A method for calibrating sensitivity of an outdoor static measurement MEMS sensor

Abstract

This invention discloses a method for calibrating the sensitivity of an outdoor static measurement MEMS sensor, belonging to the field of sensor calibration technology. The specific steps of this method are as follows: First, a dual benchmark is constructed, including both factory benchmarks and field-specific data; then, multi-source data is synchronously collected at a set period, and a valid sequence is obtained after preprocessing; subsequently, a dual-verification chain is started in parallel to fit key indicators; the results are collaboratively judged, and advanced diagnostics are triggered if there are conflicts; after diagnosis, the collection period is extended for sensitivity drift and parameters are recalculated; finally, if the conditions are met, the calibration parameters are fixed, a fault output alarm is installed, and the entire process data is recorded and uploaded to the cloud. This invention ensures sensor sensitivity calibration by constructing a dual benchmark, relying on it to synchronously collect multi-source data, and evaluating performance through a dual-verification chain. Efficiency is improved through time-slice rotation scheduling, valid data is retained during data preprocessing, differentiated processing is performed after fault diagnosis, the entire process data is recorded and uploaded, calibration parameters are fixed, ensuring the validity of the results and broadening the application scope.

Description

Technical Field

[0001] This invention relates to the field of sensor calibration technology, specifically to a method for calibrating the sensitivity of an outdoor static measurement MEMS sensor. Background Technology

[0002] MEMS sensors, with their advantages of small size, low power consumption, and controllable cost, are increasingly widely used in outdoor static measurement, covering multiple scenarios such as environmental monitoring and engineering structure monitoring. The measurement accuracy of these sensors directly affects the reliability of subsequent data analysis and decision-making, and sensitivity, as a core performance indicator, plays a crucial role in the measurement results. However, in outdoor environments, sensors must withstand the influence of complex environmental factors such as temperature changes and humidity fluctuations over a long period of time, and also undergo an environmental adaptation process after installation. These factors can all cause shifts in sensor sensitivity, thereby affecting the validity of measurement data. Therefore, establishing a scientific and reasonable sensor sensitivity calibration mechanism for outdoor static measurement scenarios has become an important prerequisite for ensuring the long-term stable operation of the measurement system and provides fundamental support for accurate monitoring in related fields.

[0003] Traditional sensor sensitivity calibration methods often rely on factory calibration data, which only reflects the sensor's performance under standard laboratory conditions. This makes it difficult to adapt to complex and ever-changing outdoor usage scenarios. Some calibration methods use a single benchmark for parameter correction, lacking consideration for dynamic changes in the field environment, leading to deviations between calibration results and actual working conditions. Furthermore, traditional methods handle outliers and data gaps in a simplistic manner during data processing, making it difficult to guarantee data quality. They also lack effective fault diagnosis mechanisms, failing to accurately distinguish between different problem types such as sensitivity drift and installation malfunctions. In addition, traditional calibration processes lack full-cycle data traceability and dynamic optimization capabilities. Once calibration parameters are fixed, they are difficult to adjust according to environmental changes and sensor aging. Measurement accuracy tends to decline after long-term use, failing to meet the long-term stability and reliability requirements of outdoor static measurements. Summary of the Invention

[0004] The purpose of this invention is to overcome the shortcomings of existing technologies and provide a method for calibrating the sensitivity of outdoor static measurement MEMS sensors. This method constructs a dual benchmark system, consisting of a factory benchmark and a field-related benchmark. The calibration is completed through steps such as multi-source data acquisition and preprocessing, parallel fitting of dual-calibration chains, conflict collaborative judgment, fault diagnosis and optimization, and calibration solidification and traceability. Scientific data processing strategies ensure data quality, parallel computing improves calibration efficiency, multi-feature fault discrimination accurately distinguishes fault types, and dynamic calibration algorithms that integrate dual benchmarks optimize parameters. The entire process achieves calibration accuracy, dynamic adaptability, and traceability, effectively coping with the impact of complex outdoor environments and solving problems such as insufficient adaptability and poor accuracy in traditional calibration, ensuring the long-term stable and reliable operation of the sensor.

[0005] To solve the above-mentioned technical problems, the present invention provides the following technical solution: a method for calibrating the sensitivity of an outdoor static measurement MEMS sensor, the specific steps of which are as follows:

[0006] S1, Constructing a dual benchmark: Obtain the sensor's factory calibration parameters, collect high-quality data from the installed sensor and the reference sensor during the initial installation phase, analyze the field correspondence between the output physical quantities of the installed sensor and the reference sensor, determine the field coupling coefficient and confidence interval, and form a dual benchmark that includes the factory benchmark and the field relationship.

[0007] The dual benchmarks of factory reference and field relationship are as follows: the factory reference is a standard calibration benchmark composed of the sensor's factory calibration parameters, specifically including factory linearity, temperature coefficient, constants, and the standard deviation of the Gaussian distribution corresponding to the above parameters; the field relationship benchmark is a dynamic benchmark adapted to the field environment, constructed based on high-quality data collected in the early stage of installation, specifically composed of the field correspondence model of the physical quantities output by the installed sensor and the reference sensor, the field coupling coefficient, and the confidence interval of the coupling coefficient.

[0008] S2, Collect and preprocess multi-source data: Based on dual benchmarks, collect target physical quantity data of the installed sensor and reference sensor synchronously according to a set cycle, and use the 3σ criterion to remove outliers and fill gaps to obtain effective data sequences;

[0009] S3, Perform parallel fitting of dual verification chains: Using the effective data sequence obtained in step S2, start the factory benchmark comparison chain and the field coupling relationship comparison chain in parallel to fit, and calculate the corresponding linearity, temperature coefficient and coupling deviation respectively.

