Rapid chromatographic analysis method, device and system based on neural network

Abstract

The invention discloses a rapid chromatographic analysis method, device and system based on a neural network, and the method comprises the steps: collecting historical chromatographic data, and carrying out the data preprocessing, so as to obtain a training data set; and based on the neural network architecture and the training data set, constructing and training to obtain a chromatographic analysis model. And carrying out enrichment treatment and impurity removal treatment on the original sample to obtain a sample to be detected. And correspondingly setting the chromatographic instrument according to the characteristics and analysis requirements of the target substance in the sample to be detected. Injecting a to-be-detected sample into the chromatographic instrument, performing real-time acquisition and real-time processing, and performing data characteristic processing in real time to obtain real-time chromatographic data meeting the input requirements of the chromatographic analysis model. And inputting real-time chromatographic data into the neural network model to realize forward propagation rapid calculation, and outputting a chromatographic analysis result. Through the assistance of the neural network, the chromatographic analysis effect can be quickly and accurately obtained.

Description

Technical Field

[0001] The present invention relates to the technical field of chromatographic analysis, and in particular to a neural network-based rapid chromatographic analysis method, device and system. Background Art

[0002] In today's chemical analysis field, rapid and accurate chromatographic analysis is crucial for numerous industries, such as oil and gas exploration, chemicals, pharmaceuticals, and environmental monitoring. Traditional chromatographic analysis methods often expose numerous limitations when dealing with complex samples or when rapid analytical results are required.

[0003] For example, hydrocarbons and hydrogen sulfide, which are flammable, colorless, highly toxic, and odorless at high concentrations, are released into formations during oil drilling. These gases are widely present in industrial production and environmental monitoring, and failure to identify them early can have serious consequences. Therefore, accurate and rapid analysis of the content and characteristics of these flammable and harmful gases is crucial during chromatographic analysis. However, traditional gas chromatography methods suffer from low efficiency and poor accuracy when performing gas analysis.

[0004] In view of this, it is necessary to provide an accurate and rapid chromatographic analysis method. Summary of the Invention

[0005] The purpose of the present invention is to overcome the shortcomings of the existing technology and provide a neural network-based rapid chromatography analysis method, device and system, which constructs and trains an efficient and accurate chromatography analysis model through historical chromatography data to achieve the purpose of quickly and accurately obtaining chromatography analysis results.

[0006] The technical solution of the present invention provides a rapid chromatographic analysis method based on a neural network, comprising:

[0007] Collect historical chromatographic data and perform data preprocessing to obtain a training data set;

[0008] Constructing an original model based on a neural network architecture, and training the initial model according to the training data set to obtain a chromatographic analysis model;

[0009] The original sample is enriched and impurity-removed to obtain the sample to be tested;

[0010] Setting the chromatograph accordingly according to the characteristics of the target substance in the sample to be tested and the analysis requirements;

[0011] Injecting the sample to be tested into the chromatograph and performing real-time acquisition and real-time processing, and performing real-time data feature processing to obtain real-time chromatographic data that meets the input requirements of the chromatographic analysis model;

[0012] The real-time chromatographic data is input into the neural network model to realize fast forward propagation calculation and output chromatographic analysis results.

[0013] In one of the optional technical solutions, the collecting of historical chromatographic data and performing data preprocessing to obtain a training data set includes:

[0014] Collect historical chromatographic data containing target substances from different industrial scenarios and laboratory simulation environments;

[0015] Recording the target substance concentration range, sample collection environment, and sample pretreatment conditions of each of the historical chromatographic data, and combining the historical chromatographic data to form a training data set;

[0016] The training data set is processed to meet the requirements of training the chromatographic analysis model, and the training data set is divided into a training set for training model update learning, a validation set for monitoring model performance during training, and a test set for evaluating model capabilities.

[0017] In one of the optional technical solutions, an original model is constructed based on a neural network architecture, and the initial model is trained according to the training data set to obtain a chromatographic analysis model, including:

[0018] Build a multi-layer perceptron neural network architecture and ensure that the number of neurons in the input layer is consistent with the dimensions of the pre-processed feature vector to ensure that the input data can be fully received;

[0019] Set up multiple hidden layers in the neural network architecture, and determine the number of neurons in each hidden layer through repeated experiments and verification to balance the complexity of the model and computational efficiency;

[0020] The number of neurons in the output layer is set according to the specific analysis task, and the initial weights and bias values ​​are assigned to the connections between the output layers to obtain the initial model of the neural network architecture.

[0021] Selecting a loss function and an optimizer based on the characteristics of the target substance and analysis requirements to optimize the initial model;

[0022] The sample data and chromatographic data in the training set are sequentially input into the initial model for training to obtain a chromatographic analysis model.

[0023] In one of the optional technical solutions, the chromatographic analysis model further optimizes weights and bias values ​​through a genetic algorithm.

[0024] In one of the optional technical solutions, the chromatographic analysis model further optimizes weights and bias values ​​using a genetic algorithm, including:

[0025] Adopting a hierarchical and segmented coding strategy, the weights and bias parameters of the neural network are layered according to the network layer, and each layer is segmented and coded according to the connection relationship of neurons;

[0026] Set the fitness function according to the requirements of chromatographic analysis accuracy, stability and analysis speed;

[0027] The fitness value is calculated by segmented coding and fitness function, and genetic operation is performed according to the fitness value to optimize the weight and bias value in the chromatographic analysis model.

