HPLC mobile phase gradient intelligent proportioning online purification system

Abstract

The present application relates to the technical field of HPLC system state monitoring and intelligent control, in particular to an HPLC mobile phase gradient intelligent proportioning online purification system, which comprises a data acquisition module, a first processing module, a second processing module, a third processing module, a decision module and a system control module.The data acquisition module acquires the baseline signal of a high performance liquid chromatography detector in real time.The first processing module determines noise fluctuation parameters.The second processing module calculates pollution entropy based on the noise fluctuation parameters and a preset system pollution sensitivity coefficient.The third processing module calculates the actual pollution entropy accumulation rate and determines an anomaly index in combination with a preset pollution entropy accumulation rate baseline model.The decision module generates a running mode decision based on the anomaly index.The system control module executes a corresponding running mode in response to the running mode decision and determines whether to switch back to the optimal analysis efficiency mode after executing the diagnostic survival mode.The present application improves the long-term running stability, reliability and analysis result accuracy of the HPLC system and avoids experimental interruption caused by accidental pollution.

Description

Technical Field

[0001] This invention relates to the field of HPLC system status monitoring and intelligent control technology, specifically to an online purification system for intelligent gradient proportioning of HPLC mobile phase. Background Technology

[0002] In the field of analytical chemistry, HPLC systems are key precision analytical instruments, and their long-term stability and the accuracy of analytical results are of paramount importance. Traditional HPLC systems usually operate in a passive response mode, lacking the ability to proactively diagnose and maintain the system's health status, which is mainly caused by chemical contamination. This mode relies on fixed maintenance cycles or reactive measures after significant deviations in analytical results, which can easily lead to decreased system reliability and experimental interruptions.

[0003] In existing technologies, system status monitoring mainly relies on the experience and judgment of operators, lacking effective means to conduct real-time and quantitative assessment of the system's micro-health status; it is impossible to establish an accurate mathematical model between subtle baseline signal fluctuations and cumulative system damage, nor can it form an autonomous control closed loop from signal perception and risk assessment to decision execution; when faced with pollution accumulation, traditional systems cannot make a dynamic and intelligent trade-off between maximizing analysis throughput and minimizing equipment damage risk, resulting in shortened system lifespan and increased operation and maintenance costs;

[0004] Therefore, how to transform a traditional passive HPLC system into an intelligent analytical platform with self-sensing, forward-looking diagnostic, and autonomous decision-making capabilities to proactively manage and maintain system health is a problem that urgently needs to be solved by those skilled in the art.

[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 an HPLC mobile phase gradient intelligent proportioning online purification system. Specifically, the technical solution of this invention includes:

[0007] The data acquisition module is used to acquire the baseline signal of the high-performance liquid chromatography detector in real time;

[0008] The first processing module is used to determine noise fluctuation parameters based on the baseline signal;

[0009] The second processing module is used to calculate the pollution entropy based on the noise fluctuation parameters and the preset system pollution sensitivity coefficient;

[0010] The third processing module is used to calculate the actual pollution entropy accumulation rate based on pollution entropy, and determine the anomaly index by combining it with the preset pollution entropy accumulation rate baseline model.

[0011] The decision-making module is used to generate operating mode decisions based on anomaly indices.

[0012] The system control module is used to execute the corresponding operating mode in response to the operating mode decision, and after the diagnostic survival mode is executed, to determine whether to switch back to the optimal analysis efficiency mode.

[0013] Preferably, the first processing module is specifically used for:

[0014] Perform drift removal processing on the baseline signal within a time window;

[0015] Calculate the standard deviation of the baseline signal after drift removal to generate noise fluctuation parameters.

[0016] Preferably, the second processing module is specifically used for:

[0017] The noise fluctuation parameter is used as the instantaneous rate of pollution accumulation;

[0018] The contamination entropy is generated by integrating the instantaneous rate over time.

[0019] Preferably, the third processing module is specifically used for:

[0020] The actual pollution entropy accumulation rate is calculated by numerically differentiating the pollution entropy data sequence.

[0021] The predicted pollution entropy accumulation rate under normal operating conditions is obtained by using a baseline model of pollution entropy accumulation rate.

[0022] The standardized deviation of the actual pollution entropy accumulation rate relative to the predicted pollution entropy accumulation rate is calculated to generate an anomaly index.

[0023] Preferably, the decision module is specifically used for:

[0024] Based on the anomaly index and preset risk preference parameters, the first utility value representing the utility of the optimal analysis efficiency model is calculated;

[0025] Based on the anomaly index and preset diagnostic cost parameters, a second utility value characterizing the utility of diagnostic survival mode is calculated.

