HPLC system failure early warning self-calibration method

Abstract

The present application relates to the technical field of intelligent control of HPLC analysis instrument, in particular to a HPLC system fault early warning self-calibration method, comprising: S1, collecting pump post high-frequency pressure signals and detector chromatogram signals in real time; based on the pump post high-frequency pressure signals, calculating relative physical disturbance quantity; based on the detector chromatogram signals, calculating chemical state deviation degree; S2, fusing the relative physical disturbance quantity and the chemical state deviation degree to construct an instantaneous health index; S3, comparing the instantaneous health index with a preset first early warning threshold and a second early warning threshold, and executing a closed-loop control strategy corresponding to the comparison result. The present method realizes accurate early fault warning by constructing a physical-chemical dual-dimension monitoring system, and the real-time monitoring and fusion evaluation of this dual-dimension overcome the defects of offline and lag of traditional system applicability test, and significantly improve the accuracy and timeliness of early warning.

Description

Technical Field

[0001] This invention relates to the field of intelligent control technology for HPLC analytical instruments, specifically a self-calibration method for fault early warning in HPLC systems. Background Technology

[0002] As a key analytical tool, the stability and reliability of the HPLC system are crucial to ensuring the data quality of long-term series analysis.

[0003] Currently, the performance evaluation of HPLC systems mainly relies on system suitability testing before the start of the analytical task. This method provides an offline, lagging status assessment, which can only confirm the system's availability at the initial moment. However, traditional methods cannot perform real-time, continuous health monitoring during the analysis sequence. Problems caused by slow system degradation or sudden physical disturbances, such as decreased column efficiency or minor pump system anomalies, are often difficult to detect in a timely manner. This can lead to unexpected interruptions in the analytical sequence, decreased data quality, and difficulties in subsequent troubleshooting.

[0004] Therefore, how to achieve real-time monitoring and dynamic evaluation of the health status of HPLC systems and solve the problems of analysis interruption and data quality degradation caused by system performance drift or sudden anomalies has become a technical problem that urgently needs to be solved in this field.

[0005] The information disclosed in the background section above is only intended to enhance the understanding of the background of this disclosure, and therefore may include information that does not constitute prior art known to those skilled in the art. Summary of the Invention

[0006] To address the aforementioned technical problems, this invention discloses a self-calibration method for fault early warning in HPLC systems. Specifically, the technical solution of this invention includes:

[0007] S1. Real-time acquisition of high-frequency pressure signal after pump and chromatographic signal from detector; calculation of relative physical disturbance based on high-frequency pressure signal after pump; calculation of chemical state deviation based on chromatographic signal from detector.

[0008] S2. By integrating the relative physical disturbance and the deviation of the chemical state, an instantaneous health index is constructed;

[0009] S3. Compare the instantaneous health index with the preset first warning threshold and second warning threshold, and execute the closed-loop control strategy corresponding to the comparison result.

[0010] The closed-loop control strategy includes:

[0011] When the instantaneous health index exceeds the second warning threshold, the analysis sequence is stopped and an alarm is triggered;

[0012] When the instantaneous health index is higher than the first warning threshold but not higher than the second warning threshold, perform feedforward dynamic self-calibration of chromatographic parameters;

[0013] When the instantaneous health index is not higher than the first warning threshold, the data is released and the analysis sequence continues.

[0014] Furthermore, the calculation of the relative physical disturbance includes:

[0015] During the system suitability testing phase, a baseline pressure signal is acquired, and the baseline physical disturbance is calculated.

[0016] Real-time pressure signals are collected during the analysis process, and real-time physical disturbances are calculated.

[0017] Calculate the ratio of the real-time physical disturbance to the baseline physical disturbance to generate the relative physical disturbance.

[0018] Furthermore, the calculation of chemical state deviation includes:

[0019] During the system suitability testing phase, standard samples are run to obtain baseline slope and baseline kurtosis.

[0020] During the analysis, the real-time slope and real-time peak value of the current sample chromatographic peak are extracted;

[0021] A weighted linear sum model is used, which combines real-time slope, real-time kurtosis, baseline slope, and baseline kurtosis to calculate the chemical state deviation.

