Intelligent control method and system for automatic sample injector of liquid chromatograph

Abstract

The present invention provides an intelligent control method and system for an automatic sampler of a liquid chromatograph, which relates to the technical field of data processing, performs data monitoring on a sample injection chamber of the automatic sampler of a liquid chromatograph, performs quantum chaos mapping on monitoring parameters, and then performs data analysis, so as to drive and control the automatic sampler based on the obtained state analysis results. The state analysis results include the intervention state of the sample bottle, the placement state, the type of the sample bottle, the sample label state of the sample bottle, and the temperature distribution state of the hole position of the sample bottle. The present invention can align the monitoring data of multiple tasks in the feature space, capture the dynamic global and local associations existing in the monitoring data, and output accurate state analysis results based on the influence of the monitoring data of the remaining tasks for any monitoring task, thereby improving the analysis accuracy.

Description

Technical Field

[0001] The invention relates to the technical field of data processing, and in particular to an intelligent control method and system for an automatic sample injector of a liquid chromatograph. Background Art

[0002] In the field of liquid chromatography analysis, the autosampler is one of the key equipment, and its performance directly affects the accuracy and efficiency of sample analysis. However, there are many problems in the operation of existing autosamplers, including sample placement, temperature control management, manual intervention detection, and sample status identification, which can lead to sample extraction failure, increased analysis errors, and even equipment damage. Specifically, the placement of the sample bottle, whether the label is correct, the uniformity of the temperature distribution in the sample chamber, and the real-time detection of manual intervention all place high demands on the reliability and safety of the autosampler.

[0003] In data-driven control systems, the training quality of machine learning models directly depends on the scale and diversity of sample data. However, due to the difficulty of collection, uneven distribution and scarcity of samples, the limitations of insufficient training data in existing technologies have seriously affected the generalization ability of the model, especially when dealing with high-dimensional complex data and abnormal samples. In addition, traditional nonlinear feature extraction algorithms are difficult to fully capture the complex dynamic relationships in the injector data, resulting in low data analysis accuracy. Summary of the invention

[0004] In view of this, the object of the present invention is to provide a liquid chromatograph automatic sample injector intelligent control method and system, capture the global and local correlation relationship of monitoring parameters, and realize more intelligent automatic sample injector control.

[0005] In a first aspect, an embodiment of the present invention provides an intelligent control method for an automatic sampler of a liquid chromatograph, wherein the method comprises: performing data monitoring on an injection chamber of the automatic sampler of the liquid chromatograph to obtain monitoring parameters of a sample bottle to be sampled in the injection chamber; performing quantum chaos mapping on the monitoring parameters to obtain test data corresponding to the monitoring parameters; the quantum chaos mapping is based on superposition calculation of multiple fluctuation terms of the monitoring parameters; analyzing the test data to determine a state analysis result corresponding to the sample bottle; wherein the state analysis result of the sample bottle includes an intervention state, a placement position state, a type of the sample bottle, a sample label state of the sample bottle, and a hole temperature distribution state of the sample bottle; and according to the state analysis result, driving and controlling the automatic sampler to sample the sample bottle to be sampled.

[0006] In combination with the first aspect, an embodiment of the present invention provides a first implementation manner of the first aspect, wherein the step of performing quantum chaotic mapping on the monitoring parameters to obtain the data to be tested corresponding to the monitoring parameters includes: standardizing the monitoring parameters to obtain initial data to be tested; performing quantum mapping processing on the initial data to be tested to obtain quantum mapping results corresponding to the initial data to be tested; and performing convolution calculation on the initial data to be tested using a preset machine learning unit; performing chaotic mapping on the quantum mapping results and the convolution calculation to obtain the data to be tested corresponding to the monitoring parameters.

[0007] In combination with the first aspect, an embodiment of the present invention provides a second implementation of the first aspect, wherein the step of analyzing the data to be tested and determining the state analysis result corresponding to the sample bottle includes: using a preset machine learning unit to perform a multi-scale pooling operation on the data to be tested, performing quantum chaos pooling on the pooled data to be tested, and obtaining the target parameters corresponding to the data to be tested; calculating the class probability corresponding to the target parameter, and determining the state analysis result corresponding to the sample bottle based on the class probability.

[0008] In combination with the first aspect, an embodiment of the present invention provides a third implementation of the first aspect, wherein the weights of the machine learning unit are initialized based on a preset quantum mapping function; wherein the quantum mapping function includes multiple nonlinear quantum chaos terms, and the multiple nonlinear quantum chaos terms include quantum chaos feedback terms and quantum oscillation control terms.

[0009] In combination with the first aspect, an embodiment of the present invention provides a fourth implementation of the first aspect, wherein a method for constructing a machine learning unit includes: training the machine learning unit using a preset training sample set to calculate an adaptive loss function corresponding to the training sample set; determining a cosine loss term corresponding to the adaptive loss function; and optimizing the learning rate of the machine learning unit based on the cosine loss term to update the weight of the machine learning unit.

[0010] In combination with the first aspect, an embodiment of the present invention provides a fifth implementation of the first aspect, wherein a method for constructing a training sample set includes: obtaining pre-collected sample bottle status monitoring samples, annotating the sample bottle status monitoring samples, and constructing an initial training sample set; based on the normal distribution of the initial training sample set, generating an adaptive noise vector corresponding to the initial training sample set; and performing data expansion on the initial training sample set based on the adaptive noise vector to construct a training sample set.

[0011] In combination with the first aspect, an embodiment of the present invention provides a sixth implementation of the first aspect, wherein the step of performing data expansion on the initial training sample set based on an adaptive noise vector to construct a training sample set includes: performing data expansion on the initial training sample set based on an adaptive noise vector through a pre-constructed generative adversarial network to construct a training sample set.

[0012] In combination with the first aspect, an embodiment of the present invention provides a seventh implementation of the first aspect, wherein the method also includes: the generative adversarial network trains the generator and the discriminator of the generative adversarial network based on a preset optimization algorithm; wherein the preset optimization algorithm includes at least one of the following optimization parameters: a dynamic adjustment factor of weights, a learning rate adjustment factor based on an adaptive gradient, and an adaptive incremental learning comprehensive loss function.

[0013] In a second aspect, an embodiment of the present invention provides an intelligent control device for an automatic sampler of a liquid chromatograph, wherein the device includes: a data acquisition module, used to perform data monitoring on an injection chamber of the automatic sampler of the liquid chromatograph, and obtain monitoring parameters of a sample bottle to be sampled in the injection chamber; a data processing module, used to perform quantum chaos mapping on the monitoring parameters, and obtain test data corresponding to the monitoring parameters; the quantum chaos mapping is based on the superposition calculation of multiple fluctuation terms of the monitoring parameters; an execution module, used to analyze the test data through a preset machine learning unit, and determine a state analysis result corresponding to the sample bottle; wherein the state analysis result of the sample bottle includes the intervention state of the sample bottle, the placement position state, the type of the sample bottle, the sample label state of the sample bottle, and the hole temperature distribution state of the sample bottle; a control module, used to drive and control the automatic sampler according to the state analysis result to sample the sample bottle to be sampled.

[0014] In a third aspect, an embodiment of the present invention provides an intelligent control system for an automatic sample injector of a liquid chromatograph, wherein the system is configured with the apparatus of the above embodiment 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:

[0016] The embodiment of the present invention provides a liquid chromatograph automatic sample injector intelligent control method and system, which can align the monitoring data of multiple tasks in the feature space through quantum chaos mapping of monitoring parameters, so that the knowledge learned from one task can be more easily transferred to another new task that is related but not completely the same, and the dynamic global and local correlations existing in the monitoring data can be captured. It can better capture the global and local nonlinear patterns of the data, such as the subtle coupling between temperature fluctuations and displacement offsets, and can enhance the model's perception of complex monitoring data. Based on this, no matter what kind of monitoring task is targeted, it can output accurate state analysis results based on the influence of the monitoring data of other tasks, thereby improving the analysis accuracy.

