Data Dynamic Fusion Method and System for Similar Monitoring Sensors

Abstract

The present invention discloses a method and system for data dynamic fusion of similar monitoring sensors, including obtaining the monitoring data of similar monitoring sensors to be fused, performing data cleaning and anomaly processing to obtain continuous monitoring data; initializing a basic matrix for state prediction and error correction; predicting the state at the current time step and calculating the prediction error covariance; calculating the mean square error between the observed value and the predicted value of each monitoring sensor, and calculating the weight factor; determining the fault state of each monitoring sensor and adjusting the measurement noise covariance; calculating the Kalman gain of each monitoring sensor and obtaining the total Kalman gain; and completing the data dynamic fusion of similar monitoring sensors by using a weighted summation scheme. The present invention adjusts the weighting factor based on Kalman filtering and dynamic weighted fusion, and at the same time adopts a fault diagnosis scheme to isolate abnormal monitoring sensors, not only realizing the data dynamic fusion of similar monitoring sensors, but also having high reliability and good accuracy.

Description

Technical Field

[0001] The present invention belongs to the field of data processing, and particularly relates to a method and system for dynamically fusing data of similar monitoring sensors. Background Art

[0002] With the development of economy and technology, sensor networks have been widely applied in people's production and life, such as in the fields of Internet of Things, autonomous driving, industrial monitoring, etc., bringing endless convenience to people's production and life. Therefore, ensuring the accuracy and reliability of the data in the sensor network has become one of the research focuses of researchers.

[0003] In current practical applications, sensor data may be missing or sparse due to reasons such as unstable network transmission, environmental interference, equipment aging, etc., thus affecting the continuity and integrity of the data in the sensor system. In addition, the sensor system is prone to failures or measurement errors after long-term operation, which will also lead to a decline in the quality of sensor data. In this case, after the sensor system performs data fusion, the poor-quality data will directly affect the reliability and accuracy of the fused data, and further affect the reliability and accuracy of the sensor system.

[0004] Existing sensor data fusion schemes, such as weighted average method, Bayesian filtering, etc., although can effectively improve the accuracy of data, have deficiencies in dealing with data missing, outliers and sensor failures, especially are weak in terms of real-time performance and fault detection. Summary of the Invention

[0005] One object of the present invention is to provide a method for dynamically fusing data of similar monitoring sensors with high reliability and good accuracy.

[0006] Another object of the present invention is to provide a system for implementing the method for dynamically fusing data of similar monitoring sensors.

[0007] The method for dynamically fusing data of similar monitoring sensors provided by the present invention includes the following steps:

[0008] S1. Obtain the monitoring data of similar monitoring sensors to be fused;

[0009] S2. Based on the sparse signal recovery method, regularization constraint and linear interpolation method, perform data cleaning and abnormal data processing on the monitoring data obtained in step S1 to obtain continuous monitoring data;

[0010] S3. Initialize the basic matrices for state prediction and error correction; the basic matrices include state transition matrix, process noise covariance matrix, observation matrix and measurement noise covariance matrix;

[0011] S4. Based on the data information obtained in steps S2 and S3, use the Kalman filter prediction model to predict the state at the current time step and calculate the prediction error covariance;

[0012] S5. For each monitoring sensor, calculate the mean square error between the observed value and the predicted value of the monitoring sensor, and calculate the corresponding weight factor based on the obtained mean square error;

[0013] S6. Based on the state prediction value and prediction error covariance of each monitoring sensor, determine the fault state of the monitoring sensor and adjust the corresponding measurement noise covariance;

[0014] S7. According to the measurement noise covariance and weight factor of each monitoring sensor, calculate the Kalman gain of each monitoring sensor, and use weighted summation for combination to obtain the total Kalman gain;

[0015] S8. According to the total Kalman gain obtained in step S7, use the weighted summation scheme to complete the data dynamic fusion of the same type of monitoring sensors.

[0016] The specific steps of step S2 are as follows:

[0017] For the data of each time step of each monitoring sensor, use the sparse signal recovery method to process the missing values, and realize data sparse recovery through L1 regularization constraint;

[0018] After sparse signal recovery, if there are still missing data, use the linear interpolation method to interpolate the missing data;

[0019] Finally, obtain continuous monitoring data.

