Potentiometric titration end point prediction and abnormal value correction method based on deep learning
By using deep learning methods to determine the type of anomalies in potentiometric titration and make corrections, the problem of inaccurate judgment of weak anomalies in potentiometric titration technology is solved, and the accuracy and precision of the analysis results are improved.
Patent Information
- Application Number
- CN202510772873.2
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-06-11
- Publication Date
- 2025-09-23
- Estimated Expiration
- 2045-06-11
AI Technical Summary
Existing potentiometric titration technology is difficult to accurately judge weak anomalies, resulting in large errors in analysis results, and lacks an effective multi-dimensional anomaly detection and adaptive correction mechanism.
A deep learning-based method is used to accurately determine whether a potential is a suspected abnormal potential through indicators such as potential change slope, monotonicity, mutation amplitude, local curvature change rate and time series correlation. Appropriate correction methods, such as the K-nearest neighbor algorithm or window sliding, are selected according to the type of anomaly to reduce analysis errors.
The accuracy of the potentiometric titration technology in judging weak abnormal situations is improved, the error of the analysis results is reduced, and real-time abnormality detection and accurate endpoint prediction are achieved.
Smart Images
Figure CN120685847A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of potentiometric titration prediction and correction, and in particular to a method for potentiometric titration endpoint prediction and outlier correction based on deep learning. Background Art
[0002] In the field of potentiometric titration analysis, traditional methods rely on manual observation of potential jumps or simple first-order derivative calculations to determine the titration endpoint. This results in poor endpoint accuracy and susceptibility to human interference. Furthermore, abnormal potential data is often generated during the experiment due to electrode fluctuations, environmental noise, and instrument failures. Existing technologies lack effective multi-dimensional anomaly detection and adaptive correction mechanisms, making it difficult to distinguish strong from weak anomalies. The ability to handle complex anomalies is insufficient, resulting in low repeatability and large errors in analytical results. With the increasing demand for analytical accuracy and automation in the chemical, environmental, and pharmaceutical industries, a potentiometric titration method that can achieve real-time anomaly detection, accurate endpoint prediction, and intelligent data correction is urgently needed.
[0003] For example, Chinese patent application publication number: CN114414648A discloses an automatic potentiometric titration method and system based on machine learning, which relates to the field of automatic potentiometric titration technology. The method includes the following steps: obtaining the initial potential of the solution to be tested; identifying the label information of a burette filled with a titrant, the label information including the titration concentration of the titrant; predicting the titration end point based on the titration concentration and the initial potential of the solution through a prediction model; titrating the solution to be tested through the burette and collecting real-time titration data of the solution to be tested during the titration analysis; judging whether the titration analysis has reached the termination condition based on the real-time titration data and the titration end point; if the titration analysis has reached the termination condition, the titration analysis is terminated, and the ion concentration of the ion to be tested in the solution to be tested is calculated based on the real-time titration data at the time of the titration analysis termination. This application has the effect of predicting the titration end point based on a machine learning prediction model to improve the accuracy of ion concentration detection.
[0004] However, the existing technology has the problem that the potentiometric titration technology is difficult to accurately judge weak abnormal situations, resulting in large errors in the analysis results. Summary of the Invention
[0005] To this end, the present invention provides a method for predicting endpoints and correcting outliers in potentiometric titration based on deep learning, so as to overcome the problem in the prior art that potentiometric titration technology is difficult to accurately judge weak anomalies, resulting in large errors in analysis results.
[0006] To achieve the above objectives, the present invention provides a method for predicting endpoints and correcting outliers in potentiometric titration based on deep learning, comprising:
[0007] Step S1, obtaining potential change data during potentiometric titration, basic data of the titrant, and parameters of the sample to be tested;
[0008] Step S2, determining whether the potential is a suspected abnormal potential based on whether the potential change slope exceeds a preset change slope and / or whether the potential change violates the monotonicity of the potential change;
[0009] Step S3, predicting the abnormal type of the suspected abnormal potential based on a comparison result of the mutation amplitude of the suspected abnormal potential with the first preset mutation amplitude and the second preset mutation amplitude;
[0010] Step S4, predicting whether the current potential is a weak abnormal potential or a normal potential based on whether the local curvature change rate of the adjacent potential titrated by the weak abnormal potential is greater than a preset curvature change rate and time series correlation;
[0011] Step S5 , based on the comparison result of the abnormal potential duration with the preset duration and the comparison result of the deviation degree of the abnormal potential titration abnormal value with the preset deviation degree, determining to correct the abnormal potential titration using different correction methods.
[0012] Further, in step S2, determining whether the potential is a suspected abnormal potential based on whether the potential change slope exceeds a preset change slope and whether the potential change violates the monotonicity of the potential change includes:
[0013] If the potential change slope exceeds a preset change slope or the potential change violates the monotonicity of potential change, determining that the potential is a suspected abnormal potential;
[0014] If the potential change slope does not exceed the preset change slope and the potential change conforms to the monotonicity of potential change, it is determined that the potential is a normal potential.
