CN114295704A - Micro-concentration gradient solution electrochemical determination method based on characteristic parameter extraction - Google Patents
Micro-concentration gradient solution electrochemical determination method based on characteristic parameter extraction Download PDFInfo
- Publication number
- CN114295704A CN114295704A CN202111671385.0A CN202111671385A CN114295704A CN 114295704 A CN114295704 A CN 114295704A CN 202111671385 A CN202111671385 A CN 202111671385A CN 114295704 A CN114295704 A CN 114295704A
- Authority
- CN
- China
- Prior art keywords
- peak
- curve
- reduction
- oxidation
- potential
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
- 238000000034 method Methods 0.000 title claims abstract description 63
- 238000000605 extraction Methods 0.000 title claims abstract description 37
- 238000001514 detection method Methods 0.000 claims abstract description 82
- 238000002484 cyclic voltammetry Methods 0.000 claims abstract description 43
- 238000001903 differential pulse voltammetry Methods 0.000 claims abstract description 34
- 238000000835 electrochemical detection Methods 0.000 claims abstract description 27
- 238000000970 chrono-amperometry Methods 0.000 claims abstract description 23
- 238000012549 training Methods 0.000 claims abstract description 7
- 238000006722 reduction reaction Methods 0.000 claims description 75
- 238000007254 oxidation reaction Methods 0.000 claims description 64
- 230000003647 oxidation Effects 0.000 claims description 60
- 238000001914 filtration Methods 0.000 claims description 13
- -1 potassium ferricyanide Chemical compound 0.000 claims description 11
- 239000000276 potassium ferrocyanide Substances 0.000 claims description 11
- XOGGUFAVLNCTRS-UHFFFAOYSA-N tetrapotassium;iron(2+);hexacyanide Chemical compound [K+].[K+].[K+].[K+].[Fe+2].N#[C-].N#[C-].N#[C-].N#[C-].N#[C-].N#[C-] XOGGUFAVLNCTRS-UHFFFAOYSA-N 0.000 claims description 11
- 229910021607 Silver chloride Inorganic materials 0.000 claims description 5
- 239000000376 reactant Substances 0.000 claims description 5
- 238000011946 reduction process Methods 0.000 claims description 5
- HKZLPVFGJNLROG-UHFFFAOYSA-M silver monochloride Chemical compound [Cl-].[Ag+] HKZLPVFGJNLROG-UHFFFAOYSA-M 0.000 claims description 5
- 238000001318 differential pulse voltammogram Methods 0.000 claims description 4
- BASFCYQUMIYNBI-UHFFFAOYSA-N platinum Chemical compound [Pt] BASFCYQUMIYNBI-UHFFFAOYSA-N 0.000 claims description 4
- 239000010931 gold Substances 0.000 claims description 3
- 229910052737 gold Inorganic materials 0.000 claims description 3
- 238000005259 measurement Methods 0.000 abstract description 7
- 238000005065 mining Methods 0.000 abstract description 4
- 239000000243 solution Substances 0.000 description 37
- 238000009499 grossing Methods 0.000 description 13
- 238000005070 sampling Methods 0.000 description 12
- 238000006479 redox reaction Methods 0.000 description 11
- 238000005516 engineering process Methods 0.000 description 8
- 238000012360 testing method Methods 0.000 description 7
- 238000004458 analytical method Methods 0.000 description 6
- 238000006243 chemical reaction Methods 0.000 description 5
- 238000010586 diagram Methods 0.000 description 5
- 230000035945 sensitivity Effects 0.000 description 5
- 239000000284 extract Substances 0.000 description 4
- 241000588724 Escherichia coli Species 0.000 description 3
- 238000004364 calculation method Methods 0.000 description 3
- 238000009826 distribution Methods 0.000 description 3
- 230000000694 effects Effects 0.000 description 3
- 238000002848 electrochemical method Methods 0.000 description 3
- 238000012545 processing Methods 0.000 description 3
- 238000011160 research Methods 0.000 description 3
- 239000000126 substance Substances 0.000 description 3
- 230000000007 visual effect Effects 0.000 description 3
- 102100035932 Cocaine- and amphetamine-regulated transcript protein Human genes 0.000 description 2
- 101000715592 Homo sapiens Cocaine- and amphetamine-regulated transcript protein Proteins 0.000 description 2
- 239000011159 matrix material Substances 0.000 description 2
- 238000007781 pre-processing Methods 0.000 description 2
- 239000003115 supporting electrolyte Substances 0.000 description 2
- 239000013598 vector Substances 0.000 description 2
- 238000012800 visualization Methods 0.000 description 2
- 229910004042 HAuCl4 Inorganic materials 0.000 description 1
- 238000007605 air drying Methods 0.000 description 1
- 239000003242 anti bacterial agent Substances 0.000 description 1
- 229940088710 antibiotic agent Drugs 0.000 description 1
- 230000001580 bacterial effect Effects 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 239000000872 buffer Substances 0.000 description 1
- 239000007853 buffer solution Substances 0.000 description 1
- 238000007405 data analysis Methods 0.000 description 1
- 230000003247 decreasing effect Effects 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 238000000151 deposition Methods 0.000 description 1
- 230000008021 deposition Effects 0.000 description 1
- 238000009792 diffusion process Methods 0.000 description 1
- 238000000840 electrochemical analysis Methods 0.000 description 1
- 238000004070 electrodeposition Methods 0.000 description 1
- 239000008151 electrolyte solution Substances 0.000 description 1
- 239000002778 food additive Substances 0.000 description 1
- 235000013373 food additive Nutrition 0.000 description 1
- 229910021397 glassy carbon Inorganic materials 0.000 description 1
- PCHJSUWPFVWCPO-UHFFFAOYSA-N gold Chemical compound [Au] PCHJSUWPFVWCPO-UHFFFAOYSA-N 0.000 description 1
- 238000011478 gradient descent method Methods 0.000 description 1
- 239000003112 inhibitor Substances 0.000 description 1
- 230000010354 integration Effects 0.000 description 1
- 238000004519 manufacturing process Methods 0.000 description 1
- 230000033116 oxidation-reduction process Effects 0.000 description 1
- 230000000737 periodic effect Effects 0.000 description 1
- 230000010287 polarization Effects 0.000 description 1
- 238000004445 quantitative analysis Methods 0.000 description 1
- 238000012546 transfer Methods 0.000 description 1
- 238000005406 washing Methods 0.000 description 1
- XLYOFNOQVPJJNP-UHFFFAOYSA-N water Substances O XLYOFNOQVPJJNP-UHFFFAOYSA-N 0.000 description 1
Images
Landscapes
- Investigating Or Analysing Biological Materials (AREA)
Abstract
The invention discloses a micro-concentration gradient solution electrochemical determination method based on characteristic parameter extraction. The method comprises the following steps: firstly, cyclic voltammetry detection, chronoamperometry detection and differential pulse voltammetry detection are carried out on a plurality of solutions to be detected with micro concentration gradients, and cyclic voltammetry detection data, chronoamperometry detection data and differential pulse voltammetry detection data of different solutions to be detected are obtained. And secondly, automatically extracting the graphic features of the obtained detection data. And thirdly, obtaining an input data set by using the characteristic value and the corresponding concentration label. And fourthly, constructing a concentration measurement model, and training the concentration measurement model by using the input data set obtained in the third step. The trained concentration determination model is used for detecting the concentration of the solution to be detected with unknown concentration. Aiming at the problems of overlapping electrochemical detection curves and high noise of a micro-concentration gradient detector, the invention adopts a plurality of detection methods to carry out parallel comparison, thereby realizing effective information mining and accurate concentration determination of detection signals.
