Humidity drift correction method and system for a mems sensor

Abstract

This invention relates to the field of MEMS sensor calibration technology, and discloses a method and system for humidity drift correction of MEMS sensors. The method includes tracking and identifying external humidity sources and extracting their features, establishing a time-varying model to predict humidity drift values, and simultaneously monitoring the internal operating state of the sensor to obtain internal feature parameters. The predicted values ​​are fused with the internal parameters to construct a drift influence quantity and generate an initial compensation quantity. Then, a sensor output feedback sequence is introduced to dynamically iteratively optimize the initial compensation quantity, generating an optimized drift compensation control sequence, which is finally applied to the sensor output to complete the correction. This method improves the accuracy of humidity drift correction and its adaptability to time-varying environments through feature fusion modeling of the external environment and internal state, and a closed-loop dynamic optimization mechanism based on output feedback.

Description

Technical Field

[0001] This invention relates to the field of MEMS sensor calibration technology, specifically to a method and system for humidity drift correction of MEMS sensors. Background Technology

[0002] In practical applications, the accuracy and stability of MEMS sensors are highly susceptible to changes in ambient humidity, making humidity drift a key issue limiting their high-precision applications. Existing calibration techniques primarily rely on direct measurements from external humidity sensors, using fixed mapping relationships or compensation curves to statically correct the sensor output. These methods treat external humidity as an independent source of interference, failing to deeply characterize the dynamic interaction between humidity and the sensor's own physical state.

[0003] Current technical solutions have shortcomings. Static compensation models cannot adapt to the rapid time-varying characteristics of humidity environments and the dynamic changes in the internal state of sensors, resulting in insufficient long-term stability. Relying solely on external humidity measurements cannot reflect the differences in the internal effects of humidity on the sensor's sensitive units. Open-loop compensation mechanisms lack the ability to self-verify and adaptively adjust the correction effect; when sensor characteristics change due to aging or sudden environmental changes, the original compensation model is prone to failure, leading to a decrease in correction accuracy. Existing methods fail to integrate the characteristics of external humidity effects with the sensor's internal operating state during modeling and lack a mechanism for closed-loop optimization of compensation quantities using the sensor's own output, limiting the accuracy and robustness of the correction. Summary of the Invention

[0004] The purpose of this invention is to provide a humidity drift correction method and system for MEMS sensors to solve the problems mentioned in the background art.

[0005] To achieve the above objectives, the present invention provides a humidity drift correction method for a MEMS sensor, the method comprising:

[0006] Track and identify external humidity sources acting on the target MEMS sensor, and extract the humidity effect characteristics of the external humidity sources;

[0007] Based on the aforementioned humidity characteristics, a time-varying humidity feature model is established for the target MEMS sensor, and the humidity drift prediction value is calculated in real time using the time-varying humidity feature model.

[0008] Monitor the internal operating state of the target MEMS sensor and obtain internal characteristic parameters that reflect the current state of the sensor's sensitive unit;

[0009] By combining the humidity drift prediction value with the internal feature parameters, the sensor drift impact value under the current environment is constructed;

[0010] An initial humidity drift baseline compensation amount is generated for the target MEMS sensor based on the sensor drift influence amount;

[0011] By introducing a sensor output feedback sequence, the initial humidity drift reference compensation amount is dynamically iteratively optimized to generate an optimized drift compensation control sequence.

[0012] The optimized drift compensation control sequence is applied to the original output of the target MEMS sensor to correct the humidity drift.

[0013] Preferably, the step of tracking and identifying the external humidity source acting on the target MEMS sensor and extracting the humidity effect characteristics of the external humidity source includes:

[0014] Scan the deployment space of the target MEMS sensor to locate at least one major physical source of humidity release;

[0015] Within a preset humidity sampling period, the surface humidity data of the physical humidity release source and the background humidity data of the ambient space are recorded simultaneously.

[0016] By performing differential analysis and pattern recognition on the surface humidity data and background humidity data, humidity effect characteristics that characterize the intensity change and diffusion law of the external humidity source are extracted.

[0017] Preferably, the step of establishing a time-varying humidity characteristic model for the target MEMS sensor based on the humidity effect characteristics, and using the time-varying humidity characteristic model to calculate the humidity drift prediction value in real time includes:

[0018] The intensity variation law in the humidity effect characteristics is mapped to a set of time-varying weight coefficients, and the diffusion law is mapped to a spatial influence function;

[0019] By combining the time-varying weighting coefficients and the spatial influence function, a time-varying humidity feature model is constructed to describe the dynamic process of humidity propagating from the source to the sensor location.

[0020] The real-time acquired environmental parameters are input into the time-varying humidity feature model to calculate the predicted humidity drift value that the target MEMS sensor is expected to experience in the next calibration cycle.

[0021] Preferably, the monitoring of the internal operating state of the target MEMS sensor and the acquisition of internal feature parameters reflecting the current state of the sensor's sensing unit include:

[0022] Acquire the intrinsic electrical response signal of the target MEMS sensor's sensitive structure under conditions without external excitation;

[0023] The intrinsic electrical response signal is analyzed in both the frequency domain and amplitude domain to separate the resonant characteristic parameters related to the mechanical properties of the sensitive structure and the impedance characteristic parameters related to the material properties.

