CN105424634A - Water quality COD detector based on optical fiber coupling ultraviolet light source and prediction model optimization system of water quality COD detector - Google Patents

Abstract

The invention relates to a water quality COD detector based on an optical fiber coupling ultraviolet light source and a prediction model optimization system of the water quality COD detector. The water quality COD detector is characterized by comprising an excitation light source and signal emission module (1), a detection pool (2) and a receiving module (3); the LED ultraviolet light source with the central wavelength of 304 nm is coupled through an optical fiber, and light emitted by the LED ultraviolet light source enters the detection pool through a lens; the detection pool is composed of two cuvettes with reflectors arranged on the side faces, and the reflectors can reflect exciting fluorescence and improve the light intensity of output fluorescence; fluorescence signals are transmitted to the receiving module through a receiving optical fiber; the included angle between the receiving optical fiber and an excitation light source is 90 degrees so that disturbance caused by exciting light to the fluorescence signals can be reduced as far as possible; the fluorescence signals enter a spectrograph through a silica optical fiber to be detected, and spectroscopic data are transmitted into a computer in real time to be processed. A spectroscopic method is adopted for the detector, and the detector has the advantages of being fast to use and capable of performing detection repeatedly.

Description

A kind of water-quality COD detecting device based on coupling fiber ultraviolet source and forecast model optimization system thereof

Technical field

The present invention relates to a kind of water-quality COD detecting device based on coupling fiber ultraviolet source and forecast model optimization system thereof, described detecting device can be used for water-quality COD determination and analysis in water body, belongs to sensor field.

Background technology

Chemical oxygen demand (COD) (ChemicalOxygenDemand, COD) be when chemically measuring that in water sample, organism is oxidized a great deal of of oxygen that consumes by strong oxidizer, in order to represent the number of Organic substance in water amount, it reflects in water by the degree that reducing substances pollutes.According to China's water quality inspection technique requirement, COD is one of most important index weighing water quality condition at present, is also must survey project in water quality monitoring.Along with printing and distributing of 2015 " carrying out most stringent water resources management system assessment mode embodiment ", complete comprehensively and carry out most stringent water resources management system assessment mode 2013 years, COD discharge capacity obtains effective control, but its larger discharge capacity still can not be ignored, about the principle of COD detection technique and the research and development of technology remain the pith of modern water quality detection research.

The detection of current domestic water chemical oxygen demand remains according to adopting GB11914-89 " mensuration of dichromate titration water chemical oxygen demand " method to measure. and it carries out chemistry titration after measuring and needing sampling, the content of determinand is indirectly obtained based on the reaction between material, testing process very complicated, also can cause secondary pollution simultaneously.In recent years, many optical detecting methods based on pure physics had carried out large quantifier elimination both at home and abroad, and confirmed feasible by the experimental verification of some reality.Secondary pollution can not be produced owing to not needing chemical reagent, there is again the feature of the fast and duplicate detection of detection speed, be applicable to real-time online detection.The technology detecting organic contamination according to the difference of material absorbing spectrum is come out already, and ultraviolet absorptivity is also by overall target that some country is detected as organic contamination.

The present invention devises a coupling fiber ultraviolet LED light source on the basis of the technology reported, intend selecting centre wavelength to be that the ultraviolet B radiation wave band high-capacity LED of 307nm is as light source, the content of COD in water is measured by the principle that the absorption spectrum of composition each in water body is different, add again forecast model optimized algorithm thus be expected to develop convenient, efficiently, the COD detection system that precision is high, and by comparing with existing standard COD value that method is surveyed and calibrate, obtain complete comparing system, to carry out the detection of COD content in water, thus build this invention.Detecting device provided by the present invention can sample on a small quantity when detecting and just obtain higher accuracy of measurement, and analytic process takes shorter simultaneously.

Summary of the invention

The object of the present invention is to provide a kind of water-quality COD detecting device based on coupling fiber ultraviolet LED and forecast model optimization system thereof, the detecting device provided is for the detection of COD in natural water, its forecast model optimization system then provides quick computing system accurately, and the convenient degree of height of formation, the detection system of pinpoint accuracy have good promotional value.