Description
Technical Field
The invention belongs to a rapid detection method of a micro-concentration gradient solution, and particularly relates to a rapid detection method of a micro-concentration gradient solution based on an electrochemical detection technology and a characteristic parameter extraction integrated model.
Background
In recent years, in the fields of chemicals, pharmaceutical production, food safety and the like, on-site concentration measurement of various micro-concentration gradient solutions, such as antibiotics, bacterial inhibitors, food additives and the like, is required. Electrochemical detection technology is increasingly applied to feature substance concentration determination by virtue of the advantages of simple operation, high sensitivity and the like. The method is a technology for analyzing and processing electric signals by collecting electrochemical response signals, converting the electrochemical response signals into the electric signals which can be identified and detected, and commonly used electrochemical detection methods comprise cyclic voltammetry, chronoamperometry, differential pulse voltammetry and the like. In the electrochemical detection application process, the portable potentiostat is an indispensable instrument in electrochemical field test, and can control the electrode potential as a set value so as to achieve the purpose of constant potential polarization. Compared with the traditional electrochemical workstation, the portable potentiostat system based on the method can realize the rapid detection of cyclic voltammetry of trace features, and simultaneously, the problem of low sensitivity coefficient of the functional module is solved. The low sensitivity of the portable potentiostat system based on field detection causes the detection signal to have high background noise, and simultaneously causes the problems of overlapping detection curves of micro-concentration gradient characteristic substance solutions, unclear peak shape and peak height characteristics and the like in the operation process of different detection personnel. The above problems present a significant challenge to achieving rapid identification of microscale feature solutions.
In order to solve the problems and realize the on-site rapid determination of the micro-concentration gradient solution, an integrated model for extracting the characteristic parameters of electrochemical detection signals is provided. The model integrates data preprocessing, feature extraction, feature reduction and concentration measurement, and compared with the traditional peak height and concentration fitting method, the method improves the analysis efficiency of electrochemical detection signals, obtains more information contained in the detection signals, and improves the accuracy of concentration measurement. At present, concentration determination research aiming at electrochemical detection data signals is more, but information mining about on-site rapid electrochemical detection signals and electrochemical rapid determination research in micro-concentration gradient solution still have more quantitative analysis needs to be perfected.
Disclosure of Invention
The invention mainly aims to realize a method for rapidly detecting a micro-concentration gradient solution on site based on an electrochemical detection technology and a feature extraction integrated model.
The method for determining the micro-concentration gradient solution based on the feature extraction integrated model comprises the following steps:
step one, performing cyclic voltammetry detection, chronoamperometry detection and differential pulse voltammetry detection on multiple detected solutions with micro concentration gradients to obtain cyclic voltammetry detection data, chronoamperometry detection data and differential pulse voltammetry detection data of different multiple detected solutions.
And step two, carrying out graphic feature extraction on the cyclic voltammetry detection data, the timing current detection data and the differential pulse voltammetry detection data obtained in the step one. Wherein the feature extracted from the cyclic voltammetry curve comprises the oxidation peak current IopReduction peak current IrpOxidation peak potential EopReduction peak potential ErpBaseline slope K of the oxidation curveoReduction curve baseline slope KrArea of oxidation peak SoReduction peak area SrInitial reduction potential VirAnd initial oxidation potential Vio. The characteristics extracted from the chronoamperometric curve include the initial steady-state current time tmAnd steady state current Is. The features extracted from the differential pulse voltammogram include the peak potential EpSum peak current Ip。
Thirdly, sorting the characteristic values extracted in the second step and the corresponding concentration labels into an initial data set; and reducing the dimension of the obtained initial data set to obtain an input data set.
And step four, constructing a concentration measurement model, and training the concentration measurement model by using the input data set obtained in the step three. The trained concentration determination model is used for detecting the concentration of the solution to be detected with unknown concentration.
Preferably, the oxidation peak current I is the characteristic extracted from the cyclic voltammetry curveopThe height from the anode current peak value on the cyclic voltammetry I-V curve to the oxidation curve baseline; reduction peak current IrpAs the cathode current peak to reduction curve baselineThe height of (d); oxidation peak potential EopIs the anode current peak position; reduction peak potential ErpIs the cathode current peak position; baseline slope K of oxidation curveoThe slope of the baseline for oxidation; reduction curve baseline slope KrThe slope of the baseline of the reduction curve; area of oxidation peak SoThe area enclosed by the oxidation curve, the base line and the distance from the oxidation peak to the base line; reduction peak area SrThe area is the area enclosed by the reduction curve, the base line and the distance from the reduction peak to the base line; initial reduction potential VirIs the initial potential at which the reactants undergo a reduction reaction; initial oxidation potential VioIs the initial potential at which the reactants undergo oxidation reactions.
