A method and system for calibrating a humidity sensor

Abstract

The application discloses a humidity sensing calibration method and system, and belongs to the technical field of humidity measurement calibration. The method is used for a micro-mechanical sensor in a meteorological automatic monitoring instrument, and comprises the following steps: acquiring an initial response sequence in a current acquisition period, and synchronously acquiring an environmental temperature fluctuation sequence and a reference capacitance value sequence; performing amplitude normalization on the registration response sequence by using the reference capacitance value sequence to obtain a normalized response sequence; constructing a response difference feature vector according to the normalized response sequence and the initial response sequence, and performing modal decomposition to obtain a plurality of intrinsic modal components; calculating an energy proportion of each intrinsic modal component, screening out a target modal component and reconstructing to obtain a reconstructed response feature; and combining a peak value of the initial response sequence to generate a calibration offset factor, compensating for a current output value of the sensor, and obtaining a calibrated humidity value. The system comprises a memory and a processor, and is used for executing the method.

Description

Technical Field

[0001] This invention relates to the field of humidity measurement and calibration technology, specifically to a humidity sensing calibration method and system. Background Technology

[0002] In practical applications, the output signal of micromechanical humidity sensors in automatic weather monitoring instruments is highly susceptible to the coupling effects of ambient temperature fluctuations and their own capacitance reference drift. Temperature changes cause a hysteresis in the humidity response of the sensor's sensitive element through heat transfer, resulting in a dynamic deviation between the measured humidity value and the actual ambient humidity. Current technologies often employ temperature compensation methods based on fixed coefficients to address this issue. This involves pre-determining a temperature-humidity correction curve and then calculating the compensation amount by substituting the ambient temperature into a formula. However, this approach ignores the temporal asynchrony between temperature changes and humidity response. During rapid temperature fluctuations, the compensation timing and response timing become misaligned, introducing additional errors. Some solutions incorporate reference capacitance detection to correct sensor drift, but these typically only use the reference capacitance as a static reference for ratio calculations, failing to consider the dynamic mapping relationship between capacitance fluctuation amplitude and humidity response amplitude. When the sensor is contaminated or ages, causing nonlinear changes in sensitivity, static normalization cannot accurately track the true shift in the response amplitude. Furthermore, the differential signals output by sensors under the cross-effect of temperature and humidity exhibit complex mode superposition across multiple time scales. Traditional methods relying on single-scale filtering or statistical feature extraction struggle to separate the dominant interference components, easily resulting in the removal of a portion of the useful signal or the retention of residual interference. This leads to fluctuations or even a decline in accuracy in the calibrated humidity values. The fundamental reason for these shortcomings lies in the lack of methods capable of simultaneously addressing time-domain registration, dynamic amplitude normalization, and accurate extraction and reconstruction of multi-scale interference modes. Summary of the Invention

[0003] A humidity sensor calibration method and system are provided to overcome the problem of insufficient calibration accuracy of micromechanical humidity sensors in meteorological monitoring caused by temperature and humidity time-domain mismatch, capacitance reference drift and multimodal interference aliasing.

[0004] To achieve the above objectives, the present invention provides the following technical solution: The present invention provides a humidity sensing calibration method and system for a micromechanical sensor in an automatic weather monitoring instrument. This method fuses synchronously acquired information from temperature fluctuations and reference capacitance, performs time-domain registration and amplitude normalization on the sensor response sequence, constructs a response difference feature vector, and performs multimodal decomposition. From this vector, the energy-dominant target mode component is extracted and reconstructed. A calibration offset factor is generated using the reconstructed features and the initial sequence peak value, achieving dynamic compensation for the sensor output value. This effectively suppresses measurement errors caused by temperature and humidity cross-sensitivity and nonlinear drift, improving humidity detection accuracy and long-term stability.

[0005] As a technical solution of the present invention, an initial response sequence output by a micromechanical sensor within the current acquisition cycle is obtained, along with an ambient temperature fluctuation sequence and a reference capacitance value sequence acquired synchronously with the initial response sequence. Using the peak and valley temperature times in the ambient temperature fluctuation sequence, and comparing them with the peak and valley response times of the initial response sequence, a first time deviation and a second time deviation are calculated. This determines the time-domain registration offset, and the initial response sequence is then shifted as a whole to align the peak response times with the peak temperature times, thus obtaining a registered response sequence. This time-domain registration process eliminates the hysteresis effect of temperature fluctuations on the humidity response, ensuring the correlation of the sequences on the time axis.

[0006] Preferably, the registration response sequence is amplitude normalized based on the reference capacitance value sequence to obtain a normalized response sequence. Specifically, the maximum and minimum capacitance values ​​in the reference capacitance value sequence are obtained and the capacitance fluctuation range is calculated. Simultaneously, the maximum and minimum response amplitudes in the registration response sequence are obtained and the response amplitude range is calculated. The ratio of the capacitance fluctuation range to the response amplitude range is used as the amplitude normalization coefficient. Each response amplitude in the registration response sequence is multiplied by this coefficient to obtain the normalized response sequence. Amplitude normalization effectively compensates for gain fluctuations in sensor sensitivity caused by environmental changes, ensuring that subsequent analysis is based on a unified energy benchmark.

[0007] Subsequently, each normalized response value in the normalized response sequence is subtracted from the initial response value at the corresponding time in the initial response sequence to obtain a difference value sequence. The mean of this difference value sequence is calculated as the difference mean. The absolute value of the difference between two adjacent difference values ​​is calculated to obtain a difference fluctuation subsequence. The difference mean and the difference fluctuation values ​​in the difference fluctuation subsequence are used to construct a response difference feature vector. This feature vector reflects the response distortion information caused by factors such as temperature and humidity cross-effects and material aging.

[0008] Based on this, mode decomposition is performed on the response difference feature vector to obtain multiple intrinsic mode components. Preferably, empirical mode decomposition, local mean decomposition, and variational mode decomposition are performed simultaneously on the response difference feature vector to obtain three sets of intrinsic mode components. These are then fused through selection to obtain a set of intrinsic mode components that characterize the fluctuation characteristics at different scales. This hybrid mode decomposition method can more comprehensively capture the multi-scale instantaneous features in the response difference and avoid mode aliasing or information omission that may be caused by a single decomposition method.

