A gas monitoring system based on infrared photoelectric technology and a data processing method thereof

Abstract

This invention relates to the field of electro-gas monitoring technology, specifically to a gas monitoring system and its data processing method based on infrared photoelectric technology. The system includes: simultaneously acquiring the infrared spectral signal of the gas, ambient light compensation signal, monitoring area temperature data, and hardware link characteristic parameters through a multi-dimensional acquisition module. By incorporating hardware link characteristic parameters into the data processing system, and combining multi-dimensional signal synchronous fusion, dynamic hardware link adaptation calibration, and coupling interference separation and reconstruction, this invention specifically compensates for signal distortion caused by core temperature deviation, turntable vibration, cable loss, and connector impedance changes. It accurately separates the coupling interference formed by dynamic hardware changes and environmental parameters, solving the problem in existing technologies where the dynamic characteristics of the hardware link are not considered, leading to a deviation in the mapping between the spectral signal and the true absorption characteristics of the gas.

Description

Technical Field

[0001] This invention relates to the field of electro-gas monitoring technology, specifically to a gas monitoring system and its data processing method based on infrared photoelectric technology. Background Technology

[0002] Infrared photoelectric gas monitoring technology captures characteristic spectral signals of gases using infrared detection components and combines this with data processing algorithms to analyze gas composition and concentration. It has become a key technology for industrial and environmental monitoring. The core of this technology lies in the quality of spectral signal acquisition and the accuracy of data processing. The synergy between the hardware acquisition link and the data processing algorithm directly determines the monitoring effect. The hardware link is responsible for signal acquisition and transmission, while the data processing algorithm performs signal analysis and concentration retrieval.

[0003] Existing patent CN120260734A, entitled "Data Processing System and Method for Intelligent Multi-Gas Detection Module Based on NDIR," focuses on optimizing data processing algorithms. It improves concentration prediction accuracy through wavelet transform denoising, recursive least squares method for updating model parameters, and federated learning for optimizing the global model. However, it fails to consider the impact of the dynamic characteristics of the hardware acquisition link on the spectral signal. It treats the hardware merely as a signal acquisition tool, ignoring the fact that changes in the working state of various components in the hardware link can lead to signal distortion, thereby affecting the accuracy of the algorithm.

[0004] Specifically, during the hardware acquisition and transmission of spectral signals, changes in the operating state of hardware components generate dynamic interference, which couples with environmental parameters. Temperature variations in the infrared detection component's core cause spectral response shifts; minute vibrations during turntable movement introduce signal jitter; cable shielding losses fluctuate with ambient temperature and humidity; and connector contact impedance changes slightly due to environmental corrosion. These dynamic hardware-level changes cause the acquired spectral signals to carry coupling interference. Current technologies do not incorporate the dynamic characteristics of the hardware link into the data processing system, relying solely on independent hardware protection or general algorithm denoising. This fails to accurately identify and compensate for this coupling interference, leading to a deviation in the mapping between the spectral signal and the true absorption characteristics of the gas. Even with optimized data processing algorithms, it is difficult to completely eliminate the impact of this deviation on monitoring accuracy. Summary of the Invention

[0005] To address the shortcomings of existing technologies, this invention provides a gas monitoring system and its data processing method based on infrared photoelectric technology, which solves the problems of spectral signal coupling interference and limited monitoring accuracy caused by the failure to consider the dynamic characteristics of hardware links when using existing technologies.

[0006] To achieve the above objectives, the present invention provides a gas monitoring data processing method based on infrared photoelectric technology, comprising the following steps:

[0007] S1: The infrared spectrum signal of the gas, the ambient light compensation signal, the temperature data of the monitoring area, and the hardware link characteristic parameters are acquired synchronously through the multi-dimensional acquisition module. The hardware link characteristic parameters include the working temperature of the infrared core, the vibration intensity of the turntable movement, the contact impedance of the connector, and the loss of the cable shielding layer.

[0008] S2: Based on the infrared spectral signal, ambient light compensation signal and temperature data of the monitoring area, multi-link signals are synchronously fused to generate a multi-dimensional fused signal;

[0009] S3: Use the hardware link characteristic parameters to dynamically adapt and calibrate the multi-dimensional fused signal, and output the calibrated initial spectral signal;

[0010] S4: Combining the hardware link characteristic parameters and environmental parameters, the calibrated initial spectral signal is subjected to coupling interference separation and signal reconstruction to obtain a clean spectral signal;

[0011] S5: Based on the pure spectral signal, environmental parameters, hardware link characteristic parameters, and hardware operating condition adaptation parameters, construct and dynamically update a dynamic fusion model for concentration inversion to obtain the real-time concentration value of the gas.

[0012] Furthermore, the implementation of S1 includes the following specific steps:

[0013] S11: Controls the turntable to adjust its posture, driving the infrared and multispectral sensors to work together;

[0014] S12: The infrared sensor uses a non-thermalized lens to capture the characteristic infrared spectrum of the gas and simultaneously collects temperature data of the monitoring area;

[0015] S13: The multispectral core collects ambient light signals as ambient light compensation signals;

[0016] S14: Transmit the signals of the infrared and multispectral sensors to the control box respectively through transmission cables that are compatible with the signal transmission of infrared and multispectral sensors.

[0017] S15: Real-time acquisition of infrared core operating temperature, turntable vibration intensity, connector contact impedance and cable shielding loss parameters via hardware link characteristic sensing module.

[0018] S16: Collect temperature, humidity and air pressure parameters as environmental parameters through the environmental sensing unit;

[0019] S17: All collected signals and parameters are marked with a unified timestamp and then stored synchronously.

[0020] Furthermore, the implementation of S2 includes the following specific steps:

[0021] S21: Based on the RS422 control serial port, send synchronous trigger commands to the infrared and multispectral cores to correct the acquisition time difference;

[0022] S22: Based on cable specifications and link loss parameters, time axis alignment of each link signal is performed using a signal delay compensation algorithm;

[0023] S23: Perform fusion processing based on complementary signal features on the aligned infrared spectral signal, multispectral compensation signal and infrared temperature measurement data to generate the multidimensional fused signal.