Initial steady state current time t in features extracted from chronoamperometric curvesmThe initial time for reaching the steady-state current on the I-t curve of the timing current method; steady state current IsIs the median after the initial steady state current on the I-t curve.
Among the features extracted from the differential pulse voltammogram, the peak potential EpThe anode current peak position on the differential pulse voltammetry I-V curve is shown; peak current IpThe height from the anode current peak to the baseline of the curve.
Preferably, in the second step, the process of automatically extracting the features of the cyclic voltammetry data is as follows:
the first step is as follows: the cyclic voltammetry detection data is equally divided into an upper part and a lower part, wherein the upper part is oxidation process data, and the lower part is reduction process data.
The second step is that: separately determining a fitting function f for the oxidation process data and the reduction process datao(x) And fr(x) Obtaining an oxidation curve and a reduction curve; respectively taking the maximum value and the minimum value in a preset range on an oxidation curve and a reduction curve, wherein the maximum value potential of the oxidation curve is the oxidation peak potential IopThe minimum potential of the reduction curve is the reduction peak potential Irp。
The third step: respectively solving first derivative of fitting function before peak potential in oxidation curve and reduction curve, and taking median point of two first derivatives as baseline slope, respectively oxidation curveLine base slope KoAnd reduction curve baseline slope Kr. Meanwhile, the original potentials of the two median points are initial response potentials which are respectively initial oxidation potentials VioAnd initial reduction potential Vir。
The fourth step: determining the oxidation curve baseline L in the smooth curve by taking the initial response potential as the tangent point and adding the baseline slope obtained in the third stepo(x) And reduction curve baseline Lr(x)。
The fifth step: the height from the peak position to the base line is taken as the peak current and is taken as the oxidation peak current IopAnd reduction of peak current Irp。
And a sixth step: the difference between the fitting function and the base line is integrated into peak areas from the initial response potential to the peak position, and the peak areas are respectively the oxidation peak area SoAnd reduction peak area Sr。
Preferably, in the step two, the process of automatically extracting the characteristics of the timing current data is as follows:
the first step is as follows: a fitting function f (x) is calculated for the chronoamperometric detection data.
The second step is that: and solving the first derivative of the fitting function to obtain a differential array.
The third step: the differential value of the next data point is used to subtract the previous one to form a differential difference array.
The fourth step: and acquiring a first negative value in the differential difference value array, wherein the point indicates that the detected data is accelerated and decelerated, and meanwhile, the current value starts to reach a steady state, and the position of the data point is determined as initial steady-state current time.
The fifth step: and taking the median of the current values in the range from the initial steady-state current time to the response end as the steady-state current.
Preferably, in the second step, the process of automatically extracting the features of the differential pulse voltammetry data is as follows:
the first step is as follows: a fitting function f (x) is calculated for the differential pulse voltammetry data.
The second step is that: and taking the corresponding position of the maximum value of the current in a preset range as the peak potential.
The third step: and solving a first derivative of the fitting function before the peak potential, taking a median point of the first derivative as a slope of the baseline, and setting the original point as a tangent point between the baseline and the fitting function.
The fourth step: and determining a base line in the smooth curve, and taking the height from the peak position point to the base line as the peak current.
Preferably, the solution to be tested is a potassium ferricyanide/potassium ferrocyanide solution.
Preferably, in the first step, a three-electrode system is constructed in the solution to be tested; the three-electrode system uses a nano-gold modified electrode as a working electrode, an Ag/AgCl electrode as a reference electrode, a platinum wire electrode as an auxiliary electrode, and a portable potentiostat to form a field rapid electrochemical detection platform, so that cyclic voltammetry detection, chronoamperometry detection and differential pulse voltammetry detection are realized.
Preferably, aiming at the cyclic voltammetry I-V curve, the chronoamperometry I-t curve and the differential pulse voltammetry I-V curve obtained in the first step, a recursive average filtering method is used for smoothing and denoising, so that the signal-to-noise ratio of original data is improved, and subsequent feature extraction operation is facilitated.
Preferably, the dimension reduction method in the third step specifically adopts a nonlinear dimension reduction algorithm t-SNE.
Preferably, the concentration determination model described in step four is constructed by the XGBoost algorithm.
The invention has the beneficial effects that:
the invention provides a feature extraction integrated model based on an electrochemical detection technology, aiming at the problems of overlapping electrochemical detection curves and high noise of a micro concentration gradient detector, a plurality of detection methods are adopted for parallel comparison, and effective information mining and accurate concentration determination of detection signals are realized. The electrochemical detection signal based on the field portable electrochemical workstation has high background noise and reduces the detection sensitivity, so that the recursive average filtering method is adopted to carry out smooth denoising treatment on the original detection signal so as to improve the signal-to-noise ratio of the signal, and the subsequent detection data analysis is facilitated. And automatically extracting the graphic features of the original signal after the smoothing treatment, and mining more useful detection signal data information on the basis of the previous research. And (3) performing dimensionality reduction operation on the feature set by using the t-SNE, and finally inputting the dimensionality reduced data set into a prediction model to accurately measure the concentration of the micro-concentration gradient solution. Compared with the traditional electrochemical analysis method, the method has the advantages that data in the test process of various electrochemical methods are quantified innovatively, signal detection, signal preprocessing, signal feature extraction, feature dimension reduction and concentration determination are integrated, the analysis efficiency of electrochemical detection signals is improved through the feature extraction and visual analysis tools in the multi-electrochemical method, more useful information contained in the detection signals is obtained, and the rapid and accurate determination of the micro-concentration gradient solution is realized.
Drawings
FIG. 1 is an overall frame diagram of the present invention;
FIG. 2a is a flowchart of a Cyclic Voltammetry (CV) algorithm;
FIG. 2b is a flowchart of a Chronoamperometry (CA) algorithm;
FIG. 2c is a flow chart of a Differential Pulse Voltammetry (DPV) algorithm;
FIG. 3 is a flowchart of a recursive average filtering algorithm;
FIG. 4a is a characteristic parameter representation of cyclic voltammetry;
FIG. 4b is a graph showing characteristic parameters of a chronometric current curve;
FIG. 4c is a characteristic parameter representation of a differential pulse voltammetry curve;
FIG. 5 is a flow chart of an electrochemical signal feature automatic extraction algorithm;
FIG. 6 is a schematic view of a micro concentration gradient CV curve of potassium ferricyanide/potassium ferrocyanide solution;
FIG. 7a is a schematic representation of a pre-smoothing operation;
FIG. 7b is a schematic diagram after smoothing;
FIG. 8 is a diagram of the visualization effect of t-SNE dimension reduction;
FIG. 9 is a graph of the results of the XGboost model in predicting the concentration of a micro concentration gradient solution;
Detailed Description
The invention is further described below with reference to the accompanying drawings:
the invention provides a method for rapidly detecting a micro-concentration gradient solution on site based on an electrochemical detection technology and a feature extraction integrated model.