[0020] The specific steps of step S4 are as follows:

[0021] Use the following formula for state prediction: In the formula is the predicted state at time step k; is the predicted state at time step k - 1; is the current state transition matrix;

[0022] Use the following formula for error covariance prediction: In the formula is the predicted error covariance matrix at time step k; is the predicted error covariance matrix at time step k - 1; Q is the current process noise covariance matrix.

[0023] The specific steps of step S5 are as follows:

[0024] The mean square error of the j-th monitoring sensor is calculated using the following formula : where is the observed value of the j-th monitoring sensor; is the j-th row of the observation matrix of the j-th monitoring sensor;

[0025] The weight factor of the j-th monitoring sensor is calculated using the following formula : where is a set first minimum constant used to prevent the denominator from being 0;

[0026] Finally, the weight factors of all monitoring sensors are normalized to ensure that the sum of all weight factors is 1.

[0027] The step S6 described above includes the following steps:

[0028] For each monitoring sensor, according to the corresponding state prediction value and prediction error covariance, calculate the Mahalanobis distance of the monitoring sensor;

[0029] Based on the obtained Mahalanobis distance, judge the monitoring sensor:

[0030] If the Mahalanobis distance is not greater than the set value, it is determined that the monitoring sensor is normal;

[0031] If the Mahalanobis distance is greater than the set value, it is determined that the monitoring sensor is abnormal, and the measurement noise covariance of the monitoring sensor is amplified to reduce the influence of the monitoring sensor on the fusion result.

[0032] The step S6 described above specifically includes the following steps:

[0033] According to the state prediction value and prediction error covariance, the Mahalanobis distance of the j-th monitoring sensor is calculated using the following formula : where is the current measurement noise covariance of the j-th monitoring sensor; is a set second minimum constant used to prevent the denominator from being 0;

[0034] Based on the obtained Mahalanobis distance, judge the j-th monitoring sensor:

[0035] If the Mahalanobis distance is not greater than the set value, it is determined that the j-th monitoring sensor is normal;

[0036] If the Mahalanobis distance is greater than the set value, it is determined that the j-th monitoring sensor is abnormal, and the current measurement noise covariance of the j-th monitoring sensor is amplified times and used as the current measurement noise covariance of the new j-th monitoring sensor; 。

[0037] The said step S7 specifically includes the following steps:

[0038] Calculate the Kalman gain of the j-th monitoring sensor by using the following formula: In the formula is the Kalman gain of the j-th monitoring sensor;

[0039] Based on the weighted summation scheme, calculate the total Kalman gain by using the following formula : In the formula, N is the total number of monitoring sensors.

[0040] The said step S8 specifically includes the following steps:

[0041] Based on the weighted summation scheme, calculate the state estimate at the current time step by using the following formula: In the formula is the state estimate at the current time step;

[0042] Calculate the error covariance at the current time step by using the following formula: In the formula is the error covariance at the current time step; S is the total matrix composed of the observation matrices of all monitoring sensors; R is the total matrix composed of the current measurement noise covariances of all monitoring sensors;

[0043] Take the obtained state estimate at the current time step as the data fusion result of the final same type of monitoring sensors.

[0044] The present invention also provides a system for implementing the data dynamic fusion method of the same type of monitoring sensors, including a data acquisition module, a data processing module, a matrix initialization module, a prediction variance calculation module, a weight calculation module, a measurement variance calculation module, a gain calculation module, and a data fusion module; the data acquisition module, the data processing module, the matrix initialization module, the prediction variance calculation module, the weight calculation module, the measurement variance calculation module, the gain calculation module, and the data fusion module are connected in series in sequence; the data acquisition module is used to acquire the monitoring data of the same type of monitoring sensors to be fused, and upload the data information to the data processing module; the data processing module is used to perform data cleaning and abnormal data processing on the acquired monitoring data based on the sparse signal recovery method, regularization constraint, and linear interpolation method according to the received data information, so as to obtain continuous monitoring data, and upload the data information to the matrix initialization module; the matrix initialization module is used to initialize the basic matrices for state prediction and error correction according to the received data information, and the basic matrices include a state transition matrix, a process noise covariance matrix, an observation matrix, and a measurement noise covariance matrix, and upload the data information to the prediction variance calculation module; the prediction variance calculation module is used to predict the state of the current time step using the Kalman filter prediction model according to the received data information, and calculate the prediction error covariance, and upload the data information to the weight calculation module; the weight calculation module is used to calculate the mean square error between the observed value and the predicted value of each monitoring sensor according to the received data information, and calculate the corresponding weight factor according to the obtained mean square error, and upload the data information to the measurement variance calculation module; the measurement variance calculation module is used to determine the fault state of each monitoring sensor based on the state prediction value and the prediction error covariance of each monitoring sensor according to the received data information, and adjust the corresponding measurement noise covariance, and upload the data information to the gain calculation module; the gain calculation module is used to calculate the Kalman gain of each monitoring sensor according to the measurement noise covariance and the weight factor of each monitoring sensor according to the received data information, and perform weighted summation for combination to obtain the total Kalman gain, and upload the data information to the data fusion module; the data fusion module is used to complete the data dynamic fusion of the same type of monitoring sensors according to the received data information and the obtained total Kalman gain by using a weighted summation scheme.