[0022] Furthermore, the method for constructing the instantaneous health index is as follows:

[0023] A pre-defined multiplication model is used to perform nonlinear fusion calculations on relative physical disturbances and chemical state deviations in order to construct an instantaneous health index.

[0024] Furthermore, feedforward dynamic self-calibration of chromatographic parameters includes:

[0025] A calibration factor is generated based on the instantaneous health index and the benchmark index under ideal health conditions;

[0026] The calibration factor is applied to the original retention time of the instrument measurement to obtain the calibrated retention time.

[0027] Furthermore, the method for generating the calibration factor is as follows:

[0028] A first-order linear approximation model is used to calculate and generate a calibration factor based on the deviation between the instantaneous health index and the benchmark index.

[0029] Furthermore, the baseline physical disturbance and the real-time physical disturbance are determined by performing energy spectrum analysis on the corresponding pressure signal within a preset chaotic oscillation characteristic frequency band.

[0030] Furthermore, strategies for suspending sequence analysis and issuing alerts include:

[0031] Immediately stop the injection of subsequent samples, put the instrument into a safe state, and send an alarm message to the operator.

[0032] Compared with the prior art, the present invention has the following beneficial effects:

[0033] 1. This method achieves precise early warning of faults by constructing a physical-chemical dual-dimensional monitoring system. At the physical level, energy dispersive spectroscopy analysis of specific chaotic frequency bands of pressure signals can extremely sensitively capture precursor signals of catastrophic faults such as microbubbles within the pump. At the chemical level, utilizing higher-order moments such as slope and kurtosis, which are more sensitive to chromatographic peak shape, can reveal slow degradation problems such as column aging at an earlier stage. This dual-dimensional real-time monitoring and fusion assessment overcomes the shortcomings of traditional offline and delayed system suitability testing, significantly improving the accuracy and timeliness of early warning.

[0034] 2. This method constructs a complete technical closed loop of real-time monitoring, dynamic evaluation, hierarchical early warning, and closed-loop control, transforming traditional HPLC from a passive analytical tool into an intelligent platform with self-awareness of its status. The system can automatically execute three strategies—data release, dynamic self-calibration, or alarm termination—based on dynamically calculated instantaneous health indices. This intelligent closed-loop control enables real-time, automated intervention in the instrument's analytical process, resolving analytical interruptions and data quality degradation caused by changes in system status, and significantly improving the automation and intelligence of the analysis.

[0035] 3. This method's unique feedforward dynamic self-calibration mechanism maximizes instrument operating efficiency while ensuring data quality. When the system enters a sub-healthy state with predictable performance drift, this method does not simply terminate the sequence. Instead, it dynamically generates calibration factors based on the deviation of the health index, performing real-time correction on key data such as retention time. This proactively offsets system drift errors, ensuring high accuracy and consistency of data in long-sequence analysis, effectively extending the runtime of the analysis sequence, and reducing efficiency losses caused by downtime and recalibration.

[0036] 4. This method significantly enhances the reliability and safety of analytical work through standardized risk assessment and emergency response procedures. By performing nonlinear fusion calculations of physical perturbations and chemical deviations and comparing them with graded thresholds, a quantitative assessment of the overall system risk is achieved. When the system is on the verge of failure, an automatic stop alarm strategy is triggered, immediately stopping sample injection and placing the instrument in a safe state, while providing the operator with clear diagnostic information. This not only avoids wasting samples and time and protects the instrument hardware, but also provides clear guidance for subsequent manual troubleshooting, minimizing the loss from failure. Attached Figure Description

[0037] The present invention will be further explained below with reference to the accompanying drawings and embodiments:

[0038] Figure 1 This is a flowchart of the method of the present invention. Detailed Implementation

[0039] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to specific embodiments.

[0040] Example 1:

[0041] Please see Figure 1 HPLC system fault early warning self-calibration method, including:

[0042] S1. Real-time acquisition of high-frequency pressure signal after pump and chromatographic signal from detector; calculation of relative physical disturbance based on high-frequency pressure signal after pump; calculation of chemical state deviation based on chromatographic signal from detector.