[0010] S4, Collaborative Judgment and Conflict Handling: Gather the key indicators of the two verification chains obtained in step S3, verify their respective judgment conditions, and if the judgment results are consistent, proceed to the subsequent process; if the results conflict, trigger advanced diagnosis.

[0011] S5, perform fault diagnosis and data optimization: after triggering advanced diagnosis, use a multi-feature fault intelligent discrimination algorithm to determine the fault type; if it is determined to be sensor sensitivity drift, extend the data acquisition cycle and return to step S2, and calculate the total calibration parameters through a dynamic calibration algorithm that integrates dual benchmarks.

[0012] S6, complete calibration, solidification, and information traceability: if the total calibration parameters meet the overall judgment conditions, they are written into the sensor's non-volatile memory for solidification; if an installation fault is determined, an alarm message is output; at the same time, all relevant data throughout the process are recorded and uploaded to the cloud platform.

[0013] Furthermore, the sensor's factory calibration parameters include factory linearity, factory temperature coefficient, factory constant, and the Gaussian distribution standard deviation corresponding to the above three parameters. The Gaussian distribution standard deviation is obtained in the following way: before leaving the factory, the sensor is placed in a temperature chamber ranging from -40℃ to 85℃, and the temperature is increased and decreased in 5℃ increments. After each step of temperature is maintained for 30 minutes, linearity, temperature coefficient, and constant data are collected. The cumulative cycle is no less than 30 times. All data of each type of parameter are fitted with a normal distribution, and the corresponding linearity standard deviation, temperature coefficient standard deviation, and constant standard deviation are calculated.

[0014] Furthermore, the high-quality data acquisition window during the initial installation period is set to be no less than 5 days of continuous acquisition after installation, with an acquisition frequency of once per hour; the fluctuation range of the output modulus of the installed sensor does not exceed 0.01mV, and the fluctuation range of the physical quantity output by the reference sensor does not exceed 0.1% of the range of the reference sensor.

[0015] Furthermore, the analysis of the on-site correspondence between the physical quantities output by the installed sensor and the reference sensor adopts either a structural mechanics model or a data-driven model. When using a structural mechanics model, the mapping relationship between the output physical quantities of the two is established based on the inherent physical coupling characteristics of the measured structure. When using a data-driven model, the cross-correlation coefficient between the output physical quantities of the two is calculated, and a sample set with a cross-correlation coefficient of not less than 0.98 is selected for modeling. After modeling is completed, the effectiveness of the model needs to be verified. The absolute value of the model fitting residual should not exceed 0.002mV, the goodness of fit should not be less than 0.99, and 30% of high-quality data samples are randomly selected for verification. The mean value of the verification deviation should not exceed 0.0015mV.

[0016] Furthermore, the set cycle is once per hour; data acquisition synchronization is achieved through the synchronous trigger signal output by the sensor main control unit; each set of temperature, output modulus, and reference physical quantity data corresponds to a unique timestamp; when using the 3σ criterion to remove outliers, the mean and standard deviation of each type of data are first calculated, and data exceeding the range of the mean plus or minus 3 times the standard deviation are judged as outliers and removed; when filling data gaps, if the gap is only 1 data point, linear interpolation of two adjacent valid data points is used; if the gap is 2 consecutive data points, polynomial fitting interpolation of the first 3 valid data points is used; if the gap reaches or exceeds 3 data points, it is marked as a data outlier segment and removed.

[0017] Furthermore, the fitting of the factory benchmark comparison chain and the field coupling relationship comparison chain is achieved through parallel computation using a time-slice round-robin scheduling mechanism by the sensor main control unit; the factory benchmark comparison chain adopts the least squares fitting method, and the iteration termination condition for fitting is that the linearity difference between two adjacent iterations does not exceed The temperature coefficient difference does not exceed mV / ℃, the maximum number of iterations is no more than 50; when calculating the coupling deviation of the on-site coupling relationship comparison chain, the output modulus data of the installed sensors and the target physical quantity data of the reference sensor are collected synchronously for each group on the day, and the theoretical modulus value is obtained by substituting them into the on-site correspondence model. The absolute difference between the actual modulus value and the theoretical modulus value is calculated. After removing outliers that exceed 3 times the average of all differences, the arithmetic mean of the remaining differences is taken as the coupling deviation for the day.

[0018] Furthermore, the key indicators include the mean linearity, mean temperature coefficient, mean coupling deviation, and mean coupling linearity, where the mean coupling linearity is the mean linearity when fitting the field correspondence model; the judgment condition for the factory benchmark comparison chain is that the difference between the mean linearity and the factory linearity does not exceed 3 times the standard deviation of the factory linearity, and the difference between the mean temperature coefficient and the factory temperature coefficient does not exceed 3 times the standard deviation of the factory temperature coefficient; the judgment condition for the field coupling relationship comparison chain is that the mean coupling deviation is within the deviation range corresponding to the confidence interval of the field coupling coefficient, and the mean coupling linearity is not lower than 0.995; specific situations of double-chain conflict include the factory chain meeting the judgment condition while the field chain does not, the factory chain not meeting the judgment condition while the field chain does, and the two chains having opposite judgment logics and the mean coupling linearity being lower than 0.99.

