An adaptive gas sensor calibration system based on environmental compensation

Abstract

This invention belongs to the field of gas sensor calibration technology. It discloses an adaptive gas sensor calibration system based on environmental compensation. The system extracts the asymmetric features of the transient response slope through a signal analysis module, achieving precise separation between environmentally actuated components and the bulk decay components. A false adaptive interception module establishes a baseline floating anchor point, effectively blocking erroneous zeroing operations caused by rising environmental background. An innovative temperature drift coupling evolution module is introduced to calculate the elastic modulus of the temperature drift response, generating a temperature-compensated evolution surface that dynamically updates with the sensor's aging state. Finally, an adaptive reconstruction output module completes temperature error elimination and lower bound clamping. This invention achieves deep decoupling between environmental factors and sensor aging, significantly improving the measurement accuracy and stability of gas sensors in complex environments.

Description

Technical Field

[0001] This invention relates to the field of gas sensor calibration technology, and more specifically, to an adaptive gas sensor calibration system based on environmental compensation. Background Technology

[0002] Traditional adaptive calibration systems for gas sensors commonly employ "automatic zero-point tracking" or "baseline regression" algorithms, which, in practical applications, fail to effectively distinguish between actual changes in environmental background concentration and sensor drift. Environmental changes such as long-term accumulation of pollutants in industrial areas and gas accumulation from human activity in enclosed spaces are often misinterpreted by calibration systems as sensor zero-point drift, forcing a "zeroing" operation and leading to a significant underestimation of actual gas concentrations. For example, in boundary monitoring of chemical industrial parks, when background pollutants slowly accumulate, sensor readings continuously rise. The adaptive algorithm mistakenly interprets this as sensor drift and performs baseline pulldown, incorrectly "correcting" concentrations above the warning threshold to a safe range, masking potential risks. After long-term operation, the physicochemical properties of sensors undergo subtle changes, altering their temperature response characteristics, and the carefully calibrated temperature compensation coefficient gradually becomes distorted. In environments with seasonal changes or large diurnal temperature variations, compensation model failure leads to periodic errors in the system, sometimes even resulting in reading trends opposite to actual concentration changes. The coupling effect of this temperature drift characteristic change and baseline drift makes traditional single compensation strategies inadequate, and the calibration effect continues to deteriorate with prolonged use. When sensor networks are deployed on a large scale in complex environments, these problems lead to a serious decline in data reliability, affecting not only the accuracy of environmental monitoring but also potentially causing safety hazards and economic losses due to false alarms or missed alarms. Existing technologies lack comprehensive calibration methods that fully consider environmental factors, sensor aging status, and the evolution of response characteristics, making it difficult for sensors to maintain long-term stable measurement performance in complex and ever-changing real-world application environments.

[0003] In view of this, the present invention proposes an adaptive gas sensor calibration system based on environmental compensation to solve the above problems. Summary of the Invention

[0004] To overcome the aforementioned deficiencies of the prior art and to achieve the above objectives, the present invention provides the following technical solution: an adaptive gas sensor calibration system based on environmental compensation, comprising: The signal analysis module is used to acquire the original concentration response sequence of the gas sensor and the synchronous ambient temperature and humidity sequence, extract the transient response slope asymmetry feature of the original concentration response sequence when the concentration changes stepwise, and separate the environmental actuation component and the bulk decay component based on the transient response slope asymmetry feature. The false adaptive interception module is used to calculate the background concentration fluctuation envelope based on the environmental actuation component, establish a baseline floating bottom anchor point, and block zero-point tracking operations for rising environmental background. The temperature drift coupling evolution module is used to map the bulk decay components to the temperature gradient space of the ambient temperature and humidity sequence, calculate the temperature drift response elastic modulus, and generate a dynamic temperature compensation evolution surface based on the temperature drift response elastic modulus. The adaptive reconstruction output module is used to apply the dynamic temperature compensation evolution surface to the original concentration response sequence to eliminate temperature errors, and uses the baseline floating bottom anchor point to clamp the data after temperature error elimination, and outputs the calibrated concentration data.

[0005] Furthermore, the transient response slope asymmetry feature of the original concentration response sequence during a concentration step change is extracted. Based on the transient response slope asymmetry feature, the environmental actuation component and the bulk decay component are separated, including: Detect the rising and falling edges of concentration in the original concentration response sequence, and calculate the rising edge response time constant and the falling edge recovery time constant, respectively. The ratio between the rising edge response time constant and the falling edge recovery time constant is calculated as a characteristic of transient response slope asymmetry. When the ratio remains within the historical baseline ratio range, the current baseline shift is determined to be caused by the accumulation of external environmental background concentration, and the high-frequency synchronous fluctuation component in the original concentration response sequence is extracted as the environmental actuation component. When the ratio deviates from the historical benchmark ratio range and is accompanied by an increase in the response time constant, the current baseline offset is determined to be caused by the aging of the sensor body due to the decrease in sensor sensitivity, and it is extracted as the decay component of the body.

[0006] Furthermore, based on the environmentally driven components, the background concentration fluctuation envelope is calculated, and a baseline floating support anchor point is established, including: Extract the local maximum and local minimum sequences of the environmental actuation components within a sliding time window; The local minimum value sequence is subjected to exponential smoothing to generate a lower bound of the background concentration fluctuation envelope that dynamically increases with the environmental background concentration. The values ​​of the lower boundary at each time point are defined as the baseline floating bottom anchor points, which adaptively rise as the actual environmental background concentration increases.

[0007] Furthermore, blocking zero-point tracking operations that target increased environmental background noise includes: Real-time monitoring of the baseline correction amount generated by the adaptive calibration algorithm; When the direction of the baseline correction is to pull the baseline down, and the target baseline value after the pull-down is lower than the current baseline floating bottom anchor point, the interception mechanism is triggered. The target baseline value is forcibly clamped to the value of the baseline floating bottom anchor point, preventing the adaptive calibration algorithm from performing a zeroing operation due to misjudging the increase in environmental background as a sensor zero-point drift, thus eliminating the risk of negative concentration output.