As shown in fig. 1, the method for rapidly determining a micro concentration gradient solution by using the feature extraction integration model comprises the following steps:
the method comprises the following steps: a nanogold modified electrode is used as a working electrode, an Ag/AgCl electrode is used as a reference electrode, a platinum wire electrode is used as an auxiliary electrode, the three electrodes form a three-electrode system, and a portable potentiostat is matched to form a field rapid electrochemical detection platform.
Step two: and (3) repeatedly performing cross cyclic voltammetry detection, chronoamperometry detection and differential pulse voltammetry detection on the micro-concentration gradient potassium ferricyanide/potassium ferrocyanide solution by using 0.1M PBS buffer solution as a supporting electrolyte to obtain experimental data.
Step three: aiming at the detection signal, the smoothing and denoising processing is carried out by utilizing a recursive average filtering (MAF) method, so that the signal-to-noise ratio of the original data is improved, and the subsequent feature extraction operation is facilitated.
Step four: and automatically extracting the graphic features of the smoothed original data, and containing more data information as much as possible while ensuring the relative independence between the features.
Step five: and (4) sorting the extracted characteristic values and the corresponding concentrations into a data set, reducing the characteristic dimension by using a nonlinear dimension reduction algorithm t-SNE, and sorting into a model input data set.
Step six: and (5) training a concentration determination model by using the data set processed in the step five, and accurately determining the concentration based on the trained model.
The specific process of the invention is as follows:
the method comprises the following steps: the cleaned Glassy Carbon Electrode (GCE) was immersed in a solution containing 0.1% HAuCl40.1MH of2SO4In the electrolyte solution, electrochemical deposition was carried out in a single potential mode of-200 mV (vs. Ag/AgCl) for a deposition time of 30 s. Taking out, washing with water, and air drying to obtain nanometer gold modified electrode (AuNPs/GCE).
Step two: a micro-concentration gradient potassium ferricyanide/potassium ferrocyanide solution was prepared with 0.1M PBS buffer as supporting electrolyte, with gradient intervals of 1 mM. A nanogold modified electrode is used as a working electrode, an Ag/AgCl electrode is used as a reference electrode, a platinum wire electrode is used as an auxiliary electrode, the three electrodes form a three-electrode system, repeated cyclic voltammetry, a chronoamperometry and differential pulse voltammetry are carried out on a micro-concentration gradient potassium ferricyanide/potassium ferrocyanide solution on the basis of a portable potentiostat electrochemical detection platform, and the three repeated methods are carried out on each concentration under the same condition for 50 times.
Fig. 2 shows a flowchart of cyclic voltammetry, chronoamperometry and differential pulse voltammetry algorithms. The specific algorithm steps for carrying out cyclic voltammetry response by utilizing the portable potentiostat system are as follows:
1) selecting a cyclic voltammetry method on a PC upper computer interface, and inputting relevant parameters, wherein the parameters comprise: initial potential Init E, upper limit potential High E, lower limit potential Low E, Scan speed Scan Rate, Scan segment number Sweep Segments, sampling Interval Sample Interval, etc.;
2) applying a linear alternating initial voltage to the two ends of the working electrode and the auxiliary electrode through a DAC module;
3) and (3) carrying out forward scanning on the oxidation-reduction reaction in the escherichia coli solution at a preset scanning speed V (V/s), and recording current and potential information of the oxidation-reduction reaction through the ADC module.
4) Continuously judging whether the current potential reaches the upper limit potential, and if the current potential does not reach the upper limit potential, continuously recording the current and potential information of the oxidation-reduction reaction; and if the current potential is detected to reach the upper limit potential, reversely scanning the oxidation-reduction reaction in the detection solution by a negative value of the scanning speed V (V/s), and recording the current and potential information of the oxidation-reduction reaction through the ADC module.
5) Continuously judging whether the current potential reaches a preset lower limit potential, and if the current potential does not reach the lower limit potential, continuously recording the current and potential information of the oxidation-reduction reaction; if the current potential is detected to reach the lower limit potential, the data is saved, and the detection is quit;
6) and displaying the detected CV curve in a PC upper computer, and storing the detected data.
The algorithm for testing the concentration and analyzing the components of the escherichia coli solution by using the chronoamperometry comprises the following steps:
1) selecting a timing current method on a PC upper computer interface, wherein the input of related parameters comprises the following steps: initial potential Init E, upper limit potential High E, lower limit potential Low E, step times Stepnum, Pulse Width Width, sampling Interval Sample Interval and the like;
2) constant potential, namely upper limit potential, is applied to the working electrode through the DAC module, so that oxidation-reduction reaction is caused to occur on the surface of the electrode, reactants are gradually consumed, the thickness of a diffusion layer near the electrode is increased, and a larger current density is generated;
3) in the oxidation-reduction process, sampling reaction current and potential for n times, wherein the sampling interval is constant, and recording the current and potential information of the oxidation-reduction reaction through an ADC (analog-to-digital converter) module;
4) after sampling in the first section of step pulse is finished, the step number is reduced by one, if the step number is not zero at the moment, the DAC module is used for applying a lower limit potential to the working electrode, the surface of the electrode can also generate an oxidation-reduction reaction and generate a larger current density, the reaction current and the reaction potential are sampled for n times, the sampling interval is unchanged, and the ADC module is used for recording the current and potential information of the oxidation-reduction reaction;
5) at the moment, continuously judging whether the step frequency is zero, if not, restarting to continuously detect from 2) until the step frequency is zero; if the step number is zero, storing the data and exiting the detection;
6) and displaying the measured CA curve in a PC upper computer, and analyzing the CA curve.