[0045] The data dynamic fusion method and system of the same type of monitoring sensors of the present invention, by processing the data of the same type of monitoring sensors, adaptively adjusting the weighting factors of the monitoring sensors based on the sparse signal recovery and dynamic weighted fusion algorithm of the Kalman filter, and isolating the abnormal monitoring sensors by adopting a fault diagnosis scheme, not only realizes the data dynamic fusion of the same type of monitoring sensors, but also has higher reliability and better accuracy. Brief Description of the Drawings

[0046] Figure 1 It is a schematic diagram of the method flow of the method of the present invention.

[0047] Figure 2 It is a schematic diagram of the functional modules of the system of the present invention. Detailed implementation manners

[0048] As Figure 1 shown in the following is the schematic diagram of the method flow of the method of the present invention: The method for dynamic data fusion of the same type of monitoring sensors disclosed in the present invention includes the following steps:

[0049] S1. Obtain the monitoring data of the same type of monitoring sensors to be fused;

[0050] In specific implementation, the data of multiple monitoring sensors can be stored in a file according to time steps, so as to facilitate data fusion and time series analysis; the file structure is designed as follows: The first column of the file is the time step (Time Step), which is used to identify the acquisition time of each row of monitoring data; the subsequent columns are arranged in descending order according to the number (or data quality) of data collected by each monitoring sensor; the column of the sensor with the largest amount of observed data or the highest data quality is arranged in the second column and is set as the reference sensor.

[0051] S2. Based on the sparse signal recovery method, regularization constraint and linear interpolation method, perform data cleaning and abnormal data processing on the monitoring data obtained in step S1 to obtain continuous monitoring data; specifically, it includes the following steps:

[0052] For the data of each time step of each monitoring sensor, use the sparse signal recovery method to process the missing values, and realize data sparse recovery through L1 regularization constraint;

[0053] After sparse signal recovery, if there are still missing data, use the linear interpolation method to interpolate the missing data;

[0054] Finally, obtain continuous monitoring data.

[0055] S3. Initialize the basic matrices for state prediction and error correction; the basic matrices include the state transition matrix, process noise covariance matrix, observation matrix and measurement noise covariance matrix.

[0056] S4. According to the data information obtained in step S2 and step S3, use the Kalman filter prediction model to predict the state of the current time step and calculate the prediction error covariance; specifically, it includes the following steps:

[0057] Use the following formula for state prediction: In the formula is the predicted state at time step k; is the predicted state at time step k-1; is the current state transition matrix;

[0058] The error covariance is predicted using the following equation: where is the predicted error covariance matrix at time step k; is the predicted error covariance matrix at time step k-1; Q is the current process noise covariance matrix.

[0059] S5. For each monitoring sensor, calculate the mean square error between the observed value and the predicted value of the monitoring sensor, and calculate the corresponding weight factor based on the obtained mean square error; specifically, it includes the following steps:

[0060] The mean square error of the j-th monitoring sensor is calculated using the following equation : where is the observed value of the j-th monitoring sensor; is the j-th row of the observation matrix of the j-th monitoring sensor;

[0061] The weight factor of the j-th monitoring sensor is calculated using the following equation : where is a set first minimum constant used to prevent the denominator from being zero; the calculation method of the weight factor can make the sensor with smaller error obtain a higher weight to dynamically reflect the change of data quality;

[0062] Finally, normalize the weight factors of all monitoring sensors to ensure that the sum of all weight factors is 1.

[0063] S6. Based on the state prediction value and the predicted error covariance of each monitoring sensor, determine the fault state of the monitoring sensor and adjust the corresponding measurement noise covariance; it includes the following steps:

[0064] For each monitoring sensor, calculate the Mahalanobis distance of the monitoring sensor according to the corresponding state prediction value and predicted error covariance.