[0026] If the second utility value is greater than the first utility value, a decision to switch to diagnostic survival mode is generated.

[0027] If the second utility value is not greater than the first utility value, then an operating mode decision is generated to maintain the optimal analysis efficiency mode.

[0028] Preferably, the system control module is specifically used for:

[0029] After implementing the diagnostic survival mode, the post-diagnostic contamination entropy was measured.

[0030] The system resilience index is calculated based on the post-diagnosis pollution entropy and the preset reference pollution entropy.

[0031] If the system resilience index exceeds the preset recovery threshold, an instruction to switch back to the optimal analysis efficiency mode will be generated.

[0032] If the system resilience index does not exceed the recovery threshold, instructions are generated to maintain diagnostic survival mode or to perform further diagnostics.

[0033] Preferably, the first utility value includes a benefit term representing the value of the analytical throughput and a risk cost term that grows non-linearly with the anomaly index.

[0034] Preferably, the second utility value includes an information gain term proportional to the anomaly index and a direct diagnostic cost term.

[0035] Compared with the prior art, the present invention has the following beneficial effects:

[0036] 1. This invention enhances system stability and proactive maintenance capabilities: This system constructs an autonomous control closed loop from signal perception and risk quantification to decision execution. By introducing the concept of contamination entropy, instantaneous noise is integrated into traceable cumulative health indicators. This enables the system to shift from passive response to proactive management, proactively diagnosing and maintaining system degradation caused by chemical contamination. This significantly improves the long-term operational stability, reliability, and accuracy of analytical results of the HPLC system, avoiding experimental interruptions due to accidental contamination.

[0037] 2. This invention improves the accuracy and specificity of anomaly detection: By establishing a baseline model of the accumulation rate of contamination entropy, this system provides a dynamic reference standard that varies with operating conditions for normal aging. The anomaly index calculated based on this model effectively eliminates the legitimate influence of changes in normal operating parameters such as flow rate and gradient on the contamination rate. This enables the system to highly specifically identify abnormal events caused by unknown chemical contamination, greatly reducing the false alarm rate and ensuring the reliability of subsequent decisions and the necessity of intervention measures.

[0038] 3. This invention achieves economically rational intelligent decision-making: The system innovatively introduces a decision-making model based on utility functions, replacing the traditional fixed threshold alarm mechanism. This model can dynamically balance maximizing analytical output with minimizing pollution risk, and comprehensively consider the information value and execution cost of diagnostic operations. This economically rational decision-making logic enables the system to intervene decisively when risks are still in their infancy but growing rapidly, achieving a highly intelligent and adaptive balance between ensuring analytical efficiency and system survival.

[0039] 4. This invention ensures quantitative evaluation and closed-loop control of intervention effects: After executing the diagnostic survival mode, this system introduces an innovative indicator—the system resilience index—to objectively and quantitatively evaluate the effectiveness of maintenance operations. Only when the system's health is confirmed to have recovered to above a preset threshold will it switch back to analysis mode. This design ensures a complete closed loop in the entire perception-analysis-decision-execution-evaluation process, preventing premature resumption of work before contamination is completely eliminated, fundamentally guaranteeing the quality of subsequent analysis data and the robust operation of the system. Attached Figure Description

[0040] The present invention will be further explained below with reference to the accompanying drawings and embodiments:

[0041] Figure 1 This is a system structure block diagram of the present invention. Detailed Implementation

[0042] 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.

[0043] Example 1:

[0044] Please see Figure 1 The HPLC mobile phase gradient intelligent proportioning online purification system includes:

[0045] The data acquisition module is used to acquire the baseline signal of the high-performance liquid chromatography detector in real time;

[0046] The first processing module is used to determine noise fluctuation parameters based on the baseline signal;

[0047] The second processing module is used to calculate the pollution entropy based on the noise fluctuation parameters and the preset system pollution sensitivity coefficient;

[0048] The third processing module is used to calculate the actual pollution entropy accumulation rate based on pollution entropy, and determine the anomaly index by combining it with the preset pollution entropy accumulation rate baseline model.

[0049] The decision-making module is used to generate operating mode decisions based on anomaly indices.

[0050] The system control module is used to execute the corresponding operating mode in response to the operating mode decision, and after the diagnostic survival mode is executed, to determine whether to switch back to the optimal analysis efficiency mode.