The method for performing concentration test and component analysis on the escherichia coli solution by using the differential pulse voltammetry comprises the following steps:
1) selecting a differential pulse voltammetry on a PC upper computer interface, and inputting relevant parameters, wherein the parameters comprise: selecting a timing current method on a PC upper computer interface, wherein the input of related parameters comprises the following steps: initial potential Init E, end potential Final E, potential increment Incr E, Amplitude, Pulse Width, sampling interval, sampling Period, Pulse Period and the like;
2) during the detection process, the ADC module measures the Faraday current generated by the oxidation-reduction reaction at the later stage of each pulse voltage. An initial potential is applied to the working electrode through the DAC module, the Faraday current and the potential are sampled after time delay (Pulse Period-Pulse Width-Sample Width) s of a timer, and the sampling time length is (Sample Width) s. After sampling is finished, the DAC module applies a step potential to the working electrode, and the step potential is formed by adding a preset amplitude potential on the basis of the initial potential;
3) after delaying by a timer with a step potential (Pulse Width-Sample Width) s, the faraday reaction current is measured and sampled at the later stage, and the sampling time length is (Sample Width) s. After sampling, the DAC module applies a new step potential to the working electrode, and the step potential is formed by adding a preset potential increment on the basis of the initial potential;
4) judging whether the new initial potential is equal to the termination potential or not, and restarting detection from the step 2) if the new initial potential is not equal to the termination potential; if the detected data are equal, the detection data are saved, and the detection is quitted.
Step three: after the original electrochemical detection signal data is obtained, due to the defects that the portable potentiostat system is not high enough in sensitivity and is easy to be interfered by the environment, the smooth denoising processing is carried out by utilizing the recursive average filtering algorithm. In the continuous domain, the expression of the recursive average filtering algorithm is:
wherein, TwFor the sliding window length, this parameter is an important parameter that affects the performance of the smoothing filter.
The transfer function of the recursive average filtering algorithm obtained from the above equation is:
the amplitude-frequency expression of the recursive average filter is obtained by equation (2):
and dividing the discrete electrochemical detection data into N small intervals by using a window, and carrying out average calculation in the intervals. A flowchart of a specific recursive average filtering algorithm is shown in fig. 3.
The smoothing filtering operation can improve the signal-to-noise ratio of the original data, and is convenient for subsequent feature extraction operation.
Step four: and (4) automatically extracting the characteristics of the electrochemical detection data obtained after the smoothing treatment, such as an electrochemical measurement curve characteristic parameter representation diagram shown in figure 4.
The selection of the characteristic parameters follows the following principles: (1) the information difference of the experimental result is reflected; (2) easy to calculate and analyze; (3) the parameters are independent of each other. According to the requirements, the characteristics of peak current, peak potential, peak area, baseline slope, initial response potential and the like in the cyclic voltammetry are extracted, initial steady-state current time and steady-state current in the chronoamperometry are extracted, and peak potential and peak current characteristics in the differential pulse voltammetry are extracted. Table 1 shows the extracted feature parameters and the detailed description.
TABLE 1 characteristic parameters extracted by three detection methods and detailed description
And constructing an automatic feature extraction algorithm according to the feature representation diagram, wherein the algorithm flow chart is shown in FIG. 5.
The automatic extraction process of the cyclic voltammetry curve features is developed into the following steps:
the first step is as follows: and (3) equally dividing the cyclic voltammetry detection data subjected to denoising and smoothing into an upper part and a lower part, wherein the upper part is an oxidation curve, and the lower part is a reduction curve.
The second step is that: separately determining a fitting function f for the oxidation curve and the reduction curveo(x) And fr(x) In that respect The maximum value and the minimum value in a certain range are respectively taken on an oxidation curve and a reduction curve, and the maximum value potential of the oxidation curve is the oxidation peak potential (I)op) The minimum potential of the reduction curve is the reduction peak potential (I)rp)。
The third step: respectively calculating the first derivative of the fitting function before the peak position, and taking the median point of the first derivative as the baseline slope and the oxidation curve baseline slope (K)o) And reduction curve baseline slope (K)r). At the same time, the original potential of the point is the initial response potential, and is respectively the initial oxidation potential (V)io) And initial reduction potential (V)ir)。
The fourth step: determining the oxidation curve baseline L in the smooth curve by taking the initial response potential as the tangent point and adding the baseline slope obtained in the third stepo(x) And reduction curve baseline Lr(x)。
The fifth step: the peak current was determined by taking the height from the peak point to the base line as the peak current (I)op) And reduction of peak current (I)rp)。
And a sixth step: the difference between the fitting function and the base line is integrated into peak areas from the initial response potential to the peak position, and the peak areas are oxidation peak areas (S)o) And oxidation peak area (S)r)。
The automatic extraction process of the characteristics of the timing current curve is developed into the following steps:
the first step is as follows: and solving a fitting function f (x) of the timing current detection data after the denoising and smoothing treatment.
The second step is that: and solving the first derivative of the fitting function to obtain a differential array.
The third step: the differential value of the next data point is used to subtract the previous one to form a differential difference array.
The fourth step: the first negative value in the differential difference array is obtained, which indicates that the detected data is increasing and decreasing, and also indicates that the current value begins to reach a steady state. The position of the data point is determined as the initial steady state current time.
The fifth step: and taking the median of the current values in the range from the initial steady-state current time to the response end as the steady-state current.
The automatic extraction process of the differential pulse voltammetry curve features is expanded into the following steps:
the first step is as follows: and (4) solving a fitting function f (x) of the denoised and smoothed differential pulse voltammetry detection data.
The second step is that: the corresponding position of the maximum value of the current within the range is taken as the peak potential.
The third step: and solving a first derivative of the fitting function before the peak potential, taking a median point of the first derivative as a slope of the baseline, and setting the original point as a tangent point between the baseline and the fitting function.
The fourth step: a baseline is determined within the smoothed curve, and the height from the peak-off location point to the baseline is the peak current.
According to the steps, all important characteristics of the cyclic voltammetry, the timing current and the differential pulse voltammetry curve can be automatically extracted, and a characteristic set is generated.