[0043] S2. By integrating the relative physical disturbance and the deviation of the chemical state, an instantaneous health index is constructed;

[0044] S3. Compare the instantaneous health index with the preset first warning threshold and second warning threshold, and execute the closed-loop control strategy corresponding to the comparison result.

[0045] The closed-loop control strategy includes:

[0046] When the instantaneous health index exceeds the second warning threshold, the analysis sequence is stopped and an alarm is triggered;

[0047] When the instantaneous health index is higher than the first warning threshold but not higher than the second warning threshold, perform feedforward dynamic self-calibration of chromatographic parameters;

[0048] When the instantaneous health index is not higher than the first warning threshold, the data is released and the analysis sequence continues;

[0049] This embodiment provides a fault early warning self-calibration method for HPLC systems, aiming to transform traditional high-performance liquid chromatography (HPLC) systems from passive analytical tools into intelligent analytical platforms with self-state perception, early fault warning, and proactive data calibration capabilities. This method constructs a complete technical closed loop, solving problems such as analytical sequence interruption, data quality degradation, and difficulty in subsequent troubleshooting caused by slow system state deterioration or sudden physical disturbances in existing technologies through real-time monitoring, dynamic evaluation, hierarchical early warning, and closed-loop control.

[0050] The specific steps of this method include:

[0051] Step S1: Real-time acquisition of multi-dimensional state signals and feature extraction;

[0052] From the massive amounts of raw data generated during instrument operation, quantitative indicators that can accurately characterize the state of the core subsystems of the system are extracted; in this embodiment, this step is achieved through two parallel channels:

[0053] Physical state channel: A high-frequency pressure sensor installed between the infusion pump outlet and the injection valve of the HPLC system acquires the high-frequency pressure signal after the pump in real time. The selection of this signal is based on technical insight: the smooth operation of the pump is the cornerstone of the physical stability of the entire system, and its slight abnormal fluctuations are key precursor signals indicating an impending catastrophic failure. Based on this pressure signal, a standardized relative physical disturbance is calculated through subsequent processing. This is used to quantify instability at the physical level;

[0054] Chemical state channel: Real-time acquisition of chromatographic signals from the detector at the end of the HPLC system. This signal directly reflects the core of the analytical method: the effectiveness of chemical separation. Changes in the morphology of chromatographic peaks are a comprehensive indicator of column efficiency, accuracy of mobile phase ratio, and the degree of system contamination. Based on this chromatographic signal, a standardized chemical state deviation is calculated through morphological analysis of key chromatographic peaks. This is used to quantify the degree of degradation in chemical separation performance;

[0055] Step S2: Construct an instantaneous health index;

[0056] The purpose of this step is to integrate state indicators from two different dimensions, physical and chemical, into a single, comprehensive system health assessment indicator. In this embodiment, this step uses a specific nonlinear fusion algorithm to fuse relative physical perturbations. Deviation from chemical state Constructing an instantaneous health index The index This is the first time for each sample analysis. The dimensionless parameters calculated dynamically directly reflect the extent to which the HPLC system deviates from its ideal optimal operating state at the current moment.

[0057] Step S3: Implement a tiered closed-loop control strategy based on the health index;

[0058] The purpose of this step is to transform the quantified health assessment results into real-time, automated interventions in the instrument analysis process, thereby achieving closed-loop intelligent control; in this embodiment, this step will convert the instantaneous health index... Compared with the preset first warning threshold and the second warning threshold The comparison is performed, and a closed-loop control strategy corresponding to the comparison result is executed.

[0059] First warning threshold This refers to the critical point at which a system transitions from a healthy state to a sub-healthy state; its value is determined by calculations based on a large amount of historical normal operating data. Statistical analysis is performed on the values, and the upper limit of the 95% or 99% confidence interval of their statistical distribution is taken to ensure that only statistically significant deviations are judged as abnormal.

[0060] Second warning threshold This refers to the critical point at which a system transitions from a sub-healthy state to a state on the verge of failure or already failed; its value is determined by the analysis of the conditions under which it occurs. At this level, key quality control parameters, such as the relative standard deviation of retention time (RSD%), will exceed the limits specified in the pharmacopoeia or quality standards.