[0019] Furthermore, the mathematical expression of the multi-feature fault intelligent discrimination algorithm is: ,in, This represents a comprehensive index for multi-feature fault discrimination. This represents the average temperature coefficient of the dual-check chain coupling during the conflict phase. This refers to the factory temperature coefficient. The standard deviation of the Gaussian distribution of the factory temperature coefficient. The coupling bias mutation rate, As the baseline mutation rate, For dynamic weights in the supply chain, For on-site dynamic weighting, , , The feature is adaptive weight; the fault discrimination comprehensive index is compared with the preset multiple thresholds one by one. When the index is in the first threshold range, it is determined to be sensitivity drift caused by sensor material degradation. When it is in the second threshold range, it is determined to be signal abnormality caused by sensor manufacturing defects. When it is in the third threshold range, it is determined to be coupling failure caused by installation deviation.

[0020] Furthermore, the initial duration of the extended data acquisition period is set to 7 days. If the bus linearity is lower than 0.995 or the temperature coefficient variance exceeds 0.8 times the factory standard deviation of the temperature coefficient after 7 days, the period is extended to 10 days. If the stability condition is still not met after 10 days, the period is extended in increments of 2 days, with the longest extension period not exceeding 15 days. The number of valid data sets within the extended period is not less than 168 sets, and the number of valid data sets per day is not less than 22 sets. The total calibration parameters include bus linearity, total temperature coefficient, and total constant. The bus linearity is obtained by fitting all temperature-output modulus data within the extended period, and the total constant is obtained by taking the average of the calculated results of the output modulus of all valid data and the temperature corresponding to the total temperature coefficient.

[0021] Furthermore, the mathematical expression for the dynamic calibration algorithm is: ,in, This represents the total temperature coefficient in the total calibration parameters. The confidence level of the comparison chain is based on the factory baseline. To determine the confidence level of the on-site coupling relationship comparison chain. The temperature coefficient of the supply chain is the one fitted to the factory condition. The equivalent temperature coefficient of the on-site chain. As an environmental adaptation factor, This is an environmental drift compensation item.

[0022] Compared with existing technologies, this outdoor static measurement MEMS sensor sensitivity calibration method has the following advantages:

[0023] I. This invention provides dual assurance for sensor sensitivity calibration by constructing a dual benchmark—the factory benchmark and the field benchmark—effectively adapting to the dynamic changes in complex outdoor environments. It relies on the dual benchmarks to synchronously collect multi-source data and undergoes rigorous preprocessing to ensure data validity and reliability. Then, through parallel fitting of dual verification chains, it comprehensively mines key indicators in the data, achieving multi-dimensional evaluation of sensor performance. The collaborative processing mechanism for the dual-chain judgment results can accurately identify potential conflicts and trigger advanced diagnostics. Combined with a multi-feature fault intelligent discrimination algorithm, it accurately distinguishes fault types, avoiding blind calibration. The dynamic calibration algorithm integrating dual benchmarks fully combines the advantages of both types of benchmarks, achieving dynamic optimization of calibration parameters, significantly improving the accuracy and adaptability of calibration results, and ensuring stable measurement performance of the sensor during long-term use.

[0024] Second, this invention achieves parallel operation of dual verification chains through a time-slice round-robin scheduling mechanism, improving the operational efficiency of the calibration process and shortening the calibration cycle. In the data preprocessing stage, a scientific outlier removal and gap filling strategy is employed to retain the maximum amount of valid data, providing high-quality data support for subsequent fitting and judgment. After fault diagnosis, differentiated processing methods are adopted according to different fault types. For sensitivity drift, the data acquisition cycle is extended and dynamically adjusted to ensure the stability and reliability of calibration parameters. For installation faults, alarm information is output promptly for quick troubleshooting. The full-process data recording and cloud upload functions enable traceability and review of the calibration process, providing data support for subsequent optimization of calibration strategies. Simultaneously, the design of non-volatile memory for storing calibration parameters ensures the long-term validity of calibration results, broadening the application scope of the sensor in outdoor static measurement scenarios.

[0025] Other advantages, objectives and features of the invention will be set forth in part in the description which follows, and in part will be apparent to those skilled in the art from the following examination or study, or may be learned from the practice of the invention. Attached Figure Description

[0026] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the accompanying drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are merely some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without any creative effort.

[0027] Figure 1 A flowchart of the sensitivity calibration method for outdoor static measurement MEMS sensors;

[0028] Figure 2 A graph showing the output relationship of each step in the sensitivity calibration of an outdoor static measurement MEMS sensor.

[0029] Figure 3 This is a flowchart of the parallel fitting and judgment process for dual-calibration chains of MEMS sensors used for outdoor static measurement. Detailed Implementation

[0030] To further illustrate the technical means and effects of the present invention in achieving its intended purpose, the following detailed description of the specific implementation methods, structures, features, and effects of the present invention, in conjunction with the accompanying drawings and preferred embodiments, is provided below.

[0031] Example 1:

[0032] Implementation of MEMS sensor sensitivity calibration in bridge structural health monitoring scenarios.