[0015] Furthermore, the potential change slope is determined based on the ratio of the potential difference between two adjacent sampling points to the difference in the amount of titrant added, the preset change slope is determined based on the average potential change slope during several potentiometric titrations of the same type, and the monotonicity of the potential change is determined based on the potential value change trend after the titrant is added.
[0016] Furthermore, in step S3, based on the comparison result of the suspected abnormal potential and the mutation amplitude of the suspected abnormal potential with the first preset mutation amplitude and the second preset mutation amplitude, predicting the abnormal type of the suspected abnormal potential includes:
[0017] If the mutation amplitude of the suspected abnormal potential is greater than the first preset mutation amplitude, it is predicted that the abnormal type of the suspected abnormal potential is a strong abnormal potential;
[0018] If the mutation amplitude of the suspected abnormal potential is less than or equal to the first preset mutation amplitude and greater than the second preset mutation amplitude, it is predicted that the abnormal type of the suspected abnormal potential is a weak abnormal potential;
[0019] If the mutation amplitude of the suspected abnormal potential is less than or equal to the second preset mutation amplitude, it is predicted that the potential is not abnormal.
[0020] Furthermore, the mutation amplitude of the suspected abnormal potential is determined based on the absolute value of the potential difference between the suspected abnormal potential and the adjacent normal point, the first preset mutation amplitude is determined based on the maximum amplitude of normal fluctuations in historical experimental data, and the second preset mutation amplitude is half of the first preset amplitude.
[0021] Further, in step S4, predicting whether the current potential is a weak abnormal potential or a normal potential based on whether the local curvature change rate of the adjacent potential titrated by the weak abnormal potential is greater than a preset curvature change rate and time series correlation includes:
[0022] If the local curvature change rate of the adjacent potentials titrated by the weak abnormal potential is greater than the preset curvature change rate and the temporal correlation of the weak abnormal potential is less than the preset temporal correlation, it is predicted that the current potential is a weak abnormal potential;
[0023] If the local curvature change rate of the adjacent potentials titrated by the weak abnormal potential is less than or equal to the preset curvature change rate and the timing correlation of the weak abnormal potential is greater than or equal to the preset timing correlation, it is predicted that the current potential is a normal potential.
[0024] Furthermore, the local curvature change rate of adjacent potentials of the weak abnormal potential titration is determined based on the first-order derivative and the second-order derivative of the potential with respect to time, and the preset curvature change rate is determined based on the average local curvature change rate of adjacent potentials of the same type of weak abnormal potential titration.
[0025] Furthermore, the timing correlation is determined based on a dynamic time warping distance, and the preset timing correlation is determined based on a timing correlation distribution of all sequence pairs in historical normal data.
[0026] Furthermore, in step S5, determining to correct the abnormal potential titration using different correction methods based on whether the duration of the abnormal potential is greater than a preset duration and a comparison result of the deviation degree of the abnormal value of the abnormal potential titration with a preset deviation degree includes:
[0027] If the duration of the abnormal potential is less than a preset duration and the deviation degree of the abnormal value is less than a preset deviation degree, it is determined that the abnormal potential titration does not need to be corrected;
[0028] If the duration of the abnormal potential is less than a preset duration and the deviation degree of the abnormal value of the abnormal potential titration is greater than or equal to the preset deviation degree, determining to use the K nearest neighbor algorithm to correct the abnormal potential titration;
[0029] If the duration of the abnormal potential is greater than or equal to a preset duration and the deviation degree of the abnormal value of the abnormal potential titration is less than a preset deviation degree, determining to correct the abnormal potential titration by a window sliding method;
[0030] If the duration of the abnormal potential is greater than or equal to a preset duration and the deviation degree of the abnormal value of the abnormal potential titration is greater than or equal to a preset deviation degree, it is determined that the potentiometric titration is to be re-prepared.
[0031] Furthermore, the preset duration is determined based on the historical average duration of abnormal potential under standard conditions, the degree of deviation of the abnormal value is determined based on the absolute value of the potential difference between the abnormal value and the ideal titration curve at the corresponding point, and the preset deviation degree is determined based on the average deviation degree of several same type of potentiometric titrations under standard conditions.
[0032] Compared with the prior art, the beneficial effect of the present invention lies in that the present invention determines whether the potential is a suspected abnormal potential by whether the potential change slope exceeds the preset change slope and whether the potential change violates the monotonicity of the potential change. According to the fact that the potential change slope exceeds the preset change slope or the potential change violates the monotonicity of the potential change, it is indicated that a potential mutation or an abnormal trend occurs, and the potential is accurately determined to be a suspected abnormal potential. According to the fact that the potential change slope does not exceed the preset change slope and the potential change complies with the monotonicity of the potential change, it is indicated that the potential change rate is normal and the trend is consistent, and the potential is accurately determined to be a normal potential. The above content improves the accuracy of the potentiometric titration technology in judging weak abnormal situations, thereby reducing the error of the analysis results.