[0028] In one of the optional technical solutions, after constructing and training the chromatographic analysis model based on the neural network architecture and the training data set, a model evaluation step is further performed, specifically comprising:

[0029] Set identification tasks and analysis tasks and corresponding indicators through the data of the test set;

[0030] enabling the chromatographic analysis model to perform an identification task and an analysis task respectively, evaluating the identification performance and the analysis performance of the chromatographic analysis model in chromatographic analysis, and obtaining an evaluation result;

[0031] If the evaluation result does not meet the preset requirements, the cause is analyzed in depth to obtain analysis results, and the chromatographic analysis model is adjusted in a targeted manner based on the analysis results.

[0032] In one of the optional technical solutions, the chromatograph is configured accordingly according to the characteristics of the target substance in the sample to be tested and the analysis requirements, including:

[0033] Select the chromatographic column and aging plan based on the characteristics of the target substance and the analysis requirements, and perform the corresponding aging plan on the chromatographic column;

[0034] Calculate the target intercolumn temperature, target intracolumn pressure and target carrier gas flow rate of the chromatograph based on the properties of the target substance, carrier gas type, chromatographic column specifications and target analytical accuracy;

[0035] Adjust the status of the chromatograph until the target intercolumn temperature, target intracolumn pressure and target carrier gas flow rate are met.

[0036] In one of the optional technical solutions, a chromatographic column and an aging scheme are selected based on the characteristics of the target substance and the analytical requirements, and the corresponding aging scheme is performed on the chromatographic column, including:

[0037] Multiple sensors are set up in the chromatograph and combined with the control module to form a control system based on real-time feedback;

[0038] Real-time monitoring of the chromatograph's column pressure, intercolumn temperature, and carrier gas flow rate;

[0039] Based on the data of the sample to be tested and the chromatograph, a prediction model is established to simulate and calculate the optimal separation conditions of the chromatograph;

[0040] Compare and analyze the real-time monitoring data with the optimal separation conditions, obtain the comparison results and feed them back to the control system;

[0041] The control system adjusts the column oven heating power, pressure regulating valve and carrier gas flow regulating valve of the chromatograph according to the comparison results to achieve coordinated control of inter-column temperature, intra-column pressure and carrier gas flow.

[0042] The present invention provides a device comprising a memory, a processor and an electronic device program on the memory, wherein the processor executes the electronic device program to implement any step of the aforementioned neural network-based rapid chromatographic analysis method.

[0043] The present invention provides a system comprising any of the aforementioned devices, wherein when the system is executed, the system implements the steps of any of the aforementioned neural network-based rapid chromatographic analysis methods.

[0044] The above technical solution has the following beneficial effects:

[0045] The neural network-based rapid chromatography analysis method provided by the present invention constructs and trains an efficient and accurate chromatography analysis model by collecting and preprocessing historical chromatography data, then constructs and trains the model based on the neural network and the training data set, and then enriches and removes impurities from the original sample to obtain a test sample with reduced noise and redundancy. The chromatograph is specifically adjusted according to the characteristics of the target substance in the test sample and the analysis requirements to optimize the separation conditions of the chromatograph, and then the real-time chromatography data produced by the chromatograph is input into the chromatography analysis model, so as to achieve the effect of quickly and accurately obtaining the chromatographic analysis results. BRIEF DESCRIPTION OF THE DRAWINGS

[0046] The disclosure of the present invention will become more easily understood with reference to the accompanying drawings. It should be understood that these drawings are for illustrative purposes only and are not intended to limit the scope of protection of the present invention. In the drawings:

[0047] Figure 1 A flowchart of a rapid chromatographic analysis method based on a neural network according to an embodiment of the present invention;

[0048] Figure 2 A schematic structural diagram of a device provided in one embodiment of the present invention. DETAILED DESCRIPTION

[0049] The specific embodiments of the present invention will be further described below with reference to the accompanying drawings.

[0050] like Figure 1FIG. 1 is a diagram illustrating a neural network-based rapid chromatography analysis method according to an embodiment of the present invention. The technical solution of the present invention provides a neural network-based rapid chromatography analysis method, comprising:

[0051] Step S101: Collect historical chromatographic data and perform data preprocessing to obtain a training data set. The historical chromatographic data may be regional stratum historical chromatographic data, such as chromatographic data obtained when a target substance exists in a stratum during oil drilling.

[0052] Step S102: constructing an original model based on a neural network architecture, and training the initial model according to a training data set to obtain a chromatographic analysis model.

[0053] Step S103: The original sample is enriched and impurity-removed to obtain a sample to be tested. The impurity-removal process may be a heat vacuum treatment or other impurity-removal process.

[0054] Step S104: The chromatograph is configured based on the characteristics of the target substance in the sample to be tested and the analysis requirements. The characteristics of the target substance can be formation characteristics. For example, during oil drilling, the sample to be tested is a soil sample collected from the formation in the area. In this case, the characteristics of the target substance include formation characteristics. It is necessary to record whether the target substance is soft soil, clay soil, sandy soil, gravel soil, etc., to ensure that the chromatograph can be configured accordingly and the chromatographic data produced by the chromatograph is accurate.