[0024] The resonance characteristic parameter and the impedance characteristic parameter together constitute the internal characteristic parameter reflecting the current state of the sensor's sensitive unit.

[0025] Preferably, the step of fusing the humidity drift prediction value with the internal feature parameters to construct the sensor drift impact amount under the current environment includes:

[0026] Establish a coupling relationship mapping table between humidity stress and sensor status, and use the coupling relationship mapping table to convert the humidity drift prediction value into the corresponding theoretical state offset;

[0027] The theoretical state offset is correlated with the actual state change calculated from the internal feature parameters;

[0028] Based on the results of the correlation matching, a comprehensive weighting factor is calculated, and the theoretical state offset is weighted and fused using the weighting factor to generate the sensor drift impact in the current environment.

[0029] Preferably, generating the initial humidity drift reference compensation amount for the target MEMS sensor based on the sensor drift influence amount includes:

[0030] The preset lookup table for drift amount to compensation amount is queried, and a preliminary compensation amount reference value is obtained based on the amplitude of the sensor drift influence amount.

[0031] Based on the changing trend of the sensor drift influence, a trend adjustment factor is applied to the preliminary compensation reference value to obtain the adjusted preliminary compensation amount.

[0032] Based on the historical calibration records of the target MEMS sensor, the adjusted preliminary compensation amount is empirically calibrated to generate the final initial humidity drift reference compensation amount.

[0033] Preferably, the step of introducing a sensor output feedback sequence to dynamically iteratively optimize the initial humidity drift reference compensation amount and generate an optimized drift compensation control sequence includes:

[0034] The actual output sequence of the target MEMS sensor after the initial humidity drift reference compensation amount is continuously acquired and used as the sensor output feedback sequence.

[0035] Calculate the statistical stability measure of the sensor output feedback sequence and compare it with a preset stability threshold to generate a stability deviation signal;

[0036] The stability deviation signal is input into a gradient optimizer, which calculates the incremental adjustment amount of the initial humidity drift reference compensation.

[0037] The initial humidity drift baseline compensation amount is iteratively corrected multiple times based on the incremental adjustment amount until the statistical stability measure meets the requirements. The compensation amounts generated by each iterative correction are arranged in chronological order to generate an optimized drift compensation control sequence.

[0038] The step of calculating the statistical stability measure of the sensor output feedback sequence and comparing it with a preset stability threshold to generate a stability deviation signal includes:

[0039] Within the sliding time window, calculate the variance and Allan variance of the sensor output feedback sequence;

[0040] The variance and the Allan variance are weighted and summed to obtain a comprehensive statistical stability measure of the sensor output feedback sequence;

[0041] The difference between the comprehensive statistical stability measure and the preset stability threshold is used as the stability deviation signal.

[0042] Preferably, the step of extracting humidity effect features characterizing the intensity change and diffusion law of the external humidity source through differential analysis and pattern recognition of the surface humidity data and background humidity data includes:

[0043] Calculate the difference between the surface humidity data and the background humidity data at the same timestamp to form a humidity difference sequence;

[0044] Time-domain analysis was performed on the humidity difference sequence to extract its periodic and trend components, which were used as characteristic components of humidity effect to characterize the intensity change law.

[0045] By analyzing the gradient distribution of background humidity data at different spatial locations and combining it with environmental fluid parameters, we fit the attenuation model parameters of humidity diffusion, which serve as the characteristic components of humidity effect characterizing the diffusion law.

[0046] Preferably, the step of performing frequency domain and amplitude domain analysis on the intrinsic electrical response signal to separate resonant characteristic parameters related to the mechanical properties of the sensitive structure and impedance characteristic parameters related to material properties includes:

[0047] Perform a spectrum transformation on the intrinsic electrical response signal to identify the main resonant frequency and quality factor in the signal spectrum, and record the main resonant frequency and quality factor as resonant characteristic parameters;

[0048] Under a specific DC bias, the voltage-current relationship curve of the intrinsic electrical response signal is measured, and the equivalent DC impedance and equivalent capacitance are calculated from the voltage-current relationship curve. The equivalent DC impedance and equivalent capacitance are recorded as impedance characteristic parameters.

[0049] Preferably, the present invention also includes a humidity drift correction system for a MEMS sensor, the system including a memory, a processor, and a computer program stored in the memory and running on the processor, wherein the processor, when executing the computer program, implements the steps of the humidity drift correction method for a MEMS sensor as described above.

[0050] Compared with the prior art, the beneficial effects of the present invention are:

[0051] By introducing a feedback sequence based on the actual sensor output, the initial humidity drift baseline compensation is dynamically and iteratively optimized to generate a continuously updated drift compensation control sequence. This method constructs a closed-loop correction system, enabling the compensation amount to adaptively adjust and optimize based on the sensor's real-time output performance. This overcomes the shortcomings of traditional open-loop compensation models, which remain fixed once set and cannot respond to sensor characteristic drift or sudden environmental changes. The online iterative update of the compensation control sequence ensures that the correction action closely follows changes in the sensor's actual operating state, thereby improving the adaptability of the correction strategy and its ability to maintain high accuracy during long-term use.