[0033] Furthermore, the present invention predicts the abnormal type of the suspected abnormal potential by comparing the suspected abnormal potential and the mutation amplitude of the suspected abnormal potential with the first preset mutation amplitude and the second preset mutation amplitude. According to the fact that the mutation amplitude of the suspected abnormal potential is greater than the first preset mutation amplitude, it indicates that the potential value has undergone a significant jump and exceeds the upper limit of the normal fluctuation range. The abnormal type of the suspected abnormal potential is accurately predicted to be a strong abnormal potential. According to the fact that the mutation amplitude of the suspected abnormal potential is less than or equal to the first preset mutation amplitude and greater than the second preset mutation amplitude, it indicates that there is a detectable abnormal fluctuation in the potential value but has not reached the level of a significant jump. The abnormal type of the suspected abnormal potential is accurately predicted to be a weak abnormal potential. According to the fact that the mutation amplitude of the suspected abnormal potential is less than or equal to the second preset mutation amplitude, it indicates that the potential fluctuation is within the normal noise range or only the inherent error of the instrument exists. The potential is accurately predicted to be not abnormal. The above content improves the accuracy of the potentiometric titration technology in judging weak abnormal situations, thereby reducing the error of the analysis results.
[0034] Furthermore, the present invention predicts whether the current potential is a weak abnormal potential or a normal potential by whether the local curvature change rate of the adjacent potentials of the weak abnormal potential titration is greater than the preset curvature change rate and the time series correlation. According to the fact that the local curvature change rate of the adjacent potentials of the weak abnormal potential titration is greater than the preset curvature change rate and the time series correlation of the weak abnormal potential is less than the preset time series correlation, it indicates that the local curvature degree of the potential curve exceeds the normal range, and there is a regular difference between the current fluctuation pattern and the historical normal sequence. The current potential is accurately predicted to be a weak abnormal potential. According to the fact that the local curvature change rate of the adjacent potentials of the weak abnormal potential titration is less than or equal to the preset curvature change rate and the time series correlation of the weak abnormal potential is greater than or equal to the preset time series correlation, it indicates that the potential curve is smooth, the trend is consistent, and the fluctuation pattern is highly similar to the historical normal sequence. The current potential is accurately predicted to be a normal potential. The above content improves the accuracy of the potential titration technology in judging weak abnormal situations, thereby reducing the error of the analysis results.
[0035] Furthermore, the present invention determines whether the duration of the abnormal potential is greater than a preset duration and the deviation degree of the abnormal value of the abnormal potential titration is compared with the preset deviation degree to determine whether the abnormal potential titration is corrected by different correction methods. According to the fact that the duration of the abnormal potential is less than the preset duration and the deviation degree of the abnormal value is less than the preset deviation degree, it indicates that a normal noise phenomenon of slight fluctuation and self-recovery occurs, and it is accurately determined that the abnormal potential titration does not need to be corrected. According to the fact that the duration of the abnormal potential is less than the preset duration and the deviation degree of the abnormal value of the abnormal potential titration is greater than or equal to the preset deviation degree, it indicates that a sudden strong interference but not continuous isolated abnormal phenomenon occurs, and it is accurately determined that the K nearest neighbor is used. The algorithm corrects the abnormal potentiometric titration. Based on the fact that the duration of the abnormal potential is greater than or equal to the preset duration and the degree of deviation of the abnormal value of the abnormal potentiometric titration is less than the preset deviation degree, it indicates that a long-standing but small-amplitude systematic drift phenomenon has occurred. The algorithm accurately determines to correct the abnormal potentiometric titration by a window sliding method. Based on the fact that the duration of the abnormal potential is greater than or equal to the preset duration and the degree of deviation of the abnormal value of the abnormal potentiometric titration is greater than or equal to the preset deviation degree, it indicates that a serious and persistent fatal abnormal phenomenon has occurred. The algorithm accurately determines to re-prepare the potentiometric titration. The above content improves the accuracy of the potentiometric titration technology in judging weak abnormal situations, thereby reducing the error of the analysis results. BRIEF DESCRIPTION OF THE DRAWINGS
[0036] Figure 1 This is a workflow diagram of a method for predicting endpoints and correcting outliers in potentiometric titration based on deep learning according to an embodiment of the present invention;
[0037] Figure 2A flowchart of a method for predicting endpoints and correcting outliers in potentiometric titrations based on deep learning according to an embodiment of the present invention for determining whether the potential is a suspected abnormal potential.
