Liquid Chromatograph Status Monitoring and Predictive Maintenance Method and System

Abstract

The present invention provides a method and system for monitoring the status and predictive maintenance of a liquid chromatography instrument, which relates to the technical field of data processing. By jointly modeling the sensor data of the liquid chromatography instrument (such as continuous parameters like flow rate, temperature, pressure, etc.) and the log data (such as discrete information like alarm status, equipment events, etc.), and performing data dimensionality reduction based on the geometric similarity of data points, the local structure and global information of the liquid chromatography instrument data are maintained, while reducing the information loss during the dimensionality reduction process. Further, a status monitoring model is used to process the data, enhancing the context awareness ability of the model. When anomalies or faults occur, it can comprehensively consider various factors for accurate diagnosis, showing higher accuracy and reliability in monitoring the complex status changes of the liquid chromatography instrument.

Description

Technical Field

[0001] The present invention relates to the technical field of data processing, and in particular, to a method and system for monitoring the state and predictive maintenance of a liquid chromatography instrument. Background Art

[0002] As a precision analytical instrument, the liquid chromatography instrument is widely used in fields such as chemical analysis, biopharmaceuticals, and food safety detection. With the continuous development of technology, the application scope and usage frequency of the liquid chromatography instrument are increasing continuously. However, equipment failures and downtime issues are still the main factors affecting its operating efficiency and reliability. The liquid chromatography instrument involves real-time monitoring of multiple parameters during operation, such as flow rate, pressure, temperature, solvent concentration, etc. Minor fluctuations in these parameters may affect the performance of the equipment and even lead to failures.

[0003] With the progress of technology, modern liquid chromatography instruments can monitor multiple parameters in real time (such as flow rate, pressure, temperature, solvent concentration, etc.), and these data constitute multi-dimensional time series. However, most of the existing methods rely on manual experience judgment and lack an automated data analysis and warning mechanism, resulting in untimely problem discovery and missed best treatment opportunities. In addition, the amount of data generated by each operation of the liquid chromatography instrument is huge, especially in the case of high-frequency sampling. Traditional storage and processing means cannot efficiently manage these data. Moreover, the interaction between multiple parameters makes data analysis more complex. For example, a change in flow rate may affect pressure and temperature, and the changes in these variables in turn affect the separation effect. In summary, traditional data processing methods face challenges in processing high-dimensional data. Summary of the Invention

[0004] In view of this, the purpose of the present invention is to provide a method and system for monitoring the state and predictive maintenance of a liquid chromatography instrument, which can realize intelligent operation and maintenance management of the liquid chromatography instrument, so as to improve the service life of the equipment, optimize the allocation of maintenance resources, reduce operating costs, and ensure the stability and reliability of analytical work.

[0005] In a first aspect, an embodiment of the present invention provides a method for monitoring the state and predictive maintenance of a liquid chromatography instrument, the method including: obtaining operation monitoring data of the liquid chromatography instrument to be monitored; performing distance measurement on data points of the operation monitoring data to determine the geometric similarity of the data points corresponding to the operation monitoring data; using a pre-trained autoencoder to perform dimensionality reduction processing on the operation monitoring data based on the geometric similarity of the data points to determine a dimensionality-reduced data representation corresponding to the operation monitoring data; loading the dimensionality-reduced data representation into a pre-trained state monitoring model, and outputting a state analysis result corresponding to the dimensionality-reduced data representation; wherein, the state monitoring model updates model parameters based on a fractional-order modified cross-entropy loss function; and based on the state analysis result, evaluating the working state of the liquid chromatography instrument.

[0006] In combination with the first aspect, the embodiment of the present invention further provides a first implementation manner of the first aspect. Among them, the state monitoring model is constructed based on a high-order neural network; the step of loading the dimensionality-reduced data representation into a pre-trained state monitoring model and outputting the state analysis result corresponding to the dimensionality-reduced data representation includes: performing forward propagation on the dimensionality-reduced data representation through multiple hidden layers of the high-order neural network; and using the deep structure of the high-order neural network to perform layer-by-layer fusion on the dimensionality-reduced data representation to determine the state analysis result corresponding to the dimensionality-reduced data representation.

[0007] In combination with the first aspect, the embodiment of the present invention further provides a second implementation manner of the first aspect. Among them, the method further includes: training a preset initial high-order neural network using a preset training sample set and determining the cross-entropy loss function corresponding to the initial high-order neural network; performing fractional-order correction on the cross-entropy loss function using a preset fractional-order calculus operator to determine the fractional-order loss gradient; based on the fractional-order loss gradient and the cross-entropy loss function, performing backpropagation update on the model parameters of the initial high-order neural network; until the initial high-order neural network meets the preset iteration condition, constructing a state monitoring model based on the initial high-order neural network.

[0008] In combination with the first aspect, the embodiment of the present invention further provides a third implementation manner of the first aspect. Among them, the step of performing distance measurement on the data points of the operation monitoring data and determining the geometric similarity of the data points corresponding to the operation monitoring data includes: measuring the distance of the data points of the operation monitoring data based on the per-dimensional difference of the data points of the operation monitoring data; calculating the geometric similarity matrix corresponding to the operation monitoring data based on the data point distance and a preset similarity weight coefficient, and determining the geometric similarity of the data points corresponding to the operation monitoring data.

[0009] In combination with the first aspect, the embodiment of the present invention further provides a fourth implementation manner of the first aspect. Among them, the step of using a pre-trained autoencoder to perform dimensionality reduction processing on the operation monitoring data based on the geometric similarity of the data points and determining the dimensionality-reduced data representation corresponding to the operation monitoring data includes: based on the geometric similarity of the data points, using a preset autoencoder to perform weighted mapping on the operation monitoring data to determine the initial dimensionality-reduced features; measuring the interaction of the data points of the initial dimensionality-reduced features and determining the feature mapping from the initial dimensionality-reduced features; performing adaptive smoothing processing on the feature mapping to obtain the dimensionality-reduced data representation corresponding to the operation monitoring data.

[0010] Combined with the first aspect, the embodiments of the present invention further provide a fifth implementation manner of the first aspect. Among them, the steps of measuring the interaction of the data points of the initial dimensionality-reduced features and determining the feature mapping from the initial dimensionality-reduced features include: determining the gradient change corresponding to the data points of the initial dimensionality-reduced features and the geometric distance between the data points; determining the interaction energy of the data points of the initial dimensionality-reduced features based on the gradient change and the geometric distance; minimizing the interaction energy to determine the feature mapping corresponding to the initial dimensionality-reduced features.

[0011] Combined with the first aspect, the embodiments of the present invention further provide a sixth implementation manner of the first aspect. Among them, the training method of the autoencoder includes: training a preset autoencoder through a preset training sample set and calculating the loss function of the autoencoder; introducing a constraint condition of minimizing the path difference into the loss function and calculating the gradient of the loss function with respect to the weights of the autoencoder; updating the parameters of the autoencoder based on the gradient and the pre-calculated adaptive weight adjustment coefficient.

[0012] Combined with the first aspect, the embodiments of the present invention further provide a seventh implementation manner of the first aspect. Among them, the calculation method of the adaptive weight adjustment coefficient includes: measuring the local density of each data point of the operation monitoring data in its neighborhood and dynamically calculating the adaptive weight adjustment coefficient.

[0013] In the second aspect, the embodiments of the present invention further provide a liquid chromatography instrument state monitoring and predictive maintenance device, which includes: a data acquisition module for acquiring the operation monitoring data of the liquid chromatography instrument to be monitored; a calculation module for measuring the distance of the data points of the operation monitoring data to determine the geometric similarity of the data points corresponding to the operation monitoring data; a data processing module for performing dimensionality reduction processing on the operation monitoring data based on the geometric similarity of the data points by using a pre-trained autoencoder to determine the dimensionality-reduced data representation corresponding to the operation monitoring data; an execution module for loading the dimensionality-reduced data representation into a pre-trained state monitoring model and outputting the state analysis result corresponding to the dimensionality-reduced data representation; wherein, the state monitoring model updates the model parameters based on the fractional-order modified cross-entropy loss function; a predictive maintenance module for evaluating the working state of the liquid chromatography instrument based on the state analysis result.