[0017] Other features and advantages of the present invention will be described in the following description, and partly become apparent from the description, or understood by practicing the present invention. The purpose and other advantages of the present invention are realized and obtained by the structures particularly pointed out in the description, claims and drawings.

[0018] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, preferred embodiments are given below and described in detail with reference to the accompanying drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] In order to more clearly illustrate the specific implementation methods of the present invention or the technical solutions in the prior art, the drawings required for use in the specific implementation methods or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are some implementation methods of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0020] Figure 1 A schematic diagram of the structure of a liquid chromatograph provided in an embodiment of the present invention;

[0021] Figure 2 A schematic diagram of the structure of an automatic sample injector for a liquid chromatograph provided in an embodiment of the present invention;

[0022] Figure 3 A flow chart of an intelligent control method for an automatic sample injector of a liquid chromatograph provided by an embodiment of the present invention;

[0023] Figure 4 A schematic diagram of intelligent control logic of an automatic sample injector provided by an embodiment of the present invention;

[0024] Figure 5 A flow chart of another liquid chromatograph automatic sample injector intelligent control method provided by an embodiment of the present invention;

[0025] Figure 6 A schematic diagram of the structure of an intelligent control device for an automatic sample injector of a liquid chromatograph provided in an embodiment of the present invention;

[0026] Figure 7 A schematic structural diagram of an electronic device provided by an embodiment of the present invention is shown. DETAILED DESCRIPTION

[0027] In order to make the purpose, technical solution and advantages of the embodiments of the present invention clearer, the technical solution of the present invention will be clearly and completely described in combination with the embodiments below. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0028] The present invention provides an intelligent control method and system for an automatic sample injector of a liquid chromatograph, which can realize accurate identification even when faced with high-dimensional complex data and abnormal samples corresponding to the sample bottle status of the liquid chromatograph, thereby significantly improving the reliability and safety of the automatic sample injector.

[0029] Liquid chromatograph is an analytical instrument used to separate, identify and quantify the components in a mixture. It uses the difference in the distribution of different substances between the stationary phase and the mobile phase to decompose complex mixtures into single components through column chromatography. Separation: Separate the components in a complex mixture. Identification: Determine the identity of each component through a detector. Quantitative analysis: Measure the concentration or content of each component.

[0030] Liquid chromatographs are based on "chromatography" and use the difference in distribution coefficients between the stationary phase and the mobile phase to separate different substances. The mobile phase (usually a solvent) is delivered to the system at a constant pressure and flow rate through an infusion pump. The sample is injected into the mobile phase through an automatic sampler. The sample enters the chromatographic column with the mobile phase and is separated under the action of the stationary phase. The separated components pass through the detector in turn, and the detector converts the signal into an electrical signal and records it. Furthermore, the detected signal is processed and analyzed by the data processing system to generate a report. Figure 1 Shows a schematic diagram of the structure of a liquid chromatograph, referring to Figure 1 , from top to bottom are the infusion pump, automatic sampler, chromatographic column and column oven, and detector.

[0031] Infusion Pump 10:

[0032] The infusion pump is one of the core components of the liquid chromatograph. Its main function is to deliver the mobile phase (usually the solvent) into the chromatographic system at a constant pressure and flow rate. The infusion pump usually adopts a high-pressure pump, which can provide a pressure of hundreds or even thousands of atmospheres to ensure that the mobile phase can pass through the chromatographic column smoothly and the flow rate is stable.

[0033] Autosampler 20:

[0034] An autosampler is used to automatically inject samples into the mobile phase. It usually consists of a sample tray and an injection needle. The sample tray can hold multiple sample bottles, and the injection needle is responsible for extracting the sample from the sample bottle and injecting it into the mobile phase. An autosampler can greatly improve analysis efficiency and reduce human errors.

[0035] Chromatographic column and column oven 30:

[0036] The chromatographic column is a key component of a liquid chromatograph. It contains a stationary phase, which is used to separate different components in a mixture. The stationary phase is usually a granular substance, such as silica gel, alumina, etc. When the mobile phase passes through the chromatographic column, different components in the mixture will interact with the stationary phase differently, thereby achieving separation. The column oven is used to control the temperature of the chromatographic column to optimize the separation effect.

[0037] Detector 40:

[0038] The detector is used to detect the components flowing out of the chromatographic column. Common detectors include UV-visible light detectors, fluorescence detectors, electrochemical detectors, etc. The detector converts the detected signal into an electrical signal and transmits it to the data processing system for analysis.

[0039] Among them, there are many problems in the operation of existing automatic samplers, including sample placement, temperature control management, manual intervention detection and sample status identification, which can lead to sample extraction failure, increased analysis errors and even equipment damage. Specifically, the placement of the sample bottle, whether the label is correct, the uniformity of the temperature distribution in the sample chamber, and the real-time detection of manual intervention all put forward high requirements on the reliability and safety of the automatic sampler.

[0040] The embodiment of the present invention collects data from the automatic sampler to intelligently control the automatic sampler, thereby improving the reliability and safety of the automatic sampler and avoiding equipment failure problems caused by sample extraction. Figure 2The schematic diagram of the structure of the automatic sampler is shown, including an XYZ sampling mechanism, a power supply, a control unit, a sample tray unit, a syringe pump or a metering pump unit, and an injection valve. These components work together to enable the automatic sampler to efficiently and accurately complete the sampling and injection process of the sample. Power supply 20: Provides power support for the entire system to ensure that all components can work properly. XYZ sampling mechanism 21: This part is responsible for moving in three-dimensional space to accurately locate the sample. It usually includes moving components of the X-axis, Y-axis and Z-axis, which can achieve accurate positioning and sampling of the sample. Sample tray unit 22: Used to place the sample to be analyzed. It can be a rotating disk or other form of sample carrying device to facilitate the sampling operation of the XYZ sampling mechanism. Among them, the sample tray unit has a hole position for placing each sample bottle. Each hole position has a clear number or mark to ensure that the sample bottle can be accurately located and identified. Among them, the design of the hole position ensures that the sample bottle can be placed firmly and the XYZ sampling mechanism can accurately access the sample bottle in each hole position. Control unit 23: It is the "brain" of the system, responsible for coordinating and controlling the work of all other components. It receives instructions and sends signals to control the operation of the XYZ sampling mechanism, syringe pump or metering pump unit, etc. Syringe pump or metering pump unit 24: used to inject the sample into the analytical instrument. It can be a syringe pump (disposable) or a metering pump (reusable), and the appropriate type is selected according to the specific application requirements. Injection valve 25: controls the passage of the sample into the analytical instrument. It can be opened or closed when needed to ensure that the sample enters the analytical instrument accurately.

[0041] During the experiment, the sample bottle may be subject to human or mechanical intervention (such as moving, opening, closing, etc.), which may cause sample contamination, leakage or other abnormal conditions. Incorrect placement or omission of the sample bottle will cause the experimental process to be interrupted or the results to be inaccurate. Different types of samples require different treatment methods. If the type of sample bottle in the corresponding position is wrong, improper treatment may result. The state of the sample (such as liquid level, color, turbidity, etc.) may affect the experimental results. In some application scenarios (such as high performance liquid chromatography), it is very important to keep the temperature of the sample bottle stable to prevent the sample from degrading or other chemical changes due to temperature changes. In response to these potential problems, the embodiment of the present invention monitors the intervention state, placement position, type, temperature distribution and sample state of the sample bottle in the injection chamber of the automatic sampler in real time, and accurately identifies each state to achieve accurate intelligent control to ensure that each injection meets the requirements and can timely and accurately avoid errors caused by human factors or environmental changes.