[0038] Figure 3 A flowchart of a method for predicting the abnormal type of the suspected abnormal potential based on deep learning for predicting the endpoint of potentiometric titration and correcting abnormal values according to an embodiment of the present invention;
[0039] Figure 4 This is a workflow diagram for predicting whether the current potential is a weak abnormal potential or a normal potential based on the deep learning-based potentiometric titration endpoint prediction and outlier correction method in an embodiment of the present invention. DETAILED DESCRIPTION
[0040] In order to make the objects and advantages of the present invention more clearly understood, the present invention is further described below in conjunction with embodiments; it should be understood that the specific embodiments described herein are merely used to explain the present invention and are not intended to limit the present invention.
[0041] The preferred embodiments of the present invention are described below with reference to the accompanying drawings. It should be understood by those skilled in the art that these embodiments are only used to explain the technical principles of the present invention and are not intended to limit the scope of protection of the present invention.
[0042] Furthermore, it should be noted that, in the description of the present invention, unless otherwise expressly specified or limited, the terms "mounted," "connected," and "connected" should be understood in a broad sense. For example, they may refer to fixed connections, detachable connections, or integral connections; mechanical connections or electrical connections; direct connections or indirect connections through an intermediate medium; and internal communication between two components. Those skilled in the art will understand the specific meanings of the above terms in the present invention based on specific circumstances.
[0043] See also Figures 1-4 As shown, Figure 1 This is a workflow diagram of a method for predicting endpoints and correcting outliers in potentiometric titration based on deep learning according to an embodiment of the present invention; Figure 2 A flowchart of a method for predicting endpoints and correcting outliers in potentiometric titrations based on deep learning according to an embodiment of the present invention for determining whether the potential is a suspected abnormal potential. Figure 3 A flowchart of a method for predicting the abnormal type of the suspected abnormal potential based on deep learning for predicting the endpoint of potentiometric titration and correcting abnormal values according to an embodiment of the present invention; Figure 4 This is a workflow diagram for predicting whether the current potential is a weak abnormal potential or a normal potential based on the deep learning-based potentiometric titration endpoint prediction and outlier correction method in an embodiment of the present invention.
[0044] The method for predicting endpoints and correcting outliers in potentiometric titrations based on deep learning in an embodiment of the present invention includes:
[0045] Step S1, obtaining potential change data during potentiometric titration, basic data of the titrant, and parameters of the sample to be tested;
[0046] Step S2, determining whether the potential is a suspected abnormal potential based on whether the potential change slope exceeds a preset change slope and / or whether the potential change violates the monotonicity of the potential change;
[0047] Step S3, predicting the abnormal type of the suspected abnormal potential based on a comparison result of the mutation amplitude of the suspected abnormal potential with the first preset mutation amplitude and the second preset mutation amplitude;
[0048] Step S4, predicting whether the current potential is a weak abnormal potential or a normal potential based on whether the local curvature change rate of the adjacent potential titrated by the weak abnormal potential is greater than a preset curvature change rate and time series correlation;
[0049] Step S5 , based on the comparison result of the abnormal potential duration with the preset duration and the comparison result of the deviation degree of the abnormal potential titration abnormal value with the preset deviation degree, determining to correct the abnormal potential titration using different correction methods.
[0050] The potential change data during the potentiometric titration process described in the embodiments of the present invention include but are not limited to "real-time potential value, potential change rate and second-order derivative", the basic data of the titrant include but are not limited to "titrant concentration, titrant type and titrant addition rate", and the parameters of the sample to be tested include but are not limited to "sample initial volume, sample initial pH value and sample temperature".
[0051] Specifically, in step S2, when determining whether the potential is a suspected abnormal potential, whether the potential is a suspected abnormal potential is determined based on whether the potential change slope exceeds a preset change slope and whether the potential change violates the monotonicity of the potential change;
[0052] If the potential change slope exceeds a preset change slope or the potential change violates the monotonicity of potential change, determining that the potential is a suspected abnormal potential;
[0053] If the potential change slope does not exceed the preset change slope and the potential change conforms to the monotonicity of potential change, it is determined that the potential is a normal potential.
[0054] In the embodiment of the present invention, the potential change slope is determined according to the ratio of the potential difference between two adjacent sampling points to the difference in the amount of titrant added, and is calculated as the difference between the current potential and the adjacent potential divided by the amount of titrant added at the current potential minus the amount of titrant added at the adjacent potential. For example, the current potential value E i =250mV, titrant addition amount V i=10.00mL, adjacent potential value E i+1 =265mV, titrant addition amount V i+1 =10.10mL, and the potential change slope is (265mV-250mV) / (10.10mL-10mL)=150mV / mL. The preset change slope is determined based on the average potential change slopes of several groups of the same type of potentiometric titrations. For example, the average potential change slopes of two groups of the same type of potentiometric titrations are 140mV / mL and 160mV / mL, respectively, resulting in a preset change slope of 150mV / mL. The monotonicity of the potential change is determined based on the consistency of the potential value change trend after the titrant is added. For example, as the titrant is continuously added, the potential value is detected once per minute. The potential values within three minutes are 450mV, 420mV, and 390mV, respectively, and the potential always maintains a downward trend. It is determined that the potential change meets the monotonicity of the potential change. However, the above value is not limited to this, and those skilled in the art can also adjust the value according to actual needs.