[0051] This embodiment provides an HPLC mobile phase gradient intelligent ratio online purification system, which aims to transform the traditional, passively responding HPLC system into an intelligent analysis platform with self-sensing, forward-looking diagnosis, and autonomous decision-making capabilities. By constructing a complete technical closed loop from microscopic signal perception to macroscopic decision execution, the system can proactively manage and maintain the system's health status degradation, which is mainly caused by chemical contamination, while ensuring the throughput of analytical tasks, thereby improving the long-term stability, reliability, and accuracy of analytical results.

[0052] In a specific implementation scenario, the system includes six collaborative core modules: data acquisition module, first processing module, second processing module, third processing module, decision-making module, and system control module;

[0053] The data acquisition module provides the most original, high-fidelity data input for the intelligent analysis of the entire system. In this embodiment, the data acquisition module is specifically implemented through a high-frequency A / D converter directly connected to the signal output terminal of the HPLC detector. This module continuously acquires the baseline signal of the HPLC detector in real time at a sampling frequency of, for example, 100Hz. The unit is volts (V); this high-frequency sampling ensures that extremely subtle and transient baseline fluctuations caused by the adsorption-desorption process of trace pollutants can be captured, laying the foundation for subsequent microtexture analysis.

[0054] The first processing module extracts characteristic parameters from the macroscopic baseline signal that can accurately characterize the microscopic contamination state of the system; in this embodiment, the first processing module is based on the real-time acquired baseline signal. This is processed to determine a core noise fluctuation parameter. ;

[0055] The second processing module integrates instantaneous, discrete noise fluctuation information into a macroscopic indicator that can quantify the cumulative and irreversible damage to the system. In this embodiment, the second processing module is based on the noise fluctuation parameters calculated by the preceding module. With a preset system pollution sensitivity coefficient Pollution entropy was calculated using an innovative modeling method. ;

[0056] The third processing module distinguishes between normal and abnormal increases in pollution entropy and generates a standardized risk signal. In this embodiment, the third processing module is based on the real-time updated pollution entropy. The data sequence was used to calculate the actual pollution entropy accumulation rate. And combined with a pre-defined baseline model of pollution entropy accumulation rate. Ultimately, an anomaly index was determined to quantify the risk level. ;

[0057] The decision-making module mimics the trade-off logic of human experts, making the optimal choice between the potentially conflicting goals of pursuing analytical efficiency and ensuring system survival. In this embodiment, the decision-making module is based on anomaly indices. By comparing the utility function values ​​of two operating modes, namely the optimal analytical efficiency mode and the diagnostic survival mode, a clear operating mode decision is generated.

[0058] The system control module translates the abstract instructions generated by the decision module into specific operations on the HPLC system hardware, and is responsible for evaluating the effectiveness of intervention measures after their execution, thus achieving a closed loop of control logic. In this embodiment, the system control module responds to the operating mode decision by executing the corresponding operating mode. If the decision is to switch to diagnostic survival mode, the module will automatically control the mobile phase mixing unit to execute a preset online cleaning procedure. After the diagnostic survival mode is executed, the module will use a quantitative evaluation mechanism to determine whether the system has recovered its health and decide whether to switch back to the optimal analytical efficiency mode.

[0059] This embodiment constructs a complete autonomous control closed loop of perception-analysis-decision-execution-evaluation through the collaborative work of the above modules. By introducing innovative concepts such as contamination entropy and anomaly index, it achieves a deep quantitative understanding of the system's health status and a forward-looking prediction of future risks. Ultimately, this invention can autonomously and intelligently achieve a dynamic balance between maximizing analytical output and minimizing contamination risk, improving the automation level, operational reliability, and service life of the HPLC system, and reducing experimental interruptions and economic losses caused by accidental contamination.

[0060] Example 2:

[0061] The first processing module is specifically used for:

[0062] Perform drift removal processing on the baseline signal within a time window;

[0063] Calculate the standard deviation of the baseline signal after drift removal to generate noise fluctuation parameters;

[0064] Based on Example 1, this embodiment limits the specific implementation of the first processing module in order to ensure the extraction accuracy and physical meaning of noise fluctuation parameters.

[0065] The specific workflow of the first processing module is divided into two core steps;

[0066] Drift removal processing is performed on the baseline signal within a time window; time window This refers to a very short data segment, such as 1 second, used to calculate noise fluctuation parameters. Its function is to ensure the instantaneity of the calculation, reflect the system state at the current moment, and is a preset parameter of the system. The inherent logic of the drift removal process is to eliminate the slowly changing baseline background caused by factors such as temperature changes and solvent gradients, thereby highlighting the high-frequency noise caused by trace contaminants that truly reflects the microscopic state. In this embodiment, this process is carried out through a time window. The process involves applying a moving average filter or a polynomial fitting algorithm to the baseline signal data points to obtain the baseline drift trend line, and then subtracting the trend line from the original signal.