[0033] S1. Establishing a Dual Benchmark: First, the factory calibration parameters of the MEMS sensor to be installed are obtained, including factory linearity, temperature coefficient, constant, and the standard deviation of the Gaussian distribution corresponding to each of these three parameters. This constitutes the factory benchmark, providing an initial standard basis for subsequent calibration. After the sensor is installed on the key monitoring section of the bridge main beam, a high-quality data acquisition phase begins in the initial installation stage. The acquisition window is set for 6 consecutive days, with an acquisition frequency of once per hour. Sufficient and regular acquisition ensures the comprehensiveness and representativeness of the data. During the acquisition process, it is ensured that the fluctuation range of the output modulus of the installed MEMS sensor does not exceed 0.01mV, while the fluctuation range of the physical quantity output by the reference high-precision strain sensor does not exceed 0.1% of its range. Strict control of the data fluctuation range ensures the high quality of the acquired data. Based on the acquired high-quality data, a structural mechanics model is used to analyze the field correspondence between the physical quantities output by the two sensors. A mapping relationship between the two is established based on the inherent physical coupling characteristics of the bridge main beam, making the model more consistent with the actual physical conditions of the bridge structure. After modeling is completed, the model's effectiveness is verified, ensuring that the absolute value of the model fit residuals does not exceed 0.002mV and the goodness of fit is not lower than 0.99. Furthermore, 30% of high-quality data samples are randomly selected for verification, and the mean of the verification deviation does not exceed 0.0015mV. This rigorous verification process ensures the model's reliability, thereby determining the field coupling coefficients and their confidence intervals, forming a dual benchmark that includes both factory benchmarks and field relationships. Combining these two benchmarks improves the adaptability and accuracy of calibration. Figure 1 As shown.

[0034] S2. Data Acquisition and Preprocessing from Multiple Sources: Based on the established dual benchmarks, target physical quantity data from the installed MEMS sensors and reference sensors are synchronously acquired at a set cycle of once per hour. Fixed-cycle acquisition ensures the timeliness and continuity of the data. Synchronization of data acquisition is achieved through a synchronization trigger signal output by the MEMS sensor's main control unit. Each set of temperature, output modulus, and reference physical quantity data corresponds to a unique timestamp. Synchronous acquisition and unique timestamps ensure data correspondence and avoid confusion. After acquisition, outlier removal is performed using the 3σ criterion. The mean and standard deviation of each data category are calculated first. Data exceeding the mean plus or minus three times the standard deviation are identified as outliers and removed, effectively eliminating data interference that does not conform to normal patterns. For data gaps, if the gap is only one data point, linear interpolation of two adjacent valid data points is used; if the gap consists of two consecutive data points, polynomial fitting interpolation of the first three valid data points is used; if the gap reaches or exceeds three data points, it is marked as an outlier segment and removed. This scientific gap filling method ensures the integrity of the data sequence, ultimately obtaining a valid data sequence and providing reliable data support for subsequent fitting calculations.

[0035] S3, Parallel Fitting of Dual-Verification Chains: Utilizing the obtained effective data sequence, the sensor main control unit employs a time-slice round-robin scheduling mechanism to initiate parallel fitting of the factory benchmark comparison chain and the field coupling relationship comparison chain. Parallel computation improves calibration efficiency and shortens overall calibration time. The factory benchmark comparison chain uses a least-squares fitting method. The iteration termination condition is that the linearity difference between two adjacent iterations does not exceed a specified threshold, the temperature coefficient difference does not exceed a specified threshold, and the maximum number of iterations does not exceed 50. Strict iteration termination conditions ensure the accuracy of the fitting results. The corresponding linearity and temperature coefficient are calculated through this fitting. When calculating the coupling deviation of the field coupling relationship comparison chain, the output modulus data of each synchronously acquired MEMS sensor and the target physical quantity data of the reference sensor are substituted into the field correspondence model to obtain the theoretical modulus value. The absolute difference between the actual modulus value and the theoretical modulus value is calculated. After removing outliers exceeding three times the mean of all differences, the arithmetic mean of the remaining differences is taken as the coupling deviation for the day. A reasonable deviation calculation method can accurately reflect the degree of agreement between the field data and the model. Figure 3 As shown.

[0036] S4, Collaborative Judgment and Conflict Handling: Key indicators from both verification chains are aggregated, including the mean linearity, mean temperature coefficient, mean coupling deviation, and mean coupling linearity. The mean coupling linearity is the mean linearity measured during the fitting of the field-based correspondence model. Comprehensive aggregation of key indicators ensures the comprehensiveness of the judgment. The judgment conditions for each chain are verified. For the factory benchmark comparison chain, the judgment condition is that the difference between the mean linearity and the factory linearity does not exceed three times the standard deviation of the factory linearity, and the difference between the mean temperature coefficient and the factory temperature coefficient does not exceed three times the standard deviation of the factory temperature coefficient. For the field-based coupling relationship comparison chain, the judgment condition is that the mean coupling deviation is within the deviation range corresponding to the confidence interval of the field coupling coefficient, and the mean coupling linearity is not lower than 0.995. Clear judgment conditions provide a clear standard for result determination. If the judgment results of the two chains are consistent, the process will proceed to the next step. If there are conflict situations such as the factory chain meeting the judgment conditions but the field chain not meeting them, the factory chain not meeting the judgment conditions but the field chain meeting them, or the two chains having opposite judgment logics and an average coupling linearity of less than 0.99, advanced diagnosis will be triggered. Timely handling of conflicts can prevent deviations in subsequent calibration.

