Plant calibration method for humidity drift of MEMS sensor in outdoor cold state measurement

Abstract

The invention relates to the field of MEMS sensors, and discloses a factory calibration method for humidity drift of an MEMS sensor in outdoor cold-state measurement, and the factory calibration method comprises the following steps: S1, high-humidity environment construction and natural attenuation, and S2, dynamic data synchronous acquisition. The technical problem of a traditional constant humidity box calibration method is fundamentally solved, coupling interference of vibration of equipment such as a fan and a compressor on output signals of the MEMS sensor is thoroughly eliminated, the purity and authenticity of drift data are ensured, and the calibration coefficient can accurately reflect physical correlation between humidity and sensor output; the natural humidity attenuation process simulates natural scenes such as day and night circulation and weather change experienced by the sensor during actual outdoor application, and essential difference exists between the continuously and gradually changed humidity path and the jumping type steady-state control of the constant humidity box, so that the calibrated humidity drift coefficient has higher generalization ability and compensation applicability in actual working conditions.

Description

Technical Field

[0001] This invention relates to the field of MEMS sensors, and more particularly to a factory calibration method for humidity drift of MEMS sensors in outdoor cold-state measurements. Background Technology

[0002] MEMS (Micro-Electro-Mechanical Systems) humidity sensors are widely used in meteorological observation, smart agriculture, warehouse management, and outdoor IoT devices to monitor ambient relative humidity. With the expansion of application scenarios, more and more devices need to operate for extended periods in harsh outdoor environments, especially in winter or high-altitude regions where ambient temperatures are often below 0°C, or even as low as -40°C. In these cold-state outdoor measurement scenarios, the measurement accuracy of MEMS humidity sensors faces significant challenges. The moisture-sensitive materials of humidity sensors (such as polyimide and porous silicon) are highly sensitive to temperature. In low-temperature environments, the adsorption / desorption kinetics of water molecules in the moisture-sensitive materials change significantly, leading to a severe deviation in the correspondence between the sensor's capacitance or resistance value and the actual relative humidity—a phenomenon known as humidity temperature drift.

[0003] In existing technologies, the calibration of humidity drift of MEMS sensors is usually performed in a constant temperature and humidity test chamber. This method has significant drawbacks: the fans, compressors and other equipment built into the constant humidity chamber generate continuous vibrations of 5-15mg. These vibrations couple into the sensor output signal, creating noise interference and altering the stress distribution of the chip and PCB, making it impossible to accurately extract pure humidity drift characteristics. The test chamber maintains constant humidity through active humidification / dehumidification. Its step steady-state control is fundamentally different from actual outdoor natural gradual change conditions (such as day-night cycles and weather changes), and it cannot simulate the nonlinear effect of humidity change rate on drift, resulting in a decrease in the practical applicability of the calibration coefficients. In addition, the multi-point steady-state calibration strategy takes 4-6 hours, and the cost of high-precision equipment is as high as 150,000-250,000 yuan per unit. The number of calibrations per batch is limited, making it difficult to meet the rapid delivery requirements of large-scale production. Summary of the Invention

[0004] To address the shortcomings of existing technologies, this invention provides a factory calibration method for humidity drift of MEMS sensors in outdoor cold-state measurements, thus solving the above problems.

[0005] To achieve the above objectives, the present invention provides the following technical solution: a factory calibration method for humidity drift of MEMS sensors in outdoor cold-state measurements, comprising the following steps: S1. High humidity environment construction and natural decay: Place the MEMS sensor to be calibrated in a sealed calibration chamber, first raise the humidity inside the chamber to the target high value, then stop all active humidity control and vibration generation equipment, and let the humidity inside the chamber decay to the ambient humidity under natural conditions. S2. Dynamic data synchronous acquisition: During the natural decay of humidity, the raw output data of the MEMS sensor is synchronously and continuously acquired and recorded as Y, and the ambient humidity data inside the cavity is recorded as H. S3. Data preprocessing and feature construction: Denoising and aligning the collected raw data Y and H, and calculating the humidity change rate, denoised as... Construct a dataset for modeling, including humidity H and the rate of humidity change. And the corresponding sensor output change is denoted as ,in , This refers to the sensor's output under known reference conditions. S4. Decoupling Physical-Mathematical Model Construction and Fitting: Based on the physical laws governing humidity and drift, a correction model integrating linear and nonlinear relationships is established.

