Chromatography column housing unit and optimization method

Abstract

The application discloses a chromatographic column box unit and an optimization method, and belongs to the technical field of chromatographic detection. The application provides a chromatographic column box unit, which comprises a data acquisition board, and the data acquisition board is integrated with the following units: a signal isolation unit for blocking digital circuit interference; a power supply isolation unit for providing independent pure power supply; a precision amplification circuit composed of an instrument amplifier and a high-stability reference voltage source; and a high-precision ADC acquisition module adopting a sigma-delta technology. The application further comprises the following: a microprocessor in the data acquisition board executes an optimized Kalman filtering algorithm to perform real-time low-delay denoising processing on data output by the ADC acquisition module. Through the collaborative design of hardware physical isolation and software algorithm optimization, the application systematically suppresses digital, power supply and random noise, significantly reduces baseline noise, and thus can realize high-sensitivity and high-precision detection of low-concentration (≤1 μL / L) gas, and improves the anti-interference ability and stability of the system.

Description

TECHNICAL FIELD

[0001] The application relates to the technical field of oil chromatographic detection, and in particular to a chromatographic column tank unit and an optimization method. BACKGROUND

[0002] The oil chromatographic analysis technology determines whether there is partial overheating, discharge and other faults in the internal part of the power equipment by detecting the components and content of the dissolved gas in the insulating oil, and is a key technology for power operation and maintenance. The column tank unit is the core of the oil chromatographic analyzer, and the performance thereof directly determines the accuracy and stability of the detection result. However, the existing column tank unit has the following technical bottlenecks:

[0003] Large baseline noise fluctuation: the digital part and the analog part in the circuit system are integrated on one circuit board, and the interference such as the high-speed switching pulse of the digital circuit and the power ripple causes the AD collected data to fluctuate in a sawtooth shape, and the baseline noise is usually greater than or equal to 500nVrms, which is easy to misjudge the low-concentration (less than or equal to 1 muL / L) gas detection. Figure 1 As shown in the figure, the X-axis represents time (minute), and the Y-axis represents signal value (count), and the baseline of the Y-axis fluctuates greatly in the interval of 7674-7700, and the low-concentration gas peak signal cannot be accurately identified. Low algorithm processing efficiency: the traditional filtering algorithm has a lag of greater than or equal to 50ms, which is easy to smooth the chromatographic peak signal, and causes the low-concentration gas peak to be covered.

[0004] The existing technology usually adopts the method of improving the filtering algorithm for optimization, but the optimization effect is limited, and it is difficult to meet the high-precision detection requirement. Therefore, it is urgent to provide a chromatographic column tank unit and an optimization method to fundamentally solve the above-mentioned defects. SUMMARY

[0005] The application provides a chromatographic column tank unit and an optimization method, which are suitable for high-precision chromatographic analysis of the dissolved gas (such as H2, CO, CO2, CH4, C2H5, C2H6 and C2H2) in the insulating oil of the power equipment such as transformer oil and reactor oil, and in particular solve the problems of large detection error of low-concentration gas components and serious baseline noise interference, and can be widely applied to the power equipment state monitoring and fault diagnosis scene.

[0006] The application is realized by the following technical scheme:

[0007] A chromatographic column tank unit, comprising a column tank main body, a double-channel chromatographic column arranged in the column tank main body, a heating unit, a temperature detection unit and a gas sensor, and further comprising a data acquisition board, wherein the data acquisition board is integrated with:

[0008] A signal isolation unit connected between the digital control circuit and the analog signal acquisition circuit, used for blocking the coupling interference of the high-speed switching pulse of the digital control circuit on the analog signal;

[0009] A power isolation unit is configured to provide an independent power supply for the analog signal acquisition circuit, and to realize isolation of the mainboard power supply and the analog power supply.

[0010] A precision amplification circuit is connected to the output of the gas sensor, and includes an instrument amplifier and a reference voltage source.

[0011] An ADC acquisition module is connected to the output of the instrument amplifier, and is configured to convert the analog signal into a digital signal.

[0012] The microprocessor in the data acquisition board is configured to execute an optimized Kalman filtering algorithm, including the following steps:

[0013] The data output by the ADC acquisition module is taken as an observation value to construct a one-dimensional state space model.

[0014] Based on the one-dimensional state space model, a forgetting factor is introduced to update a state covariance matrix P, a prediction residual is used to dynamically adjust a process noise covariance matrix Q, and a residual sliding window variance is used to dynamically adjust an observation noise covariance matrix R.

[0015] Based on the adjusted state covariance matrix P, the process noise covariance matrix Q, and the observation noise covariance matrix R, a Kalman gain K is calculated.

[0016] The Kalman gain K is subjected to amplitude limiting processing, and the data output by the ADC acquisition module is subjected to real-time filtering using the optimized Kalman filtering algorithm.

[0017] By adopting the above technical solution, the signal and power isolation unit, the precision amplification circuit, and the high-precision ADC acquisition module are integrated on the data acquisition board, and the optimized Kalman filtering algorithm is combined, so that the coupling interference of the digital circuit on the analog signal, the power noise interference, and the random noise problem in the data acquisition process are systematically solved from the two dimensions of hardware physical isolation and software algorithm optimization. This scheme realizes high signal-to-noise ratio, high-precision acquisition, and low-delay processing of the weak signal of the gas sensor, can significantly reduce the baseline noise from ≥500nVrms in the traditional technology to within 61.3nVrms, and provides a solid foundation for accurately identifying gas components with a concentration as low as 1 μL / L. This scheme fundamentally solves the core technical pain point that the low-concentration component detection is inaccurate or missed due to large baseline noise in the prior art.

[0018] Optionally, the step of constructing a one-dimensional state space model comprises: constructing a one-dimensional state equation x k =x k-1 +w k and an observation equation z k =x k +v k , wherein x k is a real signal state value at a current time, x k-1 is a real signal state value at a previous time, w k is a process noise value ， z k is an ADC acquisition value, and v k is a measurement noise value.

[0019] The step of updating the state covariance matrix P by introducing a forgetting factor comprises: introducing an adaptive forgetting factor λ, wherein 0.95≤λ≤1; constructing an update formula of the state covariance matrix P: P k - =λP k-1 +Q; and updating the state covariance matrix P k - =λP k-1 +Q k - , wherein P k-1 is a state covariance matrix at a previous time.

[0020] The step of dynamically adjusting the process noise covariance matrix Q by a prediction residual comprises: calculating a residual ε k in real time, wherein ε k =z k -x k - When ε , the value of Q is increased according to an exponential function, wherein the observation value is an ADC acquisition value z k , and the prediction value is a priori estimation value x k - at a current time.

[0021] The step of dynamically adjusting the observation noise covariance matrix R by a sliding window variance of a residual comprises: calculating a sliding window variance of a residual ε k ; and dynamically adjusting the observation noise covariance matrix R k according to a ratio of the sliding window variance to a reference noise standard deviation.

[0022] The step of calculating a Kalman gain K based on the adjusted state covariance matrix P, the adjusted process noise covariance matrix Q, and the adjusted observation noise covariance matrix R comprises: a Kalman gain K calculation formula: K=P k -I(P k - +R k ).

[0023] By adopting the technical scheme, the key parameters P, Q and R of Kalman filtering are dynamically and adaptively adjusted by introducing the forgetting factor, the prediction residual and the sliding window variance, so that the filtering algorithm can intelligently adapt to the dynamic change of the signal. When the signal is stable, strong smoothing is provided to reduce the baseline noise; when the rapidly changing chromatographic peak appears, it can quickly respond to accurately track the peak shape, thereby achieving an optimal balance between efficient denoising and maintaining signal authenticity, ensuring high-fidelity filtering effect under low delay (≤10 ms).

[0024] Optionally, the data acquisition board adopts a multi-layer PCB design, which is divided into an analog signal area, a digital signal area and a power supply area, an AGND isolation line with a width of ≥0.5 mm is arranged between the analog signal area and the digital signal area, and a PE ground protection line with a width of ≥1 mm is arranged between the power supply area and other areas.

[0025] By adopting the technical scheme, the data acquisition board is finely divided into multiple layers, and isolation lines and ground protection lines with specific physical widths are arranged, which effectively isolates the high-frequency digital signal, power supply noise and extremely sensitive analog signal path from the physical space. This design maximizes the reduction of cross-coupling and electromagnetic crosstalk between signals, and constitutes another key defense line against interference at the hardware level, providing a solid hardware guarantee for achieving ultra-low baseline noise, and enhancing the stability and reliability of the overall system in complex electromagnetic environments.