[0052] This study integrates predicted values ​​from a time-varying humidity characteristic model of an external humidity source with internal characteristic parameters reflecting the current state of the sensor's sensitive unit to construct the sensor drift impact under the current environment. It utilizes both external humidity prediction information and internal sensor operating status information. The external model predicts the trend and characteristics of humidity changes, while the internal parameters directly characterize the degree of physical impact of humidity on the sensor's core sensitive unit. Fusing these two types of heterogeneous information allows for a more refined and comprehensive characterization of the complex drift mechanism caused by the interaction between humidity and the sensor's internal physical state. Compared to traditional methods that rely solely on a single external humidity value or internal temperature, this fusion modeling approach more accurately characterizes the nature of drift, laying a more reliable model foundation for generating more precise initial compensation values. Attached Figure Description

[0053] Figure 1 This is a schematic diagram illustrating the working principle of the humidity drift correction method for MEMS sensors described in this invention.

[0054] Figure 2 A flowchart for tracking and identifying external humidity sources and extracting the characteristics of humidity effects;

[0055] Figure 3A flowchart for establishing a time-varying humidity characteristic model and calculating humidity drift prediction values;

[0056] Figure 4 The time-varying curves of the resonant characteristic parameters of the MEMS sensor;

[0057] Figure 5 The figure shows the effect of sliding window size on stability metrics in MEMS sensor drift compensation optimization. Detailed Implementation

[0058] 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.

[0059] Please see Figure 1 This invention provides a humidity drift correction method for MEMS sensors. The method includes: tracking and identifying external humidity sources acting on a target MEMS sensor, and extracting the humidity effect characteristics of the external humidity sources. Based on the humidity effect characteristics, establishing a time-varying humidity feature model for the target MEMS sensor, and using the time-varying humidity feature model to calculate the predicted humidity drift value in real time. Monitoring the internal operating state of the target MEMS sensor and obtaining internal feature parameters reflecting the current state of the sensor's sensitive unit. Fusing the predicted humidity drift value and the internal feature parameters to construct the sensor drift influence quantity under the current environment. Generating an initial humidity drift benchmark compensation quantity for the target MEMS sensor based on the sensor drift influence quantity. Introducing a sensor output feedback sequence to dynamically iteratively optimize the initial humidity drift benchmark compensation quantity, generating an optimized drift compensation control sequence. Applying the optimized drift compensation control sequence to the original output of the target MEMS sensor to complete the humidity drift correction.

[0060] Example 1: See Figure 2The system scans the deployment space of the target MEMS sensor to locate at least one major physical humidity release source. Within a preset humidity sampling period, it simultaneously records surface humidity data from the physical humidity release source and background humidity data from the surrounding environment. The difference between the surface humidity data and the background humidity data at the same timestamp is calculated, forming a humidity difference sequence. A time-domain analysis is performed on the humidity difference sequence to extract its periodic and trend components, which serve as characteristic components representing the intensity variation of humidity. The gradient distribution of the background humidity data at different spatial locations is analyzed, and combined with environmental fluid parameters, attenuation model parameters for humidity diffusion are fitted, serving as characteristic components representing the diffusion law of humidity. Through difference analysis and pattern recognition of the surface humidity data and background humidity data, humidity characteristics representing the intensity variation and diffusion law of the external humidity source are extracted.

[0061] In specific implementations, the deployment space of the target MEMS sensor is scanned to locate at least one major physical humidity release source. This physical humidity release source can be a device that generates water vapor, an open liquid surface, or a damp surface. In some embodiments, an auxiliary temperature and humidity sensor network deployed in the space is used for environmental scanning. This network measures the relative humidity and temperature gradient at different locations. Based on the distribution pattern of the humidity gradient field, a spatial region with a significantly higher and persistent humidity value than the environmental background is identified, and this region is determined as the location of the physical humidity release source. In specific implementations, within a preset humidity sampling period, surface humidity data of the physical humidity release source and background humidity data of the surrounding space are recorded simultaneously. Surface humidity data is collected by a humidity sensor close to the surface of the physical humidity release source, while background humidity data is collected by an environmental humidity sensor deployed away from the physical humidity release source in a location with good air circulation.