[0037] S5, perform fault diagnosis and data optimization: After triggering advanced diagnosis, a multi-feature intelligent fault discrimination algorithm is used to determine the fault type. The mathematical expression of the multi-feature intelligent fault discrimination algorithm is: ,in, This represents a comprehensive index for multi-feature fault discrimination. This represents the average temperature coefficient of the dual-check chain coupling during the conflict phase. This refers to the factory temperature coefficient. The standard deviation of the Gaussian distribution of the factory temperature coefficient. The coupling bias mutation rate, As the baseline mutation rate, For dynamic weights in the supply chain, For on-site dynamic weighting, , , For feature-adaptive weighting, the fault discrimination comprehensive index is compared one by one with preset multiple thresholds. When the index is in the first threshold range, it is determined to be sensitivity drift caused by sensor material degradation; in the second threshold range, it is determined to be signal abnormality caused by sensor manufacturing defects; and in the third threshold range, it is determined to be coupling failure caused by installation deviation. If it is determined to be sensor sensitivity drift, the data acquisition period is extended, initially set to 7 days. Extending the acquisition period allows for the acquisition of more data for accurate calibration. During the extended acquisition period, at least 168 valid data sets are ensured, and at least 22 valid data sets are obtained daily. Sufficient valid data can improve the accuracy of calibration parameter calculation. After 7 days, if the overall stability is lower than 0.995 or the temperature coefficient variance exceeds 0.8 times the factory temperature coefficient standard deviation, the period is extended to 10 days. If the stability conditions are still not met after 10 days, the period is extended in increments of 2 days, with a maximum extension period of 15 days. The acquisition period is flexibly adjusted according to data stability to ensure that the data meets calibration requirements. Then, returning to the data acquisition and preprocessing steps, the total calibration parameters are calculated using a dynamic calibration algorithm that integrates dual benchmarks. The mathematical expression for the dynamic calibration algorithm is: ,in, This represents the total temperature coefficient in the total calibration parameters. The confidence level of the comparison chain is based on the factory baseline. To determine the confidence level of the on-site coupling relationship comparison chain. The temperature coefficient of the supply chain is the one fitted to the factory condition. The equivalent temperature coefficient of the on-site chain. As an environmental adaptation factor, For environmental drift compensation, the total calibration parameters include bus linearity, total temperature coefficient, and total constant. The bus linearity is obtained by fitting all temperature-output modulus data within the extended period. The total constant is obtained by averaging the output modulus of all valid data and the temperature corresponding to the total temperature coefficient. The algorithm that integrates dual benchmarks allows the calibration parameters to better fit the actual use scenario.

[0038] S6, Completion of Calibration and Information Traceability: If the calculated total calibration parameters meet the overall judgment conditions, they are written into the non-volatile memory of the MEMS sensor for solidification, completing the sensitivity calibration. The solidified parameters ensure that the sensor always uses accurate calibration parameters during subsequent stable operation. If the multi-feature fault intelligent discrimination algorithm determines that it is an installation fault, such as a loose sensor installation, an alarm message is output to remind staff to handle it in a timely manner and avoid the fault affecting the accuracy of monitoring data. At the same time, all relevant data of the entire calibration process, including data acquisition records, outlier removal, fitting results, fault diagnosis process, and calibration parameters, are recorded and uploaded to the cloud monitoring platform for subsequent information traceability and query. The complete data record provides strong support for subsequent maintenance and problem troubleshooting.

[0039] In summary, this embodiment focuses on the sensitivity calibration of MEMS sensors at key sections of bridge main beams. By constructing a dual benchmark combining factory and field data, a foundation for accurate calibration is laid. Multi-source data is collected synchronously according to specified cycles, and data quality is ensured through 3σ criterion preprocessing. Parallel fitting with dual-calibration chains further improves computational efficiency and result accuracy. Conflicts are handled through collaborative judgment, and problems are located using a multi-feature fault intelligent discrimination algorithm. To address sensitivity drift, the acquisition cycle is extended, and parameters are optimized using a dynamic calibration algorithm integrating dual benchmarks. Finally, calibration is solidified, and full-process data traceability is achieved. The entire process strictly adheres to predetermined parameters and standards, ensuring sensor compatibility with bridge monitoring needs and guaranteeing accurate and reliable monitoring data.

[0040] Example 2:

[0041] Implementation of MEMS sensor sensitivity calibration in building structure vibration monitoring scenarios.

[0042] S1. Establishing a dual benchmark: First, acquire the factory calibration parameters of the MEMS sensors used for building structure vibration monitoring, covering factory linearity, temperature coefficient, constants, and the standard deviation of the Gaussian distribution corresponding to each parameter, forming a factory benchmark to lay the foundation for calibration. After the sensors are installed at key locations on the building frame columns, high-quality data acquisition is initiated during the initial installation phase. The acquisition window is 5 consecutive days, with an acquisition frequency of once per hour. Reasonable acquisition duration and frequency ensure data coverage of various environmental conditions during the initial installation phase. During the acquisition period, the fluctuation range of the output modulus of the installed MEMS sensor is strictly controlled to not exceed 0.01mV, and the fluctuation range of the physical quantity output by the reference sensor does not exceed 0.1% of its range. High-standard control of data fluctuations ensures the reliability of the acquired data. Based on the acquired high-quality data, a data-driven model is used to analyze the on-site correspondence between the output physical quantities of the installed sensor and the reference sensor. By calculating the cross-correlation coefficient between the two output physical quantities, a sample set with a cross-correlation coefficient of not less than 0.98 is selected for modeling. A sample set with a high cross-correlation coefficient can improve the relevance and accuracy of the model. After modeling, the effectiveness of the model is verified. The absolute value of the model fit residuals must not exceed 0.002 mV, and the goodness of fit must not be lower than 0.99. 30% of the high-quality data samples are randomly selected for verification, and the mean of the verification deviation must not exceed 0.0015 mV. Comprehensive verification ensures that the model meets the requirements of practical applications. The field coupling coefficient and its confidence interval are determined, and a dual benchmark combining the factory benchmark and the field relationship is constructed. The complementarity of these two benchmarks improves the accuracy and adaptability of calibration. Figure 2 As shown.