The ultimate principle of detector portion of the present invention is Lambert-Beer's law, refer to when a branch of collimated monochromatic ligth by evenly, the lean solution of non-scatter time, solution is directly proportional to the product of the degree of absorption of light to the concentration of solution and liquid layer thickness.Most of organism in water body has absorption characteristic in ultraviolet region, there is conjugated double bond, the compound of phenyl ring equiconjugate system has absorption at ultraviolet band, as benzene homologues, pyridine, phenol etc.Although the method cannot measure all organism, the stupid system thing etc. that some chemical titrations cannot record can be detected, be the more objective comprehensive Organic substance in water detection method of one.

A kind of uv absorption water-quality COD detecting device provided by the present invention is primarily of excitation source and signal emission module and the composition such as detection cell and receiver module, wherein excitation source and signal emission module (), comprise ultraviolet LED drives plate, 304nm bandpass filter, signal optical fibre; Detection cell (two), comprises 304nm bandpass filter, the quartz colorimetric utensil of 10mm × 10mm, catoptron, catoptron; Receiver module (three), comprises 304nm bandpass filter and receives optical fiber.Wherein excitation source and signal emission module () and receiver module (three) are in stainless steel casing, the 304nm bandpass filter of detection cell (two) and two panels catoptron are also in stainless steel casing, cuvette is design for disassembly, can be unloaded rear loading sample in reinstall.

Concrete each parts are described below:

Ultraviolet LED driving circuit:

LED needs to adopt constant current source to drive, and system adopts 12V DC power supply, carries out voltage-regulation by LM7806.Constant current output is realized by MHL7136.MHL7136 is a LED linear step-down constant current integrated circuit, and its input voltage, at 2.7 ~ 18V, can provide output current adjustable between 10mA ~ 1A.The both positive and negative polarity of LED is connected on vdd terminal and the LED end of MHL7136 respectively.Ultraviolet LED purchased from American SETI company, model is LED-BL-305.

Reflective mirror:

Two proceeds posterolateral of cuvette add the reflective mirror of ultraviolet high reflectance to reflect the fluorescence inspired, and strengthen the intensity exporting fluorescence.

Stainless steel waterproof case:

(1), (three) module of detecting device and the partial content of (two) module encapsulate with stainless steel casing, connect with the optical fiber with SMA interface between each module, and in stainless interface waterproof rubber seal in addition, guarantee safety and the accuracy rate of instrument.

Forecast model of the present invention is optimized traditional algorithm, propose a kind of optimization method predicting water-quality COD model based on particle cluster algorithm joint least-squares support vector (PSO_LSSVM), and then by introducing PCA algorithm, dimensionality reduction pre-service is carried out to improve the speed of convergence of model to mode input spectroscopic data.The precision overcoming LSSVM model is lower, generalization ability and the shortcoming such as robustness is poor.

4) water-quality COD detecting device of the present invention is owing to adopting ultraviolet LED as light source, and the intensity of light source is stablized, can repetitive measurement, and antijamming capability is strong, and long service life is adaptable, has very wide prospect.

Accompanying drawing explanation

Fig. 1 be the embodiment of the present invention based on coupling fiber ultraviolet LED detecting device general structure schematic diagram, in figure, example is described as follows:

1. ultraviolet LED drives plate 2.304nm bandpass filter 3. signal optical fibre

Quartz colorimetric utensil 6. catoptron of 4.304nm bandpass filter 5.10mm × 10mm

7. catoptron 8.304nm bandpass filter 9. receives optical fiber

Fig. 2 is the ultraviolet LED driving circuit figure of the embodiment of the present invention

Fig. 3 is PSO_LSSVM algorithm flow schematic diagram

Fig. 4 is the curve of actual data measured and forecast model

Embodiment

Set forth substantive distinguishing features of the present invention and significant progress further below in conjunction with accompanying drawing, but the present invention is absolutely not only confined to embodiment.

Embodiment 1:

The structure of the water-quality COD detecting device based on coupling fiber ultraviolet LED provided by the present invention as shown in Figure 1.The described water-quality COD detecting device based on coupling fiber ultraviolet LED comprises primarily of described excitation source and signal emission module unit: ultraviolet LED drives plate 1,304nm bandpass filter 2, signal optical fibre 3; Detection cell module comprises: the quartz colorimetric utensil 5 of 304nm bandpass filter 4,10mm × 10mm, catoptron 6, catoptron 7; Receiver module comprises lens 8, receives optical fiber 9; Signal optical fibre 3 and reception optical fiber 9 angle in 90 °; Whole parts of excitation source and signal emission module and receiver module and detection cell be partially encapsulated in stainless steel casing, by the Fiber connection with sub-miniature A connector between each module; Sample to be tested is inserted in the cuvette of detection cell module, after switching on power, signal optical fibre coupling excitation source also enters cuvette, receives optical fiber and collects the ultraviolet light after Organic substance in water absorbs, be conducted through 304nm bandpass filter and enter spectrometer, be transferred in computing machine.

