Humidity sensor fault detection method based on PCA and Jaccard distance

Abstract

The present invention discloses a humidity sensor fault detection method based on PCA and Jaccard distance, and relates to the field of humidity sensor fault detection. The present invention extracts the essential features of the original time series data of the sensor through principal component analysis, then converts the principal component matrix into a set, and then calculates the Jaccard distance between the principal component set corresponding to the sensor to be tested and the normal sensor, and determines whether the humidity sensor has a fault by whether the Jaccard distance exceeds a preset threshold. The present invention combines the feature extraction capability of principal component analysis with the advantages of Jaccard distance similarity measurement, realizes an effective humidity sensor fault detection method, can accurately identify the fault of the humidity sensor, and improves the accuracy of fault detection.

Description

Technical Field

[0001] The present invention relates to the field of humidity sensor fault detection, and in particular to a humidity sensor fault detection method based on PCA and Jaccard distance. Background Art

[0002] Humidity sensors are widely used in many fields, such as meteorological monitoring, industrial automation, agricultural production, smart home, etc. However, humidity sensors are prone to failure during long-term use due to environmental factors and component aging, resulting in inaccurate measurement results, which affects production and life. Therefore, it is necessary to monitor and troubleshoot humidity sensor failures. Currently, common humidity sensor fault detection methods have some limitations. For example, the detection method based on fixed thresholds cannot dynamically adapt to environmental changes, resulting in a high false alarm rate; the detection method based on machine learning requires high computing power and is costly, making it difficult to apply on a large scale. Summary of the invention

[0003] The purpose of the present invention is to improve the accuracy of fault detection and reduce the false alarm rate while reducing the cost.

[0004] To achieve the above object, the present invention provides a humidity sensor fault detection method based on PCA and Jaccard distance, the method comprising:

[0005] Step 1: With a fixed sampling period , continuously collect normal humidity sensor in a period of time The humidity measurement values ​​and environmental background parameters in the room are used to obtain the original time series data;

[0006] Step 2: constructing an observation matrix with the original time series data, and performing principal component analysis to obtain a principal component matrix of a normal humidity sensor as a normal principal component matrix;

[0007] Step 3: Repeat steps 1 and 2 to obtain a number of normal principal component matrices, and convert all the normal principal component matrices into sets to obtain a number of normal principal component sets;

[0008] Step 4: Calculate the Jaccard distance between any two normal principal component sets to obtain a normal Jaccard distance array, and select the maximum value in the normal Jaccard distance array as the Jaccard distance threshold;

[0009] Step 5: Sampling period , real-time collection of humidity sensors to be tested in the time period The humidity measurement value and environmental background parameters are used to obtain the time series data to be tested;

[0010] Step 6: construct an observation matrix with the time series data to be tested, and perform principal component analysis to obtain the principal component matrix of the humidity sensor to be tested as the principal component matrix to be tested, and convert the principal component matrix to be tested into a set to obtain the principal component set to be tested;

[0011] Step 7: Calculate the Jaccard distance between the principal component set to be tested and any one of the normal principal component sets respectively, obtain the Jaccard distance array to be tested, select the maximum value in the Jaccard distance array to be tested as the fault judgment value, and if the fault judgment value is greater than the Jaccard distance threshold, it is preliminarily determined that the humidity sensor is faulty.

[0012] The principle of the present invention is that the present invention can capture the dynamic relationship between the measured value of the humidity sensor under normal conditions and the environmental background parameters by collecting the original time series data of the normal humidity sensor. Through principal component analysis, the key structure and potential laws of the original time series data can be extracted. The normal principal component matrix is ​​converted into a set so that the similarity between different principal component matrices can be quantified by the Jaccard distance. The present invention first calculates the Jaccard distance of any two normal principal component sets and finds the maximum value therein, which represents the maximum difference allowed between the measured data of the humidity sensor under normal conditions. If the Jaccard distance between the principal component set of the sensor to be tested and any normal principal component set exceeds this maximum value, it means that the measured data of the sensor to be tested has a time series anomaly and the sensor may have a fault. The present invention uses principal component analysis to identify and remove redundant information in the original time series data, obtains the principal component that maximizes the reflection of data variability, can reduce the complexity and storage requirements of data calculation, and reduces costs. At the same time, the Jaccard distance can accurately reflect the similarities and differences between sets. By comparing the size of the principal component matrix to be tested with the normal principal component matrix, it can show the difference between the sensor to be tested and the normal sensor, thereby improving the accuracy of fault detection.

