CN104020124A - Spectral wavelength screening method based on preferential absorptivity - Google Patents

Abstract

The invention discloses a spectral wavelength screening method based on preferential absorptivity, and the methos is as follows: S1, testing a sample to get spectral data and index determination values; S2, selecting wavelength screening delta, determining maximum absorptivity value Amax and minimum absorptivity value Amin; S3, setting absorptivity step length epsilon, equally dividing (Amin, Amax) into n parts; S4, arbitrarily taking two points from Amin corresponding starting point, Amax corresponding end point, and n-1 equal diversion points for combination to obtain an absorptivity interval (A *, A *); S5, determining a wavelength combination corresponding to the (A *, A *); S6, in accordance with the above steps S4 and S5, exhausting all the absorptivity interval (A *, A *), establishing a scaling prediction model for the wavelength combination corresponding to each absorptivity interval, calculating mean square root errors or correlation coefficients; and S7, finding the absorptivity interval corresponding to mean square root error minimum value or correlation coefficient maximum value to obtain spectral wavelength screening results, namely the wavelength combination corresponding to the absorptivity interval. The spectral wavelength screening method has the advantages of less calculation amount and good predicting effect.

Description

Spectral wavelength screening method based on absorption rate preference

technical field

The invention relates to the technical field of wavelength screening in spectroscopic system design, in particular to a spectroscopic wavelength screening method based on absorption rate preference.

Background technique

Infrared spectroscopic analysis is a method to identify substances and determine their chemical composition and content. It does not require biochemical reagents. It is simple, fast, non-destructive, and easy to analyze in real time. It has application advantages in many fields. At present, the technology of developing a full-band general-purpose infrared spectrometer has been relatively mature abroad, but it has the disadvantages of bulky instruments and expensive prices, which are not suitable for popularization and application. Therefore, the research and development of low-priced small-scale special-purpose infrared analysis instruments has application prospects.

In the infrared spectrum analysis, the spectral wavelength screening method with high signal-to-noise ratio is a key technology, which is of great significance for establishing high-precision analysis models, reducing model complexity, and designing spectroscopic systems for small special-purpose spectroscopic instruments. However, there are a lot of wavelengths in the infrared band. If the wavelengths are selected by random combination and modeled separately, the existing computer computing speed is far from enough. Therefore, there are still difficulties in the selection of wavelengths for infrared spectroscopic analysis and the design of spectroscopic systems for small special spectroscopic instruments, and there is a lack of effective screening methods for spectroscopic wavelengths.

Contents of the invention

The main purpose of the present invention is to overcome the shortcomings and deficiencies of the prior art, and provide a method for screening spectral wavelengths based on absorptivity preference. This method can effectively screen out the wavelength combinations with high signal-to-noise ratios corresponding to the analysis objects, and has a wide range of applications. Wide range, simple model, less calculation, good prediction effect, etc., it proposes an effective solution for the design of the spectroscopic system in small special analytical instruments.

The object of the present invention is achieved through the following technical solutions: the optimal spectral wavelength screening method based on absorbance comprises the following steps:

S1. Test the sample to obtain the spectral data and the measured index value of the sample;

S2. In the measured spectral band, according to the physical and chemical characteristics of the measured object and the performance of the spectroscopic instrument, select the range Δ for wavelength screening that has low noise and covers index information, and at the same time determine the average value of the sample in this wavelength range. The maximum value A _max and the minimum value A _min of the absorption rate corresponding to the spectrum;

S3. Set an appropriate absorption rate step size ε, divide the full absorption rate range (A _min , A _max ) into n equal parts, and obtain n-1 equal parts;

S4. Combining any two points from the starting point corresponding to the minimum value of the absorption rate, the end point corresponding to the maximum value of the absorption rate, and n-1 equal points, to obtain an absorption rate interval (A _* , A ^* ), where ( A _* ,A ^* ) (A _min ,A _max );

S5. According to the corresponding relationship between the wavelength and the absorptivity of the spectral data, within the wavelength screening range Δ, determine the wavelength combination corresponding to the absorptivity interval (A _* , A ^* );

S6. According to the above steps S4 and S5, exhaustively enumerate all the absorptivity intervals (A _* , A ^* ), establish a calibration prediction model for the wavelength combination corresponding to each absorptivity interval, and calculate the mean square of the spectral prediction value and the measured value root error (RMSEP) or correlation coefficient;

S7. Find the absorption rate interval corresponding to the root mean square error minimum value or the correlation coefficient maximum value, determine it as the optimal absorption rate interval, and then find the wavelength combination corresponding to the optimal absorption rate interval, and complete the screening of the spectral wavelength .