[0014] In the third aspect, the embodiments of the present invention further provide a liquid chromatography instrument state monitoring and predictive maintenance system, which is configured with the above device and is used to execute the method of any of the above embodiments.

[0015] The embodiments of the present invention bring the following beneficial effects: The present invention provides a method and system for liquid chromatograph status monitoring and predictive maintenance, which relates to the technical field of data processing. By jointly modeling the liquid chromatograph sensor data (such as continuous parameters like flow rate, temperature, pressure, etc.) and log data (such as discrete information like alarm status, equipment events, etc.), and performing data dimensionality reduction based on the geometric similarity of data points, the local structure and global information of the liquid chromatograph data are maintained, while the information loss during the dimensionality reduction process is reduced. Further using the status monitoring model for data processing improves the context awareness ability of the model. When abnormalities or faults occur, it can comprehensively consider various factors for accurate diagnosis, showing higher accuracy and reliability when monitoring the complex status changes of the liquid chromatograph.

[0016] Other features and advantages of the present invention will be described in the following specification, and, in part, will be obvious from the specification, or will be understood by implementing the present invention. The objectives and other advantages of the present invention are realized and obtained by the structures particularly pointed out in the specification, claims, and drawings.

[0017] To make the above objectives, features, and advantages of the present invention more obvious and understandable, the following specifically enumerates preferred embodiments and, in conjunction with the accompanying drawings, provides a detailed description as follows. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] In order to more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the following will briefly introduce the drawings required for the description of the specific embodiments or the prior art. Obviously, the following drawings are some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can also be obtained based on these drawings.

[0019] Figure 1 It is a flowchart of a method for liquid chromatograph status monitoring and predictive maintenance provided by an embodiment of the present invention;

[0020] Figure 2 It is a flowchart of another method for liquid chromatograph status monitoring and predictive maintenance provided by an embodiment of the present invention;

[0021] Figure 3 It is a schematic diagram showing the performance of the dimensionality reduction algorithm under different hyperparameter settings provided by an embodiment of the present invention;

[0022] Figure 4 It is a schematic diagram comparing the classification accuracies of the dimensionality reduction algorithms provided by an embodiment of the present invention;

[0023] Figure 5 It is a schematic diagram of the structure of a liquid chromatograph status monitoring and predictive maintenance device provided by an embodiment of the present invention;

[0024] Figure 6 It is a logical schematic diagram of a liquid chromatograph status monitoring and predictive maintenance system provided by an embodiment of the present invention;

[0025] Figure 7 It is a structural schematic diagram of an electronic device provided by an embodiment of the present invention. Detailed implementation manners

[0026] To make the objectives, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions of the present invention will be clearly and completely described below in conjunction with the embodiments. Obviously, the described embodiments are some, but not all, of the embodiments of the present invention. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0027] Traditional maintenance methods for liquid chromatographs usually rely on regular inspections and manual interventions, and cannot timely and accurately detect potential equipment failures and performance degradation, resulting in the risks of over-maintenance or under-maintenance. This not only increases the maintenance cost, but also may lead to unplanned downtime of the equipment, affecting the overall production efficiency and the accuracy of analysis results. Therefore, how to achieve real-time status monitoring and predictive maintenance of liquid chromatographs has become a key issue for improving equipment operation efficiency, extending service life and reducing maintenance costs.

[0028] With the rapid development of big data and artificial intelligence technologies, especially the application of machine learning and deep learning algorithms, new ideas and solutions have been provided for the status monitoring and fault prediction of liquid chromatographs. By collecting the operation data of the equipment and using machine learning methods to establish an equipment status model, not only can the working status of the liquid chromatograph be monitored in real time, but also potential fault problems can be warned in advance through the analysis of historical data, so as to achieve more intelligent and precise maintenance management. However, there are still some problems in existing monitoring and maintenance methods, such as computational redundancy caused by high data dimensions, traditional dimensionality reduction methods being unable to retain local and global information in the data, overfitting in the model training process, etc. These problems limit the effectiveness of existing technologies in practical applications.

[0029] The embodiments of the present invention provide a liquid chromatograph status monitoring and predictive maintenance method and system, which can realize the intelligent operation and maintenance management of liquid chromatographs, so as to improve the service life of the equipment, optimize the allocation of maintenance resources, reduce the operation cost, and ensure the stability and reliability of analysis work.

[0030] For the convenience of understanding this embodiment, first, a liquid chromatograph status monitoring and predictive maintenance method disclosed in the embodiments of the present invention will be introduced in detail. See Figure 1Schematic flowchart of a method for monitoring the state and predictive maintenance of a liquid chromatography instrument, the method comprising the following steps:

[0031] Step S102, obtaining operation monitoring data of the liquid chromatography instrument to be monitored.

[0032] Most of the existing technologies are aimed at processing a single data source and cannot effectively reflect the complex patterns in the operation process of the liquid chromatography instrument. The operation monitoring data in the embodiments of the present invention includes the equipment operation parameters of the liquid chromatography instrument and the status operation information in the log of the liquid chromatography instrument. At the same time, real-time physical parameters provided by sensors and event information in the log records are used to obtain a more comprehensive and richer data set. By jointly modeling the two types of data, abnormal situations can be identified more accurately. For example, a small change in a certain parameter may not be noticed alone, but if accompanied by an abnormal record in the log (such as an alarm or a fault), potential problems can be discovered earlier. Among them, data collection and storage can be performed through a data collection unit, including sensor data collection and log data parsing. In addition, the data can be preprocessed and used as operation monitoring data. Among them, sensor data collection is used to collect equipment operation parameters such as flow rate, pressure, temperature, and solvent composition; log data parsing is used to extract operation status information from the instrument's own log, including start time, stop time, fault records, alarm information, etc.; data preprocessing is used for denoising, normalization, and outlier processing to improve data quality; the storage function ensures that the data is stored in a database or a data lake in an efficient and secure manner for subsequent analysis.

[0033] Step S104, performing distance measurement on the data points of the operation monitoring data to determine the geometric similarity of the data points corresponding to the operation monitoring data.

[0034] Step S106, using a pre-trained autoencoder to perform dimensionality reduction processing on the operation monitoring data based on the geometric similarity of the data points to determine the dimensionality-reduced data representation corresponding to the operation monitoring data.

[0035] In the embodiments of the present invention, distance measurement is performed on each data point, and the calculation results of the distances between each data point and all other data points are combined into a similarity matrix, which reflects the geometric similarity between the data points and quantifies the spatial relationship between different data points. And based on this, data dimensionality reduction is performed, which can ensure that the data points after dimensionality reduction still maintain the original geometric structure in the low-dimensional space. These low-dimensional data not only retain the key features of the original data, but also reduce redundant information, significantly reducing storage requirements. And it can be more easily visualized to help researchers better understand the distribution and characteristics of the data.

[0036] Step S108, loading the dimensionality-reduced data representation into a pre-trained state monitoring model and outputting the state analysis result corresponding to the dimensionality-reduced data representation.

[0037] Step S110, based on the state analysis result, evaluate the working state of the liquid chromatograph.

[0038] With the rapid development of big data and artificial intelligence technologies, the state monitoring model established by the machine learning and deep learning algorithms in the embodiments of the present invention is used to realize the state monitoring and fault prediction of the liquid chromatograph. It can not only monitor the working state of the liquid chromatograph in real time, but also, through the analysis of historical data, give early warnings of potential fault problems, so as to realize more intelligent and precise maintenance management.