[0055] Step S105: injecting the sample to be tested into the chromatograph and performing real-time acquisition and real-time processing, and performing real-time data feature processing to obtain real-time chromatographic data that meets the input requirements of the chromatographic analysis model.

[0056] Step S106: input the real-time chromatographic data into the neural network model to implement fast forward propagation calculation and output the chromatographic analysis results.

[0057] Before model building and training, the target substances and analysis requirements need to be known in advance, so that historical chromatographic data can be obtained in a targeted manner. For example, if a model needs to be built and trained to analyze the concentration of hydrocarbons or hydrogen sulfide in gases, liquids, or solids, a large amount of historical chromatographic data containing hydrocarbons or hydrogen sulfide in chromatographic analysis results from various industrial scenarios (such as refineries, natural gas purification plants, sewage treatment plants, etc.) and laboratory simulation environments needs to be collected. After the historical chromatographic data is collected in a targeted manner, it is initially cleaned to remove obvious erroneous data records caused by instrument failure or operational errors to ensure the basic quality of the data. The historical chromatographic data is then preprocessed to achieve feature standardization, feature extraction, and time series enhancement of the target substances in the historical chromatographic data, thereby obtaining a training dataset suitable for neural network training. An initial model of the neural network architecture is then built. Then, an efficient and accurate chromatographic analysis model is trained using this initial model and the training dataset. The original sample is then enriched and cleaned to obtain the test sample with reduced noise and redundancy.

[0058] Among them, the pretreatment of the original sample can be done through a high-precision microporous filter to effectively remove solid impurities in the sample, preventing them from clogging or interfering with subsequent chromatographic analysis instruments. The pore size of the filter is precisely selected according to the nature of the sample and the size of the impurities, ensuring that impurities are filtered out to the maximum extent possible without losing the target substance, thereby ensuring the accuracy of subsequent identification and analysis. Then, an enrichment device based on a special chemically modified porous carbon material is used. The weak interaction formed between the functional groups on the surface of the material and the molecules of the target substance preferentially adsorbs the target substance. During the adsorption process, the adsorption efficiency and selectivity are optimized by precisely controlling parameters such as temperature, pressure, and adsorption time.

[0059] Cold traps and photocatalysis can also be used for drying and impurity removal. The original sample is introduced into a photocatalytic reaction device. Under the irradiation of light of a specific wavelength, the electron-hole pairs generated by the loaded titanium dioxide photocatalyst undergo redox reactions with the impurity gas in the sample, converting the impurities into harmless substances or easily separated compounds.

[0060] Then, based on the characteristics of the target substance in the sample to be tested and the analysis requirements, the chromatograph is adjusted in a targeted manner to optimize the separation conditions of the chromatograph. The pre-processed chromatographic data of the sample to be tested is input into the trained neural network model. The model quickly calculates and outputs the identification and analysis results through forward propagation. Finally, the analysis results are presented to the user in an intuitive and easy-to-understand manner. For example, the identification results of the target substance (such as presence or absence, possible interfering gases) and content information are displayed in digital and graphical form on the display screen, and compared with the standard limits. At the same time, a detailed analysis report is provided, including the source of the sample, analysis method, uncertainty assessment of the analysis results, and trend analysis based on historical data, providing a comprehensive and powerful basis for the user's decision-making. In addition, the analysis results are compared with the relevant database in real time to provide users with historical identification and analysis data of similar samples and processing suggestions to help users better understand and apply the analysis results.

[0061] In one embodiment, step S101 includes:

[0062] Collect historical chromatographic data containing target substances from different industrial scenarios and laboratory simulation environments.

[0063] The target substance concentration range, sample collection environment and sample pretreatment conditions of each historical chromatographic data are recorded, and the training data set is formed after combining the historical chromatographic data.

[0064] The training dataset is processed to meet the requirements of training the chromatography analysis model, and is divided into a training set for training model update learning, a validation set for monitoring model performance during training, and a test set for evaluating model capabilities.

[0065] In this embodiment, the preprocessing of the historical chromatographic data includes feature normalization, wavelet transformation, feature extraction, and time series enhancement.

[0066] Among them, feature standardization uses standardization methods to unify different features in historical chromatographic data (such as peak height, peak area, retention time, etc.) to a scale of zero mean and unit variance, so that the weights of each feature are updated more balanced during neural network training, improving training stability and efficiency, so as to better learn the feature patterns used for identification and analysis.

[0067] Wavelet Transform and Feature Extraction uses wavelet transform to decompose historical chromatographic data into sub-signals of varying frequencies, deeply exploring the inherent connections between each frequency component and the characteristics of the target substance. Leveraging in-depth domain knowledge and extensive historical data analysis, we extract wavelet coefficients that are closely related to the target substance's onset, peak, end time, and peak shape within a specific frequency range, constructing highly recognizable feature vectors.

[0068] Time series enhancement uses the retention time range of target substances under different chromatographic conditions to target and enhance chromatographic data. It not only captures data segments related to the target substance's potential peak time, but also utilizes techniques such as local amplification and smoothing to highlight the target substance's detailed features, while reducing data redundancy and lowering the computational burden of the neural network.