[0062] In specific implementations, time-domain analysis is performed on the humidity difference sequence to extract its periodic and trend components, which serve as characteristic components of humidity effect representing the intensity variation law. The time-domain analysis includes applying a moving average filter to the humidity difference sequence to extract low-frequency trend components, and performing a fast Fourier transform on the filtered sequence to identify its significant periodic components. The extracted trend components describe the long-term, slow changes in the intensity of physical humidity release source activity, while the periodic components describe the regular fluctuations of source activity. In some embodiments, the gradient distribution of background humidity data at different spatial locations is analyzed, and combined with environmental fluid parameters, the attenuation model parameters for humidity diffusion are fitted, serving as characteristic components of humidity effect representing the diffusion law. Environmental fluid parameters include air velocity and prevailing wind direction. It can be understood that by correlating the values ​​of background humidity data at multiple spatial measurement points with the distance from the measurement points to the physical humidity release source, and fitting them using a pre-defined diffusion model, key parameters in the model are obtained. These parameters describe the rate and morphology of humidity diffusion from the source to the surrounding space. In practical implementation, the pre-defined diffusion model uses a decay function based on physical laws to describe the propagation process of humidity from the physical humidity release source to the surrounding space. This model establishes a mathematical relationship between the measured values ​​of background humidity data at multiple spatial measurement points and the distance from the corresponding measurement point to the source. During the calibration phase, multiple pre-defined measurement points at different distances from the physical humidity release source are selected within the deployment space of the target MEMS sensor. The same acquisition sensor is used to sequentially stay at each pre-defined measurement point for a predetermined duration, synchronously collecting background humidity data at each point and recording the spatial coordinates of the corresponding measurement points. After collecting data from all measurement points, model parameters can be fitted based on this dataset. Using the collected dataset, a curve fitting algorithm, such as least squares regression, is employed to substitute the distance variable and humidity measurement values ​​into the pre-defined model structure for parameter identification. In practical implementation, a model relationship describing humidity diffusion decay can be expressed as: Where: symbol This indicates the distance from the physical source of humidity release is Background humidity data values ​​measured or calculated at the location, symbol This represents a reference humidity value at or very close to the surface of a physical humidity release source, indicated by the symbol [symbol missing]. This represents the humidity diffusion attenuation coefficient obtained through data fitting, with the symbol... This represents the spatial distance from the observation location to the physical source of humidity release. The humidity diffusion attenuation coefficient is calculated by fitting background humidity data to distance data. Humidity diffusion attenuation coefficient That is, the humidity effect characteristic component, which is the core component characterizing the diffusion law.

[0063] Example 2: See Figure 3The intensity variation law of the humidity effect characteristics is mapped to a set of time-varying weight coefficients, and the diffusion law is mapped to a spatial influence function. Combining the time-varying weight coefficients and the spatial influence function, a time-varying humidity feature model is constructed to describe the dynamic process of humidity propagation from the source to the sensor location. Real-time acquired environmental parameters are input into the time-varying humidity feature model to calculate the predicted humidity drift value expected to be experienced by the target MEMS sensor within a future calibration cycle.

[0064] In specific implementations, the intensity variation pattern of humidity effects is mapped to a set of time-varying weighting coefficients. This intensity variation pattern includes trend and periodic components extracted from the humidity difference sequence. In some embodiments, the trend component is described by a time-dependent polynomial function, the coefficients of which determine the long-term baseline of the time-varying weighting coefficients. The periodic component is described by a superposition of sine and cosine functions, the amplitude and frequency parameters of which determine the periodic modulation portion of the time-varying weighting coefficients. The time-varying weighting coefficients are a comprehensive representation of the mapping results of the trend and periodic components; their values ​​change dynamically over time, reflecting the time-varying characteristics of the external humidity source's activity intensity. In specific implementations, the diffusion pattern of humidity effects is mapped to a spatial influence function. The core of the diffusion pattern is the humidity diffusion attenuation coefficient. The spatial influence function defines the attenuation relationship of humidity influence from the physical humidity release source to the location of the target MEMS sensor. It can be understood that the spatial influence function is a function of distance and the humidity diffusion attenuation coefficient, used to quantify the loss of humidity during spatial propagation.

[0065] By combining time-varying weighting coefficients and a spatial influence function, a time-varying humidity characteristic model is constructed to describe the dynamic process of humidity propagation from the source to the sensor location. This model expresses the theoretical humidity influence intensity of the physical humidity release source on the target MEMS sensor at any given time. In practical implementation, the time-varying humidity characteristic model can be characterized as the product of the time-varying weighting coefficients and the spatial influence function. This product indicates that the theoretical humidity influence at the sensor location is simultaneously constrained by both the temporal variation and spatial attenuation of the source intensity. The mathematical expression of the time-varying humidity characteristic model can be understood as follows: Where: symbol Indicates at time The theoretical influence of humidity at the target MEMS sensor location, symbol Indicates at time Time-varying weighting coefficients, sign This indicates the distance from the target MEMS sensor to the physical source of humidity release. This is the spatial influence function of the independent variable. (Symbol) Represents a time variable, symbol This represents the distance variable. In some embodiments, real-time acquired environmental parameters are input into the time-varying humidity feature model. These environmental parameters include the current time information and the real-time distance information between the target MEMS sensor and the physical humidity release source. The corresponding time-varying weighting coefficients are calculated from the time information based on the current time. The corresponding spatial influence function value is calculated based on real-time distance information. By substituting the time-varying weighting coefficients and spatial influence function values ​​into the time-varying humidity characteristic model, the predicted humidity drift values ​​expected to be experienced by the target MEMS sensor in the current and future calibration cycles are calculated.