[0013] Furthermore, in step 2, the original time series data is standardized before constructing the observation matrix;

[0014] Furthermore, in step six, the time series data to be measured is standardized before constructing the observation matrix.

[0015] Among them, the purpose of standardization is to eliminate the impact of dimensions between data and improve the stability of principal component analysis.

[0016] Furthermore, the standardization process adopts a Z-score standardization method.

[0017] Furthermore, in step 2, the principal component analysis includes:

[0018] Calculate the covariance matrix of the observation matrix;

[0019] Solve the eigenvalues ​​of the covariance matrix and calculate the cumulative contribution rate;

[0020] The number of principal components is determined based on the cumulative contribution rate, and the corresponding number of principal component vectors are selected to construct the principal component matrix.

[0021] Among them, principal component analysis is a statistical dimensionality reduction method, which achieves dimensionality reduction by mapping the original high-dimensional data to a low-dimensional space through linear transformation. The data structure after dimensionality reduction is simplified, but the main information of the original data is retained.

[0022] Furthermore, the calculation formula of the cumulative contribution rate is:

[0023] ;

[0024] in, is the cumulative contribution rate, is the sum of the eigenvalues ​​of the first k principal components, is the sum of all eigenvalues, is the ith eigenvalue.

[0025] Furthermore, in step 4, the calculation formula of the Jaccard distance between the two normal principal component sets is:

[0026] ;

[0027] in, is the Jaccard distance between the pth normal principal component set and the qth normal principal component set, is the pth normal principal component set, is the qth normal principal component set, for and The Jaccard similarity coefficient.

[0028] The Jaccard distance is an indicator used to measure the difference between two sets. The larger the Jaccard distance, the greater the difference between the two sets; conversely, the smaller the difference. In the present invention, the Jaccard distance between the principal component sets reflects the difference between the original time series of the corresponding sensors.

[0029] Among them, in step 7, the calculation formula of the Jaccard distance between the principal component set to be tested and any of the normal principal component sets is:

[0030]

[0031] in, is the Jaccard distance between the principal component set to be tested and the nth normal principal component set, is the principal component set to be tested, is the nth normal principal component set, for and The Jaccard similarity coefficient.

[0032] Further, the environmental background parameters include: ambient temperature and dew point;

[0033] The method further comprises:

[0034] Step 8: Compare the ambient temperature and dew point of the humidity sensor to be tested. If the ambient temperature is higher than the dew point, it is confirmed that the humidity sensor to be tested is faulty.

[0035] The dew point refers to the temperature at which the air is cooled to saturation when the water vapor content in the air remains unchanged. When the temperature of the environment where the humidity sensor is located is lower than the dew point, condensation may form on the surface of the humidity sensor. Condensation will cause the measurement data of the humidity sensor to continuously shift. This shift will cause the Jaccard distance between the principal component set to be measured and the normal principal component set to be larger, which may easily cause a false alarm of a fault. Therefore, in order to further improve the accuracy of detection, the present invention also investigates the temperature and dew point of the environment where the humidity sensor is located. If the temperature of the environment is higher than the dew point, it means that the possibility of condensation is low, and condensation interference is eliminated, confirming the sensor failure.

[0036] Further, if the ambient temperature is not higher than the dew point, step nine is executed;

[0037] Step nine includes:

[0038] Again, the sampling period , real-time collection of humidity sensors to be tested in multiple consecutive time periods The humidity measurement values ​​and environmental background parameters in the room are used to obtain some verification time series data;

[0039] Standardize any of the verification time series data, project it onto the first principal component vector in the most matching normal principal component matrix, and calculate the projection score;

[0040] The proportion of negative samples in all projection scores is counted. If the proportion exceeds the preset condensation threshold, it is determined to be condensation interference. Otherwise, it is confirmed that the humidity sensor to be tested is faulty.

[0041] Among them, if the temperature of the environment is not higher than the dew point, further confirmation is required. The present invention confirms by continuing to collect a number of verification time series data, and projecting these verification time series data onto the first principal component in the most matching historical normal principal component matrix. Specifically, when the humidity sensor is in a normal state, its response mode is stable, the projection score is concentrated in the positive area, and the proportion of positive samples is relatively high. When condensation occurs, condensation interference causes a sudden change in the characteristics of the sensor's sensitive elements. For example, when condensation occurs on a capacitive sensor, the capacitance suddenly increases, causing the projection to reverse, and the proportion of negative samples to increase significantly. Therefore, the condensation interference is identified based on whether the proportion of negative samples in the projection score exceeds the preset condensation threshold, which has high accuracy.