[0026] Optionally, a PI type filter circuit is connected in series at the output end of the power isolation unit, the PI type filter circuit includes a tantalum capacitor and a metal film resistor connected in series, and the PI type filter circuit has a ripple rejection ratio of ≥40 dB for a frequency band of 100 Hz-1 MHz.

[0027] By adopting the above technical scheme, the PI type filter circuit is further added on the basis of power isolation, which can efficiently filter out the residual power ripple noise of a specific frequency band (100 Hz-1 MHz) at the output end of the isolated power isolation module. This design provides more pure and stable power supply (ripple noise ≤5 mV) for the analog signal acquisition circuit (especially the precision amplifier and ADC), effectively avoiding the interference of power noise on the precision amplification and high-precision ADC conversion process, thereby further improving the precision and stability of signal acquisition, which is an important link in reducing the overall noise of the system.

[0028] Optionally, the data acquisition board is further integrated with a silicon-controlled temperature circuit, including a bidirectional silicon-controlled rectifier and a zero-crossing detection module, the zero-crossing detection module is based on an optical coupling isolation design and is used for detecting the zero-crossing point of an alternating current power supply; the microprocessor on the data acquisition board is configured to generate a trigger pulse with an adjustable width according to a temperature deviation and control the conduction of the bidirectional silicon-controlled rectifier at the zero-crossing point.

[0029] By adopting the technical solution, the silicon-controlled temperature circuit based on zero-crossing detection is integrated, the silicon-controlled rectifier is triggered to conduct at the moment when the alternating current power supply voltage is zero, and the "soft switching" control of the heating unit power is realized. This control method effectively avoids a large number of high-order harmonics and strong electromagnetic interference (EMI) generated by traditional phase control or random triggering, eliminates the possibility of secondary pollution of the temperature control system itself to the precision signal acquisition circuit from the source, and ensures that the precise temperature control is realized while maintaining the low-noise characteristics of the entire system.

[0030] Optionally, the microprocessor in the data acquisition board is configured to execute the following algorithm:

[0031] The incremental PID temperature control algorithm combines the double-point feedback of the sample inlet temperature and the sample outlet temperature collected by the temperature detection unit, dynamically adjusts the output of the heating unit to eliminate the temperature difference when the temperature difference between the sample inlet temperature and the sample outlet temperature exceeds a preset threshold, and the preset threshold is 0.03℃.

[0032] The temperature compensation algorithm is based on a preset temperature-output characteristic curve model, and compensates the output data of the gas sensor; the preset temperature-output characteristic curve model is y=a0+a1T+a2T 2 +a3T 3 , where y is the compensated output, T is the temperature, a 0、 a 1、 a 2、 a3 is a fitting coefficient.

[0033] By adopting the above technical solution, by executing the incremental PID temperature control algorithm and combining the double-point temperature feedback, high-precision and uniformity control (temperature difference ≤0.03℃, fluctuation ≤±0.02℃) of the temperature field inside the column oven is realized, and the temperature gradient in the column oven is effectively eliminated; at the same time, by executing the temperature compensation algorithm based on the polynomial model, the output drift of the gas sensor caused by the change of the environmental temperature is actively corrected. The double optimization at the algorithm level cooperatively solves the two key problems of inaccurate temperature control and temperature drift that affect the accuracy of the detection result, and ensures the high accuracy and repeatability (RSD ≤1.5%) of the detection result under different working conditions.

[0034] Optionally, the inner wall of the column box body is attached with a high-temperature-resistant glass wool insulation cotton layer, and a temperature-resistant silica gel sealing ring is arranged at the sealing gap between the column box body and the box cover.

[0035] By adopting the above technical scheme, the heat preservation and insulation performance of the column box is physically strengthened by arranging the insulation cotton layer in the column box body and efficiently sealing the gap, which effectively reduces the loss of internal heat to the outside and the impact of external temperature fluctuations on the inside of the column box, creates extremely favorable passive thermodynamic conditions for the stable operation of the high-precision temperature control algorithm, further improves the stability and uniformity of the column box temperature, and effectively reduces the energy consumption of the system to maintain constant temperature.

[0036] Optionally, it further comprises a carrier gas treatment unit connected in series in the carrier gas pipeline, comprising:

[0037] A dehumidification column filled with molecular sieve is used to adsorb moisture in the carrier gas;

[0038] A decontamination column filled with a composite filler of activated carbon and alumina is used to adsorb organic and inorganic impurities in the carrier gas.

[0039] By adopting the above technical scheme, by adding a series of dehumidification and decontamination units in the carrier gas pipeline, the trace amount of moisture and organic / inorganic impurities that may exist in the carrier gas can be effectively removed, ensuring that the carrier gas entering the chromatographic column has high purity (≥99.999%) and low humidity (dew point ≤-40℃). This avoids a series of problems such as loss of stationary phase of the chromatographic column, decrease in sensitivity of the gas sensor, broadening of the chromatographic peak, tailing, baseline drift, etc. caused by poor quality of the carrier gas, ensuring the efficiency of the chromatographic separation process and the accuracy and reliability of the analysis results.

[0040] The application also provides an optimization method of a chromatographic column box unit, comprising the following steps:

[0041] By arranging a signal isolation unit between the digital control circuit and the analog signal acquisition circuit, the coupling interference of the high-speed switching pulse of the digital control circuit on the analog signal is blocked, and the analog signal acquisition circuit is powered separately through a power isolation unit;

[0042] A precision amplification circuit is arranged at the output end of the gas sensor, the precision amplification circuit comprises an instrument amplifier and a reference voltage source connected in series, the resistance value of the gain resistor of the instrument amplifier is adjusted to achieve a predetermined gain, and a high-low frequency composite filter capacitor is connected in parallel to the output end of the reference voltage source;

[0043] An ADC acquisition module is arranged at the output end of the instrument amplifier, the ADC acquisition module adopts Σ-Δ modulation technology and supports multiple filtering modes;

[0044] Performing optimized Kalman filtering algorithm on the data output by the ADC acquisition module to remove noise, the algorithm comprising: taking the data output by the ADC acquisition module as an observation value, and constructing a one-dimensional state space model;

[0045] Based on the one-dimensional state space model, the state covariance matrix P is updated by introducing a forgetting factor, the process noise covariance matrix Q is dynamically adjusted by the prediction residual, and the observation noise covariance matrix R is dynamically adjusted by the residual sliding window variance;

[0046] Based on the adjusted state covariance matrix P, process noise covariance matrix Q and observation noise covariance matrix R, the Kalman gain K is calculated;

[0047] The Kalman gain K is limited in amplitude, and the data output by the ADC acquisition module is filtered in real time using the optimized Kalman filtering algorithm.

[0048] Optionally, the following steps are further included:

[0049] Temperature control step: obtaining temperature data of the sample inlet and outlet of the column box; calculating the temperature difference between the sample inlet and outlet; when the temperature difference between the sample inlet and outlet exceeds a preset threshold, dynamically adjusting the output of the heating unit to eliminate the temperature difference using an incremental PID algorithm, the preset threshold is 0.03℃;

[0050] Temperature compensation step: a temperature-output characteristic curve model of the gas sensor is established in advance: y=a0+a1T+a2T 2 +a3T 3 , wherein y is the compensated output, T is the temperature, a 0、 a 1、 a 2、 a3 is the fitting coefficient; according to the real-time collected temperature data, the output of the gas sensor is automatically corrected using the temperature-output characteristic curve model.

[0051] In summary, the present application includes at least one of the following beneficial technical effects:

[0052] 1. Achieving ultra-low baseline noise and high sensitivity detection: through the cooperative optimization of hardware (signal isolation, power isolation, PCB partitioning) and software (optimized Kalman filtering), various noise sources are systematically suppressed, the baseline noise is reduced to an industry-leading level (61.3nVrms), thereby enabling accurate identification and quantitative analysis of low-concentration (≤1μL / L) gas components overwhelmed by noise, significantly improving the sensitivity and accuracy of detection.

[0053] 2. High precision and high stability temperature control is achieved: through high precision PID temperature control algorithm combined with double-point feedback, active temperature compensation algorithm and strengthened physical insulation sealing structure, the column oven internal temperature field is extremely uniform (temperature difference ≤0.03℃) and stable (fluctuation ≤±0.02℃), the influence of gas sensor temperature drift is eliminated, and the high repeatability (RSD ≤1.5%) and accuracy of the analysis results in a wide temperature range and complex environment are ensured.

[0054] A comprehensive anti-interference system is constructed, and the system is reliable: the application not only solves the electromagnetic interference in the circuit, but also avoids the interference generated by the heating system through zero-crossing temperature control technology, and eliminates the interference from the gas source through the carrier gas purification unit. This multi-dimensional and systematic anti-interference design enables the device to maintain high performance and stable operation in harsh industrial sites such as strong electric fields, and has strong environmental adaptability and reliability.

