by subjective factors. In recent years, although the oil source identification technology based on machine learning improves the identification efficiency, the technology mainly focuses on asphalt oil source identification, and a solution for identifying the modifier type, judging the doping amount interval and evaluating the confidence of the result is still lacking in a system. In the past research, the infrared spectrum characteristic is utilized, and the distinction of different oil source asphalts is realized through pattern recognition. However, the method still has the obvious limitations that whether the modifier and the specific type (such as SBS, rubber powder and the like) are added in the asphalt cannot be effectively identified, the distinguishing capability of the modifier doping amount interval is lacking, and the confidence output and the uncertain sample prompting mechanism are not introduced, so that complex and changeable field samples are difficult to deal with in actual engineering. Therefore, development of a field application method capable of rapidly and accurately identifying asphalt types and modification states and having the capability of evaluating result credibility is needed. Disclosure of Invention In order to solve the technical problems, the invention provides a method for rapidly identifying asphalt materials based on infrared spectrum. In order to achieve the above object, the method specifically comprises the following steps: S1, acquiring infrared spectrum data by adopting an attenuated total reflection-Fourier transform infrared spectrum ATR-FTIR spectrometer based on a preset class of asphalt sample, and constructing an infrared spectrum fingerprint library; the preset category specifically comprises matrix asphalt, SBS modified asphalt, rubber powder modified asphalt and composite modified asphalt, and the infrared spectrum fingerprint library comprises infrared spectrum data and a label vector; the invention adopts an ATR-FTIR spectrometer, the set parameters comprise a wave number range of 4000-600 cm -1, a resolution of 4 cm -1, a scanning frequency of 32 times and a temperature control range of 25+/-1 ℃, each asphalt sample is collected for 3 times, an average spectrum is calculated by an average spectrum, infrared spectrum data are constructed, and category labels are marked (0 Is matrix asphalt, 1 is SBS modified asphalt, 2 is rubber powder modified asphalt, and 3 is composite modified asphalt), and constructing a label vector; s2, outputting a preprocessed spectrum through preprocessing operation based on an infrared spectrum fingerprint library; the preprocessing operation specifically comprises polynomial fitting baseline correction, derivative transformation enhancement characteristics and classification guide normalization; In the invention, a polynomial fitting baseline correction method is adopted, and the spectrum intensity after baseline correction The expression is as follows: In the formula, For the original spectral intensity at the wavenumber v,For k times polynomial function, the estimated baseline is represented, the fitting capacity and stability are considered to be compatible in the embodiment, k is taken as 3, and in the embodiment, 2500-2000 cm -1 and 1800-1700 cm -1 are selected for the fitting interval of the infrared spectrum data. And finally, carrying out classification guide normalization by adopting maximum-minimum normalization, and reserving the relative peak intensity proportion to obtain a preprocessed spectrum, wherein the expression is as follows: In the formula, For the spectrum intensity value, scale_factor is a dynamic weight coefficient with a preset range, and in the embodiment, the dynamic weight coefficient ranges from 0.8 to 1.5, and the characteristic peak of the modifier is given high weight based on the known chemical knowledge. S3, based on the preprocessed spectrum, performing feature fusion to output a spliced f