[0006] in, The change in sensor output For ambient humidity, The rate of change of humidity, , , , Here are the humidity drift correction coefficients to be fitted. For model residuals; A nonlinear regression algorithm is used to evaluate the model parameters. , , , Perform fitting; S5. Parameter Verification and Storage: The fitted humidity drift correction coefficients are then stored. , , , The parameters are written into the sensor's built-in non-volatile memory as factory calibration parameters.

[0007] Preferably, in step S1, after the active humidity control device is stopped, the calibration chamber maintains limited gas exchange with the outside world so that the humidity decreases gradually, simulating the natural humidity change process outdoors.

[0008] Preferably, in step S3, the data preprocessing includes using a moving average filter to remove high-frequency noise and using the 3σ criterion to remove data anomalies caused by accidental interference.

[0009] Preferably, in step S4, when building the model, temperature T is also introduced as a compensation variable to construct a temperature-related humidity coefficient model:

[0010] in, , , , Here, T is the coefficient that varies with temperature, and T is the ambient temperature. By repeating steps S1-S3 at multiple temperature points, a functional relationship between the coefficients and temperature is fitted.

[0011] Preferably, in step S5, the variance of the residual distribution of the fitted model is also stored as prior knowledge for use in setting the process noise of the subsequent online correction algorithm.

[0012] Preferably, the method further includes step S6: long-term correction interface: providing an updateable interface for the calibration parameters, so that the sensor can dynamically correct the factory calibration parameters based on the online updated coefficients during subsequent long-term use.

[0013] Preferably, the MEMS sensor is a MEMS accelerometer or a MEMS tilt sensor, and the unit of the raw output data Y is one-thousandth of the gravitational acceleration in mg.

[0014] Preferably, in step S1, the target high value is above 95%RH, and the ambient humidity is 40%-60%RH.

[0015] Preferably, in step S2, the synchronous acquisition frequency is once per minute, and the acquisition duration is 2-4 hours until the humidity inside the cavity stabilizes within the range of ±2%RH of the ambient humidity.

[0016] Preferably, in step S4, the nonlinear regression algorithm is the least squares method or the maximum likelihood estimation method, and the determination coefficient R of the model fit is... 2 Not less than 0.99.