[0055] The method is versatile: the optimized method can be extended to other types of column oven units (such as column oven optimization of gas chromatography and liquid chromatography), and has wide application prospects. BRIEF DESCRIPTION OF DRAWINGS

[0056] Figure 1 is the baseline spectrum in the related art;

[0057] Figure 2 is the overall structure diagram of the column oven unit in the application;

[0058] Figure 3 is the principle diagram of the precision amplification circuit of the application;

[0059] Figure 4 is the PCB layout diagram of a data acquisition board provided by the application;

[0060] Figure 5 is the hardware structure diagram of a high-precision data acquisition board provided by the application;

[0061] Figure 6 is the flowchart of the Kalman filtering algorithm provided by the application;

[0062] Figure 7 is the comparison diagram of the output of the thyristor temperature control circuit with and without temperature compensation algorithm provided by the application;

[0063] Figure 8 is the chromatogram containing CO2, C2H5, C2H6, C2H2 components detected by the application;

[0064] Figure 9 is the chromatogram containing H2, CO, CH4 components detected by the application.

[0065] Reference Signs List:

[0066] 100, column box body; 110, double-channel chromatographic column; 120, heating unit; 130, temperature detection unit; 140, data acquisition board; 200, precision amplification circuit; 210, input filter network; 220, instrument amplifier; 230, reference voltage generation circuit; 240, ADC acquisition module; 300, analog signal area; 310, digital signal area; 320, power supply area. DETAILED DESCRIPTION

[0067] To make the implementation technical means, creative features, and purposes and effects of the present application easy to understand, the following will combine specific embodiments and refer to the accompanying drawings to make a further detailed description of the present application. Figures 1-9 The embodiments in the present application and the features in the embodiments can be combined with each other without conflict.

[0068] The core idea of the embodiments of the present application is that, through the deep anti-interference design (including signal isolation, power isolation, and fine PCB layout) at the hardware level on the data acquisition board of the chromatographic column box unit, and combined with the high-precision signal processing algorithm (optimized Kalman filter, incremental PID temperature control, and temperature compensation algorithm) at the software level, the system noise is cooperatively suppressed from multiple dimensions, the temperature control precision is improved, and the temperature drift is eliminated, so as to systematically improve the detection sensitivity, accuracy, and stability of the chromatographic column box unit to low-concentration dissolved gas.

[0069] Embodiment 1. The present embodiment provides a chromatographic column box unit. Please refer to Figure 2 The chromatographic column box unit includes a column box body 100, a double-channel chromatographic column 110 arranged in the column box body 100, a heating unit 120, a temperature detection unit 130, a gas sensor, and a data acquisition board 140.

[0070] The column box body 100 is a box body with good heat preservation performance, used for accommodating the core components of chromatographic analysis. Preferably, the column box body 100 is a 304 stainless steel sealed box body with a thickness of 1.5 mm, which has high temperature resistance (≥200℃).

[0071] The double-channel chromatographic column 110 is arranged in the column box body 100. For example, one channel is configured as a packed column for separating H2, O2, N2, CH4, and CO, and the other channel is configured as a capillary column for separating CO2, C2H4, C2H6, and C2H2, so as to realize the full-component separation of common fault gases in insulating oil. Preferably, the double-channel chromatographic column 110 selects a type HP-5 column with a length of 30 m, an inner diameter of 0.32 mm, and a fixed phase of 5% phenylmethyl polysiloxane, and the maximum use temperature is 325℃.

[0072] Preferably, the heating unit 120 is a 220V / 50W stainless steel heating tube installed at the bottom of the column box body 100, and power adjustment is achieved through a silicon-controlled temperature circuit. The temperature detection unit 130, for example, can be a high-precision platinum resistance (such as Pt100). The temperature detection unit 130 includes two Class A PT100 probes, which are respectively installed at the sample inlet end (50 mm away from the sample inlet) and the sample outlet end (50 mm away from the sample outlet) of the double-channel chromatographic column 110. The detection range is -200℃-650℃, the accuracy is ≤0.01℃, and the temperature data is transmitted through an RS485 bus.

[0073] Optionally, the gas sensor is a double-channel catalytic combustion sensor (not shown in the figure). The double-channel catalytic combustion sensor is arranged inside the column box body 100. It has two detection units, one for measurement (active beads), and one for reference (compensation beads without catalyst, also known as inert beads or white beads). By measuring the difference between the two, more accurate signal output can be achieved, while compensating for the effects of environmental temperature and humidity changes. Preferably, the double-channel catalytic combustion sensor is of MQ-4 type, with a detection range of 0-100% LEL, a response time of ≤10s, and a recovery time of ≤30s. The bridge power supply end is connected to an ultra-high PSRR LDO of ADI company, which can effectively suppress the influence of power fluctuations on the balance of the bridge. In theory, when there is no combustible gas on the surface of the sensor, the output voltage difference of the bridge is ≤0.1mV, which greatly reduces the noise of the sensor itself.

[0074] The key improvement of the present application lies in the integrated design of the data acquisition board 140. The data acquisition board 140 is integrated with a signal isolation unit, a power isolation unit, a precision amplification circuit 200, an ADC acquisition module 240, and a microprocessor and other digital control circuits. The digital circuit generates high-speed switching pulse signals when working. In order to prevent these digital noises from coupling to the extremely weak analog signal (usually microvolt level) path through conduction or radiation, the signal isolation unit is provided in the embodiment.

[0075] Preferably, the signal isolation unit can use a high-speed digital optocoupler (such as ADuM series chip), which is physically connected between the digital control circuit and the analog signal acquisition circuit. Its isolation voltage is ≥2500Vrms, which blocks the high-speed switching pulses (frequency ≥100MHz) generated by the CPU, clock chip, etc. in the digital control circuit from coupling and interfering with the analog signal, so that the interference noise of the analog signal acquisition circuit is reduced by ≥80%.

[0076] Preferably, the power isolation unit adopts a 1500V power isolation module, which separately supplies power to the analog signal acquisition circuit. It receives the general power supply (such as +5V) of the mainboard and outputs one or more independent power supplies (such as ±15V) for the analog circuit, which is completely isolated from the power supply of the mainboard. This design achieves complete isolation of the mainboard power supply and the analog power supply, providing a pure energy foundation for subsequent precision amplification and acquisition. This hardware isolation design is the key line of defense for significantly reducing the baseline noise from ≥500nVrms to below 61.3nVrms.

[0077] In a preferred embodiment, a PI type filter circuit is connected in series with the output of the power isolation unit. The PI type filter circuit includes a connected tantalum capacitor and a metal film resistor. Preferably, the tantalum capacitor is a 100μF / 16V tantalum capacitor, and the metal film resistor is a 10Ω / 1W metal film resistor. The PI type filter circuit has a ripple rejection ratio ≥40dB for the 100Hz-1MHz frequency band, making the DC power supply to the analog signal acquisition circuit more pure (e.g. ripple noise ≤5mV), thereby further improving the stability and accuracy of signal acquisition.

[0078] Referring to Figure 3 , the input end of the precision amplification circuit 200 is differentially connected to the output of the gas sensor. Specifically, the precision amplification circuit 200 includes an input filter network 210, an instrument amplifier 220, and a reference voltage generation circuit 230.

[0079] Referring to Figure 3 , the input filter network 210 includes an input end IN+, an input end IN-, a first input resistor R1-1, a second input resistor R2-1, a first common-mode filter capacitor C1-1, a second common-mode filter capacitor C2-1, and a differential-mode filter capacitor C3-1. The input end IN+ and the input end IN- are respectively connected to the differential signal output ends of the dual-channel catalytic combustion sensor. The input end IN+ is connected to the positive input end (+) of the instrument amplifier 220 through the first input resistor R1-1, and the input end IN- is connected to the negative input end (-) of the instrument amplifier 220 through the second input resistor R2-1.

[0080] One end of the first common-mode filter capacitor C1-1 is connected between the first input resistor R1-1 and the positive input end of the instrument amplifier 220, and the other end is grounded. One end of the second common-mode filter capacitor C2-1 is connected between the second input resistor R2-1 and the negative input end of the instrument amplifier 220, and the other end is grounded. The differential-mode filter capacitor C3-1 is connected across the positive input end and the negative input end of the instrument amplifier 220.