[0061] The specific closed-loop control strategies include the following three scenarios:

[0062] Instantaneous health index Not higher than the first warning threshold When the system is deemed to be in a healthy state, the strategy of releasing data and continuing the analysis sequence is executed. This means that the quality of the current and previous analysis data is reliable, and the system can continue to analyze subsequent samples according to the predetermined sequence.

[0063] Instantaneous health index Higher than the first warning threshold And not higher than the second warning threshold When the system is in a sub-healthy, first-level warning state, the system performance has shown observable and systematic drift, but this drift is predictable and compensable. Therefore, a feedforward dynamic self-calibration strategy for chromatographic parameters is implemented. This strategy will correct the relevant chromatographic data of the current and subsequent samples in real time without interrupting the analysis sequence, so as to offset the error caused by the system drift.

[0064] Instantaneous health index Higher than the second warning threshold When the system is in a critical, level-two warning state, it is determined that the system is on the verge of failure. At this point, the system condition has deteriorated significantly, the generated analytical data is no longer reliable, and a hardware failure may be imminent. Therefore, a strategy to terminate the analysis sequence and trigger an alarm is implemented. This strategy aims to immediately stop invalid analyses, protect the sample and instrument, and notify the operator for manual intervention.

[0065] The method disclosed in this embodiment achieves precise, real-time, and automated management of the health status of the HPLC system by constructing a complete technical system of physical-chemical dual-dimensional state monitoring, single health index fusion assessment, and three-level closed-loop control. It overcomes the shortcomings of traditional methods that rely on SST for offline and delayed judgment, enabling real-time detection and response to slow drift or sudden anomalies in system performance during long analytical sequences. Through a self-calibration mechanism under first-level early warning, it maximizes instrument operating efficiency and analytical throughput while ensuring data quality. Through a stop alarm mechanism under second-level early warning, it avoids resource waste and instrument damage. This elevates the reliability, automation level, and intelligence of HPLC analysis to a new level.

[0066] This method focuses on the core issues caused by changes in system physical stability and column performance, achieving efficient early warning. For issues caused by other factors, such as column temperature fluctuations and detector aging, the current framework can be further expanded by integrating sensor signals from more dimensions to build a more comprehensive health assessment model.

[0067] Example 2:

[0068] The calculation of relative physical disturbances includes:

[0069] During the system suitability testing phase, a baseline pressure signal is acquired, and the baseline physical disturbance is calculated.

[0070] Real-time pressure signals are collected during the analysis process, and real-time physical disturbances are calculated.

[0071] Calculate the ratio of the real-time physical disturbance to the baseline physical disturbance to generate the relative physical disturbance.

[0072] The baseline physical disturbance and the real-time physical disturbance are determined by performing energy spectrum analysis on the corresponding pressure signal within the preset chaotic oscillation characteristic frequency band.

[0073] Based on Example 1, this embodiment further defines the specific calculation method for the relative physical disturbance. The core of this method is to achieve highly sensitive capture of predictive fault signals by analyzing the energy in a specific frequency domain.

[0074] Calculate the relative physical disturbance The process requires obtaining the physical disturbance quantities in the baseline state and the real-time state; these two physical disturbance quantities are the baseline physical disturbance quantities. and real-time physical disturbance quantity It is determined by performing energy spectrum analysis on the corresponding pressure signal within a preset chaotic oscillation characteristic frequency band;

[0075] Chaotic oscillation characteristic frequency band It refers to a specific high-frequency range; the technical motivation for its setting is that the present invention has found that early physical faults such as microbubbles in the pump and micro-blockage of the screen plate in front of the column do not immediately cause a significant change in the average pressure, but under the action of nonlinear fluid, they will excite the normal pulsating energy of the pump to a specific high-frequency region, forming chaotic oscillations; this frequency band is a characteristic feature of the sub-healthy state of the system.