Embodiment 2:

1) to model PCA dimension-reduction treatment

Spectrometer is the USB4000 micro fiber spectrometer of OceanOptics company, and every bar spectroscopic data point is 1204.After obtaining data, first, denoising, turbidity correction process are carried out to the spectroscopic data obtained; Then, PCA algorithm is adopted to carry out redundant information rejecting to the input amendment matrix formed.PCA analyze concrete steps as shown in Figure 3:

For given sample set (x ₁, y ₁), (x ₂, y ₂), (x ₃, y ₃) ..., (x _n, y _n), x _i∈ R ⁿfor n ties up input vector, y _ifor one dimension output vector, n is sample number, and LSSVM is constructed as follows and minimizes objective function and constraint condition thereof:

$\{\begin{array}{c}\min j(\mathrm{\ω},e)=\frac{1}{2}{\mathrm{\ω}}^T\mathrm{\ω}+\mathrm{\γ}\frac{1}{2}\underset{K=1}{\overset{N}{\Σ}}{e}_k^2\\ y_k={\mathrm{\ω}}^T\mathrm{\φ}\left(x_k\right)+b+e_k,k=1,2,...,N\end{array}---\left(1\right)$

In formula (1), ω ∈ R ⁿ, be weight vector, ω ^tfor the transposition of ω; γ is regularization parameter; e _kfor error variance; φ (x _k) be function input vector being mapped to high-dimensional feature space; J (ω, e) is Lagrange function; B ∈ R is offset parameter. the system of linear equations under the KKT of corresponding Lagrange function is:

$\left[\begin{array}{cc}0& I^T\\ I& K+{\mathrm{\γ}}^{-1}I\end{array}\right]\left[\begin{array}{c}b\\ \mathrm{\α}\end{array}\right]=\left[\begin{array}{c}0\\ Y\end{array}\right]---\left(2\right)$

In formula (2), a=[a ₁..., a _n] ^t; Y=[y ₁..., y _n] ^t; K=φ (x) ^tφ (x _k)=K (x, x _k), k=1,2 ... n; I=[1,1 ..., 1] ^t, I ^tfor the transposition of I; Utilize least-squares algorithm to solve formula (3) system of linear equations, obtain the value of a and b, thus, obtain the decision function of forecast model:

$y_k=\underset{k=1}{\overset{n}{\Σ}}{\mathrm{\α}}_{k}K(x,x_k)+b---\left(3\right)$

(3) in formula, k must meet

$\underset{i=1}{\overset{k}{\Σ}}{\[\mathrm{\φ}\left(x_i\right)-y_i\]}^2=\underset{i=1}{\overset{k}{\Σ}}{\[\underset{j=0}{\overset{n}{\Σ}}{\mathrm{\α}}_j{\mathrm{\φ}}_J\left(x_i\right)-y_i\]}^2---\left(4\right)$

2) precision of forecasting model analysis

Precision of forecasting model dissecting needle is to 50 groups of experiment samples, and random selecting 40 groups is as model training sample mould, and 10 groups as forecast model.In the target search space of a D dimension, form colony by N number of particle, suppose that the position vector of i-th particle is x _i=(x _i1, x _i2... x _iD), velocity vector is v _i=(v _i1, v _i2..., v _iD), evaluate its quality according to the fitness value of each particle, and find the individual extreme value p of current time _i=(p _i1, p _i2... p _iD) and global extremum p _g=(p _g1, p _g2... p _gD); For the t time iteration, its d ties up (1≤d≤D) according to lower column format iteration:

$x_i^{t+1}=\frac{1}{\mathrm{\α}}\[b_i-\underset{j=1}{\overset{i-1}{\Σ}}{\mathrm{\α}}_{ij}{x}_{ij}-\underset{j=i+1}{\overset{n}{\Σ}}{\mathrm{\α}}_{ij}{x_j}^{\left(t\right)}\]---\left(5\right)$

