A gas sensor calibration method and system

Abstract

The application belongs to the technical field of gas detection, and discloses a gas sensor calibration method and system, six kinds of characteristic data of sensor individual fingerprint, environment coupling response, cross interference sensitivity, full life cycle aging, metrological traceability and working condition adaptation are collected through the reference traceability stage, a single sensor patent multidimensional reference characteristic library is constructed, and the value traceability binding with the national first-class standard gas is completed, and is synchronously stored to the local, cloud and legal metrological traceability node; in the zero point decoupling stage, the self-adaptive weighted environment coupling decoupling algorithm is used to remove the environment coupling interference to obtain the real zero point drift; in the range decoupling stage, the double decoupling linkage algorithm is used to complete the full working condition interference removal and the nonlinear correction of the range response curve, the response enhancement factor is configured for the ppb level low concentration interval, the saturation correction function is set for the industrial high concentration scene, and the scene-based reference adaptation coefficient is fine-tuned, so that the in-situ self-adaptive calibration is realized throughout the scene.

Description

Technical Field

[0001] This invention belongs to the field of gas detection technology, specifically a gas sensor calibration method and system. Background Technology

[0002] Currently, mainstream electrochemical, semiconductor, infrared absorption, and photoionization gas sensors all generally suffer from core problems such as zero-point drift, range nonlinearity, temperature / humidity / atmospheric pressure coupling interference, cross-sensitivity between the target gas and coexisting interfering gases, and sensitivity decay due to aging over long-term use. Regular calibration is essential to ensure the accuracy and metrological legitimacy of the detection data. Existing gas sensor calibration methods are mainly divided into two categories: laboratory static calibration and simple on-site calibration, both of which have some technical issues. Laboratory static calibration: The sensor is placed in a standard gas environment, and response values ​​are collected by introducing standard gases of different concentrations to fit the zero point and range calibration coefficients. Although this method has high measurement accuracy, it is heavily dependent on high-precision standard gas cylinders and professional gas mixing equipment, and cannot achieve on-site calibration. The calibration process is a static single-point / multi-point linear fitting, which can only adapt to a single environmental condition at the time of calibration. When the temperature, humidity, pressure, and gas composition change, the calibration accuracy decays rapidly. It cannot dynamically assess the aging status of the sensor, the calibration validity period is extremely short, and frequent recalibration is required, resulting in high maintenance costs.

[0003] Simple on-site calibration: It often uses zero-point air zeroing combined with single-point range calibration, which can only complete basic linear correction. It cannot decouple the systematic errors caused by environmental coupling interference and cross interference. In low-concentration detection scenarios, the calibration accuracy is difficult to meet the requirements of high-precision detection.

[0004] Both types of calibration methods are open-loop calibrations. After the coefficients are written, the validity of the calibrated data cannot be self-verified in real time. When the sensor experiences sudden poisoning, a sharp drop in sensitivity, or hardware failure, the calibration failure cannot be identified in time, which can easily lead to distorted monitoring data or even safety accidents. At the same time, on-site calibration cannot achieve traceability of metrological values, and the calibration results do not have legal metrological validity and cannot meet the compliance requirements of the regulatory field. Summary of the Invention

[0005] The purpose of this invention is to provide a gas sensor calibration method and system to solve one or more problems mentioned in the background art.

[0006] To achieve the above objectives, the present invention provides the following technical solution: a gas sensor calibration method, comprising the following specific steps: Furthermore, in the benchmark traceability stage, the gas sensor to be calibrated is placed in a dynamic gas distribution system with controllable temperature, humidity, pressure, and gas composition. A benchmark feature dynamic update interface is set up, and six types of feature data are collected to construct a multi-dimensional benchmark feature library: sensor individual fingerprint features, environmental coupling response features, cross-interference sensitivity features, full life cycle aging features, metrological traceability features, and operating condition adaptation features. The response data corresponding to each type of feature is bound one-to-one with the value of the national first-level standard gas for traceability. A dedicated benchmark library uniquely corresponding to a single sensor is generated and synchronously stored in the sensor's local non-volatile storage unit, cloud metrological platform, and legal metrological traceability node. The sensor's individual fingerprint features include baseline response values, baseline noise, response recovery time, and baseline features of the sensor's individual electrochemical impedance spectroscopy or the temperature coefficient of resistance of the sensitive material under full operating conditions of zero gas environment and temperature and humidity and pressure gradient. The environmental coupling response features include static response values, dynamic response curves, and temperature, humidity and pressure coupling interaction response features under different operating conditions of temperature, humidity and pressure under the full range concentration gradient of the target gas. The cross-interference sensitivity features include the cross-response values ​​of common coexisting interfering gases at different concentration gradients under a fixed concentration of the target gas, as well as the nonlinear correlation features between the concentration of interfering gases and the response values. The full life cycle aging features include the sensitivity decay rate, electrochemical impedance spectroscopy features, response characteristic shift data, and the correlation features between aging stress and decay rate under different aging cycles in accelerated aging tests. The metering traceability features include the deviation value between each set of response data and the national first-level standard gas quantity value, traceability link information, and traceability deviation dynamic correction factor. The operating condition adaptation features include the benchmark response threshold and scenario-based benchmark adaptation coefficient under different application scenarios.

[0007] Furthermore, the trigger acquisition phase is based on the multi-dimensional benchmark feature library and benchmark feature dynamic update interface built in the benchmark tracing phase. During the real-time operation of the sensor, the local processing unit continuously and synchronously acquires three types of core parameters: environmental operating condition parameters, sensor operating status parameters, and detection scene feature parameters. The acquired parameters also include interfering gas prediction parameters and aging trend parameters, and a predictive triggering mechanism is set. The environmental operating parameters include real-time ambient temperature, relative humidity, atmospheric pressure data and environmental parameter change rate data collected by the built-in temperature and humidity sensor and pressure sensor. The sensor operating status parameters include real-time baseline response value, baseline fluctuation amplitude, response recovery time, working electrical parameters, impedance characteristic value and impedance change rate, and response peak offset. The detection scene characteristic parameters include extreme values ​​of target gas detection concentration, fluctuation frequency, coexisting interfering gas components and concentration data, and interfering gas concentration change trend data within a continuous period. The triggering conditions adopt a mechanism that combines passive triggering and active prediction triggering, including baseline drift exceeding the preset threshold, continuous running time reaching the calibration warning cycle, environmental conditions exceeding the original calibration adaptation range, continuous abnormality in detection data, sensor aging characteristics exceeding the preset threshold, environmental parameter change rate exceeding the prediction threshold, and interference gas concentration trend reaching the interference warning threshold. After triggering, all currently collected multi-source status parameters and prediction data are locked simultaneously, and the prediction risk type is marked.

[0008] Furthermore, the zero-point decoupling stage is based on the multi-source state parameters locked in the trigger acquisition stage and the predicted risk type. Combined with the six-dimensional multi-dimensional benchmark feature library constructed in the benchmark tracing stage, it matches the real-time environmental operating parameters and the scenario-based benchmark adaptation coefficient. Through the adaptive weighted environmental coupling decoupling algorithm, the real-time acquired sensor baseline response value is decoupled and calculated to eliminate the interference of environmental coupling on the baseline and obtain the true zero-point drift of the sensor. At the algorithm level, the decoupling weights of temperature, humidity and pressure are dynamically adjusted based on the rate of change of environmental parameters and the predicted interference risk. In the anomaly identification stage, the corrected zero-point baseline is compared with the zero-point noise threshold and impedance change threshold in the benchmark feature library. If either threshold exceeds the range, it is determined that there is a hardware anomaly in the sensor baseline, triggering a graded anomaly warning and deciding whether to terminate the calibration process based on the warning level. At the same time, the cause of the anomaly is located by combining the individual fingerprint features in the benchmark feature library. After completing the zero-point adaptive calibration, the zero-point correction data and decoupling weights are recorded synchronously and updated to the local and cloud-based benchmark feature libraries.