[0024] Furthermore, the implementation of S3 includes the following specific steps:

[0025] S31: Establish a real-time reference benchmark for the spectral response of the infrared core by utilizing the ambient light signal after multispectral fusion;

[0026] S32: Based on the operating temperature data of the infrared sensor, correct the spectral response shift of the infrared sensor.

[0027] S33: Dynamically adjust the acquisition frame rate of the infrared and multispectral sensors based on the vibration intensity parameters of the turntable motion;

[0028] S34: Based on connector contact impedance and cable shielding loss data, dynamically adjust the signal transmission gain of each link to compensate for signal attenuation.

[0029] Furthermore, the implementation of S4 includes the following specific steps:

[0030] S41: Perform secondary calibration on the calibrated initial spectral signal using the ambient light compensation signal;

[0031] S42: Construct a dynamic interference feature library based on hardware link characteristic parameters. The dynamic interference feature library stores interference features corresponding to core temperature deviation, turntable vibration, cable loss and connector impedance change, and associates and labels them with environmental parameter interference features.

[0032] S43: A wavelet transform algorithm based on threshold adjustment of dynamic interference feature library is adopted. The feature vectors in the dynamic interference feature library are used as the basis for dynamic threshold adjustment. Multi-scale hierarchical denoising is performed on the spectral signal to separate coupled interference components.

[0033] S44: Combine the standard gas characteristic spectral library to reconstruct the denoised signal and obtain the pure spectral signal.

[0034] Furthermore, the method also includes hardware condition adaptive learning S50, wherein S50 includes:

[0035] S501: Extract historically acquired hardware characteristic parameters, calibration data, and concentration detection results to establish a mapping model between hardware status and detection accuracy;

[0036] S502: Analyze the signal distortion patterns under different hardware loss levels and generate an adaptation strategy library;

[0037] S503: Monitor the current hardware characteristic parameters in real time, match them with the mapping model, and call the corresponding optimization parameters from the adaptation strategy library;

[0038] S504: Dynamically adjust the delay compensation coefficient of multi-link synchronous fusion, the threshold adjustment range of wavelet denoising, and the compensation weight of the dynamic fusion model using the optimization parameters.

[0039] Furthermore, the construction of the dynamic fusion model in S5 includes:

[0040] S51: Based on the Beer-Lambert law, a mapping relationship is established between pure spectral signals and environmental parameters based on correlation analysis, and hardware link attenuation compensation factor and hardware operating condition adaptation parameters are introduced to construct a multivariate nonlinear regression model.

[0041] S52: Extract the mapping relationship between hardware link characteristic parameters, operating condition adaptation parameters and spectral signal distortion through feature engineering, and embed this mapping relationship into the middle layer of the model;

[0042] S53: The recursive least squares method with weighted operating condition adaptation parameters is adopted. Newly acquired hardware characteristic parameters, environmental parameters, operating condition adaptation parameters and pure spectral signals are synchronously input into the model to dynamically update the model's attenuation compensation factor, operating condition adaptation weight and regression coefficient.

[0043] Furthermore, the concentration inversion in S5 includes:

[0044] S54: Input the updated model's processed spectral feature parameters into a hardware-linked non-negative matrix factorization algorithm to correct the absorption coefficient;

[0045] S55: Combining the gas absorption coefficient after hardware link characteristics and operating condition adaptation correction, the spectral signal of the mixed gas is separated and the characteristic spectrum of each gas is extracted.

[0046] S56: The real-time concentration value of each gas is obtained by inversion based on the extracted characteristic spectra.

[0047] Furthermore, the method also includes edge-cloud collaborative optimization S60, wherein S60 includes:

[0048] S601: The edge device packages and stores locally collected hardware link characteristic data, multi-dimensional fusion signals, concentration inversion results, model parameters, hardware calibration records and operating condition adaptation strategies, and dynamically adjusts the upload frequency based on the loss quantification evaluation results based on hardware link characteristic parameters.

[0049] S602: The cloud receives uploaded data from multiple edge devices and builds a three-dimensional adaptation database of hardware, algorithm, and working conditions;

[0050] S603: Employs federated learning algorithms to globally optimize the dynamic fusion model, generating an optimized model that adapts to different hardware link characteristics and operating parameters;

[0051] S604: The cloud sends the optimized model parameters, the updated operating condition adaptation strategy library, and hardware maintenance suggestions to each edge device. After receiving the information, the edge device updates its local model and adaptation strategy.

[0052] The present invention also provides a gas monitoring system based on infrared photoelectric technology, comprising:

[0053] A multi-dimensional acquisition module is used to simultaneously acquire infrared spectral signals of the gas, ambient light compensation signals, and temperature data of the monitoring area;

[0054] The hardware link characteristic sensing module is used to collect parameters such as the operating temperature of the infrared core, the vibration intensity of the turntable, the contact impedance of the connector, and the loss of the cable shielding layer in real time.

[0055] The multi-link signal synchronization and fusion module is used to synchronize and fuse infrared spectral signals, ambient light compensation signals, and temperature data of the monitoring area;

[0056] The hardware link dynamic adaptation and calibration module is used to calibrate multi-dimensional fused signals based on hardware link characteristic parameters;

[0057] The coupling interference processing module is used to separate and reconstruct coupling interference from the calibrated signal by combining hardware link characteristic parameters and environmental parameters.

[0058] The hardware condition self-adaptive learning module is used to establish a mapping model between hardware status and detection accuracy and generate adaptation strategies.

[0059] The dynamic fusion model module is used to build and update the concentration inversion model;

[0060] The edge-cloud collaborative optimization module is used to optimize model parameters and adaptation strategies in the cloud and update them locally.

[0061] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0062] This invention incorporates hardware link characteristic parameters into the data processing system, combining multi-dimensional signal synchronous fusion, dynamic hardware link adaptation calibration, and coupling interference separation and reconstruction to specifically compensate for signal distortion caused by core temperature deviation, turntable vibration, cable loss, and connector impedance changes. It accurately separates coupling interference caused by dynamic hardware changes and environmental parameters, solving the problem of mapping deviation between spectral signals and the true absorption characteristics of gases caused by the failure to consider the dynamic characteristics of the hardware link in existing technologies. Simultaneously, it leverages hardware condition self-adaptive learning to dynamically optimize processing parameters, constructs and updates a dynamic fusion model through improved algorithms, and achieves global model optimization through edge-cloud collaboration. This effectively improves the quality of spectral signal acquisition and the accuracy of data processing, enhances the system's adaptability to hardware losses and complex environments, extends the system's stable operating cycle, ensures the accuracy and reliability of gas concentration monitoring, and meets the practical application needs of industrial environments, environmental monitoring, and other scenarios. Attached Figure Description

[0063] Figure 1 This is a flowchart illustrating the overall gas monitoring process of the present invention.