[0017] This invention provides a factory calibration method for humidity drift of MEMS sensors in outdoor cold-state measurements. Compared with existing technologies, it has the following advantages: This invention fundamentally solves the technical difficulties of traditional constant humidity chamber calibration methods by constructing a calibration environment free from vibration interference and with natural humidity changes. Firstly, after shutting down all active vibration sources, the vibration acceleration within the calibration chamber decreases from 5-15 mg in the constant humidity chamber to below 0.5 mg, completely eliminating the coupling interference of vibrations from equipment such as fans and compressors on the MEMS sensor output signal. This ensures the purity and authenticity of drift data, enabling the calibration coefficients to accurately reflect the physical correlation between humidity and sensor output. Secondly, the natural humidity decay process simulates the day-night cycle experienced by the sensor in actual outdoor applications. The continuous and gradual humidity path in natural scenarios such as weather changes is fundamentally different from the jump-type steady-state control of the constant humidity chamber. This makes the calibrated humidity drift coefficient have stronger generalization ability and compensation applicability in actual working conditions. In addition, continuous data from high humidity to ambient humidity can be obtained in a single natural decay process without the need to maintain multiple steady-state points for a long time. The calibration cycle is shortened to less than 3 hours, and the equipment requirements are extremely low. Only a simple sealed chamber and a wet steam generator are needed. The cost of a single set of equipment is 97% lower than that of a high-precision constant humidity chamber, providing an economical and feasible technical solution for rapid batch calibration of the production line. 2. In this invention, by employing a decoupled physical-mathematical model that integrates linear and nonlinear relationships, and by introducing humidity, the rate of humidity change, and their quadratic terms, the complex coupled influence of humidity and its rate of change on sensor output can be more accurately characterized. The model's coefficient of determination R... 2 The standard deviation of the zero-point drift reaches 0.996, a 5.2% improvement over the traditional linear model, laying a theoretical foundation for high-precision compensation. Simultaneously, the variance of the residual distribution of the fitted model is stored as prior knowledge, which can be directly used for setting the process noise of the subsequent online correction algorithm, achieving seamless integration between factory calibration and field applications. Furthermore, the long-term correction interface reserved in this invention enables the sensor to have lifelong learning capabilities. During long-term use, the factory calibration parameters can be dynamically optimized based on online monitoring data, effectively adapting to the slow changes in sensor performance over time. This significantly extends the high-precision working life of the MEMS sensor in complex and humid outdoor environments. Actual outdoor verification shows that after compensation using the calibration coefficients of this invention, the standard deviation of the sensor's zero-point drift is reduced by 91.2%, a 20.9 percentage point improvement over the traditional constant humidity chamber calibration method, fully demonstrating the technical superiority of this invention in outdoor static measurement applications. Attached Figure Description

[0018] Figure 1 This is a flowchart of the factory calibration method for humidity drift of MEMS sensors in outdoor cold-state measurement proposed in this invention; Figure 2 This diagram illustrates a comparison of sensor output signals during the calibration process of a traditional constant humidity chamber (with vibration) and the natural decay calibration process of this invention (without vibration). Detailed Implementation

[0019] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0020] Please see Figures 1-2 The present invention provides the following technical solutions, specifically including the following embodiments: Example 1: The factory calibration method for humidity drift of MEMS sensors in outdoor cold-state measurements includes the following steps: S1. High Humidity Environment Construction and Natural Attenuation: The MEMS sensor to be calibrated is placed in a sealed calibration chamber. The humidity inside the chamber is first raised to the target high value. Then, all active humidity control and vibration-generating devices are stopped, allowing the humidity inside the chamber to decay to the ambient humidity under natural conditions. After the active humidity control devices are stopped, the calibration chamber maintains limited gas exchange with the outside world to make the humidity decrease slowly, simulating the natural humidity change process outdoors. The MEMS sensor is a MEMS accelerometer or a MEMS tilt sensor. The unit of the original output data Y is one-thousandth of the gravitational acceleration in mg. The target high value is above 95%RH, and the ambient humidity is 40%-60%RH. S2. Dynamic data synchronous acquisition: During the natural decay of humidity, the raw output data of the MEMS sensor is synchronously and continuously acquired and recorded as Y, and the ambient humidity data inside the cavity is recorded as H. The synchronous acquisition frequency is once per minute, and the acquisition duration is 2-4 hours until the humidity inside the cavity stabilizes within the range of ±2%RH of the ambient humidity. S3. Data preprocessing and feature construction: Denoising and aligning the collected raw data Y and H, and calculating the humidity change rate, denoised as... Construct a dataset for modeling, including humidity H and the rate of humidity change. And the corresponding sensor output change is denoted as ,in , For the sensor output under known reference conditions, the data preprocessing includes using a moving average filter to remove high-frequency noise and using the 3σ criterion to remove data outliers caused by accidental interference. S4. Decoupling Physical-Mathematical Model Construction and Fitting: Based on the physical laws governing humidity and drift, a correction model integrating linear and nonlinear relationships is established.