[0069] The preprocessing of historical chromatographic data is also applicable to the processing of real-time chromatographic data produced by the chromatograph, so as to make the data obtained by the chromatographic analysis model uniform, and further improve the accuracy of the analysis results of the chromatographic analysis data and the efficiency of the analysis process.

[0070] In one embodiment, step S102 includes:

[0071] Build a multi-layer perceptron neural network architecture and ensure that the number of neurons in the input layer is consistent with the dimension of the preprocessed feature vector to ensure that the input data can be fully received.

[0072] Set up multiple hidden layers in the neural network architecture, and determine the number of neurons in each hidden layer through repeated experiments and verification to balance the complexity and computational efficiency of the model.

[0073] The number of neurons in the output layer is set according to the specific analysis task, and the connections between the output layers are assigned initialized weights and bias values ​​to obtain the initial model of the neural network architecture.

[0074] Select a loss function and optimizer based on the characteristics of the target substance and analysis requirements to optimize the initial model.

[0075] The sample data and chromatographic data in the training set are sequentially input into the initial model for training to obtain a chromatographic analysis model.

[0076] Because they use a neural network architecture, both the original model and the chromatography analysis model have an input layer, hidden layers, and an output layer. The loss function measures the degree of discrepancy between the model's predicted value and the true value. For example, when predicting hydrocarbon or hydrogen sulfide gas content, if the model's predicted value deviates significantly from the actual content, the loss function value will be high; otherwise, it will be low. Common loss functions, such as the mean squared error (MSE), calculate the mean of the squared differences between the predicted and true values, providing a direct indicator of the model's prediction accuracy. The loss function provides guidance for model training. The core goal of model training is to minimize the loss function value by continuously adjusting parameters, bringing the model's predicted value as close to the true value as possible. The optimizer automatically adjusts the parameters of the neural network (such as weights and biases) based on the error calculated by the loss function to minimize the loss function. For example, the Adam optimizer combines the advantages of the Adagrad and RMSProp algorithms to adaptively adjust the learning rate of each parameter. During training, the optimizer updates parameters in the direction that minimizes the loss function based on the gradient information calculated by the backpropagation algorithm. Different optimizers differ in parameter update methods, learning rate adjustment strategies, etc. A suitable optimizer can accelerate model convergence, prevent the model from falling into a local optimal solution, and help the model reach the state of minimizing the loss function faster and more stably, thereby improving the model's training efficiency and performance.

[0077] Weights represent the strength of the connections between neurons in a neural network and determine the degree to which an input signal influences the output of neurons in the next layer. A neural network is an information-transfer network, with each neuron receiving input signals from multiple neurons in the previous layer. Weights act as a regulator between these signals. A bias is an additional constant term added to each neuron, providing a benchmark for neuron activation. Even if the weighted sum of all input signals is zero, the bias can still cause a neuron to produce a non-zero output.

[0078] In this embodiment, the specific formula of the chromatographic analysis model is:

[0079]

[0080] in, It is the output value of the chromatographic analysis model, which represents the content of the target substance in the sample to be tested in the real-time chromatographic data, such as the content of hydrogen sulfide gas or the sulfur content of hydrocarbons in the sample to be tested.

[0081] σ is the activation function of the output layer of the chromatographic analysis model, which is selected from the preset functions according to the specific analysis requirements.

[0082] m is the number of neurons in the hidden layer.

[0083] is the connection weight from the jth neuron in the hidden layer to the output layer.

[0084] ReLU(z) is the rectified linear unit activation function used in the hidden layer.

[0085] n is the number of neurons in the input layer, corresponding to the number of chromatographic features input to the neural network after preprocessing.

[0086] ω ji is the connection weight from the i-th neuron in the input layer to the j-th neuron in the hidden layer.

[0087] x i is the input value of the i-th neuron in the input layer, that is, the characteristic value of the preprocessed chromatographic data.

[0088] b j is the bias value of the jth neuron in the hidden layer.

[0089] b out is the bias value of the output layer.

[0090] The chromatographic analysis model can utilize the characteristics of neural networks to quickly and accurately output chromatographic analysis results.

[0091] In one embodiment, the chromatographic analysis model further optimizes weights and bias values ​​using a genetic algorithm, which specifically includes the following sub-steps:

[0092] A hierarchical and segmented coding strategy is adopted to stratify the weights and bias parameters of the neural network according to the network layers, and then segmented coding is performed according to the connection relationship of neurons in each layer.

[0093] The fitness function is set according to the requirements of chromatographic analysis accuracy, stability and analysis speed.

[0094] The fitness value is calculated by segmented coding and fitness function, and genetic operation is performed according to the fitness value to optimize the weight and bias value in the chromatographic analysis model.

[0095] In this embodiment, the optimized chromatographic analysis model formula is as follows:

[0096]

[0097] in, is the connection weight from the jth neuron in the hidden layer to the output layer after optimization by the genetic algorithm.

[0098] is the connection weight from the i-th neuron in the input layer to the j-th neuron in the hidden layer after optimization by the genetic algorithm.

[0099] is the bias value of the jth neuron in the hidden layer after optimization by the genetic algorithm.

[0100] b out·GA·opt is the bias value of the output layer after optimization by genetic algorithm.