[0065] Judge the monitoring sensor according to the obtained Mahalanobis distance:

[0066] If the Mahalanobis distance is not greater than the set value, it is determined that the monitoring sensor is normal;

[0067] If the Mahalanobis distance is greater than the set value, it is determined that the monitoring sensor is abnormal, and the measurement noise covariance of the monitoring sensor is amplified to reduce the influence of the monitoring sensor on the fusion result;

[0068] During specific implementation: According to the state prediction value and the prediction error covariance, the Mahalanobis distance of the j-th monitoring sensor is calculated using the following formula : In the formula is the current measurement noise covariance of the j-th monitoring sensor; is the set second minimum constant, used to prevent the denominator from being 0;

[0069] Based on the obtained Mahalanobis distance, the j-th monitoring sensor is judged:

[0070] If the Mahalanobis distance is not greater than the set value, it is determined that the j-th monitoring sensor is normal;

[0071] If the Mahalanobis distance is greater than the set value, it is determined that the j-th monitoring sensor is abnormal, and the current measurement noise covariance of the j-th monitoring sensor is amplified times and used as the current measurement noise covariance of the new j-th monitoring sensor; .

[0072] S7. According to the measurement noise covariance and weight factor of each monitoring sensor, calculate the Kalman gain of each monitoring sensor, and use weighted summation for merging to obtain the total Kalman gain; specifically, it includes the following steps:

[0073] The Kalman gain of the j-th monitoring sensor is calculated using the following formula: In the formula is the Kalman gain of the j-th monitoring sensor;

[0074] Based on the weighted summation scheme, the total Kalman gain is calculated using the following formula : where N is the total number of monitoring sensors.

[0075] S8. According to the total Kalman gain obtained in step S7, complete the data dynamic fusion of similar monitoring sensors using the weighted summation scheme; specifically, it includes the following steps:

[0076] Based on the weighted summation scheme, the state estimate at the current time step is calculated using the following formula: In the formula is the state estimate at the current time step;

[0077] The error covariance at the current time step is calculated using the following formula: In the formula is the error covariance at the current time step; S is the total matrix composed of the observation matrices of all monitoring sensors; R is the total matrix composed of the current measurement noise covariances of all monitoring sensors;

[0078] The state estimation at the current time step obtained is used as the data fusion result of the final homogeneous monitoring sensors.

[0079] The following combines a comparative example to illustrate the effect of the method of the present invention:

[0080] Select the monitoring data collected during the shield tunnel construction process for experimental comparison. Part of the data set is shown in Table 1;

[0081] Table 1 Schematic table of sensor monitoring data during the shield process

[0082]

[0083] Compare the method of the present invention with the existing traditional weighted fusion method, Bayesian filtering method and isolation forest method. The final evaluation indexes obtained are shown in Table 2:

[0084] Table 2 Schematic table of evaluation indexes

[0085]

[0086] As can be seen from Table 2, the method of the present invention shows obvious advantages in all evaluation indexes. In terms of accuracy, the method of the present invention reaches 93%, which is much higher than other methods, proving that its accuracy after data fusion is higher. In terms of processing efficiency, the processing time of the method of the present invention is 102 seconds, which is significantly shorter than other methods and has higher efficiency. In terms of reliability, the method of the present invention also performs excellently, reaching 91%, which is significantly better than the weighted method (70%), Bayesian filtering (78%) and isolation forest (72%), showing its greater stability and fault tolerance when facing data loss or anomalies. In terms of data recovery ability, the recovery ability of the method of the present invention is 96%, which is higher than 74%-80% of other methods, indicating that it performs the best in repairing missing or abnormal data. To sum up, the method of the present invention is superior to the existing traditional methods in multiple dimensions such as accuracy, efficiency, reliability and data recovery ability, demonstrating its significant advantages in practical applications.

