JP2008014779A - Diagnosis method of concrete - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a diagnosis method of concrete capable of diagnosing deterioration excluding an influence of an influential factor, when detecting deterioration of a concrete surface optically. <P>SOLUTION: In this method for diagnosing deterioration of concrete by irradiating the concrete surface with a near infrared ray and performing spectroscopic analysis of light reflected by the concrete surface, a deteriorated component is detected from a collected absorption spectrum by using a chemometrics procedure. <P>COPYRIGHT: (C)2008,JPO&INPIT

Description

  The present invention relates to a concrete diagnostic method for optically diagnosing the soundness of a concrete building.

  Causes of deterioration of concrete structures such as tunnels and bridges include neutralization, salt damage, and alkali aggregate reaction in addition to poor construction. Neutralization occurs when concrete reacts with carbon dioxide (carbon dioxide) in the atmosphere and changes to calcium carbonate. When this neutralization proceeds to a part where the reinforcing bar is present, the protective coating on the surface of the reinforcing bar is lost and corrosion of the reinforcing bar occurs. Corrosion of rebars due to salt damage includes a large amount of chloride in the concrete, such as the penetration of chloride from the surface of the concrete by seawater, etc., and the protective coating of the rebar is destroyed by the action of chloride ions, resulting in rebar corrosion. End up. Alkali-aggregate reaction is a reaction in which certain reactive components in the aggregate react with the alkali contained in the cement and the product absorbs and expands with moisture in the concrete, causing cracks in the concrete. It is.

  In addition, concrete deteriorates not only due to the above-mentioned single deterioration factor, but also due to multiple deterioration factors, multiple deterioration occurs, and chloride ions concentrate on the neutralized tip portion, which accelerates corrosion of the reinforcing bars. There are things. Usually, concrete diagnosis involves visual inspection of cracks, and then cuts the specimen from the reinforced concrete and tests the specimen for compressive strength, bruises the concrete surface, measures the rebound, and estimates the strength. Etc. are performed. These methods cut the specimen from the concrete or directly damage the concrete and damage the structure. Non-destructive inspection that can be repeated is effective as a method for appropriately diagnosing a structure without damaging the structure.

  In order to optically detect the deterioration of a concrete building, as shown in Patent Document 1, the concrete surface is irradiated with near infrared light, and the near infrared light reflected from the concrete surface is spectroscopically analyzed. The degree of deterioration is diagnosed by detecting absorption spectra of calcium carbonate and chloride generated in the process.

  That is, calcium carbonate has an absorption peak at 1.42 μm and chloride has an absorption peak at 2.26 μm. By analyzing this wavelength peak using a difference spectrum, the degree of deterioration is diagnosed from the difference in absorbance.

JP 2005-218881 A JP 2001-188039 A

  However, as concrete deterioration factors, particularly when concrete deteriorates in combination due to a plurality of deterioration factors, it is impossible to inspect by which influence factor (water content, unevenness, etc.) the deterioration is caused by analysis of only a single wavelength. In the case of a single wavelength, the wavelength peak of one target degradation factor cannot be accurately obtained, and the wavelength peak of the degradation factor may not be obtained. Shows wrong tendency and cannot be diagnosed accurately.

  Therefore, the object of the present invention is to solve the above-mentioned problems, and not to damage the concrete surface, and to perform a deterioration diagnosis of concrete only by a method of optically detecting the deterioration of the concrete surface. It is an object of the present invention to provide a concrete diagnostic method capable of diagnosing deterioration without the influence of the above.

  In order to achieve the above object, the invention of claim 1 is a method for diagnosing deterioration of concrete by irradiating a concrete surface with near infrared rays and spectroscopically analyzing light reflected from the concrete surface in a predetermined wavelength range. This is a concrete diagnostic method characterized in that a deterioration component is detected from a collected absorption spectrum using a chemometrics method.

  The invention according to claim 2 is the concrete diagnosis method according to claim 1, wherein the deterioration is diagnosed by detecting the concentration of calcium carbonate from the absorption spectrum.

  The invention according to claim 3 is the method for diagnosing concrete according to claim 1, wherein the deterioration is diagnosed by detecting the concentration of chloride ions from the absorption spectrum.