[0039] Among them, the results output by the model include evaluations of different states such as normal operation, early warning of potential problems, and fault detection. According to the state analysis result, it is possible to judge the current working state (such as normal, warning, fault) of the liquid chromatograph, and put forward corresponding maintenance suggestions or warning information. In one embodiment, the method for state monitoring and predictive maintenance of the liquid chromatograph proposed in the embodiments of the present invention relies on the state monitoring and predictive maintenance system of the liquid chromatograph to realize the real-time monitoring of the equipment operation state, abnormal detection and fault prediction. The system ensures the stable operation of the liquid chromatograph through the collaborative work of multiple units, improves the maintenance efficiency, and reduces the risk of unplanned downtime. For example, the state monitoring unit can perform state analysis based on the real-time collected data to judge whether the equipment is in a normal working state; the state monitoring unit can perform data comparison, abnormal detection and alarm processing. Data comparison is used to monitor the deviation between the current operating parameters and the historical normal state in real time; abnormal detection determines whether there is an operating abnormality based on a machine learning model or a rule threshold; alarm processing is used to trigger the alarm mechanism and remind the operation and maintenance personnel to take corresponding measures through a visual interface or a notification system.

[0040] Furthermore, fault prediction, remaining life estimation and maintenance plan recommendation can also be carried out through the predictive maintenance unit. Among them, based on the historical operation data and the prediction model, the health status of the equipment can be evaluated and maintenance suggestions can be provided. Fault prediction speculates the possible future fault types and times based on time series analysis and machine learning methods; remaining life estimation calculates the expected service life of key components based on the equipment aging model and historical fault data; maintenance plan recommendation combines the equipment usage, inventory situation and maintenance resources to generate an optimized maintenance plan to avoid over-maintenance or under-maintenance.

[0041] Furthermore, status display, alarm management, and interaction control can also be performed through the visualization and user interaction unit. It provides intuitive monitoring results and interaction functions, enhancing the user experience and decision-making efficiency. Among them, the status display is used to present the operating status of the device in the form of charts, dashboards, or heat maps; the alarm management provides functions such as historical alarm record query, alarm confirmation, and processing tracking; the interaction control allows users to adjust the monitoring threshold, select different prediction models, or export analysis reports to support personalized needs.

[0042] In summary, through the collaborative operation of each functional unit in the embodiments of the present invention, intelligent operation and maintenance management of the liquid chromatograph is realized, the service life of the device is increased, the maintenance resource configuration is optimized, the operation cost is reduced, and the stability and reliability of the analysis work are ensured.

[0043] Furthermore, the status monitoring and detection model in the embodiments of the present invention updates the model parameters based on the fractional-order modified cross-entropy loss function, which can enhance the convergence stability and suppress overfitting. When updating the model parameters, not only the instantaneous gradient information is considered, but also the memory characteristics of the historical gradients are used to improve the fitting stability.

[0044] In one implementation, the status monitoring model is constructed based on a high-order neural network. Forward propagation is performed on the dimensionality-reduced data representation through multiple hidden layers of the high-order neural network; and, the deep structure of the high-order neural network is used to perform layer-by-layer fusion on the dimensionality-reduced data representation to determine the status analysis result corresponding to the dimensionality-reduced data representation.

[0045] In specific implementation, the liquid chromatograph data is loaded into the high-order neural network model, and the liquid chromatograph data undergoes forward propagation through multiple hidden layers to calculate the output result of the network according to the input data. Among them, by passing the input data layer by layer to each layer in the neural network and applying the activation function, the output result of the network is finally obtained. This output result can be a classification label, a regression value, or other forms of predicted values.

[0046] Specifically, each hidden layer adopts the ReLU activation function to increase the non-linear processing ability, so as to better depict complex mapping relationships. For example, the flow rate and pressure parameters of the liquid chromatograph often show piecewise non-linear changes during operation (such as sudden changes in flow rate during the gradient elution stage), and ReLU can effectively capture such local features through its sparse activation characteristics. At the same time, the deep structure of the high-order neural network can fuse sensor data (continuous numerical type) and log parsing data (discrete status labels) layer by layer. For example, the continuous fluctuations of the temperature sensor and the "column temperature over-limit alarm" event recorded in the log are jointly modeled, so as to improve the context awareness ability of status classification. Expressed as:

[0047]

[0048] In the formula, is the linear transformation output of the q-th layer of the high-order neural network; is the output of the q-th layer of the high-order neural network after being processed by the ReLU activation function, is the output of the (q - 1)-th layer of the high-order neural network after being processed by the ReLU activation function; is the weight matrix of the q-th layer of the high-order neural network; represents the maximum value function.

[0049] In traditional neural network training methods, in the long-term trend prediction of equipment degradation or fault detection, especially when dealing with complex and non-linear data, problems such as slow convergence or overfitting may occur. The embodiments of the present invention update model parameters based on a fractional-order modified cross-entropy loss function to construct a state monitoring model, so as to enhance convergence stability and suppress overfitting. The network can remember historical gradient information and perform excellently when there are feature offsets caused by equipment aging, avoiding the instability caused by relying only on instantaneous gradients in the traditional gradient descent method, and showing higher accuracy and reliability when monitoring the complex state changes of liquid chromatography instruments. The specific steps are as follows:

[0050] 1) Use a preset training sample set to train a preset initial high-order neural network, and determine the cross-entropy loss function corresponding to the initial high-order neural network.

[0051] Initialize the parameters of the high-order neural network model, including the weights and biases of the high-order neural network. Use the Glorot method to initialize the weights and biases of the high-order neural network to ensure the rationality of the initial distribution, thereby improving the convergence speed and the stability of the fitting performance. For example, solvent composition parameters usually have non-linear coupling relationships. A reasonable initial weight distribution can accelerate the model's ability to extract features from complex solvent concentration changes in the initial stage of training, avoiding gradient disappearance or explosion caused by too large or too small initial parameter values. The initialization method is expressed as:

[0052]

[0053] In the formula, is the element in the i-th row and j-th column of the weight matrix of the q-th layer of the high-order neural network; is the bias vector of the q-th layer of the high-order neural network; represents the parameter update operation; and are the numbers of neurons in the (q - 1)-th layer and the q-th layer of the high-order neural network; r is a random number drawn from the uniform distribution U(-1, 1); i is a positive integer, j is a positive integer, and q is a positive integer.

[0054] Further, forward propagation is performed on the training data, which can refer to the above steps. Further, the cross-entropy loss function is used to evaluate the difference between the model output and the actual label, and the fractional calculus is combined to improve the backpropagation update of the model parameters to achieve a more stable fitting process.

[0055] Specifically, first define the conventional cross-entropy loss function and its calculation formula for the integer gradient. In the embodiments of the present invention, the fractional calculus operator is used to perform fractional correction on the gradient during the backpropagation process to enhance the convergence stability of the model and suppress the overfitting trend. The calculation method of the conventional cross-entropy loss function is expressed as:

[0056]

[0057] In the formula, p c is the probability distribution predicted by the model; y c is the true label distribution; C is the total number of categories, representing the number of categories in the classification problem; L r is the cross-entropy loss function, which is used to evaluate the difference between the predicted probability and the actual probability; is the partial derivative symbol.

[0058] Among them, the integer gradient of the loss function with respect to the network output is expressed as:

[0059]

[0060] In the formula, p c is the probability distribution predicted by the model, and y c is the true label distribution.

[0061] Moreover, the integer-order derivative of the ReLU activation function is expressed as:

[0062]

[0063] In the formula, is the linear transformation output of the q-th layer of the high-order neural network.

[0064] Moreover, the integer gradient of the linear transformation with respect to the weight is approximately equal to the output of the previous layer, which is expressed as:

[0065]

[0066] In the formula, h (q-1) is the output of the (q-1)-th layer of the high-order neural network processed by the ReLU activation function.

[0067] 2) Use the preset fractional calculus operator to perform fractional correction on the cross-entropy loss function to determine the fractional loss gradient.

[0068] To enhance the fitting stability of high-order neural networks, the present invention further adopts a fractional calculus operator on the basis of the original gradient calculation. The memory characteristic of historical gradient information in parameter update is utilized to improve the convergence stability of the model. Since the operating parameters of a liquid chromatograph (such as flow rate and temperature) have time continuity and state inertia, traditional integer-step gradients only rely on the current instantaneous gradient, while fractional calculus enhances the fitting stability of the model to the gradual change process of equipment state by integrating historical gradient information (such as the gradient change trend of pressure parameters in the past 100 iterations). For example, when detecting slow cumulative faults such as "column aging", fractional correction can suppress misclassification caused by fluctuations in single-sampling data and smooth the impact of instantaneous noise through the historical gradient memory characteristic.