[0042] To facilitate understanding of this embodiment, firstly, an intelligent control method for an automatic sample injector of a liquid chromatograph disclosed in an embodiment of the present invention is described. Figure 3A flow chart of an intelligent control method for an automatic sample injector of a liquid chromatograph provided by an embodiment of the present invention is shown, referring to Figure 3 , including the following steps:

[0043] Step S102, monitoring the data of the sample injection chamber of the automatic sample injector of the liquid chromatograph to obtain monitoring parameters of the sample bottle to be sampled in the sample injection chamber.

[0044] In one embodiment, the data of the sample chamber is monitored by setting a corresponding data acquisition module to capture the physical information in the sample chamber. To ensure that the sample bottle meets the preset task, the position information, status and whether there is external interference of the sample bottle can be monitored. In one embodiment, the position of the sample bottle can be detected by an infrared detection module to provide data to the subsequent analysis module; it can also detect whether there is human intervention in the sample chamber (such as a hand reaching into the sample chamber). At the same time, the temperature distribution data of the hole position can be collected to provide a basis for uniformity analysis. Among them, it can be determined whether there is human intervention by sensing the movement of objects in the sample chamber.

[0045] Step S104, performing quantum chaos mapping on the monitoring parameters to obtain the data to be measured corresponding to the monitoring parameters.

[0046] Step S106, analyzing the data to be tested to determine the state analysis result corresponding to the sample bottle.

[0047] By processing and analyzing the monitoring parameters, the state analysis results characterized by the monitoring parameters are determined. Specifically, different monitoring tasks correspond to different monitoring results, such as whether a person's hand reaches into the sample chamber to determine whether there is human intervention; sample bottle hole coordinate correction to confirm whether the sample bottle is placed correctly; sample bottle type analysis to identify the sample bottle model and material and match the corresponding injection procedure; sample label position and accuracy verification to detect whether the label meets the specifications and is correctly attached; temperature uniformity analysis of different hole positions to determine whether the temperature distribution in the sample chamber is uniform.

[0048] In the intelligent control task of the liquid chromatograph autosampler, the data monitored by the autosampler involves different monitoring types and has complex patterns. The existing technology usually performs corresponding analysis for each monitoring task, and it is easy to ignore the global and local correlation when processing high-dimensional and non-uniformly distributed features, resulting in insufficient feature extraction. In an embodiment of the present invention, by quantum chaos mapping of monitoring parameters, the monitoring data of multiple tasks can be aligned in the feature space, making it easier to transfer the knowledge learned from one task to another related but not completely the same new task, capturing the dynamic global and local correlations existing in the monitoring data, and better capturing the global and local nonlinear patterns of the data, such as the subtle coupling between temperature fluctuations and displacement offsets, which can enhance the model's perception of complex monitoring data.

[0049] The quantum chaos mapping of the embodiment of the present invention is based on the superposition calculation of multiple fluctuation items of the monitoring parameters, and the state analysis results of the sample bottle include the intervention state of the sample bottle, the placement state, the type of the sample bottle, the sample label state of the sample bottle, and the hole temperature distribution state of the sample bottle. The monitoring parameters of the embodiment of the present invention include data of multiple tasks obtained by comprehensive monitoring of the sample bottles, such as type, temperature, label, interference, etc., and there is a dynamic relationship between local or global features of each task data. Among them, multiple fluctuation items include sine and cosine fluctuations corresponding to the monitoring parameters, and the combination of the two is used to represent the nonlinear characteristics of the monitoring data and its dynamic relationship, which can enhance the extraction of nonlinear dynamic relationships of complex monitoring data. The embodiment of the present invention combines and transforms multiple fluctuation items corresponding to the monitoring data of multiple tasks to generate a new data representation with complex dynamic characteristics, which can not only capture the interaction between each fluctuation item, but also introduce nonlinearity and randomness, enhance the recognition of the diversity and complexity of the data, and capture the dynamic relationship of the data.

[0050] Through the above processing, no matter which monitoring task is performed, accurate status analysis results can be output based on the influence of the monitoring data of other tasks, thereby improving the analysis accuracy.

[0051] Step S108, according to the state analysis result, the automatic sampler is driven and controlled to sample the sample bottle to be sampled.

[0052] The status analysis results include the results corresponding to each monitoring task. For the intervention status of the sample bottle, the embodiment of the present invention determines it by detecting whether a human hand has been inserted into the sample chamber. The intervention detection result includes intervention or no intervention. If human intervention is detected, safety protection measures are activated, mechanical action is immediately suspended, an alarm is issued, and re-testing is performed after the intervention is completed to ensure safety; if no human intervention is detected, the sample injection process operates normally.

[0053] For the placement status of the sample bottle, the corresponding sample bottle at the hole coordinates of the sample tray unit is monitored to analyze the placement status of the sample bottle, including the correct state and the wrong state. If the placement position is correct, the sample extraction process is directly entered; if the placement position is wrong, the user can be notified to adjust the position of the sample bottle through the preset warning unit, and the system will pause the operation and wait for the user to adjust and re-test. Among them, when the placement position is correct, the drive module can be driven and controlled based on the injection task instruction to drive the bearings of different hole positions according to the sample position and task requirements, and accurately move the sampling needle or sample bottle to the task position.

[0054] For the sample bottle type, the machine learning unit analyzes the sample bottle type in the injection chamber, such as size, shape, etc. Then, the corresponding injection needle parameters, such as depth and speed, are selected according to the determined sample bottle type. If the sample bottle type does not match the task requirements, the preset notification module can be used to notify the user to change the sample bottle, suspend the injection process, and wait for the user to adjust.

[0055] The sample label status of the sample bottle is identified by whether the sample label of the sample bottle is correct. The status analysis results of the sample label include correct, wrong, and missing. If the label is correct, continue to perform the injection task; if the label is wrong or missing, the notification module can notify the user to check and re-attach the label, suspend the injection operation, and re-test after the label is correct.

[0056] According to the temperature distribution state of the sample bottle's hole positions, the temperature uniformity of different hole positions is analyzed to determine the temperature distribution corresponding to each hole position, including uniform or uneven. If the temperature distribution is uniform, proceed to the next step of injection operation. If the temperature is uneven, the preset control strategy can be used to adjust the heating or cooling equipment to balance the temperature of each hole position. Continue the operation after the detection temperature returns to normal. In the specific implementation, the area where the sample bottle hole position is located is equipped with a temperature control device, such as a constant temperature box or a heating system, a cooling system, so that the hole position maintains a stable temperature environment. For example, call the PID algorithm for control, the input is the temperature distribution data and the target temperature difference, according to the temperature control requirements, adjust the output power of the heating or cooling module, and dynamically adjust until the temperature difference is controlled within the allowable range.

[0057] In one embodiment, each of the above analysis results can be output by a preset machine learning unit, including intervention detection, sample position status, sample bottle type, sample label status, well temperature status, etc. These results can be sent by the machine learning unit to the main control processor of the liquid chromatograph. The main control processor module dynamically adjusts the control logic according to the analysis results of each state, such as prompting, pausing, sampling processing, etc. Correspondingly, Figure 4 The figure shows the intelligent control logic schematic diagram of the automatic sample injector.

[0058] In summary, the embodiments of the present invention can identify complex data patterns of monitoring parameters, capture global and local correlations between data, extract features more fully, improve analysis accuracy, and achieve more intelligent automatic sampler control.

[0059] Furthermore, based on the above embodiment, the embodiment of the present invention also provides another intelligent control method for an automatic sample injector of a liquid chromatograph. Figure 5 A flow chart of another liquid chromatograph automatic sample injector intelligent control method provided by an embodiment of the present invention is shown, referring to Figure 5 , including the following steps:

[0060] Step S202, monitoring the data of the sample injection chamber of the automatic sample injector of the liquid chromatograph to obtain monitoring parameters of the sample bottle to be sampled in the sample injection chamber.