[0066] Example 3: Acquiring the intrinsic electrical response signal of the target MEMS sensor's sensitive structure under conditions without external excitation. A spectral transformation is performed on the intrinsic electrical response signal to identify the main resonance peak frequency and quality factor in the signal spectrum, which are recorded as resonance characteristic parameters. Under a specific DC bias, the voltage-current relationship curve of the intrinsic electrical response signal is measured, and the equivalent DC impedance and equivalent capacitance are calculated from the voltage-current relationship curve, recorded as impedance characteristic parameters. Frequency and amplitude domain analyses are performed on the intrinsic electrical response signal to separate the resonance characteristic parameters related to the mechanical properties of the sensitive structure and the impedance characteristic parameters related to material properties; the resonance characteristic parameters and the impedance characteristic parameters together constitute internal characteristic parameters reflecting the current state of the sensor's sensitive unit.

[0067] In specific implementations, the intrinsic electrical response signal of the target MEMS sensor's sensitive structure is acquired under conditions of no external excitation. No external excitation means the sensor is not in an active measurement state, and its sensitive structure is only affected by environmental factors and its own electrical bias. In some embodiments, by disconnecting the target MEMS sensor from an external signal source, only a low-level AC scanning signal with constant amplitude or a step test signal is applied to the sensor's drive or detection port, thereby acquiring the sensor's electrical output as the intrinsic electrical response signal when there is no input of the measured physical quantity. A spectrum transformation is performed on the acquired intrinsic electrical response signal using a Fast Fourier Transform algorithm to convert the time-domain intrinsic electrical response signal to the frequency domain. In specific implementations, the main resonance frequency and quality factor in the signal spectrum are identified. The main resonance frequency corresponds to the frequency value of the peak with the largest amplitude in the spectrum, and the quality factor is obtained by calculating the ratio of the main resonance frequency to the half-power bandwidth. It can be understood that the main resonance frequency and quality factor directly reflect the mechanical dynamic characteristics of the sensitive structure, such as equivalent stiffness and damping. The identified main resonance peak frequency and quality factor values ​​are recorded as resonance characteristic parameters. Under a specific DC bias, the voltage-current relationship curve of the intrinsic electrical response signal is measured. The specific DC bias is a pre-set constant voltage or current applied across the sensor's sensitive structure. Optionally, a series of stepped DC voltages are applied across the sensitive structure using a precision source measurement unit, while simultaneously measuring the current flowing through the sensitive structure, thereby obtaining a set of voltage-current data pairs.

[0068] In specific implementations, the equivalent DC impedance and equivalent capacitance are calculated from the voltage-current relationship curve. The equivalent DC impedance is obtained by selecting the slope of the voltage-current relationship curve near a certain operating point or by calculating the voltage-to-current ratio. In some embodiments, while applying a DC bias, a small-amplitude AC signal is superimposed. By measuring the amplitude and phase relationship of the AC voltage and AC current, the complex impedance of the sensitive structure under a specific bias is calculated, and the equivalent capacitance value is then derived from its imaginary part. The measurement relationship can be expressed as: Where: symbol Represents the calculated equivalent capacitance, symbol The frequency of the superimposed small-amplitude AC signal is indicated by the symbol. Indicates frequency The complex impedance obtained by measurement The imaginary part. (Symbol) This represents the constant pi. It can be understood that the equivalent DC impedance and equivalent capacitance reflect the conductivity and dielectric properties of the sensitive structural material. The calculated equivalent DC impedance and equivalent capacitance values ​​are recorded as impedance characteristic parameters. Optionally, the acquired intrinsic electrical response signal is analyzed in both the frequency domain and amplitude domain. Frequency domain analysis, i.e., the aforementioned spectral transformation, is used to separate the resonant characteristic parameters, while amplitude domain analysis, i.e., the aforementioned voltage-current relationship measurement, is used to separate the impedance characteristic parameters.

[0069] See Figure 4 This is a time-varying curve of the resonant characteristic parameters of a MEMS sensor. The main resonance peak frequency and the quality factor show a negative correlation trend; both fluctuate over time, reflecting the dynamic characteristics of the MEMS sensor's sensitive structure under no external excitation; the error bars reflect the discreteness of the measurement data. This type of chart is used for internal characteristic parameter analysis in humidity drift correction of MEMS sensors. By monitoring the changes in the main resonance peak frequency and quality factor, the mechanical properties of the sensor's sensitive structure can be evaluated, providing data for subsequent drift compensation.

[0070] Example 4: Establish a coupling relationship mapping table between humidity stress and sensor state, and use the coupling relationship mapping table to convert the humidity drift prediction value into the corresponding theoretical state offset. Perform correlation matching between the theoretical state offset and the actual state change calculated from the internal feature parameters. Based on the correlation matching result, calculate a comprehensive weighting factor, and use the weighting factor to perform weighted fusion of the theoretical state offset to generate the sensor drift impact amount under the current environment. Query a preset drift amount to compensation amount lookup table, and obtain a preliminary compensation amount reference value based on the amplitude address of the sensor drift impact amount. Apply a trend adjustment factor to the preliminary compensation amount reference value according to the changing trend of the sensor drift impact amount to obtain the adjusted preliminary compensation amount. Based on the historical calibration record of the target MEMS sensor, perform empirical calibration on the adjusted preliminary compensation amount to generate the final initial humidity drift benchmark compensation amount.