[0064] Figure 2 This is a detailed flowchart of the hardware link characteristic perception and signal synchronization fusion of the present invention;

[0065] Figure 3 This is a flowchart of the dynamic adaptation calibration and coupling interference separation process of the present invention;

[0066] Figure 4 This is a flowchart illustrating the construction and updating process of the dynamic fusion model of the present invention.

[0067] Figure 5 This is a flowchart of the hardware operating condition self-learning and edge-cloud collaborative optimization process of the present invention;

[0068] Figure 6 This is a hardware architecture connection diagram of the gas monitoring system of the present invention;

[0069] Figure 7 This is a system module diagram of the present invention. Detailed Implementation

[0070] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0071] Example 1

[0072] Please see Figures 1-5This invention provides a gas monitoring data processing method based on infrared photoelectric technology, the method comprising the following steps:

[0073] S1: The infrared spectrum signal of the gas, the ambient light compensation signal, the temperature data of the monitoring area, and the hardware link characteristic parameters are acquired synchronously through the multi-dimensional acquisition module. The hardware link characteristic parameters include the operating temperature of the infrared core, the vibration intensity of the turntable, the contact impedance of the connector, and the loss of the cable shielding layer.

[0074] The implementation of S1 includes the following specific steps:

[0075] S11: Controls the turntable to adjust its attitude, driving the infrared and multispectral sensors to work together. The azimuth angle of the turntable can be adjusted from -100° to +100°, and the pitch angle can be adjusted from +35° to -85° (positive when tilted up). Its maximum azimuth speed and maximum pitch speed are both no less than 60° / s. This attitude adjustment capability ensures that the sensor can cover the monitoring area in all directions and acquire gas-related signals at different angles.

[0076] S12: The infrared sensor uses a calorimetric lens to capture the characteristic infrared spectrum of gases and simultaneously collect temperature data of the monitoring area. The infrared sensor uses a temperature-measuring sensor model COIN612R, which does not require blackbody calibration. It is equipped with a calorimetric lens with a focal length of 13mm. The sensor has a resolution of 640X512@12um, a field of view of 32.9°x26.59°, dimensions of Φ36*48.4mm, and a weight of 93g. Its compact and lightweight design does not affect the flexible movement of the turntable, while accurately capturing the characteristic infrared spectrum of gases and simultaneously collecting temperature data of the monitoring area.

[0077] S13: The multispectral core collects ambient light signals as ambient light compensation signals; the multispectral core uses a product with model number CA020A, which is specifically used to collect ambient light signals as ambient light compensation signals to offset the interference of ambient light changes on the acquisition of infrared spectral signals.

[0078] S14: After signal acquisition, the signals from the infrared and multispectral sensors are transmitted to the control chassis via multi-specification compatible infrared and multispectral sensors. The cable plug is YMA22T11K1P40+AP40, and the cable is TRVVSP high-flexibility shielded twisted pair drag chain / tank chain cable TRVVPS2, 12*0.2 specification, with a single finished cable outer diameter of 9mm. The cable length for different links is set according to actual deployment requirements. For example, some links have a cable length of 7.5 meters, while others have 7 meters, 5.0 meters, or 4.5 meters. All cables are waterproof, dustproof, salt spray resistant, shielded, and twisted pair, which can effectively reduce external interference and signal loss during signal transmission.

[0079] S15: During signal transmission, the infrared core operating temperature, turntable vibration intensity, connector contact impedance, and cable shielding loss parameters are collected in real time through the hardware link characteristic sensing module. The connector adopts aviation connector YMA22T1K1P40+AP40 with an operating voltage and current of 24V / 2A. The change in its contact impedance will directly affect the signal transmission quality, while the cable shielding loss is related to factors such as ambient temperature and humidity.

[0080] S16: At the same time, temperature, humidity and air pressure parameters are collected as environmental parameters through the environmental sensing unit.

[0081] S17: All collected signals and parameters are marked with a unified timestamp and then stored synchronously; the unification of timestamps can provide a basis for subsequent signal synchronization and fusion, ensuring that data from different sources remain consistent in the time dimension.

[0082] S2: Based on infrared spectral signals, ambient light compensation signals and temperature data of the monitoring area, multi-link signals are synchronously fused to generate multi-dimensional fused signals.

[0083] The implementation of S2 includes the following specific steps:

[0084] S21: In this process, a synchronization trigger command is first sent to the infrared and multispectral cores via the RS422 control serial port to correct the acquisition time difference; the control chassis is equipped with 4 RS422 control serial ports, which can meet the needs of synchronous control of multiple cores.

[0085] S21: Subsequently, based on the cable specifications and link loss parameters, the time axis of each link signal is aligned using a signal delay compensation algorithm; the specific formula for the signal delay compensation algorithm is as follows: in, The compensated signal timestamp, This is the original signal timestamp. This is a cable loss correction factor, the value of which is obtained by fitting experimental data, for example, in scenarios using TRVVSP high-flexibility shielded twisted pair drag chain tank chain cables. 1.03 is acceptable. For cable length, The signal transmission speed in the cable is 2.0 × 10⁻⁶. 8 m / s, The time delay caused by link loss is calculated from data collected by the hardware link characteristic sensing module.

[0086] S23: After time axis alignment, the aligned infrared spectral signal, multispectral compensation signal, and infrared thermometry data are fused based on complementary signal features to generate a multi-dimensional fused signal. The fusion process uses a weighted summation method, and the specific formula is as follows: in, For multi-dimensional fusion signals, , , These are the weighting coefficients for the infrared spectral signal, the multispectral compensation signal, and the infrared thermometry data, respectively. These weighting coefficients are determined using the analytic hierarchy process (AHP) and satisfy the following conditions: For example, in routine industrial environmental monitoring, 0.6 is acceptable. 0.3 is acceptable. 0.1 is acceptable. It is an infrared spectral signal. For multispectral compensation signals, This is infrared temperature measurement data.