[0067] The standard deviation of the baseline signal after drift removal is calculated to generate noise fluctuation parameters; after drift removal is completed, the first processing module applies this time window... The statistical standard deviation of the residual signal within the range is calculated; this standard deviation is defined as the noise fluctuation parameter. Noise fluctuation parameters The choice of standard deviation is a key technical consideration in this invention. Compared with simple peak-to-peak noise, standard deviation can more robustly and comprehensively reflect the fluctuation range of the signal near the mean. Its physical meaning is directly related to the microscopic energy fluctuations caused by the adsorption-desorption process of chemical substances in the system. Therefore, it can serve as a sensitive characterization of the instantaneous rate of pollution accumulation in the system.

[0068] By employing drift-reduction processing, this embodiment effectively avoids the interference of macroscopic environmental factors on microscopic noise measurements, ensuring the signal-to-noise ratio for subsequent analysis; furthermore, the standard deviation is used as the noise fluctuation parameter. It provides a more robust quantitative indicator in terms of statistics and a more explicit quantitative indicator in terms of physics, enabling it to capture the micro-contamination dynamics within the system with high sensitivity and accuracy, and providing high-quality and reliable data input for the subsequent accurate modeling of contamination entropy.

[0069] Example 3:

[0070] The second processing module is specifically used for:

[0071] The noise fluctuation parameter is used as the instantaneous rate of pollution accumulation;

[0072] The contamination entropy is generated by integrating the instantaneous rate over time.

[0073] Based on Example 1, this embodiment specifies the method for calculating pollution entropy in the second processing module. Its core lies in establishing an innovative model that can quantify the cumulative and irreversible damage of the system.

[0074] The specific workflow of the second processing module includes two core steps;

[0075] Using noise fluctuation parameters as the instantaneous rate of pollution accumulation; this step is the theoretical cornerstone of the pollution entropy conceptual model of this invention; the second processing module innovatively uses noise fluctuation parameters , The physical meaning of the time variable is elevated from a simple statistic describing signal fluctuations to a characterizing the system's physical properties within its time frame. A key proxy indicator of the deterioration of the health status at any given time; in the model, this deterioration is mainly quantified as the equivalent irreversible pollution accumulation caused by the frequent adsorption-desorption processes of trace pollutants on the surface of the stationary phase; the model is particularly suitable for detecting such dynamic pollution processes;

[0076] This design draws on the damage accumulation theory in reliability engineering, which states that the ultimate failure of a device is the result of the accumulation of minute damage over time. This invention extends this idea to the field of chemical contamination, considering that contamination in HPLC systems is also a continuous cumulative process. It is precisely this process that embodies speed at every instant;

[0077] Pollution entropy is generated by integrating the instantaneous rate over time; based on the above definition, the second processing module analyzes the noise fluctuation parameters... Perform the initial cleaning process from the initial cleaning moment of the system. up to the current moment The total cumulative pollution amount, i.e., pollution entropy, is calculated by integrating the time components. ;

[0078] To achieve more accurate modeling, this embodiment introduces pollution entropy. The calculation method is as follows:

[0079] ;

[0080] in, Let be the cumulative pollution entropy at time t, with the dimension V·s, which can be understood as a custom unit of pollution quantity, calculated by this module; This is the start time of the system's initial clean state. Its data type is time unit such as seconds. It is the time point recorded when the system runs for the first time or after a thorough maintenance is completed. In time The noise fluctuation parameter, which is used as the instantaneous rate input of pollution accumulation and has the dimension V, is calculated by the first processing module. The system contamination sensitivity coefficient is a dimensionless, adjustable parameter used to characterize the differences in sensitivity to chemical contamination response of a specific HPLC system configuration and the current analytical method category. Its function is to calibrate the model and ensure its universality across different hardware configurations. This parameter can be determined through specialized calibration experiments or selected from a pre-defined lookup table based on the chemical properties of the analyte.

[0081] This parameter is determined through specialized calibration experiments; for example, in a clean system, a standard contaminant of known mass that causes baseline noise is injected through an injection valve. And continuously monitor the increase in pollution entropy caused by this. Until the system reaches a new equilibrium; the calibration experimental dataset is a set of data with different injection qualities. and the corresponding increase in pollution entropy, The data points constituted; system pollution sensitivity coefficient The model is then fitted by linear regression analysis of the dataset to ensure the accuracy of its physical meaning and the universality of the model.