Step five: and D, performing dimensionality reduction operation on the characteristic parameter set generated in the step four by using a t-SNE algorithm, and performing visualization. The specific dimension reduction process can be expanded to the following steps:
the first step is as follows: and calculating Euclidean distances among the characteristics of different detection samples. Assuming that the feature set for all samples is a two-dimensional matrix of m × n, expressed as:
According to the formula | | xi-xj||2Calculating the Euclidean distance between every two groups of samples to obtain a new m multiplied by n two-dimensional matrix, wherein the expression is as follows:
wherein d isijRepresenting the euclidean distance of the feature row vector between the ith sample and the jth sample.
The second step is that: the euclidean distance is converted to a feature conditional distribution probability. The Euclidean distance between the feature vectors obtained in the first step is converted into a conditional probability p representing similarityi|j,pi|jThe calculation formula is as follows:
wherein λ isiIs given by xiA gaussian variance at the center.
The third step: for conditional probability pi|jSumming and normalizing to obtain symmetrical joint probability pijThe conversion formula is as follows:
the fourth step: calculating the conditional distribution probability q of the low-dimensional featureijRepresenting y for a sample with a t distribution in a low-dimensional subspaceiAnd yjThe similarity between them. q. q.sijThe calculation formula of (a) is as follows:
the fifth step: solving a matching cost function of the t-SNE by using a gradient descent method, wherein the expression is as follows:
C=∑iKL(Pi|Qi) (7)
finally obtaining the corresponding detection sample Y ═ Y (Y) of the low-dimensional subspace1,y2,…,yN)。
Step six: and (5) training a concentration determination model according to the data set subjected to the dimensionality reduction treatment in the step five, and accurately determining the concentration based on the trained model. And training a concentration determination model by adopting an XGboost algorithm, extracting a part of data set from the concentration determination model, testing the trained model, and verifying the performance of the model.
The target function of the XGboost algorithm is as follows:
Using a second order taylor expansion objective function to obtain:
the objective function that can be obtained is:
wherein,of the final objective functionThe value is the score of one CART tree, while the predicted result of the entire XGBoost is the sum of the scores of all CARTs.
Example (b):
a portable potentiostat system is used as a detection platform, and the on-site rapid determination of 51-60 mM micro-concentration gradient potassium ferricyanide/potassium ferrocyanide solution is realized on the basis of an electrochemical detection technology and a characteristic extraction parameter integrated model, wherein the process is as follows:
(1) firstly, repeatedly performing cross detection on 51-60 mM potassium ferricyanide/potassium ferrocyanide solution with 1mM equal concentration interval under a portable potentiostat system by using a cyclic voltammetry method, a chronoamperometry method and a differential pulse voltammetry method to obtain 1500 groups of experimental data. As shown in fig. 6, a CV curve of a group of micro-concentration gradient potassium ferricyanide/potassium ferrocyanide solutions is schematically shown, and an inset shows that a lower peak position is amplified ten times, and a detection curve has periodic signal interference, so that a detection curve aliasing phenomenon is caused under certain conditions.
(2) The obtained detection data is automatically transmitted to a smoothing module, the electrochemical detection data is smoothed by a recursive average filtering module, and by taking a cyclic voltammetry curve of a micro-concentration gradient potassium ferricyanide/potassium ferrocyanide solution as an example, as shown in fig. 7, the detection data after recursive average filtering is smoother, and the gradient is more obvious.
(3) And then, carrying out corresponding feature extraction on the detection signal after the smoothing treatment by adopting an automatic feature extraction module in the integrated model. The cyclic voltammetry curve automatically extracts features such as peak current, peak potential, peak area, baseline slope, initial response potential and the like, extracts initial steady-state current time and steady-state current in a chronoamperometry, extracts peak potential and peak current features in a differential pulse voltammetry, and arranges the peak potential and peak current features into a feature set.
(4) In order to improve the accuracy of concentration determination and reduce the redundancy of the feature set, the two-dimensional t-SNE dimension reduction is performed on the generated feature set, and fig. 8 is a visual result after the dimension reduction, so that it can be seen that detection data with the same concentration are gathered together, and a clear boundary is formed between the concentration and the concentration.
(5) And constructing a concentration prediction model by using an XGboost algorithm. 80% of the feature set after the t-SNE dimensionality reduction is a training set, and the rest of the data set is a testing set, wherein the data sets with the concentrations of 53mmol/L, 56mmol/L and 59mmol/L are set as an untrained blind concentration prediction set. The prediction results of the test set are shown in fig. 9, the predicted values and the actual values of the samples at each concentration are basically consistent, wherein 3 groups of untrained blind concentration sets also have a good prediction effect (as indicated by arrows), and the degree of fitting (R-Squared) between the actual concentration and the predicted concentration is 0.957.
In conclusion, the embodiment provides a method for rapidly determining a micro-concentration gradient solution based on an electrochemical detection technology and a feature extraction parameter integrated model. The integrated model operation steps are as follows: firstly, repeatedly performing cross detection on a micro-concentration gradient potassium ferricyanide/potassium ferrocyanide solution by using a cyclic voltammetry method, a chronoamperometry method and a differential pulse voltammetry method to obtain 1500 groups of data in total; secondly, smoothing the detection data by using a recursion average filtering module; then, automatic feature extraction operation is carried out on detection data of the three electrochemical detection methods, wherein the cyclic voltammetry curve automatically extracts features such as peak current, peak potential, peak area, baseline slope, initial response potential and the like, initial steady-state current time and steady-state current in a chronoamperometry are extracted, peak potential and peak current features in a differential pulse voltammetry are extracted, and the features are arranged into a feature set; then, reducing the feature set to two dimensions by using t-SNE dimension reduction, performing visual analysis, and reducing the complexity of a prediction model; and finally, inputting the processed feature set into a prediction model, so that a better concentration prediction effect is realized, and the fitting degree (R-Squared) between the actual concentration and the predicted concentration is 0.957. Therefore, the method realizes the rapid characteristic analysis of electrochemical detection data and the on-site rapid determination of the micro-concentration gradient solution.