[0042] Furthermore, the most matching normal principal component matrix is: the principal component matrix of the normal humidity sensor that is closest to the environmental background parameter value of the humidity sensor to be measured in step three; the first principal component is the principal component with the highest variance contribution rate.

[0043] Among them, the most matching normal principal component matrix is ​​selected to maximize the consistency of the projection direction with the current physical conditions of the humidity sensor to be tested, avoiding misjudgment caused by environmental offset. Selecting the first principal component for projection can retain the variation information of the verification time series data to the greatest extent, capture its main features, and reduce information loss.

[0044] Furthermore, the environmental background parameters also include: PM2.5, PM10 and air velocity;

[0045] The environmental background parameter value is obtained in the following way:

[0046] Weights are set for ambient temperature, dew point, PM2.5, PM10 and air velocity respectively, and the environmental background parameter value is calculated using a weighted method.

[0047] One or more technical solutions provided by the present invention have at least the following technical effects or advantages:

[0048] 1. The present invention combines the feature extraction capability of principal component analysis with the advantages of Jaccard distance similarity measurement, and provides an effective fault detection method for humidity sensors, which can accurately identify the faults of humidity sensors and improve the accuracy of fault detection.

[0049] 2. The present invention extracts the essential characteristics of the original time series data of the sensor through principal component analysis, improves the limitation of the single detection channel of the traditional method, and takes the maximum value of the Jaccard distance between the normal principal component sets corresponding to different environments as the threshold for fault judgment, so that the present invention can adapt to environmental fluctuations and reduce the false alarm rate.

[0050] 3. The present invention takes into account the actual situation of humidity sensors operating in certain high-humidity environments and adopts a secondary verification mechanism to reduce the problem of misjudgment of faults caused by condensation and prevent unnecessary sensor replacement or calibration.

[0051] 4. The method of the present invention has low demand for computing resources and does not need to rely on a large number of hardware devices. It can be effectively operated in low-configuration monitoring equipment or embedded systems, reducing hardware procurement and maintenance costs, and has high economy and practicality. BRIEF DESCRIPTION OF THE DRAWINGS

[0052] The drawings described herein are used to provide a further understanding of the embodiments of the present invention, constitute a part of the present invention, and do not constitute a limitation on the embodiments of the present invention;

[0053] Figure 1 It is a flow chart of the humidity sensor fault detection method based on PCA and Jaccard distance in the present invention. DETAILED DESCRIPTION

[0054] In order to more clearly understand the above-mentioned purpose, features and advantages of the present invention, the present invention is further described in detail below in conjunction with the accompanying drawings and specific embodiments. It should be noted that the embodiments of the present invention and the features in the embodiments can be combined with each other without conflict.

[0055] In the following description, many specific details are set forth to facilitate a full understanding of the present invention. However, the present invention may also be implemented in other ways different from those within the scope of this description. Therefore, the protection scope of the present invention is not limited to the specific embodiments disclosed below.

[0056] Embodiment 1

[0057] Please refer to Figure 1 Embodiment 1 of the present invention provides a humidity sensor fault detection method based on PCA and Jaccard distance, characterized in that the method includes:

[0058] Step 1: With a fixed sampling period , continuously collect normal humidity sensor in a period of time The humidity measurement values ​​and environmental background parameters in the room are used to obtain the original time series data;

[0059] Step 2: constructing an observation matrix with the original time series data, and performing principal component analysis to obtain a principal component matrix of a normal humidity sensor as a normal principal component matrix;

[0060] Step 3: Repeat steps 1 and 2 to obtain a number of normal principal component matrices, and convert all the normal principal component matrices into sets to obtain a number of normal principal component sets;

[0061] Step 4: Calculate the Jaccard distance between any two normal principal component sets to obtain a normal Jaccard distance array, and select the maximum value in the normal Jaccard distance array as the Jaccard distance threshold;

[0062] Step 5: Sampling period , real-time collection of humidity sensors to be tested in the time period The humidity measurement value and environmental background parameters are used to obtain the time series data to be tested;

[0063] Step 6: construct an observation matrix with the time series data to be tested, and perform principal component analysis to obtain the principal component matrix of the humidity sensor to be tested as the principal component matrix to be tested, and convert the principal component matrix to be tested into a set to obtain the principal component set to be tested;

[0064] Step 7: Calculate the Jaccard distance between the principal component set to be tested and any one of the normal principal component sets respectively, obtain the Jaccard distance array to be tested, select the maximum value in the Jaccard distance array to be tested as the fault judgment value, and if the fault judgment value is greater than the Jaccard distance threshold, it is preliminarily determined that the humidity sensor is faulty.