[0009] The energy percentage of each intrinsic mode component is calculated. The sum of the squares of the amplitudes at all sampling points is used as the component energy value, and the sum of the component energy values ​​of all intrinsic mode components is used as the total energy value. The energy percentage is obtained by dividing the energy value of each component by the total energy value. Intrinsic mode components with an energy percentage greater than a preset energy threshold are selected as candidate mode components, and the one with the largest energy percentage is selected as the target mode component. Through energy-dominant screening, the core mode that best represents the current sensor deviation state is automatically locked, eliminating interference from low-energy noise components.

[0010] The target modal components are reconstructed to obtain reconstructed response characteristics. The locations and amplitudes of extreme points in the target modal components can be obtained, and based on this, the target modal components are divided into rising and falling subsequences. The average amplitudes of the rising and falling subsequences are calculated separately, and the reconstructed response characteristics are constructed by combining the extreme point amplitudes. This reconstruction process preserves the temporal profile and amplitude evolution of the bias principal mode, providing a precise shape and amplitude reference for generating the calibration offset factor.

[0011] When generating the calibration offset factor, the peak response value of the initial response sequence and the peak reconstructed value of the reconstructed response features are obtained, and their ratio is calculated as the first ratio parameter. Simultaneously, the interval lengths of the extreme point distribution in the reconstructed response features and the extreme point distribution in the initial response sequence are obtained, and the difference between their interval lengths is calculated. The calibration offset factor is generated based on the first ratio parameter and the interval length difference. This offset factor integrates the effects of amplitude perturbation and time scale distortion, and can comprehensively reflect the current measurement deviation of the sensor.

[0012] Finally, the current output value of the micromechanical sensor is compensated based on the calibration offset factor to obtain the calibrated humidity value. The current output value and nominal sensitivity value of the sensor are obtained. The calibration offset factor is multiplied by the nominal sensitivity value to obtain the sensitivity adjustment amount. The current output value is added to the sensitivity adjustment amount to obtain the preliminary compensation value. Then, the preliminary compensation value is limited by the humidity reference range boundary to obtain the final calibrated humidity value. Therefore, this method can adaptively adjust the calibration amount in complex environments while ensuring that the output result does not exceed the effective range of the sensor, significantly improving the accuracy, anti-interference ability, and dynamic adaptability of humidity measurement.

[0013] This invention also provides a humidity sensing calibration system, including a memory, a processor, and a computer program stored in the memory and running on the processor. When executing the computer program, the processor implements the steps of the aforementioned humidity sensing calibration method. This system can run the aforementioned calibration algorithm online in real time to continuously calibrate the micromechanical sensors in automatic weather monitoring instruments, ensuring the quality of humidity data under long-term unattended conditions.

[0014] The technical effects and advantages provided by the present invention in the above technical solution are as follows:

[0015] When performing time-domain registration on the initial response sequence, the peak and trough temperature times in the ambient temperature fluctuation sequence are obtained, along with the peak and trough response times in the initial response sequence. The time deviations between peak-to-peak values ​​and between trough values ​​are calculated respectively, and the time-domain registration offset is determined accordingly. The entire initial response sequence is then shifted to align the peak and trough times of the registered response sequence with those of the temperature fluctuation sequence. This processing method directly compensates for the temperature and humidity response time difference caused by heat conduction delay, eliminates the inherent time lag variable in the temperature compensation path, and ensures that the temperature and humidity response information in subsequent processing strictly correspond in physical time sequence. This avoids humidity measurement overshoot or lag residuals during rapid temperature rises and falls, improving the real-time performance of transient humidity measurements. During amplitude normalization, the maximum and minimum capacitance values ​​of the reference capacitance value sequence are obtained, and the capacitance fluctuation span is calculated. Simultaneously, the maximum and minimum response amplitudes of the registration response sequence are obtained, and the response amplitude span is calculated. The ratio of the capacitance fluctuation span to the response amplitude span is taken as the amplitude normalization coefficient, and each amplitude point of the registration response sequence is multiplied by this coefficient. This operation dynamically maps the change in sensor sensitivity through the real-time fluctuation of the capacitance reference. When the sensor undergoes long-term use or becomes contaminated, causing a change in response gain, the normalization coefficient is automatically adjusted. This ensures that the response difference characteristics only reflect the nonlinear disturbance component in the humidity measurement error, eliminating the linear interference of sensitivity drift on the difference values ​​and guaranteeing the purity of the input sequence for subsequent mode decomposition. When constructing the response difference feature vector, the normalized response sequence is subtracted point by point from the initial response sequence to obtain the difference value sequence. Then, the mean difference and the absolute value of the difference between adjacent difference values ​​are extracted to form a joint feature containing both steady-state offset and transient fluctuation amplitude. Performing mode decomposition on this joint feature can simultaneously capture both slowly varying trends and rapidly varying spikes in the error. Then, by calculating the energy proportion of each intrinsic mode component, the one with the largest energy proportion exceeding a preset threshold is selected as the target mode component. This target mode component corresponds to the main interference mode in the error. Extreme points are located for the target mode component, and the rising and falling segments are distinguished. The average amplitude of the rising and falling segments is calculated separately, and the extreme point amplitudes are combined to construct a reconstructed response feature. This energy-based screening and segmented morphological reconstruction method accurately extracts the nonlinear disturbance waveforms that have the most significant impact on the measurement results from multi-scale mixed interference. The generated calibration offset factor can adaptively follow changes in environmental conditions, correcting only the main error components during compensation without constraining the sensor's true dynamic response. Therefore, under wide temperature and high humidity weather conditions, it can significantly suppress zero drift and range drift while preserving the original measurement resolution and response speed. By simultaneously applying empirical mode decomposition, local mean decomposition, and variational mode decomposition to the response difference feature vector, three sets of intrinsic mode components are obtained, which are then screened and fused.Empirical Mode Decomposition (EMD) can adaptively generate mode functions based on the signal's own time scale, exhibiting good matching properties for nonlinear and non-stationary differential signals. Local Mean Decomposition (LMD) uses smoothed local means and envelope estimation to iteratively separate modes, making it insensitive to random noise. Variational Mode Decomposition (VMD) decomposes the signal into bandwidth-limited modes through variational optimization, effectively limiting mode aliasing and spurious component generation. These three decomposition methods perform joint decomposition from different signal representation perspectives. The fused intrinsic mode components retain adaptive multi-scale characteristics while suppressing endpoint effects and noise residues. This makes the target mode components separated from the differential features closer to the real physical interference patterns, resulting in a more complete reconstructed response waveform. The final calibration offset factor exhibits higher consistency and reliability in complex field weather conditions with coexisting interference. Attached Figure Description

[0016] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in this invention. For those skilled in the art, other drawings can be obtained based on these drawings.