[0008] Furthermore, the bulk decay components are mapped to the temperature gradient space of the ambient temperature and humidity sequence, and the temperature drift response elastic modulus is calculated, including: The drift acceleration of the bulk decay component within a unit temperature change range is extracted and used as an indicator of the bulk decay component's sensitivity to temperature changes. In the temperature gradient space, the differential rate of change of sensitivity index at adjacent temperature nodes is calculated, and a second derivative matrix reflecting the change of sensor aging degree in temperature compensation response is constructed. The element representing the degree of nonlinear mutation in the second derivative matrix is ​​extracted as the temperature drift response elastic modulus. The temperature drift response elastic modulus quantifies the degree of distortion of the original temperature compensation coefficient after long-term operation of the sensor.

[0009] Furthermore, a dynamic temperature-compensated evolution surface is generated based on the temperature drift response elastic modulus, including: Retrieve the initial static temperature compensation surface calibrated at the sensor's factory settings; The elastic modulus of temperature drift response is used as a morphological operator to locally correct the curvature of the corresponding temperature nodes in the initial static temperature-compensated surface. Integral constraints are applied to the local warping correction in the time dimension to ensure the smoothness and continuity of the surface evolution, generating a dynamically temperature-compensated evolution surface that is updated in real time with the aging state of the sensor.

[0010] Furthermore, the dynamic temperature-compensated evolution surface is applied to the original concentration response sequence to eliminate temperature errors, and the data after temperature error elimination is clamped by a lower bound using a baseline floating anchor point, outputting calibrated concentration data, including: Input the current temperature and humidity data into the dynamic temperature compensation evolution surface to obtain the real-time temperature compensation coefficient; The original concentration response sequence was adjusted for sensitivity gain and temperature drift offset was subtracted using a real-time temperature compensation coefficient to obtain an intermediate calibration sequence. The intermediate calibration sequence is compared with the baseline floating bottom anchor point at the current time. If the intermediate calibration sequence is lower than the baseline floating bottom anchor point, the value of the baseline floating bottom anchor point is used as the calibrated concentration data at that time. If it is higher, the intermediate calibration sequence is directly used as the calibrated concentration data.

[0011] Furthermore, before extracting the asymmetric features of the transient response slope during a step change in concentration in the original concentration response sequence, the following steps are also included: The ratio of the signal from the target gas response channel to the interfering gas response channel in the original concentration response sequence is extracted and used as the cross-sensitivity scaling factor. After separating the environmental actuation component and the bulk decay component, the bulk decay component is cross-validated using a cross-sensitivity scaling factor. If the cross-sensitivity ratio factor remains unchanged while the bulk decay component increases, it is determined that the bulk decay component is mixed with interference from the same source gas in the environment, and the corresponding proportion of the bulk decay component is reassigned to the environmental actuation component.

[0012] Furthermore, after calculating the elastic modulus of temperature drift response, the following also includes: A time decay damping constraint is applied to the temperature drift response elastic modulus. When the change in the temperature drift response elastic modulus in adjacent periods exceeds the physical aging limit threshold, it is determined that the temperature drift response elastic modulus is disturbed by environmental transient thermal shock. The damping smoothing mechanism is triggered to replace the temperature drift response elastic modulus of the current period with the exponential moving average of the temperature drift response elastic modulus of the previous period, preventing distortion of the dynamic temperature compensation evolution surface caused by transient thermal shock.

[0013] Furthermore, this system also includes a calibration closed-loop verification module, used for: Calculate the compensated residual sequence of the calibrated concentration data relative to the original concentration response sequence; Extract the low-frequency trend term from the compensated residual sequence and calculate the mutual information entropy between the low-frequency trend term and the bulk decay component; If the mutual information entropy is lower than the preset information entropy threshold, it is determined that the temperature error is not completely eliminated. The low-frequency trend term of the compensation residual sequence is fed back to the temperature drift coupling evolution module as the prior compensation amount for calculating the temperature drift response elastic modulus in the next cycle.

[0014] The technical effects and advantages of the environmental compensation-based adaptive gas sensor calibration system of this invention are as follows: This invention accurately identifies real changes in environmental background concentration, preventing measurement distortion caused by overcompensation, and simultaneously overcomes the technical bottleneck of temperature compensation failure after long-term sensor operation. The dynamic evolution mechanism ensures high accuracy of temperature error compensation throughout the sensor's entire lifespan, overcoming the limitations of traditional static temperature compensation. Closed-loop verification and iterative optimization endow the system with continuous learning capabilities, preventing calibration accuracy from decaying over time and extending the sensor's effective lifespan. The deep decoupling method of this invention achieves highly reliable and stable gas concentration measurement, providing reliable technical support for fields such as environmental monitoring, industrial safety, and public health. Attached Figure Description

[0015] Figure 1 This is a schematic diagram of an adaptive gas sensor calibration system based on environmental compensation. Detailed Implementation

[0016] 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.

[0017] With the increasing demands for accuracy in gas sensors from environmental monitoring, industrial safety, and smart home applications, sensor calibration technology faces increasingly severe challenges. When traditional gas sensors operate for extended periods in complex environments, their performance is affected by multiple factors, including fluctuations in ambient temperature and humidity, background gas accumulation, and sensor aging. This leads to decreased detection accuracy, increased zero drift, and even abnormal phenomena such as negative output values.

[0018] Existing gas sensor calibration technologies mainly employ the following technical approaches: 1. Static Temperature Compensation Method: Before leaving the factory, sensor response curves are collected at different temperature points to generate a static temperature compensation lookup table. In actual use, the corresponding compensation coefficient is selected according to the ambient temperature. This method cannot cope with the aging effect after long-term operation of the sensor, and the compensation parameters do not adaptively adjust with the usage environment.

[0019] 2. Zero-point tracking method: Based on the assumption that the sensor output should be zero when no target gas is present, this method continuously monitors the sensor baseline and corrects for drift. However, this method is prone to misinterpreting a true increase in ambient background concentration as zero drift, leading to negative output values.

[0020] 3. Multi-sensor cross-calibration method: This method utilizes the cross-response characteristics of multiple sensors for complementary calibration. However, it requires the deployment of additional sensor arrays, is costly, and struggles to address systematic offsets caused by shared environmental factors.