[0009] Adaptive weighted environment coupling and decoupling algorithm: It represents the true zero-point drift of the sensor, which is the zero-point offset value generated by the sensor's own hardware after completely eliminating environmental coupling interference; This represents the original baseline response value of the sensor, which is the real-time baseline data acquired on-site without any interference removal processing. This represents the coupling and decoupling weights for temperature and humidity, which are adaptively adjusted based on the rate of change of on-site environmental parameters and the level of interference risk. The zero-point offset, representing the interference caused by temperature and humidity coupling, is directly retrieved from the sensor's dedicated reference feature library. This represents the pressure coupling and decoupling weight, which is dynamically adjusted based on the on-site working conditions and disturbance prediction results. The zero-point offset, representing the pressure-coupled disturbance, originates from the environmental coupling response characteristics constructed during the benchmark tracing phase. Represents the residual interference correction weight, a fixed set value, used to correct the trace environmental interference components remaining after decoupling calculation; The zero-point offset represents the residual environmental interference caused by environmental disturbances, which is a value of environmental interference that has not been completely eliminated, obtained through real-time calculation.

[0010] Furthermore, the range decoupling stage is based on the sensor's real zero-point drift and dynamic iterative data of the reference features obtained in the zero-point decoupling stage. Combined with the multi-dimensional reference feature library of the reference tracing stage, it matches the real-time environmental conditions, coexisting interfering gas parameters and prediction data obtained in the trigger acquisition stage. It calls the corresponding environmental coupling response matrix, cross-interference sensitivity matrix and nonlinear correlation features between interfering gas concentration and response value, and uses a dual decoupling linkage algorithm to achieve full-condition interference elimination. First, the cross-interference sensitivity matrix is ​​inversely operated on using a nonlinear cross-interference decoupling algorithm. Combined with the trend data of the concentration change of interfering gases, the cross-interference of coexisting interfering gases on the sensor response value is eliminated, and the pure response value corresponding to the target gas is obtained. Then, based on the decoupling results of the zero-point decoupling stage, combined with the environmental coupling response matrix and the temperature, humidity and pressure coupling interaction response characteristics, the pure response value of the target gas is decoupled in the second environment to eliminate the secondary interference of environmental parameters. Subsequently, based on the response characteristics of the full-range concentration gradient in the benchmark feature library, a piecewise nonlinear weighted fitting algorithm is used to nonlinearly correct the range response curve of the sensor to generate a range calibration coefficient that is adapted to the current working conditions. For the ppb-level low-concentration detection range, a low-concentration response enhancement factor is set based on the weighted calculation of the low-concentration benchmark response value in the benchmark feature library. For industrial high-concentration scenarios, a high-concentration saturation response correction function is set. Finally, the scenario-based benchmark adaptation coefficient in the benchmark feature library is called to fine-tune the range calibration coefficient according to the current application scenario.

[0011] Nonlinear cross-interference decoupling algorithm: The pure response value of the target gas is the actual response value produced by the target gas after completely eliminating the influence of coexisting interfering gases. Represents the original range response value of the sensor, which is the real-time data collected within the range before removing cross-interference and environmental interference; The cross-interference sensitivity matrix represents multiple sets of interfering gas response correlation matrices calibrated through standard tests during the baseline tracing phase. The inverse matrix representing the cross-interference sensitivity matrix is ​​used to reverse-calculate the response components generated by various coexisting interfering gases; The vector representing the concentration of coexisting interfering gases is a set of data on all coexisting interfering gas components and their corresponding concentrations collected in real time on-site.

[0012] Furthermore, the aging correction stage is based on the range calibration coefficient adapted to the current working condition generated in the range decoupling stage, combined with the full life cycle aging feature sequence constructed in the benchmark tracing stage and the correlation features between aging stress and decay rate, matched with the real-time operating status parameters of the sensor obtained in the trigger acquisition stage, and adopts a multi-factor aging quantification evaluation model, which integrates four major evaluation factors: sensor running time, environmental stress, response characteristic offset, and impedance change rate, and calculates the current aging quantification value, real-time sensitivity decay rate, and aging trend prediction value of the sensor through weighted calculation. Multifactor aging quantitative assessment model: This represents a quantitative value for sensor aging, ranging from zero to one. A larger value indicates a more severe degree of sensor aging. The runtime weight is a fixed weight preset by the model and is used to measure the impact of the sensor's natural runtime on aging. The normalized value representing the runtime is the value obtained by linearly mapping the actual runtime of the sensor to a standard range according to the physical range. The environmental stress weights are fixed weights preset in the model and are used to measure the accelerating effect of on-site environmental stress on sensor aging. The normalized value of the cumulative integral of environmental stress is the value after mapping the cumulative integral data of environmental stress to a standard range. The weight representing the shift in response characteristics is a fixed weight preset by the model and is used to measure the degree to which the shift in sensor response characteristics characterizes aging. The normalized value of the response characteristic offset is the value after mapping the sensor response characteristic offset data to a standard range. The weight representing the rate of change of impedance is a fixed weight preset in the model and is used to measure the degree to which the sensor impedance changes respond to aging. The normalized value represents the impedance change rate, which is the value after mapping the sensor impedance change rate data to a standard range.

[0013] Based on the aging quantification value and sensitivity decay rate, the zero point and range calibration coefficients generated in the zero point decoupling stage and the range decoupling stage are dynamically iteratively corrected. The number of iterations is adaptively adjusted according to the degree of aging to compensate for the sensitivity decay, response characteristic shift and baseline drift caused by sensor aging. At the same time, the aging quantification value, sensitivity decay rate and correction data are updated to the full life cycle aging feature sequence of the benchmark feature library. The scrapping early warning process compares the quantitative value of aging, the predicted value of aging trend, and the preset aging scrapping threshold. If any indicator exceeds the threshold, a sensor replacement early warning is triggered, and an analysis of the cause of aging is pushed out simultaneously.

[0014] Furthermore, the traceability verification stage is based on the calibration coefficient iterative correction and aging feature dynamic update data completed in the aging correction stage. Combined with the national metrological benchmark traceability data and traceability deviation dynamic correction factor in the benchmark traceability stage, as well as the environmental operating condition parameters, prediction data, and decoupling weights of the trigger acquisition stage, zero-point decoupling stage, and range decoupling stage, the zero-point and range calibration coefficients generated in this calibration, as well as the corresponding environmental operating condition parameters, aging quantified values, calibration timestamps, prediction data, and decoupling weights, are compared and verified with the national first-level standard gas traceability data in the benchmark feature library. The relative deviation between the calibration coefficients in this calibration and the national first-level standard gas traceability benchmark is calculated through the preset value traceability deviation algorithm. The graded comparison and verification is divided into three levels. The first level verifies the relative deviation between the calibration coefficient and the national first-level standard gas traceability benchmark. If the relative deviation is within the preset allowable measurement error range, the calibration coefficient is determined to have legal measurement traceability. The second level verifies the compatibility between the decoupling weight, aging correction factor and benchmark feature library of the calibration. If the compatibility meets the standard, the reliability of the calibration result is further confirmed. The third level verifies the consistency between the predicted data and the actual calibration result, which is used to optimize the accuracy of subsequent prediction triggers. If all three levels of verification are up to standard, the calibration coefficient, calibration log, traceability verification results, and anomaly records will be synchronously and permanently stored in the sensor's local non-volatile storage unit, and simultaneously synchronized to the cloud-based metrology platform and the legal metrology traceability node. If any verification fails to meet the standard, the calibration will be deemed invalid, triggering a calibration failure warning. At the same time, the cause of the failure will be analyzed and a targeted retry plan will be pushed out.