[0055] The present invention determines whether the potential is a suspected abnormal potential by whether the potential change slope exceeds a preset change slope and whether the potential change violates the monotonicity of the potential change. According to the fact that the potential change slope exceeds the preset change slope or the potential change violates the monotonicity of the potential change, it is indicated that a potential mutation or an abnormal trend occurs, and the potential is accurately determined to be a suspected abnormal potential. According to the fact that the potential change slope does not exceed the preset change slope and the potential change complies with the monotonicity of the potential change, it is indicated that the potential change rate is normal and the trend is consistent, and the potential is accurately determined to be a normal potential. The above content improves the accuracy of the potentiometric titration technology in judging weak abnormal situations, thereby reducing the error of the analysis results.
[0056] Specifically, in step S3, when predicting the abnormal type of the suspected abnormal potential, the abnormal type of the suspected abnormal potential is predicted based on a comparison result of the suspected abnormal potential and the mutation amplitude of the suspected abnormal potential with a first preset mutation amplitude and a second preset mutation amplitude;
[0057] If the mutation amplitude of the suspected abnormal potential is greater than the first preset mutation amplitude, it is predicted that the abnormal type of the suspected abnormal potential is a strong abnormal potential;
[0058] If the mutation amplitude of the suspected abnormal potential is less than or equal to the first preset mutation amplitude and greater than the second preset mutation amplitude, it is predicted that the abnormal type of the suspected abnormal potential is a weak abnormal potential;
[0059] If the mutation amplitude of the suspected abnormal potential is less than or equal to the second preset mutation amplitude, it is predicted that the potential is not abnormal.
[0060] The sudden change amplitude of the suspected abnormal potential in the embodiment of the present invention is determined according to the absolute value of the potential difference between the suspected abnormal potential and the adjacent normal point, and is calculated as the sum of the potential value of the suspected abnormal potential minus the potential value of the previous normal potential minus the potential value of the next normal potential divided by 2. For example, the potential value of the previous normal point is: prev =300mV; suspected abnormal point: potential value E suspected =350mV; the next normal point: potential value E next =305mV, 350mV-(305mV+300mV) / 2=47.5mV, the first preset mutation amplitude range is 30mV-50mV, and the preferred value is 40mV. The reason for selecting this preferred value is based on the statistical characteristics of a large amount of experimental data. This preferred value can not only ensure sensitivity to real anomalies, but also avoid overreaction to normal fluctuations. The second preset mutation amplitude is half of the first preset amplitude. For example, the first preset mutation amplitude is 40mV, and the second preset mutation amplitude is 40mV / 2=20mV, but the above values are not limited to this. Those skilled in the art can also adjust the value according to actual needs.
[0061] The present invention predicts the abnormal type of the suspected abnormal potential through a comparison result of the suspected abnormal potential and the mutation amplitude of the suspected abnormal potential with the first preset mutation amplitude and the second preset mutation amplitude. According to the fact that the mutation amplitude of the suspected abnormal potential is greater than the first preset mutation amplitude, it indicates that a significant jump in the potential value occurs and exceeds the upper limit of the normal fluctuation range. The abnormal type of the suspected abnormal potential is accurately predicted to be a strong abnormal potential. According to the fact that the mutation amplitude of the suspected abnormal potential is less than or equal to the first preset mutation amplitude and greater than the second preset mutation amplitude, it indicates that a perceptible abnormal fluctuation in the potential value occurs but does not reach the level of a significant jump. The abnormal type of the suspected abnormal potential is accurately predicted to be a weak abnormal potential. According to the fact that the mutation amplitude of the suspected abnormal potential is less than or equal to the second preset mutation amplitude, it indicates that the potential fluctuation is within the normal noise range or only the inherent error of the instrument exists. The potential is accurately predicted to be not abnormal. The above content improves the accuracy of the potentiometric titration technology in judging weak abnormal situations, thereby reducing the error of the analysis results.
[0062] Specifically, in step S4, when the current potential is predicted to be a weak abnormal potential or a normal potential, the current potential is predicted to be a weak abnormal potential or a normal potential based on whether the local curvature change rate of the adjacent potential titrated by the weak abnormal potential is greater than the preset curvature change rate and the time series correlation;
[0063] If the local curvature change rate of the adjacent potentials titrated by the weak abnormal potential is greater than the preset curvature change rate and the temporal correlation of the weak abnormal potential is less than the preset temporal correlation, it is predicted that the current potential is a weak abnormal potential;
[0064] If the local curvature change rate of the adjacent potentials titrated by the weak abnormal potential is less than or equal to the preset curvature change rate and the timing correlation of the weak abnormal potential is greater than or equal to the preset timing correlation, it is predicted that the current potential is a normal potential.