  According to the present invention, by analyzing an absorption spectrum using a chemometrics method, an excellent effect of accurately diagnosing deterioration of concrete is exhibited.

  A preferred embodiment of the present invention will be described below in detail with reference to the accompanying drawings.

  FIG. 1 shows an optical system used in the concrete diagnosis method of the present invention. A near-infrared light source 10 having a wavelength range of 0.9 to 2.5 μm, such as a halogen lamp, is applied to a concrete surface 11 and the reflected light is irradiated. The light enters the optical fiber 13 through the condensing lens 12, the reflected light is emitted from the optical fiber 13 to the fixed grating 14, and is reflected by a mirror 15, and is reflected by a MEMS (Micro-Electro Mechanical Mechanical) (MEMS) 16. Spectral scanning is performed to detect an absorption spectrum for each specific wavelength by the photodetector 17, and the concentration of calcium carbonate and chloride ions is detected from the absorption spectrum to deteriorate the concrete surface 11. Diagnose.

  The deterioration degree is detected from the absorption spectrum of the concrete by using a chemometrics method.

  First, as a MEMS 16, a PCX (Polychrom) spectroscope was used, and as a concrete surface 11, the degree of deterioration of a specimen simulating neutralization and salt damage of crushed and unground cement paste, mortar, concrete was measured. .

  That is, a halogen lamp was applied to the concrete surface, and the reflected light was scanned with the MEMS 16 and scanned.

  In this case, in order to detect the neutralization peak (1.42 μm) mainly, the spectral range of 0.9 to 1.7 μm and to detect the chloride ion peak (2.26 μm) mainly. The wavelength region was measured separately for 1.7 to 2.5 μm.

Outline of specimen:
a) Specimens for neutralization and salt damage measurement Specimen type: Cement paste, mortar, concrete Water-cement ratio: 48.5%
Shape: 4 × 4 × 16 cm (cement paste, mortar), 10 × 10 × 40 cm (concrete) used in pulverized and unground respectively Chloride ion concentration: 0, 1, ³ , 5, 10, 20 kg / m ³
Neutralization promotion period: 0, 1, 3 months (carbon dioxide concentration: 7%, temperature 40 ° C., humidity 50%) b) Real structure simulation specimen Compounding strength: 30 N / mm ²
Slump: 10cm
Air volume: 3.9%
Shape: 0.5 × 0.5 × 0.1m 4 bodies, 1 × 1 × 0.1m 1 body Chloride ion concentration: 0, 1, 5, 10 kg / m ³
(Shape: mixed in 0.5 × 0.5 × 0.1m)
Test items:
a) Neutralization measurement The degree of neutralization of cement paste, mortar, and concrete was measured. Further, the degree of neutralization was evaluated by quantifying the depth of neutralization and the amount of calcium hydroxide with TG / DTA using phenolphthalene. For neutralization, the specimen was mainly unground.

b) Measurement of salt damage Chloride ion concentration in cement paste, mortar and concrete was measured.

  For the actual structure simulation test specimen, the total salinity contained in the concrete was measured by potentiometric titration (JCI-SC4), and the relationship with absorbance was examined.

FIG. 12 shows the absorption spectrum of the suspended surface of the cement paste with the chloride ion concentration changed to 0 (cp0), 5 (cp5), 10 (cp10), 20 (cp20) kg / m ^3, and FIG. Similarly, the absorption spectrum of cement paste powder is shown.

analysis method:
The analysis method was carried out by both the conventional difference spectrum and the analysis using the chemometrics method of the present invention, and evaluation was performed from the measurement results.

a) Difference spectrum analysis Since the neutralization and chloride ion absorption peaks are known values, the peak of the chloride ion concentration in Fig. 12 is obtained using the straight line passing through the bottom of the peak with the absorption band as the baseline. If there is, the spectrum was rewritten so that the absorbance of the absorption peak was 0, and the absorbance at 2265 nm was determined from the difference between the baseline passing through the tail and the absorption peak.

b) Chemometric analysis The absorption spectrum was analyzed by a technique (tool) for statistically analyzing the spectrum, which is a method of the present invention.

Test results Neutralization measurement results The following reactions are known to occur when concrete is neutralized.