[0015] Furthermore, in the closed-loop verification stage, based on the calibration validity determination and related data solidification results completed in the traceability verification stage, the sensor enters the normal detection operation state. The local processing unit performs full-cycle validity self-verification on the real-time detection data based on the applicable operating conditions range, calibration coefficient, and decoupling weight of this calibration. It sets up feedback optimization function and multi-source data fusion verification function. The real-time operating condition verification link monitors environmental operating condition parameters in real time. When the environmental temperature, humidity, and pressure exceed the applicable range of this calibration, it automatically triggers an operating condition over-limit warning and simultaneously starts dynamic fine-tuning of calibration parameters. The calibration validity verification process collects sensor response characteristic parameters in real time. When the baseline drift, response recovery time, or sensitivity change exceeds the allowable threshold after calibration, the calibration is automatically determined to be invalid and a recalibration process is triggered. At the same time, the timing of recalibration is optimized by combining the predicted data. The multi-source data fusion verification process integrates local sensor detection data, cloud benchmark data, and legal metrological traceability data to cross-verify the real-time detection data. The feedback optimization process continuously collects sensor feature data during the testing process, updates it to the benchmark feature library regularly, optimizes benchmark features, decoupling algorithm weights, and aging assessment model parameters. The full-cycle iteration process combines aging trend prediction, operating condition change prediction, and interference prediction data to plan the next calibration time in advance.

[0016] The present invention also provides a gas sensor calibration system based on the above method, including a reference tracing unit, a trigger acquisition unit, a decoupling calibration unit, an aging correction unit, a tracing verification unit, a closed-loop verification unit, and a data storage and transmission unit; The reference traceability unit is deployed in the statutory metrology laboratory and includes a dynamic gas mixing system with controllable temperature, humidity, pressure and gas composition. It is used to collect six types of sensor characteristic data, build a multi-dimensional reference feature library, complete the traceability binding with the national first-level standard gas, generate a dedicated reference library for a single sensor, and synchronously store it to the local, cloud and statutory metrology traceability nodes through a triple synchronous storage mechanism, and set a dynamic update interface for reference features. The trigger acquisition unit is integrated into the sensor local processing module. It has a built-in temperature and humidity acquisition module and a pressure acquisition module. It continuously and synchronously acquires three types of core parameters: environmental conditions, sensor operating status, and detection scenario. It adopts a dual triggering mechanism of passive and active prediction to lock the multi-source parameters and predict the risk type after triggering. The decoupling calibration unit includes a zero-point decoupling module and a range decoupling module. Through an adaptive weighted decoupling algorithm and a dual decoupling linkage algorithm, the modules eliminate environmental coupling interference and cross interference, complete zero-point drift correction and range nonlinear calibration, generate calibration coefficients adapted to the current working conditions, and perform scenario-based fine-tuning. The aging correction unit is based on a multi-factor aging quantification assessment model to quantify the aging status of the sensor, iteratively correct the calibration coefficient, and trigger a scrap warning. The traceability verification unit uses a three-level comparison and verification mechanism to compare the calibration coefficient with the national first-level standard gas traceability data to determine the calibration validity. If the data meets the standard, it is solidified to meet the regulatory traceability requirements; if it does not meet the standard, a retry plan is pushed. The closed-loop verification unit realizes full-cycle self-verification after calibration, dynamically fine-tunes parameters, iteratively optimizes the benchmark and algorithm, and plans the next calibration time. The data storage and transmission unit ensures real-time synchronization, traceable storage, and transmission of data at each stage.

[0017] The beneficial effects of this invention are as follows: 1. This invention collects six types of characteristic data from sensors during the benchmark traceability stage: individual sensor fingerprint, environmental coupling response, cross-interference sensitivity, full life cycle aging, metrological traceability, and operating condition adaptation. This data forms a multi-dimensional benchmark feature library for a single sensor and is bound to the metrological traceability of national first-level standard gases. The data is simultaneously stored locally, in the cloud, and at legal metrological traceability nodes, ensuring traceability and verifiability of metrological data. During the trigger acquisition stage, a dual triggering mechanism of passive and active prediction is employed to accurately lock the calibration timing and multi-source operating parameters. In the zero-point decoupling stage, an adaptive weighted environmental coupling decoupling algorithm is used to eliminate environmental coupling interference and obtain the true zero-point drift. In the range decoupling stage, a dual decoupling linkage algorithm is used to eliminate interference under all operating conditions and correct the nonlinearity of the range response curve. A response enhancement factor is configured for the ppb-level low-concentration range, and a saturation correction function is set for industrial high-concentration scenarios. Combined with scenario-based benchmark adaptation coefficient fine-tuning, the entire process achieves on-site adaptive calibration.

[0018] 2. This invention establishes a full lifecycle aging quantitative assessment and closed-loop calibration verification system. Through a multi-factor aging quantitative assessment model, it integrates four core factors: sensor operating time, environmental stress, response characteristic offset, and impedance change rate. This accurately quantifies the sensor's aging degree, sensitivity decay rate, and aging trend. Based on the quantification results, it adaptively and iteratively corrects the zero-point and range calibration coefficients, effectively compensating for sensitivity decay and baseline drift caused by aging. Simultaneously, it triggers aging and scrap warnings and locates the causes of aging. In the traceability verification stage, a three-level comparison verification mechanism verifies the deviation of calibration coefficients from national metrological standards, the compatibility of decoupling parameters with the benchmark library, and the consistency between predicted data and calibration results. Once the standards are met, the entire process data is solidified. In the closed-loop verification stage, the detection data achieves full-cycle self-verification, dynamically fine-tunes calibration parameters, iteratively optimizes benchmarks and algorithms, and intelligently plans the next calibration time.

[0019] 3. This invention achieves deep decoupling of multi-source interference and real-time identification and early warning of hardware anomalies. In the zero-point decoupling stage, the decoupling weights are dynamically adjusted according to the rate of change of environmental parameters and interference risk to remove the influence of temperature, humidity, and pressure coupling interference on the baseline and obtain true zero-point data. In the range decoupling stage, a nonlinear cross-interference decoupling algorithm is first used to eliminate cross-interference of coexisting gases, and then a secondary environmental decoupling is completed by combining environmental coupling characteristics to eliminate the influence of various interferences on the detection results. At the same time, in the decoupling calibration process, a dual threshold comparison mechanism of zero-point noise threshold and impedance change threshold is adopted to identify sensor hardware anomalies in real time, trigger graded early warnings and locate the cause of the fault, and avoid invalid calibration. The entire process continuously optimizes the benchmark features and algorithm parameters through data synchronization storage and dynamic update interfaces. Attached Figure Description

[0020] Figure 1 This is a flowchart illustrating the entire calibration process of the gas sensor of this invention. Figure 2 This is a flowchart of the zero-point and range decoupling calibration sub-process of the present invention; Figure 3 This is a flowchart of the closed-loop verification process for aging assessment and traceability in this invention. Detailed Implementation

[0021] 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.

[0022] like Figures 1 to 3 As shown, this embodiment of the invention provides a gas sensor calibration method, including the following specific steps: In this embodiment of the invention, the reference traceability stage is carried out in a legal metrology laboratory environment. The gas sensor to be calibrated is placed in a dynamic gas mixing system with controllable temperature, humidity, pressure, and gas composition. A dynamic update interface for reference features is set up. Six types of feature data are collected, namely, sensor individual fingerprint features, environmental coupling response features, cross-interference sensitivity features, full life cycle aging features, metrology traceability features, and operating condition adaptation features, to build a multi-dimensional reference feature library. The response data corresponding to each type of feature is bound one-to-one with the value of the national first-level standard gas for traceability. A dedicated reference library uniquely corresponding to a single sensor is generated and synchronously stored in the sensor's local non-volatile storage unit, cloud metrology platform, and legal metrology traceability node. The dynamic gas distribution system integrates a high-precision temperature and humidity control module, a pressure stabilization module, and a standard gas distribution module. It can independently control the ambient temperature, relative humidity, and atmospheric pressure to the target operating point. At the same time, it can accurately configure standard gases and mixed gases of different concentrations and components. The system has the ability to maintain stable operating conditions for a long time and can control the fluctuation of environmental parameters within the extremely small error range allowed by measurement.