[0101] In this example, the genetic algorithm employed a hierarchical and segmented encoding strategy. The neural network's weights and bias parameters were stratified by network layer, and within each layer, segmented encoding was performed based on the neuronal connectivity. This encoding approach not only preserves the structural information between parameters but also facilitates targeted optimization of parameters at different levels and functions during operation, improving optimization efficiency and adapting to the complex demands of chromatographic data identification and analysis.

[0102] This embodiment also designs the fitness function with the accuracy, stability and analysis speed of chromatographic analysis as comprehensive considerations. For the identification task, in addition to taking the classification accuracy of the neural network on the validation set as the main measurement indicator, indicators such as the false positive rate and the missed judgment rate are also introduced; for the analysis task, in addition to the mean square error (MSE), the stability index of chromatographic peak recognition (such as the coefficient of variation of the peak characteristics when repeatedly analyzing the same standard sample) and the analysis time index (the average time required for the model to process the data once) are introduced. By reasonably setting the weights of each indicator, a fitness function that comprehensively reflects the comprehensive performance of the model in chromatographic analysis is constructed to ensure that the neural network parameters optimized by the genetic algorithm can not only improve the accuracy of identification, but also ensure the stability and speed of the analysis results.

[0103] A chromatographic knowledge database can also be incorporated into genetic manipulations for assistance. Genetic algorithms require chromosome configuration. For example, for the identification or analysis of hydrocarbons or hydrogen sulfide, chromosomes that demonstrate excellent performance under specific chromatographic conditions, such as intercolumn temperature, intracolumn pressure, and carrier gas flow, are given an additional selection advantage. This allows the genetic algorithm to prioritize and propagate parameter combinations that meet the actual requirements of chromatographic analysis during the evolutionary process, accelerating convergence to optimal solutions and improving hydrogen sulfide identification and analysis capabilities. This allows for better optimization of weights and biases within the chromatographic analysis model.

[0104] Specific genetic operations include:

[0105] Crossover operation, a crossover method based on the similarity of chromatographic features is designed. Before performing the crossover operation, the similarity analysis of the output features of the neural network corresponding to the two chromosomes involved in the crossover on some typical chromatographic samples is performed.

[0106] Mutation: The mutation operation is improved by taking into account uncertainties in chromatographic analysis (such as instrument noise and minor sample differences). During mutation, the probability and amplitude of the mutation are adaptively adjusted based on the diversity of the current population and the noise level of the chromatographic data. When population diversity is low or the chromatographic data is noisy, the probability and amplitude of the mutation are appropriately increased to introduce more new parameter combinations and help the algorithm escape local optima. Conversely, the intensity of the mutation is reduced to maintain the algorithm's stability and convergence, ensuring accurate identification and analysis even in complex chromatographic data environments.

[0107] In one embodiment, after constructing and training a chromatographic analysis model based on a neural network architecture and a training dataset, a model evaluation step is performed, specifically comprising:

[0108] The identification tasks, analysis tasks and corresponding indicators are set through the data of the test set.

[0109] The chromatographic analysis model is made to perform identification tasks and analysis tasks respectively, and the identification performance and analysis performance of the chromatographic analysis model in chromatographic analysis are evaluated to obtain evaluation results.

[0110] If the evaluation results do not meet the preset requirements, the reasons are analyzed in depth, the analysis results are obtained, and the chromatographic analysis model is adjusted based on the analysis results.

[0111] In this example, after each round of training, the performance of the chromatographic analysis model was evaluated using a validation set, and changes in the loss function value and other evaluation indicators (such as accuracy and recall in the identification task, mean absolute error (MAE) and peak identification accuracy in the analysis task) were observed. If the performance on the validation set no longer improved within a certain number of rounds, the model was considered to be overfitting and training was stopped prematurely. Simultaneously, during the training process, the model's identification and analysis results for various chromatographic samples at different training stages were regularly recorded to analyze the model's learning process and performance trends.

[0112] The trained neural network model was comprehensively evaluated using the test data. For identification tasks, metrics such as accuracy, recall, and F1 score were calculated. For analysis tasks, multiple metrics such as root mean square error (RMSE), mean absolute error (MAE), peak identification accuracy, analysis result stability indicators (such as standard deviation), and average analysis time were calculated to comprehensively quantify the model's identification and analysis performance in chromatographic analysis.

[0113] If the evaluation results do not meet expectations, conduct an in-depth analysis of the reasons. Check multiple aspects, including data preprocessing, neural network structure, genetic algorithm parameter settings, and training process. For example, check whether the data preprocessing has fully extracted the characteristic information used to identify and analyze hydrogen sulfide, whether the neural network structure can effectively learn these characteristics, whether the genetic algorithm encoding, fitness function, and genetic operations are reasonable, and whether the hyperparameter settings during the training process are optimal. Based on the analysis results, adjust the model in a targeted manner, such as optimizing the data preprocessing method, adjusting the neural network structure, redesigning the genetic algorithm parameters or training hyperparameters, and then retrain and evaluate until the model performance meets the requirements for hydrogen sulfide chromatographic data identification and analysis.

[0114] In one embodiment, step S104 includes:

[0115] Select the chromatographic column and aging plan based on the characteristics of the target substance and analysis requirements, and perform the corresponding aging plan on the chromatographic column.

