Calibration method and system of online gas chromatograph

Abstract

The invention relates to a calibration method and system for an online gas chromatograph, and the system comprises a quantum dot gas sensing array module which is arranged at the front end of a chromatographic sample injection port and comprises at least three core-shell structure quantum dots with different surface modifications; the dual-channel data acquisition module comprises a main detection channel and a calibration channel, and the calibration channel comprises a fiber optic spectrometer and a high-speed ADC converter; the environmental parameter monitoring module is used for acquiring temperature, humidity and pressure data in real time; and the AI dynamic correction engine is used for receiving the signals and the environmental parameters of the quantum dot array response module and outputting real-time correction factors. The quantum dots with different surface modifications generate differential response through specific interaction to form unique gas molecule fingerprints, compared with a traditional electrochemical sensor, the mixed gas recognition accuracy is improved by 3-5 times, and through an AI dynamic compensation algorithm, the gas molecule fingerprints can be identified by 3-5 times. The LSTM-GAN model analyzes the coupling relationship between the quantum dot response matrix and the environmental parameters, so that the calibration accuracy is effectively improved.

Description

Technical Field

[0001] The present invention relates to the technical field of gas chromatographs, and particularly to a calibration method and system for an on-line gas chromatograph. Background Art

[0002] A gas chromatograph is an instrument that uses chromatographic separation technology and detection technology to qualitatively and quantitatively analyze complex mixtures of multiple components. It can generally be used to analyze organic substances in soil that are thermally stable and have a boiling point not exceeding 500 °C, such as volatile organic compounds, organochlorines, organophosphorus, polycyclic aromatic hydrocarbons, phthalic acid esters, etc. There are many types of gas chromatographs with different functions, but their basic structures are similar. A gas chromatograph generally consists of a gas path system, an injection system, a separation system (chromatographic column system), a detection and temperature control system, and a recording system.

[0003] For example, a calibration method, device, system, and storage medium for an on-line gas chromatograph proposed in Chinese Patent No. 202410483805.X can calibrate multi-channel chromatographs simultaneously, improving the calibration work efficiency, fully meeting on-site use, and at the same time, fully evaluating the performance of all channels and detectors of the chromatograph, improving the accuracy of on-line gas chromatograph calibration.

[0004] However, the existing calibration of on-line gas chromatographs in the prior art has the following defects: 1. The traditional standard gas calibration method cannot achieve real-time dynamic correction; 2. Baseline drift problems caused by temperature drift and chromatographic column aging; 3. The response speed of the existing automatic calibration system is slow.

[0005] Therefore, a calibration method and system for an on-line gas chromatograph are proposed to solve the above problems. Summary of the Invention

[0006] In view of the deficiencies of the prior art, the present invention provides a calibration method and system for an on-line gas chromatograph, which has the advantages of reducing the calibration response time, being able to complete calibration without interrupting the detection process, and having a self-compensation function for chromatographic column aging, solving the problems that the traditional standard gas calibration method cannot achieve real-time dynamic correction, the baseline drift problems caused by temperature drift and chromatographic column aging, and the slow response speed of the existing automatic calibration system.

[0007] To achieve the above object, the present invention provides the following technical solution: An on-line gas chromatograph calibration system, comprising: A quantum dot gas sensing array module, arranged at the front end of the chromatograph injection port, comprising at least three kinds of core-shell structure quantum dots with different surface modifications; A dual-channel data acquisition module, including a main detection channel and a calibration channel, and the calibration channel includes an optical fiber spectrometer and a high-speed ADC converter; An environmental parameter monitoring module that collects temperature, humidity, and pressure data in real time; An AI dynamic correction engine that receives the signals and environmental parameters of the quantum dot array response module and outputs a real-time correction factor; A chromatographic parameter compensation module that injects the correction factor into the chromatographic workstation for real-time compensation. It is characterized in that.

[0008] Furthermore, the quantum dot gas sensing array module uses CdSe / ZnS core-shell structure quantum dots. The ligands modified on the surface of the quantum dots include three types of functional groups: carboxyl, amino, and mercapto. Each quantum dot is arranged in concentric circles according to the energy band gradient.