[0081] The input filter network 210 collectively constitutes a differential low-pass filter, which is used to preliminarily filter out common-mode and differential-mode high-frequency noise in the input signal, and prevent external electromagnetic interference from entering the amplification link. In a preferred embodiment, the resistance values of R1-1 and R2-1 are 53.6 Ω, and the capacitance values of C1-1, C2-1, and C3-1 are all 10 μF.

[0082] Preferably, the instrumentation amplifier 220 is selected to be a chip with ultra-low noise and high common-mode rejection ratio, which has a common-mode rejection ratio ≥ 120 dB @ 50 / 60 Hz and an input offset voltage ≤ 10 μV. The positive-phase input end and the negative-phase input end of the instrumentation amplifier 220 are respectively connected to the output ends of the input filter network 210 described above.

[0083] The gain of the instrumentation amplifier 220 is determined by the gain setting resistor R1-2 connected between the two RG pins thereof. In the present embodiment, the resistance value of R1-2 is preferably 200 Ω, so as to set the gain of the instrumentation amplifier 220 to 100 times, thereby realizing high-multiplication and high-precision amplification of the weak differential signal.

[0084] The output end (OUT) of the instrumentation amplifier 220 is used to output the amplified analog signal. The output end is connected to the analog signal input end of the ADC acquisition module 240 through a third resistor R1-3. The reference voltage pin (REF) of the instrumentation amplifier 220 is connected to the output end of the reference voltage generation circuit 230 described below, which is used to set the direct current bias level of the amplifier output signal.

[0085] The reference voltage generation circuit 230 is used to provide a high-stability and low-noise reference voltage for the instrumentation amplifier 220 and the ADC acquisition module 240. The circuit includes a reference voltage source U1ref-2 and a voltage buffer U2ref-2.

[0086] In a preferred embodiment, the reference voltage source U1ref-2 is selected to be an ADR4525A chip of ADI Company, which can output a 2.5 V precision voltage with extremely low temperature drift (≤ 5 ppm / °C). The output end thereof is connected to the positive-phase input end of the voltage buffer U2ref-2 through a first reference resistor R1ref-2, and is connected to ground through a second reference resistor R2ref-2. A first reference filter capacitor C1ref-2 (e.g., 1 μF) is connected in parallel between the output end of the reference voltage source U1ref-2 and the ground, which is used to filter out noise in the reference voltage.

[0087] In a more preferred embodiment, in order to further improve the stability of the reference voltage, a large-capacity tantalum capacitor (e.g., 10 μF, not shown in the figure) and a small-capacity ceramic capacitor (e.g., 4.7 μF) can be connected in parallel to the output end of the reference voltage source U1ref-2, forming a high-low frequency composite filter capacitor combination, which effectively suppresses power supply ripple and noise in the full frequency band.

[0088] Optionally, the voltage buffer U2ref-2 is selected as the MAX40023 operational amplifier chip of ADI Company, the positive input end of which is connected to the output of the reference voltage source U1ref-2, and the negative input end is directly connected to the output end, forming a voltage follower. The output end serves as the output of the entire reference voltage generation circuit 230 and is connected to the REF pin of the instrument amplifier 220. The introduction of the voltage buffer U2ref-2 enhances the driving capability of the reference voltage, avoids the load effect of the subsequent circuit on the reference voltage source, and ensures the stability and consistency of the reference voltage.

[0089] In a preferred embodiment, the ADC acquisition module 240 is selected as the AD7175-2 chip of ADI Company with low-noise Σ-Δ modulation technology. The chip supports a 1 ksps sampling rate and a multi-stage filtering mode, and preferably the multi-stage filtering mode is a Sinc5+1 filtering mode. The chip acquisition noise is ≤14 nVrms, ensuring high-precision conversion of analog signals to digital signals.

[0090] The analog signal input end of the ADC acquisition module 240 is connected to the output end of the instrument amplifier 220 through a third resistor R1-3. An ADC input filter capacitor C1-3 is connected between the analog signal input end of the ADC acquisition module 240 and the ground. The third resistor R1-3 and the ADC input filter capacitor C1-3 together form an RC low-pass filter, which is used for the last filtering before the signal enters the ADC, further filtering out high-frequency noise that may be left or generated in the signal link, and ensuring the purity of the input ADC signal.

[0091] The REF pin of the ADC acquisition module 240 is connected to the output end of a reference voltage generation circuit similar to the above-described reference voltage generation circuit 230, which has a similar circuit as described above and is not described here to obtain a stable 2.5V reference voltage, thereby ensuring the quantization accuracy of the analog-to-digital signal conversion.

[0092] The digital output end (OUT) of the ADC acquisition module 240 outputs the converted digital signal to the single-chip microcomputer (such as STM32F103) on the data acquisition board 140 for subsequent Kalman filtering algorithm and temperature compensation algorithm processing.

[0093] The working principle of the precision amplification circuit 200 is as follows: when the weak differential voltage signal carrying the gas concentration information from the double-channel catalytic combustion sensor is input into the precision amplification circuit 200, first, the high-frequency interference is filtered out through the input filter network 210; then, the clean differential signal is amplified by 100 times by the instrument amplifier 220 on the basis of the super-stable reference voltage (provided by the reference voltage generation circuit 230) with low noise; after the amplified signal is filtered by a first-order RC low-pass filter, it is sent to the high-precision ADC acquisition module 240 for digital conversion. Through the selection of precision devices, multi-stage filtering design, and the introduction of high-stability reference voltage, the whole signal link forms a full-process low-noise processing scheme from analog signal acquisition, amplification to digital signal conversion, ensuring accurate capture and quantization of the μL / L level low-concentration gas signal, and providing high-quality raw data for subsequent algorithm processing.

[0094] Further, the data acquisition board 140 adopts a multi-layer PCB design, such as a four-layer board design, and the four layers are signal layer, ground, power supply, and signal layer.

[0095] Referring to Figure 4 , the top and bottom layers of the PCB include three independent regions: an analog signal region 300, a digital signal region 310, and a power supply region 320. Only sensitive devices such as the precision amplification circuit 200, the ADC acquisition module 240, and the reference voltage source are placed in the analog signal region 300; high-frequency noise sources such as microprocessors, memories, and communication interfaces are placed in the digital signal region 310; and power supply isolation modules and low-dropout linear voltage stabilizers (LDOs) are placed in the power supply region 320.

[0096] Among them, a partition trench is formed between the analog signal region 300 and the digital signal region 310 by hollowing out the middle ground layer and the power supply layer, and an analog ground (AGND) isolation line with a width greater than or equal to 0.5 mm is laid on the top and bottom layers. The "ground" of the two regions is only connected at one point through an isolation device (such as a power supply isolation module), completely cutting off the path of digital ground noise polluting the analog ground. At the same time, between the power supply region 320 and other regions, a PE ground protection line with a width greater than or equal to 1 mm is provided to absorb and shield external electromagnetic interference (EMI), enhancing the overall electromagnetic compatibility (EMC). The signal lines in each region adopt a short-path, low-crossing layout, with an analog signal line spacing of ≥0.2 mm to avoid signal crosstalk and further reduce circuit noise.

[0097] Referring to Figure 5 , Figure 5 A hardware structure block diagram of a high-precision data acquisition board provided by the present application is provided, and the specific structure principle is as follows: the main board interface is the input end of the external power supply, which receives the 12V DC power supply provided by the external main control board.

[0098] The 12V power supply of the mainboard interface is divided into two paths: one is electrically connected to the input end of the 12V to 5V power isolation module. The power isolation module converts the 12V power supply into an isolated 5.0V power supply, which serves as the "dirty power supply" (i.e., the primary isolation power supply) for the entire analog signal processing area, achieving electrical isolation between the analog ground and the digital ground. The other is electrically connected to the input end of the 12V to 3.3V power conversion module, and the 3.3V power supply output by this module is specifically used to power the isolated main chip described later.

[0099] The isolated 5.0V power supply is electrically connected to the input end of a negative voltage low-noise linear voltage regulator LT3094 and the input end of a positive voltage ultra-low noise linear voltage regulator LT3045. The output end of the LT3094 outputs a stable negative voltage of -4.8V, and the output end of the first LT3045 outputs a stable positive voltage of +4.8V. The -4.8V voltage and the +4.8V voltage together form a set of high-precision positive and negative dual power supplies, which are specifically used to power the operational amplifier in the operational amplifier channel described later.

[0100] The +4.8V voltage output by the first LT3045 is also electrically connected to the input end of a second positive voltage ultra-low noise linear voltage regulator LT3045, and the output end of this LT3045 outputs a stable voltage of 2.7V. This 2.7V voltage is specifically used to provide bias or operating voltage for the gas sensor (such as a dual-channel catalytic combustion sensor) at the chromatographic column interface.