[0087] Such as Figure 2The following is a schematic diagram of the functional modules of the system of the present invention: The system for implementing the data dynamic fusion method of the same type of monitoring sensors disclosed in the present invention includes a data acquisition module, a data processing module, a matrix initialization module, a prediction variance calculation module, a weight calculation module, a measurement variance calculation module, a gain calculation module, and a data fusion module; the data acquisition module, the data processing module, the matrix initialization module, the prediction variance calculation module, the weight calculation module, the measurement variance calculation module, the gain calculation module, and the data fusion module are connected in series in sequence; the data acquisition module is used to acquire the monitoring data of the same type of monitoring sensors to be fused and upload the data information to the data processing module; the data processing module is used to perform data cleaning and abnormal data processing on the acquired monitoring data based on the sparse signal recovery method, regularization constraint, and linear interpolation method according to the received data information to obtain continuous monitoring data and upload the data information to the matrix initialization module; the matrix initialization module is used to initialize the basic matrices for state prediction and error correction according to the received data information, and the basic matrices include a state transition matrix, a process noise covariance matrix, an observation matrix, and a measurement noise covariance matrix, and upload the data information to the prediction variance calculation module; the prediction variance calculation module is used to predict the state of the current time step using the Kalman filter prediction model according to the received data information, calculate the prediction error covariance, and upload the data information to the weight calculation module; the weight calculation module is used to calculate the mean square error between the observed value and the predicted value of each monitoring sensor according to the received data information, and calculate the corresponding weight factor according to the obtained mean square error, and upload the data information to the measurement variance calculation module; the measurement variance calculation module is used to determine the fault state of each monitoring sensor based on the state prediction value and prediction error covariance of each monitoring sensor according to the received data information, adjust the corresponding measurement noise covariance, and upload the data information to the gain calculation module; the gain calculation module is used to calculate the Kalman gain of each monitoring sensor according to the measurement noise covariance and weight factor of each monitoring sensor according to the received data information, and perform weighted summation for merging to obtain the total Kalman gain, and upload the data information to the data fusion module; the data fusion module is used to complete the data dynamic fusion of the same type of monitoring sensors according to the received data information and the obtained total Kalman gain using a weighted summation scheme.

Claims

1. A method for dynamically fusing data of homogeneous monitoring sensors, characterized in that It includes the following steps: S1. Obtain the monitoring data of homogeneous monitoring sensors to be fused; S2. Based on the sparse signal recovery method, regularization constraint, and linear interpolation method, perform data cleaning and abnormal data processing on the monitoring data obtained in step S1 to obtain continuous monitoring data; S3. Initialize the basic matrices for state prediction and error correction; the basic matrices include the state transition matrix, process noise covariance matrix, observation matrix, and measurement noise covariance matrix; S4. According to the data information obtained in steps S2 and S3, use the Kalman filter prediction model to predict the state at the current time step and calculate the prediction error covariance; specifically, it includes the following steps: The state prediction is carried out using the following formula: In the formula is the predicted state at time step k; is the predicted state at time step k-1; is the current state transition matrix; The error covariance prediction is carried out using the following formula: In the formula is the predicted error covariance matrix at time step k; is the predicted error covariance matrix at time step k-1; Q is the current process noise covariance matrix; S5. For each monitoring sensor, calculate the mean square error between the observed value and the predicted value of the monitoring sensor, and calculate the corresponding weight factor according to the obtained mean square error; specifically, it includes the following steps: The mean square error of the j-th monitoring sensor is calculated using the following formula : In the formula is the observed value of the j-th monitoring sensor; is the j-th row of the observation matrix of the j-th monitoring sensor; The weight factor of the j-th monitoring sensor is calculated using the following formula : where is a set first minimum constant used to prevent the denominator from being zero; Finally, normalize the weight factors of all monitoring sensors to ensure that the sum of all weight factors is 1; S6. Based on the state prediction value and prediction error covariance of each monitoring sensor, determine the fault state of the monitoring sensor and adjust the corresponding measurement noise covariance; S7. According to the measurement noise covariance and weight factor of each monitoring sensor, calculate the Kalman gain of each monitoring sensor, and use weighted summation for combination to obtain the total Kalman gain; S8. According to the total Kalman gain obtained in step S7, use the weighted summation scheme to complete the data dynamic fusion of homogeneous monitoring sensors.

2. The data dynamic fusion method of the same kind of monitoring sensors according to claim 1, characterized in that The specific steps of step S2 are as follows: For the data of each time step of each monitoring sensor, use the sparse signal recovery method to process the missing values, and realize data sparse recovery through L1 regularization constraint; After sparse signal recovery, if there are still missing data, use the linear interpolation method to interpolate the missing data; Finally, obtain continuous monitoring data.