[0009] Preferably, the surface modification of the quantum dots adopts one of the quantum dots modified by -mercapto propionic acid (MPA) that are sensitive to aromatic hydrocarbons, the quantum dots modified by cysteamine hydrochloride (Cys) that are sensitive to sulfur-containing compounds, and the quantum dots modified by -mercapto undecanoic acid (MUA) that are sensitive to alkanes.

[0010] Furthermore, the main detection channel and the calibration channel of the dual-channel data acquisition module achieve time synchronization through FPGA. The synchronization accuracy reaches ±ns, and the data of the main detection channel and the calibration channel are associated by means of timestamp matching.

[0011] Furthermore, the AI dynamic correction engine includes: An LSTM time series prediction sub-network, the input of which is the quantum dot response time series data; A GAN noise feature extraction sub-network, the input of which is the environmental parameter matrix; The fusion layer dynamically weights the outputs of the two sub-networks using an attention mechanism.

[0012] Furthermore, the sub-network of the LSTM time series prediction sub-network includes a layer of bidirectional LSTM structures, the number of neurons in each layer is --, and an anti-overfitting design with a dropout rate of. is adopted.

[0013] Furthermore, the chromatographic parameter compensation module (500) realizes: Nonlinear correction of the retention time Rt: ; Dynamic compensation of the peak area A: A' = A / exp(-γt) ; Where α, β, and γ are the dynamic coefficients output by the AI model.

[0014] A calibration method for an on-line gas chromatograph, the calibration method is as follows: S1. The quantum dot gas sensing array module obtains the fingerprint information of the gas to be measured; S2. Synchronously collect the chromatographic data of the main detection channel and the response matrix of the calibration channel; S3. The environmental parameter monitoring module collects temperature, humidity, and pressure data; S4. The AI dynamic correction engine generates real-time correction factors; S5. The chromatographic parameter compensation module dynamically compensates the chromatographic peak parameters.

[0015] Preferably, the step S4 specifically includes: 1) Update the correction factor every 50 ms; 2) Trigger emergency calibration when the environmental parameter mutation exceeds the threshold; 3) Store the calibration log through the blockchain to ensure the data cannot be tampered with.

[0016] Furthermore, it also includes a model update mechanism, specifically as follows: 1) Automatically generate a calibration effect evaluation report every week; 2) Start model retraining when the evaluation accuracy rate is lower than 95% for three consecutive times; 3) Adopt federated learning to achieve multi-device collaborative optimization.

[0017] Compared with the prior art, the technical solution of this application has the following beneficial effects: 1. The present invention uses the molecular fingerprint recognition of quantum dot arrays. Quantum dots with different surface modifications produce differential responses through specific interactions to form unique gas molecule fingerprints. Compared with traditional electrochemical sensors, the recognition accuracy of the present invention for mixed gases is increased by 3-5 times. Through the AI dynamic compensation algorithm, the LSTM-GAN model analyzes the coupling relationship between the quantum dot response matrix and environmental parameters, thereby effectively improving the calibration accuracy.

[0018] 2. The present invention adopts dual-channel ns-level synchronization. The FPGA hardware synchronization ensures that the calibration signal and the chromatographic signal are strictly aligned, avoiding the millisecond-level delay of traditional software synchronization. The quantum dot fluorescence signal is collected in real time by a high-speed fiber optic spectrometer, and the AI model adopts a lightweight design and can perform real-time operations on embedded devices.

[0019] 3. Through the technical collaboration of quantum dot sensing - dual-channel synchronization - AI dynamic compensation, the present invention achieves laboratory-level accuracy meeting the NIST traceability standard. The millisecond-level response meets the requirements of industrial real-time control, and the operation and maintenance cost is reduced by more than 8%. The built-in blockchain evidence storage meets strict regulations such as GMP / EPA. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] Figure 1 It is a framework diagram of a calibration system for an on-line gas chromatograph of the present invention; Figure 2 It is a flowchart of a calibration method for an on-line gas chromatograph of the present invention.

[0021] In the figure: 100, quantum dot gas sensing array module; 200, dual-channel data acquisition module; 210, main detection channel; 220, calibration channel; 221, fiber optic spectrometer; 222, high-speed ADC converter; 300, environmental parameter monitoring module; 400, AI dynamic correction engine; 410, LSTM time series prediction sub-network; 420, GAN noise feature extraction sub-network; 500, chromatographic parameter compensation module. Detailed implementation

[0022] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative work shall fall within the protection scope of the present invention.