[0023] The benchmark feature dynamic update interface is a standardized data interaction interface that supports bidirectional data interaction and incremental updates between the sensor local, cloud-based metrology platform, and legal metrology traceability node. It can receive new feature data, correction parameters, and verification results generated during the calibration process in real time, automatically complete the iterative update of the benchmark feature library, and retain historical benchmark data versions.

[0024] The individual fingerprint features of the sensor include the baseline response value, baseline noise, response recovery time, and baseline features of the individual electrochemical impedance spectroscopy of the sensor or the temperature coefficient of resistance of the sensitive material under all operating conditions in a zero-air environment, thereby distinguishing the individual differences of different sensors. The individual fingerprint feature serves as a unique identifier for each sensor, used to accurately match dedicated benchmark data during calibration, avoiding feature confusion and parameter misuse between different sensors. It also provides a dedicated comparison basis for hardware anomaly identification and fault location, ensuring that the calibration, decoupling, and aging assessment of each sensor are performed based on its own inherent characteristics.

[0025] The environmental coupling response characteristics include static response values, dynamic response curves, and temperature, humidity and pressure coupling interaction response characteristics under different temperature, humidity and pressure conditions under the full range concentration gradient of the target gas, thereby adapting to the environmental interference response requirements under complex working conditions. The environmental coupling response matrix is ​​constructed in two dimensions using the full-range concentration gradient and the full-condition temperature, humidity and pressure gradient. It includes the real response data of the sensor and the coupling interference rules under different environmental combinations. The matrix data is calibrated in one go during the benchmark traceability stage and stored in a dedicated benchmark feature library. In the zero-point decoupling and range decoupling stages of on-site calibration, the system can automatically match the corresponding coupling response data in the matrix according to the real-time collected environmental parameters, quickly remove the interference components caused by environmental coupling, and improve the execution efficiency of on-site calibration and the accuracy of environmental interference removal.

[0026] The cross-interference sensitive feature includes the cross-response value of common coexisting interfering gases at different concentration gradients under a fixed concentration of the target gas, as well as the nonlinear correlation feature between the concentration of interfering gas and the response value, thereby realizing interference elimination in complex interference scenarios. The full life cycle aging characteristics include sensitivity decay rate, electrochemical impedance spectroscopy characteristics, response characteristic shift data, and aging stress and decay rate correlation characteristics under different aging cycles in accelerated aging tests, thereby achieving accurate prediction and compensation of aging state. The metrological traceability features include the deviation value between each set of response data and the national first-level standard gas value, traceability link information, and traceability deviation dynamic correction factor. The operating condition adaptation features include the benchmark response threshold and scenario-based benchmark adaptation coefficient under different application scenarios such as industrial high-temperature scenarios and indoor low-concentration scenarios, thereby adapting to the calibration requirements of different application scenarios.

[0027] In this embodiment of the invention, the trigger acquisition phase is based on the multi-dimensional benchmark feature library and benchmark feature dynamic update interface built in the benchmark tracing phase. During the real-time operation of the sensor, the local processing unit continuously and synchronously acquires three types of core parameters: environmental operating condition parameters, sensor operating status parameters, and detection scene feature parameters. The acquired parameters also include interfering gas prediction parameters and aging trend parameters. A predictive triggering mechanism is set to ensure accurate calibration timing, reduce invalid calibration, and lower operation and maintenance costs. The environmental operating parameters include real-time ambient temperature, relative humidity, atmospheric pressure data and environmental parameter change rate data collected by built-in temperature and humidity sensors and pressure sensors, such as the sudden change rate of temperature and humidity, so as to predict the risk of environmental coupling interference; the sensor operating status parameters include the sensor's real-time baseline response value, baseline fluctuation amplitude, response recovery time, working electrical parameters, impedance characteristic value and impedance change rate, response peak offset, so as to predict the risk of sensor aging and poisoning. The detection scenario characteristic parameters include the extreme value of the target gas detection concentration, fluctuation frequency, coexisting interfering gas components and concentration data, and interfering gas concentration change trend data within a continuous period, thereby predicting the intensity of cross-interference. The triggering conditions adopt a mechanism that combines passive triggering and active prediction triggering, including baseline drift exceeding the preset threshold, continuous running time reaching the calibration warning cycle, environmental conditions exceeding the original calibration adaptation range, continuous abnormality in detection data, sensor aging characteristics exceeding the preset threshold, environmental parameter change rate exceeding the prediction threshold, and interference gas concentration trend reaching the interference warning threshold. After triggering, all currently collected multi-source status parameters and prediction data are locked simultaneously, and the prediction risk type is marked.

[0028] The proactive prediction trigger identifies calibration needs in advance based on historical sensor operating data, operating condition change trends, and aging development patterns, and initiates calibration preparation before parameters exceed limits. The passive trigger directly initiates the calibration process based on real-time collected parameters exceeding preset thresholds. The two triggering methods complement each other. After the trigger command takes effect, the system automatically locks multi-source operating data and prediction information for the complete time period before and after the trigger to avoid invalid triggering, repeated triggering, and data loss issues.

[0029] In this embodiment of the invention, the zero-point decoupling stage is based on the multi-source state parameters locked in the trigger acquisition stage and the predicted risk type. Combined with the six-dimensional multi-dimensional benchmark feature library constructed in the benchmark tracing stage, it matches the real-time environmental operating parameters and the scenario-based benchmark adaptation coefficient. The adaptive weighted environmental coupling decoupling algorithm is used to decouple the real-time acquired sensor baseline response value, eliminate the interference of environmental temperature and humidity, atmospheric pressure and their coupling effect on the baseline, and obtain the true zero-point drift of the sensor. The adaptive weighted environment coupling and decoupling algorithm is a three-layer serial decoupling calculation structure designed specifically for zero-point calibration of gas sensors. It consists of a data input layer, a dynamic weight allocation layer, and a decoupling operation output layer. The algorithm input is deeply bound to the sensor's field-acquired data and the reference feature library data. The data input layer receives five types of core data: real-time baseline response values ​​of the sensor, ambient temperature, relative humidity, atmospheric pressure, rate of change of environmental parameters, and predicted level of interference risk. The dynamic weight allocation layer uses environmental coupling response features in the benchmark feature library as the calculation basis and adaptively allocates decoupling weights according to the rate of change of environmental parameters and the level of interference risk. The decoupling weight range for temperature and humidity coupling is 0.4~0.6, the decoupling weight range for pressure coupling is 0.3~0.5, and the residual interference correction weight is fixed at 0.1. The weight parameters are dynamically adjusted according to the operating conditions. The decoupling operation output layer uses a weighted decoupling operation formula to remove the influence of temperature, humidity, and pressure coupling interference on the baseline and outputs the true zero-point drift without environmental interference. The algorithm input and output are directly related to the sensor zero-point calibration process.

[0030] The adjustment of dynamic weights is based on the rate of change of real-time environmental parameters of the sensor and the level of interference risk. The more drastic the fluctuation of environmental parameters and the higher the level of interference risk, the higher the weight ratio of temperature, humidity and pressure coupling decoupling. The weight of residual interference correction is adjusted synchronously. The weight allocation fits the actual working conditions on site and can isolate the influence of environmental coupling interference on the baseline to the greatest extent, ensuring that the decoupling calculation results truly reflect the zero-point state of the sensor itself.

[0031] At the algorithm level, the decoupling weights of temperature, humidity, and pressure are dynamically adjusted based on the rate of change of environmental parameters and the predicted interference risk. In the anomaly identification stage, the corrected zero-point baseline is compared with the zero-point noise threshold and impedance change threshold in the benchmark feature library. If either threshold exceeds the range, it is determined that there is a hardware anomaly in the sensor baseline, such as poisoning of sensitive components or circuit failure, triggering a graded anomaly warning. The anomaly warning is divided into mild, moderate, and severe, and the calibration process is terminated based on the warning level to avoid invalid calibration. At the same time, the cause of the anomaly is located by combining the individual fingerprint features in the benchmark feature library. The hardware anomalies are classified into three levels—mild, moderate, and severe—based on their impact on detection and calibration. Different levels correspond to different handling methods, such as warning prompts, suspension of calibration, termination of calibration, and forced maintenance. Anomaly localization involves comparing the real-time collected baseline, impedance, and response characteristic data with the sensor's unique fingerprint characteristics to quickly pinpoint the specific source of the anomaly, such as aging of sensitive components, poisoning of components, circuit failure, or poor contact.

