Rapid Spectroscopic Detection Method for Purity and Impurities in Industrial Hydrochloric Acid

Abstract

This invention discloses a rapid spectroscopic detection method for the purity and impurities of industrial hydrochloric acid, belonging to the field of chemical analysis and detection technology. The method uses a deuterium-tungsten composite light source and a miniature fiber optic spectrometer to collect ultraviolet absorption band spectral data and visible light absorption band spectral data. After preprocessing and characteristic wavelength screening, a dual-objective partial least squares correction model is established to achieve simultaneous prediction of purity and impurities. Mutual interference is eliminated through impurity interference compensation and purity feedback correction mechanisms. Combined with spectral quality evaluation, the reliability of the results is ensured. This invention achieves a single detection time of less than 1 minute, a purity accuracy of ±0.5%, and an iron impurity detection limit of 5 ppm.

Description

Technical Field

[0001] This invention belongs to the field of chemical analysis and detection technology, specifically a rapid spectroscopic detection method for the purity and impurities of industrial hydrochloric acid. Background Technology

[0002] As an important basic chemical raw material and a precursor chemical for toxic substances, industrial hydrochloric acid is subject to strict administrative supervision in its production, sale, and use. Hydrochloric acid production and sales units are required to establish comprehensive record-keeping systems, while public security organs are required to conduct regular or irregular inspections and verifications of the flow of hydrochloric acid. In actual regulatory work, quickly and accurately determining the purity level and impurity content of hydrochloric acid samples is a key technical means to determine the legality of its source and the compliance of its use.

[0003] Traditional methods for determining the purity of hydrochloric acid primarily employ acid-base titration. While this method offers high accuracy, it suffers from several significant drawbacks: First, titration requires precise sample weighing, preparation of standard solutions, dropwise addition of indicators, and observation of the endpoint color change. The entire process typically takes over 30 minutes, failing to meet the timeliness requirements for rapid screening of large batches of samples. Second, titration is highly dependent on the professional skills of the operators, necessitating systematically trained chemical analysts for correct operation, a skill generally lacking in the expertise of grassroots public security investigators. Third, titration consumes large quantities of chemical reagents, generating acidic waste that requires proper disposal, which is detrimental to on-site testing and environmental protection.

[0004] For the detection of heavy metal impurities such as iron and arsenic in hydrochloric acid, the current national standard methods mainly include atomic absorption spectrometry and colorimetry. While atomic absorption spectrometry offers high sensitivity and selectivity, the equipment is expensive, bulky, and requires a dedicated gas source, making it unsuitable for rapid on-site detection. Colorimetry, although requiring relatively simple equipment, involves cumbersome pretreatment steps, long color development times, and different impurities require different color development systems, making simultaneous rapid detection of multiple impurities difficult. This situation leads to the need in actual regulatory work for purity testing and impurity testing to be performed separately in different laboratories using different methods, resulting in testing cycles lasting several days and severely hindering regulatory efficiency.

[0005] In recent years, ultraviolet-visible absorption spectroscopy combined with chemometrics has made significant progress in the field of rapid quantitative analysis. Chinese patent CN101893576A discloses a heavy metal test strip, its preparation method, and its applications. This method utilizes the displacement reaction between chloride and a heavy metal chelating agent to achieve qualitative or semi-quantitative detection. However, this method can only detect the total amount of heavy metals and cannot distinguish specific impurity types, nor can it simultaneously obtain sample purity information. Partial least squares regression has been successfully applied in the determination of effective components in traditional Chinese medicine and rapid evaluation of food quality, but systematic research on its application to the simultaneous detection of purity and impurities in industrial hydrochloric acid remains lacking.

[0006] In summary, there is an urgent need to develop a new method that can simultaneously and rapidly detect the purity and heavy metal impurity content of industrial hydrochloric acid, in order to meet the needs of efficient, accurate, and portable detection in the control of the flow of industrial hydrochloric acid, and to provide effective technical support for the verification of ledgers of hydrochloric acid production and operation units and public security inspections. Summary of the Invention

[0007] To address the technical problems in existing technologies, such as long detection cycles for industrial hydrochloric acid purity, low efficiency due to the need for separate methods for purity and impurities, and the difficulty in achieving laboratory-level quantitative accuracy with portable equipment, this invention provides a rapid spectroscopic detection method for industrial hydrochloric acid purity and impurities.

[0008] The present invention adopts the following technical solution:

[0009] A rapid spectroscopic method for detecting the purity and impurities of industrial hydrochloric acid includes the following steps:

[0010] A diluted industrial hydrochloric acid sample solution was irradiated using a deuterium-tungsten composite light source. Ultraviolet (UV) and visible light absorption (VLS) spectral data were simultaneously acquired using a miniature fiber optic spectrometer. The acquired UV and VLS spectral data underwent baseline correction, scattering correction, and smoothing / denoising preprocessing to obtain standardized spectral data. Based on correlation analysis between UV spectral data and hydrochloric acid concentration, and between VLS spectral data and impurity content, the variable importance projection method was used to screen characteristic wavelength ranges for purity and impurities from the standardized spectral data. The selected characteristic wavelength range spectral data were then used as the input... The system takes hydrochloric acid purity and impurity content as inputs and outputs as outputs. It employs partial least squares regression to establish purity correction and impurity correction models, respectively, and jointly optimizes the model parameters within a unified multi-task learning framework. The system calculates the interference contribution of impurity absorption to the purity detection band to obtain an interference compensation factor. The prediction results of the impurity correction model are used to correct the prediction results of the purity correction model, resulting in a compensated hydrochloric acid purity value. The system also performs a joint evaluation of the spectral data of the sample under test using residual statistics and leverage values ​​to identify abnormal spectra. For the spectral data that passes the quality evaluation, the purity correction and impurity correction models are applied for prediction, outputting the purity value and impurity content of the industrial hydrochloric acid sample.

[0011] The beneficial effects of this invention are as follows:

[0012] This invention combines ultraviolet-visible absorption spectroscopy with partial least squares chemometrics to achieve simultaneous and rapid detection of the purity and heavy metal impurity content of industrial hydrochloric acid. Compared with traditional titration and atomic absorption spectrometry, the detection time of this invention is reduced from several hours to less than 1 minute, significantly improving detection efficiency. By innovatively introducing impurity interference compensation and purity feedback correction mechanisms, the interference of impurity absorption on purity detection is effectively eliminated, achieving near-laboratory-level detection accuracy while maintaining portability. This method requires no complex sample pretreatment or chemical reagent consumption, is simple to operate, and is particularly suitable for applications requiring rapid on-site screening for industrial hydrochloric acid flow control and for police investigation and evidence collection. Attached Figure Description

[0013] Figure 1 This is an overall flowchart of the rapid spectroscopic detection method for the purity and impurities of industrial hydrochloric acid according to the present invention.