[0061] In combination with the above embodiments, the sample status, label status, placement position and temperature control data collected by the infrared detection module through the infrared detection head, position sensor and temperature sensor may include the placement position of the sample bottle, the sample label status, the temperature of each hole position, etc. In addition, the status of the mechanical operation area is detected to record the human intervention behavior data (such as the sample compartment door is open, the hand is inserted into the sample compartment, etc.), and the characterization content includes whether human intervention occurs, the behavior type, and the trigger time. The collected sample injection process data can also be recorded through the experimental system, and the characterization content includes whether the sample extraction is successful and the status data during extraction.

[0062] The collected sensor data is in JSON format. The sample data is as follows:

[0063] {timestamp,position_x,position_y,label_status,temperature,intervention_status, extraction_success;2025-01-08T10:00:00,12,34, correct,25.3, no_intervention, success};

[0064] The format of the collected injection record data is JSON format. The sample data is as follows:

[0065] {timestamp,sample_id,sample_position,label_status,extraction_success;2025-01-08T10:05:00,S001,P1,correct,success};

[0066] The collected human intervention data is in JSON format. The sample data is as follows:

[0067] {timestamp:2025-01-8T10:00:00,behavior:hand_intervention,location:sample_chamber}.

[0068] It should be noted that this embodiment is only used to illustrate one data format and type of the present invention. In actual applications, the attributes of data are usually more than 10 attributes, and the number of attributes of data may reach dozens or even hundreds.

[0069] Step S204, performing quantum chaos mapping on the monitoring parameters to obtain the data to be measured corresponding to the monitoring parameters.

[0070] The specific implementation includes the following steps:

[0071] 1) Standardize the monitoring parameters to obtain the initial test data.

[0072] Standardizing the data can solve the problem of unstable training caused by inconsistent feature scales. The multidimensional features of the sampler data have significant scale differences, such as temperature features (20-40°C), displacement features (0-100mm) and time intervals (measured in milliseconds). Standardization normalizes different features to the same numerical range, which helps the neural network to assign reasonable weights to each dimension of features. In the feature extraction task, the neural network can balance the focus on each dimension of features, avoiding the influence of certain large numerical features (such as displacement) covering up small numerical features (such as time changes), thereby improving the stability of the model in extracting multidimensional sampler features, expressed as:

[0073]

[0074] In the formula, is the monitoring parameter input to the neural network; is the mean value of the monitoring parameters input to the neural network, is the standard deviation of the monitoring parameters input to the neural network, The results are normalized for the monitoring parameters input to the neural network.

[0075] 2) Performing quantum mapping processing on the initial data to be tested to obtain quantum mapping results corresponding to the initial data to be tested; and performing convolution calculation on the initial data to be tested using a preset machine learning unit.

[0076] 3) Perform quantum chaos mapping on the monitoring parameters to obtain the measured data corresponding to the monitoring parameters.

[0077] Among them, the embodiment of the present invention uses quantum state superposition and interference principles to strengthen the global feature learning ability of the neural network, thereby extracting complex nonlinear patterns. In the process of feature extraction of the monitoring data of the automatic sampler, the features of multiple dimensions usually have dynamic global and local associations, such as the periodic changes of the temperature control characteristics will affect the slight offset of the sample displacement, and these associations may be ignored by traditional methods in high-dimensional space. Quantum state superposition strengthens the dynamic relationship between different features through sine and cosine interference, so that the network can capture the global and local nonlinear patterns of the sampler data. For example, the subtle coupling between temperature fluctuations and displacement offsets is extracted as a high-dimensional feature map under the action of quantum state superposition to enhance the model's perception of complex sampler data, expressed as:

[0078]

[0079] In the formula, is the data feature of the sampler output after convolution, For the convolution kernel function, is the convolution operator, is the number of convolution kernels. is the quantum chaos mapping operator, is the quantum mapping function; are the parameters of the neural network, including the weight matrix of the neural network and the bias term of the neural network. Preferably, Set to 4.

[0080] The quantum chaos mapping operator adopts the implementation method of quantum interference fusion, using the principle of quantum state superposition and interference, and represents the sine and cosine relationship between features and quantum perturbations through the combination of sine and cosine fluctuations, making the network more capable of nonlinear feature extraction. Quantum chaos mapping can be well applied in some data features with time series attributes. For example, there may be a weak correlation between temperature and displacement features, and time series features may fluctuate periodically. Through quantum chaos mapping, weight initialization contains more dynamic perturbations, helping the network to capture these complex distribution relationships faster, thereby improving the accuracy of feature extraction. Among them, by superimposing different fluctuation terms (sine and cosine), the correlation between features is strengthened. For the learning of complex patterns, the dynamic relationship between global and local features can be captured. The implementation method is:

[0081]

[0082]

[0083] in, It is the intermediate process data feature of the convolution operation.

[0084] Step S206, using a preset machine learning unit to perform a multi-scale pooling operation on the data to be tested, performing quantum chaos pooling on the multi-scale pooled data to be tested, and obtaining target parameters corresponding to the data to be tested.

[0085] In the specific implementation, the convolution layer of the machine learning unit can be used to execute step S204, and the pooling layer of the machine learning unit is used to perform a pooling operation on the convolved features. In an embodiment of the present invention, by using chaotic perturbations and different scale pooling strategies in the pooling layer, the data is multi-scale fused, the features are reduced in dimension and more abundant key information is retained, and the expressive ability of the features is improved. The key information of the monitoring data of the automatic sampler is often distributed in local areas (such as abnormally high temperature or sudden displacement offset), while the global data may contain more useless features or noise. Chaotic perturbations and multi-scale pooling strategies capture features at different scales, reduce the global dimension while retaining key local information. For example, the nonlinear weight adjustment method based on chaotic perturbations makes abnormal features (such as sharp temperature control fluctuations in a short period of time) preferentially retained in the pooling process, and multi-scale pooling ensures that the temperature trend of a long time period and the fluctuation of a short time period are expressed at the same time through different window scales, so as to enhance the expression ability of the neural network for the abnormal features of the sampler, which is expressed as:

[0086]

[0087]

[0088] In the formula, is the data feature of the test data after pooling (i.e., the target parameter), is the data feature of the sampler output after convolution, is the pooling operator, is the quantum chaos pooling mapping function; are the parameters of the neural network; For the The result of pooling scale; For the Pooling windows; For the The weight coefficient of the pooling window size, is the total number of pooling scales. Preferably, the pooling operation is implemented using the maximum pooling method. Refer to the quantum chaos map operator above.

[0089] In order to enhance the nonlinear expression of the pooling layer, the quantum chaotic pooling mapping function contains a variety of chaotic perturbations, and the calculation method is expressed as:

[0090]

[0091] In the formula, is the quantum chaos pooling coefficient; is the adaptive adjustment coefficient in the pooling process and is the training parameter; is the smoothing coefficient. Preferably, Set to 0.3, Set to 0.95. Refer to the quantum chaos map operator above.

[0092] High-dimensional data usually contains complex nonlinear patterns, such as weak correlations or periodic fluctuations between variables. The chaotic perturbation of the embodiment of the present invention enhances the system's ability to process complex data by utilizing nonlinear dynamic mechanisms, and combines adaptive adjustment parameters based on a variety of nonlinear mapping items such as sine and logarithmic. The periodic behavior is captured by sine mapping, and the logarithmic mapping can smooth the data with drastic changes. These nonlinear mapping items can accurately describe the complex relationship between features, and achieve accurate modeling of high-dimensional and non-uniformly distributed features. In addition, the sine item provides periodic fluctuations, simulates complex dynamic behaviors, and avoids falling into local optimality, while the logarithmic item ensures the numerical stability of the calculation by smoothing the data input. Through chaotic perturbation, the model can more comprehensively extract the dynamic correlation of data in nonlinear complex features, thereby improving the control accuracy and operation stability of the liquid chromatograph automatic sampler.

[0093] Step S208, calculating the class probability corresponding to the target parameter, and determining the state analysis result corresponding to the sample bottle based on the class probability.

[0094] The preset Softmax function can be used to calculate the class probability of the target parameter, and the class with the largest class probability is taken as the final classification class to determine the state analysis result.