Ca (OH) ₂ + CO ₂ → CaCO ₃ + H ₂ O
As neutralization progresses, calcium hydroxide decreases and calcium carbonate increases. As a result of measurement by FTIR (Fourier Transform Infrared Spectrometer), as neutralization progresses, the absorbance of 1.42 μm indicating the characteristics of calcium hydroxide decreases and 3.98 μm indicating the characteristics of calcium carbonate increases. It could be confirmed.

  FIG. 14 shows the measurement results.

  If the amount of calcium hydroxide is large, that is, if the neutralization of concrete has not progressed, it can be seen that the absorbance at 1.42 μm is large. From these results, it was confirmed that neutralization can be measured with a PCX spectrometer without destroying the concrete.

Salt damage measurement results;
Specimens for salt damage measurement a) Measurement results of cement pace and mortar obtained by difference spectrum are shown in FIGS.

  As shown in FIGS. 17 and 18, in the difference spectrum, it can be understood that the chloride ion concentration and the absorbance do not have a proportional relationship due to the influencing factors, and particularly when the chloride ion concentration is high, the absorbance is low and measurement is impossible.

b) Analysis by chemometrics FIGS. 15 and 16 show the output results by chemometrics on the suspended surface of cement paste powder and cement paste.

  The cement paste powder and the cement paste both showed a result that the correlation spectrum was high at a wavelength of 2.27 μm on both surfaces.

  From the above results, it is difficult to measure the degree of deterioration of calcium carbonate and chloride ions in the conventional difference spectrum, but the degree of deterioration can be accurately measured by using the chemometrics method of the present invention.

  The relationship between neutralization, salt damage and deterioration of this concrete will be explained.

Regarding neutralization;
Neutralization is a reaction in which hydroxide ions (OH ⁻ ) in calcium hydroxide are lost. The fact that the pH of the concrete is lowered means that a protective coating is not easily formed on the surface of the reinforcing bar, so that corrosion of the reinforcing bar is likely to occur. The absorption peak at 1.42 μm belongs to the first overtone of the stretching vibration of the —OH group, and it is known that the absorption peak in this wavelength region is lowered by neutralization. In other words, it can be said that measurement of neutralization can directly measure the amount of calcium hydroxide with a spectroscope.

Regarding salt damage;
Unlike the case of neutralization, the absorption peak due to chloride ions is thought to depend on C ₃ A (tricalcium aluminate) in the cement. For this reason, it is considered that C ₃ A hydrate fixes chloride ions and a peak appears at 2.26 μm. Moreover, present as Friedel said salt is immobilized to have a chloride ion _{(3CaO · Al2O 3 · CaCl 2} · 10H 2 O) in the concrete. However, the following chemical reaction occurs by intervening with carbon dioxide.

3CaO · Al ₂ O ₃ · CaCl ₂ · 10H ₂ O + 3CO ₂ →
3CaCO ₃ + 2Al (OH) ₃ + CaCl ₂ + 7H ₂ O
This reaction is a mechanism by which Friedel's salt near the concrete surface reacts with carbon dioxide and dissolves by neutralization.

  Next, a chemometric method using the PLS regression analysis method of the present invention will be described.

Outline of PLS regression analysis method;
The PLS regression analysis method is an abbreviation for partial least squares, and is shortened to “partial least square method” in Japanese.

  In the PLS regression analysis method, model creation (also referred to as calibration curve creation, calibration, and training) is important, and a large number of standard spectrum data in which the combination is changed depending on the concentration of the target substance and the component that becomes noise is required.

  In order to create an optimal model using such a large amount of spectrum data, a considerable amount of calculation is required. However, recent advances in computers have made it possible to perform calculations even in a small computer in a short time. Currently used for many analyses.

  PLS methods have been announced in various ways, not only when there is only one objective variable, but also a method for obtaining a plurality of objects simultaneously (PLS2 method), a two-dimensional extension method, and a nonlinear PLS method. In addition to absorbance spectra, it can be applied to regression analysis in various fields such as economics and biology.

Basic principle of PLS regression analysis;
In the PLS regression analysis method, it is easier to understand if the collected absorption spectrum is converted into a vector display. The vector display is to indicate the collected wavelengths (200 nm, 201 nm..., Etc.) with the respective axes. In other words, if the wavelength resolution is 100 data, one point can be obtained in a 100-dimensional space. However, since the space of 4 dimensions or more is very difficult to understand, FIG. 2 shows the 3 wavelength data in 3 dimensions.