[0014] Figure 2 This is a detailed flowchart of the spectral data acquisition and preprocessing steps of the present invention.

[0015] Figure 3 This is a detailed flowchart of the characteristic wavelength range screening steps of the present invention.

[0016] Figure 4 This is a detailed flowchart of the steps for establishing the dual-objective partial least squares correction model of the present invention.

[0017] Figure 5 This is a detailed flowchart of the impurity interference compensation and purity feedback correction steps of the present invention. Detailed Implementation

[0018] The technical solution of the present invention will be clearly and completely described below with reference to the accompanying drawings and embodiments. It should be noted that the following embodiments are only used to illustrate the technical solution of the present invention, and are not intended to limit the scope of protection of the present invention.

[0019] like Figure 1As shown, the overall process of the rapid spectroscopic detection method for the purity and impurities of industrial hydrochloric acid of this invention includes six main steps: Step S1: Spectral data acquisition and preprocessing. Ultraviolet-visible full-band spectra are acquired from diluted industrial hydrochloric acid samples, followed by baseline correction, scattering correction, and smoothing / denoising to obtain standardized spectral data. Step S2: Characteristic wavelength range screening. Based on spectral-concentration correlation analysis and variable importance projection method, characteristic wavelength ranges for purity and impurities are screened respectively. Step S3: Establishment of a dual-objective partial least squares correction model. A multi-task learning framework is used to jointly optimize the purity correction model and the impurity correction model. Step S4: Impurity interference compensation and purity feedback correction. An interference compensation factor is calculated to eliminate the interference of impurities on purity detection, and a concentration correction factor is introduced to provide feedback correction for impurity prediction. Step S5: Spectral quality evaluation and anomaly identification. Residual statistics and leverage values ​​are used to jointly evaluate and identify abnormal spectra. Step S6: Comprehensive prediction and result output. The correction model is applied to the spectra that pass the quality evaluation to predict the purity value and impurity content. These six steps constitute a complete detection process, enabling simultaneous and rapid detection of the purity and impurities of industrial hydrochloric acid.

[0020] The core technical idea of ​​the rapid spectroscopic detection method for purity and impurities of industrial hydrochloric acid provided by this invention is: to achieve quantitative purity by utilizing the good linear relationship between the characteristic absorption of hydrochloric acid molecules in the ultraviolet region and concentration, and to simultaneously detect impurities by utilizing the characteristic absorption peaks of heavy metal impurities such as iron and arsenic in the visible light region. Furthermore, through an innovative dual-objective collaborative modeling strategy and an impurity interference compensation mechanism, high-precision synchronous determination of purity and impurity content is achieved in a unified detection system.

[0021] Step S1: Spectral data acquisition and preprocessing.

[0022] The first core step of this invention is the acquisition and preprocessing of spectral data. For example... Figure 2 As shown, the purpose of this step is to obtain high-quality standardized spectral data to lay the foundation for subsequent modeling and analysis.

[0023] Regarding the configuration of the spectral acquisition system, this invention preferably uses a deuterium-tungsten composite light source as the illumination source. The deuterium lamp portion covers the ultraviolet band from 190nm to 380nm, while the tungsten lamp portion covers the visible light band from 380nm to 800nm. The combination of these two components achieves continuous spectral output across the entire wavelength range of 190nm to 800nm. The light source output power is maintained within the range of 15W to 25W, preferably 20W, to ensure a sufficient signal-to-noise ratio. In one embodiment of this invention, a DH-2000 type deuterium-tungsten composite light source manufactured by a company in Shenzhen is used, with a spectral range of 190nm to 2500nm, which fully meets the detection requirements.

[0024] Regarding the spectral detector, this invention employs a miniature fiber optic spectrometer based on a linear CCD or CMOS image sensor, with a preferred spectral resolution of 0.5 nm to 2.0 nm and an adjustable integration time ranging from 1 ms to 10 s. In one embodiment, the invention uses a Marine Optics USB4000 miniature fiber optic spectrometer, with a spectral range of 200 nm to 850 nm, a spectral resolution of approximately 1.5 nm, and a single full-spectrum acquisition time as low as 3 ms, fully meeting the requirements for millisecond-level rapid acquisition. The fiber optic transmission system uses quartz fiber with a core diameter of 400 μm to 600 μm and a numerical aperture of 0.22 to 0.25 to ensure effective transmission of ultraviolet light energy.

[0025] Regarding sample preparation, since the concentration of industrial hydrochloric acid is typically between 30% and 37%, direct measurement will lead to absorbance saturation in the ultraviolet region; therefore, appropriate dilution is necessary. The preferred dilution method of this invention is: take 1.0 mL of industrial hydrochloric acid sample and dilute to 100 mL with deionized water to obtain a diluted solution of approximately 0.30% to 0.37%. In one embodiment of this invention, a quartz cuvette with a 10 mm optical path is used as the sample cell, and the absorbance value measured at 205 nm is approximately 0.8 to 1.2 AU, which is within the linear response range of the spectrometer. The sample solution temperature should be controlled within the range of 20 ± 2 °C to minimize the influence of temperature on the spectrum.

[0026] Regarding the spectral acquisition parameter settings, the preferred integration time of this invention is 50ms to 200ms, and the number of repeated scans per measurement is 5 to 20, preferably 10, to improve the signal-to-noise ratio. Dark current subtraction is achieved using dark spectra acquired under shading conditions, and deionized water from the same batch as the sample is used as a blank reference. The raw spectral data are stored in wavelength-absorbance pairs, with data point intervals of approximately 0.3nm to 0.5nm, and each spectrum contains approximately 2000 to 4000 data points.

[0027] Spectral preprocessing is a crucial step in ensuring the accuracy of subsequent modeling. The preprocessing workflow of this invention includes three sub-steps: baseline correction, scattering correction, and smoothing and denoising.