[0101] Through the above structure, a complete power supply link from the external power supply to the precision analog circuit is formed, which realizes the functions of power supply isolation, positive and negative dual power supply generation, multi-stage LDO filtering and noise reduction, and dedicated sensor power supply, ensuring the stability and low noise characteristics of the analog circuit from the source.

[0102] At the same time, the chromatographic column interface receives 2.7V power supply provided by the power isolation module and connects the gas sensor. The weak analog signal (e.g., differential signal) output by the gas sensor is output from the chromatographic column interface and is electrically connected to the signal input end of the operational amplifier channel 1 and the operational amplifier channel 2, respectively.

[0103] The reference module (e.g., built-in ADR4525A reference voltage source) is powered by the +4.8V power supply provided by the power isolation module, which outputs a highly stable reference voltage (e.g., 2.5V). The output end of the reference voltage is electrically connected to the reference voltage input end (REF pin) of the operational amplifier channel 1 and the operational amplifier channel 2, as well as the reference voltage input end of the ADC channel.

[0104] The operational amplifier channel 1 and the operational amplifier channel 2 (for example, the built-in AD8422 instrument amplifier 220) receive analog signals from the column interface, and perform differential amplification with high gain and high common-mode rejection ratio by taking the voltage provided by the reference module as the reference and using the ±4.8V dual power supply provided by the power supply isolation module. The amplified analog signals are output from the respective output terminals.

[0105] The temperature module (for example, containing a PT100 detection unit) is used to collect the environment or column box temperature, and the output temperature signal is also an analog signal.

[0106] The ADC channel (for example, the built-in AD7175-2 high-precision ADC) is provided with multiple analog input channels. The first analog input channel thereof is electrically connected to the output terminal of the operational amplifier channel 1, the second analog input channel is electrically connected to the output terminal of the operational amplifier channel 2, and the third analog input channel is electrically connected to the output terminal of the temperature module. The ADC channel takes the voltage provided by the reference module as the reference, and converts the received analog signals of multiple channels into high-resolution digital signals.

[0107] Meanwhile, the digital signal output terminal of the ADC channel is electrically connected to the input terminal of the slave end, and the slave end and the master end (for example, jointly forming an ADUM1400 signal isolation unit) realize electrical isolation transmission of signals through magnetic coupling or optical coupling. The slave end works in the analog power supply domain, and the master end works in the digital power supply domain.

[0108] The output terminal of the master end is electrically connected to the mainboard interface, and the collected digital signals are transmitted to the external host system (for example, a single-chip microcomputer or a processor) through the mainboard interface for subsequent data analysis and processing.

[0109] In summary, through the above hardware structure, the full-link isolation and low-noise design from power input to signal output are realized. The power isolation module provides a pure energy basis for signal processing; the instrument amplifier 220 effectively amplifies and accurately digitizes the weak signal under the constraint of high-precision reference; and the signal isolation module builds a solid “firewall” to prevent digital domain noise from polluting the analog domain, thereby ensuring high signal-to-noise ratio and high measurement accuracy of the entire data acquisition system.

[0110] In a preferred embodiment, the microprocessor on the data acquisition board receives digital signals from the ADC acquisition module. In order to further filter out random noise introduced during the acquisition process, while avoiding the signal lag and peak flattening problems caused by traditional filtering algorithms (such as moving average filtering), the microprocessor is configured to execute an optimized Kalman filtering algorithm, which is deeply improved for the characteristics of the chromatographic signal (slow baseline drift and fast peak change), and refers to Figure 6 , and the specific implementation logic is as follows:

[0111] S41, the data output by the ADC acquisition module is taken as an observation value to construct a one-dimensional state space model.

[0112] In order to meet the stringent requirement that the filtering lag time is less than or equal to 10 ms, the embodiment simplifies the traditional high-dimensional matrix operation into scalar operation. A one-dimensional state space model is constructed, and the state vector only contains the ADC acquisition value.

[0113] Preferably, the step of constructing the one-dimensional state space model comprises: constructing a one-dimensional state equation x k =x k-1 +w k and an observation equation z k =x k +v k , wherein x k is a real signal state value at a current time, x k-1 is a real signal state value at a previous time, w k is a process noise value ， z k is an ADC acquisition value, and v k is a measurement noise value.

[0114] Suppose that the ADC sampling frequency is fs, and the execution time of the single filtering algorithm controlled by the microprocessor satisfies ≤10 ms-1 / fs, so as to reserve sufficient sampling transmission time and ensure the real-time performance of data processing. For example, if fs=1 kHz, the execution time needs to be ≤9 ms.

[0115] S42, based on the one-dimensional state space model, the state covariance matrix P is updated by introducing a forgetting factor, the process noise covariance matrix Q is dynamically adjusted through the prediction residual, and the observation noise covariance matrix R is dynamically adjusted through the residual sliding window variance.

[0116] Preferably, the step of updating the state covariance matrix P by introducing a forgetting factor comprises:

[0117] The microprocessor sets the initial value P0 according to the range [0, V max ] of the ADC acquisition module during initialization, P0=(V max / 2) 2 , so as to cover the maximum error range of the initial estimation.

[0118] An adaptive forgetting factor λ is introduced, wherein 0.95≤λ≤1; the update formula of the state covariance matrix P is constructed: P k - =λP k-1 +Q, P k-1 is the state covariance matrix at a previous time; and P k - =λPk-1 +Q update state covariance matrix P k - When the signal is detected to be stationary, λ takes a value close to 1, making P decay slowly; when the signal changes rapidly, λ decreases, accelerating the decay of P, thus increasing the sensitivity of the algorithm to changes in the system state.

[0119] Preferably, the step of dynamically adjusting the process noise covariance matrix Q through the prediction residual comprises:

[0120] Real-time calculation of the residual ε of the observation value and the prediction value k , ε k =z k -x k - When , the value of Q is increased exponentially, wherein the observation value is the ADC acquisition value z k , and the prediction value is the prior estimate value x k - .

[0121] Specifically, the value of Q represents the noise of the "unmodeled dynamics" of the system. The embodiment adopts the strategy of "offline calibration + online adjustment":

[0122] Offline calibration: collect ADC static data (such as short-circuit ADC input, collect pure noise), calculate the variance σ 2 static of the data; collect ADC dynamic calibration data (input known sine / jump signal), calculate the variance σ 2 dynamic of the prediction error; combine σ 2 static and σ 2 dynamic , calculate the basic Q value: Q base =ασ 2 dynamic +(1-α)σ 2 static , wherein α is preferably 0.7.

[0123] Online adjustment: the microprocessor calculates the residual ε k =z k -x k - in real time. When , it is determined that the signal has a sudden change (such as a chromatographic peak), at which time the value of Q is temporarily increased according to the formula (β is an adjustment coefficient, for example, 0.1), so that the filter quickly tracks the signal change; when the residual returns to normal, the value of Q is exponentially decayed back to Q base .

[0124] Preferably, the step of dynamically adjusting the observation noise covariance matrix R through the residual sliding window variance includes:

[0125] Calculate the residual ε k The sliding window variance; based on the ratio of the sliding window variance to the standard deviation of the reference noise, the observation noise covariance matrix R is dynamically adjusted. k .

[0126] Specifically, the baseline noise standard deviation σ of the hardware noise is calculated through multiple sets of static sampling by the ADC. 2 ADC Set the basic observation noise covariance R of the Kalman filter. base R is usually configured base Equal to or slightly greater than σ 2 ADC To ensure filtering stability, during online dynamic adjustment, the microprocessor maintains a residual sliding window of length N (e.g., N=5~10), and calculates the variance Var(ε) within the residual sliding window. k ) win The formula for calculating the real-time R value is: R k =R base Var(ε k ) win / σ 2 ADC The physical meaning of this formula is: using the ratio of the current residual fluctuation to the hardware standard noise floor as the gain coefficient, the basic parameter R is adjusted. base Perform dynamic scaling. When noise within the window increases, R... k Automatically increase R, reducing the weight of the current observation in the update process, thereby suppressing sudden disturbances; conversely, decrease R. k This improves tracking accuracy.

[0127] S43. Preferably, the step of calculating the Kalman gain K based on the adjusted state covariance matrix P, the process noise covariance matrix Q, and the observation noise covariance matrix R includes:

[0128] The Kalman gain K is calculated using the formula: K = P k - / (P k - +R k ).

[0129] S44. Preferably, the step of limiting the Kalman gain K includes:

[0130] Set the upper limit of gain K max For slowly varying signals (such as a baseline), K max Set to 0.8; for rapidly changing signals (such as the peak-forming phase), Kmax Relaxed to 0.95.