[0023] Embodiment 1 Please refer to Figure 1-2 , a calibration system for an on-line gas chromatograph in this embodiment, characterized by including: Quantum dot gas sensing array module 100, arranged at the front end of the chromatographic injection port, including at least three kinds of core-shell structure quantum dots with different surface modifications. CdSe / ZnS core-shell structure quantum dots are used. The surface modification ligands of the quantum dots include three types of functional groups: carboxyl, amino, and mercapto. Each quantum dot is arranged in concentric circles according to the energy band gradient; Dual-channel data acquisition module 200, including main detection channel 210 and calibration channel 220. Calibration channel 220 includes fiber optic spectrometer 221 and high-speed ADC converter 222; Environmental parameter monitoring module 300, which collects temperature, humidity, and pressure data in real time; AI dynamic correction engine 400, which receives the signals of the quantum dot array response module 100 and environmental parameters and outputs real-time correction factors, including: LSTM time series prediction sub-network 410, used for inputting quantum dot response time series data; GAN noise feature extraction sub-network 420, used for inputting environmental parameter matrices; The fusion layer dynamically weights the outputs of the two sub-networks using an attention mechanism; Chromatographic parameter compensation module 500, which injects the correction factor into the chromatographic workstation for real-time compensation, characterized in that.

[0024] Specifically, the quantum dot surface modification uses quantum dots modified with 3-mercaptopropionic acid (MPA) to be sensitive to aromatic hydrocarbons.

[0025] In this embodiment, the main detection channel 210 and the calibration channel 220 of the dual-channel data acquisition module 200 achieve time synchronization through FPGA, with a synchronization accuracy of ±10 ns. The data of the main detection channel 210 and the calibration channel 220 are associated by means of timestamp matching.

[0026] Specifically, the sub-network of the LSTM time series prediction sub-network 410 includes a three-layer bidirectional LSTM structure, with the number of neurons in each layer being 128 - 256 - 128, and an anti-overfitting design with a dropout rate of 0.2 is adopted.

[0027] In this embodiment, the chromatographic parameter compensation module (500) realizes: Nonlinear correction of the retention time Rt: ; Dynamic compensation of the peak area A: A' = A / exp(-γt) ; Among them, α, β, and γ are dynamic coefficients output by the AI model.

[0028] In this embodiment, the quantum dot modification concentration: the concentration of the MPA modification solution must be controlled at 0.08 - 0.12 M. Too low will result in incomplete modification, and too high will cause quantum dot aggregation; AI model update threshold: When the accuracy of the validation set is lower than 95% ± 2% for three consecutive times and the confidence interval p < 0.05, retraining is triggered; Emergency calibration trigger condition: The temperature change rate > 2 °C / min or the pressure fluctuation > 5 kPa / min.

[0029] A calibration method for an on-line gas chromatograph is as follows: S1. The quantum dot gas sensing array module 100 obtains the fingerprint information of the gas to be measured; S2. Synchronously collect the chromatographic data of the main detection channel 210 and the response matrix of the calibration channel 220; S3. The environmental parameter monitoring module 300 collects temperature, humidity, and pressure data; S4. The AI dynamic correction engine 400 generates a real-time correction factor, specifically including: 1) Update the correction factor every 50 ms; 2) Trigger emergency calibration when the environmental parameters mutate beyond the threshold; 3) Store the calibration log through the blockchain to ensure the data cannot be tampered with; S5. The chromatographic parameter compensation module 500 performs dynamic compensation on the chromatographic peak parameters.

[0030] Specifically, it also includes a model update mechanism, as follows: 1) Automatically generate a calibration effect evaluation report every week; 2) Retrain the model when the accuracy rate in three consecutive evaluations is lower than 95%. 3) Use federated learning to achieve multi-device collaborative optimization. Example 2 In this example, the hardware implementation of a calibration system for an on-line gas chromatograph is specifically as follows: 1) Quantum dot sensing array module CdSe / ZnS core-shell quantum dots are synthesized by the thermal injection method, with a diameter of 5 ± 0.2 nm and an adjustable PL peak value of 520 - 620 nm.

[0031] Surface modification process: MPA modification: The quantum dots are dispersed in a 0.1 M MPA ethanol solution and refluxed for 6 h under nitrogen protection. Cys modification: The quantum dots are oscillated with a 10 mM Cys aqueous solution at pH = 9 for 24 h.