Claims (10)
1. A micro-concentration gradient solution electrochemical determination method based on characteristic parameter extraction is characterized in that: performing cyclic voltammetry detection, chronoamperometry detection and differential pulse voltammetry detection on a plurality of detected solutions with micro concentration gradients to obtain cyclic voltammetry detection data, chronoamperometry detection data and differential pulse voltammetry detection data of different plurality of detected solutions;
secondly, extracting the graph characteristics of the cyclic voltammetry detection data, the timing current detection data and the differential pulse voltammetry detection data obtained in the first step; wherein the feature extracted from the cyclic voltammogram comprises the oxidation peak current IopReduction peak current IrpOxidation peak potential EopReduction peak potential ErpBaseline slope K of the oxidation curveoReduction curve baseline slope KrArea of oxidation peak SoReduction peak area SrInitial reduction potential VirAnd initial oxidation potential Vuo(ii) a The extracted features in the chronoamperometric curve include an initial steady-state current time tmAnd steady state current Is(ii) a The features extracted from the differential pulse voltammogram include the peak potential EpSum peak current Ip;
Thirdly, sorting the characteristic values extracted in the second step and the corresponding concentration labels into an initial data set; performing dimensionality reduction on the obtained initial data set to obtain an input data set;
step four, constructing a concentration determination model, and training the concentration determination model by using the input data set obtained in the step three; the trained concentration determination model is used for detecting the concentration of the solution to be detected with unknown concentration.
2. The method for electrochemical determination of micro concentration gradient solution based on characteristic parameter extraction as claimed in claim 1, wherein: among the features extracted from the cyclic voltammetry curve, the oxidation peak current IopThe height from the anode current peak value on the cyclic voltammetry I-V curve to the oxidation curve baseline; reduction peak current IrpThe height from the cathode current peak value to the reduction curve base line; oxidation peak potential EopIs the anode current peak position; reduction peak potential ErpIs the cathode current peak position; baseline slope K of oxidation curveoThe slope of the baseline for oxidation; reduction curve baseline slope KrThe slope of the baseline of the reduction curve; area of oxidation peak SoIs the oxidation curve, base line and distance from oxidation peak to base lineThe area enclosed by the three parts; reduction peak area SrThe area is the area enclosed by the reduction curve, the base line and the distance from the reduction peak to the base line; initial reduction potential VirIs the initial potential at which the reactants undergo a reduction reaction; initial oxidation potential VioIs the initial potential at which the reactants undergo oxidation reactions;
initial steady state current time t in features extracted from chronoamperometric curvesmThe initial time for reaching the steady-state current on the I-t curve of the timing current method; steady state current IsIs the median after the initial steady state current on the I-t curve;
among the features extracted from the differential pulse voltammogram, the peak potential EpThe anode current peak position on the differential pulse voltammetry I-V curve is shown; peak current IpThe height from the anode current peak to the baseline of the curve.
3. The method for electrochemical determination of micro concentration gradient solution based on characteristic parameter extraction as claimed in claim 1, wherein: in the second step, the process of automatically extracting the features of the cyclic voltammetry data is as follows:
the first step is as follows: equally dividing cyclic voltammetry detection data into an upper part and a lower part, wherein the upper part is oxidation process data, and the lower part is reduction process data;
the second step is that: separately determining a fitting function f for the oxidation process data and the reduction process datao(x) And fr(x) Obtaining an oxidation curve and a reduction curve; respectively taking the maximum value and the minimum value in a preset range on an oxidation curve and a reduction curve, wherein the maximum value potential of the oxidation curve is the oxidation peak potential IopThe minimum potential of the reduction curve is the reduction peak potential Irp;
The third step: respectively solving first derivative of fitting function before peak potential on oxidation curve and reduction curve, taking median point of two first derivatives as baseline slope, and respectively taking baseline slope K of oxidation curveoAnd reduction curve baseline slope Kr(ii) a Meanwhile, the original potentials of the two median points are initial response potentials which are respectively initial oxidation potentials VioAnd initial reduction potentialVir;
The fourth step: determining the oxidation curve baseline L in the smooth curve by taking the initial response potential as the tangent point and adding the baseline slope obtained in the third stepo(x) And reduction curve baseline Lr(x);
The fifth step: the height from the peak position to the base line is taken as the peak current and is taken as the oxidation peak current IopAnd reduction of peak current Irp;
And a sixth step: the difference between the fitting function and the base line is integrated into peak areas from the initial response potential to the peak position, and the peak areas are respectively the oxidation peak area SoAnd reduction peak area Sr。
4. The method for electrochemical determination of micro concentration gradient solution based on characteristic parameter extraction as claimed in claim 1, wherein: in the second step, the process of automatically extracting the characteristics of the timing current data is as follows:
the first step is as follows: solving a fitting function f (x) of the timing current detection data;
the second step is that: solving a first derivative of the fitting function to obtain a differential array;
the third step: subtracting the previous data point from the differential value of the next data point to form a differential difference value array;
the fourth step: acquiring a first negative value in the differential difference value array, wherein the point indicates that the detected data is accelerated and decelerated, and meanwhile, the current value starts to reach a steady state, and the position of the data point is determined as initial steady-state current time;
the fifth step: and taking the median of the current values in the range from the initial steady-state current time to the response end as the steady-state current.
5. The method for electrochemical determination of micro concentration gradient solution based on characteristic parameter extraction as claimed in claim 1, wherein: in the second step, the process of automatically extracting the characteristics of the differential pulse voltammetry data is as follows:
the first step is as follows: solving a fitting function f (x) of the differential pulse voltammetry detection data;
the second step is that: taking the corresponding position of the maximum value of the current in a preset range as a peak potential;
the third step: solving a first derivative of a fitting function before peak potential, taking a median point of the first derivative as a baseline slope, and setting an original point as a tangent point between the baseline and the fitting function;
the fourth step: and determining a base line in the smooth curve, and taking the height from the peak position point to the base line as the peak current.
6. The method for electrochemical determination of micro concentration gradient solution based on characteristic parameter extraction as claimed in claim 1, wherein: the solution to be detected is potassium ferricyanide/potassium ferrocyanide solution.
7. The method for electrochemical determination of micro concentration gradient solution based on characteristic parameter extraction as claimed in claim 1, wherein: in the first step, a three-electrode system is constructed in a solution to be detected; the three-electrode system uses a nano-gold modified electrode as a working electrode, an Ag/AgCl electrode as a reference electrode, a platinum wire electrode as an auxiliary electrode, and a portable potentiostat to form a field rapid electrochemical detection platform, so that cyclic voltammetry detection, chronoamperometry detection and differential pulse voltammetry detection are realized.