[0076] To further clarify, the method for determining its range is as follows: preliminary experimental calibration is performed on a specific model of HPLC pump system. For example, a trace amount of air is artificially introduced into the mobile phase to induce a resonant failure in the system, and a spectrum analyzer is used to record the frequency range in which the pressure signal energy is abnormally concentrated at this time, thereby determining the range. ;

[0077] The specific calculation steps are as follows:

[0078] Reference pressure signals were collected during the System Suitability Testing (SST) phase. At this stage, the system is considered to be in a stable and healthy optimal operating state; subsequently, by performing energy spectrum analysis on the reference pressure signal, the reference physical disturbance is calculated. The calculation formula is as follows:

[0079] ;

[0080] in, The pressure signal was acquired during the SST phase. For the Fourier transform operator, this formula is used to calculate the energy integral of the reference pressure signal within a preset chaotic oscillation characteristic frequency band;

[0081] During the analysis, real-time pressure signals were collected for each sample. And using the same energy spectrum analysis method, the real-time physical disturbance was calculated. :

[0082] ;

[0083] in, The pressure signal is collected in real time during the analysis process;

[0084] Calculate real-time physical disturbances Compared with the baseline physical disturbance The ratio of the two values ​​is used to generate the relative physical disturbance. :

[0085] ;

[0086] To ensure computational stability, the baseline physical perturbation is used here. The value must be measured during the system suitability testing phase and be significantly greater than the effective value of the instrument's background noise; if the value is too small, it indicates that the reference state calibration is incorrect and needs to be repeated.

[0087] By introducing a calculation method based on specific frequency domain energy spectrum analysis, this invention can extremely sensitively capture weak signals that indicate physical faults in the system, which cannot be identified by conventional time domain analysis or broadband analysis methods. By calculating the ratio of real-time values ​​to reference values, a dimensionless relative physical disturbance is generated. This index eliminates the influence of different individual instruments, different pump stroke settings, and background noise fluctuations, achieving standardization and normalization of physical disturbance assessment. This makes the setting of warning thresholds more universal and greatly improves the accuracy and reliability of fault warning.

[0088] Example 3:

[0089] The calculation of chemical state deviation includes:

[0090] During the system suitability testing phase, standard samples are run to obtain baseline slope and baseline kurtosis.

[0091] During the analysis, the real-time slope and real-time peak value of the current sample chromatographic peak are extracted;

[0092] A weighted linear sum model is used, which combines real-time slope, real-time kurtosis, baseline slope, and baseline kurtosis to calculate and generate the chemical state deviation.

[0093] Based on Example 1, this embodiment further defines the specific calculation method for chemical state deviation. This method achieves continuous and quantitative monitoring of the deterioration of chemical separation performance by weighted evaluation of the higher-order statistical moments of chromatographic peaks.

[0094] Calculate the deviation of chemical state The process utilizes the principle that chromatographic peak morphology is highly correlated with chemical state indicators such as chromatographic column efficiency and peak tailing; the specific steps are as follows:

[0095] During the System Suitability Testing (SST) phase, a standard sample is run to obtain a baseline slope. Compared with the baseline kurtosis ;

[0096] slope It is a third-order statistical moment describing the symmetry of chromatographic peaks;

[0097] Kudo It is a fourth-order statistical moment that describes the sharpness of a chromatographic peak;

[0098] and Together they constitute the morphological benchmark of a chemical separation system in its initial optimal state;

[0099] During the analysis, the real-time slope of the key component chromatographic peaks obtained after separation of the current sample is extracted using a standard algorithm. With real-time kurtosis ;

[0100] A weighted linear sum model is used, combined with real-time slope. Real-time peak Reference slope Compared with the baseline kurtosis Calculate the deviation of the chemical state of formation ;

[0101] The weighted linear sum model is a mathematical model used to construct a multi-indicator comprehensive evaluation system. Its purpose is to integrate multiple sub-indicators with different sensitivities and importance into a single comprehensive indicator. In this embodiment, its calculation formula is as follows:

[0102] ;

[0103] in, and For real-time extraction of chromatographic peak slope and peak size; and This is the baseline value obtained during the SST phase; and These are preset weighting coefficients, whose function is to adjust the relative importance of slope and kurtosis in the comprehensive evaluation. They are based on historical experimental data analysis to determine which parameters are more sensitive to predicting specific separation failure modes or show changes earlier, thus assigning higher weight to those parameters. Typically, these two weighting coefficients satisfy... ;

[0104] This method transforms the traditional discrete, qualitative system state assessment performed only in the SST phase into a continuous, quantitative chemical performance tracking throughout the entire analytical sequence. Compared to relying solely on traditional parameters such as retention time or peak area, higher-order moments such as slope and kurtosis are more sensitive to minute deformations of chromatographic peaks, enabling earlier detection of slow-progressing problems such as column aging and mobile phase deterioration. The introduction of weighted linearity and models allows the method to be customized and optimized based on major failure modes, improving the specificity of chemical state assessment and the accuracy of early warning.