[0065] Among them, the sampling period and the time period Those skilled in the art can determine the sampling period according to actual needs. For example, It can be 10 minutes or 30 minutes. It can be 6 hours, 8 hours or 12 hours. The environmental background parameters include ambient temperature, dew point PM2.5, PM10, air pressure and air velocity.

[0066] Wherein, in step 2, the original time series data is standardized before constructing the observation matrix;

[0067] Furthermore, in step six, the time series data to be measured is standardized before constructing the observation matrix.

[0068] The standardization process adopts the Z-score standardization method, and the specific formula is:

[0069] ;

[0070] in, is the i-th measurement data, is the mean of the original time series data, is the standard deviation of the original time series data, for The standardized value of .

[0071] When constructing the observation matrix, the humidity measurement value and the environmental background parameter of each collection point are used as a row of the observation matrix, so that each column of the observation matrix corresponds to a parameter.

[0072] Take a normal humidity sensor as an example. Assuming a sampling period of 1 hour, the following raw time series data is obtained after 5 hours of continuous sampling (it should be noted that, in order to simplify the calculation, only three parameters, humidity measurement value, ambient temperature and air pressure, are set for each sample here):

[0073]

[0074] After standardizing the above original time series data, construct the observation matrix X, where:

[0075] ;

[0076] Wherein, in step 2, the principal component analysis includes:

[0077] Calculate the covariance matrix of the observation matrix;

[0078] Solve the eigenvalues ​​of the covariance matrix and calculate the cumulative contribution rate;

[0079] The number of principal components is determined based on the cumulative contribution rate, and the corresponding number of principal component vectors are selected to construct the principal component matrix.

[0080] The calculation formula of the covariance matrix is:

[0081] ;

[0082] in, is the covariance matrix, is the observation matrix, is the transpose of the observation matrix, is the sample size.

[0083] Taking the above observation matrix X as an example, its covariance matrix is ​​calculated as:

[0084] ;

[0085] The steps of solving the eigenvalues ​​of the covariance matrix include:

[0086] According to the formula Covariance matrix Perform eigendecomposition, where is the eigenvector matrix, is the eigenvalue diagonal matrix;

[0087] Then by solving the characteristic equation Get all eigenvalues, where is the characteristic value, is the identity matrix.

[0088] The eigenvalues ​​of the above covariance matrix are solved as follows: , , .

[0089] The calculation formula of the cumulative contribution rate is:

[0090] ;

[0091] in, is the cumulative contribution rate, is the sum of the eigenvalues ​​of the first k principal components, is the sum of all eigenvalues, is the ith eigenvalue.

[0092] because and The cumulative contribution rate exceeds 90%, so the number of principal components is determined to be 2, and the calculations are and The corresponding principal component vectors are PC1:[0.53,0.57,-0.63] and PC2:[-0.72,0.28,0.63].

[0093] Project the standardized original time series data onto PC1 and PC2 to obtain the principal component matrix, where:

[0094] ;

[0095] The steps of converting all the normal principal component matrices into a set are as follows:

[0096] Discretize and bin the numerical range of each principal component (such as PC1, PC2) in the principal component matrix to divide the corresponding intervals;

[0097] Map the principal component values ​​of each sample to the corresponding intervals and combine them to generate a composite category label;

[0098] The composite category labels of all samples are used as set elements to construct a principal component set.

[0099] For example, the numerical range corresponding to PC1 is [-2.43, 2.96], which is divided into three intervals: interval 1: [-3, -1), interval 2: [-1, 1), interval 3: [1, 3), the numerical range corresponding to PC2 is [-1.10, 0.87], which is also divided into three intervals: interval 4: [-2, -0.5), interval 5: [-0.5, 0.5), interval 6: [0.5, 2). It should be noted that this is only a simple division for the sake of example. In actual application, the intervals can be divided more finely, or dynamically divided based on quantiles or clusters. Those skilled in the art can adjust according to actual needs.