[0071] In practical implementation, a coupling relationship mapping table between humidity stress and sensor state is established. This mapping table is constructed based on prior calibration data or theoretical models of the target MEMS sensor, establishing a correspondence between external humidity levels and theoretical changes in the state of the sensor's sensitive units. In some embodiments, the coupling relationship mapping table is stored in the form of a data table, where the row or column indexes represent different humidity conditions, and the table content represents the theoretical offset of each state parameter of the sensor's sensitive units under those humidity conditions. A simplified coupling relationship mapping table is shown in Table 1.

[0072] Table 1: Humidity-State Theoretical Offset Mapping Table

[0073]

[0074] In practical implementation, a coupling relationship mapping table is used to convert the humidity drift prediction value into the corresponding theoretical state offset. For example, for the theoretical humidity condition corresponding to the humidity drift prediction value calculated at a certain moment, a set of values, including the theoretical offset of the resonant frequency and the theoretical offset of the equivalent DC impedance under that humidity condition, are obtained by searching or interpolating in the coupling relationship mapping table. This set of values ​​constitutes the theoretical state offset. The theoretical state offset is then correlated with the actual state change calculated from internal characteristic parameters. The internal characteristic parameters include the currently monitored resonant characteristic parameters and impedance characteristic parameters. The actual state change is calculated by comparing the difference between the current internal characteristic parameters and the internal characteristic parameters under a reference state. It can be understood that the purpose of correlation matching is to evaluate the consistency between the theoretical state offset and the actual state change. The matching operation can calculate the cosine similarity or correlation coefficient between the two sets of offset vectors. Based on the result of the correlation matching, a comprehensive weighting factor is calculated. The comprehensive weighting factor is a scalar coefficient between 0 and 1. The higher the correlation matching result, the closer the value of the comprehensive weighting factor is to 1. By using a comprehensive weighting factor to weight and fuse the theoretical state offsets, the sensor drift impact in the current environment is generated. The mathematical expression of the weighted fusion is: Where: symbol This represents the sensor drift effect after fusion, with the symbol... Represents the overall weighting factor, symbol This represents the theoretical state offset. Sensor drift effect. It is a vector or scalar that represents the total expected offset of the sensor state caused by the current humidity effect after internal state verification and weighting.

[0075] A preliminary compensation reference value is obtained by querying a pre-defined drift-to-compensation lookup table and addressing the magnitude of the sensor drift influence. The drift-to-compensation lookup table establishes a mapping relationship between the sensor drift influence and the required preliminary electrical or numerical compensation. In some embodiments, the magnitude of the sensor drift influence refers to the magnitude of its vector or the absolute value of its key components. Based on the changing trend of the sensor drift influence, a trend adjustment factor is applied to the preliminary compensation reference value to obtain the adjusted preliminary compensation value. The changing trend of the sensor drift influence is determined by the first difference or derivative of its historical sequence. The trend adjustment factor is used to add or subtract an adjustment amount proportional to the trend strength based on the preliminary compensation reference value. Optionally, the adjusted preliminary compensation value is empirically calibrated based on the historical calibration records of the target MEMS sensor. The historical calibration records store past similar sensor drift influences and the ultimately verified effective actual compensation values. Empirical calibration generates the final initial humidity drift reference compensation by finding the entry in the historical record that is closest to the current situation and fine-tuning the adjusted preliminary compensation amount with its actual compensation amount.

[0076] Example 5: Continuously acquire the actual output sequence of the target MEMS sensor after applying the initial humidity drift benchmark compensation amount, as the sensor output feedback sequence. Calculate the statistical stability measure of the sensor output feedback sequence and compare it with a preset stability threshold to generate a stability deviation signal. Within a sliding time window, calculate the variance and Allan variance of the sensor output feedback sequence; perform a weighted sum of the variance and Allan variance to obtain the comprehensive statistical stability measure of the sensor output feedback sequence; subtract the comprehensive statistical stability measure from the preset stability threshold, and use the difference as the stability deviation signal. Input the stability deviation signal into a gradient optimizer, which calculates the incremental adjustment amount for the initial humidity drift benchmark compensation amount. Iterate and correct the initial humidity drift benchmark compensation amount multiple times based on the incremental adjustment amount until the statistical stability measure meets the requirements. Arrange the compensation amounts generated by each iteration in chronological order to generate an optimized drift compensation control sequence.

[0077] In specific implementations, the actual output sequence of the target MEMS sensor after applying an initial humidity drift baseline compensation is continuously acquired as the sensor output feedback sequence. The actual output sequence refers to the original reading of the target MEMS sensor after initial compensation or the output data stream after preliminary processing. In some embodiments, the acquisition operation is performed at fixed time intervals, forming a discrete data sequence arranged in chronological order. The length of the sensor output feedback sequence covers one or more preset evaluation windows. The statistical stability measure of the sensor output feedback sequence is calculated and compared with a preset stability threshold to generate a stability deviation signal. In specific implementations, within a sliding time window, the variance and Allan variance of the sensor output feedback sequence are calculated. The length of the sliding time window is preset according to the sensor characteristics and application requirements. The variance calculates the degree of dispersion of the sensor output feedback sequence around its mean within the sliding time window. The Allan variance calculates the frequency stability characteristics of the sensor output feedback sequence at different clustering times within the sliding time window. The variance of the sensor output feedback sequence and the Allan variance of the sensor output feedback sequence are weighted and summed to obtain a comprehensive statistical stability measure of the sensor output feedback sequence. The weighting coefficients of the variance and Allan variance are assigned based on prior knowledge of the sensor noise type.