[0087] This fusion process integrates the advantages of different types of signals, reduces the limitations of a single signal, improves signal integrity and reliability, and effectively solves the problem of poor signal coordination caused by the time difference and characteristic differences in the acquisition of signals from different links.

[0088] S3: Utilize hardware link characteristic parameters to dynamically adapt and calibrate the multi-dimensional fused signal, and output the calibrated initial spectral signal.

[0089] The implementation of S3 includes the following specific steps:

[0090] S31: First, a real-time reference benchmark for the spectral response of the infrared core is established using the ambient light signal after multispectral fusion. The establishment of the reference benchmark is based on the response characteristics of the multispectral signal under different ambient light intensities, ensuring that the reference benchmark can reflect the impact of ambient light changes on infrared spectrum acquisition in real time.

[0091] S32: Next, combining the operating temperature data of the infrared sensor, the spectral response shift of the infrared sensor is corrected; the formula for correcting the spectral response shift is: in, This is the signal after temperature offset correction. The temperature response coefficient of the infrared sensor is obtained from the sensor's factory calibration data, such as that of the COIN612R infrared sensor. 0.02 is acceptable. This refers to the operating temperature of the infrared sensor. The standard operating temperature for the infrared sensor is 25℃.

[0092] S33: Based on the vibration intensity parameters of the turntable movement, dynamically adjust the acquisition frame rate of the infrared and multispectral sensors. The frame rate adjustment is based on the mapping relationship between vibration intensity and acquisition frame rate. When the vibration intensity is low, the acquisition frame rate can be appropriately increased to obtain more data. When the vibration intensity is high, the acquisition frame rate is reduced to avoid aggravating signal jitter. For example, if the vibration intensity threshold is set to 0.5g, the acquisition frame rate is set to 30fps when the detected vibration intensity is less than this threshold, and the acquisition frame rate is adjusted to 15fps when the vibration intensity is greater than or equal to this threshold. The vibration intensity is acquired by the vibration sensor.

[0093] S34: Based on connector contact impedance and cable shielding loss data, dynamically adjust the signal transmission gain of each link to compensate for signal attenuation; the signal transmission gain adjustment formula is: in, For the adjusted signal transmission gain, The initial gain is set to 1. The standard contact impedance for the connector is 50Ω. This represents the actual contact impedance of the connector. The cable shielding loss factor is determined by the cable specification parameters, such as the TRVVSP type cable. 0.001m⁻¹ can be taken. This refers to the cable shielding loss length, which is related to the actual cable length.

[0094] Through this series of dynamic adaptation and calibration operations, signal distortion caused by changes in hardware link characteristics can be compensated in a targeted manner, making the signal closer to the true value and laying a good foundation for subsequent signal processing.

[0095] S4: Then, by combining the hardware link characteristic parameters and environmental parameters, the initial spectral signal after calibration is subjected to coupling interference separation and signal reconstruction to obtain a pure spectral signal.

[0096] The implementation of S4 includes the following specific steps:

[0097] S41: First, the initial spectral signal after calibration is recalibrated using the ambient light compensation signal; this further counteracts the interference caused by changes in ambient light. The recalibration formula is: in, This is the signal after secondary calibration. The ambient light reference signal is used, and the value is a multispectral signal collected under standard ambient light conditions. This is the actual ambient light compensation signal collected.

[0098] S42: A dynamic interference feature library is constructed based on hardware link characteristic parameters. The dynamic interference feature library stores the interference features corresponding to core temperature deviation, turntable vibration, cable loss and connector impedance changes, and associates and labels them with environmental parameter interference features. The interference features are extracted from a large amount of experimental data. For example, the interference feature corresponding to core temperature deviation is manifested as a change in amplitude of a specific frequency band of the spectral signal, and the interference feature corresponding to turntable vibration is manifested as periodic jitter of the signal.

[0099] S43: A wavelet transform algorithm based on threshold adjustment using a dynamic interference feature library is adopted. The feature vectors in the dynamic interference feature library are used as the basis for dynamic threshold adjustment to perform multi-scale hierarchical denoising of the spectral signal and separate coupled interference components. The wavelet transform algorithm based on threshold adjustment using the dynamic interference feature library is an improved wavelet transform algorithm, and the improved wavelet transform threshold formula is as follows: in, The wavelet transform threshold, The noise standard deviation is estimated using the high-frequency components of the signal. For signal length, The adjustment coefficient is dynamically determined by the feature vectors in the dynamic interference feature library. For example, when interference such as movement temperature deviation is detected, A value of 1.2 can be chosen when turntable vibration interference is detected. 1.5 is acceptable;

[0100] The wavelet denoising formula is: in, These are the denoised wavelet coefficients. These are the original wavelet coefficients.

[0101] S44: After denoising, the denoised signal is reconstructed by combining the standard gas characteristic spectrum library to obtain a pure spectral signal. The signal reconstruction process involves matching and aligning the denoised signal with the standard gas characteristic spectrum to fill in the missing parts of the signal, ensuring that the reconstructed signal can accurately reflect the true spectral characteristics of the gas.

[0102] This process effectively separates the coupling interference caused by dynamic changes in the hardware link and environmental parameters, solving the problem of accurately eliminating such interference in existing technologies and effectively improving the purity of spectral signals.

[0103] Furthermore, the method also includes hardware condition adaptive learning S50, wherein S50 includes:

[0104] S501: After obtaining the pure spectral signal, perform hardware condition adaptive learning. First, extract historically acquired hardware characteristic parameters, calibration data, and concentration detection results to establish a mapping model between hardware status and detection accuracy. The expression of the mapping model is: in, For detection accuracy, This is a vector of hardware characteristic parameters, including parameters such as the operating temperature of the infrared mechanism, the vibration intensity of the turntable movement, the contact impedance of the connectors, and the loss of the cable shielding layer. To calibrate the data vector, This is a vector of concentration detection results. The mapping function is obtained through training using the support vector machine algorithm.