[0116] The target intercolumn temperature, target intracolumn pressure and target carrier gas flow rate of the chromatograph are calculated based on the properties of the target substance, carrier gas type, chromatographic column specifications and target analytical accuracy.

[0117] Adjust the status of the chromatograph until the target intercolumn temperature, target intracolumn pressure and target carrier gas flow rate are met.

[0118] In this embodiment, the aging process of the chromatographic column is to connect the chromatographic column to the chromatograph, and in the absence of sample injection, introduce carrier gas at a low flow rate, set the initial temperature to be lower than the maximum operating temperature of the chromatographic column, and then slowly increase the temperature at a preset heating rate to a temperature lower than the maximum operating temperature of the chromatographic column, and maintain the temperature at this temperature for a preset aging time to remove residual impurities in the chromatographic column, stabilize the stationary phase, and improve the column efficiency and stability.

[0119] In one embodiment, a chromatographic column and an aging protocol are selected based on the characteristics of the target substance and the analytical requirements, and the corresponding aging protocol is performed on the chromatographic column, including the following sub-steps:

[0120] Multiple sensors are set up in the chromatograph and combined with the control module to form a control system based on real-time feedback.

[0121] Real-time monitoring of the chromatograph's column pressure, intercolumn temperature, and carrier gas flow rate.

[0122] A prediction model is established based on the data of the sample to be tested and the chromatograph, and the optimal separation conditions of the chromatograph are simulated and calculated.

[0123] Compare and analyze the real-time monitoring data with the optimal separation conditions, obtain the comparison results and feed them back to the control system.

[0124] The control system adjusts the column oven heating power, pressure regulating valve and carrier gas flow regulating valve of the chromatograph according to the comparison results to achieve coordinated control of inter-column temperature, intra-column pressure and carrier gas flow.

[0125] In this embodiment, by setting up multiple sensors in the chromatograph, the state of the internal environment of the chromatograph can be monitored in real time, so that real-time regulation and control can be performed based on the real-time feedback data of the chromatograph to ensure that the chromatograph meets the optimal separation conditions of the target substance.

[0126] As a specific application of this method, during the analysis of hydrocarbon and hydrogen sulfide content in petroleum samples at a refinery, personnel collected historical chromatographic data containing hydrocarbons and hydrogen sulfide from various past production scenarios and laboratory simulations of similar environments. Detailed records of hydrocarbon and hydrogen sulfide concentration ranges, collection environmental parameters, and sample pretreatment conditions were also collected to form a training dataset. Subsequently, using methods such as Z-score normalization, the dataset was divided into training, validation, and test sets with a ratio of 70%, 15%, and 15%, respectively, laying the foundation for training the chromatographic analysis model.

[0127] A neural network architecture based on a multilayer perceptron was constructed, with the number of neurons in the input layer consistent with the dimensions of the preprocessed feature vector. The number of neurons in each hidden layer was determined through repeated experiments. To predict hydrogen sulfide concentration, the number of neurons in the output layer was set to 1. Initial weights and biases were assigned to each layer to create an initial model. A mean squared error (MSE) loss function and the Adam optimizer were used. A genetic algorithm was employed to optimize the weights and biases through a hierarchical segmented encoding strategy, setting a fitness function, and performing genetic operations. A chromatographic analysis model was then trained using the training set data.

[0128] Using the test data set, the identification system is set up to determine the presence of hydrocarbons and hydrogen sulfide. The hydrocarbon and hydrogen sulfide concentrations are then analyzed separately, followed by testing for identification accuracy, root mean square error (RMS), and other relevant metrics. The chromatographic analysis model is then run to evaluate its identification and analysis performance. If the evaluation results do not meet the pre-set requirements, further investigation is conducted on data preprocessing, neural network structure, and genetic algorithm parameters. Targeted adjustments to the model are made, such as optimizing preprocessing methods, adjusting the network structure, or resetting genetic algorithm parameters, until the requirements are met.

[0129] Raw samples are collected from drilling fluids or specific production processes in refineries. Hydrocarbons and hydrogen sulfide are enriched using a specially chemically modified porous carbon material enrichment device, and impurities are removed using a photocatalytic reaction device to obtain the test sample. The chromatograph's intercolumn temperature, intracolumn pressure, and carrier gas flow rate are calculated and set based on the properties of the hydrocarbons and hydrogen sulfide, carrier gas type, chromatographic column specifications, and target accuracy. A control system based on real-time feedback, coupled with real-time sensor monitoring and predictive modeling to simulate optimal separation conditions, allows for coordinated control and adjustment, and simultaneously selects and implements an appropriate column aging protocol.

[0130] The sample to be tested is injected into the gas chromatograph using a high-precision microsyringe. After carrier gas separation, it is detected in real time by a flame photometric detector. A high-speed data acquisition card combined with digital filtering technology dynamically adjusts the acquisition frequency based on the retention time characteristics of hydrocarbons and hydrogen sulfide and the neural network prediction model. The same preprocessing methods used for the training data are then applied to obtain real-time chromatographic data that meets the model requirements. This data is then input into the chromatographic analysis model, and through rapid forward propagation calculations, information is output, including the presence, concentration, and whether hydrogen sulfide exceeds the standard. This provides a powerful basis for monitoring and controlling hydrocarbon and hydrogen sulfide levels in refineries.