Preferably, in the step S2, the wavelength screening range Δ is set according to the infrared light absorption range of the physical and chemical properties of the measurement object, and the noise band range caused by the operation of the instrument is excluded.

Preferably, in the step S3, the absorbance step size ε is set according to the spectral overall absorbance value (including the minimum and maximum absorbance values) obtained from the spectral experiment, model accuracy and modeling operation efficiency.

Preferably, in the step S6, the calibration prediction model is based on partial least squares (PLS), multiple linear regression (MLR), principal component analysis (PCA) and the like.

Preferably, in the step S7, the optimal absorptivity interval and the corresponding wavelength combination are the most suitable wavelength models for quantitative analysis obtained according to the effect of spectral calibration prediction.

Compared with the prior art, the present invention has the following advantages and beneficial effects:

1. The present invention selects the optimal absorption rate interval by segmenting the absorption rate, and then using partial least squares (PLS), multiple linear regression (MLR), principal component analysis (PCA) and other calibration prediction models, Then effectively screen out the high signal-to-noise ratio wavelength combination corresponding to the analysis object, which has the advantages of wide application range, simple model, less calculation amount, good prediction effect, etc., and proposes an effective solution for the design of the spectroscopic system in small special analysis instruments .

2. After several experiments, it has been confirmed that the wavelength combination selected by the present invention is significantly better than the wavelength screening range Δ where they are located, and the prediction effect is significantly improved. Since the number of wavelengths used in the present invention is greatly reduced, it is very important for the establishment of a high-precision analysis model It is of great significance to reduce the complexity of the model and design the spectroscopic system of small special spectroscopic instruments.

3. The present invention screens the wavelength model according to the magnitude of the absorption rate, avoids the shortcomings of high noise at high absorption rate wavelengths and weak information at low absorption rate wavelengths, and has obvious physical and chemical significance. It overcomes the deficiency of conventional single continuous waveband sorting and screening based on wavelength size, and can select multiple high signal-to-noise ratio wavebands according to the absorption rate, and has a wider application range.

Description of drawings

Fig. 1 is the method flowchart of embodiment 1.

Fig. 2 is a schematic diagram of the wavelength combinations in the optimal absorption rate range (0.42, 1.00) in Example 1, taking the near-infrared analysis of human serum cholesterol as an example.

FIG. 3 shows band 1 in the wavelength combination selected in FIG. 2 .

FIG. 4 shows the second band in the wavelength combination selected in FIG. 2 .

Detailed ways

The present invention will be further described in detail below in conjunction with the embodiments and the accompanying drawings, but the embodiments of the present invention are not limited thereto.

Example 1

In this embodiment, the near-infrared transmission spectrum analysis of human serum cholesterol is taken as an example, with reference to FIG. 1 , the specific steps of the spectroscopic wavelength screening method based on the absorption rate preference proposed by the present invention are described.

S1. The sample is tested to obtain the spectral data and the clinical measurement value of the cholesterol index of the sample.

S2. In this embodiment, the full near-infrared band (780-2498nm) is selected as the wavelength screening range Δ. Within this wavelength screening range, the maximum and minimum absorbance values of the average spectra of all samples are close to 5 and 0 respectively, so the A _max and A _min are set to 5 and 0, respectively.

S3. Set the absorption rate step size ε to 0.01, divide the full absorption rate range (0,5) into 500 equal divisions, and obtain 499 equal division points, and the corresponding absorption rate of each equal division point is 0.01, 0.02, ... , 4.99.

S4, from 0, 0.01, 0.02, ..., 4.99, 5 arbitrarily select two points to combine to obtain an absorption rate interval, such as shown in Figure 2, select these two absorption rates of 0.42 and 1.00 to form an absorption rate interval (0.42 , 1.00).

S5. It can be seen from Fig. 2 that the absorptivity interval corresponds to two bands A and B, as shown in Fig. 3 and Fig. 4 respectively. In the absorptivity interval (0.42, 1.00), there are wavelengths of 1374-1392nm Band A, and band B with a wavelength of 1544-1852nm. These two bands are the wavelength combinations corresponding to the absorptivity interval (0.42, 1.00).

S6. According to the above steps S4 and S5, enumerate all the absorption rate intervals (A _* , A ^* ), for example, you can first fix a starting point, then exhaustively enumerate all other equally divided points, starting points and end points, and then change the starting point in turn , that is, start from 0, take (0,0.01), (0,0.02), (0,0.03), ..., (0,5), and then start from the equal point of 0.01, take (0.01,0.02 ), (0.01,0.03), ..., (0.01,5), select in turn until all points are combined in pairs and end.