[0017] Figure 1 This is a flowchart of the humidity sensor calibration method;

[0018] Figure 2 This is a flowchart of the response difference feature vector construction process;

[0019] Figure 3 It is a calibration comparison curve of the ambient temperature fluctuation sequence and the response sequence;

[0020] Figure 4 This is a schematic diagram of constructing response difference feature vectors;

[0021] Figure 5 This is a curve comparing the humidity output values ​​before and after calibration of the micromechanical sensor. Detailed Implementation

[0022] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, 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, 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.

[0023] See Figure 1This invention provides a humidity sensor calibration method and system, which is used in a micromechanical sensor in an automatic weather monitoring instrument. The method involves acquiring the initial response sequence output by the micromechanical sensor within the current acquisition period, as well as the ambient temperature fluctuation sequence and reference capacitance value sequence acquired synchronously with the initial response sequence. Based on the ambient temperature fluctuation sequence, the initial response sequence is time-domain registered to obtain a registered response sequence. Based on the reference capacitance value sequence, the registered response sequence is amplitude-normalized to obtain a normalized response sequence. Based on the normalized response sequence and the initial response sequence, a response difference feature vector is constructed. The response difference feature vector is modally decomposed to obtain multiple intrinsic mode components. The energy proportion of each intrinsic mode component is calculated, and a target mode component is selected from the multiple intrinsic mode components based on the energy proportion. The target mode component is reconstructed to obtain reconstructed response features. Based on the reconstructed response features and the peak value of the initial response sequence, a calibration offset factor is generated. Based on the calibration offset factor, the current output value of the micromechanical sensor is compensated to obtain a calibrated humidity value.

[0024] Example 1:

[0025] In practical implementation, the process of obtaining the peak and trough temperature moments in the ambient temperature fluctuation sequence is as follows: The temperature value of each sampling point in the ambient temperature fluctuation sequence is scanned point by point. When the temperature value of a sampling point is greater than the temperature values ​​of its preceding and following sampling points, the corresponding moment of that sampling point is recorded as a peak temperature moment; when the temperature value of a sampling point is less than the temperature values ​​of its preceding and following sampling points, the corresponding moment of that sampling point is recorded as a trough temperature moment. Similarly, the process of obtaining the peak and trough response moments in the initial response sequence is as follows: The response value of each sampling point in the initial response sequence is scanned point by point. When the response value of a sampling point is greater than the response values ​​of its preceding and following sampling points, the corresponding moment of that sampling point is recorded as a peak response moment; when the response value of a sampling point is less than the response values ​​of its preceding and following sampling points, the corresponding moment of that sampling point is recorded as a trough response moment.

[0026] It is understandable that, when both the ambient temperature fluctuation sequence and the initial response sequence are discrete time sequences, the above extreme value detection is accomplished by comparing the values ​​of three adjacent points. If consecutive equal extreme values ​​occur, the midpoint of the consecutive equal segment is taken as the corresponding extreme value point.

[0027] After obtaining the peak temperature and trough temperature times, the first time deviation between the peak temperature time and the peak response time is calculated. In one embodiment, the peak temperature time with the largest amplitude is selected from the ambient temperature fluctuation sequence as the reference peak temperature time, and the peak response time with the largest amplitude is selected from the initial response sequence as the reference peak response time. The time difference between the reference peak temperature time and the reference peak response time is calculated to obtain the first time deviation. When there are multiple peak temperature times and multiple peak response times, the average peak temperature time and the average peak response time can be calculated to obtain the average peak response time. The time difference between the average peak temperature time and the average peak response time can then be calculated as the first time deviation.

[0028] The second time deviation between the valley temperature time and the valley response time is calculated in a similar way to the first time deviation calculation: the valley temperature time with the smallest amplitude is selected as the reference valley temperature time, and the valley response time with the smallest amplitude is selected as the reference valley response time. The time difference between the reference valley temperature time and the reference valley response time is calculated to obtain the second time deviation; or the time difference is calculated by averaging all valley temperature times and valley response times.

[0029] The time-domain registration offset is determined based on the first time deviation and the second time deviation. The time-domain registration offset is determined by calculating the arithmetic mean of the first time deviation and the second time deviation, and using this arithmetic mean as the time-domain registration offset. In some embodiments, the time-domain registration offset can also be determined based on a weighted average of the first time deviation and the second time deviation, with the weights set according to the amplitude ratio of the temperature fluctuation peaks and troughs.

[0030] Based on the time-domain registration offset, the initial response sequence is shifted globally to align the peak response time with the peak temperature time, thus obtaining a registered response sequence. During the global shift, if the first time deviation indicates that the peak temperature time is earlier than the peak response time, the absolute value of the time-domain registration offset is uniformly subtracted from the times of all sampling points in the initial response sequence, causing the initial response sequence to shift negatively along the time axis. If the first time deviation indicates that the peak temperature time is later than the peak response time, the absolute value of the time-domain registration offset is uniformly added to the times of all sampling points in the initial response sequence, causing the initial response sequence to shift positively along the time axis. After the shift, the sequence is resampled to maintain equal time intervals consistent with the original sampling rate. Linear interpolation is used to obtain the response values ​​at the corresponding times during resampling, forming the registered response sequence.

[0031] Obtain the maximum and minimum capacitance values ​​in the reference capacitance value sequence. By iterating through all capacitance values ​​in the reference capacitance value sequence, the maximum value is retained as the maximum capacitance value, and the minimum value is retained as the minimum capacitance value. Obtain the maximum and minimum response amplitudes in the registration response sequence. By iterating through all response amplitudes in the registration response sequence, the maximum value is retained as the maximum response amplitude, and the minimum value is retained as the minimum response amplitude.

[0032] Calculate the capacitance fluctuation range based on the maximum and minimum capacitance values. The capacitance fluctuation range is the difference between the maximum and minimum capacitance values. Similarly, calculate the response amplitude range based on the maximum and minimum response amplitudes. The response amplitude range is the difference between the maximum and minimum response amplitudes.

[0033] The ratio of the capacitance fluctuation range to the response amplitude range is calculated and used as the amplitude normalization coefficient. The amplitude normalization coefficient is calculated using the following formula:

[0034]

[0035] in, Indicates the maximum capacitance value. Indicates the minimum capacitance value. This represents the capacitance fluctuation range value. Indicates the maximum response amplitude. Indicates the minimum response amplitude. For the response amplitude span value; This represents the amplitude normalization coefficient, which takes the value of a real number greater than zero. When the response amplitude span is zero, let... The value is set to the preset value of 1.0.