[0065] The local curvature change rate of adjacent potentials in the weak abnormal potential titration described in the embodiment of the present invention is determined based on the first and second derivatives of the potential with respect to time. The calculation method is the absolute value of the second derivative of the potential with respect to time, divided by (1 plus the square of the first derivative of the potential with respect to time) to the power of three times. For example, the first derivative of the potential with respect to time is 5mV / s, and the second derivative of the potential with respect to time is 3mV / s. 2 , and the curvature change rate is 0.02. The preset curvature change rate is determined by the average local curvature change rate of adjacent potentials of the same type of weak abnormal potential titration. For example, the local curvature change rates of adjacent potentials of two groups of the same type of weak abnormal potential titration are 0.03 and 0.04, and the preset curvature change rate is 0.035. However, the above value is not limited to this, and those skilled in the art can also adjust the value according to actual needs.
[0066] The timing correlation in the embodiment of the present invention is determined according to the dynamic time warping distance. For example, the current potential sequence is detected to be [300, 305], and the normal positioning sequence is [295, 300]. The 2×2 matrix is obtained: D=[|300-295|||305-295|||300-300|||305-300|]=
[51005] . The shortest path is (1,1)→(1,2)→(2,2). The dynamic time warping distance is The distance is 10, and the dynamic regularization distance is normalized to 1 / (1+10) to obtain the time series correlation of 0.09. The preset time series correlation range is set to 0.07-0.09, and the preferred value is 0.08. The reason for selecting this preferred value is that if the preferred value is too small, early minor anomalies may be missed, and if the preferred value is too large, critical anomalies may be missed. However, the above values are not limited to this, and those skilled in the art can also adjust the value according to actual needs.
[0067] The present invention predicts whether the current potential is a weak abnormal potential or a normal potential by whether the local curvature change rate of the adjacent potentials of the weak abnormal potential titration is greater than the preset curvature change rate and the time series correlation. According to the fact that the local curvature change rate of the adjacent potentials of the weak abnormal potential titration is greater than the preset curvature change rate and the time series correlation of the weak abnormal potential is less than the preset time series correlation, it is indicated that the local curvature degree of the potential curve exceeds the normal range, and there is a regular difference between the current fluctuation pattern and the historical normal sequence. The current potential is accurately predicted to be a weak abnormal potential. According to the fact that the local curvature change rate of the adjacent potentials of the weak abnormal potential titration is less than or equal to the preset curvature change rate and the time series correlation of the weak abnormal potential is greater than or equal to the preset time series correlation, it is indicated that the potential curve is smooth, the trend is consistent, and the fluctuation pattern is highly similar to the historical normal sequence. The current potential is accurately predicted to be a normal potential. The above content improves the accuracy of the potential titration technology in judging weak abnormal situations, thereby reducing the error of the analysis results.
[0068] Specifically, in step S5, when it is determined to correct the abnormal potential titration using different correction methods, it is determined to correct the abnormal potential titration using different correction methods based on whether the duration of the abnormal potential is greater than a preset duration and a comparison result of the deviation degree of the abnormal value of the abnormal potential titration with the preset deviation degree;
[0069] If the duration of the abnormal potential is less than a preset duration and the deviation degree of the abnormal value is less than a preset deviation degree, it is determined that the abnormal potential titration does not need to be corrected;
[0070] If the duration of the abnormal potential is less than a preset duration and the deviation degree of the abnormal value of the abnormal potential titration is greater than or equal to the preset deviation degree, determining to use the K nearest neighbor algorithm to correct the abnormal potential titration;
[0071] If the duration of the abnormal potential is greater than or equal to a preset duration and the deviation degree of the abnormal value of the abnormal potential titration is less than a preset deviation degree, determining to correct the abnormal potential titration by a window sliding method;
[0072] If the duration of the abnormal potential is greater than or equal to a preset duration and the deviation degree of the abnormal value of the abnormal potential titration is greater than or equal to a preset deviation degree, it is determined that the potentiometric titration is to be re-prepared.
[0073] In the embodiment of the present invention, the preset duration range is set to 3s-7s, and the preferred value is 5s. The reason for selecting this preferred value is that if the preferred value is too small, the system will cause the correction mechanism to be triggered frequently. If the preferred value is too large, the subsequent data may continue to deviate from the ideal curve, affecting the accuracy of the titration end point. The degree of deviation of the abnormal value is determined according to the absolute value of the potential difference between the abnormal value and the ideal titration curve at the corresponding point. The calculation method is the difference between the actual measured potential and the theoretical potential of the ideal titration curve at this point divided by the theoretical potential of the ideal titration curve at this point. For example, the actual measured potential is 325mV, and the theoretical potential of the ideal titration curve at this point is 300mV. The degree of abnormal deviation is (325mV-300mV) / 300mV=0.08, but the above value is not limited to this. Those skilled in the art can also adjust the value according to actual needs.