[0105] Example 4:

[0106] The method for constructing the instantaneous health index is as follows:

[0107] A pre-defined multiplication model is used to perform nonlinear fusion calculations on relative physical disturbances and chemical state deviations in order to construct an instantaneous health index.

[0108] Based on Example 1, this embodiment further defines the specific method for constructing the instantaneous health index. This method adopts an innovative multiplication model to reflect the synergistic amplification effect between physical disturbance and chemical aging.

[0109] Constructing an instantaneous health index The method is as follows: using a pre-defined multiplication model, the relative physical disturbance quantity is... Deviation from chemical state Perform nonlinear fusion calculations;

[0110] The multiplicative model is a custom nonlinear model constructed by this invention to achieve the fusion of physical and chemical states. Its core technical concept is that the overall failure risk of the system is not a simple linear superposition of physical disturbances and chemical state degradation, but a multiplicative mutual amplification relationship.

[0111] In this embodiment, the mathematical expression of the model is:

[0112] ;

[0113] in, For the first The instantaneous health index calculated from the needle sample is a dimensionless parameter;

[0114] and These are the relative physical disturbance and the deviation from the chemical state, respectively, calculated from the preceding steps.

[0115] and These are coupling coefficients, dimensionless preset weighting coefficients, whose function is to adjust the relative importance of physical disturbances and chemical aging in assessing overall health status; to clarify their calibration process, the observed variables in the calibration dataset are defined. , representing the actual drift of chromatographic parameters recorded under specific experimental conditions; the coupling coefficient is determined by applying physical perturbations of different intensities to chromatographic columns with different aging degrees during the system characterization stage, obtaining a set of coupling coefficients. The calibration dataset is composed of these components; through multivariate fitting analysis, the solution is obtained that allows the calculations based on this model to be performed. and The highest correlation between them and value;

[0116] Compared to a simple additive model, the multiplicative model used in this embodiment can more accurately reveal the intrinsic mechanism of system failure; it can amplify the dual risk signal, that is, when both physical and chemical states deviate simultaneously, the health index... It will grow rapidly, thus enabling earlier and more decisive early warnings; this non-linear fusion method makes the instantaneous health index a highly sensitive indicator of the overall system risk, improving the prediction accuracy of the early warning model.

[0117] Example 5:

[0118] Feedforward dynamic self-calibration of chromatographic parameters includes:

[0119] A calibration factor is generated based on the instantaneous health index and the benchmark index under ideal health conditions;

[0120] The calibration factor is applied to the original retention time of the instrument measurement to obtain the calibrated retention time;

[0121] The method for generating calibration factors is as follows:

[0122] A first-order linear approximation model is used to calculate and generate a calibration factor based on the deviation between the instantaneous health index and the benchmark index.

[0123] Based on Example 1, this embodiment further defines the specific implementation method of feedforward dynamic self-calibration of chromatographic parameters after the system enters the first-level warning state. This method corrects the chromatographic data in real time by dynamically generating calibration factors associated with health indices.

[0124] The core of feedforward dynamic self-calibration of chromatographic parameters lies in the fact that when the system is in the first-level warning range... At that time, its performance drift is systematic and regular, and therefore can be predicted and compensated for through modeling; the process includes two key steps:

[0125] Generate calibration factors ;

[0126] The purpose of this step is to calculate a specific scaling factor for correcting the data based on the current health deviation of the system. In this embodiment, the method for generating the scaling factor is as follows: using a first-order linear approximation model based on the instantaneous health index. Compared with benchmark index The deviation is used to calculate and generate a calibration factor. ;

[0127] The first-order linear approximation model is based on a fundamental assumption: when there is a small deviation from the ideal healthy state, the measurement error caused by system drift is proportional to the health index characterizing that deviation; it should be noted that this linear approximation relationship is considered to exist only when the system is in a first-level warning state, i.e. It is effective within a small deviation range, at which point the system drift is regular and predictable, thus ensuring the effectiveness of the calibration;

[0128] Benchmark Index This refers to an index indicating ideal health, with a theoretical value of 1.