[0105] S502: Analyze the signal distortion patterns under different hardware loss levels and generate an adaptation strategy library; the adaptation strategy library contains optimized parameter combinations for different hardware states, such as signal transmission gain adjustment strategies and wavelet denoising threshold adjustment strategies when the connector contact impedance increases.

[0106] S503: Monitors current hardware characteristic parameters in real time, matches them with the mapping model, and calls the corresponding optimization parameters from the adaptation strategy library; dynamically adjusts the delay compensation coefficient of multi-link synchronous fusion, the threshold adjustment range of wavelet denoising, and the compensation weight of the dynamic fusion model using the optimization parameters.

[0107] Through hardware condition self-adaptive learning, the system can automatically adjust relevant parameters according to the real-time status of the hardware, which improves the system's adaptability to hardware wear and tear and extends the system's stable working cycle.

[0108] S5: Based on pure spectral signals, environmental parameters, hardware link characteristic parameters, and hardware operating condition adaptation parameters, a dynamic fusion model for concentration inversion is constructed and dynamically updated to obtain the real-time concentration value of the gas.

[0109] The implementation of S5 includes the following specific steps:

[0110] S51: When constructing the dynamic fusion model, firstly, based on the Beer-Lambert law, a mapping relationship is established between the pure spectral signal and environmental parameters through correlation analysis. Then, a hardware link attenuation compensation factor and hardware operating condition adaptation parameters are introduced to construct a multivariate nonlinear regression model. The modified formula for the Beer-Lambert law is: in, The intensity of transmitted light. The intensity of the incident light. The gas absorption coefficient is... For gas concentration, Optical path length This is the hardware link attenuation compensation factor. Parameters adapted to hardware operating conditions This is an environmental parameter correction factor, calculated from temperature, humidity, and air pressure parameters, for example... ,in For ambient temperature, This refers to ambient air pressure.

[0111] S52: The mapping relationship between hardware link characteristic parameters, operating condition adaptation parameters and spectral signal distortion is extracted through feature engineering, and the mapping relationship is embedded into the intermediate layer of the model; the mapping relationship is extracted through principal component analysis and correlation analysis.

[0112] S53: A recursive least squares method with weighted operating condition adaptation parameters is adopted. Newly acquired hardware characteristic parameters, environmental parameters, operating condition adaptation parameters, and pure spectral signals are synchronously input into the model to dynamically update the model's attenuation compensation factor, operating condition adaptation weights, and regression coefficients. The improved recursive least squares update formula is as follows: in, These are the estimated values ​​of the model parameters for the nth iteration. Here is the gain matrix. The actual output signal is as follows. The input vector contains the pure spectral signal, environmental parameters, hardware link characteristic parameters, and hardware operating condition adaptation parameters. The forgetting factor, with a value of 0.98, is used to balance the weights of new and old data. Let covariance matrix be the variance matrix. It is an identity matrix.

[0113] S54: During concentration inversion, the updated model-processed spectral feature parameters are input into a non-negative matrix factorization algorithm that combines hardware link correction of the absorption coefficient.

[0114] S55: Combining the gas absorption coefficients corrected for hardware link characteristics and operating conditions, the spectral signals of the mixed gas are separated, and the characteristic spectra of each gas are extracted; the objective function of nonnegative matrix factorization is: in, For the spectral matrix of the mixed gas, For the characteristic spectral matrix, For the concentration matrix, For the Frobenius norm, It is an L1 norm. and This is the regularization parameter, with a value of 0.01, used to avoid overfitting.

[0115] S56: The real-time concentration value of each gas is obtained by inverting the extracted characteristic spectra. Through the construction and updating of this dynamic fusion model, the influence of hardware link characteristics, environmental parameters and operating condition adaptation parameters can be fully considered, which effectively improves the accuracy and stability of gas concentration inversion.

[0116] The method also includes edge-cloud collaborative optimization of S60, which includes:

[0117] S601: Finally, perform edge-cloud collaborative optimization. The edge device packages and stores locally collected hardware link characteristic data, multi-dimensional fusion signals, concentration inversion results, model parameters, hardware calibration records, and operating condition adaptation strategies. It also dynamically adjusts the upload frequency based on the loss quantification assessment results based on hardware link characteristic parameters. Hardware health assessment is based on the changing trend and threshold of hardware characteristic parameters. For example, when the change in hardware characteristic parameters within a certain period of time is less than the set threshold, the hardware health is determined to be good, and the upload frequency is set to once per hour. When the change is greater than or equal to the threshold, the upload frequency is adjusted to once every 15 minutes.

[0118] S602: The cloud receives uploaded data from multiple edge devices and builds a three-dimensional adaptation database of hardware, algorithm, and working conditions. The database stores monitoring data and model performance data under different hardware states, different algorithm parameter configurations, and different working conditions.

[0119] S603: Employs a federated learning algorithm to globally optimize the dynamic fusion model, generating an optimized model adapted to different hardware link characteristics and operating parameters; the federated learning model optimization formula is: in, These are the globally optimized model parameters. For the number of edge terminals, The weight of the m-th edge is determined based on the amount and quality of data at the edge, satisfying the following conditions: , These are the local model parameters for the m-th edge.

[0120] S604: The cloud distributes the optimized model parameters, updated operating condition adaptation strategy library, and hardware maintenance suggestions to each edge device. After receiving the information, the edge device updates its local model and adaptation strategy. Through collaborative optimization between the edge and the cloud, continuous iterative upgrades of the model are achieved, improving the system's adaptability and monitoring performance under different application scenarios and hardware operating conditions. At the same time, it provides data support for hardware maintenance and reduces maintenance costs.

[0121] Example 2

[0122] Please see Figures 6-7 The present invention also provides a gas monitoring system based on infrared photoelectric technology, used to implement the gas monitoring method based on infrared photoelectric technology in Embodiment 1. The system includes:

[0123] The multi-dimensional acquisition module is used to simultaneously acquire infrared spectral signals of the gas, ambient light compensation signals, and temperature data of the monitoring area. It includes a turntable, an infrared sensor, a multispectral sensor, and an environmental sensing unit. The turntable adopts a structure with azimuth and pitch adjustment functions, which can drive the infrared and multispectral sensors to achieve all-round monitoring. The infrared sensor uses the COIN612R temperature measuring sensor with a 13mm pyrometric lens, and the multispectral sensor uses the CA020A model. The environmental sensing unit uses a temperature and humidity sensor and a barometric pressure sensor to collect ambient temperature, humidity, and barometric pressure parameters.