[0043] S2. Data Acquisition and Preprocessing from Multiple Sources: Based on dual benchmarks, target physical quantity data from the installed MEMS sensors and reference sensors are synchronously acquired at a set cycle of once per hour. This regular acquisition cycle ensures data continuity and comparability. Synchronization is achieved through a synchronization trigger signal output by the sensor's main control unit. Each data set has a unique timestamp, ensuring accurate time correspondence for subsequent analysis. In the data preprocessing stage, the 3σ criterion is used to remove outliers. The mean and standard deviation of each data type are calculated, and data exceeding the mean plus or minus three times the standard deviation are removed, effectively filtering out the impact of outlier data on the calibration results. For data gaps, if only one data point is missing, linear interpolation between two adjacent valid data points is used to fill the gap; if two consecutive data points are missing, polynomial fitting interpolation between the first three valid data points is used; if the gap reaches or exceeds three data points, it is marked as an outlier segment and removed. This reasonable gap handling method ensures the integrity and validity of the data sequence, ultimately yielding a valid data sequence and providing a high-quality data foundation for dual-calibration chain fitting.

[0044] S3, Parallel Fitting of Dual-Check Chains: Utilizing effective data sequences, the fitting process of the factory benchmark comparison chain and the field coupling relationship comparison chain is run in parallel through the time-slice round-robin scheduling mechanism of the sensor main control unit. Parallel processing significantly improves fitting efficiency and reduces calibration time costs. The factory benchmark comparison chain uses the least squares fitting method, fitting according to the set iteration termination conditions until the linearity difference and temperature coefficient difference between two adjacent iterations meet the specified requirements, and the number of iterations does not exceed 50. Strict fitting standards ensure the accuracy of the fitting results, thereby calculating the corresponding linearity and temperature coefficient. When calculating the coupling deviation of the field coupling relationship comparison chain, the output modulus data of the installed sensors and the target physical quantity data of the reference sensor, which are synchronously collected for each group on the same day, are substituted into the field correspondence model to obtain the theoretical modulus value. The absolute difference between the actual modulus value and the theoretical modulus value is calculated. After eliminating abnormal differences, the arithmetic mean of the remaining differences is taken as the coupling deviation for the day. Scientific deviation calculation can accurately reflect the matching situation of the two chains.

[0045] S4. Collaborative Conflict Judgment and Handling: Collect key indicators for both calibration chains, namely the mean linearity, mean temperature coefficient, mean coupling deviation, and mean coupling linearity. Comprehensive indicator collection makes the judgment results more convincing. Verify the judgment conditions for both chains separately. For the factory benchmark comparison chain, the difference between the mean linearity and the factory linearity must not exceed 3 times the standard deviation of the factory linearity, and the difference between the mean temperature coefficient and the factory temperature coefficient must not exceed 3 times the standard deviation of the factory temperature coefficient. For the field coupling comparison chain, the mean coupling deviation must be within the deviation range corresponding to the confidence interval of the field coupling coefficient, and the mean coupling linearity must not be lower than 0.995. Clear judgment criteria ensure the consistency and fairness of the judgment results. If the judgment results of the two chains are consistent, proceed to the next process; if any dual-chain conflict occurs, such as the factory chain meeting the condition but the field chain not meeting it, the factory chain not meeting the condition but the field chain meeting it, or the judgment logic of the two chains being opposite and the mean coupling linearity being lower than 0.99, advanced diagnosis is triggered to promptly identify the conflict and initiate targeted handling procedures to avoid deviation in the calibration direction.

[0046] S5, Fault Diagnosis and Data Optimization: After triggering advanced diagnosis, a multi-feature fault intelligent discrimination algorithm is used to determine the fault type. The mathematical expression of the multi-feature fault intelligent discrimination algorithm is: ,in, This represents a comprehensive index for multi-feature fault discrimination. This represents the average temperature coefficient of the dual-check chain coupling during the conflict phase. This refers to the factory temperature coefficient. The standard deviation of the Gaussian distribution of the factory temperature coefficient. The coupling bias mutation rate, As the baseline mutation rate, For dynamic weights in the supply chain, For on-site dynamic weighting, , , For feature-adaptive weighting, the fault discrimination comprehensive index is compared one by one with preset multiple thresholds. When the index is in the first threshold range, it is determined to be sensitivity drift caused by sensor material degradation; in the second threshold range, it is determined to be signal abnormality caused by sensor manufacturing defects; and in the third threshold range, it is determined to be coupling failure caused by installation deviation. If it is determined to be sensor sensitivity drift, the data acquisition cycle is extended, initially by 7 days, during which at least 22 sets of valid data are guaranteed per day, and at least 168 sets of valid data are guaranteed in total. Sufficient valid data can provide sufficient support for calibration parameter calculation. After 7 days, the bus uniformity and temperature coefficient variance are checked. If the bus uniformity is lower than 0.995 or the temperature coefficient variance exceeds 0.8 times the factory standard deviation of the temperature coefficient, the acquisition cycle is extended to 10 days. If the stability conditions are still not met, the cycle is extended in increments of 2 days, with a maximum of 15 days. The acquisition cycle is dynamically adjusted according to the data stability to ensure that the data quality meets the calibration requirements. Then, returning to the data acquisition and preprocessing steps, the total calibration parameters are calculated using a dynamic calibration algorithm that integrates dual benchmarks. The mathematical expression for the dynamic calibration algorithm is: ,in, This represents the total temperature coefficient in the total calibration parameters. The confidence level of the comparison chain is based on the factory baseline. To determine the confidence level of the on-site coupling relationship comparison chain. The temperature coefficient of the supply chain is the one fitted to the factory condition. The equivalent temperature coefficient of the on-site chain. As an environmental adaptation factor, For environmental drift compensation, the total calibration parameters include bus linearity, total temperature coefficient, and total constant, which are obtained through corresponding calculation methods. This algorithm can integrate the advantages of dual benchmarks, making the calibration parameters more in line with the actual needs of building structure vibration monitoring.