[0078] The stability deviation signal is input to a gradient optimizer, which calculates the incremental adjustment amount for the initial humidity drift baseline compensation. The gradient optimizer is an optimization algorithm based on the gradient descent principle. It calculates the adjustment direction and step size of the compensation amount to reduce the stability deviation signal based on the input stability deviation signal and the current compensation amount. In some embodiments, the incremental adjustment amount is determined by the gradient optimizer based on the negative gradient direction of the stability deviation signal and an adaptive learning rate. The calculation relationship of the incremental adjustment amount can be expressed as: Where: symbol Represents the calculated incremental adjustment amount, symbol The learning rate parameter of the gradient optimizer is represented by the symbol [symbol missing]. Indicates the current compensation amount objective function for variables gradient, objective function Defined by the stability deviation signal. Symbol The objective function to be minimized is directly related to the stability deviation signal. The initial humidity drift baseline compensation is iteratively corrected multiple times based on the incremental adjustment. During each iteration, the stability deviation signal is recalculated based on the latest sensor output feedback sequence, and a new incremental adjustment is generated again through the gradient optimizer to update the compensation. Optionally, the iteration process continues until the comprehensive statistical stability metric meets the requirements. Meeting the requirements means that the comprehensive statistical stability metric is less than or equal to a preset stability threshold, or its change is less than a convergence criterion in consecutive iterations. The compensation values ​​generated by each iteration are arranged in chronological order to generate an optimized drift compensation control sequence, which is an ordered list containing all intermediate compensation values ​​from the initial compensation value to the final converged compensation value.

[0079] See Figure 5 This is a graph showing the impact of sliding window size on stability metrics in MEMS sensor drift compensation optimization. As the sliding window size increases, both variance and Allan variance show an upward trend; the overall statistical stability metric increases synchronously, showing a positive correlation with the window size. This type of graph is used in the iterative optimization stage of humidity drift correction compensation for MEMS sensors. By analyzing the impact of different sliding windows on stability metrics, the optimal window size can be determined, improving the accuracy and efficiency of compensation optimization. A window that is too small will lead to large statistical errors in stability metrics, while a window that is too large will reduce real-time performance; a balance must be chosen based on sensor characteristics and application requirements.

[0080] 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.

[0081] 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 method for humidity drift correction of a MEMS sensor, characterized in that, The method includes: Track and identify external humidity sources acting on the target MEMS sensor, and extract the humidity effect characteristics of the external humidity sources; Based on the aforementioned humidity characteristics, a time-varying humidity feature model is established for the target MEMS sensor, and the humidity drift prediction value is calculated in real time using the time-varying humidity feature model. Monitor the internal operating state of the target MEMS sensor and obtain internal characteristic parameters that reflect the current state of the sensor's sensitive unit; By combining the humidity drift prediction value with the internal feature parameters, the sensor drift impact value under the current environment is constructed; An initial humidity drift baseline compensation amount is generated for the target MEMS sensor based on the sensor drift influence amount; By introducing a sensor output feedback sequence, the initial humidity drift reference compensation amount is dynamically iteratively optimized to generate an optimized drift compensation control sequence. The optimized drift compensation control sequence is applied to the original output of the target MEMS sensor to correct the humidity drift.

2. The humidity drift correction method for a MEMS sensor as described in claim 1, characterized in that, The process of tracking and identifying external humidity sources acting on the target MEMS sensor and extracting the humidity characteristics of the external humidity sources includes: Scan the deployment space of the target MEMS sensor to locate at least one major physical source of humidity release; Within a preset humidity sampling period, the surface humidity data of the physical humidity release source and the background humidity data of the ambient space are recorded simultaneously. By performing differential analysis and pattern recognition on the surface humidity data and background humidity data, humidity effect characteristics that characterize the intensity change and diffusion law of the external humidity source are extracted.

3. The humidity drift correction method for a MEMS sensor as described in claim 2, characterized in that, The step of establishing a time-varying humidity characteristic model for the target MEMS sensor based on the humidity effect characteristics, and using the time-varying humidity characteristic model to calculate the humidity drift prediction value in real time includes: The intensity variation law in the humidity effect characteristics is mapped to a set of time-varying weight coefficients, and the diffusion law is mapped to a spatial influence function; By combining the time-varying weighting coefficients and the spatial influence function, a time-varying humidity feature model is constructed to describe the dynamic process of humidity propagating from the source to the sensor location. The real-time acquired environmental parameters are input into the time-varying humidity feature model to calculate the predicted humidity drift value that the target MEMS sensor is expected to experience in the next calibration cycle.