[0021] in, The change in sensor output For ambient humidity, The rate of change of humidity, , , , Here are the humidity drift correction coefficients to be fitted. For model residuals; A nonlinear regression algorithm is used to evaluate the model parameters. , , , Perform fitting; When building the model, temperature T was also introduced as a compensation variable to construct a temperature-dependent humidity coefficient model:

[0022] in, , , , Here, T is the coefficient that varies with temperature, and T is the ambient temperature. By repeating steps S1-S3 at multiple temperature points, a functional relationship between the coefficients and temperature is fitted. The nonlinear regression algorithm is either the least squares method or the maximum likelihood estimation method, and the determination coefficient R of the model fit is... 2 Not less than 0.99.

[0023] S5. Parameter Verification and Storage: The fitted humidity drift correction coefficients are then stored. , , , The data is written into the sensor's built-in non-volatile memory as factory calibration parameters. The variance of the residual distribution of the fitted model is also stored as prior knowledge for use in setting the process noise of the subsequent online calibration algorithm.

[0024] It also includes step S6: Long-term correction interface: providing an updateable interface for calibration parameters, so that the sensor can dynamically correct the factory calibration parameters based on the online updated coefficients during subsequent long-term use.

[0025] Example 2: Based on Example 1, this example uses a MEMS tilt sensor (model: TA-30, range ±30°, resolution 0.001°) for bridge health monitoring as the object. Humidity drift factory calibration is performed using both the traditional constant humidity chamber calibration method and the natural decay calibration method of this invention. The technical effect of this invention is verified through comparative experiments. Experimental conditions and equipment configuration: Twelve MEMS tilt sensors from the same batch were randomly divided into two groups, A and B, with six sensors in each group. Group A was calibrated using the traditional constant humidity chamber method, while Group B was calibrated using the natural attenuation method of this invention. Traditional constant temperature and humidity chamber calibration equipment (Group A): Programmable constant temperature and humidity test chamber: temperature range -40℃~150℃, humidity range 20%RH~98%RH, built-in circulating fan and compressor, humidity control accuracy ±2%RH, vibration characteristics: fan operation vibration acceleration is about 5~15mg (spectrum analysis shows that the main frequency is 50Hz and its harmonics). The calibration equipment of this invention (Group B) includes: a sealed calibration chamber with a volume of 10L, a thermal insulation material thickness of 50mm, and a micro-gap (0.5mm wide gap for slow gas exchange with the outside); a wet steam generator for rapid humidification to the target value; a reference sensor with a high-precision humidity sensor (accuracy ±0.8%RH, acquisition frequency 1Hz); and a design without active vibration sources: after all fans and compressors are turned off, the vibration acceleration inside the chamber is <0.5mg.

[0026] Group A: Implementation process of traditional humidity chamber calibration method: Step 1: Set steady-state humidity points: Select 6 steady-state humidity calibration points: 95%RH, 85%RH, 75%RH, 65%RH, 55%RH, 45%RH, maintain each humidity point for 30 minutes, and keep the temperature constant at 25℃; Step 2: Data Acquisition: In the last 10 minutes of each steady-state humidity point, the sensor output value is acquired at a frequency of 1Hz, and the average value is taken as the drift amount at that humidity point. Step 3: Vibration interference phenomenon: When the compressor of the constant humidity chamber starts, the sensor output shows obvious shaking, with a shaking amplitude of about ±8mg. The continuous operation of the fan causes the output signal to fluctuate periodically, with a standard deviation of about 3.5mg. In order to obtain a stable reading, it is necessary to wait for the compressor to stop, and the single calibration cycle is as long as 4 hours. Step 4: Model Fitting: A linear model was used: ΔY = a·H + b. The fitting results were: a = -0.38, b = 36.2, and the coefficient of determination R0 was 1 / 2. 2 = 0.947.