3. The method for dynamically fusing data of homogeneous monitoring sensors according to claim 2, characterized in that The steps of step S6 include the following steps: For each monitoring sensor, calculate the Mahalanobis distance of the monitoring sensor according to the corresponding state prediction value and prediction error covariance; Judge the monitoring sensor according to the obtained Mahalanobis distance: If the Mahalanobis distance is not greater than the set value, it is determined that the monitoring sensor is normal; If the Mahalanobis distance is greater than the set value, it is determined that the monitoring sensor is abnormal, and the measurement noise covariance of the monitoring sensor is amplified to reduce the influence of the monitoring sensor on the fusion result.

4. The data dynamic fusion method of the same type of monitoring sensors according to claim 3, characterized in that The specific steps of step S6 are as follows: According to the state prediction value and the prediction error covariance, the Mahalanobis distance of the j-th monitoring sensor is calculated using the following formula : In the formula is the current measurement noise covariance of the j-th monitoring sensor; is the set second minimum constant, which is used to prevent the denominator from being zero; Judge the j-th monitoring sensor according to the obtained Mahalanobis distance: If the Mahalanobis distance is not greater than the set value, it is determined that the j-th monitoring sensor is normal; If the Mahalanobis distance is greater than the set value, it is determined that the j-th monitoring sensor is abnormal, and the current measurement noise covariance of the j-th monitoring sensor is amplified by a factor, and this is used as the new current measurement noise covariance of the j-th monitoring sensor; .

5. The data dynamic fusion method of the same kind of monitoring sensors according to claim 4, characterized in that The specific steps of step S7 are as follows: Use the following formula to calculate the Kalman gain of the j-th monitoring sensor: In the formula is the Kalman gain of the j-th monitoring sensor; Based on the weighted summation scheme, the total Kalman gain is calculated using the following equation : where N is the total number of monitoring sensors.

6. The data dynamic fusion method of the same type of monitoring sensors according to claim 5, characterized in that The specific steps of step S8 are as follows: Based on the weighted summation scheme, the state estimate at the current time step is calculated using the following equation: where is the state estimate at the current time step; The error covariance at the current time step is calculated using the following equation: where is the error covariance at the current time step; S is the total matrix composed of the observation matrices of all monitoring sensors; R is the total matrix composed of the current measurement noise covariances of all monitoring sensors; The state estimate at the current time step obtained is used as the data fusion result of the final homogeneous monitoring sensors.

7. A system for implementing the data dynamic fusion method of the same type of monitoring sensors described in any one of claims 1 to 6, characterized in that It includes a data acquisition module, a data processing module, a matrix initialization module, a predicted variance calculation module, a weight calculation module, a measurement variance calculation module, a gain calculation module, and a data fusion module; the data acquisition module, the data processing module, the matrix initialization module, the predicted variance calculation module, the weight calculation module, the measurement variance calculation module, the gain calculation module, and the data fusion module are connected in series in sequence; the data acquisition module is used to acquire the monitoring data of the same type of monitoring sensors to be fused and upload the data information to the data processing module; The data processing module is used to perform data cleaning and abnormal data processing on the acquired monitoring data based on the sparse signal recovery method, regularization constraint, and linear interpolation method according to the received data information to obtain continuous monitoring data, and upload the data information to the matrix initialization module; The matrix initialization module is used to initialize the basic matrices for state prediction and error correction according to the received data information. The basic matrices include a state transition matrix, a process noise covariance matrix, an observation matrix, and a measurement noise covariance matrix, and upload the data information to the predicted variance calculation module; The predicted variance calculation module is used to predict the state at the current time step using the Kalman filter prediction model according to the received data information, calculate the predicted error covariance, and upload the data information to the weight calculation module; The weight calculation module is used to calculate the mean square error between the observed value and the predicted value of each monitoring sensor according to the received data information, and calculate the corresponding weight factor according to the obtained mean square error, and upload the data information to the measurement variance calculation module; The measurement variance calculation module is used to determine the fault state of each monitoring sensor based on the state predicted value and the predicted error covariance of each monitoring sensor according to the received data information, adjust the corresponding measurement noise covariance, and upload the data information to the gain calculation module; the gain calculation module is used to calculate the Kalman gain of each monitoring sensor according to the received data information, based on the measurement noise covariance and the weight factor of each monitoring sensor, and perform weighted summation for merging to obtain the total Kalman gain, and upload the data information to the data fusion module; The data fusion module is used to complete the data dynamic fusion of the same type of monitoring sensors according to the received data information, based on the obtained total Kalman gain, using the weighted summation scheme.

Images