Upgraded by following formula:

v _id(t+1)＝w·v _id(t)+c ₁r ₁[p _id-x _id(t)]+c ₂r ₂[g _id-x _id(t)]x _id(t+1)＝x _id(t+1)+v _id(t+1)(6)

In formula (4), r ₁and r ₂for the random number between [0,1]; c ₁and c ₂for Studying factors; W is inertia weight. at every one dimension, particle has a maximum restriction speed v _maxif the speed of certain one dimension exceedes the v of setting _max, so, the speed of this one dimension just equals v _max; Inertia weight is commonly defined as w, then

$w=w_{max}-\frac{\left(w_{\max }-w_{min}\right)\·t_{\max }}{t}---\left(7\right)$

T _maxfor total iterations, t is current iteration number of times, w _max, w _minbe respectively minimum and maximum weight factor.

In this embodiment, the optimum configurations of PSO is: r ₁and r ₂for the random number between [0,1], c ₁=c ₂=1.8, D=2, maximum iteration time t _max=1000, w _max=0.94, w _min=0.4. algorithm adopts standard C language to write, and uses MATLAB7.0 version to compile.Finally obtain C=268.354, σ=10.489, finally obtain the curve of actual data measured and forecast model as Fig. 4.

Claims (7)

1. the water-quality COD detecting device based on coupling fiber ultraviolet source and forecast model optimization system thereof, it is characterized in that it is a kind of coupling fiber burst of ultraviolel water-quality COD fluorescence detector, it is made up of excitation source and signal emission module () and detection cell (two) and receiver module (three) three part; Centre wavelength be the ultraviolet LED light source of 304nm by coupling fiber, its light sent enters detection cell through lens; The cuvette that detection cell is equipped with catoptron by two sides forms, and catoptron can reflect the fluorescence excited, and strengthens the light intensity exporting fluorescence; Fluorescence signal is by receiving Optical Fiber Transmission to receiver module; Accept optical fiber and excitation source angle in 90 °, to reduce exciting light to the interference of fluorescence signal as far as possible; Fluorescence signal is incided in spectrometer by a silica fibre and detects, and spectroscopic data is sent in computing machine in real time and processes.

2. detecting device according to claim 1, is characterized in that: described excitation source and signal emission module () comprising: ultraviolet LED drives plate 1,304nm bandpass filter 2, signal optical fibre 3;

Described detection cell unit comprises: the quartz colorimetric utensil 5 of 304nm bandpass filter 4,10mm × 10mm, catoptron 6, catoptron 7;

Described receiver module comprises lens 8, receives optical fiber 9;

Described signal optical fibre 3 and receive optical fiber 9 and be all silica fibre with SMA905 interface and mutual angle in 90 °, LED needs to adopt constant current source to drive, and system adopts 12V DC power supply, carries out voltage-regulation by LM7806.Constant current output is realized by MHL7136.MHL7136 is a LED linear step-down constant current integrated circuit, and its input voltage, at 2.7 ~ 18V, can provide output current adjustable between 10mA ~ 1A.The both positive and negative polarity of LED is connected on vdd terminal and the LED end of MHL7136 respectively.Ultraviolet LED purchased from American SETI company, model is LED-BL-305.

3., based on water-quality COD detecting device and the forecast model optimization system thereof of coupling fiber ultraviolet source, it is characterized in that comprising the following steps:

(1) the detecting device collecting sample described in claim 1 and 2 is utilized;

(2) data contained by sample are carried out data normalization;

(3) 1# selects the parameter that LSSVM prediction model parameters is optimized;

(4) train, renormalization solving model;

(5) Output rusults.

4. a kind of water-quality COD detecting device based on coupling fiber ultraviolet source according to claim 3 and forecast model optimization system thereof, it is characterized in that utilizing PCA principal component analytical method that multivariable challenge is reduced to few variable problem, concrete steps are as follows:

(1) M N is tieed up sample, composition sample matrix, carries out data normalization;

(2) sample matrix correlation matrix is calculated;

(3) proper vector and the characteristic of correspondence value thereof of correlation matrix is calculated;

(4) calculate eigenvalue contribution rate, determine major component.

5. a kind of water-quality COD detecting device based on coupling fiber ultraviolet source according to claim 3 and forecast model optimization system thereof, it is characterized in that, step 1# comprises with lower part:

(1) initializing set is carried out to population;

(2) each particle is substituted into LSSVM model to calculate according to model Output rusults, the adaptive value of each particle;

(3) more new particle extreme value and global extremum; More the inertia weight of new particle, speed and position are until meet end condition;

(4) optimized parameter is exported.