[0032] After completing the zero-point adaptive calibration, the zero-point correction data and decoupling weights are recorded synchronously and updated to the local and cloud-based benchmark feature libraries.

[0033] In this embodiment of the invention, the range decoupling stage is based on the sensor’s real zero-point drift and dynamic iterative data of the reference features obtained in the zero-point decoupling stage. Combined with the multi-dimensional reference feature library of the reference tracing stage, it matches the real-time environmental conditions, coexisting interfering gas parameters and prediction data obtained in the trigger acquisition stage. It calls the corresponding environmental coupling response matrix, cross-interference sensitivity matrix and nonlinear correlation features between interfering gas concentration and response value, and uses a dual decoupling linkage algorithm to achieve full-condition interference elimination. The dual decoupling linkage algorithm is a series dual-module dedicated decoupling algorithm designed to solve the problem of superposition of cross interference and environmental coupling interference in the range calibration of gas sensors. The overall structure is that the nonlinear cross interference decoupling module and the environmental secondary decoupling module are connected in series. The output of the former module is directly used as the input of the latter module, and the data between the modules is directly transmitted. The algorithm input data includes: real-time concentration of coexisting interfering gases, original response value of the target gas, cross-interference sensitivity matrix, environmental coupling response matrix, and temperature, humidity and pressure coupling interaction characteristics. All data comes from the trigger acquisition stage and the baseline tracing stage. The nonlinear cross-interference decoupling module performs inverse matrix operation on the cross-interference sensitivity matrix and combines it with the trend of interfering gas concentration change to complete the decoupling calculation, remove the cross-interference of coexisting gases and obtain the pure response value of the target gas. The cross-interference sensitivity matrix is ​​constructed from multiple sets of cross-response data of interfering gases collected during the benchmark tracing phase. The matrix contains the correspondence between different concentrations of interfering gases and sensor response values. During the calibration process, the system automatically calls the corresponding data in the matrix to perform decoupling operations based on the real-time detected components and concentrations of interfering gases, and removes the interference response generated by non-target gases. The secondary environmental decoupling module performs secondary decoupling on the pure response value based on the zero-point decoupling result and combined with the environmental coupling response characteristics to eliminate residual environmental parameter interference; the algorithm is based on the sensor's dedicated reference feature library and the input and output data are strongly coupled with the range calibration process.

[0034] The two decoupling modules adopt a sequential collaborative execution logic. First, they complete the comprehensive elimination of cross-interference, and then perform secondary environmental decoupling based on the pure response value. They work together to achieve deep stripping of multi-source interference, so that the range calibration calculation is based entirely on the real response data of the target gas.

[0035] First, the cross-interference sensitivity matrix is ​​inversely operated by a nonlinear cross-interference decoupling algorithm. Then, combined with the data on the trend of interfering gas concentration, the cross-interference of coexisting interfering gases on the sensor response value is eliminated to obtain the pure response value of the target gas. Based on the decoupling results of the zero-point decoupling stage, combined with the environmental coupling response matrix and the temperature, humidity and pressure coupling interaction response characteristics, the pure response value of the target gas is subjected to secondary environmental decoupling to eliminate secondary interference from environmental parameters, thereby ensuring the accuracy of the response value. Subsequently, based on the response characteristics of the full-range concentration gradient in the benchmark feature library, a piecewise nonlinear weighted fitting algorithm is used to nonlinearly correct the range response curve of the sensor and generate a range calibration coefficient that is adapted to the current working conditions. The segmented nonlinear weighted fitting algorithm is a full-range segmented fitting algorithm customized for the nonlinear characteristics of the range response of gas sensors. It divides the gas detection concentration range into three independent fitting intervals: ppb-level low concentration interval, conventional medium concentration interval, and industrial high concentration interval. Each interval is calculated independently and then spliced ​​together to form a complete calibration curve, which can adapt to the accuracy requirements of different concentration detection scenarios. Each segment employs a second-order nonlinear polynomial fitting method, with pre-set fitting weights based on detection priority: 0.4 for the ppb-level low-concentration segment, 0.3 for the conventional medium-concentration segment, and 0.3 for the industrial high-concentration segment. The algorithm inputs full-range concentration gradient response data from the benchmark feature library, performs segmented fitting calculations using the least squares method, and outputs segmented nonlinear range calibration coefficients. The algorithm combines scenario-based benchmark adaptation coefficients to simultaneously enhance low-concentration response and correct high-concentration saturation. The input and output data are directly related to sensor range calibration and scenario-based fine-tuning.

[0036] For the ppb-level low-concentration detection range, a low-concentration response enhancement factor is set based on the weighted calculation of low-concentration benchmark response values ​​in the benchmark feature library to improve the calibration resolution and accuracy in the low-concentration range. For industrial high-concentration scenarios, a high-concentration saturation response correction function is set to avoid calibration deviations caused by sensor saturation response. Finally, the scenario-based benchmark adaptation coefficients in the benchmark feature library are called to fine-tune the range calibration coefficients according to the current application scenario.

[0037] The high-concentration saturation correction function is designed based on the saturation response characteristics of the sensor under high-concentration conditions. By normalizing, compressing and linearly correcting the high-concentration response signal that exceeds the adaptive detection range, it eliminates the problems of signal top-level, response distortion and curve distortion caused by excessively high concentration. The corrected response data can smoothly connect to the calibration curve in the medium and low concentration range.

[0038] The scenario-based benchmark adaptation coefficients are pre-calibrated and generated for different application scenarios during the benchmark tracing stage. The system automatically identifies the current application scenario type through real-time collected detection scenario feature parameters and accurately matches the adaptation coefficients of the corresponding scenario. It can quickly adapt to different scenarios such as industrial high temperature and humidity, indoor low concentration, and complex outdoor working conditions.

[0039] In this embodiment of the invention, the aging correction stage is based on the range calibration coefficient adapted to the current working condition generated in the range decoupling stage, combined with the full life cycle aging feature sequence constructed in the benchmark tracing stage and the correlation features between aging stress and decay rate, and matched with the real-time operating status parameters of the sensor obtained in the trigger acquisition stage, including impedance change rate and response peak offset. A multi-factor aging quantification evaluation model is adopted, which integrates four major evaluation factors: sensor running time, environmental stress, response characteristic offset, and impedance change rate. The current aging quantification value, real-time sensitivity decay rate, and aging trend prediction value of the sensor are obtained through weighted calculation. The multi-factor aging quantitative assessment model is a four-layer forward structured AI model adapted to the full life cycle aging assessment of gas sensors. The layers are, in order, factor input layer, normalization preprocessing layer, adaptive weighted calculation layer, and result output layer. Each layer is unidirectionally serially connected and has no feedback loop, which perfectly matches the aging mechanism of gas sensors and the on-site calibration scenario. The input data of the factor input layer consists of four core quantification factors: sensor runtime, cumulative integral value of environmental stress, response characteristic offset, and impedance change rate. All of these factors are obtained from the sensor operating status parameters collected in real time during the trigger acquisition phase. The normalization preprocessing layer linearly maps each factor data to the standard interval [0, 1] according to the physical range, eliminating the differences in the dimensions and numerical scales of different factors. The adaptive weighted calculation layer presets fixed weight coefficients based on the aging sensitivity characteristics of the gas sensor: runtime weight 0.2, environmental stress weight 0.3, response characteristic offset weight 0.3, and impedance change rate weight 0.2. The aging quantification calculation is completed through a weighted summation formula. The result output layer synchronously outputs three types of results: aging quantification value, real-time sensitivity decay rate, and aging trend prediction value. The output data is directly connected to the calibration coefficient iterative correction stage and is linked in real time with the sensor's full life cycle aging feature library.