[0047] S6, Calibration and Information Traceability Completion: If the overall calibration parameters meet the overall judgment conditions, they are written into the non-volatile memory of the MEMS sensor for solidification, completing the calibration. The solidified parameters ensure that the sensor maintains accurate measurement sensitivity during long-term use. If an installation fault is detected, such as installation position deviation, an alarm message is output to promptly remind staff to rectify the issue and avoid data distortion due to installation problems. Simultaneously, all relevant data from the entire calibration process, including acquired data, preprocessing records, fitting results, fault diagnosis conclusions, and calibration parameters, are recorded and uploaded to the cloud platform, enabling full data traceability. Complete data records provide crucial information for subsequent calibration effect evaluation, sensor maintenance, and troubleshooting.

[0048] In summary, this embodiment focuses on the calibration of MEMS sensors for building frame columns. First, it establishes a dual benchmark based on the sensor's factory parameters and high-quality data from the initial installation phase, providing a solid foundation for calibration. Data is collected synchronously at a set period and preprocessed according to standardized procedures, eliminating outliers and filling data gaps to form a valid data sequence. Parallel fitting of the dual verification chains is achieved through time-slice round-robin scheduling, and conflicts are collaboratively determined after accurate calculation of key indicators. After triggering advanced diagnostics, a multi-feature fault intelligent discrimination algorithm is used to identify the fault type, adjust the acquisition period and optimize calibration parameters based on sensitivity drift dynamics, and finally complete parameter solidification or fault alarm. Simultaneously, the entire process of data traceability is uploaded, strictly adhering to established specifications and parameter requirements to ensure that the sensor meets the accuracy requirements for building vibration monitoring.

[0049] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention in any way. Although the present invention has been disclosed above with reference to preferred embodiments, it is not intended to limit the present invention. Any person skilled in the art can make some modifications or alterations to the above-disclosed technical content to create equivalent embodiments without departing from the scope of the present invention. Any simple modifications, equivalent changes and alterations made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the scope of the present invention.

Claims

1. A method for calibrating the sensitivity of an outdoor static measurement MEMS sensor, characterized in that, The specific steps of this method are as follows: S1, Constructing a dual benchmark: Obtain the sensor's factory calibration parameters, collect high-quality data from the installed sensor and the reference sensor during the initial installation phase, analyze the field correspondence between the output physical quantities of the installed sensor and the reference sensor, determine the field coupling coefficient and confidence interval, and form a dual benchmark that includes the factory benchmark and the field relationship. S2, Collect and preprocess multi-source data: Based on dual benchmarks, collect target physical quantity data of the installed sensor and reference sensor synchronously according to a set cycle, and use the 3σ criterion to remove outliers and fill gaps to obtain effective data sequences; S3, Perform parallel fitting of dual verification chains: Using the effective data sequence obtained in step S2, start the factory benchmark comparison chain and the field coupling relationship comparison chain in parallel to fit, and calculate the corresponding linearity, temperature coefficient and coupling deviation respectively. The fitting of the factory benchmark comparison chain and the field coupling relationship comparison chain is achieved through parallel computation using a time-slice round-robin scheduling mechanism by the sensor main control unit; the factory benchmark comparison chain adopts the least squares fitting method, and the iteration termination condition is that the linearity difference between two adjacent iterations does not exceed The temperature coefficient difference does not exceed mV / ℃, the maximum number of iterations is no more than 50; when calculating the coupling deviation of the on-site coupling relationship comparison chain, the output modulus data of the installed sensor and the target physical quantity data of the reference sensor are collected synchronously on the same day, and the theoretical modulus value is obtained by substituting them into the on-site correspondence model. The absolute difference between the actual modulus value and the theoretical modulus value is calculated. After removing outliers that exceed 3 times the mean of all differences, the arithmetic mean of the remaining differences is taken as the coupling deviation of the day. S4, Collaborative Judgment and Conflict Handling: Gather the key indicators of the two verification chains obtained in step S3, verify their respective judgment conditions, and if the judgment results are consistent, proceed to the subsequent process; if the results conflict, trigger advanced diagnosis. The key indicators include the mean linearity, mean temperature coefficient, mean coupling deviation, and mean coupling linearity. The mean coupling linearity is the mean linearity when fitting the field-based correspondence model. The criteria for determining the factory benchmark comparison chain are that the difference between the mean linearity and the factory linearity does not exceed three times the standard deviation of the factory linearity, and the difference between the mean temperature coefficient and the factory temperature coefficient does not exceed three times the standard deviation of the factory temperature coefficient. The criteria for determining the field-based coupling relationship comparison chain are that the mean coupling deviation is within the deviation range corresponding to the confidence interval of the field coupling coefficient, and the mean coupling linearity is not lower than 0.

995. Specific situations of double-chain conflict include the factory chain meeting the criteria while the field chain does not, the factory chain not meeting the criteria while the field chain does, and the two chains having opposite judgment logics and a mean coupling linearity lower than 0.