[0082] This embodiment uses instantaneous rate, By performing the core step of time integration, a novel technical concept of pollution entropy was successfully constructed; it transforms high-frequency, volatile microscopic system disturbances into a monotonically or trend-growing macroscopic system state variable. This allows the assessment of system health to go beyond instantaneous alarms, enabling long-term trend analysis, prediction of system lifespan, and assessment of cumulative damage, providing a stable and information-rich core basis for subsequent more complex anomaly detection and decision-making.

[0083] Example 4:

[0084] The third processing module is specifically used for:

[0085] The actual pollution entropy accumulation rate is calculated by numerically differentiating the pollution entropy data sequence.

[0086] The predicted pollution entropy accumulation rate under normal operating conditions is obtained by using a baseline model of pollution entropy accumulation rate.

[0087] Calculate the standardized deviation of the actual pollution entropy accumulation rate relative to the predicted pollution entropy accumulation rate to generate an anomaly index;

[0088] Based on Example 1, this embodiment defines the specific logic for the third processing module to determine the anomaly index, aiming to achieve accurate, sensitive and statistically significant identification of abnormal pollution events.

[0089] The specific workflow of the third processing module is divided into logically progressive steps:

[0090] The actual pollution entropy accumulation rate is calculated by numerically differentiating the pollution entropy data sequence; the third processing module maintains a time series of pollution entropy. Data; by applying a numerical differentiation algorithm to the data sequence within a sliding time window, the current actual pollution entropy accumulation rate is calculated. ; As pollution entropy, The derivative with respect to time has the physical meaning of the actual rate of increase of the system's pollution entropy at the current moment, and its dimension is volts (V).

[0091] The predicted pollution entropy accumulation rate under normal operating conditions is obtained through a pollution entropy accumulation rate baseline model; pollution entropy accumulation rate baseline model. It is an empirical formula trained using machine learning methods such as multiple linear regression, based on a large amount of historical health operation data, to predict the normal aging rate of a system under current operating parameters. This linear model is an effective approximation of complex physical processes within the normal operating range; its function is to provide a dynamic, condition-dependent standard for judging normal operation. Whether an anomaly is detected is determined during the initial system deployment phase by collecting health operation data under different analysis methods and conducting offline training. In this embodiment, the model can be represented as:

[0092] ;

[0093] in, The predicted rate of accumulation of normal pollution entropy, in volts (V), is calculated by this model. Dimensionless parameters characterizing the complexity of the mobile phase gradient, such as the sum of absolute values ​​of solvent proportion changes per unit time, are derived from current HPLC operating methods. The average flow rate of the mobile phase, in units of mL / min, is obtained from the current HPLC operating method; , The coefficients are model coefficients, representing the contribution of gradient complexity, flow velocity to the normal pollution accumulation rate, and the system's inherent fundamental accumulation rate, respectively, determined through regression analysis of historical data. To ensure dimensional consistency, the dimensions of each coefficient are as follows: Its dimension is volt (V). The dimensions are volts per minute per milliliter (V·min / mL), while Its dimension is volt (V);

[0094] The standardized deviation of the actual pollution entropy accumulation rate relative to the predicted pollution entropy accumulation rate is calculated to generate an anomaly index. To eliminate the normal influence of different operating conditions on the pollution accumulation rate and thus purely reflect the risk caused by unknown chemical attacks, this embodiment introduces an anomaly index. Calculation:

[0095] ;

[0096] in, It is a dimensionless property designed to prevent the denominator from being zero. The same tiny normal number, for example, 1% of the system's historical minimum normal accumulation rate, is used to enhance the computational stability of the algorithm under extremely low contamination rate conditions.

[0097] It is a dimensionless real-time anomaly index, which serves as the core risk signal and is calculated by this module; The actual rate of accumulation of pollution entropy, i.e., pollution entropy The derivative with respect to time, in volts (V), is calculated in the first step of this module; The predicted rate of accumulation of normal pollution entropy;

[0098] This embodiment is designed by establishing a baseline model. This enables the system to distinguish between normal aging and abnormal events; it generates anomaly indices by calculating standardized deviations. This index eliminates the legitimate influence of different flow rates, gradients, and other operating parameters on the contamination rate, making... It has become a highly specific risk indicator that is comparable under any operating condition; this improves the accuracy of anomaly detection, effectively avoids false alarms caused by changes in normal operating parameters, and ensures the reliability of subsequent decisions.

[0099] Example 5:

[0100] The decision-making module is specifically used for:

[0101] Based on the anomaly index and preset risk preference parameters, the first utility value representing the utility of the optimal analysis efficiency model is calculated;

[0102] Based on the anomaly index and preset diagnostic cost parameters, a second utility value characterizing the utility of diagnostic survival mode is calculated.

