CN-116067953-B - Method for manufacturing colorimetric sensing unit and array by using multi-target Bayesian optimization algorithm

CN116067953BCN 116067953 BCN116067953 BCN 116067953BCN-116067953-B

Abstract

The invention discloses a method for manufacturing a colorimetric sensing unit and an array by using a multi-target Bayesian optimization algorithm. The method comprises the steps of firstly preparing colorimetric sensing units of different formulas on a pore plate, drying the pore plate, introducing gas into the dried pore plate for testing to obtain values of response time, reversibility, responsivity and sensitivity of each formula, processing the values of the response time, reversibility, responsivity and sensitivity of each formula and the values of the response time, reversibility, responsivity and sensitivity through a Sigmoid function to obtain scoring values, substituting the scoring values into a multi-objective Bayesian optimization algorithm to generate a next formula, preparing the colorimetric sensing units on the pore plate according to the formulas, and repeating the steps continuously until the circulation number reaches a stop condition to obtain a global optimal formula. The method solves the problem that multiple indexes of the sensor are easy to consider when simultaneously optimizing, and can quickly optimize the sensor until the sensor array simultaneously has wide measuring range, high sensitivity, quick response and high reversibility.

Inventors

JIANG JING
ZHANG LONGHAN
CHEN YANGGUAN
LONG YIFAN
Ai Zhehong

Assignees

之江实验室

Dates

Publication Date: 20260508
Application Date: 20230106

Claims (9)