[0129] The specific calculation formula for this model is as follows:

[0130] ;

[0131] in:

[0132] In order to target the Dimensionless calibration factor generated from needle samples;

[0133] This represents the instantaneous health index of the current sample;

[0134] It should be noted that this first-order linear approximation model is only applicable when the system is in a first-level warning state, i.e. This linear relationship applies and is enabled at this time; outside this interval, the linear relationship may no longer hold, therefore when In such cases, the system will execute a stop and alarm strategy instead of applying this calibration model to ensure the validity of the results;

[0135] This is the calibration proportionality coefficient, a preset dimensionless parameter whose function is to define the conversion relationship between the deviation of the health index and the required calibration amount; to clarify its calibration process, the observed variables in the calibration dataset are defined. This represents the relative drift of retention time, precisely measured in a controlled experiment. The method for determining the value is by adjusting the system health index. exist By slowly and controllably changing within the interval, a set of data is obtained. The calibration dataset is composed of these components; by performing linear regression analysis on this dataset, the regression coefficients obtained are the optimal values. value;

[0136] Apply calibration factors to correct the data;

[0137] This step involves applying the calculated calibration factor to the actual measurement data to complete the final calibration; in this embodiment, the calibration factor is... The original retention times of all components in the current sample used for instrument measurements To obtain the retention time after calibration This ensures that the consistency of the entire chromatogram data is corrected; the calculation formula is:

[0138] ;

[0139] in, It is the raw retention time directly output by the instrument. It is the calibration factor calculated in the previous steps; both sides of the formula have the dimension of time, which satisfies the dimensionality consistency.

[0140] This self-calibration method forms a precise early warning-compensation closed loop. The first-order linear approximation model, while ensuring calibration accuracy, has the advantages of low computational cost and fast response speed, meeting the requirements of real-time processing. By dynamically generating calibration factors directly related to the instantaneous health index, this method achieves personalized and precise calibration for each data point. This not only significantly improves the accuracy and consistency of data in long-sequence analysis, but also extends the effective running time of the analysis sequence through active correction, reduces downtime caused by recalibration, and improves the efficiency of the analysis work.

[0141] Example 6:

[0142] Strategies for suspending sequence analysis and issuing alerts include:

[0143] Immediately stop the injection of subsequent samples, put the instrument into a safe state, and send an alarm message to the operator;

[0144] Based on Example 1, this embodiment further defines the specific execution details of the strategy to terminate the analysis sequence and issue an alarm after the system is determined to enter the level 2 warning state, so as to ensure that the analysis process can be terminated safely and effectively and clear instructions can be provided to the user when the system is on the verge of failure.

[0145] Instantaneous health index Exceeding the second warning threshold When the system determines that a serious fault has occurred or is about to occur, the strategy of suspending the analysis sequence and issuing an alarm is triggered. The specific operations include:

[0146] Immediately stop subsequent sample injections: The control software immediately sends a stop command to the autosampler, clearing all sample tasks to be analyzed in the injection queue; this is intended to prevent the injection of samples into an unreliable system as soon as possible, thereby avoiding waste of resources;

[0147] Put the instrument into a safe state: The system automatically executes preset safety procedures, which may include stopping the gradient elution program of the high-pressure pump, switching to a low-flow-rate, weak solvent flushing program to protect the column and tubing; turning off consumable components such as detector lamps to extend their lifespan; the core purpose of this is to protect the instrument's core hardware and prevent permanent damage caused by uncontrolled operation.

[0148] The system sends alarm messages to the operator: A clear, high-priority alarm window pops up through the human-machine interface; the alarm message not only informs the operator that the analysis sequence has been aborted, but also provides diagnostic information, such as a level-two warning: a serious physical disturbance has been detected. A suspected pump seal leak has been detected; please check immediately. This information can also be pushed to designated mobile devices via the network to ensure that operators are informed in a timely manner. This provides operators with clear fault guidance and shortens the diagnosis and troubleshooting time required for manual intervention.