  In FIG. 2A, data of four points having different concentrations are plotted.

  A is a normal spectrum display method of the horizontal axis wavelength and the vertical axis absorbance, but this can be a vector display as shown in a. (A) is a three-dimensional display for the sake of clarity, but if the number of wavelengths is increased, the number of dimensions increases by the number of wavelengths.

  For these four points, a straight line N that changes the most with respect to the density (maximizes dispersion) and minimizes the residual is obtained. This is the same as turning around Figure 2-i to find the axis where the points appear to be aligned along the density.

  In this way, when an axis as shown in FIG. 2C is found and the graph is rewritten based on the axis N, the density axis N and the noise axis E (exactly speaking, as shown in FIG. Here, the components can be separated into a two-dimensional plane).

  That is, if the N-axis is obtained first, the concentration of the target substance can be calculated without being affected by the noise component by converting the unknown concentration spectrum to the N-axis.

  The above is the basic principle of the PLS regression analysis method.

  Here, the density axis N is considered.

  In order to convert the measured data X to the N-axis, the X and the N-axis unit vector (actually, the vector may be anything other than 0, but is set to 1 for convenience of calculation) (the inner product is multiplied). Take it) The measurement data is converted to the N axis with the unit vector of the N axis as w.

t = Xw
t is the coordinate of x after conversion to the N axis. This t is called a latent constant in the PLS regression analysis. Since t is proportional to N,
y = tq
q is a proportionality constant, and y is the concentration of the target substance.
For this reason, PLS regression analysis is called “regression analysis through latent constant t”.

  w is also called a weight vector. Multiplying the spectrum X by w is equivalent to weighting in the signal processing field.

  In practice, the same calculation is repeated once more from the residual components, not just one conversion, and components proportional to the density are extracted. The first t taken is called the first component (t1), and the weight vector at this time is called (w1). The t taken next is called the second component (t2), and the weight vector at this time is called (w2).

  FIG. 3 shows a structural diagram of the PLS regression analysis.

  In PLS regression analysis, the objective variable is not directly analyzed as an explanatory variable, but once converted to the N-axis (coordinates are t1, t2,...), And the regression analysis is performed again based on that. It has become.

Cross validation:
In PLS regression analysis, the larger the number a of t components, the more linear the regression line and the smaller the prediction error. However, a reproducibility is poor for a calibration curve prepared by taking more components than necessary. This is called overfitting. This is because over-fitting is used to explain regression even for special conditional noise components that occur only when creating a calibration curve.

  Therefore, it is necessary to determine the optimum number of components that minimizes the prediction error as shown in FIG.

There are two major cross-validation methods.
(1) External validation:
In this method, the sample population is divided into two, one for creating a calibration curve and the other for checking.

  The standard error of the calibration curve is called SEC (standard error of calibration), and the standard error for checking is called SEP (standard error of prediction).

  FIG. 5 is a schematic diagram illustrating the implementation of external validation.

  If there is sample data of (1) to (16), create a calibration curve using the data of (1) to (8), and use this calibration curve to generate the data of (9) to (16). Calculate to calculate the prediction error (difference between the calculated value and the measured value).

(2) Internal validation:
This method uses sample data effectively.

  The standard error obtained by this method is called SEV (standard error of validation).

  Internal validation is convenient because it requires fewer samples.

  FIG. 6 is a schematic diagram of internal validation.

  If there is sample data of (1) to (8), first remove (1) from the calibration curve data, create a calibration curve from (2) to (8), calculate (1), and (1) Calculate the prediction error of. Next, remove (2), calculate the prediction error in the same way, and repeat this for all data.

  The optimum number of components is increased by increasing the number of components from 1 and the number of components with the smallest standard error in both methods.

NIPALS method;
There are roughly three calculation algorithms for the PLS regression analysis method. These are the NIPALS method, the solid value method, and the BIDIAGONALIZATION method.

  A general method is the NIPALS method, but Pyroet, which is general-purpose software of Infometrics, uses the BIDIAGONALZATION method.

  Here, the calculation method based on the most general NIPALS method will be described.