[0021] However, existing technologies still have the following drawbacks: 1. Inability to distinguish between true increase in ambient background concentration and sensor zero drift: When there is a long-term accumulation of low concentration of target gas in the environment, the existing algorithm often misjudges it as zero drift, forcibly lowers the baseline, and causes the measurement result to be negative.

[0022] 2. Temperature compensation parameters cannot adapt to sensor aging: After long-term operation, the sensor's response characteristics to temperature changes will change, and the temperature compensation coefficient calibrated at the factory will no longer be applicable, resulting in increased measurement errors under temperature fluctuations.

[0023] 3. The symmetry assumption of response characteristics does not hold: Existing algorithms generally assume that the response rate of the sensor to the increase and decrease of concentration is symmetrical, ignoring the difference in response dynamics of the aging sensor in different directions.

[0024] To address the aforementioned issues, this application provides an adaptive gas sensor calibration system based on environmental compensation. This system analyzes the time asymmetry characteristics of the sensor response, distinguishes between changes in the environmental background and sensor aging, establishes a dynamically evolving temperature compensation model, and achieves accurate calibration of long-term operating gas sensors.

[0025] Figure 1 This is a schematic diagram of the structure of the adaptive gas sensor calibration system based on environmental compensation provided in an embodiment of this application. Figure 1 As shown, the system includes a signal analysis module, a false adaptive interception module, a temperature drift coupling evolution module, and an adaptive reconstruction output module.

[0026] The signal analysis module acquires the raw concentration response sequence of the gas sensor and the synchronized ambient temperature and humidity sequence. It extracts the transient response slope asymmetry feature of the raw concentration response sequence during concentration step changes, and separates the environmental actuation component and the bulk decay component based on this asymmetry. The environmental actuation component represents the signal change caused by external environmental factors (such as background concentration fluctuations and interfering gases), while the bulk decay component reflects the signal drift caused by internal factors such as sensor aging and decreased sensitivity. The transient response slope asymmetry feature refers to the difference in the sensor's response rate to concentration increases and decreases; this difference exhibits different pattern characteristics under normal sensor operation and aging conditions. By analyzing this asymmetry, the system can accurately distinguish the contributions of environmental changes and sensor decay, providing a basis for subsequent calibration.

[0027] The spurious adaptive interception module calculates the background concentration fluctuation envelope based on environmental actuation components, establishes a baseline floating anchor point, and blocks zero-point tracking operations triggered by rising environmental background. This module solves the core problem of traditional adaptive algorithms failing to distinguish between a true increase in environmental background concentration and sensor zero-point drift. The baseline floating anchor point dynamically adjusts with the actual change in environmental background concentration, ensuring that the system does not mistakenly identify an increase in environmental background as zero-point drift and perform unnecessary baseline lowering operations. This effectively avoids the risk of negative concentration output and improves measurement reliability.

[0028] The temperature drift coupling evolution module maps the bulk decay components to the temperature gradient space of the ambient temperature and humidity sequence, calculates the temperature drift response elastic modulus, and generates a dynamic temperature compensation evolution surface based on the temperature drift response elastic modulus. The temperature response characteristics of a sensor change with usage time and aging, and static temperature compensation coefficients cannot adapt to this dynamic change. This module establishes a mapping relationship between the decay components and temperature changes, quantitatively describes the impact of aging on the temperature response, and enables dynamic updates of the temperature compensation parameters, solving the problem of traditional temperature compensation methods failing after sensor aging.

[0029] The adaptive reconstruction output module applies a dynamically temperature-compensated evolution surface to the original concentration response sequence to eliminate temperature errors. It then uses a baseline floating anchor point to perform lower-bound clamping on the temperature-error-eliminated data, outputting calibrated concentration data. This module integrates the processing results of the preceding modules, performing temperature error compensation and baseline adjustment on the original signal to ensure the accuracy and reliability of the output data. The lower-bound clamping mechanism effectively prevents negative outputs due to overcompensation, improving the system's stability and availability in complex environments.

[0030] The modules are connected via wired and / or wireless means to enable data transmission between them.

[0031] In this embodiment of the invention, the detailed implementation steps for extracting the transient response slope asymmetry feature of the original concentration response sequence during a concentration step change, and separating the environmental actuation component and the bulk decay component based on the transient response slope asymmetry feature, include: In the original concentration response sequence, rising and falling edges of concentration are detected, and the rising edge response time constant and falling edge recovery time constant are calculated respectively. An adaptive threshold algorithm is used in the detection process. First, the first derivative of the signal is calculated, and intervals where the derivative value exceeds the background noise level of 3σ are identified as candidate edges. Then, morphological processing is used to eliminate false edges, finally determining the valid rising and falling edges. The time constant calculation uses an exponential fitting method, and the formula is fitted to the rising edge segment data: ; in, for Concentration value at time, This is the initial concentration. For the range of concentration change, The moment when the concentration begins to change. This is the rising edge response time constant.

[0032] Fitting formula for the falling edge segment data: ; in, This is the concentration value after stabilization. The moment when the concentration begins to decrease. This is the recovery time constant for the falling edge.

[0033] The ratio of the rising edge response time constant to the falling edge recovery time constant is calculated as a characteristic of transient response slope asymmetry. The time constant ratio is defined as: ; This ratio is an important indicator of the sensor's response kinetics, reflecting the difference in response rate to increases and decreases in concentration. A properly functioning sensor exhibits relatively stable... The value is determined by environmental factors and physical aging, which can alter this ratio in different ways.

[0034] When the ratio remains within the historical baseline ratio range, the current baseline shift is determined to be caused by the accumulation of background concentration in the external environment. The high-frequency synchronous fluctuation component in the original concentration response sequence is then extracted as the environmentally actuated component. The historical baseline ratio range is determined by statistically analyzing the data under normal sensor conditions. The value distribution is determined, and it is usually taken as... Scope, of which The mean, The standard deviation is given. High-frequency synchronous wave components are extracted through wavelet decomposition. By selecting appropriate wavelet basis functions and decomposition scales, the original sequence is decomposed into sub-signals of different frequency bands. Components with energy concentrated in the high-frequency band and highly correlated with changes in environmental parameters are extracted as environmental actuation components.