[0095] Furthermore, the weights of the machine learning unit of the embodiment of the present invention are initialized based on a preset quantum mapping function, wherein the quantum mapping function includes multiple nonlinear quantum chaos terms, and the multiple nonlinear quantum chaos terms include quantum chaos feedback terms and quantum oscillation control terms.

[0096] Among them, the machine learning unit of the embodiment of the present invention adopts a neural network model based on quantum chaos. The neural network based on quantum chaos strengthens the global and local dynamic relationship of feature extraction by utilizing quantum state superposition and interference, and retains key information and improves feature expression capabilities through a pooling strategy of chaotic perturbation and multi-scale fusion, so that the network can exhibit stronger global optimization capabilities in complex pattern learning and achieve efficient learning and stable convergence of complex injector data.

[0097] In the training phase, the structure and parameters of the neural network are defined. The neural network includes convolutional layers and pooling layers. Quantum randomness is used to provide a variety of initial states for the neural network, solving the problem that traditional initialization methods are prone to fall into local optimality. It is expressed as:

[0098]

[0099] In the formula, is the initial weight matrix of the neural network; is a quantum mapping function, which generates initial weights according to quantum chaos mapping; is the parameter of the quantum mapping function.

[0100] Among them, the quantum mapping function of the embodiment of the present invention is composed of multiple nonlinear quantum chaos terms, and generates initial weights with dynamic characteristics and nonlinear distribution through quantum state superposition, interference and chaotic disturbance. In view of the characteristics of the complex distribution, high dimension and nonlinear correlation of the data features of the sampler, the initial weights generated by the quantum chaos mapping avoid the limitations of traditional random initialization, so that the neural network can be optimized from the global perspective. Based on this, the convolutional features can be further nonlinearly enhanced, and the global and local associations between features can be dynamically constructed by using quantum superposition and chaotic disturbance. Among them, the monitoring parameters of the automatic sampler usually include features such as temperature, position coordinates, and time series. The distribution of these features may have strong non-uniformity in different dimensions. For example, there may be a weak correlation between temperature and displacement features, and the time series features may fluctuate periodically. Through quantum chaos mapping, weight initialization contains more dynamic disturbances, which helps the network capture these complex distribution relationships faster, thereby improving the accuracy of feature extraction. The calculation method is expressed as:

[0101]

[0102] In the formula, is the quantum chaos feedback coefficient, which adjusts the chaos intensity; is the first oscillation control constant, is the training parameter; is the second oscillation control constant, is the training parameter; is the adaptive adjustment coefficient; is the benchmark parameter of the first quantum system, is the training parameter; is the reference parameter of the second quantum system, is the training parameter. Preferably, Set to 0.3, Set to 0.7.

[0103] Furthermore, the machine learning unit of the embodiment of the present invention is constructed by the following steps:

[0104] 1) Use the preset training sample set to train the machine learning unit to calculate the adaptive loss function corresponding to the training sample set.

[0105] The machine learning unit can be trained using a preset training sample set to optimize the model weight. After the training sample set is processed based on the above steps S104-S206, the data can be nonlinearly activated, and the loss value can be calculated and updated according to the adaptive loss adjustment mechanism to ensure the neural network's ability to express complex patterns, while allowing the loss to be dynamically adjusted with the batch to avoid overfitting or underfitting. The nonlinear activation of the data is expressed as:

[0106]

[0107] In the formula, For the activated injector data feature, is the data feature of the injector after pooling; To take the maximum value function, characterize the maximum pooling operation.

[0108] Furthermore, an adaptive loss function is used to optimize the training results. The loss function of the neural network is dynamically adjusted. The loss weight at each update is determined by an adaptive mechanism to enhance the adaptive ability of the loss. Some features in the monitoring data of the automatic sampler (such as abnormal temperature peaks or outlier displacement points) may be the focus of the task, but these features account for a low proportion in large-scale training data and are easily ignored by traditional loss functions. Adaptive activation and loss optimization ensures the network's attention to important features by adjusting the weights of these features in real time. For example, when abnormal fluctuations in temperature control features are detected during training, the loss function will dynamically increase the weights of these features, thereby prompting the network to learn the pattern of abnormal features more accurately. The calculation method is expressed as:

[0109]

[0110] In the formula, For the neural network The loss function of the iteration is, Based on the predicted results and the true labels The calculated cross entropy loss is, is the true label of the sample, which is calculated by the preset Softmax function on the output features of the neural network; is the characteristic variance of the activated injector data, is the regularization coefficient of the neural network loss function, is the smoothing coefficient of the neural network loss function, is the gradient of the activated injector data feature, represents the differentiation with respect to the number of iterations, is the maximum number of iterations. Preferably, Set to 0.3, Set to 0.2.

[0111] 2) Determine the cosine loss term corresponding to the adaptive loss function.

[0112] 3) Optimize the learning rate of the machine learning unit based on the cosine loss term to update the weight of the machine learning unit.

[0113] The neural network uses an adaptive learning rate, which is dynamically adjusted according to the training error of each iteration and the feedback of the network. In the training process, the cosine loss term is used as a function of periodic oscillation to increase the nonlinear dynamic characteristics of weight update. The calculation method is expressed as:

[0114]

[0115] In the formula, is the feedback adjustment factor, which controls the impact of the last iterative adjustment on the current adjustment; is the adaptive learning rate of the neural network at the t-1th iteration; is the learning rate balancing factor used to balance the weight updates during error backpropagation. is the cosine loss term, which is determined by calculating the cosine of the weight matrix of the neural network.

[0116] The weights of the neural network are updated through back propagation based on the calculated loss. In the monitoring data features of the automatic sampler, some dimensions (such as temperature control and time series features) may show periodic changes. By using the cosine loss term as a periodic oscillation function in training, the network can more accurately capture the periodic change characteristics. For example, when the temperature characteristics show regular fluctuations over time, the cosine loss term helps the network better fit these periodic patterns, so that the extracted features can reflect the dynamic operation status of the sampler. The update method is expressed as:

[0117]

[0118] In the formula, For the neural network The weight after the iteration update is is the weight of the neural network after the update at the t-1th iteration, For the neural network The gradient of the loss function with respect to the weight at the iteration, For the neural network Adaptive learning rate for each iteration.

[0119] Furthermore, the above process is iterated repeatedly, and the training is stopped when the loss function converges to the set threshold, which is expressed as:

[0120]

[0121] In the formula, For the neural network The loss function of the iteration; is the loss function of the neural network at the t-1th iteration; is the convergence threshold of neural network training. Preferably, Set to 0.001.

[0122] Furthermore, the training sample set of the embodiment of the present invention is constructed by the following steps:

[0123] 1) Obtain the pre-collected sample bottle status monitoring samples, annotate the sample bottle status monitoring samples, and construct an initial training sample set;

[0124] The sample bottle status monitoring sample can refer to the above-mentioned monitoring parameters, which will not be repeated here. Furthermore, the collected data is labeled. In an embodiment of the present invention, the labeling method is manual labeling. In one embodiment, the labeling categories include different monitoring states, such as: a-Sample bottle position status: Correct (such as a1): The sample bottle is placed in the specified hole position. Error (such as a2): The sample bottle is not placed, the placement is wrong or tilted. b-Sample label status: Correct label (such as b1): The label complies with the regulations and is attached to the correct position. Wrong label (such as b2): The label position is offset or unclear. Missing label (such as b3): The sample bottle has no label. c-Well position temperature status: Normal (such as c1): Within the temperature control range (for example, ±0.5°C). Abnormal (such as c2): The temperature is out of range or the temperature difference between the wells is too large.

[0125] 2) Based on the normal distribution of the initial training sample set, an adaptive noise vector corresponding to the initial training sample set is generated;

[0126] 3) Based on the adaptive noise vector, the initial training sample set is expanded to construct a training sample set.