[0036] Multiply each response amplitude in the registration response sequence by the amplitude normalization coefficient to obtain the normalized response sequence. For the i-th response in the registration response sequence... Response amplitude at each sampling point Calculate the normalized response value After traversing all sampling points, a normalized response sequence is formed.

[0037] See Figure 3 In the figure, the horizontal axis represents time in seconds, the left vertical axis represents temperature in degrees Celsius, and the right vertical axis represents the response value or capacitance value, with units not specified. The curves in the figure include the ambient temperature fluctuation sequence (red solid line), the initial response sequence (blue dashed line), the reference capacitance value sequence (green dotted line), the normalized response sequence (cyan solid line), and the registration response sequence (purple dotted line).

[0038] The ambient temperature fluctuation sequence exhibits obvious periodic fluctuations, with temperatures ranging from approximately 20°C to 32°C, and multiple peaks and troughs are clearly visible. The initial response sequence changes synchronously with the ambient temperature fluctuation sequence, but the overall response amplitude is lower than the peak ambient temperature, and there is a certain time lag and amplitude difference. The reference capacitance value sequence shows a similar trend to the initial response sequence, with amplitudes ranging from 280 to 400, reflecting changes in the capacitance sensor output.

[0039] The normalized response sequence was generated after time-domain registration and translation compensation of the initial response sequence. The curve as a whole moves closer to the ambient temperature fluctuation sequence, and the peak and trough times are basically aligned, reflecting the effect of eliminating time deviation through time-domain registration in Example 1. The amplitude of the normalized response sequence is significantly improved and the fluctuation trend tends to be smoother, indicating that the response value adjusted by the amplitude normalization coefficient is more stable and accurate.

[0040] Example 2:

[0041] In specific implementation, please refer to Figure 2 The difference value sequence is obtained by subtracting the corresponding initial response value from the initial response sequence for each normalized response value in the normalized response sequence. The normalized response sequence is denoted as... ,in, This represents the total number of sampling points in the normalized response sequence and the initial response sequence. Indicates the number of normalized response sequences. The normalized response value corresponding to each sampling point The sampling point number, The value is to An integer. The initial response sequence is denoted as... ,in, Indicates the first element in the initial response sequence. The initial response value corresponding to each sampling point. The corresponding time refers to the sequence number of the same sampling point. At the corresponding times, the sampling times of the two sequences are aligned one-to-one. The difference value sequence is denoted as... ,in, Indicates the th value in the difference value sequence The difference value corresponding to each sampling point, the difference value The calculation method is as follows:

[0042]

[0043] It is understandable that when calculating the difference value... Previously, the normalized response sequence and the initial response sequence had been aligned on the time axis, and the same index in the two sequences was... Corresponding to the same physical sampling time.

[0044] Calculate the mean of the series of differences, and use it as the difference mean. The difference mean is denoted as... Mean difference The calculation uses the following formula:

[0045]

[0046] in, This represents the total number of sampling points in the difference value sequence. Indicates the th value in the difference value sequence The difference value corresponding to each sampling point Indicates to from arrive All the differences are summed together.

[0047] Calculate the absolute value of the difference between two adjacent difference values ​​in the difference value sequence to obtain the difference fluctuation subsequence. In specific implementation, the difference value sequence... Starting with the first difference value, subtract the previous difference value from each subsequent difference value and take the absolute value to obtain... Each differential fluctuation value. The differential fluctuation subsequence is denoted as... ,in, Indicating the first subsequence of differential fluctuation Individual fluctuation values, This refers to the sequence number of the difference fluctuation value. The value is to Integer values. Difference fluctuation values. The calculation method is as follows:

[0048]

[0049] in, Indicates the th value in the difference value sequence One difference value, Indicates the th value in the difference value sequence One difference value, This indicates the operation of taking the absolute value.

[0050] Based on the mean difference and the difference fluctuation subsequence, a response difference feature vector is constructed. The response difference feature vector is a one-dimensional vector, where the elements are, in order, the mean difference. And each differential fluctuation value in the differential fluctuation subsequence. The response differential eigenvector is denoted as... Response difference feature vector The constituent forms are:

[0051]

[0052] in, The mean of the differences. For the differential fluctuation subsequence Individual fluctuation values, For the differential fluctuation subsequence Individual fluctuation values, For the differential fluctuation subsequence Individual fluctuation values. Response difference feature vector. The total dimension is The first dimension element is the mean difference. ,the remaining Each dimension element corresponds to a different fluctuation value in the differential fluctuation subsequence.

[0053] In some embodiments, when the total number of sampling points in the difference value sequence When the response difference feature vector dimension is too high due to a large difference, the difference fluctuation subsequence is downsampled, and the difference fluctuation values ​​at equally spaced positions are included in the response difference feature vector. The downsampling interval is determined based on the preset target dimension of the response difference feature vector.

[0054] See Figure 4 In the graph, the horizontal axis represents time in seconds (s), the left side of the vertical axis represents the numerical range of the differential value sequence, from -30 to 50, and the right side represents the numerical range of the differential fluctuation subsequence, from 0 to 14. The solid blue curve represents the differential value sequence, and the dashed red curve represents the differential fluctuation subsequence. The differential value sequence curve exhibits significant fluctuations over the entire time range, with a large range of values, mainly concentrated between -27 and 50, showing the dynamic trend of the differential value sequence over time. The amplitude of the differential fluctuation subsequence curve is generally smaller than that of the differential value sequence, and its fluctuation frequency is higher, reflecting the variation characteristics between adjacent differential values.

[0055] Example 3:

[0056] In practice, empirical mode decomposition (EMD) is performed on the response difference feature vector to obtain the first set of intrinsic mode components. The input to EMD is the response difference feature vector, denoted as... , The dimension is , It is a positive integer. The process of empirical mode decomposition includes the following: dividing the response difference feature vector... As the current signal to be decomposed, find all the maximum and minimum points in the current signal to be decomposed. Use cubic spline interpolation to fit all the maximum points to obtain the upper envelope. Use cubic spline interpolation to fit all the minimum points to obtain the lower envelope. Calculate the mean of the upper and lower envelopes at each sampling point to obtain the mean envelope. Subtract the mean envelope from the current signal to be decomposed to obtain candidate intrinsic mode components. Determine whether the candidate intrinsic mode components meet the preset intrinsic mode function conditions. The preset intrinsic mode function conditions are: within the entire signal length, the number of extreme points in the candidate intrinsic mode components is equal to or differs from the number of zero-crossing points by at most one, and at any sampling point, the mean of the upper and lower envelopes is zero. When a candidate intrinsic mode component (IMC) satisfies the IMC condition, it is output as an IMC. The remaining portion of the current signal to be decomposed (subtracting the output IMC) is then used as the new current signal to be decomposed, and this process is repeated. When a candidate IMC does not satisfy the IMC condition, it is used as the new current signal to be decomposed, and this process is repeated until the remaining signal is monotonic and no IMC can be extracted. The set of IMCs obtained after empirical mode decomposition is denoted as the first set of IMCs, and the number of IMCs in the first set is denoted as _____. .