[0035] When the ratio deviates from the historical baseline ratio range and is accompanied by an increase in the response time constant, the current baseline shift is determined to be caused by intrinsic aging due to decreased sensor sensitivity, and this is extracted as an intrinsic decay component. The deviation determination employs a dynamic threshold method, comprehensively considering both the magnitude and duration of the deviation to ensure sufficient tolerance for short-term fluctuations while enabling timely detection of persistent changes. The extraction of the intrinsic decay component utilizes a long-short separation technique, employing low-pass filtering and trend analysis to extract long-term, slow-changing trends from the original sequence. These trends typically manifest as unidirectional baseline drift and persistent changes in response characteristics.

[0036] In this embodiment of the invention, the detailed implementation steps for establishing a baseline floating support anchor point based on the background concentration fluctuation envelope calculated from the environmental actuation component include: Local maxima and local minima sequences of environmental actuation components are extracted within a sliding time window. The sliding window size is dynamically adjusted based on sensor response characteristics and environmental fluctuation cycles, typically 3-5 times the response time constant, to ensure the capture of the complete fluctuation cycle. Extremum extraction employs a peak detection algorithm combined with morphological filtering to eliminate noise interference and improve the accuracy of extremum point identification. Local maxima reflect short-term peaks in environmental concentration, while local minima represent the baseline level of the environmental background; together, they depict the fluctuation range and pattern of environmental concentration.

[0037] Exponential smoothing is applied to the local minimum value sequence to generate a lower bound for the background concentration fluctuation envelope that dynamically increases with the ambient background concentration. Exponential smoothing effectively captures the slow changing trend of background concentration while filtering out the influence of short-term fluctuations. The smoothing formula is: ; in, for The smoothing value at time, for Local minimum at time t, This is the smoothed value from the previous time step. This is the smoothing coefficient. The smoothing coefficient is typically set in the range of 0.1-0.3, with smaller values ​​being preferable. The value can better smooth out short-term fluctuations and highlight long-term trend changes. The lower boundary line directly reflects the bottom profile of the environmental background concentration, providing a basis for establishing a bottoming anchor point for baseline floating.

[0038] The values ​​of the lower boundary at each time point are defined as the baseline floating anchor points, which adaptively rise as the actual ambient background concentration increases. The anchor point represents the reasonable minimum value of the sensor response under current environmental conditions; readings below this value typically indicate overcompensation or algorithmic misjudgment. The adaptive rising characteristic of the anchor point ensures that the system can adapt to long-term changes in ambient background concentration, avoiding the limitations of traditional fixed baseline methods. When the ambient background concentration continues to rise (such as due to accumulated pollution in industrial areas or gas accumulation from indoor activities), the anchor point will rise accordingly to prevent the system from mistakenly interpreting this real change as zero-point drift.

[0039] In this embodiment of the invention, the detailed implementation steps for blocking zero-point tracking operations in response to increased environmental background include: This system monitors the baseline corrections generated by the adaptive calibration algorithm in real time. Traditional adaptive calibration algorithms typically estimate zero-point drift and generate corrections based on the statistical characteristics of the signal (such as the sliding minimum, quantiles, etc.). This system continuously monitors the direction and magnitude of these corrections to identify adjustments that may lead to overcompensation. The monitoring process employs multi-window comparison technology, simultaneously analyzing correction trends within short-term, medium-term, and long-term windows to improve the reliability and robustness of the judgment.

[0040] An interception mechanism is triggered when the baseline correction direction is to lower the baseline, and the target baseline value after the correction is lower than the current baseline floating support anchor point. Lowering the baseline is a standard handling method for zero-point drift in adaptive algorithms, but when the environmental background is actually rising, this operation may lead to over-compensation of the signal. The system identifies potential erroneous compensation behaviors by comparing the corrected target baseline with the current anchor point value. The comparison uses boundary-based difference analysis, taking into account measurement uncertainty and system noise, to ensure that only correction operations significantly lower than the anchor point are intercepted.

[0041] This mechanism clamps the target baseline value to the value of the baseline floating anchor point, preventing the adaptive calibration algorithm from misinterpreting the increase in environmental background as sensor zero-point drift and performing a zeroing operation, thus eliminating the risk of negative concentration output. The clamping operation is achieved by replacing the baseline value generated by the original algorithm, ensuring that the final applied baseline correction will not be lower than the actual level of the environmental background. This mechanism effectively prevents the system from erroneously identifying the background increase as zero-point drift and overcompensating under conditions of continuously increasing environmental background concentration (such as pollution accumulation in industrial areas or gas accumulation in confined spaces), leading to an underestimation of the actual concentration or even negative values. This significantly improves the reliability and accuracy of the system in complex environments.

[0042] In this embodiment of the invention, the detailed implementation steps for mapping the bulk decay component to the temperature gradient space of the ambient temperature and humidity sequence and calculating the elastic modulus of the temperature drift response include: The drift acceleration of the bulk decay component within a unit temperature change range is extracted as an indicator of the bulk decay component's sensitivity to temperature changes. The extraction process first involves arranging the temperature sequence in ascending order and segmenting it, with each segment's temperature change controlled within a 1-2°C range. Then, the rate of change of the bulk decay component within each temperature segment, i.e., the drift acceleration, is calculated. Finally, a mapping relationship between temperature and drift acceleration is constructed to reflect the sensor's sensitivity to decay at different temperatures. The drift acceleration calculation formula is: ; in, For temperature Drift acceleration, For the bulk decay component, For temperature change The change in decay component within the interval. Drift acceleration reflects the effect of temperature changes on the sensor's aging rate and is a key indicator for quantifying temperature drift response.

[0043] In the temperature gradient space, the differential rate of change of sensitivity indices at adjacent temperature nodes is calculated to construct a second-order derivative matrix reflecting the change in temperature compensation response due to sensor aging. The temperature gradient space is a multi-dimensional space describing the rate of temperature change, considering not only the absolute value of temperature but also its direction and rate of change. The differential rate of change is calculated using the central difference method, performing second-order differences on the sensitivity indices at adjacent temperature nodes to construct a matrix reflecting the relationship between the temperature gradient and decay acceleration. The formula for calculating the elements of the second-order derivative matrix is: ; in, The matrix elements represent temperature. and temperature gradient The second derivative value under the given condition, For temperature The drift acceleration is calculated. The second derivative matrix comprehensively describes the nonlinear changes in the temperature response characteristics of the sensor after aging, providing a mathematical basis for dynamic temperature compensation.