4. The humidity drift correction method for a MEMS sensor as described in claim 1, characterized in that, The monitoring of the internal operating state of the target MEMS sensor, and the acquisition of internal characteristic parameters reflecting the current state of the sensor's sensing unit, include: Acquire the intrinsic electrical response signal of the target MEMS sensor sensing unit under conditions without external excitation; The intrinsic electrical response signal is analyzed in both the frequency domain and amplitude domain to separate the resonant characteristic parameters related to the mechanical properties of the sensitive unit and the impedance characteristic parameters related to the material properties. The resonance characteristic parameter and the impedance characteristic parameter together constitute the internal characteristic parameter reflecting the current state of the sensor's sensitive unit.

5. The humidity drift correction method for a MEMS sensor as described in claim 1, characterized in that, The process of fusing the humidity drift prediction value with the internal feature parameters to construct the sensor drift impact in the current environment includes: Establish a coupling relationship mapping table between humidity stress and sensor status, and use the coupling relationship mapping table to convert the humidity drift prediction value into the corresponding theoretical state offset; The theoretical state offset is correlated with the actual state change calculated from the internal feature parameters; Based on the results of the correlation matching, a comprehensive weighting factor is calculated, and the theoretical state offset is weighted and fused using the weighting factor to generate the sensor drift impact in the current environment.

6. The humidity drift correction method for a MEMS sensor as described in claim 1, characterized in that, The step of generating the initial humidity drift reference compensation amount for the target MEMS sensor based on the sensor drift influence amount includes: The preset lookup table for drift amount to compensation amount is queried, and a preliminary compensation amount reference value is obtained based on the amplitude of the sensor drift influence amount. Based on the changing trend of the sensor drift influence, a trend adjustment factor is applied to the preliminary compensation reference value to obtain the adjusted preliminary compensation amount. Based on the historical calibration records of the target MEMS sensor, the adjusted preliminary compensation amount is empirically calibrated to generate the final initial humidity drift reference compensation amount.

7. The humidity drift correction method for a MEMS sensor as described in claim 1, characterized in that, The process of introducing a sensor output feedback sequence to dynamically iteratively optimize the initial humidity drift reference compensation amount and generate an optimized drift compensation control sequence includes: The actual output sequence of the target MEMS sensor after the initial humidity drift reference compensation amount is continuously acquired and used as the sensor output feedback sequence. Calculate the statistical stability measure of the sensor output feedback sequence and compare it with a preset stability threshold to generate a stability deviation signal; The stability deviation signal is input into a gradient optimizer, which calculates the incremental adjustment amount of the initial humidity drift reference compensation. The initial humidity drift baseline compensation amount is iteratively corrected multiple times based on the incremental adjustment amount until the statistical stability measure meets the requirements. The compensation amounts generated by each iterative correction are arranged in chronological order to generate an optimized drift compensation control sequence. The step of calculating the statistical stability measure of the sensor output feedback sequence and comparing it with a preset stability threshold to generate a stability deviation signal includes: Within the sliding time window, calculate the variance and Allan variance of the sensor output feedback sequence; The variance of the sensor output feedback sequence is weighted and summed with the Allan variance to obtain a comprehensive statistical stability measure of the sensor output feedback sequence. The difference between the comprehensive statistical stability measure and the preset stability threshold is used as the stability deviation signal.

8. The humidity drift correction method for a MEMS sensor as described in claim 2, characterized in that, The step of extracting humidity effect characteristics characterizing the intensity change and diffusion law of the external humidity source through differential analysis and pattern recognition of the surface humidity data and background humidity data includes: Calculate the difference between the surface humidity data and the background humidity data at the same timestamp to form a humidity difference sequence; Time-domain analysis was performed on the humidity difference sequence to extract its periodic and trend components, which were used as characteristic components of humidity effect to characterize the intensity change law. By analyzing the gradient distribution of background humidity data at different spatial locations and combining it with environmental fluid parameters, we fit the attenuation model parameters of humidity diffusion, which serve as the characteristic components of humidity effect characterizing the diffusion law.

9. The humidity drift correction method for a MEMS sensor as described in claim 4, characterized in that, The frequency domain and amplitude domain analysis of the intrinsic electrical response signal, separating the resonant characteristic parameters related to the mechanical properties of the sensitive unit and the impedance characteristic parameters related to the material properties, includes: Perform a spectrum transformation on the intrinsic electrical response signal to identify the main resonant frequency and quality factor in the signal spectrum, and record the main resonant frequency and quality factor as resonant characteristic parameters; Under a specific DC bias, the voltage-current relationship curve of the intrinsic electrical response signal is measured, and the equivalent DC impedance and equivalent capacitance are calculated from the voltage-current relationship curve. The equivalent DC impedance and equivalent capacitance are recorded as impedance characteristic parameters.

10. A humidity drift correction system for a MEMS sensor, comprising a memory, a processor, and a computer program stored in the memory and running on the processor, characterized in that, When the processor executes the computer program, it implements the steps of the humidity drift correction method for the MEMS sensor as described in any one of claims 1 to 9.

Images