[0074] The K-nearest neighbor algorithm described in the embodiment of the present invention is based on the interpolation or correction of normal data points around the abnormal point. For example, the theoretical potential of the ideal titration curve at t=5 is 300mV, and the actual measured value is 325mV (the deviation degree is 8%, which is greater than the preset threshold value such as 5%), but the abnormal duration is only 1s (less than the preset duration 5s). K=3 is selected, and the normal points before and after t=5s are taken: t=4s (298mV), t=6s (302mV), t=7s (301mV), and the correction value is calculated as = (298mV+302mV+301mV) / 3=300.33mV, that is, the average value of the neighboring points is used to replace the abnormal value 325mV, reducing the deviation. The window sliding method analyzes the trend of historical data in the sliding window to correct persistent anomalies. For example, the window size N is set to 5, and the potential values of the current abnormal single point t=10s and the previous 4 points are taken: 302mV, 303mV, 304mV, 305mV, and 306mV. The window average value is (302mV+303mV+304mV+305mV+306mV) / 5=304mV. If the ideal value at t=10s is 305mV and the actual measured value is 306mV, the average value of 304mV is used to correct the abnormal value. However, the above value is not limited to this, and those skilled in the art can also adjust the value according to actual needs.
[0075] The present invention determines whether the duration of the abnormal potential is greater than a preset duration and the deviation degree of the abnormal value of the abnormal potential titration is compared with the preset deviation degree to determine whether the abnormal potential titration is corrected by different correction methods. According to the fact that the duration of the abnormal potential is less than the preset duration and the deviation degree of the abnormal value is less than the preset deviation degree, it indicates that a normal noise phenomenon of slight fluctuation and self-recovery occurs, and it is accurately determined that the abnormal potential titration does not need to be corrected. According to the fact that the duration of the abnormal potential is less than the preset duration and the deviation degree of the abnormal value of the abnormal potential titration is greater than or equal to the preset deviation degree, it indicates that a sudden strong interference but not continuous isolated abnormal phenomenon occurs, and the K nearest neighbor algorithm is accurately determined. The abnormal potential titration is corrected. If the duration of the abnormal potential is greater than or equal to the preset duration and the deviation of the abnormal value of the abnormal potential titration is less than the preset deviation, it indicates that a long-standing but small-amplitude systematic drift phenomenon has occurred. The abnormal potential titration is accurately determined to be corrected by a window sliding method. If the duration of the abnormal potential is greater than or equal to the preset duration and the deviation of the abnormal value of the abnormal potential titration is greater than or equal to the preset deviation, it indicates that a serious and persistent fatal abnormal phenomenon has occurred. The potentiometric titration is accurately determined to be re-prepared. The above content improves the accuracy of the potentiometric titration technology in judging weak abnormal situations, thereby reducing the error of the analysis results.
[0076] Thus far, the technical solutions of the present invention have been described in conjunction with the preferred embodiments shown in the accompanying drawings. However, it will be readily understood by those skilled in the art that the scope of protection of the present invention is obviously not limited to these specific embodiments. Without departing from the principles of the present invention, those skilled in the art may make equivalent changes or substitutions to the relevant technical features, and the technical solutions after such changes or substitutions will fall within the scope of protection of the present invention.
Claims
1. A method for predicting endpoint and correcting outliers in potentiometric titration based on deep learning, characterized in that: include: Step S1, obtaining potential change data during potentiometric titration, basic data of the titrant, and parameters of the sample to be tested; Step S2, determining whether the potential is a suspected abnormal potential based on whether the potential change slope exceeds a preset change slope and / or whether the potential change violates the monotonicity of the potential change; Step S3, predicting the abnormal type of the suspected abnormal potential based on a comparison result of the mutation amplitude of the suspected abnormal potential with the first preset mutation amplitude and the second preset mutation amplitude; Step S4, predicting whether the current potential is a weak abnormal potential or a normal potential based on whether the local curvature change rate of the adjacent potential titrated by the weak abnormal potential is greater than a preset curvature change rate and time series correlation; Step S5 , based on the comparison result of the abnormal potential duration with the preset duration and the comparison result of the deviation degree of the abnormal potential titration abnormal value with the preset deviation degree, determining to correct the abnormal potential titration using different correction methods.
2. The method for predicting endpoint and correcting outliers of potentiometric titration based on deep learning according to claim 1, characterized in that: In the step S2, determining whether the potential is a suspected abnormal potential based on whether the potential change slope exceeds a preset change slope and whether the potential change violates the monotonicity of the potential change includes: If the potential change slope exceeds a preset change slope or the potential change violates the monotonicity of potential change, determining that the potential is a suspected abnormal potential; If the potential change slope does not exceed the preset change slope and the potential change conforms to the monotonicity of potential change, it is determined that the potential is a normal potential.
3. The method for predicting endpoint and correcting outliers of potentiometric titration based on deep learning according to claim 2, characterized in that: The potential change slope is determined based on the ratio of the potential difference between two adjacent sampling points to the difference in titrant addition amount. The preset change slope is determined based on the average potential change slope during several potentiometric titrations of the same type. The monotonicity of the potential change is determined based on the potential value change trend after titrant addition.