[0044] The element representing the degree of nonlinear abrupt change in the second derivative matrix is ​​extracted as the temperature drift response elastic modulus. This elastic modulus quantifies the distortion of the original temperature compensation coefficient after long-term sensor operation. The elastic modulus extraction employs eigenvalue decomposition, performing singular value decomposition (SVD) on the second derivative matrix to extract the eigenvalues ​​corresponding to the main eigenvectors as the elastic modulus. The calculation formula is as follows: ; in, The elastic modulus is the response modulus to temperature drift. Representation matrix The maximum eigenvalue. The elastic modulus directly reflects the degree of change in the sensor's temperature response characteristics. The larger the value, the more severe the distortion of the original temperature compensation coefficient, requiring a greater degree of dynamic adjustment. This parameter organically combines the physical process of sensor aging with the mathematical model of temperature compensation, realizing adaptive optimization of the temperature compensation parameter.

[0045] In this embodiment of the invention, the detailed implementation steps for generating a dynamic temperature-compensated evolution surface based on the temperature drift response elastic modulus include: Retrieve the initial static temperature compensation surface calibrated at the sensor's factory. This initial static temperature compensation surface is typically provided by the sensor manufacturer and is established based on the temperature response characteristics of the new sensor under standard conditions, containing a multi-dimensional mapping relationship between temperature and compensation coefficients. The surface can be represented as a function. ,in For temperature, For humidity, This corresponds to the compensation coefficient. The system first loads the base surface from the device configuration or cloud database as the starting point for dynamic adjustment.

[0046] The elastic modulus of temperature drift response is used as a morphological operator to locally correct the curvature of corresponding temperature nodes in the initial static temperature-compensated surface. The morphological operator is a mathematical transformation that can selectively deform the original surface based on the distribution characteristics of the elastic modulus. The correction process first identifies temperature regions where the elastic modulus changes significantly; these regions are typically where the sensor's temperature response changes most noticeably. Then, the local curvature correction amount is calculated, and point transformations are performed on the original surface. Finally, smoothing constraints are applied to ensure the continuity and smoothness of the corrected surface. The curvature correction calculation formula is: ; in, This is the corrected compensation coefficient. The original compensation coefficient, For temperature The elastic modulus below, The decay function controls the correction effect as it deviates from the reference point with temperature. The decay pattern is observed. This local correction method based on a physical model can accurately capture the changing patterns of the temperature response after sensor aging, enabling targeted compensation and adjustment.

[0047] Integral constraints are applied to local warpage corrections over time to ensure the smooth and continuous evolution of the surface, generating a dynamically temperature-compensated evolution surface that updates in real time with the sensor's aging state. Integral constraints are a key mechanism to prevent abrupt surface changes; by considering the historical cumulative effect of corrections, the gradual and stable evolution of the surface is ensured. The constraint process employs a time-window integration method, combining the current correction amount with the cumulative effect of historical corrections to generate a final correction value with a smooth transition. The update formula for the evolving surface is: ; in, for Dynamic temperature-compensated surface at any given time. This is the currently calculated correction value. This is the dynamically warmed surface from the previous moment. To update the coefficients, the rate of surface evolution is controlled. The dynamic temperature-compensated evolution surface breaks through the limitations of traditional static temperature compensation, and can adjust in real time according to the aging state of the sensor, ensuring the long-term effectiveness of temperature compensation and significantly improving the accuracy and stability of the sensor during long-term use.

[0048] In this embodiment of the invention, the detailed implementation steps of applying a dynamic temperature-compensated evolution surface to the original concentration response sequence to eliminate temperature errors, and using a baseline floating anchor point to clamp the data after temperature error elimination to output calibrated concentration data include: The system inputs the current temperature and humidity data into a dynamic temperature-compensated evolution surface to obtain real-time temperature compensation coefficients. First, it acquires temperature and humidity data synchronized with concentration measurements. Then, using interpolation or nearest neighbor search methods, it locates the corresponding points on the dynamic temperature-compensated evolution surface and extracts the real-time temperature compensation coefficients. The compensation coefficients consist of a gain adjustment factor and an offset correction, corresponding to the sensor sensitivity and the temperature-dependent changes at zero point, respectively. The system employs a multi-point query optimization algorithm to improve retrieval efficiency on the high-dimensional surface and ensure real-time processing performance.

[0049] An intermediate calibration sequence is obtained by adjusting the sensitivity gain and subtracting temperature drift offset from the original concentration response sequence using a real-time temperature compensation coefficient. Sensitivity gain adjustment compensates for changes in sensor sensitivity at different temperatures, while temperature drift offset subtraction corrects for zero-point changes with temperature. The calibration calculation formula is as follows: ; in, This is an intermediate calibration sequence. This is the original concentration response sequence. Temperature and humidity Offset correction amount below, Temperature and humidity The sensitivity gain factor is calculated. Both parameters are extracted from the dynamic temperature compensation evolution surface and dynamically adjusted according to the sensor's aging state and current environmental conditions to ensure the accuracy and adaptability of the compensation.

[0050] The intermediate calibration sequence is compared with the current baseline floating support anchor point. If the intermediate calibration sequence is lower than the baseline floating support anchor point, the value of the baseline floating support anchor point is used as the calibrated concentration data for that moment; otherwise, the intermediate calibration sequence is directly used as the calibrated concentration data. Lower bound clamping is a safety mechanism to prevent overcompensation, ensuring that the output data does not fall below the actual level of the environmental background. The comparison process uses real-time point-to-point comparison, setting appropriate comparison tolerances to account for measurement uncertainty and avoid frequent switching caused by noise. When the intermediate sequence value is lower than the anchor point, it indicates that temperature compensation may be excessive, and the system limits the output to the anchor point level to prevent unreasonable low or negative values; when the intermediate sequence value is higher than the anchor point, it indicates that the signal reflects the actual gas concentration change, and the system directly outputs the intermediate sequence value to ensure measurement sensitivity and accuracy. This intelligent clamping mechanism significantly improves the system's reliability in complex environments and effectively solves the overcorrection problem that may occur with traditional compensation methods when the environmental background concentration changes.