[0069] Preferably, the Caputo-type fractional derivative (or other equivalent fractional operators) is adopted to perform fractional-step gradient correction on the loss function, and the fractional loss gradient is defined as:

[0070]

[0071] where is the order of the fractional derivative, 0 < α ≤ 1, preferably, α = 0.9; represents the process of performing fractional transformation on the integer-step gradient.

[0072] To further reflect the combination of this operator and the network structure, expand the components of . Preferably, the following formula is used to perform fractional correction on the weight gradient of the i-th row and j-th column, which can retain historical information and smooth the fluctuations of instantaneous gradients during backpropagation, thereby enhancing the stability of the network fitting process, expressed as:

[0073]

[0074] In the formula, is the Gamma function; is the upper limit of discrete training steps; represents the integer-step gradient at the time of training step ; is a positive integer; represents the gradient coefficient corresponding to in the conventional integer-step gradient (which includes the derivative information of the forward propagation and the output of the previous layer); and are the predicted probability and the true probability corresponding to the -th training step respectively.

[0075] 3) Based on the fractional-order loss gradient and the cross-entropy loss function, backpropagate and update the model parameters of the initial high-order neural network.

[0076] In the embodiments of the present invention, by combining adaptive sub-gradient and fractional-order corrected parameter update, the learning rate of each parameter is dynamically adjusted to suppress overfitting and balance the convergence speed. Define the conventional adaptive sub-gradient update method, and then further add a fractional-order correction term to implement the training of a high-order neural network classifier model with enhanced fitting stability. First, give the weight update method of the conventional adaptive sub-gradient, which is expressed as:

[0077]

[0078] In the formula, is the parameter update operation; is the global learning rate; is the cumulative square of the gradient before update; is a small constant to avoid division by zero. Preferably, is set to 0.001.

[0079] Furthermore, the update method of the cumulative square of the gradient is expressed as:

[0080]

[0081] In the formula, is the parameter update operation.

[0082] However, in order to consider the gradient correction effect of fractional calculus on the basis of the adaptive sub-gradient, further is incorporated into the update formula to perform adaptive sub-gradient update with fractional-order correction, so that the high-order neural network not only considers the instantaneous integer-order gradient information during the update process, but also considers the memory and smoothing effects of the fractional-order on the historical gradient, thereby improving the overall fitting stability and generalization performance. For example, there is a dimensional difference between the numerical range of the solvent component parameter (such as the methanol proportion) and the binary state marker in the log (such as "pump start / stop"). The traditional unified learning rate may lead to unbalanced parameter updates. By adopting the adaptive mechanism with the fractional-order correction term, the model can assign different gradient weights to the high-frequency fluctuation data of the flow sensor (which requires rapid response) and the low-frequency events of the log state (which requires long-term memory). In the scenario of solvent component mutation, the fractional-order gradient correction can delay the update speed of related parameters and avoid overfitting caused by instantaneous concentration jumps, while the state markers of log parsing maintain stable feature associations through the adaptive learning rate.

[0083] In summary, the update combining fractional-order gradient correction is expressed as:

[0084]

[0085] In the formula, is the updated weight after combining the fractional gradient correction; is the balance coefficient between the fractional gradient and the integer gradient. Preferably, = 0.1.

[0086] 4) Until the initial high-order neural network meets the preset iteration conditions, a state monitoring model is constructed based on the initial high-order neural network.

[0087] Repeat the above steps iteratively until the preset stop iteration condition is met, which indicates that the model training is completed. In one embodiment, the preset stop iteration condition is to reach the preset maximum number of iterations. Preferably, the preset maximum number of iterations is set to 1000 times.

[0088] In summary, the embodiment of the present invention adopts a high-order neural network model classification algorithm based on a fractional-order update strategy, combines fractional calculus for gradient correction in the backpropagation process, and can enhance the convergence stability and suppress overfitting. At the same time, combined with the adaptive subgradient update method, the fractional gradient correction method is used, so that when the model parameters are updated, not only the instantaneous gradient information is considered, but also the memory characteristics of the historical gradient can be utilized to improve the fitting stability.

[0089] Further, on the basis of the above embodiment, the embodiment of the present invention also provides another method for state monitoring and predictive maintenance of a liquid chromatograph. Through the collaborative work of multiple functional units such as data acquisition, feature extraction, model training, prediction analysis, and maintenance suggestions, the stable operation of the liquid chromatograph is ensured, the maintenance efficiency is improved, and the risk of unplanned shutdown is reduced. Among them, feature engineering, model training, model evaluation, and model update can be realized through a machine learning modeling unit. A state monitoring and prediction model is established based on historical data to identify the operation trend of the equipment and early warning of potential faults. Among them, feature engineering is used to extract representative feature variables, including time series features, frequency domain features, and statistical features; model training uses supervised learning or unsupervised learning methods to train classification, regression, or clustering models to achieve anomaly detection, life prediction, and health assessment; model evaluation evaluates the accuracy and generalization ability of the model through indicators such as cross-validation, confusion matrix, and regression error; model update continuously optimizes the training model based on the latest operation data to adapt to the dynamic changes of the equipment state. Figure 2 shows the flowchart of the method. Referring to Figure 2 , the method includes the following steps:

[0090] Step S202, obtain the operation monitoring data of the liquid chromatograph to be monitored.

[0091] Among them, the key parameters during the operation of the liquid chromatograph can be collected in real time through the data acquisition unit of the system, that is, the equipment operation parameters of the liquid chromatograph and the status operation information in the log of the liquid chromatograph, ensuring the integrity and real-time nature of the data. The operation monitoring data can refer to the above embodiments and will not be elaborated here.

[0092] Furthermore, the data is dimensionally reduced. Among them, the liquid chromatograph is usually equipped with multiple sensors to monitor various operation parameters in real time. These sensors generate data once per second or every few seconds, resulting in a rapid increase in the data dimension. In order to capture instantaneous changes and minute fluctuations, the liquid chromatograph usually adopts a high-frequency data acquisition method, generating a large amount of time series data, increasing the data dimension and sample size. In addition, a large amount of historical data is generated during the long-term operation of the equipment, which not only contains the monitoring information of different time periods but may also involve samples of different batches, further increasing the complexity of the data. Moreover, each parameter does not exist in isolation, and there are complex interactions between them. Due to different requirements and processing methods for data in different fields, the data shows a high degree of complexity.

[0093] However, processing and analyzing high-dimensional data will significantly increase the consumption of computing resources, including memory and processing time. By dimensionally reducing the data, the computational complexity can be significantly reduced, making data analysis more efficient. Existing dimensionality reduction methods, such as PCA, mainly compress data through linear relationships and are prone to ignoring the non-linear characteristics of the data. Especially when facing high-dimensional liquid chromatograph data, local structures and global information are easily lost.

[0094] The embodiment of the present invention adopts an autoencoder algorithm based on geometric similarity. By calculating geometric similarity, the relative relationship between input data points is obtained, and a strategy of path reconstruction is combined for feature dimensionality reduction. When reducing the dimension of high-dimensional data, the local structure and global information of the liquid chromatograph data are maintained, and at the same time, the information loss during the dimensionality reduction process is reduced. Moreover, during the feature dimensionality reduction process, an adaptive weight adjustment coefficient for features is adopted. By dynamically calculating the local density, different update rates are assigned to features of different importance. For high-density regions (such as the stable operation stage), the features accelerate learning, while for features in low-density regions (such as the startup transient stage), a more conservative update strategy is adopted, enhancing the robustness and flexibility of the model when dealing with the non-linear and time-continuous characteristics of the liquid chromatograph. Specifically, refer to the following steps S204 - S212.

[0095] Step S204, based on the per-dimensional difference of the data points of the operation monitoring data, measure the distance between the data points of the operation monitoring data.

[0096] Step S206, based on the data point distance and the preset similarity weight coefficient, a geometric similarity matrix corresponding to the operation monitoring data is calculated to determine the geometric similarity of the data points corresponding to the operation monitoring data.