[0100] Furthermore, in the principal component matrix, the PC1 value of the first sample is -1.52, which belongs to interval 1, and the PC2 value is 0.87, which belongs to interval 6. Therefore, the category label of the first sample can be expressed as PC1-1_PC2-6. Similarly, the category labels of the second to fifth samples are PC1-2_PC2-5, PC1-3_PC2-4, PC1-6_PC2-3, and PC1-1_PC2-4, respectively.

[0101] Therefore, the principal component set obtained after the above principal component matrix is ​​transformed is:

[0102] ;

[0103] Among them, in step six, the step of converting the principal component matrix to be tested into a set is the same as the above step.

[0104] Among them, in step 4, the calculation formula of the Jaccard distance between the two normal principal component sets is:

[0105] ;

[0106] in, is the Jaccard distance between the pth normal principal component set and the qth normal principal component set, is the pth normal principal component set, is the qth normal principal component set, for and The Jaccard similarity coefficient.

[0107] For example, assuming that the above set S is the pth normal principal component set, assuming that the qth normal principal component set is:

[0108] ;

[0109] and The intersection of is 4 elements, and the union is 6 elements, so:

[0110] .

[0111] Among them, in step 7, the calculation formula of the Jaccard distance between the principal component set to be tested and any of the normal principal component sets is:

[0112] ;

[0113] in, is the Jaccard distance between the principal component set to be tested and the nth normal principal component set, is the principal component set to be tested, is the nth normal principal component set, for and The Jaccard similarity coefficient.

[0114] Wherein, the method further comprises:

[0115] Step 8: Compare the ambient temperature and dew point of the humidity sensor to be tested. If the ambient temperature is higher than the dew point, it is confirmed that the humidity sensor to be tested is faulty.

[0116] Among them, the dew point can be obtained by instruments such as dew point thermometer or dry-bulb thermometer, or it can be calculated using the Goff-Grech formula or the Magras formula.

[0117] If the ambient temperature is not higher than the dew point, proceed to step nine;

[0118] Step nine includes:

[0119] Again, the sampling period , real-time collection of humidity sensors to be tested in multiple consecutive time periods The humidity measurement values ​​and environmental background parameters in the room are used to obtain some verification time series data;

[0120] Standardize any of the verification time series data, project it onto the first principal component vector in the most matching normal principal component matrix, and calculate the projection score;

[0121] The proportion of negative samples in all projection scores is counted. If the proportion exceeds the preset condensation threshold, it is determined to be condensation interference. Otherwise, it is confirmed that the humidity sensor to be tested is faulty.

[0122] The projection score is calculated by performing a dot product operation on the standardized verification time series data and the first principal component. The number of verification time series data and the size of the condensation threshold can be determined by those skilled in the art according to actual conditions, and the present invention does not limit this.

[0123] Among them, the most matching normal principal component matrix is: the principal component matrix of the normal humidity sensor that is closest to the environmental parameter value of the humidity sensor to be measured in step three; the first principal component is the principal component with the highest variance contribution rate.

[0124] The environmental background parameter value is obtained in the following way:

[0125] Weights are set for ambient temperature, dew point, PM2.5, PM10 and air velocity respectively, and the environmental background parameter value is calculated using a weighted method. The specific weight setting can be determined by those skilled in the art based on experience.

[0126] Although the preferred embodiments of the present invention have been described, those skilled in the art may make other changes and modifications to these embodiments once they have learned the basic creative concept. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments and all changes and modifications that fall within the scope of the present invention.

[0127] Obviously, those skilled in the art can make various changes and modifications to the present invention without departing from the spirit and scope of the present invention. Thus, if these modifications and variations of the present invention fall within the scope of the claims of the present invention and their equivalents, the present invention is also intended to include these modifications and variations.