4. The method for predicting endpoint and correcting outliers of potentiometric titration based on deep learning according to claim 3, characterized in that: In step S3, based on the comparison result of the suspected abnormal potential and the mutation amplitude of the suspected abnormal potential with the first preset mutation amplitude and the second preset mutation amplitude, predicting the abnormal type of the suspected abnormal potential includes: If the mutation amplitude of the suspected abnormal potential is greater than the first preset mutation amplitude, it is predicted that the abnormal type of the suspected abnormal potential is a strong abnormal potential; If the mutation amplitude of the suspected abnormal potential is less than or equal to the first preset mutation amplitude and greater than the second preset mutation amplitude, it is predicted that the abnormal type of the suspected abnormal potential is a weak abnormal potential; If the mutation amplitude of the suspected abnormal potential is less than or equal to the second preset mutation amplitude, it is predicted that the potential is not abnormal.
5. The method for predicting endpoint and correcting outliers of potentiometric titration based on deep learning according to claim 4, characterized in that: The mutation amplitude of the suspected abnormal potential is determined according to the absolute value of the potential difference between the suspected abnormal potential and the adjacent normal point, the first preset mutation amplitude is determined according to the maximum amplitude of normal fluctuations in historical experimental data, and the second preset mutation amplitude is half of the first preset amplitude.
6. The method for predicting endpoint and correcting outliers of potentiometric titration based on deep learning according to claim 5, characterized in that: In the step S4, based on whether the local curvature change rate of the adjacent potential titrated by the weak abnormal potential is greater than the preset curvature change rate and the time series correlation, predicting whether the current potential is a weak abnormal potential or a normal potential includes: If the local curvature change rate of the adjacent potentials titrated by the weak abnormal potential is greater than the preset curvature change rate and the temporal correlation of the weak abnormal potential is less than the preset temporal correlation, it is predicted that the current potential is a weak abnormal potential; If the local curvature change rate of the adjacent potentials titrated by the weak abnormal potential is less than or equal to the preset curvature change rate and the timing correlation of the weak abnormal potential is greater than or equal to the preset timing correlation, it is predicted that the current potential is a normal potential.
7. The method for predicting endpoint and correcting outliers of potentiometric titration based on deep learning according to claim 6, characterized in that: The local curvature change rate of adjacent potentials of the weak abnormal potential titration is determined according to the first-order derivative and the second-order derivative of the potential with respect to time, and the preset curvature change rate is determined by the average local curvature change rate of adjacent potentials of the same type of weak abnormal potential titration.
8. The method for predicting endpoint and correcting outliers of potentiometric titration based on deep learning according to claim 7, characterized in that: The timing correlation is determined according to the dynamic time warping distance, and the preset timing correlation is determined according to the timing correlation distribution of all sequence pairs in historical normal data.
9. The method for predicting endpoint and correcting outliers of potentiometric titration based on deep learning according to claim 8, characterized in that: In the step S5, determining to correct the abnormal potential titration using different correction methods based on whether the duration of the abnormal potential is greater than a preset duration and a comparison result of the deviation degree of the abnormal value of the abnormal potential titration with a preset deviation degree includes: If the duration of the abnormal potential is less than a preset duration and the deviation degree of the abnormal value is less than a preset deviation degree, it is determined that the abnormal potential titration does not need to be corrected; If the duration of the abnormal potential is less than a preset duration and the deviation degree of the abnormal value of the abnormal potential titration is greater than or equal to the preset deviation degree, determining to use the K nearest neighbor algorithm to correct the abnormal potential titration; If the duration of the abnormal potential is greater than or equal to a preset duration and the deviation degree of the abnormal value of the abnormal potential titration is less than a preset deviation degree, determining to correct the abnormal potential titration by a window sliding method; If the duration of the abnormal potential is greater than or equal to a preset duration and the deviation degree of the abnormal value of the abnormal potential titration is greater than or equal to a preset deviation degree, it is determined that the potentiometric titration is to be re-prepared.
10. The method for predicting endpoint and correcting outliers of potentiometric titration based on deep learning according to claim 9, characterized in that: The preset duration is determined based on the historical average duration of abnormal potential under standard conditions, the degree of deviation of the abnormal value is determined based on the absolute value of the potential difference between the abnormal value and the ideal titration curve at the corresponding point, and the preset degree of deviation is determined based on the average degree of deviation of several same type of potentiometric titrations under standard conditions.
Citation Information
Patent Citations
Automatic potentiometric titration method and system based on machine learning
CN114414648A
Intelligent titration analysis system based on artificial intelligence
CN119881202A
Automatic electrochemical titration and analysis system - contains acquisition system and memory to measure and plot potential difference change with time as reaction occurs
FR2504680A1
Analyzing device, and state detecting method
WO2023218738A1