[0057] Local mean decomposition is performed on the response difference feature vector to obtain the second set of intrinsic mode components. The input to the local mean decomposition is the response difference feature vector. The process of local mean decomposition includes the following: dividing the response difference feature vector... As the current signal to be decomposed, all local maxima and local minima are identified. The local mean sequences between adjacent local maxima and local minima are calculated, and smoothed using a moving average method to obtain the local mean function. Simultaneously, the local amplitude sequences between adjacent local maxima and local minima are calculated, and smoothed using a moving average method to obtain the local envelope function. The local mean function is subtracted from the current signal to be decomposed to obtain the zero-mean signal. The zero-mean signal is then divided by the local envelope function to obtain the demodulated signal. This process of extracting the local mean and local envelope function is repeated on the demodulated signal until the local envelope function of the demodulated signal approaches 1. At this point, the zero-mean signal is a product function component, and the cumulative product of the corresponding local envelope functions is used as the amplitude envelope of this product function component. The ratio of the product function component to the amplitude envelope is output as an intrinsic mode component. Subtract the already output intrinsic mode components from the current signal to be decomposed, and repeat the above process with the remaining part as the new current signal to be decomposed until the remaining signal is monotonic. The set of intrinsic mode components obtained after local mean decomposition is denoted as the second set of intrinsic mode components, and the number of intrinsic mode components contained in the second set is denoted as... .

[0058] Variational mode decomposition (VMD) is performed on the response difference eigenvectors to obtain the third set of intrinsic mode components. The input to the VMD is the response difference eigenvectors. The variational mode decomposition process includes the following: setting the number of decomposition modes. and secondary penalty factor Decomposition of mode numbers Set to a positive integer, with a value greater than or equal to 2 and less than or equal to the response difference feature vector. Dimensions Half of the integer, quadratic penalty factor Set to a positive real number, with a value of 1000. Initialize. Each intrinsic mode component , corresponding center frequency And the Lagrange multipliers are updated iteratively using the alternating direction multiplier method, updating each eigenmode component in each iteration. , corresponding center frequency And the Lagrange multipliers, until a preset iterative convergence condition is met. The preset iterative convergence condition is that the sum of the changes in the amplitudes of all intrinsic mode components in two adjacent iterations is less than a preset convergence threshold. The preset convergence threshold is set to... The set of intrinsic mode components obtained after variational mode decomposition is denoted as the third set of intrinsic mode components. The number of intrinsic mode components contained in the third set of intrinsic mode components is... .

[0059] The first, second, and third groups of intrinsic mode components (IMCs) are filtered and fused to obtain multiple IMCs. The filtering and fusion process includes: calculating the center frequency of each IMC in the first group; calculating the center frequency of each IMC in the second group; and calculating the center frequency of each IMC in the third group. The center frequency is calculated by performing a Hilbert transform on the IMC to obtain an analytic signal; the average instantaneous frequency of the analytic signal over the entire sequence length is the center frequency of the IMC. All IMCs in the three groups are arranged in ascending order of their center frequencies to form a set of components to be fused.

[0060] The intrinsic mode components in the set of components to be fused are grouped by setting a center frequency tolerance threshold. Center frequency tolerance threshold The value is set to 0.1 Hz, ensuring the difference between center frequencies is less than the center frequency tolerance threshold. All intrinsic mode components are grouped into the same group, and the difference between center frequencies is greater than or equal to the center frequency tolerance threshold. The intrinsic mode components (IMCs) are grouped into different groups. Within the same group, if a group contains only one IMC, that IMC is directly retained as a fused IMC. If a group contains multiple IMCs, the average amplitude of each IMC at the same sampling point within the group is calculated, and the sequence of average amplitudes is used as the fused IMC. After all groups are fused, the resulting set of IMCs is the multiple IMCs obtained after fusion. The number of multiple IMCs obtained after fusion is denoted as _____. , Less than or equal to .

[0061] In some embodiments, when grouping the intrinsic mode components in the set of components to be fused, the center frequency tolerance threshold is... The setting method is as follows: calculate the standard deviation of the center frequencies of all intrinsic mode components in the component set to be fused, and take one-tenth of the standard deviation of the center frequencies as the center frequency tolerance threshold. The value of .

[0062] It is understandable that the three modal decomposition methods described above extract the oscillation modes of response difference feature vectors from different perspectives. The screening and fusion process merges the intrinsic modal components of the same physical oscillation mode extracted by different decomposition methods through center frequency matching, eliminates redundant components, and finally obtains a set of multiple intrinsic modal components without redundancy.

[0063] Example 4:

[0064] In practice, the sum of squared amplitudes of all sampling points in each intrinsic mode component is calculated as the component energy value of each intrinsic mode component. Let the multiple intrinsic mode components obtained after screening and fusion be denoted as... ,in, This represents the total number of multiple intrinsic mode components. It is a positive integer; This represents the index of the sampling point within the intrinsic modal component. The value is arrive integers, This represents the total number of sampling points contained in each intrinsic mode component; Indicates the first Each intrinsic modal component at the sampling point index The amplitude at that point, The value is arrive Integers.

[0065] No. The component energy values ​​of each intrinsic mode component are denoted as Component energy value The calculation uses the following formula:

[0066]

[0067] in, Indicates the index of the sampling point from arrive Perform cumulative summation. Indicates the first Each intrinsic modal component at the sampling point index The square of the amplitude at that point.

[0068] The sum of the component energies of all intrinsic modal components is calculated as the total energy value, denoted as . , ,in, Indicates the sequence number of the intrinsic modal components. from arrive The component energy values ​​are accumulated.

[0069] Divide the energy value of each intrinsic mode component by the total energy value to obtain the energy percentage of each intrinsic mode component. The energy percentage of each intrinsic mode component is denoted as . , .