PLS model:
When X is an explanatory variable and y is an objective variable, the spectrum is expressed by Formula 1 (In Formula 1, the spectrum is also shown in the matrix).

  In the case of the concentration estimation model based on the absorbance spectrum waveform analysis, x (n, d) is the absorbance at the wavelength d and the measurement number n. y (n) is the concentration at the time of measurement number n. N is the number of measurements (number of samples), and D is the number of wavelength divisions (number of explanatory variables).

  In the PLS method, the following two basic formulas 2 and 3 are considered.

  However, the superscript T means a transposed matrix.

  At this time, T is a latent variable, P is loading, q is a coefficient, E is a residual of X, and f is a residual of y. Furthermore, ta is a latent variable vector of the a component, and pa is a loading vector of the a component, which are expressed by equations 4 to 6, respectively.

  t (n, a) is a latent variable of measurement number n of the a component. A can be selected from 1 to N in terms of the number of components. The features of the model are the components up to the top six, and more than that reduces the prediction error. The optimum component number A is determined by performing cross validation. p (a, d) is the loading of the wavelength d of the a component. Q (a) is a coefficient of the a component.

  In the PLS method, the information of X is not directly used for modeling of y, but a part of the information of X is converted into a latent constant t and y is modeled using t.

Latent constant t:
If ta is a linear combination of X, it is expressed by Equation 7.

  Here, wa is called a weight vector.

   w (d, a) is a weighting coefficient of the wavelength d of the a component.

Calculation of the first component:
First, the case where there is one component (a = 1) is calculated.

  When there is one component a, Equations 2 and 3 become Equations 9 and 10 below.

  From Equation 7, Equation 11 is obtained.

  When the norm of w is set to 1, Equation 12 is obtained.

  The PLS model is to increase the correlation between y and t while increasing the variance of t. The condition that satisfies this is the point at which the covariance S of y and t in Equation 13 is maximized.

  Here, a condition that maximizes S under a constraint condition where the norm of w is 1 is obtained using the Lagrange's undetermined multiplier method.

  Since the function G is a function of the variable w, G is partially differentiated with respect to w (d, 1) to obtain the following relationship.

  Multiply both sides of Equation 17 by w (d, 1).

  Furthermore, the sum total about d is taken.

  Here, from the constraint condition of ‖w1‖ = 0,

Since the left side of Equation 16 is defined as S = y ^T t, 2 μ is a value of y ^T t. Therefore, the maximum value of w that maximizes S = y ^T t is given by Equation 21.

  Since the norm of w1 is 1, w is given by Equation 22.

  The latent variable t is obtained by Equation 23.

  The loading vector p1 of Equation 9 is obtained so that the sum of squares of the elements of the residual E of X is minimized.

  The coefficient qa in Expression 10 is obtained as in Expression 25 so that the sum of squares of elements of the residual vector f of y is minimized.

Calculation after the second component:
The model equation for the second component can be written as

  Here, since t1p1T of X is used for modeling of the number of components 1 and t1q1 of y is used for explanation, the remaining information can be replaced with equations 28 and 29.

  Using Xnew and ynew, equations 26 and 27 are

It becomes.

  This is the same formula as Equations 9 and 10, except that the component number is increased by one.

  Therefore, t2, p2, and q2 can be obtained similarly to the first component. By repeating this loop, the third and subsequent components can be calculated.

Calculation of regression vector b The number of required components A model equation that is repeatedly calculated A can be written as follows.

  The model formula for estimating that substituting equation 7 into the latent variable t of equation 33 is

    Substituting t1 = Xw1 into the above formula and summing up with X

  Here, as shown in Equation 36, a certain explanatory vector x is converted into a model expression for estimating the objective variable y.

  Here, b is referred to as a regression vector, and is expressed by Equation 37.

  b is obtained from Equation 35 as follows.

General algorithm:
The PLS1 method algorithm is summarized as shown in FIG.

  FIG. 7 simply describes the flow for determining the optimum number of components.