[0103] If the second utility value is greater than the first utility value, a decision to switch to diagnostic survival mode is generated.

[0104] If the second utility value is not greater than the first utility value, then an operating mode decision is generated to maintain the optimal analysis efficiency mode;

[0105] The first utility value includes a benefit term representing the value of the analytical throughput and a risk cost term that grows non-linearly with the anomaly index.

[0106] The second utility value includes an information gain term proportional to the anomaly index and a direct diagnostic cost term.

[0107] Based on Example 1, this embodiment provides a detailed and in-depth definition of how the decision-making module generates operational mode decisions. Its core is to construct an introspective antifragile decision-making model based on decision theory. This model not only considers current risks but also weighs the future utility of different decisions.

[0108] The core task of the decision-making module is based on anomaly indices. It makes a choice between two operating modes: the optimal analysis efficiency mode and the diagnostic survival mode; this decision is not based on a simple threshold trigger, but is achieved by calculating and comparing the utility functions of the two modes.

[0109] The decision module calculates the first utility value, which characterizes the utility of the optimal analytical efficiency mode, based on the anomaly index and preset risk preference parameters. The optimal analytical efficiency mode refers to the operating state of the HPLC system when it is performing analytical tasks normally, and its goal is to maximize the sample throughput.

[0110] Its utility function Defined as:

[0111] ;

[0112] in, This is the total utility value of the optimal analysis efficiency model. Its data type is a custom utility unit, which is calculated by this module. This represents the revenue item used to analyze the value of flux; For throughput analysis, the unit is samples / hour, obtained from the system operation plan; The value weight per unit of throughput is one of the risk preference parameters. Its unit is utility unit / sample / hour, which is preset by the user based on the economic value of the analysis task. This is a risk cost term that grows non-linearly with the abnormal index;

[0113] The risk of this system is not linear; small anomalies with a small A value may have limited risk, but as the degree of anomaly increases, the probability of catastrophic failure will increase exponentially; this formula can realistically simulate this nonlinear risk. The value weight of risk cost is one of the parameters of risk preference. Its data type is utility unit, which is preset by the user according to their tolerance for risk. The risk aversion coefficient is one of the parameters of risk preference. A larger value indicates that the decision-making system is more sensitive to risk. Its data type is dimensionless and is preset by the user. This is an anomaly index, dimensionless, and is input from the third processing module.

[0114] The decision module calculates a second utility value, which characterizes the utility of the diagnostic survival mode, based on the anomaly index and preset diagnostic cost parameters. The diagnostic survival mode refers to the state in which the system suspends routine analysis tasks and instead performs maintenance operations such as online cleaning and system flushing, which are aimed at identifying and eliminating sources of contamination.

[0115] Its utility function Defined as:

[0116] ;

[0117] in, This represents the total utility value for the diagnostic survival model. Its data type is a custom utility unit, calculated by this module. The information gain term is proportional to the anomaly index; the value of performing diagnostics lies in eliminating uncertainty; the anomaly index. The higher the value, the greater the uncertainty of the system state, and the more valuable it is to obtain information about the pollution source through diagnostic operations.

[0118] in The value weight of information gain is one of the risk preference parameters. Its data type is utility unit, which is preset by the user. This is a direct cost item for diagnosis; The direct cost of performing diagnostic procedures can be in time or in equivalent monetary units, and is determined based on the pre-defined diagnostic procedure. The weights for diagnostic costs are one of the risk preference parameters, and their data type is utility unit / cost unit, which is preset by the user; the values ​​of each weight parameter are set to ensure that the terms on both sides of the formula have the same utility dimension;

[0119] The decision-making module makes decisions based on the principle of maximizing expected utility.

[0120] If the second utility value is greater than the first utility value This means that the risk cost of continuing to maintain high-throughput analysis has exceeded its benefits, while the information value obtained from performing diagnostics has exceeded its direct cost; at this point, the decision module will generate an operating mode decision to switch to diagnostic survival mode.

[0121] If the second utility value is not greater than the first utility value If the current risk is within a controllable range, then maintaining the benefits of the analysis task is still a better choice; at this time, the decision module will generate an operating mode decision to maintain the optimal analysis efficiency mode.

[0122] This embodiment constructs and compares... and These two ingeniously structured utility functions enable intelligent decision-making that far surpasses traditional fixed-threshold alarms; they are economically rational, meaning that decision-making is no longer a simple yes / no judgment, but a dynamic, economically meaningful trade-off between benefits, risks, information value, and costs; they are adaptive, because by introducing risk preference parameters, the decision-making logic can be customized according to the user's strategic goals; and they are forward-looking, because the exponential risk cost term enables the system to intervene decisively when risks are still in their infancy but growing rapidly, reflecting the antifragile design philosophy, that is, proactively responding to fluctuations to avoid greater losses.