[0032] Array integration: Arrange the three modified quantum dots in concentric circles as shown Figure 2 on a quartz substrate (spacing 50 μm) and connect them to the injection port through a microfluidic channel.

[0033] Detection principle: The benzene series molecules form hydrogen bonds with the carboxyl groups of the MPA-modified quantum dots, resulting in a linear relationship between the fluorescence quenching rate and the concentration with R² > 0.99. See the data in Example 1.

[0034] 2) Dual-channel data acquisition Main channel: Agilent 7890B chromatograph, sampling rate 10 Hz; Calibration channel: OceanHDX spectrometer, wavelength resolution 0.5 nm; Hardware synchronization is achieved through Xilinx Artix-7 FPGA, and the IEEE 1588 precise time protocol is used with a synchronization error < 8 ns.

[0035] 2. Software algorithm implementation 1) AI dynamic correction engine, and the model training is specifically as follows: # Core code of the LSTM-GAN hybrid architecture (implemented in PyTorch) class LSTMGAN(nn.Module): def __init__(self): super().__init__() self.lstm = nn.LSTM(input_size = 64, hidden_size = 256, num_layers = 3, bidirectional = True) self.gan = Generator(input_dim=128, output_dim=3) # Environmental noise generation self.attention = nn.MultiheadAttention(embed_dim=256, num_heads=4) def forward(self, x): lstm_out, _ = self.lstm(x) # Temporal feature extraction gan_out = self.gan(env_params) # Noise simulation attn_out, _ = self.attention(lstm_out, gan_out, gan_out) # Dynamic weighting return attn_out Training data: Collect 500 groups of chromatographic data with temperature mutations dropping from 20°C to 50°C and humidity fluctuations dropping from 30% to 80% RH, and label the retention time of NIST standard substances as the reference value.

[0036] 2) Real-time compensation process is as follows: 1) Execute every 50 ms: Read the quantum dot fluorescence intensity matrix , representing 3 types of quantum dots × 64 wavelength channels; Input into the LSTM-GAN model to output the compensation coefficient: , representing temperature and pressure compensation.

[0037] 2) Chromatographic parameter correction: Retention time correction: , where τ is the chromatographic column aging time constant; Peak area compensation: 3. Typical example: Petrochemical VOCs detection Test conditions: Sample: Simulated oil gas containing benzene (1.2 ppm), toluene (0.8 ppm), and xylene (0.5 ppm); Environmental interference: Temperature step change (dropping from 25°C to 40°C); Detection results: In summary, the present invention uses molecular fingerprint recognition of quantum dot arrays. Quantum dots with different surface modifications produce differential responses through specific interactions to form unique gas molecule fingerprints. Compared with traditional electrochemical sensors, the recognition accuracy of the present invention for mixed gases is increased by 3 to 5 times. Through the AI dynamic compensation algorithm, the LSTM-GAN model analyzes the coupling relationship between the quantum dot response matrix and environmental parameters, thereby effectively improving the calibration accuracy. With dual-channel ns-level synchronization, the FPGA hardware synchronization ensures that the calibration signal is strictly aligned with the chromatographic signal, avoiding the millisecond-level delay of traditional software synchronization. The quantum dot fluorescence signal is collected in real time by a high-speed fiber optic spectrometer, and the AI model adopts a lightweight design and can perform real-time operations on embedded devices. Through the technical collaboration of quantum dot sensing - dual-channel synchronization - AI dynamic compensation, it achieves laboratory-level accuracy meeting the NIST traceable standard, with a millisecond-level response to meet the industrial real-time control requirements, and the operation and maintenance cost is reduced by more than 80%. The built-in blockchain evidence storage meets strict regulations such as GMP / EPA.

[0038] It should be noted that in this article, relational terms such as first and second are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations. Moreover, the term "comprising", "including" or any other variant thereof is intended to cover non-exclusive inclusion, so that a process, method, article or device comprising a series of elements not only includes those elements, but also includes other elements not expressly listed, or elements inherent to such process, method, article or device. Without further limitation, an element defined by the statement "comprising an..." does not exclude the existence of additional identical elements in the process, method, article or device comprising the element.