[0028] Baseline correction aims to eliminate dark current drift and stray light interference in the spectrometer. The baseline correction method used in this invention is as follows: The non-absorption regions at both ends of the spectrum, i.e., the 750nm to 800nm ​​band, are selected. The average absorbance value within this region is calculated as the baseline offset, and then this offset is subtracted from the absorbance value of the entire spectrum. For spectra with significant baseline tilt, a quadratic polynomial fitting is further used for correction. In one embodiment of this invention, the residual absorbance value of the spectrum in the non-absorption region after baseline correction is less than 0.005 AU.

[0029] Scattering correction aims to eliminate light scattering interference caused by suspended particles or bubbles in the sample. This invention preferably employs a standard normal transformation method for scattering correction. This method centers and normalizes the variance of each spectrum, ensuring that the spectra of different samples have the same mean and standard deviation, thereby effectively eliminating the effects of multiplicative scattering and baseline shift. In one embodiment of this invention, after performing a standard normal transformation on the spectra of 50 industrial hydrochloric acid samples, the standard deviation of the Euclidean distance between the spectra decreased from 0.85 to 0.32, significantly suppressing the scattering effect.

[0030] Smoothing and denoising aim to eliminate high-frequency random noise and improve the signal-to-noise ratio (SNR) of the spectrum. This invention preferably employs the Savitzky-Golay convolutional smoothing method, which effectively smooths high-frequency noise while preserving the peak shape and position. In one embodiment of this invention, the smoothing parameters are: a polynomial order of 2, a smoothing window width of 11 data points, corresponding to a wavelength range of approximately 5 nm. After smoothing, the SNR of the spectrum increases from approximately 200 to approximately 500, while the changes in the position and intensity of spectral characteristic peaks are less than 0.5%.

[0031] After the above preprocessing steps, the obtained standardized spectral data have good comparability and stability, providing high-quality input data for subsequent feature wavelength screening and modeling analysis.

[0032] Step S2: Feature wavelength range screening.

[0033] The second core step of this invention is the screening of characteristic wavelength ranges. For example... Figure 3 As shown, the purpose of this step is to screen out the characteristic wavelength ranges that are most relevant to the purity and impurity content of hydrochloric acid from the full-band spectral data, reduce redundant information interference, and improve the predictive performance and robustness of the model.

[0034] The theoretical basis for characteristic wavelength screening lies in the fact that the contribution of absorbance values ​​at different wavelengths to the concentration of the target analyte varies significantly. Some wavelength ranges carry crucial concentration information, while other wavelength ranges may only contain noise or interfering information. By screening the most representative characteristic wavelengths, prediction accuracy can be improved while simplifying model complexity.

[0035] This invention first performs a spectrum-concentration correlation analysis. For purity detection, each wavelength point in the preprocessed spectral data matrix is ​​used as the independent variable, and the mass fraction of hydrochloric acid determined by the standard method is used as the dependent variable. The Pearson correlation coefficient is calculated wavelength by wavelength. In one embodiment of this invention, the analysis results of 80 calibration set samples show that the absolute value of the correlation coefficient in the 200nm to 230nm band is generally greater than 0.95, indicating that the absorbance in this band is highly linearly correlated with the hydrochloric acid concentration; while the absolute value of the correlation coefficient in the 260nm to 380nm band is generally less than 0.3, indicating that this band is not sensitive to changes in hydrochloric acid concentration. For impurity detection, correlation analysis with iron impurity content as the dependent variable shows that the absolute value of the correlation coefficient in the 420nm to 450nm band is greater than 0.85, corresponding to the characteristic absorption of ferric ions; while arsenic impurities show a high correlation in the 325nm to 340nm band, corresponding to the characteristic absorption of arsenic-molybdenum heteropolyacid colorimetric complex.

[0036] Based on correlation analysis, this invention further employs the variable importance projection method for feature wavelength selection. The variable importance projection value is an important indicator in partial least squares models for evaluating the contribution of each wavelength variable to prediction. This value comprehensively considers the loading weight of the variable in each principal component and the explanatory power of each principal component to the variance of the dependent variable. This invention uses the following criteria for feature wavelength selection: for the purity detection model, wavelengths with a variable importance projection value greater than 1.0 are selected as candidate feature wavelengths; for the impurity detection model, wavelengths with a variable importance projection value greater than 0.8 are selected as candidate feature wavelengths. In one embodiment of this invention, a total of 87 candidate feature wavelengths are selected for purity detection, mainly distributed in the 203nm to 228nm range; and a total of 52 candidate feature wavelengths are selected for iron impurity detection, mainly distributed in the 418nm to 448nm range.

[0037] To further simplify the model and improve robustness, this invention integrates scattered candidate characteristic wavelengths into continuous characteristic wavelength ranges. Specifically, the candidate characteristic wavelengths are sorted by wavelength value, and wavelength points with an adjacent wavelength difference of less than 5 nm are merged into continuous ranges, while isolated ranges containing fewer than 10 wavelength points are removed. In one embodiment of this invention, the final determined purity characteristic wavelength range is 205 nm to 225 nm, containing approximately 60 wavelength points; the iron impurity characteristic wavelength range is 420 nm to 445 nm, containing approximately 50 wavelength points; and the arsenic impurity characteristic wavelength range is 328 nm to 338 nm, containing approximately 30 wavelength points.

[0038] This invention also introduces a wavelength range optimization selection mechanism to further improve model performance. This mechanism is based on a combined window moving method, expanding the search range by 10 nm both inside and outside the initially determined characteristic wavelength range, moving the window boundary with a step size of 2 nm, establishing a partial least squares model for each boundary combination, calculating the root mean square error of cross-validation, and selecting the boundary combination with the smallest error as the optimal characteristic wavelength range. In one embodiment of this invention, the optimized purity characteristic wavelength range is 207 nm to 223 nm, and the iron impurity characteristic wavelength range is 422 nm to 443 nm, reducing the model prediction root mean square error by approximately 8% and 12%, respectively.

[0039] Step S3: Establishment of the dual-objective partial least squares correction model.

[0040] The third core step of this invention is the establishment of a dual-objective partial least squares correction model. For example... Figure 4 As shown, the purpose of this step is to simultaneously establish a purity correction model and an impurity correction model within a unified multi-task learning framework, thereby achieving joint optimization of the model parameters.