[0097] If the feature dimension of the collected liquid chromatograph is too high, it will easily lead to computational redundancy. In order to solve the problem that it is difficult to retain both local and global information during the dimensionality reduction process of high-dimensional liquid chromatograph data, an autoencoder algorithm based on geometric similarity measurement is adopted. The geometric similarity measurement is combined with the path reconstruction strategy to perform feature dimensionality reduction. While compressing the dimension, the key information of the liquid chromatograph data is preserved as much as possible, thereby improving the representation ability and accuracy after dimensionality reduction.

[0098] The present invention obtains the relative relationship between input data points through geometric similarity calculation, and then uses the pre-trained autoencoder to reduce the dimension of the data based on the relative relationship. The calculation method of the geometric similarity matrix is ​​expressed as:

[0099]

[0100] In the formula, is the geometric similarity matrix, representing the input liquid chromatograph data The similarity between the data points in ; is the input liquid chromatography data of the autoencoder; is the weight coefficient of similarity, representing the data point and the relative importance of Input the i-th data point of the liquid chromatograph data to the autoencoder; Input the jth data point of the liquid chromatography data to the autoencoder; For data points and The distance between The number of data points in the liquid chromatography data input to the autoencoder; i is a positive integer; j is a positive integer.

[0101] Furthermore, in order to better characterize the relative importance between data points, the weight coefficient of similarity is calculated based on the square of the distance between data points, expressed as:

[0102]

[0103] In the formula, is a hyperparameter to control similarity decay; For data points and The square of the distance; For data points and The square of the distance; is the th data point of the input liquid chromatograph data to the autoencoder; is a positive integer. Preferably, is set to 0.95.

[0104] Furthermore, in the process of calculating geometric similarity, for example, by measuring the local structural relationship between multi-dimensional parameters such as flow rate, pressure, and temperature collected by sensors, in the dynamic change scenario of solvent composition, temperature and solvent concentration may show a non-linear association. The model can strengthen the similarity of adjacent data points and weaken the influence of outliers, so as to retain the local characteristics of the solvent gradient change during the operation of the chromatograph and solve the problem of misjudgment of similarity caused by high-dimensional noise in sensor data. The calculation method for measuring the distance between data points is expressed as:

[0105]

[0106] In the formula, is the value of the i-th data point in the k-th feature dimension; m r is the number of feature dimensions of the input liquid chromatograph data to the autoencoder; is the value of the j-th data point of the input liquid chromatograph data to the autoencoder in the k-th feature dimension.

[0107] Step S208, based on the geometric similarity of data points, use a preset autoencoder to perform weighted mapping on the operation monitoring data to determine the initial dimensionality reduction features.

[0108] Using the autoencoder to perform weighted mapping on the data based on geometric similarity to perform dimensionality reduction processing on the data can retain the geometric relationship between data points during the transfer between layers of the autoencoder, so that the dimensionality reduction process can take into account both the local structure and high-dimensional features of the liquid chromatograph data. The calculation method is expressed as:

[0109]

[0110] In the formula, is the representation of the liquid chromatograph data after weighted mapping, that is, the initial dimensionality reduction feature; is the weight matrix of the autoencoder; is the bias term of the autoencoder.

[0111] Furthermore, in order to dynamically obtain the weight matrix and reduce the similarity deviation, the weight matrix of the autoencoder is calculated through a diagonal matrix, expressed as:

[0112]

[0113] In the formula, is an adjustment factor for controlling the weight ratio; is a diagonal matrix, and its diagonal elements , is the element in the i-th row and i-th column of the diagonal matrix; is the element in the i-th row and j-th column of the geometric similarity matrix; i is a positive integer; j is a positive integer; is the number of data points in the data input to the autoencoder from the liquid chromatograph. Preferably, is set to 0.1.

[0114] Furthermore, traditional machine learning algorithms usually adopt a way of fixed learning rate and unified update rate, and the learning of the importance of different liquid chromatograph data features is relatively simple and lacks flexibility. The embodiments of the present invention not only perform data dimensionality reduction based on geometric similarity, but also optimize the path difference between data points before and after dimensionality reduction. Moreover, the autoencoder is trained through the following steps, and an adaptive weight adjustment coefficient of features is adopted. By dynamically calculating the local density, different update rates are assigned to features of different importance. For high-density regions (such as the stable operation stage), the features are accelerated in learning, while for features in low-density regions (such as the start-up transient stage), a more conservative update strategy is adopted, enhancing the robustness and flexibility of the model when dealing with the non-linear and time-continuity features of the liquid chromatograph.

[0115] The specific steps are as follows:

[0116] 1) Train a preset autoencoder through a preset training sample set, and calculate the loss function of the autoencoder.

[0117] 2) Introduce a constraint condition of minimizing the path difference into the loss function, and calculate the gradient of the loss function with respect to the weights of the autoencoder.

[0118] Among them, the above steps for calculating the initial dimensionality-reduced features can be used to process the data as the training sample set for training the autoencoder. To achieve the path reconstruction and dimensionality reduction optimization of the high-dimensional mapping to the low-dimensional, the global structure and topological relationship of the liquid chromatograph data are retained after dimensionality reduction, reducing the loss of global information when mapping from high-dimensional to low-dimensional. The embodiments of the present invention optimize the path difference between data points before and after dimensionality reduction, improving the overall consistency of dimensionality reduction. For example, the instrument operation state (such as the pump pressure fluctuation period) parsed from the log data usually has a global topological relationship in time series. By minimizing the path difference between data points before and after dimensionality reduction, the model can maintain the overall shape of the pressure-flow curve and avoid the breakage of the time series pattern caused by dimensionality reduction. During the iterative training process of the autoencoder, the constraint condition is:

[0119]

[0120] In the formula, Denote the weight matrix and bias term of the autoencoder when minimizing the condition; Is the i-th liquid chromatography data representation after weighted mapping; Is the j-th liquid chromatography data representation after weighted mapping; Is the L2 norm.

[0121] Furthermore, calculate the gradient of the loss function to update the weights of the autoencoder. The calculation method of the gradient of the loss function with respect to the weights is expressed as:

[0122]

[0123] In the formula, Is the partial derivative symbol; Is the reconstruction loss function of the autoencoder.

[0124] 3) Update the parameters of the autoencoder based on the gradient and the pre-computed adaptive weight adjustment coefficient.

[0125] Traditional machine learning algorithms usually adopt the method of fixed learning rate and unified update rate, and learn the importance of different liquid chromatography data features relatively simply, lacking flexibility. Combining the above steps, the embodiments of the present invention adjust the feature weights through dynamic learning and update the parameters in combination with the adaptive weight adjustment coefficient. Specifically, through the adaptive weight adjustment coefficient of the feature, different update rates are adaptively assigned to features of different importance during the training process, enabling the model to accelerate learning for key features and adopt more stable updates for secondary features, thereby improving the flexibility and accuracy of dimensionality reduction. For example, in solvent component analysis, the absorbance features at certain specific wavelengths may be crucial for the detection of target substances, while the baseline drift state recorded in the log may correspond to secondary features. By accelerating the learning of features in high-density regions (such as the stable operation stage) and conservatively updating the features in low-density regions (such as the transient state of instrument startup), the model can more accurately capture the principal component changes of the solvent components while suppressing the temporary device noise recorded in the log. The calculation method is expressed as:

[0126]

[0127] In the formula, Is the updated weight matrix of the autoencoder; Is the learning rate of the autoencoder; Is the adaptive weight adjustment coefficient of the feature.

[0128] Furthermore, calculate the adaptive weight adjustment coefficient of the feature through local density dynamics, and the calculation method is expressed as:

[0129]

[0130] In the formula, is the adjustment factor of the autoencoder; is the local density of the input liquid chromatograph data of the autoencoder; is the exponential function. Preferably, is set to 0.2.

[0131] Furthermore, the local density of the input liquid chromatograph data of the autoencoder measures the importance of the feature in the training liquid chromatograph data, and the calculation method is expressed as:

[0132]

[0133] In the formula, is the density adjustment factor parameter; represents the L2 norm between the i-th data point and the target data point. Preferably, is set to 0.5.