[0039] Although the embodiments of the present invention have been shown and described, it will be understood by those of ordinary skill in the art that various changes, modifications, substitutions and variations can be made to these embodiments without departing from the principles and spirit of the present invention. The scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A calibration system for an on-line gas chromatograph, characterized in that, Including: A quantum dot gas sensing array module (100) is provided at the front end of the chromatographic injection port and includes at least three kinds of core-shell structure quantum dots with different surface modifications; A dual-channel data acquisition module (200) includes a main detection channel (210) and a calibration channel (220), and the calibration channel (220) includes an optical fiber spectrometer (221) and a high-speed ADC converter (222); An environmental parameter monitoring module (300) collects temperature, humidity, and pressure data in real time; An AI dynamic correction engine (400) receives the signals of the quantum dot array response module (100) and environmental parameters and outputs a real-time correction factor; A chromatographic parameter compensation module (500) injects the correction factor into the chromatographic workstation for real-time compensation, characterized in that.

2. The calibration system of an on-line gas chromatograph according to claim 1, wherein, The quantum dot gas sensing array module (100) uses CdSe / ZnS core-shell structure quantum dots, and the ligands for surface modification of the quantum dots include three types of functional groups: carboxyl, amino, and mercapto, and each quantum dot is arranged in concentric circles according to the energy band gradient.

3. The calibration system of an on-line gas chromatograph according to claim 2, wherein, The surface modification of the quantum dots adopts one of the quantum dots modified by 3-mercaptopropionic acid (MPA) sensitive to aromatic hydrocarbons, the quantum dots modified by cysteamine hydrochloride (Cys) sensitive to sulfur-containing compounds, and the quantum dots modified by 11-mercaptoundecanoic acid (MUA) sensitive to alkanes.

4. The calibration system of an on-line gas chromatograph according to claim 1, wherein The main detection channel (210) and the calibration channel (220) of the dual-channel data acquisition module (200) achieve time synchronization through FPGA, and the synchronization accuracy reaches ±10 ns, and the data of the main detection channel (210) and the calibration channel (220) are associated by means of timestamp matching.

5. The calibration system of an on-line gas chromatograph according to claim 1, wherein, The AI dynamic correction engine (400) includes: An LSTM time series prediction sub-network (410) for inputting quantum dot response time series data; A GAN noise feature extraction sub-network (420) for inputting an environmental parameter matrix; The fusion layer dynamically weights the outputs of the two sub-networks by using an attention mechanism.

6. The calibration system of an on-line gas chromatograph according to claim 5, characterized in that, The sub-network of the LSTM time series prediction sub-network (410) includes a 3-layer bidirectional LSTM structure, and the number of neurons in each layer is 128-256-128, and an anti-overfitting design with a dropout rate of 0.2 is adopted.

7. A calibration method and system for an on-line gas chromatograph according to claim 1, characterized in that, The chromatographic parameter compensation module (500) realizes: Nonlinear correction of retention time Rt: ; Dynamic compensation of peak area A: A' = A / exp(-γt) ; where α, β, and γ are dynamic coefficients output by the AI model.

8. A calibration method for an on-line gas chromatograph, including the calibration method of the calibration system of the on-line gas chromatograph according to any one of claims 1-7, characterized in that, The calibration method is specifically as follows: S1. The quantum dot gas sensing array module (100) obtains the fingerprint information of the gas to be measured; S2. Synchronously collect the chromatographic data of the main detection channel (210) and the response matrix of the calibration channel (220); S3. The environmental parameter monitoring module (300) collects temperature, humidity, and pressure data; S4. The AI dynamic correction engine (400) generates a real-time correction factor; S5. The chromatographic parameter compensation module (500) dynamically compensates the chromatographic peak parameters.

9. The calibration method of an on-line gas chromatograph according to claim 8, characterized in that, The specific steps of step S4 include: 1) Update the correction factor every 50 ms; 2) Trigger an emergency calibration when the environmental parameter mutation exceeds the threshold; 3) Store the calibration log through the blockchain to ensure that the data cannot be tampered with.

10. A calibration method for an on-line gas chromatograph according to claim 8, characterized in that, It also includes a model update mechanism, specifically as follows: 1) Automatically generate a calibration effect evaluation report every week; 2) Retrain the model when the accuracy rate in three consecutive evaluations is lower than 95%; 3) Achieve multi-device collaborative optimization using federated learning.

Images