[0041] Partial least squares regression is the core modeling algorithm of this invention. Its basic principle is to maximize the covariance between spectral and concentration variables by simultaneously extracting latent variables from the spectral data matrix and the concentration matrix, thereby establishing an optimal prediction model. Compared with principal component regression, partial least squares regression fully considers the correlation with concentration when extracting spectral principal components, thus exhibiting better predictive performance.

[0042] The dual-objective collaborative modeling strategy of this invention innovatively places purity correction and impurity correction under a unified framework for joint optimization. Let the spectral data matrix be... Its dimensions are ,in For the number of samples, Let be the number of wavelength points; let the concentration response matrix be... Its dimensions are ,in For the target analyte quantity, in this invention These correspond to the purity of hydrochloric acid, the content of iron impurities, and the content of arsenic impurities, respectively. The partial least squares regression model can be expressed as:

[0043] ,

[0044] ,

[0045] in: The score matrix has the following dimensions: , The number of latent variables. The value of is usually in the range of 3 to 15, and the optimal value is determined by cross-validation. Here is the spectral loading matrix, with dimensions of... This reflects the contribution of each wavelength variable to the latent variable; The concentration loading matrix has the following dimensions: This reflects the relationship between each target analyte and latent variables; and These are the spectral residual matrix and the concentration residual matrix, respectively, representing the error components that the model cannot explain.

[0046] The final prediction of the model can be expressed as:

[0047] ,

[0048] in: This is the predicted concentration matrix; The regression coefficient matrix has the following dimensions: It can be calculated using the following formula:

[0049] ,

[0050] in: Here is the weight matrix, with dimension 1. The matrix, obtained by a nonlinear iterative partial least squares algorithm, reflects the importance of each wavelength variable in the prediction.

[0051] Number of latent variables in this invention It is determined using leave-one-out cross-validation. Specifically, for a given... The value is used to sequentially select each sample in the calibration set as a validation sample, and the remaining samples are used to build a model and predict the concentration of that sample. The root mean square error of prediction for all samples is then calculated. The root mean square error of prediction is plotted as a function of... The changing curve is selected when the error begins to stabilize. The value is used as the optimal number of latent variables to avoid overfitting. In one embodiment of the present invention, cross-validation analysis of 80 calibration set samples shows that: when At that time, the root mean square error of purity prediction was 0.32%, and the root mean square error of iron impurity prediction was 4.2 ppm, further increasing... The improvement in time error was not significant, therefore the optimal number of latent variables was determined to be 7.

[0052] In the model building process, this invention employs a multi-task learning strategy to jointly optimize the purity model and the impurity model. Traditional methods establish purity and impurity models independently, neglecting the correlation between the two types of models. This invention recognizes that purity and impurity content are chemically intrinsically related; high-purity samples typically correspond to lower impurity content, thus the latent variable spaces of the two types of models partially overlap. This invention simultaneously fits purity and the content of each impurity as different columns of a unified response matrix in a single partial least squares model, enabling the extracted latent variables to simultaneously capture spectral information related to both purity and impurities, thereby improving the model's generalization ability.

[0053] Model performance evaluation metrics include the coefficient of determination. Corrected root mean square error, cross-validation root mean square error, and prediction root mean square error. Coefficient of determination. The formula reflects the model's ability to explain data variation and is as follows:

[0054] ,

[0055] in: For the first Reference values ​​for each sample; For the first Predicted values ​​for each sample; This is the average of the reference values ​​for all samples; This represents the number of samples. The value of is between 0 and 1, and the closer the value is to 1, the better the model fit.

[0056] The root mean square error (RMSEP) of the prediction reflects the average deviation of the model's predictions, and is calculated using the following formula:

[0057] ,

[0058] in: This represents the number of samples in the test set. A smaller RMSEP value indicates higher model prediction accuracy.

[0059] In one embodiment of the present invention, a model is built using 80 calibration set samples and externally validated using 20 independent test set samples. The model performance indicators are as follows: purity calibration model The value was 0.992, and the RMSEP was 0.38%; the iron impurity correction model... The value was 0.985, and the RMSEP was 5.1 ppm; the arsenic impurity correction model... The result shows that the bi-objective partial least squares correction model established in this invention has excellent predictive performance, with an accuracy of 0.978 and an RMSEP of 0.8 ppm.

[0060] Step S4: Impurity interference compensation and purity feedback correction.

[0061] The fourth core step of this invention is impurity interference compensation and purity feedback correction. For example... Figure 5 As shown, the purpose of this step is to eliminate the interference of impurity absorption on purity detection and further improve the accuracy of purity prediction.

[0062] The technical background for impurity interference compensation lies in the following: Although this invention locks the purity detection band to 205nm to 225nm through characteristic wavelength range screening, and the impurity detection band to the visible light region, the two do not overlap in their main wavelength bands. However, impurities such as iron and arsenic also exhibit a certain degree of background absorption in the ultraviolet region. When the impurity content is high, this background absorption will be superimposed on the absorption of hydrochloric acid itself, leading to an overestimation of the purity prediction value. Experimental data shows that for every 100ppm increase in iron impurity content, the absorbance at 205nm increases by approximately 0.008AU, corresponding to a purity deviation of approximately 0.15%.

[0063] To address the aforementioned issues, this invention innovatively introduces an impurity interference compensation mechanism. The core idea of ​​this mechanism is as follows: first, the impurity content of the sample is predicted using an impurity correction model; then, the interference contribution of the impurities in the purity detection band is calculated based on the known impurity-absorbance relationship; finally, this interference contribution is subtracted from the original purity prediction value.

[0064] The calculation of the interference compensation factor is crucial to this mechanism. This invention defines the interference compensation factor. The coefficient representing the influence of impurity absorption on purity prediction is calculated using the following formula:

[0065] ,

[0066] in: The total interference compensation factor, expressed as a percentage, represents the amount that needs to be deducted from the purity prediction value. The number of impurity types in this invention These correspond to iron impurities and arsenic impurities, respectively. For the first The interference coefficient per unit concentration of each impurity, expressed in % / ppm, was obtained through experimental calibration. For the first The predicted content of each impurity is expressed in ppm.

[0067] This invention calibrates the unit concentration interference coefficient of each impurity using the following method. Prepare a series of samples with the same hydrochloric acid concentration but increasing impurity content, and measure their absorbance changes in the purity detection band. Perform a linear regression of the absorbance increment against the impurity concentration increment; the regression coefficient is the spectral interference coefficient of that impurity. Then, convert the spectral interference coefficient into its impact on purity prediction to obtain... In one embodiment of the present invention, the calibrated interference coefficient per unit concentration of iron impurities is... The interference coefficient per unit concentration of arsenic impurities is 0.0015% / ppm. It is 0.0008% / ppm.