[0124] The hardware link characteristic sensing module is used to collect infrared core operating temperature, turntable vibration intensity, connector contact impedance, and cable shielding loss parameters in real time. It includes temperature sensors, vibration sensors, impedance measurement sensors, and loss detection sensors, which collect different hardware link characteristic parameters respectively. The output signals of the sensors are transmitted to the subsequent processing module through a dedicated line.

[0125] The multi-link signal synchronization fusion module is used to synchronize and fuse infrared spectral signals, ambient light compensation signals, and temperature data of the monitoring area. It realizes the synchronous triggering of signal acquisition based on RS422 control serial port, completes the time axis alignment of each link signal through signal delay compensation algorithm, and realizes the complementary fusion of multi-dimensional signals through weighted fusion algorithm. The hardware carrier of this module is the main control switching board in the control chassis, which has the functions of signal processing and data transmission.

[0126] The hardware link dynamic adaptation calibration module is used to calibrate multi-dimensional fused signals based on hardware link characteristic parameters. It establishes a real-time reference benchmark for the infrared core spectral response and combines hardware link characteristic parameters to perform temperature offset correction, acquisition frame rate adjustment, and transmission gain compensation on the signal to ensure the accuracy and stability of the signal. The function of this module is implemented through a dedicated processing chip on the main control switching board.

[0127] The coupling interference processing module is used to separate and reconstruct coupling interference from the calibrated signal by combining hardware link characteristic parameters and environmental parameters. It constructs a dynamic interference feature library, uses an improved wavelet transform algorithm to perform multi-scale hierarchical denoising on the signal, and combines a standard gas feature spectrum library to complete the signal reconstruction to obtain a clean spectral signal. The hardware implementation of this module depends on the signal processing unit in the control chassis.

[0128] The hardware condition self-adaptive learning module is used to establish a mapping model between hardware status and detection accuracy and generate adaptation strategies. It trains the model by extracting historical data, matches the current hardware characteristic parameters in real time and calls the corresponding optimization parameters, and dynamically adjusts the relevant processing parameters of the system. The function of this module is implemented in the embedded processor of the control chassis through software algorithms.

[0129] The dynamic fusion model module is used to build and update the concentration inversion model. It constructs a multivariate nonlinear regression model based on the Beer-Lambert law, updates the model parameters using an improved recursive least squares method, and realizes gas concentration inversion through an optimized nonnegative matrix factorization algorithm. The core algorithm of this module runs on the high-performance processor of the control chassis to ensure the real-time performance of data processing.

[0130] The edge-cloud collaborative optimization module is used to achieve cloud-based optimization and local updates of model parameters and adaptation strategies. The edge part is integrated in the control chassis and is responsible for data storage, local model operation and data upload. The cloud part includes cloud servers and databases, which are responsible for receiving data from the edge, building a 3D adaptation database, performing global model optimization and distributing updated content. The edge and cloud interact with each other through the network to ensure the smooth implementation of collaborative optimization.

[0131] During system operation, various signals and parameters are first collected synchronously by the multi-dimensional acquisition module and the hardware link characteristic perception module. The collected data is then transmitted to the multi-link signal synchronization and fusion module for synchronization and fusion, generating a multi-dimensional fused signal. Subsequently, the hardware link dynamic adaptation and calibration module calibrates the fused signal based on the hardware link characteristic parameters, and the coupling interference processing module further separates the interference and reconstructs the signal to obtain a pure spectral signal. The hardware operating condition self-adaptive learning module adjusts the system processing parameters in real time, and the dynamic fusion model module performs concentration inversion based on the pure spectral signal and related parameters. Finally, the edge-cloud collaborative optimization module continuously optimizes the model and adaptation strategy.

[0132] Example 3

[0133] To enable those skilled in the art to fully understand and implement this invention, the specific implementation principle of this invention is further explained below in conjunction with a specific application scenario.

[0134] This embodiment applies the gas monitoring system and data processing method based on infrared photoelectric technology to a gas monitoring scenario in an industrial workshop. Industrial workshops contain various gases generated during production processes, some of which are toxic and flammable, posing a threat to the personal safety of workshop workers and the normal operation of production equipment. Therefore, it is necessary to monitor the composition and concentration of gases in the workshop in real time and accurately, so as to promptly identify potential risks and take corresponding measures.

[0135] During system deployment, four monitoring points were strategically set up based on the workshop's layout and area. Each monitoring point was equipped with a detection unit comprising a turntable, an infrared sensor, and a multispectral sensor. The detection units were connected to the control cabinet via dedicated cables, using TRVVSP high-flexibility shielded twisted-pair cable. Appropriate cable lengths were selected based on the distance between the monitoring point and the control cabinet to ensure stable and reliable signal transmission. The control cabinet was located in the workshop's control room for easy operation and maintenance. It was connected to the workshop's 220VAC / 50Hz AC power supply, received video signals from the four detection units via four video network ports, communicated with the detection units via four RS422 serial ports, transmitted processed image signals to the client via one network port, and provided local control functionality via one USB interface.

[0136] During system operation, the turntable adjusts its azimuth and elevation angles according to the preset monitoring trajectory, driving the infrared and multispectral sensors to perform a full-range scan of the monitoring area. The infrared sensor captures the characteristic infrared spectral signals of the gas and simultaneously collects the temperature data of the monitoring area. The multispectral sensor collects ambient light signals as ambient light compensation signals. The hardware link characteristic sensing module collects parameters such as the operating temperature of the infrared sensor, the vibration intensity of the turntable movement, the contact impedance of the connectors, and the loss of the cable shielding layer in real time. The environmental sensing unit collects the temperature, humidity, and air pressure parameters in the workshop. All collected signals and parameters are marked with a unified timestamp and then synchronously stored in the edge storage unit.