[0040] The model has four major evaluation factors that are directly related to the aging state of the sensor. The running time reflects the degree of natural aging, the environmental stress integral value reflects the degree of accelerated aging under working conditions, the response characteristic offset reflects the degree of performance degradation, and the impedance change rate reflects the degree of degradation of the sensitive element. The factors corroborate each other and are comprehensively weighted to comprehensively and objectively quantify the true aging state of the sensor and avoid the errors caused by single factor evaluation.

[0041] Based on the aging quantification value and sensitivity decay rate, the zero-point and range calibration coefficients generated in the zero-point decoupling stage and the range decoupling stage are dynamically iteratively corrected. The number of iterations is adaptively adjusted according to the degree of aging to compensate for the sensitivity decay, response characteristic shift and baseline drift caused by sensor aging. At the same time, the aging quantification value, sensitivity decay rate and correction data are updated to the full life cycle aging feature sequence of the benchmark feature library to achieve dynamic optimization of aging features. The full life cycle aging feature library adopts an incremental update rule. After each calibration, the aging quantification value, sensitivity decay rate, and correction data calculated in this time are appended to the feature library in time series, while historical aging data are retained to form a complete aging trajectory.

[0042] The scrapping early warning process compares the quantitative value of aging, the predicted value of aging trend, and the preset aging scrapping threshold. If any indicator exceeds the threshold, a sensor replacement warning is triggered, and an analysis of the cause of aging is pushed out simultaneously, such as environmental stress or natural aging.

[0043] The aging and scrapping threshold is comprehensively calibrated during the benchmark traceability stage by combining multiple factors such as the characteristics of the sensor's sensitive elements, the target detection accuracy requirements, the safety level of the application scenario, and industry metrological standards. The threshold value is bound to the aging characteristics of a single sensor throughout its entire life cycle. When any indicator of the aging quantification value or the aging trend prediction value exceeds the preset threshold, it indicates that the core performance of the sensor can no longer meet the requirements of normal detection and legal metrology. Continued use will cause data distortion and safety hazards, triggering a replacement warning in a timely manner.

[0044] In this embodiment of the invention, the traceability verification stage is based on the calibration coefficient iterative correction and aging feature dynamic update data completed in the aging correction stage. It combines the national metrological benchmark traceability data and traceability deviation dynamic correction factor in the benchmark traceability stage, as well as the environmental operating condition parameters, prediction data, and decoupling weights of the trigger acquisition stage, zero-point decoupling stage, and range decoupling stage. The zero-point and range calibration coefficients generated in this calibration, as well as the corresponding environmental operating condition parameters, aging quantified values, calibration timestamps, prediction data, and decoupling weights, are compared and verified with the national first-level standard gas traceability data in the benchmark feature library. The relative deviation between the calibration coefficients in this calibration and the national first-level standard gas traceability benchmark is calculated through the preset value traceability deviation algorithm. The graded comparison and verification is divided into three levels. The first level verifies the relative deviation between the calibration coefficient and the national first-level standard gas traceability benchmark. If the relative deviation is within the preset allowable measurement error range, the calibration coefficient is determined to have legal measurement traceability. The second level verifies the compatibility between the decoupling weight, aging correction factor and benchmark feature library of the calibration. If the compatibility meets the standard, the reliability of the calibration result is further confirmed. The third level verifies the consistency between the predicted data and the actual calibration result, which is used to optimize the accuracy of subsequent prediction triggers. The three-level comparison and verification process uses the national first-level standard gas traceability benchmark as the sole legal basis for judgment. The first-level verification ensures that the calibration coefficients have legal metrological traceability; the second-level verification ensures that the decoupling parameters, aging correction parameters and sensor benchmark characteristics are matched and adapted; the third-level verification continuously optimizes the triggering algorithm and prediction model by comparing the consistency between the predicted data and the actual calibration results, thereby improving the accuracy of subsequent calibration triggering; and a complete and traceable verification record is formed throughout the verification process.

[0045] The permissible error range for Level 3 calibration is pre-set in the benchmark traceability stage according to national metrological technical specifications, sensor detection accuracy level, and industry application standards. The error limit is fully matched with the sensor's own performance indicators and legal metrological requirements, and the calibration results are objective and impartial.

[0046] If all three levels of verification are up to standard, the calibration coefficient, calibration log, traceability verification results, and anomaly records will be synchronously and permanently stored in the sensor's local non-volatile storage unit, and simultaneously synchronized to the cloud-based metrology platform and the legal metrology traceability node. If any verification fails to meet the standard, the calibration will be deemed invalid, triggering a calibration failure warning. At the same time, the cause of failure will be analyzed, such as predicted deviation or mismatch of reference characteristics, and a targeted retry plan will be pushed out.

[0047] The calibration failure retry plan is automatically generated based on the reason for the failure of the three-level verification. If the measurement deviation exceeds the standard, a new benchmark comparison plan will be pushed. If the decoupling parameters do not match, a weight recalculation plan will be pushed. If the predicted data deviation is pushed, a trigger condition optimization plan will be pushed. The plan includes specific operation steps, parameter adjustment direction, and retry environment requirements.

[0048] The calibration data is solidified using a dual locking mechanism of hardware-level write protection and software encryption. The solidified calibration coefficients, traceability records, verification results, and log information can only be viewed in read-only mode and legally required for traceability verification. They cannot be manually modified, deleted, overwritten, or tampered with on-site. The data is marked with a unique device identifier and timestamp throughout the entire process.

[0049] In this embodiment of the invention, the closed-loop verification stage is based on the calibration validity determination and related data solidification results completed in the traceability verification stage. The sensor enters the normal detection operation state. The local processing unit performs full-cycle validity self-verification on the real-time detection data based on the applicable working condition range, calibration coefficient, and decoupling weight of this calibration. It sets up feedback optimization function and multi-source data fusion verification function. The working condition real-time verification link monitors the environmental working condition parameters in real time. When the environmental temperature, humidity, and pressure exceed the applicable range of this calibration, it automatically triggers the working condition over-limit warning and simultaneously starts the dynamic fine-tuning of calibration parameters based on the scenario-based adaptation coefficient of the benchmark feature library. The calibration validity verification process collects sensor response characteristic parameters in real time. When the baseline drift, response recovery time, or sensitivity change exceeds the allowable threshold after calibration, the calibration is automatically determined to be invalid and a recalibration process is triggered. At the same time, the timing of recalibration is optimized by combining the predicted data. The multi-source data fusion verification process integrates local sensor detection data, cloud benchmark data, and legal metrological traceability data to cross-verify the real-time detection data. The multi-source data fusion verification follows the execution rule of using local real-time data as the basis, cloud-based benchmark data as the reference, and legal metrological traceability data as the final criterion. The system compares the deviation range of the three data in real time. If the deviation is within the allowable range, the test data is deemed valid. If it exceeds the allowable range, the system immediately triggers the dynamic fine-tuning of calibration parameters or the automatic recalibration process.

[0050] The feedback optimization process continuously collects sensor characteristic data during the testing process, such as response values, aging data, and interference data, and updates them to the benchmark feature library regularly. It optimizes benchmark features, decoupling algorithm weights, and aging assessment model parameters. The full-cycle iteration process combines aging trend prediction, operating condition change prediction, and interference prediction data to plan the next calibration time in advance, achieving adaptive calibration throughout the sensor's life cycle, further reducing operation and maintenance costs, and improving testing reliability.

[0051] The full-cycle self-calibration adopts an operation mode that combines real-time online monitoring with periodic automatic verification. Multi-source data fusion calibration ensures the accuracy of real-time detection data by cross-comparing local sensor detection data, cloud-based benchmark data, and legal metrological traceability data. Calibration parameters are adaptively and dynamically fine-tuned according to changes in on-site operating conditions, interference intensity, and sensor aging status to always maintain optimal detection accuracy. The next calibration time is determined by comprehensively calculating the sensor aging trend, on-site operating condition stability, interference gas intensity, and historical calibration results.