[0131] The final gain value is: K = min(K max , P k - / (P k - + R k ).

[0132] Through the Kalman filtering algorithm of multi-parameter cooperative optimization described above, the application can minimize the mean square error (MSE) in a very short time window (lag ≤ 10 ms), which not only avoids the chromatographic peak broadening and lag caused by traditional low-pass filtering, but also effectively filters out high-frequency digital noise, ensuring the accurate restoration of weak signals of low-concentration (≤ 1 μL / L) gas.

[0133] Through the synergistic effect of the above hardware and software, the embodiment can realize full-link low-noise processing of gas sensor signals from source isolation, to precise amplification, to high-precision conversion and intelligent filtering, thereby providing a high-quality data basis for subsequent gas quantitative analysis.

[0134] In addition, the embodiment further integrates a high-precision temperature control system to eliminate the influence of temperature fluctuations and drift on the detection results.

[0135] Preferably, the data acquisition board 140 also integrates a silicon-controlled temperature circuit connected with the microprocessor, which includes a bidirectional thyristor and a zero-crossing detection module. The bidirectional thyristor can use BT136 type bidirectional thyristor, and the zero-crossing detection module is based on optical coupling isolation design (such as PC817), which is used to safely detect the voltage zero-crossing point of the mains AC power supply and send an interrupt signal to the microprocessor.

[0136] The microprocessor is configured to generate a trigger pulse with adjustable width (0.5-2 ms) according to the deviation of the current temperature from the set temperature, and trigger the bidirectional thyristor to conduct at the precise time after receiving the zero-crossing interrupt signal, thereby controlling the power flowing through the heating unit 120. This "soft switching" method avoids the strong electromagnetic harmonic interference generated by traditional phase-shift power regulation, eliminating the secondary pollution of the temperature control system to the signal acquisition circuit from the source, and making the heating power regulation accuracy ≤ 1%. The effect is shown in Figure 7 As shown, the fluctuation amplitude of the sensor output in the range of -30℃ to 60℃ under 20% LEL concentration is ≤ 2%.

[0137] Further, the microprocessor is configured to execute the following two algorithms:

[0138] Incremental PID temperature control algorithm: This algorithm is used to accurately control the output power of the heating unit. An incremental PID is used instead of a positional PID. A set of optimal PID parameters can be determined through experimental tuning, such as a proportional coefficient Kp = 2.5, an integral time Ti = 10 s, and a derivative time Td = 2 s.

[0139] The temperature detection unit in this embodiment is specifically two A-level PT100 probes, which are installed at the sample inlet and outlet of the double-channel chromatographic column, respectively. The microprocessor not only pays attention to whether the temperature of a single point reaches the set value (e.g., 65°C), but also calculates the temperature difference between the sample inlet and outlet in real time; when the temperature difference exceeds a very small preset threshold (e.g., 0.03°C), the algorithm dynamically adjusts the output of the heating unit to actively eliminate the temperature gradient in the column oven, so that the internal temperature of the column oven can be stabilized within the range of ±0.02°C of the set value, and the temperature fluctuation amplitude is reduced by ≥96% compared with the traditional PID.

[0140] Temperature compensation algorithm: This algorithm is based on a pre-calibrated temperature-output characteristic curve model. For example, through experimental measurement of multiple sets of baseline output y of the gas sensor at temperature T, a polynomial curve is fitted: y = a0 + a1T + a2T 2 +a3T 3 , where y is the compensated output, T is the temperature, a 0、 a 1、 a3 are fitting coefficients. In actual work, the microprocessor calculates the baseline drift amount at the current temperature using this model according to the real-time collected column oven temperature T, and subtracts the drift amount from the original output data of the gas sensor, thereby realizing automatic correction of temperature drift. Based on this model, automatic temperature compensation of sensor output is realized, so that the fluctuation amplitude of sensor output within the range of -30°C to 60°C at 20% LEL and 60% LEL concentrations is ≤2%, and the influence of temperature change on the detection result is eliminated.

[0141] Furthermore, the physical structure and gas circuit of the column oven unit are also improved and optimized, and a comprehensive optimization system is constructed.

[0142] Specifically, the inner wall of the column oven body is tightly attached to a layer of high-temperature-resistant glass wool insulation cotton layer with a thickness of, for example, 6 mm, which has a thermal conductivity ≤0.03 W / (mK) and a heat loss rate ≤5% at room temperature, greatly reducing the heat loss inside the column oven. In addition, a temperature-resistant silicone seal ring is embedded in the sealing gap between the column oven body and its cover, so that the column oven body sealing level reaches IP54, effectively isolating the influence of external environmental temperature changes (such as ±5°C fluctuation) on the internal temperature of the column oven, ensuring uniform temperature distribution inside the column oven, and the temperature difference between any two points ≤0.05°C.

[0143] In addition, the column box unit further comprises a carrier gas treatment unit connected in series in the carrier gas pipeline, the carrier gas treatment unit comprising a dehumidification column and a impurity removal column, the dehumidification column being preferably a 13X molecular sieve dehumidification column (capacity 50 mL) with a molecular sieve pore size of 0.3 nm and an adsorption capacity for water molecules of ≥20% (mass fraction), so as to control the dew point of the carrier gas (nitrogen) to ≤-40℃, thereby avoiding water from causing loss of the stationary phase of the chromatographic column and reduction of the sensitivity of the sensor. The impurity removal column is preferably an activated carbon-aluminum oxide composite impurity removal column (capacity 100 mL), in which the activated carbon adsorbs organic impurities (such as hydrocarbons and alcohols), and the aluminum oxide adsorbs inorganic impurities such as oxygen and carbon dioxide, with an impurity removal rate of ≥99.9%, so as to make the purity of the carrier gas reach ≥99.999%, thereby avoiding interference of the impurities with the separation of the chromatographic peaks and ensuring the accuracy of the detection results.

[0144] In summary, the present application, through the multi-dimensional collaborative technical solutions of hardware circuit optimization, software algorithm upgrade, physical structure enhancement and gas path purification in the above embodiments, constructs a full-range anti-interference, high-precision and high-stability chromatographic column box unit system. It not only solves the problems of digital / analog crosstalk and power noise at the circuit level, solves the problems of signal processing delay and temperature control inaccuracy at the algorithm level, and eliminates the interference introduced by the environment and gas source at the physical and gas path levels. The finally achieved technical effect is: the baseline noise is extremely low (≤61.3nVrms), the temperature field is uniform and stable (temperature difference ≤0.03℃, fluctuation ≤±0.02℃), it can accurately and reliably detect and quantify gas components as low as 1μL / L, and it still maintains high performance in complex industrial environments, with extremely strong application value.

[0145] The test results are as follows:

[0146] Baseline noise comparison: the baseline noise before optimization is ≥500nVrms, and the baseline noise after optimization is reduced to 61.3nVrms, with an attenuation amplitude of ≥87.7%.

[0147] Temperature control precision comparison: the column box temperature fluctuation before optimization is ±0.5℃, and the fluctuation after optimization is ±0.02℃, with a precision improvement of ≥96%.

[0148] Low concentration detection capability comparison: before optimization, 1μL / L gas peaks cannot be accurately identified, and after optimization, each gas peak (such as Figure 8 、 9 After optimization, the spectrum is shown, the X axis represents time (minutes), and the Y axis represents signal value (counts), and the detection error is ≤5%.

[0149] Anti-interference capability comparison: under the interference of a 10kV strong electric field, the detection error before optimization is ≥20%, and the detection error after optimization is ≤3%, with a significant improvement in anti-interference capability.

[0150] Repeatability comparison: The same sample was detected for 10 times in succession, the relative standard deviation (RSD) before optimization was greater than or equal to 8%, and the RSD after optimization was less than or equal to 1.5%, and the repeatability was greatly improved.

[0151] Embodiment 2. The embodiment discloses an optimization method of a chromatographic column box unit, which cooperates through four core aspects of hardware optimization, algorithm optimization, structure optimization and carrier gas treatment optimization to systematically improve the performance of the chromatographic column box unit, especially to reduce the baseline noise, improve the temperature control precision and the overall anti-interference ability. The method of the embodiment can be applied to the chromatographic column box unit described in embodiment 1.

[0152] The optimization method specifically comprises the following steps:

[0153] S1, a signal isolation unit is arranged between the digital control circuit and the analog signal acquisition circuit to block the coupling interference of the high-speed switching pulse of the digital control circuit on the analog signal, and a power supply isolation unit is arranged to independently supply power for the analog signal acquisition circuit.