8. The method for electrochemical determination of micro concentration gradient solution based on characteristic parameter extraction as claimed in claim 1, wherein: and (4) aiming at the cyclic voltammetry I-V curve, the chronoamperometry I-t curve and the differential pulse voltammetry I-V curve obtained in the step one, a recursive average filtering method is utilized to carry out smooth denoising treatment, so that the signal-to-noise ratio of original data is improved, and the subsequent feature extraction operation is facilitated.
9. The method for electrochemical determination of micro concentration gradient solution based on characteristic parameter extraction as claimed in claim 1, wherein: the dimension reduction method in the third step specifically adopts a nonlinear dimension reduction algorithm t-SNE.
10. The method for electrochemical determination of micro concentration gradient solution based on characteristic parameter extraction as claimed in claim 1, wherein: and the concentration determination model in the fourth step is constructed by an XGboost algorithm.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202111671385.0A CN114295704B (en) | 2021-12-31 | 2021-12-31 | Micro-concentration gradient solution electrochemical determination method based on characteristic parameter extraction |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202111671385.0A CN114295704B (en) | 2021-12-31 | 2021-12-31 | Micro-concentration gradient solution electrochemical determination method based on characteristic parameter extraction |
Publications (2)
Publication Number | Publication Date |
---|---|
CN114295704A true CN114295704A (en) | 2022-04-08 |
CN114295704B CN114295704B (en) | 2023-11-14 |
Family
ID=80974053
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202111671385.0A Active CN114295704B (en) | 2021-12-31 | 2021-12-31 | Micro-concentration gradient solution electrochemical determination method based on characteristic parameter extraction |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN114295704B (en) |
Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
WO2004053476A1 (en) * | 2002-12-09 | 2004-06-24 | Otre Ab | Simplified signal processing method for voltammetry |
CN101614690A (en) * | 2009-07-27 | 2009-12-30 | 中国科学院长春应用化学研究所 | A kind of multi-channel electrochemical analyzer |
CN111537576A (en) * | 2020-04-16 | 2020-08-14 | 中国科学院化学研究所 | Method for detecting neurochemical substance in brain |
CN111563558A (en) * | 2020-05-13 | 2020-08-21 | 宿州学院 | Rapid identification method for producing area and brand of wine |
WO2021237974A1 (en) * | 2020-05-26 | 2021-12-02 | 江苏大学 | Volt-ampere electronic tongue-based device and method for rapid testing of physical and chemical indicators of soybean sauce |
-
2021
- 2021-12-31 CN CN202111671385.0A patent/CN114295704B/en active Active
Patent Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
WO2004053476A1 (en) * | 2002-12-09 | 2004-06-24 | Otre Ab | Simplified signal processing method for voltammetry |
CN101614690A (en) * | 2009-07-27 | 2009-12-30 | 中国科学院长春应用化学研究所 | A kind of multi-channel electrochemical analyzer |
CN111537576A (en) * | 2020-04-16 | 2020-08-14 | 中国科学院化学研究所 | Method for detecting neurochemical substance in brain |
CN111563558A (en) * | 2020-05-13 | 2020-08-21 | 宿州学院 | Rapid identification method for producing area and brand of wine |
WO2021237974A1 (en) * | 2020-05-26 | 2021-12-02 | 江苏大学 | Volt-ampere electronic tongue-based device and method for rapid testing of physical and chemical indicators of soybean sauce |
Also Published As
Publication number | Publication date |
---|---|
CN114295704B (en) | 2023-11-14 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
JP7090184B2 (en) | Biometric concentration measurement method using artificial intelligence deep learning learning | |
US8303787B2 (en) | Method and apparatus for processing electrochemical signals | |
KR102142647B1 (en) | Artificial Neural Network Model-Based Methods, Apparatus, Learning starategy and Systems for Analyte Analysis | |
US6054038A (en) | Portable, hand-held, in-situ electrochemical sensor for evaluating corrosion and adhesion on coated or uncoated metal structures | |
Ye et al. | Analyzing the anodic stripping square wave voltammetry of heavy metal ions via machine learning: Information beyond a single voltammetric peak | |
CN111579637B (en) | Nondestructive testing method and device for detecting and distinguishing internal and external defects of steel wire rope | |
CN114295704B (en) | Micro-concentration gradient solution electrochemical determination method based on characteristic parameter extraction | |
CN111812158A (en) | Frequency conversion detection method for urine conductivity and osmotic pressure | |
CN118038981A (en) | Method and measuring instrument for extracting Cq value based on curvature change of qPCR amplification curve | |
CN110084420B (en) | Method for detecting total sugar, total acid and alcoholic strength of yellow water in strong aromatic Chinese spirit fermentation | |
CN116929438A (en) | Performance test method and device for sensor | |
CN105606670A (en) | Electrochemical method for detecting polycyclic aromatic hydrocarbons in high-sensitivity manner | |
Asir et al. | Electrochemical determination of potassium ferricyanide using artificial intelligence | |
CN111537576A (en) | Method for detecting neurochemical substance in brain | |
CN114609028A (en) | Portable device and method for in-situ detection of corrosion resistance of organic coating | |
Luo et al. | A portable impedance-based electrochemical measurement device | |
Economou et al. | Wiener filtering of electroanalytical data by means of fast Fourier transform | |
US20230011245A1 (en) | Method and apparatus for detecting corrosion | |
Kumar et al. | A New Approach of Modeling the Electronic Tongue Sensors for Classification | |
Chady et al. | Data fusion of eddy current NDT signals | |
Pandelova et al. | Determination of Limit of Detection in Electrochemical Biosensors | |
CN114004304A (en) | Multi-feature target identification method | |
CN116858920A (en) | Water quality anomaly detection method based on dynamic window correlation detection and feature fusion | |
Ramos et al. | Research Article LabVIEW 2010 Computer Vision Platform Based Virtual Instrument and Its Application for Pitting Corrosion Study | |
Sokolkov | Electrochemical Sensors for Digital Medicine and Biology Electrochemical Digitization of Analytical Information Biological Fluids Samples |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
GR01 | Patent grant | ||
GR01 | Patent grant |