[0134] Step S210, measure the interaction of the data points of the initial dimensionality-reduced features, and determine the feature mapping from the initial dimensionality-reduced features.

[0135] Combined with the above embodiments, the embodiments of the present invention also define an energy function during the feature dimensionality reduction process to measure the interaction energy of the data points with other data points during the dimensionality reduction process. This energy function considers the effects of distance, similarity, and local density to maintain the local structure of the liquid chromatograph data and ensure the smoothness of the dimensionality reduction process, avoiding excessive stretching or compression. For example, if there are short-term abnormal jumps in the temperature data collected by the sensor (such as sensor transient faults), the gradient smoothing strategy in the energy function can suppress the stretching effect of such abnormal points on the overall dimensionality reduction result.

[0136] In specific implementation, determine the gradient change corresponding to the data points of the initial dimensionality-reduced features, and the geometric distance between the data points; based on the gradient change and the geometric distance, determine the interaction energy of the data points of the initial dimensionality-reduced features; minimize the interaction energy to determine the feature mapping corresponding to the initial dimensionality-reduced features. Specifically, the calculation method is expressed as:

[0137]

[0138] In the formula, is the energy function; is the weight coefficient of similarity, representing the relative importance of the data points and ; represents the Euclidean distance between two points, is the density adjustment factor parameter; is the gradient of the data point , representing the local deformation of the liquid chromatograph data.

[0139] Further, by minimizing the energy function as the objective, a feature mapping that minimizes the energy is sought to optimize the feature dimensionality reduction effect. In one implementation, the gradient descent method is used to optimize the energy function, and by updating the gradient of the energy function, the positions of the data points are gradually adjusted to achieve the overall minimum energy. In the gradient descent method, the calculation method of the gradient of the energy function with respect to the data points is expressed as:

[0140]

[0141] In the formula, represents the gradient of the energy function with respect to the th data point; represents the second derivative of the th data point, which characterizes the acceleration of the local change of the liquid chromatograph data.

[0142] Further, the embodiments of the present invention can also measure the interaction of data points based on the energy function to train the autoencoder. Among them, the embodiments of the present invention also adopt a dynamic adaptive adjustment mechanism to adjust the learning rate, automatically adjust the learning rate of the autoencoder according to the current energy value and gradient change, and avoid the convergence problem caused by a fixed learning rate, which is expressed as:

[0143]

[0144] In the formula, is the initial learning rate of the autoencoder; is a hyperparameter that controls the learning rate decay. Preferably, is set to 0.95, is set to 0.01.

[0145] Further, the convergence condition of the autoencoder adopted in the embodiments of the present invention is as follows:

[0146] Dynamically determine whether to meet the convergence according to the loss change and weight stability, and prevent premature stop or overfitting. At the same time, in order to examine the stability of the weights at the same time, the measurement of the weight difference is increased when calculating the loss change. For example, when storing the historical data of the chromatograph for several months, the degradation of the device performance may cause the feature distribution to drift slowly. The traditional fixed-threshold convergence is prone to early stop, while after adopting the weight difference term, the model can continuously track the small changes of the weight matrix. Based on this, when the change of the loss function is very small, it is determined that the training of the model reaches a relatively stable process, and the performance of the model is basically stable, and the training can be stopped.

[0147] In the scenario of gradual change of data distribution, extend the number of training rounds to ensure that the dimensionality reduction model fully adapts to the feature shift caused by device aging. The judgment condition is:

[0148]

[0149] In the formula, is the weight matrix of the autoencoder in the previous iteration; is the convergence adjustment factor; is the loss function value of the i-th iteration of the autoencoder; is the loss function value of the (i - 1)-th iteration of the autoencoder. Preferably, is set to 0.1.

[0150] In summary, repeat the above steps iteratively until the convergence condition is met, which indicates that the training of the autoencoder model is completed.

[0151] Step S212: Perform adaptive smoothing processing on the feature mapping to obtain a reduced-dimensional data representation corresponding to the operation monitoring data.

[0152] In the embodiment of the present invention, the parameters of the trained autoencoder are applied to the data of the liquid chromatograph to obtain a more compact and effective feature representation. The output reduced-dimensional feature representation is as follows:

[0153]

[0154] In the formula, is the reduced-dimensional low-dimensional feature representation; is the weight matrix of the updated autoencoder; is the bias term of the updated autoencoder, and the update method is the same as that of the weight matrix of the autoencoder; is the activation function of the autoencoder.

[0155] Among them, in the embodiment of the present invention, in order to enhance the expression ability of the non-linear liquid chromatograph data, the local density of each data point of the operation monitoring data in its neighborhood is measured, and the adaptive weight adjustment coefficient is calculated dynamically. Specifically, the smoothness is controlled by an adaptive parameter in the activation function, and the calculation method is expressed as:

[0156]

[0157] In the formula, is the adaptive parameter for controlling the smoothness of the activation function. Preferably, is set to 0.1.

[0158] Step S214: Load the reduced-dimensional data representation into a pre-trained state monitoring model, and output a state analysis result corresponding to the reduced-dimensional data representation.

[0159] Step S216: Based on the state analysis result, evaluate the working state of the liquid chromatograph.

[0160] In summary, another method for state monitoring and predictive maintenance of a liquid chromatography instrument provided by the embodiments of the present invention jointly models the sensor data of the liquid chromatography instrument (such as continuous parameters like flow rate, temperature, pressure, etc.) and the log data (such as discrete information like alarm status, equipment events, etc.), and uses a deep high-order neural network for data fusion and processing, which improves the context awareness ability of the model. When an anomaly or fault occurs, it can comprehensively consider various factors for accurate diagnosis.

[0161] Moreover, an autoencoder algorithm based on geometric similarity is adopted. The relative relationship between input data points is obtained through geometric similarity calculation, and a strategy of path reconstruction is combined for feature dimensionality reduction. When reducing the dimensionality of high-dimensional data, the local structure and global information of the liquid chromatography instrument data are maintained, and at the same time, the information loss during the dimensionality reduction process is reduced. During the feature dimensionality reduction process, an adaptive weight adjustment coefficient of the feature is adopted, and the local density is dynamically calculated to assign different update rates to features of different importance. For high-density regions (such as the stable operation stage), the features are accelerated in learning, while for features in low-density regions (such as the startup transient stage), a more conservative update strategy is adopted, enhancing the robustness and flexibility of the model when dealing with the non-linear and time-continuous characteristics of the liquid chromatography instrument.

[0162] Furthermore, a high-order neural network classification algorithm combined with fractional calculus for gradient correction is adopted. In model training, the historical gradient is corrected by fractional order to enhance the convergence stability and suppress overfitting, enabling the network to remember the historical gradient information and performing excellently when there are feature offsets caused by equipment aging, avoiding the instability caused by relying only on the instantaneous gradient in the traditional gradient descent method and showing higher accuracy and reliability when monitoring the complex state changes of the liquid chromatography instrument.

[0163] Furthermore, referring to Figure 3 , it shows the performance of the autoencoder based on geometric similarity measure (GS-AE) and the traditional principal component analysis (PCA) algorithm in the embodiments of the present invention under different hyperparameter settings. Hyperparameters such as the geometric similarity weight coefficient, feature learning rate, etc. may have a greater impact on the dimensionality reduction result. In order to evaluate the performance stability and superiority of GS-AE under different hyperparameter settings, GS-AE can better retain the global structure and local features of the liquid chromatography instrument data under different hyperparameter conditions. While PCA shows greater sensitivity to hyperparameter changes and has a larger reconstruction error after dimensionality reduction. The experimental results show that GS-AE can obtain a lower reconstruction error during the hyperparameter optimization process, demonstrating its advantage in maintaining the key information of the data.

[0164] Furthermore, Figure 4Shows the accuracy rates of various common dimensionality reduction algorithms such as comparative GS-AE, PCA, t-SNE, and Locally Linear Embedding (LLE) when performing classification tasks after dimensionality reduction of liquid chromatograph data. By verifying the effects of different dimensionality reduction algorithms in the data classification tasks after dimensionality reduction, it is indicated that GS-AE can retain more useful information during the dimensionality reduction process of complex liquid chromatograph data, improving the classification accuracy. GS-AE shows the highest classification accuracy among all methods. Especially when facing complex high-dimensional data, it can better maintain the classification structure of the data. For the traditional PCA and t-SNE methods, due to over-simplifying the data or being unable to fully capture global information, the classification accuracy after dimensionality reduction is relatively low. The advantage of GS-AE lies in its ability to better preserve the detailed features in liquid chromatograph data, thus significantly improving the classification effect.