[0127] It is understandable that in the task of the present invention, the acquisition, labeling and preprocessing of training injector data are time-consuming and labor-intensive, and insufficient training samples can easily lead to poor model generalization ability and affect the accuracy of the model. In the intelligent control task of liquid chromatograph automatic injector, for low-frequency complex samples and outlier data, traditional methods tend to ignore the importance of these samples, resulting in the expansion of the injector data set is difficult to fully cover complex distribution samples.

[0128] In an embodiment of the present invention, for the task of expanding high-dimensional, diverse and unevenly distributed autosampler data, the present invention adopts a sample expansion algorithm based on dynamic balance to achieve diversified generation and accurate discrimination of high-dimensional, nonlinear and complex injector data, thereby improving the quality of injector data expansion and the generalization ability of the model.

[0129] In one implementation, a training sample set is constructed by performing data expansion on an initial training sample set based on an adaptive noise vector using a pre-constructed generative adversarial network.

[0130] Before training begins, the generator will analyze and encode the distribution characteristics of the original sampler data, and generate an adaptive noise vector based on the correlation between high-dimensional features and the unevenness of sample distribution. In one embodiment, the noise vector can be generated based on the normal distribution of the initial training sample set. The variance of the normal distribution is the variance of the original data, which can improve the diversity of parameter initialization and be more consistent with the characteristic distribution of the current training data, thereby generating pseudo sampler data that is more consistent with the original distribution, overcoming the lack of sample diversity and improving the fit of high-dimensional sampler data. The sample expansion method is expressed as:

[0131]

[0132] In the formula, The first injector data samples; is the generator function, The noise vector and the generator parameters jointly determine the output pseudo injector data; For the A noise vector is input to the generator, The generator parameters.

[0133] The discriminator discriminates the input injector data through a multi-layer neural network structure, which can capture subtle differences in high-dimensional injector data and enhance the ability to distinguish complex injector data distribution. For edge samples or complex distribution samples in injector data (such as abnormal temperature control records, offset sample bottle positions), the discriminator can effectively distinguish the subtle features of real data and generated data. The enhanced feature capture capability enables the discriminator to guide the generator to more accurately generate pseudo samples that conform to the real distribution, thereby improving the quality of data expansion, which is expressed as:

[0134]

[0135]

[0136] In the formula, is the discriminator function, For the A real sample of injector data, are the discriminator parameters, represents the discriminator Layer activation function, represents the discriminator Layer activation function, represents the discriminator Layer activation function, For the discriminator The weight matrix of the layer, are the parameters of the first layer of the discriminator, is the Sigmoid activation function.

[0137] The embodiment of the present invention increases the number of samples through a trained data expansion model. In one embodiment, assuming that the original collected samples are 800, and the data expansion model expands and generates 200 samples, the expanded data set contains 1000 samples.

[0138] Furthermore, in the intelligent control task of the automatic sampler of the liquid chromatograph, the samples generated by the traditional sample augmentation technology are insufficient in diversity and authenticity, resulting in limited generalization ability of the model for sparse and high-dimensional complex distribution data. In addition, there is a lack of dynamic adjustment of gradients and learning rates in the data generation and discrimination process, and the training process is prone to fall into local optimality or oscillation, which affects the convergence effect of the model and the quality of the augmented sampler data.

[0139] Correspondingly, the generative adversarial network of an embodiment of the present invention trains the generator and discriminator of the generative adversarial network based on a preset optimization algorithm; wherein the preset optimization algorithm includes at least one of the following optimization parameters: a dynamic weight adjustment factor, a learning rate adjustment factor based on an adaptive gradient, and an adaptive incremental learning comprehensive loss function.

[0140] 1) Dynamic weight adjustment factor.

[0141] First, the embodiment of the present invention adopts an adaptive weighted loss function constraint in the training process of the generator and the discriminator, automatically adjusts the weight according to the distribution of the injector data, takes into account diversity and authenticity, and thus improves the learning ability of unbalanced injector data and sparse injector data. Among them, the initial loss function is expressed as:

[0142]

[0143] In the formula, is the loss function of the generator, which aims to minimize the gap between the generated injector data and the real injector data; represents the real injector data distribution, It means that it obeys a specific distribution. Express expectations, represents the generated injector data sample; is a real injector data sample, The pseudo injector data corresponding to the real injector data samples will affect the fit of the generator to the distribution.

[0144] The loss function of the discriminator takes into account the need to distinguish between real injector data and generated injector data, and the calculation method is expressed as:

[0145]

[0146] In the formula, is the loss function of the discriminator, The real samples and The generated samples are discriminated separately to maximize the discrimination; Indicates the generation of injector data distribution.

[0147] In order to prevent the generator and the discriminator from overfitting or being difficult to estimate correctly during the training process, the present invention balances the learning progress by dynamically adjusting the factor, and uses the dynamic adjustment factor to calculate the weight parameter matrix of the generator or the discriminator, so that the training of the generator and the discriminator can remain robust, which is expressed as:

[0148]

[0149] In the formula, is the balanced generator loss function, is the dynamic adjustment factor of the generator loss function, is the weight parameter matrix of the generator, is the L2 norm.

[0150] In specific implementation, dynamic adjustment is performed by comparing the size of the balanced generator loss gradient and the balanced discriminator loss gradient. For example, when the balanced generator loss gradient (such as 0.02) is less than 50% of the balanced discriminator loss gradient (such as 0.05), it means that the generator is optimized slowly. At this time, the adjustment factor is increased from the original 0.3 to 0.4 to help the generator optimize faster; if the gradient ratio reaches 80%, it remains unchanged. In one embodiment, Can be set to 0.3.

[0151] The loss function of the discriminator also uses a dynamic adjustment factor to improve the generalization ability of the discriminator. For some difficult pseudo samples generated by the generator, the adjustment factor can appropriately reduce the training pressure of the generator, while enhancing the discriminator's ability to distinguish these samples, avoiding oscillations or falling into local optimality during training, and improving the robustness and diversity of data expansion. The calculation method is expressed as:

[0152]

[0153] In the formula, is the balanced discriminator loss function, is the dynamic adjustment factor of the discriminator loss function, is the weight parameter matrix of the discriminator. In one embodiment, Can be set to 0.3.

[0154] 2) Learning rate adjustment factor based on adaptive gradient.

[0155] In generative adversarial training, the embodiment of the present invention can also adopt a learning rate dynamic adjustment strategy based on adaptive gradient, so that in each round of training, the learning rate is adaptively updated according to the gradient changes of the generator and the discriminator, so as to avoid the model falling into the local optimum or the occurrence of training oscillation. When the discriminator's ability to discriminate certain pseudo samples tends to be saturated, the learning rate will automatically decrease, which promotes the generator to better explore the high-dimensional data distribution, improves the convergence speed and quality of training, and makes the expanded data closer to the real sample, which is expressed as:

[0156]

[0157]

[0158] In the formula, is the learning rate of the generator, is the learning rate of the discriminator, is the initial learning rate of the generative adversarial network, is the learning rate adjustment factor for generating adversarial networks, is the gradient of the generator loss function with respect to its weight parameters, is the gradient of the discriminator loss function with respect to its weight parameters. Preferably, Set to 0.001.

[0159] 3) Adaptive incremental learning of comprehensive loss function.

[0160] In the expansion of the monitoring parameters of the automatic sampler, the embodiment of the present invention also designs and adds an adaptive incremental learning mechanism to focus on training difficult samples with high complexity, low frequency or outliers, avoid neglecting a small number of samples under large-scale sampler data sets, and adopts dynamic adjustment of incremental training weighting factors to enhance the learning effect of difficult samples, ensure that the expanded data covers more complex distribution samples, and improve the adaptability of the generator to the global distribution, which is expressed as:

[0161]

[0162] In the formula, For incremental learning of the comprehensive loss function, difficult samples receive higher attention under this loss function; is the incremental training weighting factor.