Establish a partial least squares (PLS) calibration prediction model for the wavelength combination corresponding to each of the above absorptivity intervals. At present, the partial least squares (PLS) method is a widely used and effective modeling method for infrared spectral analysis. Through This method calculates the root mean square error (RMSEP) and correlation coefficient between the predicted and measured values of the spectrum in the absorbance interval.

S7. A total of 500 × 499 × . The absorption rate interval is determined as the optimal absorption rate interval, and then the wavelength combination corresponding to the optimal absorption rate interval is found to complete the screening of the spectral wavelength.

The present embodiment compares the wavelength screening of the present invention in conjunction with the PLS method and the full near-infrared spectral region PLS method, and the comparison results are as follows:

PLS method in the full near-infrared spectrum region: the full spectrum band is 780-2498nm, and the predicted root mean square error and correlation coefficient obtained by using unknown test samples are 0.835 (mmol L ^-1 ) and 0.677, respectively.

Absorbance preferred wavelength screening combined with PLS method of the present invention: the determined optimal absorptivity interval is (0.42, 1.00), the corresponding band combination is 1374～1392∪1544～1852, and the predicted root mean square error obtained by using unknown test samples , and the correlation coefficients were 0.181 (mmol L ^-1 ) and 0.988, respectively.

Experimental results confirm: the wavelength combination screened out based on the absorption rate preferred wavelength screening method of the present invention is significantly better than the prediction effect of the full near-infrared spectrum region, the number of wavelengths is significantly reduced, and the method can select multiple high signal-to-noise wavelengths according to the absorption rate. It has a wider application range than the wavelength band, and is of great significance for establishing high-precision analysis models, reducing model complexity, and designing spectroscopic systems for small special-purpose spectroscopic instruments.

The above-mentioned embodiment is a preferred embodiment of the present invention, but the embodiment of the present invention is not limited by the above-mentioned embodiment, and any other changes, modifications, substitutions, combinations, Simplifications should be equivalent replacement methods, and all are included in the protection scope of the present invention.

Claims (4)

1. based on absorptivity point optical wavelength screening technique preferentially, it is characterized in that, comprise the following steps:

S1, test sample, obtain the index determining value of spectroscopic data and sample;

S2, in measured spectral band, according to the performance of the physics of determination object, chemical characteristic and spectral instrument, select low noise and cover the range delta for wavelength screening of indication information, determine absorption maximum A corresponding to sample average spectrum in this wavelength coverage simultaneously _maxwith minimum value A _min;

S3, suitable absorptivity step-length ε is set, by hypersorption rate scope (A _min, A _max) n decile, obtain n-1 Along ent;

S4, from terminal corresponding to starting point corresponding to absorptivity minimum value, absorption maximum and n-1 Along ent, get arbitrarily at 2 and combine, obtain an absorptivity interval (A _*, A ^*), wherein (A _*, A ^*) (A _min, A _max);

S5, according to the corresponding relation of the wavelength of spectroscopic data and absorptivity, in wavelength screening range delta, determine this absorptivity interval (A _*, A ^*) corresponding wavelength combinations;

S6, according to above-mentioned steps S4, S5, exhaustive all absorptivity interval (A _*, A ^*), the interval corresponding wavelength combinations of each absorptivity is set up to calibration forecast model, calculate root-mean-square error or the related coefficient of Forecast of Spectra value and measured value;

S7, find root-mean-square error minimum value or the corresponding absorptivity of related coefficient maximal value interval, be defined as optimal absorption rate interval, and and then find the interval corresponding wavelength combinations of this optimal absorption rate, complete the screening of point optical wavelength.

2. according to claim 1 based on absorptivity point optical wavelength screening technique preferentially, it is characterized in that, in described step S2, wavelength screening range delta is to set for the absorption region of infrared light according to the physics of determination object, chemical characteristic, and gets rid of instrument and move the noise wavelength band causing.

3. according to claim 1 based on absorptivity point optical wavelength screening technique preferentially, it is characterized in that, in described step S3, absorptivity step-length ε is that the spectrum overall absorption rate value, model accuracy and the modeling operational efficiency that obtain according to spectrum experiment are set.

4. according to claim 1ly it is characterized in that based on absorptivity point optical wavelength screening technique preferentially, in described step S6, calibration forecast model adopts based on offset minimum binary, multiple linear regression, principal component analytical method.