  In FIG. 7, first, from the calculation start 30 by the PLS method, the component is set to a = 1 to obtain the first component, and then the weight vector wa of the first component described in Equation 22 is calculated 32, and then the wa The latent variable t is calculated 33 based on the above, the loading vector Pa of Expression 24 is calculated 34, the coefficient qa is calculated from Expression 25, and then the second component model described in Expressions 28 and 29 is set 36. Then, the component a is incremented to a = a + 1, and at step 1, it is determined whether there is an operation for the next component. If yes (yes), the operation returns to the operation of the weight vector wa of the second component. After performing the operations 32 to 35, the next component setting 36 is performed and the increment is sequentially performed 37. After step1 and the necessary number of component operations are performed (no), the regression vector b described in the equation 38 is calculated. End 39 That.

  FIGS. 8 to 11 show details of the flowchart of FIG. 7. FIG. 8 is a flowchart for setting at the start, FIG. 9 is a flowchart for moving spectral data to a calculation array, and a PSL weighting function w (a, d). FIG. 10 is a calculation flowchart of the latent variable t, a calculation flowchart of the coefficient q, a calculation flowchart of the loading vector P, and FIG. 11 is a regression vector calculation flowchart.

  The concentration of chloride ions and calcium carbonate can be measured by taking a product with an unknown spectrum from the regression vector b obtained by the above flowchart.

It is a figure which shows one Embodiment of the optical system used for the concrete diagnostic method of this invention. In this invention, it is a figure which shows the principle of PLS. In this invention, it is a figure which shows the structure figure of PLS regression analysis. In this invention, it is a figure which shows the plot of the relationship between the prediction error in PLS regression analysis, and the number of components. In this invention, it is explanatory drawing of the external validation in PLS regression analysis. In this invention, it is explanatory drawing of the internal validation in PLS regression analysis. In this invention, it is a figure which shows the flowchart of the algorithm of PLS1 method. In this invention, it is a figure which shows the setting flowchart at the time of the start of the algorithm of PLS1 method. In this invention, it is a figure which shows the flowchart which transfers the spectrum data of the algorithm of PLS1 method to the arrangement | sequence for calculation, and the calculation flowchart of PSL weighting function w (a, d). In the present invention, it is a figure which shows the calculation flowchart of the latent variable t of the algorithm of PLS1 method, and the calculation flowchart of the coefficient q. In this invention, it is a figure which shows the calculation flowchart of the regression vector of the algorithm of PLS1 method. In this invention, it is a figure which shows the absorption spectrum of the hanging surface of the cement paste which changed the chloride ion density | concentration with 0-20 kg / m < ³ >. In this invention, it is a figure which shows the absorption spectrum of the cement paste powder which changed the chloride ion density | concentration with 0-20 kg / m < ³ >. In this invention, it is a figure which shows the relationship between a calcium carbonate density | concentration of 1.42 micrometers, and a light absorbency. In this invention, it is a figure which shows the analysis result of the chloride ion density | concentration which analyzed the absorption spectrum of the cement-paste powder which changed the chloride ion density | concentration into 0-20 kg / m < ³ > with the chemometrics method. In the present invention, the cement paste with the chloride ion concentration changed from 0 to 20 kg / m ³ is a diagram showing the analysis result of the chloride ion concentration obtained by analyzing the absorption spectrum of the suspended surface by a chemometric method. It is a figure which shows the analysis result of the chloride ion density | concentration which analyzed the absorption spectrum of the cement-paste powder which changed the chloride ion density | concentration with 0-20 kg / m < ³ > conventionally by the difference spectrum technique. FIG. 5 is a diagram showing an analysis result of chloride ion concentration obtained by analyzing the absorption spectrum of a suspended surface with a difference spectrum method in a conventional cement paste with a chloride concentration changed from 0 to 20 kg / m ³ .

Explanation of symbols

DESCRIPTION OF SYMBOLS 10 Light source 11 Concrete surface 13 Optical fiber 16 MEMS
17 Photodetector

Claims (3)

  In a method for diagnosing concrete deterioration by irradiating the concrete surface with near-infrared light and spectroscopically analyzing the light reflected from the concrete surface in a predetermined wavelength range, the collected absorption spectrum is used to determine the deterioration component using a chemometrics method. A method for diagnosing concrete characterized by detecting.   The method for diagnosing concrete according to claim 1, wherein the deterioration is diagnosed by detecting the concentration of calcium carbonate from the absorption spectrum. The method for diagnosing concrete according to claim 1, wherein the deterioration is diagnosed by detecting the concentration of chloride ions from the absorption spectrum.

Images