[0051] In this embodiment of the invention, before the signal analysis module extracts the asymmetric features of the transient response slope during a step change in concentration of the original concentration response sequence, it further includes: The signal ratio between the target gas response channel and the interfering gas response channel in the original concentration response sequence is extracted as the cross-sensitivity scaling factor. The cross-sensitivity scaling factor reflects the sensor's relative sensitivity to the target gas and interfering gases and is an important indicator for evaluating signal purity. The extraction process first identifies the target channel and potential interfering channels in the sensor array, then calculates the signal ratio of the two types of channels within a synchronization time window, and constructs a statistical distribution model of the ratio. A dynamic window method is used for ratio calculation; the window size is adaptively adjusted according to the signal change rate to ensure complete capture of response characteristics. A normally functioning sensor has a relatively stable cross-sensitivity scaling factor; abnormal changes in this factor usually indicate the presence of a specific interfering gas in the measurement environment or a change in sensor selectivity.

[0052] After separating the environmentally actuated component and the bulk decay component, the bulk decay component is cross-validated using a cross-sensitivity scaling factor. Cross-validation is a crucial step in improving separation accuracy, verifying the rationality of the separation results through consistency checks of independent features. The validation process first analyzes the stability of the cross-sensitivity scaling factor during the variation of the bulk decay component to determine whether the decay is caused by a single mechanism; then, it calculates the temporal correlation between the two components to identify potential association patterns; finally, it uses statistical significance tests to determine whether there is any contamination from environmentally sourced gases. This multi-dimensional cross-validation mechanism significantly improves the system's ability to identify complex environmental interferences, ensuring the accuracy and reliability of the separation results.

[0053] If the bulk decay component increases while the cross-sensitivity scaling factor remains unchanged, it is determined that environmental co-source gas interference has entered the bulk decay component. The corresponding proportion of this interference is then reclassified to the environmental actuation component. This reclassification is a correction mechanism to rectify separation errors, ensuring accurate classification of environmental and bulk factors. The reclassification process first establishes a mapping relationship between the cross-sensitivity scaling factor and the concentration of environmental co-source gas. Then, based on this mapping, the proportion of mixed environmental gas is estimated. Finally, the corresponding proportion of the signal is transferred from the bulk decay component to the environmental actuation component. The transfer calculation uses a weighted allocation method to ensure that the total signal amount is conserved while achieving reasonable distribution. This dynamic correction mechanism based on cross-validation effectively solves the technical challenge of accurately distinguishing between environmental and sensor bulk factors in complex gas environments, improving the system's adaptability to multi-source interference and calibration accuracy.

[0054] In this embodiment of the invention, after the temperature drift coupling evolution module calculates the temperature drift response elastic modulus, it further includes: A time-decay damping constraint is applied to the elastic modulus of the temperature drift response. When the change in the elastic modulus of the temperature drift response within adjacent periods exceeds the physical aging limit threshold, it is determined that the elastic modulus of the temperature drift response is disturbed by transient thermal shock from the environment. The time-decay damping constraint is a stabilizing mechanism to prevent model overfitting, and a reasonable rate of change limit is set based on the physical process of sensor aging. The constraint process first establishes a theoretical model of sensor aging to determine the maximum physically possible rate of change; then, it monitors the change in elastic modulus in adjacent periods and compares it with the theoretical limit; when the rate of change exceeds the physically possible range, it is determined to be an abnormal aging phenomenon, possibly caused by the temporary effects of environmental thermal shock (such as rapid temperature changes or heat source radiation). This anomaly detection mechanism based on a physical model effectively distinguishes between real aging and environmental disturbances, improving the stability and reliability of system parameter updates.

[0055] A damping smoothing mechanism is triggered, replacing the current period's temperature drift response elastic modulus with the exponential moving average of the temperature drift response elastic modulus from previous periods. This prevents distortion of the dynamic temperature compensation evolution surface caused by transient thermal shock. Damping smoothing is a crucial step in anomaly handling, eliminating the impact of anomalous fluctuations through smoothing constraints on historical data. The smoothing process employs the exponentially weighted moving average (EWMA) method, applying a decay-weighted average to the historical elastic modulus sequence. The smoothing formula is as follows: ; in, for The smooth elastic modulus at time t. This represents the original elastic modulus at the previous moment. The smooth elastic modulus of the previous moment. This is the smoothing coefficient (usually set in the range of 0.1-0.3). Smoothing effectively suppresses the influence of transient disturbances, ensures the stable and gradual change of the elastic modulus, prevents the temperature-compensated surface from undergoing severe distortion due to short-term environmental fluctuations, and further improves the stability and reliability of the system under complex temperature environments.

[0056] In this embodiment of the invention, the system further includes a calibration closed-loop verification module, used for: The compensated residual sequence of the calibrated concentration data relative to the original concentration response sequence is calculated. The compensated residual sequence is a crucial indicator for evaluating the calibration effectiveness, reflecting signal components that were not fully compensated during the calibration process. The calculation process first restores the calibrated data to the original signal space through an inverse transform, then calculates the difference between the calibrated data and the actual original signal to obtain a point-to-point residual sequence. Normalization is applied to the residual calculation to eliminate the influence of signal amplitude differences, facilitating comparative analysis of results under different conditions. The complete residual sequence contains all the information from the calibration process, providing detailed evidence for subsequent analysis.

[0057] The low-frequency trend term of the compensated residual sequence is extracted, and the mutual information entropy between the low-frequency trend term and the bulk decay component is calculated. The low-frequency trend term extraction employs a low-pass filtering method, with the cutoff frequency set to 1 / 10 of the signal's main variation period, effectively preserving long-term trends while filtering short-term fluctuations. The mutual information entropy calculation, based on information theory principles, quantifies the statistical dependence between two sequences; the formula is: ; in, For mutual information entropy, This is a low-frequency trend term. For the bulk decay component, For joint probability distribution, and This represents a marginal probability distribution. The lower the mutual information entropy value, the weaker the correlation between the two sequences, meaning that the ontological decay component has been more completely compensated; conversely, a high mutual information entropy indicates incomplete compensation, and the residual still contains significant ontological decay information.