[0068] The formula for calculating the compensated purity prediction value is:

[0069] ,

[0070] in: The purity value of hydrochloric acid after compensation is expressed in % (%). This is the predicted purity value, in percent.

[0071] In one embodiment of the present invention, an industrial hydrochloric acid sample with an iron impurity content of 200 ppm and an arsenic impurity content of 30 ppm was tested. The original predicted purity value was 35.86%, the standard method determination value was 35.52%, and the deviation was 0.34%. After impurity interference compensation, the predicted value was corrected to 35.86% - 0.0015×200 - 0.0008×30 = 35.53%, and the deviation was reduced to 0.01%. This result fully demonstrates the effectiveness of the impurity interference compensation mechanism.

[0072] Building upon impurity interference compensation, this invention further introduces a purity feedback correction mechanism. The technical background of this mechanism is that the prediction accuracy of the impurity correction model is affected by the sample matrix concentration. When the hydrochloric acid concentration is within the boundary region of the calibration set concentration range, the impurity prediction may have a certain deviation. The purity feedback correction mechanism uses the purity prediction value to correct the impurity prediction, forming a closed-loop feedback optimization.

[0073] The specific implementation of purity feedback correction is as follows: A purity correction model is established at a certain standard concentration. When the actual concentration of the sample deviates from this standard concentration, a concentration correction factor is introduced to correct the predicted impurity value. Concentration correction factor The calculation formula is:

[0074] ,

[0075] in: This is a concentration correction factor, dimensionless. The sensitivity coefficient, which is dimensionless, is obtained through experimental calibration and typically ranges from 0.1 to 0.5. The standard concentration used in model establishment is 35% in this invention.

[0076] The corrected predicted impurity content is:

[0077] ,

[0078] in: The first after purity feedback correction The predicted content of various impurities, in ppm; For the first The original predicted value of each impurity, that is, the predicted content directly output by the impurity correction model, in ppm; The concentration correction factor is dimensionless, and its calculation method is shown in the aforementioned formula.

[0079] In one embodiment of the present invention, the calibration obtained correction sensitivity coefficient The value is 0.25. For an industrial hydrochloric acid sample with a purity of 32%, the original predicted value of iron impurities is 85 ppm. After purity feedback correction, the predicted value is 85 × [1 + 0.25 ×(32-35) / 35] = 83 ppm. The deviation from the value of 82 ppm determined by the standard method is reduced from 3.7% to 1.2%.

[0080] Through a dual mechanism of impurity interference compensation and purity feedback correction, this invention achieves deep coupling and collaborative optimization of the purity model and the impurity model, effectively eliminating mutual interference between the two types of detection, and achieving near-laboratory-level detection accuracy while ensuring rapid detection.

[0081] Step S5: Spectral quality evaluation and anomaly identification.

[0082] The fifth core step of this invention is spectral quality evaluation and anomaly identification. The purpose of this step is to identify and eliminate abnormal spectral data to ensure the reliability of the prediction results.

[0083] The sources of anomalous spectra mainly include: strong light scattering caused by bubbles or suspended particles in the sample; abnormal absorbance caused by contamination or fingerprints on the cuvette surface; signal distortion caused by light source fluctuations or detector saturation; and extrapolation errors caused by sample concentrations exceeding the model's applicable range. If these anomalous factors are not identified and addressed, they will severely affect the accuracy and reliability of the prediction results.

[0084] This invention employs a joint evaluation method using residual statistics and leverage values ​​for anomaly spectrum identification. Residual statistics The formula for calculating the deviation between the sample spectrum and the model-fitted spectrum is as follows:

[0085] ,

[0086] in: This is the residual sum of squares statistic, dimensionless; For the first Measured absorbance at each wavelength point; For the first Model reconstructed absorbance at each wavelength point; Let be the residual vector, and its elements ; This represents the number of wavelength points. The larger the value, the greater the deviation between the sample spectrum and the model, and the higher the probability of an anomaly.

[0087] Leverage value Reflects the distance of the sample to the center of the calibration set in the latent variable space, also known as Hotelling. The statistic is calculated using the following formula:

[0088] ,

[0089] in: This is a Hotelling statistic, dimensionless; For the sample at the Scores on each latent variable; For the correction set in the th The variance of scores on each latent variable; The sample score vector; To correct the set score covariance matrix; This represents the number of latent variables. The larger the value, the farther the sample is from the center of the calibration set, and the higher the risk of extrapolation.

[0090] This invention uses a 95% confidence limit as the anomaly detection threshold. threshold Calculated using the F-distribution of the corrected set residuals:

[0091] ,

[0092] in: , , To measure the eigenvalue statistics of the residual matrix of the correction set; For standard normal distribution Quantiles, take hour .

[0093] threshold Calculated using the F-distribution:

[0094] ,

[0095] in: To determine the number of samples in the calibration set; The number of latent variables; For degrees of freedom and The F-distribution at confidence level The critical value at that point.

[0096] This invention classifies spectra that meet any of the following conditions as anomalous spectra: or For spectra deemed abnormal, the system will issue a warning and suggest that the user re-acquire the sample spectrum or check the sample preparation process.

[0097] In one embodiment of the present invention, the calculated It is 0.0125. The value was 15.2. In practical testing applications, the quality of 1000 industrial hydrochloric acid sample spectra was evaluated, and 27 abnormal spectra (accounting for 2.7%) were identified, of which 18 were due to... Exceeding limits was identified (mainly due to air bubbles and suspended particles), 9 items were due to Exceeding limits was identified (mainly because the concentration exceeded the model range). After removing the abnormal spectra, the overall bias of the prediction results decreased from 0.58% to 0.41%, demonstrating the necessity and effectiveness of the anomaly identification mechanism.

[0098] Step S6: Comprehensive prediction and result output.

[0099] The sixth core step of this invention is comprehensive prediction and result output. After processing through the above five steps, the system applies a calibration model to predict the purity value and impurity content of the industrial hydrochloric acid sample based on the spectral data that has passed the quality evaluation, and outputs the purity value and impurity content of the sample.