[0070] The intrinsic mode components whose energy proportion exceeds a preset energy threshold are selected as candidate mode components. The preset energy threshold is a pre-defined critical ratio value used to distinguish between effective signal components and noise components. The preset energy threshold is set to... The setting is based on the following: After mode decomposition, the intrinsic mode components caused by noise in the response difference feature vector are usually widely distributed in terms of energy. The energy of a single noise component generally accounts for a low proportion of the total energy. Under typical signal-to-noise ratio conditions, the energy proportion of the noise component generally does not exceed [a certain percentage]. Therefore, the preset energy threshold is set to This results in a higher energy content than The intrinsic modal components are considered as effective components containing meaningful response differences, with an energy percentage less than or equal to The intrinsic mode components are treated as noise components and are removed.

[0071] The candidate modal component with the largest energy proportion is selected as the target modal component. If multiple candidate modal components have the same maximum energy proportion, the candidate modal component with the lowest center frequency is selected as the target modal component. The target modal component is a single eigenmode component used for subsequent reconstruction, denoted as . ,in, Indicates the sampling point index. .

[0072] Obtain the location and magnitude of extreme points in the target modal components. For the target modal components... The amplitude sequence is scanned point by point, when and At that time, sampling point index The location of a maximum point is recorded as the amplitude of the target modal component at that sampling point index. It is recorded as the extreme point amplitude corresponding to the maximum point; when and At that time, sampling point index The location of a local minimum is recorded, and the amplitude of the target modal component at that sampling point index is recorded as the amplitude of the local minimum. At the sequence boundary, if... and Greater than Then the sampling point index will be... Record the location of the maximum point; if and Less than Then the sampling point index will be... Record the location of the minimum point; in Similar boundary processing is applied at each location. All extreme point locations are arranged in ascending order of sampling point index, forming an extreme point location sequence, and the extreme point amplitude corresponding to each extreme point location forms an extreme point amplitude sequence.

[0073] Based on the location of the extreme points, the target modal components are divided into rising and falling subsequences. The division rule is: starting from the index of the first sampling point... The segment between the first and second extreme points is defined as follows: if the first extreme point is a maximum, the segment is defined as an ascending segment; if the first extreme point is a minimum, the segment is defined as a descending segment. Similarly, the segment between the first and second extreme points is defined as follows: if the first extreme point is a maximum and the second extreme point is a minimum, the segment is defined as a descending segment; if the first extreme point is a minimum and the second extreme point is a maximum, the segment is defined as an ascending segment. This process continues until all adjacent extreme point positions are classified. The index from the last extreme point to the last sampling point is also defined. The intervals between points are defined as follows: if the last extreme point is a maximum, it is defined as a falling segment; if the last extreme point is a minimum, it is defined as a rising segment. The amplitudes of all sampling points of the target modal component within all segments marked as rising segments are assigned to the rising segment subsequence; the amplitudes of all sampling points of the target modal component within all segments marked as falling segments are assigned to the falling segment subsequence.

[0074] Calculate the average amplitude of all sampled points in the rising segment subsequence, and use this as the rising segment average amplitude. The rising segment average amplitude is denoted as... , The calculation method is as follows: count the total number of sampling points contained in the rising segment subsequence, denoted as Calculate the sum of the amplitudes of all sampled points in the rising segment subsequence, and divide the sum by . The average amplitude of the rising segment is obtained. .

[0075] Calculate the average amplitude of all sampled points in the falling segment subsequence, and use this as the average amplitude of the falling segment. The average amplitude of the falling segment is denoted as... , The calculation method is as follows: count the total number of sampling points contained in the descent segment subsequence, denoted as Calculate the sum of the amplitudes of all sampled points in the falling segment subsequence, and divide the sum by . The average amplitude of the falling segment is obtained. .

[0076] Based on the average amplitude of the rising segment, the average amplitude of the falling segment, and the amplitude of the extreme points, reconstructed response features are constructed. These reconstructed response features are generated using piecewise linear interpolation. The construction process is as follows: the target modal component sequence is divided into multiple segments according to the order of the extreme point positions. The first segment is indexed from the sampling point... The process begins at the first extreme point and ends at the first extreme point; the middle section begins at one extreme point and ends at the next adjacent extreme point; the last section begins at the last extreme point and ends at the sampling point index. Finish.

[0077] For each segment, the segment reconstruction slope is determined based on the segment type. When the segment type is an ascending segment, the segment reconstruction slope is... Set as ,in, This indicates the number of sampling points spanning this section. It equals the index of the end sampling point of the segment minus the index of the start sampling point. When the segment type is a descending segment, the segment reconstruction slope is... Set as .

[0078] After determining the reconstruction slope of a segment, the amplitude at the starting point of that segment is used as the initial amplitude, and the reconstruction response features are generated point by point along the sampling point index sequence. The starting amplitude of the segment is determined as follows: the starting amplitude of the first segment is taken from the target modal component at the sampling point index. The amplitude at the starting point; for subsequent segments, the amplitude at the starting point of the segment is taken as the reconstructed amplitude generated at the end point of the previous segment, to ensure the continuity of the reconstructed response characteristics throughout the sequence. At the... Within each segment, the reconstructed response characteristics are in the first... Values ​​at each sampling point It is generated recursively by the following formula: ,in, Indicates the first Within each section The index of each sampling point from Starting from, when hour It equals the amplitude at the starting point of the segment. Calculate point by point until the end point of the segment to complete the reconstruction of the segment.

[0079] After all segments are generated, a complete reconstructed response feature sequence is obtained. The number of sampling points in the reconstructed response feature sequence is the same as that of the target modal components. Each sample value in the reconstructed response feature sequence is the value of the reconstructed response feature at the corresponding sample point index.

[0080] It is understandable that in the above piecewise linear interpolation process, the rising segment uses the average amplitude of the rising segment. The upward slope of the reconstructed curve was determined, and the average amplitude of the downward segment was used for the downward segment. By determining the descending slope of the reconstruction curve, the reconstruction response feature sequence highlights the overall changing trend of the rising and falling segments while maintaining the extreme point positions and amplitude boundary relationships of the target modal components.

[0081] Example 5:

[0082] In practice, the peak response value in the initial response sequence is obtained as follows: Iterate through the response values ​​of all sampled points in the initial response sequence, select the sampled point with the largest response value, and use the response value corresponding to this sampled point as the peak response value of the initial response sequence, denoted as . The method for obtaining the peak reconstruction value in the reconstruction response feature is as follows: traverse the reconstruction values ​​of all sampling points in the reconstruction response feature sequence, select the sampling point with the largest reconstruction value, and take the reconstruction value corresponding to this sampling point as the peak reconstruction value in the reconstruction response feature, denoted as . .