[0137] The multi-link signal synchronization fusion module sends synchronization trigger commands to four detection units via an RS422 control serial port to correct acquisition time differences. Based on cable specifications and link loss parameters, it aligns the time axes of each link signal using a signal delay compensation algorithm. Then, it complementaryly fuses the infrared spectral signal, multispectral compensated signal, and infrared temperature measurement data to generate a multi-dimensional fused signal. The hardware link dynamic adaptation and calibration module uses the multispectral fused ambient light signal to establish a real-time reference benchmark for the infrared sensor's spectral response. It corrects spectral response offsets by combining the infrared sensor's operating temperature data, adjusts the acquisition frame rate based on the turntable's vibration intensity, and adjusts the signal transmission gain based on connector contact impedance and cable shielding loss data, effectively compensating for signal distortion caused by changes in hardware link characteristics.

[0138] The coupling interference processing module uses the ambient light compensation signal to perform secondary calibration on the calibrated initial spectral signal. Based on the hardware link characteristic parameters and environmental parameters, a dynamic interference feature library is constructed. An improved wavelet transform algorithm is used to perform multi-scale hierarchical denoising on the signal, separating the coupling interference components. The denoised signal is reconstructed by combining the standard gas characteristic spectral library to obtain a pure spectral signal. This successfully solves the coupling interference problem caused by the dynamic changes of hardware links and environmental parameters in the complex environment of industrial workshops.

[0139] The hardware condition self-adaptive learning module extracts historically collected hardware characteristic parameters, calibration data, and concentration detection results, establishes a mapping model between hardware status and detection accuracy, generates an adaptation strategy library, monitors current hardware characteristic parameters in real time and matches them with the mapping model, and calls corresponding optimization parameters to dynamically adjust the system's delay compensation coefficient, wavelet denoising threshold adjustment range, and compensation weight of the dynamic fusion model, enabling the system to adapt to hardware wear and tear during long-term operation and maintain stable monitoring accuracy.

[0140] The dynamic fusion model module constructs a multivariate nonlinear regression model based on pure spectral signals, environmental parameters, hardware link characteristic parameters, and hardware operating condition adaptation parameters. It dynamically updates the model parameters using an improved recursive least squares method and employs an optimized nonnegative matrix factorization algorithm to separate the spectral signals of the mixed gas, extract the characteristic spectra of each gas, and invert them to obtain real-time concentration values. The monitored gas concentration data is transmitted to the client via the network port of the control chassis, allowing staff to view the gas concentration at each monitoring point in the workshop in real time through the client.

[0141] The edge devices dynamically adjust the data upload frequency based on hardware health assessment results, packaging and uploading hardware link characteristic data, multi-dimensional fusion signals, concentration inversion results, model parameters, hardware calibration records, and operating condition adaptation strategies to the cloud. The cloud receives the uploaded data from multiple edge devices, constructs a hardware-algorithm-operating condition three-dimensional adaptation database, and uses a federated learning algorithm to globally optimize the dynamic fusion model, generating a personalized optimized model adapted to the industrial workshop scenario. The optimized model parameters, the updated operating condition adaptation strategy library, and hardware maintenance suggestions are then distributed to the edge devices. Upon receiving these, the edge devices update their local models and adaptation strategies, achieving continuous improvement in system performance.

[0142] Table 1 shows a partial monitoring data example of a certain monitoring point in the workshop during a certain period. This example shows that the system can accurately collect various signals and parameters, and obtain stable and reliable gas concentration monitoring results through data processing methods, effectively identifying the gas composition and concentration changes in the workshop.

[0143] Table 1

[0144] In this industrial workshop application scenario, the system demonstrated excellent practicality and reliability. It was able to adapt to the complex environmental conditions and hardware operating conditions within the workshop, accurately monitor the composition and concentration of gases, and promptly detect abnormal gas concentrations, providing workers with ample reaction time and effectively reducing the risk of safety accidents. At the same time, the system's edge-cloud collaborative optimization function enabled the model to continuously iterate and upgrade, further improving monitoring performance and meeting the long-term, stable gas monitoring needs of the industrial workshop.

[0145] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus.

[0146] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A gas monitoring data processing method based on infrared photoelectric technology, characterized in that, The method includes the following steps: S1: The infrared spectrum signal of the gas, the ambient light compensation signal, the temperature data of the monitoring area, and the hardware link characteristic parameters are acquired synchronously through the multi-dimensional acquisition module. The hardware link characteristic parameters include the working temperature of the infrared core, the vibration intensity of the turntable movement, the contact impedance of the connector, and the loss of the cable shielding layer. S2: Based on the infrared spectral signal, ambient light compensation signal and temperature data of the monitoring area, multi-link signals are synchronously fused to generate a multi-dimensional fused signal; S3: Use the hardware link characteristic parameters to dynamically adapt and calibrate the multi-dimensional fused signal, and output the calibrated initial spectral signal; S4: Combining the hardware link characteristic parameters and environmental parameters, the calibrated initial spectral signal is subjected to coupling interference separation and signal reconstruction to obtain a clean spectral signal; S5: Based on the pure spectral signal, environmental parameters, hardware link characteristic parameters, and hardware operating condition adaptation parameters, construct and dynamically update a dynamic fusion model for concentration inversion to obtain the real-time concentration value of the gas.

2. The gas monitoring data processing method based on infrared photoelectric technology according to claim 1, characterized in that, The implementation of S1 includes the following specific steps: S11: Controls the turntable to adjust its posture, driving the infrared and multispectral sensors to work together; S12: The infrared sensor uses a non-thermalized lens to capture the characteristic infrared spectrum of the gas and simultaneously collects temperature data of the monitoring area; S13: The multispectral core collects ambient light signals as ambient light compensation signals; S14: Transmit the signals of the infrared and multispectral sensors to the control box respectively through transmission cables that are compatible with the signal transmission of infrared and multispectral sensors. S15: Real-time acquisition of infrared core operating temperature, turntable vibration intensity, connector contact impedance and cable shielding loss parameters via hardware link characteristic sensing module. S16: Collect temperature, humidity and air pressure parameters as environmental parameters through the environmental sensing unit; S17: All collected signals and parameters are marked with a unified timestamp and then stored synchronously.