[0163] In order to measure the difficulty of generating a certain injector data sample, the embodiment of the present invention selects key training objects based on their information entropy, and the calculation method is expressed as:

[0164]

[0165] In the formula, For the The information entropy of samples is For the samples belong to the category The conditional probability of is the total number of categories. Preferably, the information entropy threshold is 0.3 to identify key samples, thereby achieving refined incremental training.

[0166] The incremental training weighting factor is dynamically adjusted according to the entropy value of the generated injector data, and the calculation method is expressed as:

[0167]

[0168] Where: To adjust the intensity, control the influence of entropy value on the incremental training weighting factor. In summary, the above training steps are repeated and iterated until the incremental learning comprehensive loss function reaches convergence.

[0169] Step S210: According to the state analysis result, the automatic sampler is driven and controlled to sample the sample bottle to be sampled.

[0170] In summary, the embodiments of the present invention bring the following beneficial effects:

[0171] 1. In the intelligent control task of the liquid chromatograph automatic sampler, a neural network based on quantum chaos is adopted, and quantum state superposition and interference are used to strengthen the dynamic relationship between global and local features. At the same time, key sampler data features are retained through quantum chaos perturbation and multi-scale pooling, which solves the problems of difficult extraction of complex nonlinear sampler data features and unstable data pattern learning.

[0172] 2. In the intelligent control task of the liquid chromatograph automatic sampler, a generative adversarial network based on dynamic balance is adopted: by dynamically adjusting the training factors of the generator and the discriminator and the adaptive weighted loss function, the problems of insufficient diversity of the sampler data and uneven distribution of high-dimensional features are solved, and the authenticity and comprehensiveness of the expanded sampler data are enhanced.

[0173] 3. In the intelligent control task of the liquid chromatograph automatic sampler, an adaptive gradient learning rate dynamic adjustment strategy is adopted to adjust the learning rate in real time according to the gradient changes of the generator and the discriminator, which solves the problem of falling into local optimality and oscillation in the training of the generative adversarial network, and improves the convergence speed of the model and the efficiency of sampler data expansion.

[0174] 4. In the intelligent control task of the liquid chromatograph automatic sampler, an information entropy-driven incremental learning mechanism is adopted. By calculating the sample information entropy, key samples are selected for training, and the weighting factors are dynamically adjusted. This solves the problem of poor learning effect of low-frequency complex samples and outlier samples, and ensures that the sampler data expansion covers more complex distribution samples.

[0175] Furthermore, in the above-mentioned implementation manner, the present invention also provides an intelligent control device for an automatic sample injector of a liquid chromatograph. Figure 6 The structure diagram of an intelligent control device for an automatic sample injector of a liquid chromatograph provided by an embodiment of the present invention is shown. Figure 6 The device includes: a data acquisition module 100, which is used to monitor the data of the sample injection chamber of the automatic sampler of the liquid chromatograph to obtain the monitoring parameters of the sample bottle to be sampled in the sample injection chamber; a data processing module 200, which is used to perform quantum chaos mapping on the monitoring parameters to obtain the data to be measured corresponding to the monitoring parameters; the quantum chaos mapping is based on the superposition calculation of multiple fluctuation terms of the monitoring parameters; an execution module 300, which is used to analyze the data to be measured and determine the state analysis result corresponding to the sample bottle; wherein the state analysis result of the sample bottle includes the intervention state of the sample bottle, the placement position state, the type of the sample bottle, the sample label state of the sample bottle and the hole temperature distribution state of the sample bottle; a control module 400, which is used to drive and control the automatic sampler according to the state analysis result to sample the sample bottle to be sampled.

[0176] An intelligent control device for an automatic sample injector of a liquid chromatograph provided in an embodiment of the present invention has the same technical features as an intelligent control method for an automatic sample injector of a liquid chromatograph provided in the above embodiment, and therefore can also solve the same technical problems and achieve the same technical effects.

[0177] Furthermore, the data processing module 200 is also used to standardize the monitoring parameters to obtain the initial data to be tested; perform quantum mapping on the initial data to be tested to obtain the quantum mapping results corresponding to the initial data to be tested; and perform convolution calculation on the initial data to be tested using a preset machine learning unit; perform chaotic mapping on the quantum mapping results and the convolution calculation to obtain the data to be tested corresponding to the monitoring parameters. The execution module 300 is also used to perform multi-scale pooling operations on the data to be tested using a preset machine learning unit, perform quantum chaotic pooling on the pooled data to be tested, and obtain the target parameters corresponding to the data to be tested; calculate the class probability corresponding to the target parameter, and determine the state analysis results corresponding to the sample bottle based on the class probability.

[0178] The device also includes a construction module for initializing the weights of the machine learning unit based on a preset quantum mapping function; wherein the quantum mapping function includes multiple nonlinear quantum chaos terms, and the multiple nonlinear quantum chaos terms include quantum chaos feedback terms and quantum oscillation control terms. The above construction module is also used to train the machine learning unit using a preset training sample set to calculate the adaptive loss function corresponding to the training sample set; determine the cosine loss term corresponding to the adaptive loss function; and optimize the learning rate of the machine learning unit based on the cosine loss term to update the weight of the machine learning unit.

[0179] The above-mentioned construction module is also used to obtain pre-collected sample bottle status monitoring samples, annotate the sample bottle status monitoring samples, and construct an initial training sample set; based on the normal distribution of the initial training sample set, generate an adaptive noise vector corresponding to the initial training sample set; based on the adaptive noise vector, perform data expansion on the initial training sample set to construct a training sample set. The above-mentioned construction module is also used to perform data expansion on the initial training sample set based on the adaptive noise vector through a pre-constructed generative adversarial network to construct a training sample set. The generative adversarial network trains the generator and discriminator of the generative adversarial network based on a preset optimization algorithm; the preset optimization algorithm includes at least one of the following optimization parameters: a dynamic weight adjustment factor, a learning rate adjustment factor based on an adaptive gradient, and an adaptive incremental learning comprehensive loss function.

[0180] Furthermore, on the basis of the above embodiments, the embodiments of the present invention also provide an intelligent control system for an automatic sample injector of a liquid chromatograph, which is configured with an intelligent control device for an automatic sample injector of a liquid chromatograph of any of the above embodiments, and is used to execute an intelligent control method for an automatic sample injector of a liquid chromatograph of any of the above embodiments. The intelligent control system for an automatic sample injector of a liquid chromatograph provided by the embodiments of the present invention has the same technical features as the intelligent control method for an automatic sample injector of a liquid chromatograph provided by the above embodiments, so it can also solve the same technical problems and achieve the same technical effects.

[0181] An embodiment of the present invention further provides an electronic device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the above Figures 1 to 5 The embodiment of the present invention also provides a computer-readable storage medium having a computer program stored thereon, and the computer program is executed by a processor to perform the above Figures 1 to 5 The embodiment of the present invention also provides a structural diagram of an electronic device, such as Figure 7FIG. 1 is a schematic diagram of the structure of the electronic device, wherein 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, and the processor 71 executes the computer executable instructions to implement the above Figures 1 to 7 Any of the methods shown.