[0027] Group B: Implementation process of the natural attenuation calibration method of this invention: Step S1: High humidity environment construction and natural decay: Place the sensor in a sealed calibration chamber, start the wet steam generator, and raise the humidity in the chamber to 98%RH within 15 minutes. Stabilize the temperature at 25℃. After reaching the target humidity, immediately turn off the wet steam generator and all active devices, leaving only the data acquisition device powered. The chamber slowly exchanges gas with the outside world (ambient humidity 52%RH) through a reserved micro-gap, and the humidity decays naturally. Step S2: Dynamic data synchronous acquisition: Acquisition frequency: 1 time / minute, acquisition duration: 3 hours (humidity naturally decays from 98%RH to 51%RH), synchronous recording: sensor output Y (unit: mg), ambient humidity H (unit: %RH); Step S3: Data Preprocessing and Feature Construction: High-frequency noise is removed using a 5-point moving average filter, outliers are removed based on the 3σ criterion (removal rate <0.3%), and the humidity change rate is calculated. The unit is %RH / minute. Calculate the drift amount: ,in The sensor output at the initial moment (98%RH); Step S4: Decoupling the physical and mathematical model construction and fitting: Establishing a calibration model that integrates linear and nonlinear relationships.

[0028] in, The change in sensor output For ambient humidity, The rate of change of humidity, , , , Here are the humidity drift correction coefficients to be fitted. For model residuals; A nonlinear regression algorithm is used to evaluate the model parameters. , , , By performing a fitting, the calibration coefficients are obtained: = -0.4521, = 0.0102, = 0.00458, = 0.00011, coefficient of determination R 2 = 0.996; Step S5: Parameter Verification and Storage: ... , , , and residual variance σ 2 =0.0012 is written to the sensor's Flash memory, and the variance of the residual distribution is used as the noise prior value of the Kalman filtering process.

[0029] Actual working condition verification: Three sensors from each of groups A and B, after calibration, were deployed at the same outdoor bridge monitoring point (ambient humidity range 30%RH~90%RH, day-night temperature difference 10℃). After 30 days of continuous operation, all vibration sources were turned off, humidity was allowed to decrease naturally, and the drift curve smoothed out. The actual compensation effect was then compared. Verification metric: Standard deviation of zero drift

[0030] Comparative analysis table:

[0031] Conclusion: Raw data collected under vibration-free conditions accurately reflects humidity drift characteristics, avoiding the contamination of model parameters by vibration noise. The decoupled physical-mathematical model integrating humidity H, humidity change rate dH / dt, and their quadratic terms has a high coefficient of determination R. 2 The accuracy reached 0.996, a 5.2% improvement over the traditional linear model. The humidity change path (98%RH→50%RH) covered by the natural decay process is consistent with the actual outdoor day-night cycle and weather change patterns. The calibration coefficient has strong generalization ability. The simple sealed cavity can replace high-precision constant humidity equipment. Multiple sensors can be calibrated in batches by a single cavity, meeting the needs of large-scale production. In long-term outdoor monitoring, the standard deviation of sensor zero drift decreased from 12.5mg to 1.1mg, with an accuracy improvement of 91.2%, which is 20.9 percentage points higher than the traditional method. The cost of a single calibration equipment set decreased from 150,000 to 250,000 yuan to less than 5,000 yuan, a reduction of 97%. Therefore, this invention provides a low-cost, high-efficiency, and high-precision factory calibration technology path for the large-scale application of MEMS sensors in outdoor static health monitoring of bridges, slopes, buildings, etc., effectively solving the problem of long-term measurement accuracy degradation caused by humidity drift, and has significant engineering practical value and application prospects.

[0032] The above description is merely a preferred embodiment of this application and is not intended to limit this application. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of this application should be included within the protection scope of this application.