6. a kind of water-quality COD detecting device based on coupling fiber ultraviolet source according to claim 3 and forecast model optimization system thereof, is characterized in that for given sample set (x ₁, y ₁), (x ₂, y ₂), (x ₃, y ₃) ..., (x _n, y _n), x _i∈ R ⁿfor n ties up input vector, y _ifor one dimension output vector, n is sample number, and LSSVM is constructed as follows and minimizes objective function and constraint condition thereof:

$\left\{\begin{array}{c}\min J(\mathrm{\ω},e)=\frac{1}{2}{\mathrm{\ω}}^T\mathrm{\ω}+\mathrm{\γ}\frac{1}{2}\underset{K=1}{\overset{N}{\Σ}}{e}_k^2\\ y_k={\mathrm{\ω}}^T\mathrm{\φ}\left(x_k\right)+b+e_k,k=1,2,...,N\end{array}\right.---\left(1\right)$

In formula (1), ω ∈ R ⁿ, be weight vector, ω ^tfor the transposition of ω; γ is regularization parameter; e _kfor error variance; φ (x _k) be function input vector being mapped to high-dimensional feature space; J (ω, e) is Lagrange function; B ∈ R is offset parameter. the system of linear equations under the KKT of corresponding Lagrange function is:

$\left[\begin{array}{cc}0& I^T\\ I& K+{\mathrm{\γ}}^{-1}I\end{array}\right]\left[\begin{array}{c}b\\ \mathrm{\α}\end{array}\right]=\left[\begin{array}{c}0\\ Y\end{array}\right]---\left(2\right)$

In formula (2), α=[α ₁..., α _n] ^t; Y=[y ₁..., y _n] ^t; K=φ (x) ^tφ (x _k)=K (x, x _k), k=1,2 ... n; I=[1,1 ..., 1] ^t, I ^tfor the transposition of I; Utilize least-squares algorithm to solve formula (2) system of linear equations, obtain the value of α and b, thus, obtain the decision function of forecast model:

(3) in formula, k must meet

$\underset{i=1}{\overset{k}{\Σ}}{\[\mathrm{\φ}\left(x_i\right)-y_i\]}^2=\underset{i=1}{\overset{k}{\Σ}}{\[\underset{j=0}{\overset{n}{\Σ}}{\mathrm{\α}}_j{\mathrm{\φ}}_j\left(x_i\right)-y_i\]}^2---\left(4\right)$

7. the optimization system of a kind of water-quality COD forecast model according to claim 3, in the target search space of a D dimension, forms colony by N number of particle, supposes that the position vector of i-th particle is x _i=(x _i1, x _i2... x _iD), velocity vector is v _i=(v _i1, v _i2..., v _iD), evaluate its quality according to the fitness value of each particle, and find the individual extreme value p of current time _i=(p _i1, p _i2... p _iD) and global extremum p _g=(p _g1, p _g2... p _gD); For the t time iteration, its d ties up (1≤d≤D) according to lower column format iteration:

$x_i^{t+1}=\frac{1}{\mathrm{\α}}\[b_i-\underset{j=1}{\overset{i-1}{\Σ}}{\mathrm{\α}}_{ij}{x}_{ij}-\underset{j=i+1}{\overset{n}{\Σ}}{\mathrm{\α}}_{ij}{x_j}^{\left(t\right)})---\left(5\right)$

Upgraded by following formula:

v _id(t+1)＝w·v _id(t)+c ₁r ₁[p _id-x _id(t)]+c ₂r ₂[g _id-x _id(t)]x _id(t+1)＝x _id(t+1)+v _id(t+1)(6)

In formula (4), r ₁and r ₂for the random number between [0,1]; c ₁and c ₂for Studying factors; W is inertia weight. at every one dimension, particle has a maximum restriction speed v _maxif the speed of certain one dimension exceedes the v of setting _max, so, the speed of this one dimension just equals v _max; Inertia weight is commonly defined as w, then

$w=w_{max}-\frac{(w_{\max }-w_{\min })\·t_{max}}{t}---\left(7\right)$

T _maxfor total iterations, t is current iteration number of times, w _max, w _minbe respectively minimum and maximum weight factor.