[0154] Specifically, a signal isolation unit (such as ADUM1400 chip) is arranged between the digital control circuit and the analog signal acquisition circuit to physically block the coupling interference of the high-speed switching pulse generated by the digital control circuit (such as microprocessor, clock, etc.) on the analog signal. At the same time, a power supply isolation unit (such as R-78E5.0-0.5 module) is arranged to independently supply power for the analog signal acquisition circuit, to realize the electrical isolation of the mainboard power supply and the analog power supply and cut off the noise conduction path from the power supply.

[0155] S2, a precision amplification circuit 200 is arranged at the output end of the gas sensor, the precision amplification circuit 200 comprises an instrument amplifier 220 and a reference voltage source connected in series, the resistance value of the gain resistor of the instrument amplifier 220 is adjusted to realize a predetermined gain, and a high-low frequency composite filter capacitor is connected in parallel to the output end of the reference voltage source.

[0156] Specifically, a precision amplification circuit 200 is arranged at the output end of the sensor (such as MQ-4 type catalytic combustion sensor). The core of the circuit is an instrument amplifier 220 (such as AD8422) and a reference voltage source (such as ADR4525A) connected in series. Preferably, the resistance value of the gain resistor (RG) of the instrument amplifier 220 is adjusted to 200Ω to realize a preset 100 times signal gain. At the same time, a high-low frequency composite filter capacitor composed of a 10μF tantalum capacitor and a 4.7μF ceramic capacitor is connected in parallel to the output end of the reference voltage source to provide an extremely stable reference voltage.

[0157] S3, setting an ADC acquisition module 240 at the output end of the instrument amplifier 220, the ADC acquisition module 240 adopts Σ-Δ modulation technology and supports multi-stage filtering mode.

[0158] Specifically, the ADC acquisition module 240 (such as AD7175-2) is arranged at the output end of the instrument amplifier 220. The module adopts Σ-Δ modulation technology and is configured to work in a mode supporting multi-stage filtering (such as Sinc5+1 filtering mode) to convert the amplified analog signal into a digital signal with high precision.

[0159] S4, performing denoising processing on the data output by the ADC acquisition module 240 by using an optimized Kalman filtering algorithm, the algorithm including: taking the data output by the ADC acquisition module 240 as an observation value, and constructing a one-dimensional state space model;

[0160] Based on the one-dimensional state space model, the state covariance matrix P is updated by introducing a forgetting factor, the process noise covariance matrix Q is dynamically adjusted by a prediction residual, and the observation noise covariance matrix R is dynamically adjusted by a residual sliding window variance;

[0161] Based on the adjusted state covariance matrix P, process noise covariance matrix Q and observation noise covariance matrix R, the Kalman gain K is calculated;

[0162] The Kalman gain K is limited in amplitude, and the data output by the ADC acquisition module 240 is filtered in real time by using the optimized Kalman filtering algorithm.

[0163] The specific implementation logic is as described above, and will not be repeated here.

[0164] S5, temperature control step: acquiring temperature data of the sample inlet end and the sample outlet end of the column box; calculating the temperature difference between the sample inlet end temperature and the sample outlet end temperature; when the temperature difference between the sample inlet end temperature and the sample outlet end temperature exceeds a preset threshold, dynamically adjusting the output of the heating unit 120 to eliminate the temperature difference by using an incremental PID algorithm, and the preset threshold is 0.03℃.

[0165] Specifically, the temperature data of at least two key points in the column box, i.e., the sample inlet end temperature and the sample outlet end temperature, are acquired in real time by the temperature detection unit 130 (such as two PT100 probes installed at the sample inlet end and the sample outlet end of the column).

[0166] Then, the temperature difference between the sample inlet end temperature and the sample outlet end temperature is calculated in the microprocessor.

[0167] When the temperature difference exceeds a preset threshold (e.g. 0.03℃), or any point temperature deviates from the set value, the temperature control algorithm is triggered immediately for adjustment. This embodiment uses an incremental PID algorithm (whose parameters are optimized and set as Kp=2.5, Ti=10s, Td=2s) to dynamically adjust the output power of the heating unit 120 to eliminate the temperature difference and temperature deviation. The combination of this double-point feedback and high-precision algorithm ensures the high uniformity (temperature difference ≤0.03℃) and stability (fluctuation ≤±0.02℃) of the temperature field inside the column oven.

[0168] More preferably, the power adjustment of the heating unit 120 adopts a zero-crossing trigger control method, i.e. triggering the silicon-controlled rectifier to conduct at the zero-crossing point of the alternating current power supply to avoid electromagnetic interference.

[0169] S6, temperature compensation step: a temperature-output characteristic curve model between temperature and gas sensor output is established in advance: y=a0+a1T+a2T 2 +a3T 3 where y is the compensated output, T is the temperature, a 0、 a 1、 a 2、 a3 is the fitting coefficient. According to the real-time collected temperature data, the output of the gas sensor is automatically corrected using the temperature-output characteristic curve model.

[0170] Specifically, a mathematical relationship between temperature and sensor output, i.e. a temperature-output characteristic curve model, is established in advance through experimental calibration. In this embodiment, the model is constructed as a cubic polynomial function: y=a0+a1T+a2T 2 +a3T 3 where y is the compensated output, T is the temperature, a 0、 a 1、 a 2、 a3 is the fitting coefficient.

[0171] In actual operation, according to the real-time collected column oven temperature data T, the microprocessor uses the preset temperature-output characteristic curve model to perform real-time automatic correction calculation on the original sensor output data obtained by the ADC acquisition module 240, and obtains the final compensated result. This step effectively eliminates the influence of temperature drift on detection accuracy.

[0172] By further performing steps S5 and S6, this method cooperatively solves the two key problems of temperature control inaccuracy and temperature drift that affect the accuracy of the detection result, providing stable and reliable operating conditions for the entire chromatographic analysis process and ensuring the high accuracy and repeatability (RSD can be ≤1.5%) of the analysis result.

[0173] Compared with the prior art, the optimized method and structure have the following significant beneficial effects:

[0174] Baseline noise is greatly reduced: By hardware isolation, precision amplifier circuit optimization and PCB layout improvement, the total circuit noise is controlled within 61.3nVrms, which can accurately identify low concentration gas peaks ≤1μL / L and avoid misjudgment.

[0175] High temperature control accuracy: The optimized incremental PID algorithm combined with structure insulation design, the column oven temperature fluctuation is ≤±0.02℃, and the temperature compensation algorithm is used to eliminate the influence of temperature on the sensor output, and the detection repeatability RSD is ≤1.5%.

[0176] Strong anti-interference ability: Signal and power isolation, PCB partition layout, and high-purity carrier gas treatment make the system still work stably in strong electric field and temperature and humidity fluctuation environment, and the detection error is ≤3%.

[0177] Stable and reliable structure: 304 stainless steel column oven main body and IP54 sealing design, high temperature resistance and corrosion resistance, heat loss rate of insulation cotton layer ≤5%, equipment service life extension ≥50%.

[0178] Method versatility: The optimized method can be extended to other types of column oven units (such as gas chromatography, liquid chromatography column oven optimization), which has broad application prospects.

[0179] The above only describes the preferred embodiments of the present application, and is not intended to limit the protection scope of the present application. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principles of the present application shall be included in the protection scope of the present application. Those skilled in the art can make various combinations, modifications or equivalent replacements to the above embodiments, and these combinations, modifications or equivalent replacements are all within the scope of the present application.