1. A method for rapidly manufacturing a colorimetric sensing unit by utilizing a multi-target Bayesian optimization algorithm is characterized in that: 1) Initially preparing N colorimetric sensing units with different formulas on an N-hole plate; 2) Drying the N pore plate, and introducing the dried N pore plate into a gas test to obtain values of response time, reversibility, responsivity and sensitivity of each colorimetric sensing unit corresponding to each formula; 3) The response time, reversibility, responsivity and sensitivity values of each formula and the corresponding colorimetric sensing unit are processed by a Sigmoid function to obtain scoring values, and the scoring values are substituted into a multi-target Bayesian optimization algorithm to generate N next-round formulas; In the step 3), the response time, reversibility, responsivity and sensitivity values of the colorimetric sensing unit are processed by a Sigmoid function to obtain a grading value, specifically, the response time, reversibility and responsivity values are respectively processed by adopting a Sigmoid function conversion calculation, and then are multiplied by the sensitivity value weighting to obtain the grading value, which is specifically expressed as: Wherein Obj is a grading value, obj ΔE 、Obj Δt 、Obj reversibility is a conversion target value of the responsivity, response time and reversibility of the colorimetric sensing unit, obj sensitivity is the sensitivity of the colorimetric sensing unit, b 2 is a first scale factor and a second scale factor, Δt represents the time that the colorimetric sensing unit passes from exposure to the gas of which the concentration is to be measured until the response value reaches 95% of a final stable value, represents the response time of the sensing unit, ΔE is the response value exposed to the CO 2 atmosphere, and ΔE 0 is the response value after stopping exposure to the CO 2 atmosphere for 240 seconds; th ΔE 、th Δt 、th reversibility represents the threshold value of the sigmoid function acting on the three targets of the response value, the response time and the reversibility, s ΔE 、s Δt 、s reversibility represents the adjustment parameter of the slope around the control threshold value of the sigmoid function acting on the three targets of the response value, the response time and the reversibility, slope (CO 2 , delta E) represents the slope obtained by linear regression of the response data collected by taking the carbon dioxide concentration C CO2 as an independent variable and the response value delta E as a dependent variable, and C CO2 represents the carbon dioxide concentration; 4) Preparing N colorimetric sensing units on an N-well plate according to N formulas; 5) Continuously repeating the steps 2) to 4) for circulation, and obtaining a global optimal formula after a plurality of circulation rounds reach a stop condition, and preparing a colorimetric sensing unit according to the global optimal formula; The iteration is combined with the search space, the search round and the number of data points contained in the round at early stage or later stage to comprehensively determine, and the processing is set according to the following formula: Wherein S stage is an index for evaluating early or later optimization rounds, N batch is the number of experimental data points of each round, k is the number of rounds completed, m is the dimension of the search space, and d i is the width of the ith dimension variable; Parameters for controlling bias exploration and bias mining in the acquisition function are updated in each round of search by combining the index S stage : Wherein X, Y is input and output variables of the function to be optimized, ζ and λ are super parameters for obtaining the function to regulate and control exploration and mining balance, μ (X) and σ (X) are mean and variance of the agent model estimation, Φ (X) and Φ (X) are cumulative distribution function and probability density function of the variables respectively; the formula for updating xi and lambda with the iteration round number is as follows: 。
2. A method of manufacturing a colorimetric sensing unit using a multi-objective bayesian optimization algorithm according to claim 1, wherein: Each colorimetric sensing unit is mainly prepared by adding Polytetrafluoroethylene (PTFE) film into the hole site bottom of a hole site of a hole plate, adding source solution into the hole site of the hole plate, and uniformly mixing, wherein the source solution comprises tetrabutylammonium hydroxide, ethylcellulose, polyethylene glycol, a pH indicator and a solvent, the solvent comprises Isopropanol (IPA) and deionized water, and the pH indicator is one of thymol blue, m-cresol purple, phenol red and cresol red.
3. A method of manufacturing a colorimetric sensing unit using a multi-objective bayesian optimization algorithm according to claim 1, wherein: The formula comprises tetrabutyl ammonium hydroxide, ethylcellulose, polyethylene glycol, a pH indicator and the concentration of a solvent, wherein the concentration of the solvent comprises the concentration of isopropanol and deionized water.
4. A method of manufacturing a colorimetric sensing unit using a multi-objective bayesian optimization algorithm according to claim 1, wherein: In the colorimetric sensing unit, a porous polytetrafluoroethylene film is used as a matrix, and tetrabutylammonium hydroxide, cresol red, polyethylene glycol and ethylcellulose are used for constructing the colorimetric sensing unit.
5. A method of manufacturing a colorimetric sensing unit using a multi-objective bayesian optimization algorithm according to claim 1, wherein: in the step 2), the N pore plate is dried for 1 hour by using an oven at 50 ℃, then the dried N pore plate is fixed in an air chamber, gas to be detected is introduced for a period of time and then is measured, an image of the N pore plate is shot by a camera to obtain RGB values of all holes, and further an aeration curve is obtained according to the analysis and the treatment of the RGB values, and values of response time, reversibility, responsiveness and sensitivity are extracted from the aeration curve.
6. The method of claim 1, wherein in the step 3), six parameters and score values of the N formulations are input into the multi-objective Bayesian optimization algorithm to construct a seven-dimensional space, and the multi-objective Bayesian optimization algorithm is utilized to optimize iteration to generate new six parameters of each of the N formulations with larger score values as targets.
7. The method of claim 1, wherein in the step 5), parameters of the obtaining function in the multi-objective Bayesian optimization algorithm are adjusted during loop iteration, so that the multi-objective Bayesian optimization algorithm tends to explore when generating the next round of formulation in loop iteration 1-N, and tends to explore when generating the next round of formulation in later loop iteration n+1.
8. The method of claim 1, wherein in the step 5), the stopping condition means that the variance of the scoring values of the N formulas is smaller than a preset variance threshold and the average value of the scoring values of the N formulas is larger than a preset average threshold, and finally the formula with the highest scoring value is taken as the globally optimal formula.
9. A method of manufacturing a colorimetric sensor array using a multi-objective bayesian optimization algorithm, comprising: The colorimetric sensor array is divided into colorimetric sensor units with different detection ranges, and is manufactured optimally according to the method of claim 1 for each colorimetric sensor unit, and the colorimetric sensor array is formed by all colorimetric sensor unit arrays.