[0100] The comprehensive prediction process is as follows: First, the preprocessed spectrum of the sample to be tested is... and The joint evaluation process continues prediction if the spectrum is determined to be normal; otherwise, an abnormality warning is issued. Then, the spectral data of the purity characteristic wavelength range and the impurity characteristic wavelength range are input into the corresponding correction model to obtain the original predicted values. Next, the interference compensation factor is calculated based on the impurity predicted value, and the purity predicted value is corrected. Subsequently, the impurity predicted value is corrected based on the corrected purity predicted value. Finally, the compensated and corrected purity value and impurity content are output.

[0101] The output includes: hydrochloric acid mass fraction (%), iron impurity content (ppm), arsenic impurity content (ppm), spectral quality assessment results, and confidence intervals. The confidence intervals are calculated based on the root mean square error of the model prediction.

[0102] ,

[0103] in: The confidence interval is 95%. These are predicted values.

[0104] In one embodiment of the present invention, the system output example is as follows: hydrochloric acid mass fraction 35.2%±0.8% (95% CI), iron impurity content 68ppm±10ppm (95% CI), arsenic impurity content 12ppm±2ppm (95% CI), spectral quality evaluation: normal.

[0105] To verify the accuracy and reliability of the method of the present invention, systematic methodological verification experiments were conducted.

[0106] Linearity range validation: A series of standard samples with hydrochloric acid mass fractions of 30%, 32%, 34%, 35%, 36%, and 37% were prepared, and each concentration was measured in triplicate. The results showed that within the range of 30% to 37%, the linear correlation coefficient between the predicted and reference values ​​was [missing information]. The value is 0.998, which meets the linear range requirement of the analysis method.

[0107] Precision verification: The same industrial hydrochloric acid sample was tested 10 times consecutively under the same conditions, and the relative standard deviation of the results was calculated. The RSD for purity determination was 0.45%, the RSD for iron impurity determination was 3.8%, and the RSD for arsenic impurity determination was 5.2%, all of which meet the precision requirements of the analytical method.

[0108] Accuracy verification: Accuracy was verified using spiked recovery experiments. Industrial hydrochloric acid samples of known concentrations were taken, and different amounts of iron and arsenic standard solutions were added, with the spiked recoveries determined. The spiked recoveries for iron impurities ranged from 96.5% to 103.2%, and for arsenic impurities from 94.8% to 105.6%, both within the acceptable range of 95% to 105%.

[0109] Method Comparison and Validation: Thirty actual industrial hydrochloric acid samples were tested using both the method of this invention and the national standard methods (titration and atomic absorption spectrometry). Paired t-tests were used to compare the results of the two methods. Statistical analysis showed that, at a 95% confidence level, there was no significant difference in the results between the two methods. This demonstrates that the method of the present invention has good consistency with the standard method.

[0110] Detection limit validation: The detection limits of each target analyte were determined using the signal-to-noise ratio method. The detection limit for iron impurities was 5 ppm, and the detection limit for arsenic impurities was 1 ppm, meeting the national standard requirements for impurity limits in industrial hydrochloric acid.

[0111] The above verification results fully demonstrate the accuracy, precision and reliability of the method of the present invention, and fully meet the technical requirements for rapid detection in the control of industrial hydrochloric acid flow.

[0112] Example 1: Routine quality control testing in industrial hydrochloric acid production enterprises.

[0113] I. Testing equipment and parameter configuration.

[0114] A detection system was constructed using a deuterium-tungsten composite light source (model DH-2000, spectral range 190nm to 2500nm, output power 20W) and a miniature fiber optic spectrometer (model USB4000, spectral range 200nm to 850nm, spectral resolution 1.5nm). The quartz fiber core diameter was 600μm, and the numerical aperture was 0.22. A 10mm optical path length quartz cuvette was used as the sample cell.

[0115] II. Sample preparation.

[0116] Industrial hydrochloric acid samples were extracted from the production line. 1.0 mL of the sample was accurately measured into a 100 mL volumetric flask using a pipette, and then diluted to the mark with deionized water to obtain a 100-fold diluted test solution. The sample solution temperature was controlled at 20±2℃.

[0117] III. Testing Procedures.

[0118] Step 1: Collect a blank reference spectrum, using deionized water as the reference, set the integration time to 100ms, and repeat the scan 10 times and take the average.

[0119] Step 2: Inject the sample solution to be tested into a cuvette and collect the absorption spectrum across the entire wavelength range from 190 nm to 800 nm.

[0120] Step 3: Perform baseline correction on the original spectrum, select the non-absorption region from 750nm to 800nm ​​to calculate the baseline offset and subtract it; perform scattering correction using standard normal transformation; and perform smoothing and denoising using the Savitzky-Golay method (second-order polynomial, window width 11 points).

[0121] Step 4: Extract spectral data for the purity characteristic wavelength range of 207nm to 223nm and the impurity characteristic wavelength range (iron: 422nm to 443nm, arsenic: 328nm to 338nm).

[0122] Step 5: Input the spectral data of the characteristic wavelength range into the pre-established partial least squares correction model (number of latent variables k=7) to obtain the original predicted values ​​of purity and impurities.

[0123] Step 6: Calculate the interference compensation factor The purity prediction value is compensated and corrected.

[0124] Step 7: Calculate the concentration correction factor Feedback correction is performed on the predicted impurity values.

[0125] Step 8: Calculation and Statistics, and thresholds ( =0.0125, =15.2) Compare and perform anomaly identification, and output the detection results.

[0126] IV. Test Results.

[0127] Samples from the production line were tested for five consecutive working days, with five samples taken each day. Simultaneously, the national standard methods (GB / T 622 titration method for purity determination and atomic absorption spectrometry for iron impurities) were used for comparative verification.

[0128] Table 1. Routine quality control test results of industrial hydrochloric acid production enterprises

[0129]

[0130] As shown in Table 1, the purity detection deviations for samples 1-5 were 0.11%, -0.14%, -0.11%, 0.08%, and -0.08%, respectively, while the iron impurity detection deviations were 5.9%, 3.2%, 3.5%, 3.6%, and -3.7%, respectively. The detection time for a single sample was approximately 45 seconds.

[0131] Example 2: On-site rapid detection application.

[0132] I. Testing equipment and parameter configuration.