[0083] The ratio of the peak reconstructed value to the peak response value is calculated and used as the first ratio parameter, denoted as . First ratio parameter The calculation process is as follows: reconstruct the peak value Divide by peak response value ,Right now When the peak response value When the amplitude is zero, the first ratio parameter Set to 1.0.

[0084] The process of obtaining the interval length of the extreme point distribution in the reconstructed response feature is as follows: extract all extreme points in the reconstructed response feature sequence, scan the reconstructed value of each sampling point in the reconstructed response feature sequence, and when the reconstructed value of a sampling point is greater than the reconstructed values ​​of its two adjacent sampling points, record the sampling time of the sampling point as a maximum point time; when the reconstructed value of a sampling point is less than the reconstructed values ​​of its two adjacent sampling points, record the sampling time of the sampling point as a minimum point time; at the beginning and end of the sequence, if the reconstructed value of the first sampling point is greater than the reconstructed value of the second sampling point, record the sampling time of the first sampling point as a maximum point time; if the reconstructed value of the first sampling point is less than the reconstructed value of the second sampling point, record the sampling time of the first sampling point as a minimum point time; the same processing is performed at the end sampling point. Arrange all recorded maximum and minimum times in chronological order. Select the earliest extreme time as the first extreme time and the latest extreme time as the last extreme time. Subtract the first extreme time from the last extreme time to obtain the time span as the interval length of the extreme point distribution in the reconstructed response features, denoted as . .

[0085] The process of obtaining the interval length of the extreme point distribution in the initial response sequence is as follows: Using the same extreme point extraction method as the reconstructed response feature sequence, record all maximum and minimum timestamps in the initial response sequence. Select the earliest extreme point as the first extreme point of the initial response sequence, and the latest extreme point as the last extreme point. Subtract the first extreme point from the last extreme point to obtain the time span as the interval length of the extreme point distribution in the initial response sequence, denoted as . .

[0086] The difference between the interval length of the extreme point distribution in the reconstructed response features and the interval length of the extreme point distribution in the initial response sequence is calculated and denoted as the interval length difference. , .when When the value is negative, retain the negative sign.

[0087] Based on the first ratio parameter and the difference in interval length, a calibration offset factor is generated, denoted as . Calibration offset factor The calculation uses the following formula:

[0088]

[0089] in, The first preset weighting coefficient, This is the second preset weighting coefficient; The first ratio parameter, This represents the difference in interval lengths. The first preset weighting coefficient... Second preset weight coefficient The values ​​of are all real numbers between 0 and 1, and satisfy . First preset weighting coefficient Set to 0.65, second preset weighting coefficient The value is set to 0.35, based on the following: In historical calibration data, a contribution analysis of the peak ratio feature and the interval length difference feature showed that the peak ratio feature explained 65% of the variance for humidity calibration offset, while the interval length difference feature explained 35%. Therefore, the first preset weighting coefficient is set accordingly. The value is 0.65, the second preset weighting coefficient. The value is set to 0.35 to match the relative importance of the two features in the compensation.

[0090] Obtain the current output value of the micromechanical sensor and its nominal sensitivity value. The current output value of the micromechanical sensor is the uncompensated humidity measurement value directly output by the sensor at the current sampling moment, denoted as... The nominal sensitivity value of a micromechanical sensor is the output change corresponding to a unit change in humidity, determined during the sensor's factory calibration. It is denoted as... , It is a real number greater than zero, read from the datasheet of the micromechanical sensor or the calibration parameters stored internally.

[0091] Calculate the product of the calibration offset factor and the nominal sensitivity value, and use this product as the sensitivity adjustment amount, denoted as . , .

[0092] Add the sensitivity adjustment amount to the current output value to obtain the preliminary compensation value, which is denoted as . , .

[0093] Obtain the comparison results between the preliminary compensation value and the humidity reference range boundary, which includes the upper limit of the humidity reference range. and the lower limit of the humidity reference range Upper limit of humidity reference range and the lower limit of the humidity reference range These represent the maximum and minimum humidity values ​​that the micromechanical sensor can effectively measure, respectively. The initial compensation value will be... With upper limit of humidity reference range and the lower limit of the humidity reference range Compare them.

[0094] The specific method for limiting the preliminary compensation value based on the comparison results is as follows: when the preliminary compensation value... Greater than the upper limit of the humidity reference range At that time, set the upper limit of the humidity reference range. As the calibrated humidity value; when the initial compensation value Less than the lower limit of the humidity reference range At that time, set the lower limit of the humidity reference range. As the calibrated humidity value; when the initial compensation value Greater than or equal to the lower limit of the humidity reference range And less than or equal to the upper limit of the humidity reference range. At that time, the initial compensation value will be... Used directly as the calibrated humidity value.

[0095] It is understandable that the limiting process ensures that the humidity value after calibration is always within the effective measurement range of the micromechanical sensor, avoiding invalid measurement results caused by excessive compensation exceeding the physical range.

[0096] See Figure 5 In the graph, the vertical axis represents the humidity value in percentage relative humidity (%RH), and the horizontal axis represents time in minutes. The legend distinguishes four curves: the blue solid line represents the current output value of the micromechanical sensor, the red dashed line represents the calibrated humidity value, and the green and purple dotted lines represent the upper limit (80%RH) and lower limit (10%RH) of the humidity reference range, respectively.

[0097] Overall, the current output value of the micromechanical sensor exhibits periodic fluctuations over time, with the humidity value ranging from approximately 45%RH to 70%RH. The calibrated humidity value curve closely follows the sensor's current output value curve, showing a consistent overall trend. However, at the peaks and troughs of each fluctuation cycle, the calibrated curve is slightly higher than the sensor's original output, demonstrating the effectiveness of compensating for the sensor output using the method described in Example 5.

[0098] The upper and lower limits of the humidity reference range in the figure are displayed as horizontal dashed lines. After calibration, the humidity values ​​all remain within this range, indicating that the initial compensation value has been limited to prevent the output from exceeding the effective measurement range of the sensor and resulting in invalid measurement results.

[0099] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any changes or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application.