Claims

1. A chromatographic column oven unit comprising a column oven main body (100), a double passage chromatographic column (110) provided in the column oven main body (100), a heating unit (120), a temperature detecting unit (130), and a gas sensor, characterized in that, Also comprising a data acquisition board (140) integrated with: a signal isolation unit connected between the digital control circuit and the analog signal acquisition circuit, for blocking the coupling interference of the high-speed switching pulse of the digital control circuit on the analog signal; a power supply isolation unit for providing an independent power supply for the analog signal acquisition circuit, realizing the isolation of the mainboard power supply and the analog power supply; a precision amplification circuit (200) having an input end connected to the output of the gas sensor, the precision amplification circuit (200) comprising an instrument amplifier (220) and a reference voltage source; the gain resistor of the instrument amplifier (220) is configured to realize a predetermined gain, and the output end of the reference voltage source is connected in parallel with a high-low frequency composite filter capacitor combination; the reference voltage source is connected with the instrument amplifier (220); an ADC acquisition module (240) connected to the output end of the instrument amplifier (220), for converting the analog signal into a digital signal, the ADC acquisition module (240) adopts Σ-Δ modulation technology and supports multi-stage filtering mode; wherein the microprocessor in the data acquisition board (140) is configured to execute an optimized Kalman filtering algorithm, comprising the following steps: taking the data output by the ADC acquisition module (240) as an observation value, and constructing a one-dimensional state space model; based on the one-dimensional state space model, updating the state covariance matrix P by introducing a forgetting factor, dynamically adjusting the process noise covariance matrix Q through the prediction residual, and dynamically adjusting the observation noise covariance matrix R through the residual sliding window variance; based on the adjusted state covariance matrix P, process noise covariance matrix Q and observation noise covariance matrix R, calculating the Kalman gain K; limiting the Kalman gain K, and using the optimized Kalman filtering algorithm to perform real-time filtering on the data output by the ADC acquisition module (240); The step of constructing the one-dimensional state space model comprises constructing a one-dimensional state equation x k =x k-1 +w k and an observation equation z k =x k +v k , wherein x k is a real signal state value at a current moment, x k-1 is a real signal state value at a previous moment, w k is a process noise value ， z k is an ADC acquisition value, and v k is a measurement noise value; The step of updating the state covariance matrix P by introducing a forgetting factor comprises: introducing an adaptive forgetting factor λ, wherein 0.95≤λ≤1; constructing an update formula of the state covariance matrix P: P k - =λP k-1 +Q; updating the state covariance matrix P by using P k - =λP k-1 +Q k - , P k-1 is a state covariance matrix at a previous moment. The step of dynamically adjusting the process noise covariance matrix Q by the prediction residual comprises: calculating the residual ε of the observation value and the prediction value in real time k , k z k -x k - When , the Q value is increased in an exponential function, wherein the observation value is the ADC acquisition value z k , and the prediction value is the prior estimation value x k - ; The step of dynamically adjusting the observation noise covariance matrix R by the sliding window variance of the residual includes: calculating the sliding window variance of the residual ε k ; dynamically adjusting the observation noise covariance matrix R k according to the ratio of the sliding window variance to the reference noise standard deviation. The step of calculating the Kalman gain K based on the adjusted state covariance matrix P, the process noise covariance matrix Q and the observation noise covariance matrix R comprises: the formula of the Kalman gain K is K = P k - / (P k - +R k ).

2. A chromatography column enclosure unit according to claim 1, characterized in that: the data acquisition board (140) adopts a multi-layer PCB design, and its layout is divided into an analog signal area (300), a digital signal area (310) and a power area (320) which are independent of each other; an AGND isolation line with a width ≥0.5mm is arranged between the analog signal area (300) and the digital signal area (310), and a PE ground protection line with a width ≥1mm is arranged between the power area (320) and other areas.

3. A chromatography column enclosure unit according to claim 1, characterized by: The output end of the power supply isolation unit is connected in series with a PI type filter circuit, the PI type filter circuit comprises a tantalum capacitor and a metal film resistor connected in series, and the PI type filter circuit has a ripple rejection ratio ≥40dB for a frequency band of 100Hz-1MHz.

4. A chromatography column enclosure unit according to claim 1, wherein, The data acquisition board (140) is also integrated with: a silicon-controlled temperature control circuit comprising a bidirectional thyristor and a zero-crossing detection module, the zero-crossing detection module is based on an optocoupler isolation design and is used for detecting the zero-crossing point of an alternating power supply; the microprocessor on the data acquisition board (140) is configured to generate a trigger pulse with adjustable width according to the temperature deviation, and control the conduction of the bidirectional thyristor at the zero-crossing point.

5. A chromatography column enclosure unit according to claim 4, wherein, The microprocessor in the data acquisition board (140) is configured to execute the following algorithm: Incremental PID temperature control algorithm: combined with the double-point feedback of the sample inlet temperature and the sample outlet temperature collected by the temperature detection unit (130), when the temperature difference between the sample inlet temperature and the sample outlet temperature exceeds a preset threshold, dynamically adjust the output of the heating unit (120) to eliminate the temperature difference, and the preset threshold is 0.03℃; A temperature compensation algorithm: compensating the output data of the gas sensor based on a preset temperature-output characteristic curve model; the preset temperature-output characteristic curve model is y = a0 + a1T + a2T + a3T 2 +a3T 3 , wherein y is the compensated output, T is the temperature, a 0、 a 1、 a 2、 a3 is a fitting coefficient.

6. A chromatography column enclosure unit according to claim 1 or 5, characterized in that: The inner wall of the column box body (100) is attached with a high-temperature-resistant glass wool insulation layer, and a temperature-resistant silicone sealing ring is arranged at the sealing gap between the column box body (100) and its cover.

7. A chromatography column enclosure unit according to claim 1, wherein, It also includes a carrier gas treatment unit connected in series in the carrier gas pipeline, which includes: A dehumidification column filled with molecular sieve for adsorbing moisture in the carrier gas; A decontamination column filled with a composite filler of activated carbon and alumina for adsorbing organic and inorganic impurities in the carrier gas.

8. A method of optimizing a chromatography column cartridge unit, characterized by, The following steps are included: By setting a signal isolation unit between the digital control circuit and the analog signal acquisition circuit, the high-speed switching pulse of the digital control circuit is blocked from coupling and interfering with the analog signal, and the analog signal acquisition circuit is powered separately through a power isolation unit; A precision amplification circuit (200) is provided at the output end of the gas sensor, which includes an instrument amplifier (220) and a reference voltage source connected in series, the resistance value of the gain resistor of the instrument amplifier (220) is adjusted to achieve a predetermined gain, and a high-low frequency composite filter capacitor is connected in parallel to the output end of the reference voltage source; An ADC acquisition module (240) is provided at the output end of the instrument amplifier (220), which uses Σ-Δ modulation technology and supports multiple filtering modes; An optimized Kalman filtering algorithm is executed on the data output by the ADC acquisition module (240) for denoising, which includes: taking the data output by the ADC acquisition module (240) as the observation value, and constructing a one-dimensional state space model; Based on the one-dimensional state space model, the state covariance matrix P is updated by introducing a forgetting factor, the process noise covariance matrix Q is dynamically adjusted by the prediction residual, and the observation noise covariance matrix R is dynamically adjusted by the residual sliding window variance; Based on the adjusted state covariance matrix P, process noise covariance matrix Q and observation noise covariance matrix R, the Kalman gain K is calculated; The Kalman gain K is amplitude-limited, and the data output by the ADC acquisition module (240) is real-time filtered using the optimized Kalman filtering algorithm; The step of constructing the one-dimensional state space model comprises constructing a one-dimensional state equation x k =x k-1 +w k and an observation equation z k =x k +v k , wherein x k is a real signal state value at a current moment, x k-1 is a real signal state value at a previous moment, w k is a process noise value ， z k is an ADC acquisition value, and v k is a measurement noise value; The step of updating the state covariance matrix P by introducing a forgetting factor comprises: introducing an adaptive forgetting factor λ, wherein 0.95≤λ≤1; constructing an update formula of the state covariance matrix P: P k - =λP k-1 +Q; updating the state covariance matrix P by using P k - =λP k-1 +Q k - , P k-1 is a state covariance matrix at a previous moment. The step of dynamically adjusting the process noise covariance matrix Q by the prediction residual comprises: calculating the residual ε of the observation value and the prediction value in real time k , k z k -x k - When , the Q value is increased in an exponential function, wherein the observation value is the ADC acquisition value z k , and the prediction value is the prior estimation value x k - ; The step of dynamically adjusting the observation noise covariance matrix R by the sliding window variance of the residual includes: calculating the sliding window variance of the residual ε k ; and dynamically adjusting the observation noise covariance matrix R k according to the ratio of the sliding window variance to the reference noise standard deviation. The step of calculating the Kalman gain K based on the adjusted state covariance matrix P, the process noise covariance matrix Q and the observation noise covariance matrix R comprises: the formula of the Kalman gain K is K = P k - / (P k - +R k ).

9. The method of optimizing a chromatography column cartridge unit of claim 8, wherein, The following steps are also included: Temperature control step: obtain the temperature data of the column box sample inlet and outlet; calculate the temperature difference between the sample inlet temperature and the sample outlet temperature; when the temperature difference between the sample inlet temperature and the sample outlet temperature exceeds a preset threshold, use an incremental PID algorithm to dynamically adjust the output of the heating unit (120) to eliminate the temperature difference, and the preset threshold is 0.03℃; Temperature compensation step: pre-establishing a temperature-output characteristic curve model of temperature and gas sensor output: y=a0+a1T+a2T 2 +a3T 3 , wherein y is the compensated output, T is the temperature, a 0、 a 1、 a 2、 a3 is a fitting coefficient; according to the real-time collected temperature data, the output of the gas sensor is automatically corrected by using the temperature-output characteristic curve model.

Images