[0058] If the mutual information entropy is lower than the preset information entropy threshold, the temperature error is deemed incompletely eliminated. The low-frequency trend term of the compensation residual sequence is then fed back to the temperature drift coupling evolution module as a priori compensation for calculating the elastic modulus of the temperature drift response in the next cycle, achieving decoupling and closed-loop iteration. The preset information entropy threshold is determined by analyzing the statistical distribution under ideal compensation scenarios, typically set to 2-3 times the mutual information entropy at the background noise level. Closed-loop feedback is a key mechanism for improving the system's adaptive capability, gradually enhancing calibration accuracy through continuous evaluation and optimization. The feedback process introduces the residual trend as prior knowledge into the elastic modulus calculation of the next cycle, enabling parameter fine-tuning based on actual results. The calculation formula is: ; in, This is the corrected elastic modulus. These are the original calculated values. This represents the low-frequency trend term of the residual. The learning rate controls the intensity of feedback adjustments. This closed-loop optimization mechanism based on residual analysis enables the system to continuously learn and improve, constantly enhancing the accuracy of temperature compensation and effectively solving the technical challenges of traditional open-loop calibration methods in adapting to complex environments and long-term sensor degradation.

[0059] The adaptive gas sensor calibration system based on environmental compensation proposed in this application will be further explained below with specific application examples: In an industrial environmental monitoring scenario, a metal oxide semiconductor (MOS) gas sensor was deployed to monitor volatile organic compounds (VOCs) in the environment. The sensor had been operating continuously for over a year, and its temperature compensation performance began to decline. Furthermore, when VOC concentrations in the environment were low but persistent, the system frequently produced negative concentration readings.

[0060] After applying this system, the signal analysis module first acquires the sensor's raw concentration response sequence and synchronized temperature and humidity sequences over a week. Analyzing this data, the signal analysis module detects multiple concentration step events and calculates the rise-edge response time constant and fall-edge recovery time constant for each event. This ratio is approximately 0.6-0.8 for a normal sensor, but in this aged sensor, the ratio has become 1.2-1.5, and the response time is significantly prolonged, indicating that the sensor has undergone significant aging.

[0061] Based on this characteristic, the signal analysis module successfully separated the original signal into an environmentally driven component (containing the actual fluctuations in the concentration of environmental VOCs) and a bulk decay component (reflecting the aging state of the sensor). By analyzing the signal ratio between the target VOCs channel and the cross-sensitive CO channel, the system further confirmed that the bulk decay component was indeed mainly caused by sensor aging, rather than interference from the same source in the environment.

[0062] The spurious adaptive interception module analyzes the fluctuation characteristics of environmental actuated components and discovers a persistent VOCs background concentration of approximately 0.3 ppm. The module establishes a baseline floating anchor point sequence, dynamically adjusting it according to this background concentration. When the system's built-in zero-point tracking algorithm attempts to lower the baseline below 0.1 ppm, the spurious adaptive interception module triggers an interception mechanism, preventing this operation and effectively avoiding negative output.

[0063] The temperature drift coupling evolution module analyzed the response characteristics of the bulk decay component at different temperatures, finding that the sensor's sensitivity to temperature changes was significantly higher in the 15-25°C range than in other temperature ranges, exhibiting a larger temperature drift response elastic modulus. Based on this characteristic, the module dynamically corrected the sensor's original temperature compensation surface, particularly enhancing the compensation intensity within the 15-25°C range. When a sudden high temperature (32°C) occurred, the system detected a sudden change in the temperature drift response elastic modulus, promptly triggering a damping smoothing mechanism to prevent overcompensation.

[0064] The adaptive reconstruction output module applies optimized temperature compensation parameters to the raw data, while simultaneously using a baseline floating anchor point for lower bound clamping. For example, on a typical day, when the ambient temperature rose from 18°C ​​to 24°C at 9:00 AM, the uncalibrated sensor output dropped from 0.5 ppm to 0.1 ppm, exhibiting a spurious negative correlation. After calibration by this system, the output stabilized at 0.4-0.5 ppm, accurately reflecting the actual concentration level of VOCs in the environment.

[0065] The calibration closed-loop verification module, by calculating the mutual information entropy between the compensation residual and the bulk decay component, discovered that approximately 15% of the systematic error remained unresolved after the initial calibration. The system feeds this residual information back to the temperature drift coupling evolution module, which further optimizes the temperature compensation parameters in the next calibration cycle, achieving the best final calibration result.

[0066] After calibration by this system, the temperature response stability of the gas sensor improved by 85%, negative output events were completely eliminated, and the measurement accuracy was improved by 70% compared to the uncalibrated state, restoring the aging sensor, which had been running for a year, to a performance level close to that of a new sensor.

[0067] Through the above embodiments, the environmental compensation-based adaptive gas sensor calibration system provided in this application effectively distinguishes between changes in environmental background and sensor aging by analyzing the time dynamics characteristics of the sensor response, and establishes a dynamically evolving temperature compensation model, thereby achieving accurate calibration of long-term operating gas sensors. This system effectively solves the problems in existing technologies, such as the inability to distinguish between actual increases in environmental background and sensor zero-point drift, the inability of temperature compensation parameters to adapt to sensor aging, and the invalidation of the symmetry assumption of response characteristics. It significantly improves the accuracy and reliability of gas sensors during long-term operation in complex environments.

[0068] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing embodiments or make equivalent substitutions for some of the technical features. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

[0069] It should be noted that all formulas in this manual are calculated by removing dimensions and taking their numerical values. The formulas are derived from software simulations based on a large amount of collected data to obtain the most recent real-world results. The preset parameters and thresholds in the formulas are set by those skilled in the art according to the actual situation.

[0070] Although embodiments of the invention have been shown and described, those skilled in the art will understand 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 claims and their equivalents.