[0133] A portable detection system is employed, comprising: a miniature deuterium-tungsten composite light source (15W power), a portable fiber optic spectrometer (spectral range 200nm to 800nm, resolution 1.8nm, integration time adjustable from 1ms to 10s), disposable plastic cuvettes (10mm optical path), and a portable temperature-controlled sample holder (temperature control accuracy ±1℃). The overall dimensions are 250mm × 180mm × 80mm, and it weighs 1.8kg. It is powered by a lithium battery.

[0134] II. On-site sample preparation.

[0135] Using the provided quantitative sampler, take 1.0 mL of the hydrochloric acid sample to be tested and add it to a dilution bottle pre-filled with 90 mL of deionized water. Shake to mix and let stand for 1 min to obtain a solution diluted approximately 100 times. Use a portable thermometer to confirm that the solution temperature is within the range of 18℃ to 22℃.

[0136] III. Testing Procedures.

[0137] Step 1: Power on and warm up for 3 minutes. The system will automatically complete the light source stabilization and dark current correction.

[0138] Step 2: Collect a reference spectrum using randomly matched deionized water, set the integration time to 150ms, and repeat the scan 5 times.

[0139] Step 3: Transfer the sample solution to be tested to a disposable cuvette, place it in the sample compartment, and press the start detection button.

[0140] Step 4: The system automatically completes spectral acquisition (full band from 190nm to 800nm), preprocessing (baseline correction, SNV scattering correction, SG smoothing), feature extraction (purity range 207nm to 223nm, impurity range 328nm to 338nm and 422nm to 443nm), model prediction, and result correction.

[0141] Step 5: After about 40 seconds, the screen will display the test results, including hydrochloric acid purity (%), iron impurity content (ppm), arsenic impurity content (ppm), and spectral quality evaluation status.

[0142] IV. Test Results.

[0143] Eight industrial hydrochloric acid samples from different sources were tested on-site, and samples were simultaneously sent to the laboratory for verification using standard methods.

[0144] Table 2. On-site test results of industrial hydrochloric acid samples from eight different sources.

[0145]

[0146] As shown in Table 2, on-site testing was conducted on eight industrial hydrochloric acid samples from different sources. The purity of the 36% grade sample from Plant A, measured using the method of this invention and the standard method, was 36.12% and 36.08% respectively, with iron contents of 42 ppm and 45 ppm respectively. The purity of the 35% grade sample from Plant B, measured using the method of this invention and the standard method, was 34.89% and 34.95% respectively, with iron contents of 68 ppm and 65 ppm respectively. The purity of the 31% grade sample from Plant C, measured using the method of this invention and the standard method, was 31.22% and 31.35% respectively, with iron contents of 125 ppm and 118 ppm respectively. The purity of the 36% grade sample from Plant D… The measured purities were 35.76% and 35.82%, respectively, with iron contents of 38 ppm and 41 ppm. For distributor E, the measured purities were 33.45% and 33.52%, respectively, with iron contents of 156 ppm and 148 ppm. For distributor F, the measured purities were 34.28% and 34.21%, respectively, with iron contents of 89 ppm and 92 ppm. For distributor G, the measured purities were 32.86% and 32.91%, respectively, with iron contents of 178 ppm and 185 ppm. Spectral quality assessment showed... Exceeding limits warning; the purities of samples from distributor H were measured at 35.15% and 35.18%, respectively, and the iron contents were 52 ppm and 49 ppm, respectively. Sample G... The statistic (17.8) exceeds the threshold (15.2) and is therefore identified as a boundary sample. The system issues a warning and suggests resampling for verification.

[0147] Example 3: Detection and verification of samples with high impurity content.

[0148] I. Experimental Objective.

[0149] Verify the effectiveness of the impurity interference compensation mechanism of the present invention in the detection of samples with high impurity content.

[0150] II. Sample preparation.

[0151] Based on analytical grade hydrochloric acid (purity 36.0%, iron content <1ppm, arsenic content <0.5ppm), different amounts of ferric chloride standard solution and sodium arsenite standard solution were added to prepare a series of samples with iron impurity contents of 50ppm, 100ppm, 150ppm, and 200ppm, and arsenic impurity contents of 10ppm, 20ppm, 30ppm, and 40ppm.

[0152] III. Testing Methods.

[0153] The method of this invention was used for detection, and the detection parameters were the same as in Example 1. To verify the effect of impurity interference compensation, the predicted purity values ​​before and after compensation were recorded respectively.

[0154] IV. Test Results.

[0155] Table 3. Detection results of samples with high impurity content

[0156]

[0157] As shown in Table 3, the purity of the sample with an iron content of 50 ppm and an arsenic content of 10 ppm was 36.08% before compensation and 36.01% after compensation, with a deviation of 0.08% before compensation and 0.01% after compensation; the purity of the sample with an iron content of 100 ppm and an arsenic content of 10 ppm was 36.16% before compensation and 36.01% after compensation, with a deviation of 0.16% before compensation and 0.01% after compensation; the purity of the sample with an iron content of 150 ppm and an arsenic content of 20 ppm was 36.27% before compensation. The purity of the sample after compensation was 36.02%, with a deviation of 0.27% before compensation and 0.02% after compensation. For the sample with 200 ppm iron and 30 ppm arsenic, the purity before compensation was 36.37%, and the purity after compensation was 36.01%, with a deviation of 0.37% before compensation and 0.01% after compensation. For the sample with 200 ppm iron and 40 ppm arsenic, the purity before compensation was 36.40%, and the purity after compensation was 36.02%, with a deviation of 0.40% before compensation and 0.02% after compensation. The true purity of all the above samples was 36.00%. Using the interference compensation mechanism of this invention, even under conditions of high impurity content (200 ppm iron and 40 ppm arsenic), the purity detection deviation can still be controlled within 0.02%, proving the effectiveness of the impurity interference compensation mechanism.

[0158] In summary, the rapid spectroscopic detection method for the purity and impurities of industrial hydrochloric acid provided by this invention achieves simultaneous and rapid detection of industrial hydrochloric acid purity and heavy metal impurity content through an innovative combination of ultraviolet-visible absorption spectroscopy and partial least squares chemometrics. It has the advantages of fast detection speed, high accuracy, simple operation, and no need for complex pretreatment. It is particularly suitable for rapid on-site screening in the control of industrial hydrochloric acid flow and for evidence collection by public security inspections, and has good application prospects and promotion value.