[0182] exist Figure 7 In the illustrated embodiment, the electronic device further includes a bus 72 and a communication interface 73, wherein the processor 71, the communication interface 73 and the memory 70 are connected via 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. The communication connection between the system network element and at least one other network element is realized through at least one communication interface 73 (which may be wired or wireless), and the Internet, a wide area network, a local area network, a metropolitan area network, etc. may be used. The bus 72 may be an ISA (Industry Standard Architecture) bus, a PCI (Peripheral Component Interconnect) bus, or an EISA (Extended Industry Standard Architecture) bus, etc. It may also be an AMBA (Advanced Microcontroller Bus Architecture) bus, where AMBA defines three types of buses, including an APB (Advanced Peripheral Bus) bus, an AHB (Advanced High-performance Bus) bus, and an AXI (Advanced Xtensible Interface) bus. The bus 72 may be divided into an address bus, a data bus, a control bus, etc. For ease of representation, Figure 7Only one bidirectional arrow is used in the diagram, but it does not mean that there is only one bus or one type of bus. The processor 71 may be an integrated circuit chip with signal processing capabilities. In the implementation process, each step of the above method can be completed by the hardware integrated logic circuit in the processor 71 or the instructions in the form of software. The above-mentioned processor 71 can be a general-purpose processor, including a central processing unit (CPU), a network processor (NP), etc.; it can also be a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field programmable gate array (FPGA) or other programmable logic devices, discrete gates or transistor logic devices, discrete hardware components. The general-purpose processor can be a microprocessor or the processor can also be any conventional processor. The steps of the method disclosed in the embodiment of the present application can be directly embodied as a hardware decoding processor to be executed, or a combination of hardware and software modules in the decoding processor can be executed. The software module can be located in a random access memory, flash memory, read-only memory, programmable read-only memory or electrically erasable programmable memory, register or other mature storage media in the art. The storage medium is located in the memory, and the processor 71 reads the information in the memory and completes the above-mentioned Figures 1 to 5 Any of the methods shown.

[0183] A computer program product of a liquid chromatograph automatic sampler intelligent control method and system provided in an embodiment of the present invention includes a computer-readable storage medium storing a program code, and the instructions included in the program code can be used to execute the method described in the previous method embodiment. The specific implementation can refer to the method embodiment, which will not be repeated here. A person skilled in the art can clearly understand that for the convenience and simplicity of description, the specific working process of the system described above can refer to the corresponding process in the aforementioned method embodiment, which will not be repeated here. In addition, in the description of the embodiment of the present invention, unless otherwise clearly specified and limited, the terms "installation", "connection" and "connection" 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 it can be indirectly connected through an intermediate medium, or it can be the internal connection of two components. For those skilled in the art, the specific meaning of the above terms in the present invention can be understood in specific circumstances. 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 understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or the part of the technical solution, can be embodied in the form of a software product, which is stored in a storage medium and includes several instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes: a U disk, a mobile hard disk, a read-only memory (ROM, Read-Only Memory), a random access memory (RAM, Random Access Memory), a disk or an optical disk, and other media that can store program codes. In the description of the present invention, it should be noted that the orientation or position relationship indicated by the terms "center", "upper", "lower", "left", "right", "vertical", "horizontal", "inside", "outside", etc. is based on the orientation or position relationship shown in the accompanying drawings, which 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 on 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 implementations of the present invention, which are used to illustrate the technical solutions of the present invention, rather than to limit them. The protection scope of the present invention is not limited thereto. Although the present invention is described in detail with reference to the above embodiments, those skilled in the art should understand that any person skilled in the art can still modify the technical solutions recorded in the above embodiments within the technical scope disclosed by the present invention, or can easily think of changes, or make equivalent replacements for some of the technical features therein; 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 be included in the protection scope of the present invention. Therefore, the protection scope of the present invention shall be based on the protection scope of the claims.

Claims

1. A liquid chromatograph automatic sample injector intelligent control method, characterized in that: The method comprises: Performing data monitoring on a sample injection chamber of an automatic sample injector of a liquid chromatograph to obtain monitoring parameters of a sample bottle to be sampled in the sample injection chamber; Performing quantum chaos mapping on the monitoring parameter to obtain the measured data corresponding to the monitoring parameter; the quantum chaos mapping is based on superposition calculation of multiple fluctuation terms of the monitoring parameter; Analyze the data to be tested to determine the state analysis result corresponding to the sample bottle; wherein the state analysis result of the sample bottle includes the intervention state, placement position state, type of the sample bottle, sample label state of the sample bottle and hole position temperature distribution state of the sample bottle; According to the state analysis result, the automatic sampler is driven and controlled to sample the sample bottle to be sampled; The step of performing quantum chaos mapping on the monitoring parameters to obtain the data to be measured corresponding to the monitoring parameters comprises: Standardizing the monitoring parameters to obtain initial data to be measured; Performing quantum mapping processing on the initial data to be tested to obtain a quantum mapping result corresponding to the initial data to be tested; and performing convolution calculation on the initial data to be tested using a preset machine learning unit; Chaotic mapping is performed on the quantum mapping result and the convolution calculation to obtain the data to be measured corresponding to the monitoring parameter.

2. The method according to claim 1, characterized in that The step of analyzing the data to be tested to determine the state analysis result corresponding to the sample bottle includes: Using a preset machine learning unit to perform a multi-scale pooling operation on the data to be tested; Performing quantum chaotic pooling on the multi-scale pooled data to be measured to obtain target parameters corresponding to the data to be measured; The class probability corresponding to the target parameter is calculated, and based on the class probability, the state analysis result corresponding to the sample bottle is determined.

3. The method according to any one of claims 1 or 2, characterized in that: The weight of the machine learning unit is initialized based on a preset quantum mapping function; The quantum mapping function includes multiple nonlinear quantum chaos terms, and the multiple nonlinear quantum chaos terms include quantum chaos feedback terms and quantum oscillation control terms.

4. The method according to any one of claims 1 or 2, characterized in that: The method for constructing the machine learning unit comprises: Using a preset training sample set to train the machine learning unit to calculate an adaptive loss function corresponding to the training sample set; Determining a cosine loss term corresponding to the adaptive loss function; The learning rate of the machine learning unit is optimized based on the cosine loss term to update the weight of the machine learning unit.

5. The method according to claim 4, characterized in that The method for constructing the training sample set includes: Acquire pre-collected sample bottle status monitoring samples, annotate the sample bottle status monitoring samples, and construct an initial training sample set; Based on the normal distribution of the initial training sample set, generating an adaptive noise vector corresponding to the initial training sample set; The initial training sample set is expanded based on the adaptive noise vector to construct a training sample set.

6. The method according to claim 5, characterized in that The step of performing data expansion on the initial training sample set based on the adaptive noise vector to construct a training sample set includes: The initial training sample set is expanded based on the adaptive noise vector by using a pre-constructed generative adversarial network to construct a training sample set.

7. The method according to claim 6, characterized in that The method further comprises: The generative adversarial network trains the generator and discriminator of the generative adversarial network based on a preset optimization algorithm; wherein the preset optimization algorithm includes at least one of the following optimization parameters: a dynamic weight adjustment factor, a learning rate adjustment factor based on an adaptive gradient, and an adaptive incremental learning comprehensive loss function.

8. An intelligent control device for an automatic sample injector of a liquid chromatograph, characterized in that: The device comprises: A data acquisition module is used to monitor the data of the sample injection chamber of the automatic sample injector of the liquid chromatograph to obtain monitoring parameters of the sample bottle to be sampled in the sample injection chamber; A data processing module, used for performing quantum chaos mapping on the monitoring parameters to obtain the measured data corresponding to the monitoring parameters; the quantum chaos mapping is based on superposition calculation of multiple fluctuation terms of the monitoring parameters; An execution module is used to analyze the data to be tested through a preset machine learning unit to determine a state analysis result corresponding to the sample bottle; wherein the state analysis result of the sample bottle includes an intervention state, a placement position state, a type of the sample bottle, a sample label state of the sample bottle, and a hole position temperature distribution state of the sample bottle; A control module, used for driving and controlling the automatic sampler according to the state analysis result, so as to sample the sample bottle to be sampled; The data processing module is also used to standardize the monitoring parameters to obtain initial data to be tested; perform quantum mapping on the initial data to be tested to obtain quantum mapping results corresponding to the initial data to be tested; and use a preset machine learning unit to perform convolution calculation on the initial data to be tested; perform chaotic mapping on the quantum mapping results and the convolution calculation to obtain the data to be tested corresponding to the monitoring parameters.

9. An intelligent control system for an automatic sample injector of a liquid chromatograph, characterized in that: The device according to claim 8 is configured to execute the method according to any one of claims 1 to 7.

Images