[0165] Furthermore, on the basis of the above embodiments, the embodiments of the present invention also provide a device for liquid chromatograph status monitoring and predictive maintenance. Figure 5 Shows a schematic structural diagram of a device for liquid chromatograph status monitoring and predictive maintenance provided by the embodiments of the present invention, as Figure 5 shown. The device includes: a data acquisition module 100 for acquiring the operation monitoring data of the liquid chromatograph to be monitored; a calculation module 200 for performing distance measurement on the data points of the operation monitoring data to determine the geometric similarity of the data points corresponding to the operation monitoring data; a data processing module 300 for performing dimensionality reduction processing on the operation monitoring data based on the geometric similarity of the data points by using a pre-trained autoencoder to determine the dimensionality-reduced data representation corresponding to the operation monitoring data; an execution module 400 for loading the dimensionality-reduced data representation into a pre-trained status monitoring model and outputting the status analysis result corresponding to the dimensionality-reduced data representation; wherein, the status monitoring model updates the model parameters based on the fractional-order modified cross-entropy loss function; a predictive maintenance module 500 for evaluating the working status of the liquid chromatograph based on the status analysis result.

[0166] For the device for liquid chromatograph status monitoring and predictive maintenance provided by the embodiments of the present invention, its implementation principle and the technical effects produced are the same as those of the foregoing method embodiments. For the sake of brief description, for the parts not mentioned in the device embodiments, reference may be made to the corresponding contents in the foregoing method embodiments.

[0167] Among them, the state monitoring model is constructed based on a high-order neural network; the execution module 400 is further configured to perform forward propagation on the dimensionality-reduced data representation through multiple hidden layers of the high-order neural network; and, use the deep structure of the high-order neural network to perform layer-by-layer fusion on the dimensionality-reduced data representation to determine the state analysis result corresponding to the dimensionality-reduced data representation. The execution module 400 is further configured to train a preset initial high-order neural network using a preset training sample set and determine the cross-entropy loss function corresponding to the initial high-order neural network; use a preset fractional calculus operator to perform fractional correction on the cross-entropy loss function to determine the fractional loss gradient; based on the fractional loss gradient and the cross-entropy loss function, perform backpropagation update on the model parameters of the initial high-order neural network; until the initial high-order neural network meets the preset iteration condition, construct a state monitoring model based on the initial high-order neural network.

[0168] The calculation module 200 is further configured to measure the data point distance of the operation monitoring data based on the per-dimensional difference of the data points of the operation monitoring data; calculate the geometric similarity matrix corresponding to the operation monitoring data based on the data point distance and a preset similarity weight coefficient, and determine the data point geometric similarity corresponding to the operation monitoring data.

[0169] The data processing module 300 is further configured to perform weighted mapping on the operation monitoring data using a preset autoencoder based on the data point geometric similarity to determine the initial dimensionality-reduced features; measure the interaction of the data points of the initial dimensionality-reduced features, and determine the feature mapping from the initial dimensionality-reduced features; perform adaptive smoothing processing on the feature mapping to obtain the dimensionality-reduced data representation corresponding to the operation monitoring data. The data processing module 300 is further configured to determine the gradient change corresponding to the data points of the initial dimensionality-reduced features and the geometric distance between the data points; based on the gradient change and the geometric distance, determine the interaction energy of the data points of the initial dimensionality-reduced features; minimize the interaction energy to determine the feature mapping corresponding to the initial dimensionality-reduced features. The data processing module 300 is further configured to train a preset autoencoder using a preset training sample set and calculate the loss function of the autoencoder; introduce a constraint condition of minimizing the path difference to the loss function and calculate the gradient of the loss function with respect to the weights of the autoencoder; based on the gradient and a pre-calculated adaptive weight adjustment coefficient, perform parameter update on the autoencoder. The data processing module 300 is further configured to measure the local density of each data point of the operation monitoring data within its neighborhood and dynamically calculate the adaptive weight adjustment coefficient.

[0170] Furthermore, an embodiment of the present invention further provides a liquid chromatography instrument state monitoring and predictive maintenance system, which is configured with the device of the above embodiment and is used to execute Figures 1 to 2 the method of any one of the embodiments. Figure 6The logic schematic diagram corresponding to the embodiment of the present invention is shown. Among them, the liquid chromatography instrument status monitoring and predictive maintenance system ensures the stable operation of the liquid chromatography instrument, improves the maintenance efficiency, and reduces the risk of unplanned downtime through the collaborative work of multiple functional units such as data acquisition, feature extraction, model training, predictive analysis, and maintenance suggestions.

[0171] The embodiment of the present invention further provides an electronic device, including a memory, a processor, and a computer program stored on the memory and executable on the processor. When the processor executes the computer program, the steps of the method shown above are implemented. Figures 1 to 2 The embodiment of the present invention further provides a computer-readable storage medium. A computer program is stored on the computer-readable storage medium. When the computer program is run by the processor, the steps of the method shown above are executed. Figures 1 to 2 The embodiment of the present invention further provides a structural schematic diagram of an electronic device. As shown in Figure 7 shown, it is the structural schematic diagram of the electronic device. Among them, the electronic device includes a processor 71 and a memory 70. The memory 70 stores computer-executable instructions that can be executed by the processor 71. The processor 71 executes the computer-executable instructions to implement the method shown above. Figures 1 to 3 In Figure 7In the illustrated embodiment, the electronic device further includes a bus 72 and a communication interface 73. Among them, the processor 71, the communication interface 73, and the memory 70 are connected through the bus 72. The memory 70 may include a high-speed random access memory (RAM), and may also include a non-volatile memory, such as at least one disk memory. Through at least one communication interface 73 (which can be wired or wireless), a communication connection is established between this system network element and at least one other network element, and the Internet, wide area network, local area network, metropolitan area network, etc. can be used. The bus 72 can be an ISA (Industry Standard Architecture) bus, a PCI (Peripheral Component Interconnect) bus, an EISA (Extended Industry Standard Architecture) bus, etc., and can also be an AMBA (Advanced Microcontroller Bus Architecture) bus. Among them, AMBA defines three types of buses, including an APB (Advanced Peripheral Bus) bus, an AHB (Advanced High-performance Bus) bus, and an AXI (Advanced eXtensible Interface) bus. The bus 72 can be divided into an address bus, a data bus, a control bus, etc. For the sake of convenience of representation, Figure 7 only a two-way arrow is used in the figure, but it does not mean that there is only one bus or one type of bus.

[0172] The processor 71 may be an integrated circuit chip with the ability to process signals. In the implementation process, each step of the above method can be completed by the integrated logic circuit of the hardware in the processor 71 or the instructions in the form of software. The above-mentioned processor 71 may be a general-purpose processor, including a central processing unit (CPU for short), a network processor (NP for short), etc.; it may also be a digital signal processor (DSP for short), an application specific integrated circuit (ASIC for short), a field-programmable gate array (FPGA for short), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components. The general-purpose processor may be a microprocessor or the processor may also be any conventional processor, etc. The steps of the method disclosed in combination with the embodiments of the present application can be directly embodied as being executed and completed by the hardware decoding processor, or executed and completed by the combination of the hardware and software modules in the decoding processor. The software module may be located in a mature storage medium in the art such as a random access memory, a flash memory, a read-only memory, a programmable read-only memory, or an electrically erasable programmable memory, a register, etc. This storage medium is located in the memory, and the processor 71 reads the information in the memory and combines its hardware to complete the foregoing Figures 1 to 2 any of the methods shown.