[0123] Example 6:

[0124] The system control module is specifically used for:

[0125] After implementing the diagnostic survival mode, the post-diagnostic contamination entropy was measured.

[0126] The system resilience index is calculated based on the post-diagnosis pollution entropy and the preset reference pollution entropy.

[0127] If the system resilience index exceeds the preset recovery threshold, an instruction to switch back to the optimal analysis efficiency mode will be generated.

[0128] If the system resilience index does not exceed the recovery threshold, instructions are generated to maintain diagnostic survival mode or to perform further diagnosis.

[0129] Based on Example 1, this embodiment limits the operational logic of the system control module after executing the diagnostic survival mode. Its core is to establish a quantitative system recovery assessment mechanism to ensure the integrity and reliability of the decision-making closed loop.

[0130] After receiving a decision to switch to diagnostic survival mode, the system control module controls the HPLC hardware unit to execute preset diagnostic and cleaning procedures. The innovation of this embodiment lies in its subsequent evaluation and recovery strategy, the specific workflow of which is as follows:

[0131] After executing the diagnostic survival mode, the post-diagnostic contamination entropy is measured. Following completion of the diagnostic procedure, the system control module instructs the system to enter a brief stabilization period. During this period, by invoking the aforementioned data acquisition, first processing, and second processing modules, the current system contamination entropy is remeasured and calculated; this value is defined as the post-diagnostic contamination entropy. ;

[0132] Based on the post-diagnostic contamination entropy and the preset reference contamination entropy, the system resilience index is calculated; to objectively evaluate the recovery effect, this embodiment introduces the system resilience index. Reference pollution entropy This refers to the baseline value of the contamination entropy of the system under ideal clean conditions; its function is to provide a benchmark for complete recovery. It is a baseline value obtained from calibration tests conducted under clean conditions after a brand-new installation or thorough manual maintenance of the system, and is stored in the system. To ensure the validity of the calculation, if the calibration test results... If it is zero, then it is set to a preset non-zero minimum value that represents the lowest background noise level of the instrument;

[0133] System resilience index The calculation formula is as follows:

[0134] ;

[0135] in, It is a dimensionless resilience index, typically ranging from 0 to 1. The closer it is to 1, the more thorough the recovery. It is calculated by this module. The pollution entropy, measured after the diagnostic procedure, has the dimension of V·s and is determined in the first step of this module; The preset reference pollution entropy, with dimensions V·s, is a preset calibration value of the system; the logic of this formula lies in... The ratio reflects the difference between the current state and the optimal state. Subtracting this ratio from 1 gives the degree of recovery.

[0136] Decision-making is based on system resilience index and recovery threshold; recovery threshold It is a preset minimum resilience index standard for judging whether the system has fully recovered to the point where it can undertake normal analysis tasks; its function is to prevent the system from returning to a high-load working state prematurely if the contamination has not been completely removed; it is based on statistical analysis of historical data, for example, to determine the quantile corresponding to the recovery level that can ensure 99% of the analysis results are not affected by residual contamination, ensuring that the setting of this threshold has a scientific basis and statistical significance, such as 0.95;

[0137] If the system resilience index exceeds the preset recovery threshold If the system control module determines that the system has been successfully restored, it will generate an instruction to switch back to the optimal analysis efficiency mode.

[0138] If the system resilience index does not exceed the recovery threshold If the result is negative, it indicates that the single diagnostic procedure was ineffective and the system is still in a contaminated state. At this time, the system control module will generate instructions to maintain the diagnostic survival mode or perform further diagnosis, such as starting a more powerful cleaning procedure or issuing an alarm to the operator that manual intervention is required.

[0139] This embodiment introduces a system resilience index. This quantitative evaluation indicator greatly enhances the system's intelligence and reliability; it achieves a closed-loop decision-making process, enabling the system not only to make and execute decisions but also to evaluate their effectiveness, ensuring the integrity of the entire detection-decision-execution-evaluation process; and it restores the scientific nature of decision-making by using objective and quantifiable data. It replaces the traditional recovery method that relies on the operator's subjective judgment or fixed time, ensuring that the analysis task is resumed only when the system is truly healthy, thus fundamentally guaranteeing the quality of subsequent analysis data; it also improves the system's resilience, enabling the system to recover from abnormal states and verify the degree of recovery, making the entire analysis platform more robust and reliable.

[0140] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention.