[0052] On-site adaptive calibration is achieved through the collaborative efforts of the sensor's local processing unit, a dedicated reference feature library, and an adaptive decoupling algorithm. It eliminates the need to disassemble the sensor, connect to professional laboratory equipment, or introduce standard gases, and can automatically complete the entire calibration process at the actual installation and operation site of the sensor.

[0053] The calibration parameters are dynamically fine-tuned using a small-step, incremental adjustment method. Based on slight changes in real-time operating conditions, interference intensity, and aging status, the zero-point and range calibration coefficients are slightly adapted and corrected to avoid frequent and large adjustments that could cause fluctuations in the test data.

[0054] This invention also provides a gas sensor calibration system based on the above method, including a reference tracing unit, a trigger acquisition unit, a decoupling calibration unit, an aging correction unit, a tracing verification unit, a closed-loop verification unit, and a data storage and transmission unit. The reference traceability unit is deployed in the statutory metrology laboratory and includes a dynamic gas mixing system with controllable temperature, humidity, pressure and gas composition. It is used to collect six types of sensor characteristic data, build a multi-dimensional reference feature library, complete the traceability binding with the national first-level standard gas, generate a dedicated reference library for a single sensor, and synchronously store it to the local, cloud and statutory metrology traceability nodes through a triple synchronous storage mechanism, and set a dynamic update interface for reference features. The triple synchronous storage adopts real-time triggered synchronization logic. Whenever the benchmark library is generated, feature data is updated, and calibration results are solidified, the system immediately performs a three-party synchronous write operation. The stored data includes complete information such as the sensor's unique device identifier, national metrological traceability number, full-dimensional feature data, calibration parameters, timestamps, and operation logs. The data of the local, cloud, and legal metrological traceability nodes are consistent and cannot be tampered with.

[0055] The trigger acquisition unit is integrated into the sensor local processing module. It has built-in temperature and humidity sensors and pressure sensors. It continuously and synchronously collects three types of core parameters: environmental conditions, sensor operating status, and detection scenario. It adopts a dual triggering mechanism of passive and active prediction to lock the multi-source parameters and predict the risk type after triggering. The decoupling calibration unit includes a zero-point decoupling module and a range decoupling module. Through an adaptive weighted decoupling algorithm and a dual decoupling linkage algorithm, the modules eliminate environmental coupling interference and cross interference, complete zero-point drift correction and range nonlinear calibration, generate calibration coefficients adapted to the current working conditions, and perform scenario-based fine-tuning. The aging correction unit is based on a multi-factor aging quantification assessment model to quantify the aging status of the sensor, iteratively correct the calibration coefficient, and trigger a scrap warning. The traceability verification unit uses a three-level comparison and verification mechanism to compare the calibration coefficient with the national first-level standard gas traceability data to determine the calibration validity. If the data meets the standard, it is solidified to meet the regulatory traceability requirements; if it does not meet the standard, a retry plan is pushed. The closed-loop verification unit realizes full-cycle self-verification after calibration, dynamically fine-tunes parameters, iteratively optimizes the benchmark and algorithm, and plans the next calibration time. The data storage and transmission unit ensures real-time synchronization, traceable storage, and transmission of data at each stage.

[0056] 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.

[0057] 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 sensor calibration method, characterized in that, The specific steps include the following: In the benchmark traceability stage, the gas sensor to be calibrated is placed in the dynamic gas distribution system, multi-dimensional feature data is collected to build a benchmark feature library, the value traceability binding with the national first-level standard gas is completed, a dedicated benchmark library corresponding to a single sensor is generated, and a benchmark feature dynamic update interface is set and stored synchronously. During the trigger acquisition phase, based on the benchmark feature library and benchmark feature dynamic update interface, core operating parameters are continuously acquired while the sensor is running. A dual triggering mechanism combining passive triggering and active predictive triggering is adopted to lock the multi-source parameters acquired after triggering and predict the risk type. In the zero-point decoupling stage, based on the multi-source parameters and the predicted risk type, and combined with the benchmark feature library, the sensor baseline response value is decoupled and calculated using an adaptive weighted environmental coupling decoupling algorithm. Environmental interference is eliminated to obtain the true zero-point drift, zero-point calibration is completed and the benchmark feature library is updated. At the same time, hardware anomalies are identified and the cause is located. In the range decoupling stage, based on the zero drift and benchmark iteration data, combined with the benchmark feature library and trigger acquisition parameters, a dual decoupling linkage algorithm is used to eliminate interference, nonlinearly correct the range response curve, generate a range calibration coefficient that adapts to the working conditions, and perform scenario-based fine-tuning in combination with the scenario-based benchmark adaptation coefficient. During the aging correction phase, based on the range calibration coefficient, combined with the aging characteristics of the benchmark traceability phase and the trigger acquisition operation parameters, the aging status of the sensor is quantified through a multi-factor aging quantification evaluation model, the calibration coefficient is iteratively corrected and the aging feature library is optimized, and an aging scrap warning is triggered and the cause is analyzed. During the traceability and verification phase, based on the corrected calibration coefficients and aging update data, combined with relevant parameters and benchmark traceability data from each stage, the calibration validity is determined through three-level comparison and verification. If the calibration meets the standard, the calibration data is solidified; if it does not meet the standard, an early warning is triggered and a retry plan is pushed. During the closed-loop verification phase, based on the calibration validity results, the sensor's local processing unit performs full-cycle self-verification of the detection data during normal operation, sets up feedback optimization and multi-source fusion verification functions, dynamically fine-tunes calibration parameters, iteratively optimizes benchmarks and algorithms, and plans the next calibration time.

2. The gas sensor calibration method according to claim 1, characterized in that, The dynamic gas distribution system used in the benchmark traceability stage is a gas distribution system with controllable temperature, humidity, pressure, and gas composition; the multi-dimensional feature data includes six categories: sensor individual fingerprint features, environmental coupling response features, cross-interference sensitivity features, full life cycle aging features, metering traceability features, and operating condition adaptation features. The response data corresponding to each type of feature is bound one-to-one with the value of the national first-level standard gas for traceability. The dedicated benchmark library is synchronously stored in the sensor's local non-volatile storage unit, the cloud-based metrology platform, and the legal metrology traceability node.

3. The gas sensor calibration method according to claim 2, characterized in that, The sensor's individual fingerprint features include baseline response, baseline noise, response recovery time, and individual sensitive baseline features under all operating conditions of zero-air environment, temperature and humidity, and pressure gradient. The environmental coupling response characteristics include the static response, dynamic response curves, and temperature-humidity-pressure coupling interaction response characteristics under different temperature, humidity, and pressure conditions under the full range concentration gradient of the target gas. The cross-interference sensitivity features include the cross-response values ​​of interfering gases with different concentration gradients coexisting under a fixed concentration of the target gas, and the nonlinear correlation features between the concentration of the interfering gas and the response value; the full life cycle aging features include the sensitivity decay rate, response characteristic offset data, and the correlation features between aging stress and decay rate under different aging cycles. The metering traceability features include the deviation of the response data from the national first-level standard gas value, traceability link information, and dynamic correction factor for traceability deviation; the operating condition adaptation features include the benchmark response threshold and scenario-based benchmark adaptation coefficient under different application scenarios.