[0131] It should be understood that the size of the serial numbers of the steps in the above embodiments does not mean the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present invention.

[0132] like Figure 2 The figure shows a schematic structural diagram of a device of the present invention, comprising a memory 202, a processor 201 and an electronic device program on the memory 202, wherein the processor 201 executes the electronic device program to implement the steps of the neural network-based rapid chromatography analysis method of any of the above-mentioned embodiments.

[0133] Figure 2 A processor 201 is taken as an example.

[0134] The electronic device may further include an input device 203 and a display device 204 .

[0135] The processor 201, the memory 202, the input device 203 and the display device 204 may be connected via a bus or other means, with the bus connection being used as an example in the figure.

[0136] Memory 202, as a non-volatile electronic device-readable storage medium, can be used to store non-volatile software programs, non-volatile electronic device executable programs, and modules, such as the program instructions / modules corresponding to the neural network-based rapid chromatographic analysis method in the embodiments of the present application. Processor 201 executes the non-volatile software programs, instructions, and modules stored in memory 202 to perform various functional applications and data processing, thereby implementing the neural network-based rapid chromatographic analysis method in the above-mentioned embodiments.

[0137] The memory 202 may include a program storage area and a data storage area, wherein the program storage area may store an operating system and application programs required for at least one function; the data storage area may store data created based on the use of the neural network-based rapid chromatography analysis method, etc. In addition, the memory 202 may include a high-speed random access memory and may also include a non-volatile memory, such as at least one disk storage device, a flash memory device, or other non-volatile solid-state storage device. In some embodiments, the memory 202 may optionally include a memory remotely located relative to the processor 201, and these remote memories may be connected to a device that performs the neural network-based rapid chromatography analysis method via a network. Examples of the above-mentioned network include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and combinations thereof.

[0138] The input device 203 can receive user clicks and generate signal inputs related to user settings and function control of the neural network-based rapid chromatography analysis method. The display device 204 can include a display device such as a display screen.

[0139] The one or more modules are stored in the memory 202 and, when executed by the one or more processors 201 , execute the neural network-based rapid chromatographic analysis method in any of the above method embodiments.

[0140] When in operation, the electronic device disclosed in the present invention can execute all the steps of the above-mentioned neural network-based rapid chromatography analysis method, construct and train an efficient and accurate chromatography analysis model by collecting and preprocessing historical chromatography data, then construct and train the model based on the neural network and the training data set, and then enrich and remove impurities from the original sample to obtain a test sample with reduced noise and redundancy. The chromatograph is adjusted in a targeted manner according to the characteristics of the target substance in the test sample and the analysis requirements to optimize the separation conditions of the chromatograph, and then the real-time chromatography data produced by the chromatograph is input into the chromatography analysis model to achieve the purpose of quickly and accurately obtaining the chromatographic analysis results.

[0141] An embodiment of the present invention provides an electronic device readable storage medium storing an electronic device program / instruction. When the electronic device program / instruction is executed by the processor 201, all steps of the neural network-based rapid chromatography analysis method as described above are implemented.

[0142] In the context of the present disclosure, a storage medium may be a tangible medium that may contain or store a program for use by or in conjunction with an instruction execution system, device, or apparatus. The storage medium may be a machine-readable signal medium or a machine-readable storage medium. Alternatively, the storage medium may be a non-transitory electronic device readable storage medium, for example, a non-transitory electronic device readable storage medium may be a ROM, a random access memory (RAM), a compact disc read-only memory (CD-ROM), a magnetic tape, a floppy disk, and an optical data storage device.

[0143] An embodiment of the present invention provides an electronic device program product, including an electronic device program / instruction, which, when executed by a processor, implements the steps of the neural network-based rapid chromatographic analysis method as described above.

[0144] By running the above-mentioned electronic device program product, all steps of the neural network-based rapid chromatography analysis method as described above can be executed. By collecting and preprocessing historical chromatography data, an efficient and accurate chromatography analysis model is constructed and trained. Then, a model is constructed and trained based on the neural network and the training data set. The original sample is enriched and impurities are removed to obtain a test sample with reduced noise and redundancy. The chromatograph is adjusted in a targeted manner according to the characteristics of the target substance in the test sample and the analysis requirements to optimize the separation conditions of the chromatograph. Then, the real-time chromatography data produced by the chromatograph is input into the chromatography analysis model to achieve the purpose of quickly and accurately obtaining chromatographic analysis results.

[0145] The present invention also provides a system comprising any of the aforementioned devices, which, when executed, implements the steps of any of the aforementioned neural network-based rapid chromatographic analysis methods.

[0146] The above-described embodiments merely illustrate several implementations of the present invention, and while their descriptions are relatively specific and detailed, they should not be construed as limiting the scope of the present invention. It should be noted that a person skilled in the art would be able to make numerous variations and improvements without departing from the spirit of the present invention, all of which fall within the scope of protection of the present invention. Therefore, the scope of protection of the present invention shall be determined by the appended claims.