3. The gas monitoring data processing method based on infrared photoelectric technology according to claim 1, characterized in that, The implementation of S2 includes the following specific steps: S21: Based on the RS422 control serial port, send synchronous trigger commands to the infrared and multispectral cores to correct the acquisition time difference; S22: Based on cable specifications and link loss parameters, time axis alignment of each link signal is performed using a signal delay compensation algorithm; S23: Perform fusion processing based on complementary signal features on the aligned infrared spectral signal, multispectral compensation signal and infrared temperature measurement data to generate the multidimensional fused signal.

4. The gas monitoring data processing method based on infrared photoelectric technology according to claim 1, characterized in that, The implementation of S3 includes the following specific steps: S31: Establish a real-time reference benchmark for the spectral response of the infrared core by utilizing the ambient light signal after multispectral fusion; S32: Based on the operating temperature data of the infrared sensor, correct the spectral response shift of the infrared sensor. S33: Dynamically adjust the acquisition frame rate of the infrared and multispectral sensors based on the vibration intensity parameters of the turntable motion; S34: Based on connector contact impedance and cable shielding loss data, dynamically adjust the signal transmission gain of each link to compensate for signal attenuation.

5. The gas monitoring data processing method based on infrared photoelectric technology according to claim 1, characterized in that, The implementation of S4 includes the following specific steps: S41: Perform secondary calibration on the calibrated initial spectral signal using the ambient light compensation signal; S42: Construct a dynamic interference feature library based on hardware link characteristic parameters. The dynamic interference feature library stores interference features corresponding to core temperature deviation, turntable vibration, cable loss and connector impedance change, and associates and labels them with environmental parameter interference features. S43: A wavelet transform algorithm based on threshold adjustment of dynamic interference feature library is adopted. The feature vectors in the dynamic interference feature library are used as the basis for dynamic threshold adjustment. Multi-scale hierarchical denoising is performed on the spectral signal to separate coupled interference components. S44: Combine the standard gas characteristic spectral library to reconstruct the denoised signal and obtain the pure spectral signal.

6. The gas monitoring data processing method based on infrared photoelectric technology according to claim 1, characterized in that, The method also includes a hardware condition adaptive learning S50, wherein S50 includes: S501: Extract historically acquired hardware characteristic parameters, calibration data, and concentration detection results to establish a mapping model between hardware status and detection accuracy; S502: Analyze the signal distortion patterns under different hardware loss levels and generate an adaptation strategy library; S503: Monitor the current hardware characteristic parameters in real time, match them with the mapping model, and call the corresponding optimization parameters from the adaptation strategy library; S504: Dynamically adjust the delay compensation coefficient of multi-link synchronous fusion, the threshold adjustment range of wavelet denoising, and the compensation weight of the dynamic fusion model using the optimization parameters.

7. The gas monitoring data processing method based on infrared photoelectric technology according to claim 6, characterized in that, The construction of the dynamic fusion model in S5 includes: S51: Based on the Beer-Lambert law, a mapping relationship is established between pure spectral signals and environmental parameters based on correlation analysis, and hardware link attenuation compensation factor and hardware operating condition adaptation parameters are introduced to construct a multivariate nonlinear regression model. S52: Extract the mapping relationship between hardware link characteristic parameters, operating condition adaptation parameters and spectral signal distortion through feature engineering, and embed this mapping relationship into the middle layer of the model; S53: The recursive least squares method with weighted operating condition adaptation parameters is adopted. Newly acquired hardware characteristic parameters, environmental parameters, operating condition adaptation parameters and pure spectral signals are synchronously input into the model to dynamically update the model's attenuation compensation factor, operating condition adaptation weight and regression coefficient.

8. The gas monitoring data processing method based on infrared photoelectric technology according to claim 7, characterized in that, The concentration inversion in step S5 includes: S54: Input the updated model's processed spectral feature parameters into a hardware-linked non-negative matrix factorization algorithm to correct the absorption coefficient; S55: Combining the gas absorption coefficient after hardware link characteristics and operating condition adaptation correction, the spectral signal of the mixed gas is separated and the characteristic spectrum of each gas is extracted. S56: The real-time concentration value of each gas is obtained by inversion based on the extracted characteristic spectra.

9. A gas monitoring data processing method based on infrared photoelectric technology according to claim 1, characterized in that, The method also includes edge-cloud collaborative optimization S60, wherein S60 includes: S601: The edge device packages and stores locally collected hardware link characteristic data, multi-dimensional fusion signals, concentration inversion results, model parameters, hardware calibration records and operating condition adaptation strategies, and dynamically adjusts the upload frequency based on the loss quantification evaluation results based on hardware link characteristic parameters. S602: The cloud receives uploaded data from multiple edge devices and builds a three-dimensional adaptation database of hardware, algorithm, and working conditions; S603: Employs federated learning algorithms to globally optimize the dynamic fusion model, generating an optimized model that adapts to different hardware link characteristics and operating parameters; S604: The cloud sends the optimized model parameters, the updated operating condition adaptation strategy library, and hardware maintenance suggestions to each edge device. After receiving the information, the edge device updates its local model and adaptation strategy.

10. A gas monitoring system based on infrared photoelectric technology, used to implement the gas monitoring data processing method based on infrared photoelectric technology as described in any one of claims 1 to 9, characterized in that, include: A multi-dimensional acquisition module is used to simultaneously acquire infrared spectral signals of the gas, ambient light compensation signals, and temperature data of the monitoring area; The hardware link characteristic sensing module is used to collect parameters such as the operating temperature of the infrared core, the vibration intensity of the turntable, the contact impedance of the connector, and the loss of the cable shielding layer in real time. The multi-link signal synchronization and fusion module is used to synchronize and fuse infrared spectral signals, ambient light compensation signals, and temperature data of the monitoring area; The hardware link dynamic adaptation and calibration module is used to calibrate multi-dimensional fused signals based on hardware link characteristic parameters; The coupling interference processing module is used to separate and reconstruct coupling interference from the calibrated signal by combining hardware link characteristic parameters and environmental parameters. The hardware condition self-adaptive learning module is used to establish a mapping model between hardware status and detection accuracy and generate adaptation strategies. The dynamic fusion model module is used to build and update the concentration inversion model; The edge-cloud collaborative optimization module is used to optimize model parameters and adaptation strategies in the cloud and update them locally.

Images