99. S5, perform fault diagnosis and data optimization: after triggering advanced diagnosis, use a multi-feature fault intelligent discrimination algorithm to determine the fault type; if it is determined to be sensor sensitivity drift, extend the data acquisition cycle and return to step S2, and calculate the total calibration parameters through a dynamic calibration algorithm that integrates dual benchmarks. S6, complete calibration, solidification, and information traceability: if the total calibration parameters meet the overall judgment conditions, they are written into the sensor's non-volatile memory for solidification; if an installation fault is determined, an alarm message is output; at the same time, all relevant data throughout the process are recorded and uploaded to the cloud platform.

2. The outdoor static measurement MEMS sensor sensitivity calibration method according to claim 1, characterized in that, The sensor's factory calibration parameters include factory linearity, factory temperature coefficient, factory constant, and the Gaussian distribution standard deviation corresponding to the above three parameters. The Gaussian distribution standard deviation is obtained in the following way: before leaving the factory, the sensor is placed in a temperature chamber ranging from -40℃ to 85℃, and the temperature is increased and decreased in 5℃ increments. After each step of temperature is maintained for 30 minutes, linearity, temperature coefficient, and constant data are collected. The cumulative cycle is no less than 30 times. All data of each type of parameter are fitted with a normal distribution, and the corresponding linearity standard deviation, temperature coefficient standard deviation, and constant standard deviation are calculated.

3. The outdoor static measurement MEMS sensor sensitivity calibration method according to claim 1, characterized in that, In step S1, the high-quality data acquisition window period during the initial installation period is set to be no less than 5 days of continuous acquisition after installation, with an acquisition frequency of once per hour; the fluctuation range of the output modulus of the installed sensor does not exceed 0.01mV, and the fluctuation range of the physical quantity output by the reference sensor does not exceed 0.1% of the range of the reference sensor.

4. The outdoor static measurement MEMS sensor sensitivity calibration method according to claim 1, characterized in that, In step S1, the analysis of the field correspondence between the physical quantities output by the installed sensor and the reference sensor adopts either a structural mechanics model or a data-driven model. When using a structural mechanics model, the mapping relationship between the output physical quantities of the two is established based on the inherent physical coupling characteristics of the structure under test. When using a data-driven model, the cross-correlation coefficient between the output physical quantities of the two is calculated, and a sample set with a cross-correlation coefficient of not less than 0.98 is selected for modeling. After modeling is completed, the effectiveness of the model needs to be verified. The absolute value of the model fitting residuals should not exceed 0.002mV, the goodness of fit should not be lower than 0.99, and 30% of high-quality data samples should be randomly selected for verification. The mean value of the verification bias should not exceed 0.0015mV.

5. The outdoor static measurement MEMS sensor sensitivity calibration method according to claim 1, characterized in that, In step S2, the set period is once per hour; Data acquisition synchronization is achieved through the synchronous trigger signal output by the sensor main control unit; each set of temperature, output modulus, and reference physical quantity data corresponds to a unique timestamp; when using the 3σ criterion to remove outliers, the mean and standard deviation of each type of data are first calculated, and data exceeding the range of the mean plus or minus 3 times the standard deviation are judged as outliers and removed; when filling data gaps, if the gap is only 1 data point, linear interpolation of two adjacent valid data points is used; if the gap is 2 consecutive data points, polynomial fitting interpolation of the first 3 valid data points is used; if the gap reaches or exceeds 3 data points, it is marked as a data outlier segment and removed.

6. The outdoor static measurement MEMS sensor sensitivity calibration method according to claim 1, characterized in that, In step S5, the mathematical expression of the multi-feature fault intelligent discrimination algorithm is: ,in, This represents a comprehensive index for multi-feature fault discrimination. This represents the average temperature coefficient of the dual-check chain coupling during the conflict phase. This refers to the factory temperature coefficient. The standard deviation of the Gaussian distribution of the factory temperature coefficient. The coupling bias mutation rate, As the baseline mutation rate, For dynamic weights in the supply chain, For on-site dynamic weighting, , , Adaptive weights for features.

7. The outdoor static measurement MEMS sensor sensitivity calibration method according to claim 1, characterized in that, In step S5, the initial duration of the extended data acquisition period is set to 7 days. If the overall stability is lower than 0.995 or the temperature coefficient variance exceeds 0.8 times the factory standard deviation of the temperature coefficient after 7 days, the period is extended to 10 days. If the stability condition is still not met after 10 days, the period is extended in increments of 2 days, with the longest extension period not exceeding 15 days. The number of valid data sets within the extended period shall not be less than 168 sets, and the number of valid data sets per day shall not be less than 22 sets. The total calibration parameters include bus linearity, total temperature coefficient, and total constant. The bus linearity is obtained by fitting all temperature-output modulus data within the extended period, and the total constant is obtained by averaging the calculated results of the output modulus of all valid data and the temperature corresponding to the total temperature coefficient.

8. The outdoor static measurement MEMS sensor sensitivity calibration method according to claim 1, characterized in that, In step S5, the mathematical expression of the dynamic calibration algorithm is: ,in, This represents the total temperature coefficient in the total calibration parameters. The confidence level of the comparison chain is based on the factory baseline. To determine the confidence level of the on-site coupling relationship comparison chain. The temperature coefficient of the supply chain is the one fitted to the factory condition. The equivalent temperature coefficient of the on-site chain. As an environmental adaptation factor, This is an environmental drift compensation item.

Images