Claims

1. A rapid chromatographic analysis method based on neural network, characterized in that: include: Collect historical chromatographic data and perform data preprocessing to obtain a training data set; Constructing an original model based on a neural network architecture, and training the initial model according to the training data set to obtain a chromatographic analysis model; The original sample is enriched and impurity-removed to obtain the sample to be tested; Setting the chromatograph accordingly according to the characteristics of the target substance in the sample to be tested and the analysis requirements; Injecting the sample to be tested into the chromatograph and performing real-time acquisition and real-time processing, and performing real-time data feature processing to obtain real-time chromatographic data that meets the input requirements of the chromatographic analysis model; The real-time chromatographic data is input into the neural network model to realize fast forward propagation calculation and output chromatographic analysis results.

2. The neural network-based rapid chromatography analysis method according to claim 1, characterized in that: The collecting of historical chromatographic data and performing data preprocessing to obtain a training data set includes: Collect historical chromatographic data containing target substances from different industrial scenarios and laboratory simulation environments; Recording the target substance concentration range, sample collection environment, and sample pretreatment conditions of each of the historical chromatographic data, and combining the historical chromatographic data to form a training data set; The training data set is processed to meet the requirements of training the chromatographic analysis model, and the training data set is divided into a training set for training model update learning, a validation set for monitoring model performance during training, and a test set for evaluating model capabilities.

3. The neural network-based rapid chromatography analysis method according to claim 2, characterized in that: Constructing an original model based on a neural network architecture, and training the initial model according to the training data set to obtain a chromatographic analysis model, including: Build a multi-layer perceptron neural network architecture and ensure that the number of neurons in the input layer is consistent with the dimensions of the pre-processed feature vector to ensure that the input data can be fully received; Set up multiple hidden layers in the neural network architecture, and determine the number of neurons in each hidden layer through repeated experiments and verification to balance the complexity of the model and computational efficiency; The number of neurons in the output layer is set according to the specific analysis task, and the initial weights and bias values ​​are assigned to the connections between the output layers to obtain the initial model of the neural network architecture. Selecting a loss function and an optimizer based on the characteristics of the target substance and analysis requirements to optimize the initial model; The sample data and chromatographic data in the training set are sequentially input into the initial model for training to obtain a chromatographic analysis model.

4. The neural network-based rapid chromatography analysis method according to claim 3, characterized in that: The chromatographic analysis model is also optimized for weights and bias values ​​by genetic algorithm.

5. The neural network-based rapid chromatography analysis method according to claim 4, characterized in that: The chromatographic analysis model further optimizes weights and bias values ​​using a genetic algorithm, including: Adopting a hierarchical and segmented coding strategy, the weights and bias parameters of the neural network are layered according to the network layer, and each layer is segmented and coded according to the connection relationship of neurons; Set the fitness function according to the requirements of chromatographic analysis accuracy, stability and analysis speed; The fitness value is calculated by segmented coding and fitness function, and genetic operation is performed according to the fitness value to optimize the weight and bias value in the chromatographic analysis model.

6. The neural network-based rapid chromatography analysis method according to claim 2, characterized in that: After constructing and training the chromatographic analysis model based on the neural network architecture and the training data set, a model evaluation step is performed, specifically including: Set identification tasks and analysis tasks and corresponding indicators through the data of the test set; enabling the chromatographic analysis model to perform an identification task and an analysis task respectively, evaluating the identification performance and the analysis performance of the chromatographic analysis model in chromatographic analysis, and obtaining an evaluation result; If the evaluation result does not meet the preset requirements, the cause is analyzed in depth to obtain analysis results, and the chromatographic analysis model is adjusted in a targeted manner based on the analysis results.

7. The neural network-based rapid chromatography analysis method according to claim 1, characterized in that: The chromatograph is configured accordingly according to the characteristics of the target substance in the sample to be tested and the analysis requirements, including: Select the chromatographic column and aging plan based on the characteristics of the target substance and the analysis requirements, and perform the corresponding aging plan on the chromatographic column; Calculate the target intercolumn temperature, target intracolumn pressure and target carrier gas flow rate of the chromatograph based on the properties of the target substance, carrier gas type, chromatographic column specifications and target analytical accuracy; Adjust the status of the chromatograph until the target intercolumn temperature, target intracolumn pressure and target carrier gas flow rate are met.

8. The neural network-based rapid chromatography analysis method according to claim 7, characterized in that: The chromatographic column and aging scheme are selected based on the characteristics and analysis requirements of the target substance, and the corresponding aging scheme is performed on the chromatographic column, including: Multiple sensors are set up in the chromatograph and combined with the control module to form a control system based on real-time feedback; Real-time monitoring of the chromatograph's column pressure, intercolumn temperature, and carrier gas flow rate; Based on the data of the sample to be tested and the chromatograph, a prediction model is established to simulate and calculate the optimal separation conditions of the chromatograph; Compare and analyze the real-time monitoring data with the optimal separation conditions, obtain the comparison results and feed them back to the control system; The control system adjusts the column oven heating power, pressure regulating valve and carrier gas flow regulating valve of the chromatograph according to the comparison results to achieve coordinated control of inter-column temperature, intra-column pressure and carrier gas flow.

9. A device comprising a memory, a processor, and an electronic device program on the memory, characterized in that: The processor executes the electronic device program to implement the steps of the neural network-based rapid chromatography analysis method according to any one of claims 1 to 8.

10. A system, characterized in that: The system comprises the device according to claim 9, and when executed, implements the steps of the neural network-based rapid chromatography analysis method according to any one of claims 1 to 8.

Images