[0159] The embodiments of the present invention are not limited to the specific embodiments described above. Those skilled in the art can make various equivalent changes or substitutions based on the technical solutions of the present invention, and all such changes or substitutions should be included within the protection scope of the present invention.

Claims

1. A rapid spectroscopic method for detecting the purity and impurities of industrial hydrochloric acid, characterized in that, include: The diluted industrial hydrochloric acid sample solution was irradiated with a deuterium-tungsten composite light source. The ultraviolet absorption band spectral data and visible light absorption band spectral data were simultaneously acquired by a miniature fiber optic spectrometer. The wavelength range of the ultraviolet absorption band is 190nm to 280nm, and the wavelength range of the visible light absorption band is 380nm to 780nm. The collected ultraviolet absorption band spectral data and visible light absorption band spectral data are sequentially subjected to baseline correction, scattering correction and smoothing and denoising preprocessing to obtain standardized spectral data. The baseline correction is achieved by selecting the non-absorption region of the spectrum, calculating the baseline offset, and subtracting the baseline offset from the original spectral data. The scattering correction adopts the standard normal transformation method to eliminate the multiplicative scattering effect. Based on the correlation analysis between the ultraviolet absorption band spectral data and hydrochloric acid concentration, and the correlation analysis between the visible light absorption band spectral data and impurity content, the variable importance projection method is used to screen the purity characteristic wavelength range and the impurity characteristic wavelength range from the standardized spectral data, respectively. The spectral data corresponding to the purity characteristic wavelength range and the impurity characteristic wavelength range are the characteristic wavelength range spectral data. Using the spectral data corresponding to the selected purity characteristic wavelength range and the impurity characteristic wavelength range as input, and the hydrochloric acid purity and impurity content as output, a partial least squares regression algorithm is used to establish a purity correction model and an impurity correction model respectively. The model parameters are jointly optimized under a unified multi-task learning framework. By using purity and the content of each impurity as different columns of a unified response matrix, they are simultaneously fitted in a single partial least squares model, so that the extracted latent variables can simultaneously capture spectral information related to purity and impurities. The spectral data of the sample to be tested are evaluated by a combination of residual statistics and leverage values ​​to identify abnormal spectra. If the spectrum is determined to be normal, prediction continues. The spectral data of the purity characteristic wavelength range and the impurity characteristic wavelength range are respectively input into the corresponding correction model to obtain the original predicted value. The interference contribution of impurity absorption to the purity detection band is calculated to obtain the interference compensation factor. The prediction results of the impurity correction model are used to correct the prediction results of the purity correction model to obtain the compensated hydrochloric acid purity value. The compensated purity prediction value is the original purity prediction value minus the interference compensation factor. A concentration correction factor is introduced to correct the impurity prediction value. The impurity prediction value is corrected based on the corrected purity prediction value. The corrected impurity content prediction value is the original impurity prediction value multiplied by the concentration correction factor. Finally, the compensated purity and impurity content are output.

2. The rapid spectroscopic detection method for purity and impurities of industrial hydrochloric acid according to claim 1, characterized in that, The purity characteristic wavelength range is 205nm to 225nm, and the impurity characteristic wavelength range includes the iron impurity characteristic wavelength range of 420nm to 445nm and the arsenic impurity characteristic wavelength range of 328nm to 338nm.

3. The rapid spectroscopic detection method for purity and impurities of industrial hydrochloric acid according to claim 1, characterized in that, The industrial hydrochloric acid sample solution is diluted 80 to 120 times, and the mass fraction of hydrochloric acid in the diluted solution is 0.25% to 0.50%. The temperature of the sample solution is controlled within the range of 18°C ​​to 22°C.

4. The rapid spectroscopic detection method for purity and impurities of industrial hydrochloric acid according to claim 1, characterized in that, The number of latent variables in the partial least squares regression algorithm is determined by leave-one-out cross-validation. The number of latent variables ranges from 3 to 15, and the value at which the root mean square error of the prediction begins to stabilize is selected as the optimal number of latent variables.

5. The rapid spectroscopic detection method for purity and impurities of industrial hydrochloric acid according to claim 1, characterized in that, The formula for calculating the interference compensation factor is as follows: , in, The total interference compensation factor is... For the types and quantities of impurities, For the first The interference coefficient per unit concentration of the impurities, For the first Predicted content of various impurities.

6. The rapid spectroscopic detection method for purity and impurities of industrial hydrochloric acid according to claim 5, characterized in that, The unit concentration interference coefficient is calibrated by the following method: a series of samples with the same hydrochloric acid concentration but increasing impurity content are prepared, and the absorbance change in the purity detection band is measured. The spectral interference coefficient is obtained by linear regression of the absorbance increment with the impurity concentration increment. The spectral interference coefficient is then converted into the influence on purity prediction to obtain the unit concentration interference coefficient.

7. The rapid spectroscopic detection method for purity and impurities of industrial hydrochloric acid according to claim 1, characterized in that, The feedback correction includes introducing a concentration correction factor to correct the impurity prediction value. The formula for calculating the concentration correction factor is as follows: , in, For concentration correction factor, To correct the sensitivity coefficient, The purity value of hydrochloric acid after compensation. The standard concentration used when building the model.

8. The rapid spectroscopic detection method for purity and impurities of industrial hydrochloric acid according to claim 1, characterized in that, The residual statistic is the sum of squared residuals. The lever value is obtained by calculating the sum of squared deviations between the measured spectrum and the model-reconstructed spectrum; the lever value is Hotelling. The statistic is obtained by calculating the Mahalanobis distance between the sample and the center of the calibration set in the latent variable space.

9. The rapid spectroscopic detection method for purity and impurities of industrial hydrochloric acid according to claim 8, characterized in that, The determination of abnormal spectra uses a 95% confidence limit as the threshold, and will satisfy... Greater than its threshold or Greater than its threshold If the spectrum under any condition is determined to be abnormal, a warning will be given for the abnormal spectrum and it will be recommended to reacquire the sample spectrum.

10. The rapid spectroscopic detection method for purity and impurities of industrial hydrochloric acid according to claim 1, characterized in that, The acquisition parameters for the spectral data include: integration time of 50ms to 200ms, number of repeated scans per measurement of 5 to 20, acquisition wavelength range of 190nm to 800nm, and wavelength resolution of 0.5nm to 2.0nm; deionized water is used as a blank reference for reference spectral acquisition.

Images