Claims

1. A humidity sensing calibration method, characterized in that, The method is used for micromechanical sensors in automatic weather monitoring instruments, and the method includes: Acquire the initial response sequence output by the micromechanical sensor within the current acquisition cycle, and acquire the ambient temperature fluctuation sequence and reference capacitance value sequence acquired synchronously with the initial response sequence; Based on the ambient temperature fluctuation sequence, the initial response sequence is registered in the time domain to obtain the registered response sequence; Based on the reference capacitance value sequence, the registration response sequence is normalized to obtain a normalized response sequence. Based on the normalized response sequence and the initial response sequence, a response difference feature vector is constructed; Modal decomposition is performed on the response difference feature vector to obtain multiple intrinsic mode components; Calculate the energy percentage of each intrinsic mode component, and select the target mode component from the plurality of intrinsic mode components based on the energy percentage; The target modal components are reconstructed to obtain the reconstructed response features; A calibration offset factor is generated based on the reconstructed response characteristics and the peak values ​​of the initial response sequence; The current output value of the micromechanical sensor is compensated based on the calibration offset factor to obtain the calibrated humidity value.

2. The humidity sensing calibration method according to claim 1, characterized in that, Based on the environmental temperature fluctuation sequence, the initial response sequence is time-domain registered to obtain a registered response sequence, including: Obtain the peak temperature time and valley temperature time in the ambient temperature fluctuation sequence, and obtain the peak response time and valley response time in the initial response sequence; Calculate the first time deviation between the peak temperature time and the peak response time, and calculate the second time deviation between the valley temperature time and the valley response time; The time-domain registration offset is determined based on the first time deviation and the second time deviation; Based on the time-domain registration offset, the initial response sequence is shifted as a whole so that the peak response time of the initial response sequence is aligned with the peak temperature time, thereby obtaining a registered response sequence.

3. The humidity sensing calibration method according to claim 1, characterized in that, Based on the reference capacitance value sequence, the registration response sequence is normalized to obtain a normalized response sequence, including: Obtain the maximum and minimum capacitance values ​​in the reference capacitance value sequence, and obtain the maximum and minimum response amplitudes in the registration response sequence; Calculate the capacitance fluctuation range value based on the maximum capacitance value and the minimum capacitance value; Calculate the response amplitude span value based on the maximum response amplitude and the minimum response amplitude; Calculate the ratio of the capacitance fluctuation span value to the response amplitude span value, and use it as the amplitude normalization coefficient; The normalized response sequence is obtained by multiplying each response amplitude in the registration response sequence by the amplitude normalization coefficient.

4. The humidity sensing calibration method according to claim 1, characterized in that, Based on the normalized response sequence and the initial response sequence, a response difference feature vector is constructed, including: Subtract the initial response value at the corresponding time from each normalized response value in the normalized response sequence to obtain the difference value sequence; Calculate the mean of the difference value sequence, and use it as the difference mean; Calculate the absolute value of the difference between two adjacent difference values ​​in the difference value sequence to obtain the difference fluctuation subsequence; Based on the mean difference and the difference fluctuation subsequence, a response difference feature vector is constructed, which includes the mean difference and each difference fluctuation value in the difference fluctuation subsequence.

5. The humidity sensing calibration method according to claim 1, characterized in that, Modal decomposition is performed on the response difference feature vector to obtain multiple intrinsic mode components, including: Perform empirical mode decomposition on the response difference feature vector to obtain the first set of intrinsic mode components; Perform local mean decomposition on the response difference feature vector to obtain the second set of intrinsic mode components; Perform variational mode decomposition on the response difference feature vector to obtain the third set of intrinsic mode components; The first group of intrinsic modal components, the second group of intrinsic modal components, and the third group of intrinsic modal components are filtered and fused to obtain multiple intrinsic modal components.

6. The humidity sensing calibration method according to claim 1, characterized in that, The step of calculating the energy percentage of each intrinsic mode component and selecting the target mode component from the plurality of intrinsic mode components based on the energy percentage includes: Calculate the sum of squared amplitudes of all sampling points in each intrinsic mode component, and use it as the component energy value of each intrinsic mode component; Calculate the sum of the component energy values ​​of all intrinsic modal components, and use it as the total energy value; Divide the component energy value of each intrinsic mode component by the total energy value to obtain the energy percentage of each intrinsic mode component; The intrinsic mode components corresponding to the energy percentage being greater than the preset energy threshold are taken as candidate mode components; The candidate modal component with the largest energy percentage is selected as the target modal component.

7. The humidity sensing calibration method according to claim 1, characterized in that, The target modal components are reconstructed to obtain reconstructed response features, including: Obtain the location and magnitude of the extreme points in the target modal components; Based on the location of the extreme points, the target modal components are divided into rising segment sequences and falling segment sequences; Calculate the average amplitude of all sampling points in the rising segment subsequence, and use it as the average amplitude of the rising segment; Calculate the average amplitude of all sampling points in the falling segment subsequence, and use it as the average amplitude of the falling segment; The reconstructed response features are constructed based on the average amplitude of the rising segment, the average amplitude of the falling segment, and the amplitude of the extreme point.

8. The humidity sensing calibration method according to claim 1, characterized in that, Based on the reconstructed response characteristics and the peak values ​​of the initial response sequence, a calibration offset factor is generated, including: Obtain the peak response value in the initial response sequence, and obtain the peak reconstruction value in the reconstructed response features; Calculate the ratio of the peak reconstructed value to the peak response value, and use it as the first ratio parameter; Obtain the interval length of the extreme point distribution in the reconstructed response features, and obtain the interval length of the extreme point distribution in the initial response sequence; The difference between the interval length of the extreme point distribution in the reconstructed response features and the interval length of the extreme point distribution in the initial response sequence is calculated as the interval length difference. A calibration offset factor is generated based on the first ratio parameter and the difference in interval length.

9. A humidity sensing calibration method according to claim 1, characterized in that, The current output value of the micromechanical sensor is compensated based on the calibration offset factor to obtain a calibrated humidity value, including: Obtain the current output value of the micromechanical sensor and obtain the nominal sensitivity value of the micromechanical sensor; Calculate the product of the calibration offset factor and the nominal sensitivity value, and use it as the sensitivity adjustment amount; Add the sensitivity adjustment amount to the current output value to obtain a preliminary compensation value; The comparison result between the preliminary compensation value and the humidity reference range boundary is obtained, and the preliminary compensation value is limited according to the comparison result to obtain the calibrated humidity value.

10. A humidity sensing calibration system, 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 sensing calibration method according to any one of claims 1 to 9.

Images