Claims

1. A humidity sensor fault detection method based on PCA and Jaccard distance, characterized in that: The method comprises: Step 1: With a fixed sampling period , continuously collect normal humidity sensor in a period of time The humidity measurement value and environmental background parameters in the room are used to obtain the original time series data, wherein the environmental background parameters include: ambient temperature and dew point; Step 2: constructing an observation matrix with the original time series data, and performing principal component analysis to obtain a principal component matrix of a normal humidity sensor as a normal principal component matrix; Step 3: Repeat steps 1 and 2 to obtain a number of normal principal component matrices, and convert all the normal principal component matrices into sets to obtain a number of normal principal component sets; Step 4: Calculate the Jaccard distance between any two normal principal component sets to obtain a normal Jaccard distance array, and select the maximum value in the normal Jaccard distance array as the Jaccard distance threshold; Step 5: Sampling period , real-time collection of humidity sensors to be tested in the time period The humidity measurement value and environmental background parameters are used to obtain the time series data to be tested; Step 6: construct an observation matrix with the time series data to be tested, and perform principal component analysis to obtain the principal component matrix of the humidity sensor to be tested as the principal component matrix to be tested, and convert the principal component matrix to be tested into a set to obtain the principal component set to be tested; Step 7: Calculate the Jaccard distance between the principal component set to be tested and any one of the normal principal component sets respectively, obtain the Jaccard distance array to be tested, select the maximum value in the Jaccard distance array to be tested as the fault judgment value, and if the fault judgment value is greater than the Jaccard distance threshold, it is preliminarily determined that the humidity sensor is faulty.

2. The humidity sensor fault detection method based on PCA and Jaccard distance according to claim 1 is characterized in that: In step 2, the original time series data is standardized before constructing the observation matrix; Furthermore, in step six, the time series data to be measured is standardized before constructing the observation matrix.

3. The humidity sensor fault detection method based on PCA and Jaccard distance according to claim 2 is characterized in that: The standardization process adopts the Z-score standardization method.

4. The humidity sensor fault detection method based on PCA and Jaccard distance according to claim 1, characterized in that: In step 2, the principal component analysis includes: Calculate the covariance matrix of the observation matrix; Solve the eigenvalues ​​of the covariance matrix and calculate the cumulative contribution rate; The number of principal components is determined based on the cumulative contribution rate, and the corresponding number of principal component vectors are selected to construct the principal component matrix.

5. The humidity sensor fault detection method based on PCA and Jaccard distance according to claim 4 is characterized in that: The calculation formula of the cumulative contribution rate is: ； in, is the cumulative contribution rate, is the sum of the eigenvalues ​​of the first k principal components, is the sum of all eigenvalues, is the i-th eigenvalue.

6. The humidity sensor fault detection method based on PCA and Jaccard distance according to claim 1, characterized in that: In step 4, the calculation formula of the Jaccard distance between the two normal principal component sets is: ； in, is the Jaccard distance between the pth normal principal component set and the qth normal principal component set, is the pth normal principal component set, is the qth normal principal component set, for and The Jaccard similarity coefficient.

7. The humidity sensor fault detection method based on PCA and Jaccard distance according to claim 1, characterized in that: The method further comprises: Step 8: Compare the ambient temperature and dew point of the humidity sensor to be tested. If the ambient temperature is higher than the dew point, it is confirmed that the humidity sensor to be tested is faulty.

8. The humidity sensor fault detection method based on PCA and Jaccard distance according to claim 7, characterized in that: If the ambient temperature is not higher than the dew point, proceed to step nine; Step nine includes: Again, the sampling period , real-time collection of humidity sensors to be tested in multiple consecutive time periods The humidity measurement values ​​and environmental background parameters in the room are used to obtain some verification time series data; Standardize any of the verification time series data, project it onto the first principal component vector in the most matching normal principal component matrix, and calculate the projection score; The proportion of negative samples in all projection scores is counted. If the proportion exceeds the preset condensation threshold, it is determined to be condensation interference. Otherwise, it is confirmed that the humidity sensor to be tested is faulty.

9. The humidity sensor fault detection method based on PCA and Jaccard distance according to claim 8, characterized in that: The most matching normal principal component matrix is: the principal component matrix of the normal humidity sensor that is closest to the environmental background parameter value of the humidity sensor to be measured in step 3; the first principal component is the principal component with the highest variance contribution rate.

10. The humidity sensor fault detection method based on PCA and Jaccard distance according to claim 9, characterized in that: The environmental background parameters also include: PM2.5, PM10 and air velocity; The environmental background parameter value is obtained in the following way: Weights are set for ambient temperature, dew point, PM2.5, PM10 and air velocity respectively, and the environmental background parameter value is calculated using a weighted method.

Images