4. The gas sensor calibration method according to claim 3, characterized in that, The core operating parameters collected during the trigger acquisition phase include environmental operating condition parameters, sensor operating status parameters, and detection scene characteristic parameters, while interfering gas prediction parameters and aging trend parameters are collected simultaneously. The environmental operating parameters include real-time ambient temperature, relative humidity, atmospheric pressure, and the rate of change of environmental parameters; the sensor operating status parameters include real-time baseline response value, baseline fluctuation amplitude, response recovery time, operating electrical parameters, impedance characteristic value and rate of change, and response peak offset; the detection scene characteristic parameters include extreme values ​​of target gas detection concentration, fluctuation frequency, coexisting interfering gas components and concentrations, and trend data within a continuous period. The triggering conditions for the dual triggering mechanism include: baseline drift exceeding a preset threshold, continuous running time reaching the calibration warning cycle, environmental conditions exceeding the original calibration adaptation range, continuous abnormal detection data, sensor aging characteristics exceeding a preset threshold, environmental parameter change rate exceeding the interference prediction threshold, and interference gas concentration trend reaching the interference warning threshold.

5. A gas sensor calibration method according to claim 4, characterized in that, The zero-point decoupling stage combines a dedicated benchmark feature library to match real-time environmental operating parameters and scenario-based benchmark adaptation coefficients. During the decoupling calculation using the adaptive weighted environmental coupling decoupling algorithm, the decoupling weights of temperature, humidity, and pressure are dynamically adjusted based on the rate of change of environmental parameters and the predicted interference risk. The hardware anomaly identification process compares the corrected zero-point baseline with the zero-point noise threshold and impedance change threshold of the reference feature library using a dual threshold comparison. If either threshold exceeds the range, a hardware anomaly is determined, triggering a graded anomaly warning and determining whether to terminate the calibration based on the warning level. At the same time, the cause of the anomaly is located by combining individual fingerprint features. After zero-point calibration is completed, zero-point correction data and decoupling weights are recorded synchronously and updated to the local and cloud-based benchmark feature libraries.

6. The gas sensor calibration method according to claim 5, characterized in that, The dual decoupling linkage algorithm used in the range decoupling stage first performs inverse matrix operation on the cross-interference sensitivity matrix of the reference feature library through a nonlinear cross-interference decoupling algorithm. Combined with the trend of the change of the concentration of the interfering gas, the cross-interference of the coexisting interfering gas is eliminated to obtain the pure response value of the target gas. Then, based on the zero-point decoupling result, combined with the environmental coupling response matrix of the reference feature library and the temperature, humidity and pressure coupling interaction response characteristics, a second environmental decoupling is performed on the pure response value of the target gas to eliminate the secondary interference of environmental parameters. The nonlinear correction of the range response curve is based on the response characteristics of the full range concentration gradient in the benchmark feature library and is completed using a piecewise nonlinear weighted fitting algorithm to generate a range calibration coefficient that is adapted to the current working conditions. The scenario-based fine-tuning sets a low-concentration response enhancement factor for the ppb-level low-concentration detection range, sets a high-concentration saturation response correction function for industrial high-concentration scenarios, and then calls the scenario-based benchmark adaptation coefficient to complete the scenario-based fine-tuning of the range calibration coefficient.

7. A gas sensor calibration method according to claim 6, characterized in that, The multi-factor aging quantification assessment model used in the aging correction stage integrates four major assessment factors: sensor running time, environmental stress, response characteristic offset, and impedance change rate, and calculates the current aging quantification value, real-time sensitivity decay rate, and aging trend prediction value of the sensor by weighted calculation. The iterative correction of the calibration coefficients is based on the aging quantification value and the sensitivity decay rate. The zero-point and range calibration coefficients are dynamically iteratively corrected. The number of iterations of the calibration coefficients is adaptively adjusted according to the degree of aging to compensate for the sensitivity decay, response characteristic shift and baseline drift caused by aging. At the same time, the aging-related quantification values ​​and correction data are updated to the full life cycle aging feature sequence of the benchmark feature library. The aging and scrapping early warning system compares the quantitative value of aging, the predicted value of aging trend, and the preset aging and scrapping threshold. If any indicator exceeds the threshold, a sensor replacement early warning is triggered, and an analysis of the cause of aging is pushed out simultaneously.

8. A gas sensor calibration method according to claim 7, characterized in that, The three-level comparison verification in the traceability verification stage is as follows: the first-level verification compares the relative deviation of the calibration coefficient with the national first-level standard gas traceability benchmark. If the relative deviation is within the preset allowable measurement error range, the calibration coefficient is determined to have legal measurement traceability. The second-level verification checks the compatibility of the decoupling weights, aging correction factors, and benchmark feature library in this calibration. If the compatibility meets the standard, the reliability of the calibration results is confirmed. The consistency between the three-level verification prediction data and the actual calibration results is used to optimize the accuracy of subsequent prediction triggers; All three levels of verification have been completed and the calibration coefficients, calibration logs, traceability verification results, and anomaly records have been synchronously and permanently stored in the sensor’s local non-volatile storage unit and synchronized to the cloud-based metrology platform and legal metrology traceability nodes. If any verification fails, the calibration is deemed invalid, triggering a calibration failure warning. The cause of the failure is analyzed, and a targeted retry plan is provided.

9. A gas sensor calibration method according to claim 8, characterized in that, The full-cycle self-verification of the closed-loop verification stage includes real-time operating condition verification, calibration validity verification, and multi-source data fusion verification. The operating condition is monitored and verified in real time. When the ambient temperature, humidity, and pressure exceed the range of this calibration, an over-limit warning is automatically triggered, and the calibration parameters are dynamically fine-tuned simultaneously. The calibration validity verification involves real-time acquisition of sensor response characteristic parameters. When baseline drift, response recovery time, or sensitivity changes exceed the allowable threshold after calibration, the calibration is automatically determined to be invalid and a recalibration process is triggered. The triggering timing is optimized by combining the predicted data. The multi-source data fusion verification integrates local sensor detection data, cloud-based benchmark data, and legal metrological traceability data to perform cross-verification on real-time detection data. The feedback optimization function continuously collects sensor feature data during the detection process, updates it to the benchmark feature library regularly, optimizes benchmark features, decoupling algorithm weights, and aging assessment model parameters, and plans the next calibration time in advance by combining aging trends, operating condition changes, and interference prediction data.

10. A gas sensor calibration system, based on the method according to any one of claims 1-9, characterized in that, It includes a reference tracing unit, a trigger acquisition unit, a decoupling calibration unit, an aging correction unit, a tracing verification unit, a closed-loop verification unit, and a data storage and transmission unit; The reference traceability unit is deployed in a statutory metrology laboratory and includes a dynamic gas mixing system with controllable temperature, humidity, pressure and gas composition. It is used to collect multi-dimensional feature data of the sensor, build a reference feature library, complete the traceability binding with the national first-level standard gas, generate a dedicated reference library for a single sensor, set a dynamic update interface for reference features and synchronously store reference library data. The trigger acquisition unit is integrated into the sensor local processing module and has built-in temperature and humidity sensors and pressure sensors. It is used to continuously and synchronously acquire the core operating parameters of the sensor and adopts a dual triggering mechanism of passive and active prediction to lock the multi-source parameters after triggering and predict the risk type. The decoupling calibration unit includes a zero-point decoupling module and a range decoupling module. The zero-point decoupling module is used to complete the sensor zero-point calibration through an adaptive weighted environment coupling decoupling algorithm. The range decoupling module is used to complete the sensor range nonlinear calibration through a dual decoupling linkage algorithm, generate calibration coefficients adapted to the working conditions, and complete scenario-based fine-tuning. The aging correction unit is used to quantify the aging status of the sensor based on a multi-factor aging quantification evaluation model, iteratively correct the calibration coefficient, and trigger an aging and scrapping warning. The traceability verification unit is used to compare the calibration coefficient with the national first-level standard gas traceability data through a three-level comparison verification mechanism to determine the calibration validity. If the calibration is qualified, the calibration data is fixed; if it is not qualified, a retry plan is pushed. The closed-loop verification unit is used to realize full-cycle self-verification after calibration, dynamically fine-tune calibration parameters, iteratively optimize the benchmark and algorithm, and plan the next calibration time. The data storage and transmission unit is used to ensure real-time synchronization, traceable storage and transmission of data at each stage.

Images