Claims

1. An adaptive gas sensor calibration system based on environmental compensation, characterized in that, include: The signal analysis module is used to acquire the original concentration response sequence of the gas sensor and the synchronized ambient temperature and humidity sequence, extract the transient response slope asymmetry feature of the original concentration response sequence during concentration step changes, and separate the environmental actuation component and the bulk decay component based on the transient response slope asymmetry feature, including: In the original concentration response sequence, the rising edge and falling edge of concentration are detected, and the rising edge response time constant and falling edge recovery time constant are calculated respectively. The ratio between the rising edge response time constant and the falling edge recovery time constant is calculated as the transient response slope asymmetry feature; When the ratio remains within the historical baseline ratio range, it is determined that the current baseline shift is caused by the accumulation of external environmental background concentration, and the high-frequency synchronous fluctuation component in the original concentration response sequence is extracted as the environmental actuation component. When the ratio deviates from the historical benchmark ratio range and is accompanied by an increase in the response time constant, the current baseline shift is determined to be body aging caused by a decrease in sensor sensitivity, and it is extracted as the body decay component. The false adaptive interception module is used to calculate the background concentration fluctuation envelope based on the environmental actuation component, establish a baseline floating bottom anchor point, and block zero-point tracking operations for rising environmental background. The temperature drift coupling evolution module is used to map the bulk decay component to the temperature gradient space of the ambient temperature and humidity sequence, calculate the temperature drift response elastic modulus, and generate a dynamic temperature compensation evolution surface based on the temperature drift response elastic modulus. The step of mapping the bulk decay component to the temperature gradient space of the ambient temperature and humidity sequence and calculating the temperature drift response elastic modulus includes: The drift acceleration of the bulk decay component within a unit temperature change range is extracted and used as an index of the sensitivity of the bulk decay component to temperature changes. In the temperature gradient space, the differential rate of change of the sensitivity index at adjacent temperature nodes is calculated, and a second derivative matrix reflecting the change of sensor aging degree in temperature compensation response is constructed. The element representing the degree of nonlinear mutation in the second derivative matrix is ​​extracted as the temperature drift response elastic modulus; The adaptive reconstruction output module is used to apply the dynamic temperature compensation evolution surface to the original concentration response sequence to eliminate temperature errors, and to use the baseline floating bottom anchor point to clamp the data after temperature error elimination to output calibrated concentration data.

2. The system according to claim 1, characterized in that, The step of calculating the background concentration fluctuation envelope based on the environmental actuation components and establishing a baseline floating support anchor point includes: Extract the local maximum and local minimum sequences of the environmental actuation components within the sliding time window; The local minimum value sequence is subjected to exponential smoothing to generate a lower bound of the background concentration fluctuation envelope that dynamically increases with the environmental background concentration. The values ​​of the lower boundary line at each time point are defined as the baseline floating bottom anchor points, which adaptively rise as the actual concentration of the environmental background increases.

3. The system according to claim 2, characterized in that, The blocking of zero-point tracking operations in response to increased environmental background includes: Real-time monitoring of the baseline correction amount generated by the adaptive calibration algorithm; When the direction of the baseline correction is to pull the baseline down, and the target baseline value after the pull-down is lower than the current baseline floating bottom anchor point, the interception mechanism is triggered. The target baseline value is forcibly clamped to the value of the baseline floating bottom anchor point.

4. The system according to claim 3, characterized in that, The generation of a dynamic temperature-compensated evolution surface based on the temperature drift response elastic modulus includes: Retrieve the initial static temperature compensation surface calibrated at the sensor's factory settings; The temperature drift response elastic modulus is used as a morphological operator to locally correct the curvature of the corresponding temperature nodes in the initial static temperature-compensated surface. The local warpage correction is integrally constrained in the time dimension to ensure the smoothness and continuity of the surface evolution, thereby generating the dynamic temperature-compensated evolution surface that is updated in real time with the aging state of the sensor.

5. The system according to claim 4, characterized in that, The process of applying the dynamic temperature-compensated evolution surface to the original concentration response sequence to eliminate temperature errors, and using the baseline floating anchor point to clamp the data after temperature error elimination to output calibrated concentration data, includes: Input the current temperature and humidity data into the dynamic temperature compensation evolution surface to obtain the real-time temperature compensation coefficient; The original concentration response sequence is adjusted for sensitivity gain and temperature drift offset by using the real-time temperature compensation coefficient to obtain an intermediate calibration sequence. The intermediate calibration sequence is compared with the baseline floating support anchor point at the current time. If the intermediate calibration sequence is lower than the baseline floating support anchor point, the value of the baseline floating support anchor point is output as the calibrated concentration data at that time. If it is higher, the intermediate calibration sequence is directly output as the calibrated concentration data.

6. The system according to claim 2, characterized in that, Before extracting the transient response slope asymmetry feature of the original concentration response sequence during a concentration step change, the method further includes: The signal ratio of the target gas response channel to the interfering gas response channel in the original concentration response sequence is extracted and used as the cross-sensitivity scaling factor. After separating the environmental actuation component and the bulk decay component, the bulk decay component is cross-validated using the cross-sensitivity scaling factor. If the cross-sensitivity ratio factor remains unchanged while the bulk decay component increases, it is determined that the bulk decay component is mixed with interference from the same source gas in the environment, and the corresponding ratio value in the bulk decay component is reassigned to the environmental actuation component.

7. The system according to claim 4, characterized in that, After calculating the elastic modulus of temperature drift response, the method further includes: A time decay damping constraint is applied to the temperature drift response elastic modulus. When the change in the temperature drift response elastic modulus in adjacent periods exceeds the physical aging limit threshold, it is determined that the temperature drift response elastic modulus is disturbed by environmental transient thermal shock. A damping smoothing mechanism is triggered to replace the temperature drift response elastic modulus of the current period with the exponential moving average of the temperature drift response elastic modulus of the previous period, preventing distortion of the dynamic temperature compensation evolution surface caused by transient thermal shock.

8. The system according to claim 7, characterized in that, It also includes a calibration closed-loop verification module, used for: Calculate the compensated residual sequence of the calibrated concentration data relative to the original concentration response sequence; Extract the low-frequency trend term of the compensated residual sequence, and calculate the mutual information entropy between the low-frequency trend term and the bulk decay component; If the mutual information entropy is lower than the preset information entropy threshold, it is determined that the temperature error is not completely eliminated. The low-frequency trend term of the compensation residual sequence is fed back to the temperature drift coupling evolution module as the prior compensation amount for calculating the temperature drift response elastic modulus in the next cycle.

Images