Claims

1. An HPLC mobile phase gradient intelligent proportioning online purification system, characterized in that, include: The data acquisition module is used to acquire the baseline signal of the high-performance liquid chromatography detector in real time; The first processing module is used to determine noise fluctuation parameters based on the baseline signal; The second processing module is used to calculate the pollution entropy based on the noise fluctuation parameters and the preset system pollution sensitivity coefficient; The third processing module is used to calculate the actual pollution entropy accumulation rate based on pollution entropy, and determine the anomaly index by combining it with the preset pollution entropy accumulation rate baseline model. The decision-making module is used to generate operating mode decisions based on anomaly indices. The system control module is used to execute the corresponding operating mode in response to the operating mode decision, and after the diagnostic survival mode is executed, to determine whether to switch back to the optimal analysis efficiency mode. The first processing module is specifically used for: Perform drift removal processing on the baseline signal within a time window; Calculate the standard deviation of the baseline signal after drift removal to generate noise fluctuation parameters; The second processing module is specifically used for: The noise fluctuation parameter is used as the instantaneous rate of pollution accumulation; The contamination entropy is generated by integrating the instantaneous rate over time. ; The calculation method is as follows: in, Let be the cumulative pollution entropy at time t, with dimensions V·s; This is the start time of the system's initial clean state. Its data type is time unit such as seconds. It is the time point recorded when the system runs for the first time or after a thorough maintenance is completed. In time The noise fluctuation parameter is used as the instantaneous rate input of pollution accumulation, and its dimension is V; The system contamination sensitivity coefficient is a dimensionless, adjustable parameter used to characterize the differences in sensitivity to chemical contamination response of a specific HPLC system configuration and the current analytical method category. The third processing module is specifically used for: The actual pollution entropy accumulation rate is calculated by numerically differentiating the pollution entropy data sequence. The predicted pollution entropy accumulation rate under normal operating conditions is obtained by using a baseline model of pollution entropy accumulation rate. The standardized deviation of the actual pollution entropy accumulation rate relative to the predicted pollution entropy accumulation rate is calculated to generate an anomaly index. Abnormal Index Calculation: in, It is a dimensionless property designed to prevent the denominator from being zero. The same tiny positive constant; It is a dimensionless real-time anomaly index, serving as a core risk signal; The actual rate of accumulation of pollution entropy is expressed in volts (V). The predicted rate of accumulation of normal pollution entropy is expressed in volts (V).

2. The HPLC mobile phase gradient intelligent proportioning online purification system according to claim 1, characterized in that, The decision-making module is specifically used for: Based on the anomaly index and preset risk preference parameters, the first utility value representing the utility of the optimal analysis efficiency model is calculated; Based on the anomaly index and preset diagnostic cost parameters, a second utility value characterizing the utility of diagnostic survival mode is calculated. If the second utility value is greater than the first utility value, a decision to switch to diagnostic survival mode is generated. If the second utility value is not greater than the first utility value, then an operating mode decision is generated to maintain the optimal analysis efficiency mode.

3. The HPLC mobile phase gradient intelligent proportioning online purification system according to claim 1, characterized in that, The system control module is specifically used for: After implementing the diagnostic survival mode, the post-diagnostic contamination entropy was measured. The system resilience index is calculated based on the post-diagnosis pollution entropy and the preset reference pollution entropy. If the system resilience index exceeds the preset recovery threshold, an instruction to switch back to the optimal analysis efficiency mode will be generated. If the system resilience index does not exceed the recovery threshold, instructions are generated to maintain diagnostic survival mode or to perform further diagnosis. System resilience index The calculation formula is as follows: in, It is a dimensionless resilience index, with a value ranging from 0 to 1. The closer it is to 1, the more thorough the recovery. The pollution entropy, measured after performing the diagnostic procedure, has the dimension of V·s; The preset reference pollution entropy, with dimensions V·s, is a preset calibration value of the system; the logic of this formula lies in... The ratio reflects the difference between the current state and the optimal state. Subtracting this ratio from 1 gives the degree of recovery. Decision-making is based on system resilience index and recovery threshold; recovery threshold It is a preset minimum system resilience index standard for judging whether the system has been fully restored to the point where it can undertake normal analysis tasks.

4. The HPLC mobile phase gradient intelligent proportioning online purification system according to claim 2, characterized in that, The first utility value includes a benefit term representing the value of the analytical throughput and a risk cost term that grows non-linearly with the anomaly index.

5. The HPLC mobile phase gradient intelligent proportioning online purification system according to claim 2, characterized in that, The second utility value includes an information gain term proportional to the anomaly index and a direct diagnostic cost term.

Images