Description

Method for manufacturing colorimetric sensing unit and array by using multi-target Bayesian optimization algorithm Technical Field The invention belongs to an optimized manufacturing method of colorimetric sensing in the field of biochemical synthesis, and particularly relates to a method for manufacturing colorimetric sensing units and arrays by using a multi-objective Bayesian optimization algorithm in a sensor performance optimization process. Background Carbon dioxide sensing is required in many fields including building ventilation, exhaust gas treatment, carbon capture utilization and storage, deep diving and aerospace. In practical applications, the concentration of CO 2 can range from 400ppm to greater than 30% of the air concentration. Various efforts have been made to construct CO 2 sensors based on different mechanisms, such as gas chromatography, electrode, electrochemical, chemi-resistive, optical and acoustic based mechanisms. Among them, colorimetry is receiving more attention because of its low cost and ease of use, and is an ideal choice for preparing portable and disposable CO 2 sensors. Current research on colorimetric CO 2 sensors has focused on lower detection limits, liquid phase detection, or high selectivity detection. Few colorimetric sensors developed in a systematic way are capable of achieving both high sensitivity, wide range, short response time and good reversibility of detection. Theoretically, a CO 2 colorimetric sensor array with all the advantages mentioned above can be prepared by adjusting the types and proportions of the raw materials. However, the adjustment of the types and proportions of raw materials can involve approximately 10 variables, and the manual handling of such high-dimensional variables is not only cumbersome but also prone to localized optimization. To solve this problem, the raw material types and proportions of the colorimetric sensor array can be optimized by some optimization algorithms, but most of the optimization algorithms for biochemical synthesis and material preparation reported so far have only one optimization target, such as catalytic efficiency, yield, quantum efficiency and the like. In the development of the CO 2 colorimetric sensor mentioned above, various targets such as sensitivity, responsiveness, response time, and reversibility have a great influence on the usability thereof. The degree to which different objectives need to be optimized varies. The prior art lacks an implementation method capable of achieving the multi-optimization target result of the colorimetric sensing array at the same time. Disclosure of Invention In order to solve the problems in the background art, the invention provides a method for optimizing the performance of a sensor in a machine learning optimized colorimetric sensor, solves the problem that multiple indexes of the sensor are easy to lose when simultaneously optimizing the indexes, and accurately and quickly obtains the CO 2 colorimetric sensor array with wide range, high sensitivity, quick response and high reversibility. The technical scheme adopted by the invention is as follows: For colorimetric sensing units within a fixed detection range, the method is processed as follows: 1) N colorimetric sensing units with different formulas are initially prepared on an N pore plate, one colorimetric sensing unit is prepared on each pore site, and the colorimetric sensing units with the N different formulas can be randomly generated by the initial preparation. 2) Drying the N pore plate, and introducing the dried N pore plate into gas to be tested with different concentrations in a fixed detection range to obtain values of response time, reversibility, responsivity and sensitivity of each formula corresponding to the colorimetric sensing unit; 3) The response time, reversibility, responsivity and sensitivity values of each formula and the corresponding colorimetric sensing unit are processed by a Sigmoid function to obtain scoring values, and the scoring values are substituted into a multi-target Bayesian optimization algorithm to generate N next-round formulas; 4) Preparing N colorimetric sensing units on an N-well plate according to N formulas in the same manner as in the step 1); 5) And (3) continuously repeating the steps 2) to 4) to form a cycle, and obtaining a global optimal formula after a plurality of cycles reach a stop condition, and preparing the colorimetric sensing unit according to the global optimal formula. Each colorimetric sensing unit is mainly obtained by uniformly mixing Polytetrafluoroethylene (PTFE) film serving as a sensing matrix which is firstly added into the bottom of a hole site of a hole plate, a source solution which comprises tetrabutylammonium hydroxide (TBAH), ethylcellulose (EC), polyethylene glycol (PEG, mw=200), a pH indicator and a solvent, wherein the solvent comprises Isopropanol (IPA) and deionized water, and the pH indicator is selected from one of Thymol Blue (TB),