Claims

1. A factory calibration method for humidity drift of MEMS sensors in outdoor cold-state measurements, characterized in that: Includes the following steps: S1. High humidity environment construction and natural decay: Place the MEMS sensor to be calibrated in a sealed calibration chamber, first raise the humidity inside the chamber to the target high value, then stop all active humidity control and vibration generation equipment, and let the humidity inside the chamber decay to the ambient humidity under natural conditions. S2. Dynamic data synchronous acquisition: During the natural decay of humidity, the raw output data of the MEMS sensor is synchronously and continuously acquired and recorded as Y, and the ambient humidity data inside the cavity is recorded as H. S3. Data preprocessing and feature construction: Denoising and aligning the collected raw data Y and H, and calculating the humidity change rate, denoised as... Construct a dataset for modeling, including humidity H and the rate of humidity change. And the corresponding sensor output change is denoted as ,in , This refers to the sensor's output under known reference conditions. S4. Decoupling Physical-Mathematical Model Construction and Fitting: Based on the physical laws governing humidity and drift, a correction model integrating linear and nonlinear relationships is established. in, The change in sensor output For ambient humidity, The rate of change of humidity, , , , Here are the humidity drift correction coefficients to be fitted. For model residuals; A nonlinear regression algorithm is used to evaluate the model parameters. , , , Perform fitting; S5. Parameter Verification and Storage: The fitted humidity drift correction coefficients are then stored. , , , The parameters are written into the sensor's built-in non-volatile memory as factory calibration parameters.

2. The factory calibration method for humidity drift of MEMS sensors in outdoor cold-state measurement according to claim 1, characterized in that: In step S1, after the active humidity control device is stopped, the calibration chamber maintains limited gas exchange with the outside world to allow the humidity to decrease gradually, simulating the natural humidity change process outdoors.

3. The factory calibration method for humidity drift of MEMS sensors in outdoor cold-state measurement according to claim 1, characterized in that: In step S3, the data preprocessing includes using a moving average filter to remove high-frequency noise and using the 3σ criterion to remove data anomalies caused by accidental interference.

4. The factory calibration method for humidity drift of MEMS sensors in outdoor cold-state measurement according to claim 1, characterized in that: In step S4, when building the model, temperature T is also introduced as a compensation variable to construct a temperature-dependent humidity coefficient model: in, , , , Here, T is the coefficient that varies with temperature, and T is the ambient temperature. By repeating steps S1-S3 at multiple temperature points, a functional relationship between the coefficients and temperature is fitted.

5. The factory calibration method for humidity drift of MEMS sensors in outdoor cold-state measurement according to claim 1, characterized in that: In step S5, the variance of the residual distribution of the fitted model is also stored as prior knowledge for use in setting the process noise of the subsequent online correction algorithm.

6. The factory calibration method for humidity drift of MEMS sensors in outdoor cold-state measurement according to claim 1, characterized in that: It also includes step S6: Long-term correction interface: providing an updateable interface for calibration parameters, so that the sensor can dynamically correct the factory calibration parameters based on the online updated coefficients during subsequent long-term use.

7. The factory calibration method for humidity drift of MEMS sensors in outdoor cold-state measurement according to claim 1, characterized in that: The MEMS sensor is a MEMS accelerometer or a MEMS tilt sensor, and the unit of the raw output data Y is one-thousandth of the gravitational acceleration in mg.

8. The factory calibration method for humidity drift of MEMS sensors in outdoor cold-state measurement according to claim 1, characterized in that: In step S1, the target high value is above 95%RH, and the ambient humidity is 40%-60%RH.

9. The factory calibration method for humidity drift of MEMS sensors in outdoor cold-state measurement according to claim 1, characterized in that: In step S2, the synchronous acquisition frequency is once per minute, and the acquisition duration is 2-4 hours until the humidity inside the cavity stabilizes within the range of ±2%RH of the ambient humidity.

10. The factory calibration method for humidity drift of MEMS sensors in outdoor cold-state measurement according to claim 1, characterized in that: In step S4, the nonlinear regression algorithm is either the least squares method or the maximum likelihood estimation method, and the determination coefficient R of the model fit is... 2 Not less than 0.99.

Images