[0149] The specific strategy provided in this embodiment transforms the concept of alarm termination into a set of specific, executable, and standardized operating procedures that take into account sample safety, instrument safety, and rapid response. Through automated emergency handling, it minimizes the losses caused by system failure and provides clear, data-supported decision-making basis for subsequent manual maintenance, demonstrating a high degree of intelligence and robustness.

[0150] The above are merely preferred embodiments of the present invention and are not intended to limit the scope of protection of the present invention; any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A self-calibration method for fault early warning of an HPLC system, characterized in that, The specific steps include: S1. Real-time acquisition of high-frequency pressure signal after pump and chromatographic signal from detector; calculation of relative physical disturbance based on high-frequency pressure signal after pump; calculation of chemical state deviation based on chromatographic signal from detector. S2. By integrating the relative physical disturbance and the deviation of the chemical state, an instantaneous health index is constructed; S3. Compare the instantaneous health index with the preset first warning threshold and second warning threshold, and execute the closed-loop control strategy corresponding to the comparison result. The closed-loop control strategy includes: When the instantaneous health index exceeds the second warning threshold, the analysis sequence is stopped and an alarm is triggered; When the instantaneous health index is higher than the first warning threshold but not higher than the second warning threshold, perform feedforward dynamic self-calibration of chromatographic parameters; When the instantaneous health index is not higher than the first warning threshold, the data is released and the analysis sequence continues; The calculation of relative physical disturbances includes: During the system suitability testing phase, a baseline pressure signal is acquired, and the baseline physical disturbance is calculated. Calculate the baseline physical disturbance The calculation formula is as follows: in, The pressure signal was acquired during the SST phase. This is a Fourier transform operator; the formula calculates the energy integral of the reference pressure signal within a preset chaotic oscillation characteristic frequency band. It refers to a specific high-frequency range; Real-time pressure signals are collected during the analysis process, and real-time physical disturbances are calculated. Calculate the ratio of the real-time physical disturbance to the baseline physical disturbance to generate the relative physical disturbance. The calculation of chemical state deviation includes: During the system suitability testing phase, standard samples are run to obtain baseline slope and baseline kurtosis. During the analysis, the real-time slope and real-time peak value of the current sample chromatographic peak are extracted; A weighted linear sum model is used, combined with real-time slope. Real-time peak Reference slope Compared with the baseline kurtosis Calculate the deviation of the chemical state of formation ; The calculation formula is as follows: in, and For real-time extraction of chromatographic peak slope and peak size; and This is the baseline value obtained during the SST phase; and These are preset weighting coefficients; The method for constructing the instantaneous health index is as follows: A pre-defined multiplication model is used to perform nonlinear fusion calculations on relative physical disturbances and chemical state deviations in order to construct an instantaneous health index. in, For the first The instantaneous health index calculated from the needle sample is a dimensionless parameter; and These are the relative physical disturbance and the deviation from the chemical state, respectively, calculated from the preceding steps. and It is the coupling coefficient, which is a dimensionless preset weighting coefficient.

2. The HPLC system fault early warning self-calibration method according to claim 1, characterized in that, Feedforward dynamic self-calibration of chromatographic parameters includes: A calibration factor is generated based on the instantaneous health index and the benchmark index under ideal health conditions; The calibration factor is applied to the original retention time of the instrument measurement to obtain the calibrated retention time.

3. The HPLC system fault early warning self-calibration method according to claim 2, characterized in that, The method for generating calibration factors is as follows: A first-order linear approximation model is used to calculate and generate a calibration factor based on the deviation between the instantaneous health index and the benchmark index.

4. The HPLC system fault early warning self-calibration method according to claim 1, characterized in that, The baseline physical disturbance and the real-time physical disturbance are determined by performing energy spectrum analysis on the corresponding pressure signal within a preset chaotic oscillation characteristic frequency band.

5. The HPLC system fault early warning self-calibration method according to claim 1, characterized in that, Strategies for suspending sequence analysis and issuing an alert include: Immediately stop the injection of subsequent samples, put the instrument into a safe state, and send an alarm message to the operator.

Images