[0173] A computer program product of a method and system for state monitoring and predictive maintenance of a liquid chromatograph provided by an embodiment of the present invention includes a computer-readable storage medium storing program codes. The instructions included in the program codes can be used to execute the method described in the foregoing method embodiments. For the specific implementation, reference can be made to the method embodiments and will not be elaborated here. Those skilled in the art can clearly understand that for the convenience and brevity of description, the specific working process of the system described above can refer to the corresponding process in the foregoing method embodiments and will not be elaborated here. In addition, in the description of the embodiments of the present invention, unless otherwise clearly defined and limited, the terms "install", "connect", and "couple" should be understood in a broad sense. For example, it can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be directly connected or indirectly connected through an intermediate medium, and it can be the internal communication of two components. For those skilled in the art, the specific meanings of the above terms in the present invention can be understood according to specific situations. If the function is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on such an understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or a part of this technical solution, can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes several instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods described in various embodiments of the present invention. The foregoing storage medium includes: various media such as USB flash drives, mobile hard disks, read-only memories (ROMs), random access memories (RAMs), magnetic disks, or optical discs that can store program codes. In the description of the present invention, it should be noted that the orientation or positional relationship indicated by the terms "center", "upper", "lower", "left", "right", "vertical", "horizontal", "inner", "outer", etc. is based on the orientation or positional relationship shown in the drawings, and is only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be understood as a limitation to the present invention. In addition, the terms "first", "second", and "third" are only used for descriptive purposes and cannot be understood as indicating or implying relative importance.Finally, it should be noted that the above embodiments are only specific embodiments of the present invention, used to illustrate the technical solutions of the present invention, rather than limiting it. The protection scope of the present invention is not limited thereto. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that any person skilled in the art within the technical scope disclosed by the present invention can still modify the technical solutions recorded in the foregoing embodiments or can easily think of changes, or perform equivalent replacements for some of the technical features; and these modifications, changes or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present invention, and should all be covered within the protection scope of the present invention. Therefore, the protection scope of the present invention should be subject to the protection scope of the claims.

Claims

1. A method for condition monitoring and predictive maintenance of a liquid chromatograph, characterized in that: The method comprises: Acquire operation monitoring data of the liquid chromatograph to be monitored; the operation monitoring data includes equipment operation parameters of the liquid chromatograph and status operation information in the log of the liquid chromatograph; Performing distance measurement on the data points of the operation monitoring data to determine the geometric similarity of the data points corresponding to the operation monitoring data; Using a pre-trained autoencoder to perform dimensionality reduction processing on the operation monitoring data based on the geometric similarity of the data points, and determining a dimensionality reduction data representation corresponding to the operation monitoring data; Loading the reduced-dimensional data representation into a pre-trained state monitoring model, and outputting a state analysis result corresponding to the reduced-dimensional data representation; wherein the state monitoring model updates model parameters based on a fractional-order corrected cross entropy loss function; Based on the status analysis result, evaluating the working status of the liquid chromatograph; The step of performing dimensionality reduction processing on the operation monitoring data based on the geometric similarity of the data points using a pre-trained autoencoder to determine the dimensionality reduction data representation corresponding to the operation monitoring data comprises: Based on the geometric similarity of the data points, a preset autoencoder is used to perform weighted mapping on the operation monitoring data to determine an initial dimensionality reduction feature; performing interaction measurement on data points of the initial reduced-dimensionality features and determining a feature map from the initial reduced-dimensionality features; Adaptively smoothing the feature map to obtain a dimension-reduced data representation corresponding to the operation monitoring data; The condition monitoring model is constructed based on a high-order neural network; the method for constructing the condition monitoring model includes: A preset initial high-order neural network is trained using a preset training sample set, and a cross-entropy loss function corresponding to the initial high-order neural network is determined; a fractional-order correction is performed on the cross-entropy loss function using a preset fractional-order calculus operator to determine a fractional-order loss gradient; based on the fractional-order loss gradient and the cross-entropy loss function, the model parameters of the initial high-order neural network are back-propagated and updated; until the initial high-order neural network meets the preset iteration conditions, a state monitoring model is constructed based on the initial high-order neural network.

2. The method according to claim 1, characterized in that The step of loading the reduced-dimensional data representation into a pre-trained state monitoring model and outputting a state analysis result corresponding to the reduced-dimensional data representation comprises: The reduced-dimensional data representation is forward-propagated through multiple hidden layers of the high-order neural network; and the reduced-dimensional data representation is layer-by-layer fused using the deep structure of the high-order neural network to determine a state analysis result corresponding to the reduced-dimensional data representation.

3. The method according to claim 1, characterized in that The step of performing distance measurement on the data points of the operation monitoring data to determine the geometric similarity of the data points corresponding to the operation monitoring data comprises: Measuring the distance between data points of the operation monitoring data based on the dimension-by-dimensional difference between the data points of the operation monitoring data; Based on the data point distance and a preset similarity weight coefficient, a geometric similarity matrix corresponding to the operation monitoring data is calculated to determine the geometric similarity of the data points corresponding to the operation monitoring data.

4. The method according to claim 1, characterized in that: The step of performing interaction measurement on the data points of the initial dimensionality reduction features and determining a feature map from the initial dimensionality reduction features comprises: Determine the gradient change corresponding to the data points of the initial dimension reduction feature, and the geometric distance between the data points; Determining the interaction energy of the data points of the initial dimensionality reduction feature based on the gradient change and the geometric distance; The interaction energy is minimized to determine a feature map corresponding to the initial dimensionality reduction feature.

5. The method according to claim 1, characterized in that The training method of the autoencoder comprises: Training a preset autoencoder using a preset training sample set, and calculating a loss function of the autoencoder; Introducing a constraint condition of minimizing the path difference into the loss function, and calculating the gradient of the loss function with respect to the weight of the autoencoder; Based on the gradient and the pre-calculated adaptive weight adjustment coefficient, the parameters of the autoencoder are updated.

6. The method according to claim 5, characterized in that The method for calculating the adaptive weight adjustment coefficient includes: The local density of each data point of the operation monitoring data in its neighborhood is measured, and an adaptive weight adjustment coefficient is dynamically calculated.

7. A liquid chromatograph condition monitoring and predictive maintenance device, characterized in that: The device comprises: A data acquisition module, used to obtain operation monitoring data of the liquid chromatograph to be monitored; the operation monitoring data includes equipment operation parameters of the liquid chromatograph and status operation information in the log of the liquid chromatograph; A calculation module, used to perform distance measurement on the data points of the operation monitoring data to determine the geometric similarity of the data points corresponding to the operation monitoring data; A data processing module, configured to perform dimensionality reduction processing on the operation monitoring data based on the geometric similarity of the data points using a pre-trained autoencoder, and determine a dimensionality reduction data representation corresponding to the operation monitoring data; An execution module, used for loading the reduced-dimensional data representation into a pre-trained state monitoring model, and outputting a state analysis result corresponding to the reduced-dimensional data representation; wherein the state monitoring model updates model parameters based on a fractional-order corrected cross entropy loss function; A predictive maintenance module, used for evaluating the working status of the liquid chromatograph based on the status analysis result; The data processing module is further used to: perform weighted mapping on the operation monitoring data based on the geometric similarity of the data points using a preset autoencoder to determine an initial dimensionality reduction feature; perform interaction measurement on the data points of the initial dimensionality reduction feature to determine a feature map from the initial dimensionality reduction feature; perform adaptive smoothing on the feature map to obtain a dimensionality reduction data representation corresponding to the operation monitoring data; The condition monitoring model is constructed based on a high-order neural network; the execution module is also used to: train a preset initial high-order neural network using a preset training sample set, and determine the cross-entropy loss function corresponding to the initial high-order neural network; perform fractional-order correction on the cross-entropy loss function using a preset fractional-order calculus operator, and determine the fractional-order loss gradient; based on the fractional-order loss gradient and the cross-entropy loss function, back-propagate and update the model parameters of the initial high-order neural network; until the initial high-order neural network meets the preset iteration conditions, the condition monitoring model is constructed based on the initial high-order neural network.

8. A liquid chromatograph condition monitoring and predictive maintenance system, characterized in that: The system is configured with the apparatus of claim 7, and is used to execute the method of any one of claims 1 to 6.

Images