A multi-sensor data credibility assessment method and system

Abstract

The application provides a kind of multi-sensor data credibility evaluation method and system, the technical solution of the application is, based on sensor historical monitoring data, construct the slow time-varying true value estimation model of introducing weighted standard deviation, calculate the time correlation credibility of current measurement value, and determine its proportion weight;Based on the monitoring data of multi-source heterogeneous associated sensor, construct the spatial correlation model based on multi-sensor fluctuation consistency, calculate the spatial correlation credibility of current measurement value, and determine its proportion weight;According to the time and space correlation credibility and its proportion weight, the comprehensive credibility of the measurement data is obtained by linear weighted fusion calculation.This application can solve the technical problems of single dimension, weak anti-interference ability and complex calculation in the prior art.

Description

Technical Field

[0001] This invention relates to the field of data science technology, specifically to a method and system for assessing the reliability of multi-sensor data. Background Technology

[0002] Infrastructure structural health monitoring systems typically embed various types of monitoring sensors at key risk points in the structure to monitor its operational status in real time and ensure the safe operation of the infrastructure. These monitoring parameters collectively reflect the performance and structural health status of the infrastructure under external loads.

[0003] However, in practical engineering applications of structural health monitoring systems, sensors operate in complex environments for extended periods, facing various interference factors, including node failures, node aging, unstable power supply, accidental human contact, and environmental vibration. These interference factors can introduce noise, produce outliers, or cause slow drift in the raw monitoring data, thus reducing the reliability of the sensor data. Without effective data reliability assessment, the monitoring system is highly susceptible to false alarms or missed alarms after a period of time, severely reducing system reliability and operational efficiency.

[0004] Existing data credibility assessment methods have the following main drawbacks: Existing algorithms rely on the sensor's own historical data for time-series analysis or only on spatial consistency comparison with neighboring sensors, but lack the fusion of spatiotemporal dimensions. For example, relying solely on its own historical data makes it difficult to distinguish between real structural abrupt changes and the sensor's own step faults, while relying solely on spatial comparison easily ignores the impact of the sensor's slow time-varying drift characteristics, reducing the accuracy of credibility assessment.

[0005] Evaluation methods based on statistical values ​​are often sensitive to historical outliers. If there are outliers within the historical window, they can severely skew the statistical parameters, leading to inaccurate subsequent evaluations.

[0006] Many evaluation algorithms based on deep learning or complex probabilistic models require large sample databases and expensive computing resources, making it difficult to run in real time at edge acquisition terminals with limited computing power, and thus failing to meet the timeliness requirements of monitoring and early warning.

[0007] Therefore, designing a multi-sensor data reliability assessment method that can integrate time and space dimensions, is robust to historical gross errors, and has an appropriate computational cost is an urgent problem to be solved in this field. Summary of the Invention

[0008] To address the shortcomings of existing technologies, this invention proposes a multi-sensor data reliability assessment method and system, solving the technical problems of single reliability estimation dimension, weak anti-interference capability, and computational complexity in existing technologies. This solution is applicable to engineering scenarios where structural monitoring data exhibits slow time-varying characteristics.

[0009] The technical solution adopted in this invention is as follows: First, based on historical monitoring data from sensors, a slow time-varying true value estimation model incorporating weighted standard deviation is constructed to calculate the temporal correlation reliability of the current measurement value and determine its proportion weight. Further, based on monitoring data from multi-source heterogeneous correlated sensors, a spatial correlation model is constructed. The current measurement characteristics are estimated by quantifying correlation weights, and the spatial correlation reliability of the current measurement value is calculated and its proportion weight is determined. Finally, based on the temporal and spatial correlation reliability and their proportion weights, the comprehensive reliability of the measurement data is obtained through linear weighted fusion calculation.

[0010] In the first feasible approach, a method for evaluating the reliability of multi-sensor data is provided, comprising: A true value estimation model based on slow time-varying statistical parameters is constructed, and the reliability of the temporal correlation of the sensor to be evaluated is calculated using historical monitoring data of the sensor to be evaluated. A spatial correlation model based on fluctuation consistency is constructed, and the spatial correlation confidence of the sensor to be evaluated is calculated using the monitoring data of the correlated sensors. Based on the aforementioned temporal correlation reliability and spatial correlation reliability, the overall reliability of the current measurement data is calculated through linear weighted fusion: in, Indicates overall credibility. Indicates the credibility of time correlation. Indicates the reliability of spatial association. The weighting indicates the proportion of credibility of time-related relationships. The weighting represents the proportion of spatial correlation credibility.

[0011] Furthermore, the weighting of the temporal correlation credibility and the weighting of the spatial correlation credibility... satisfy: in, The weighting indicates the proportion of credibility of time-related relationships. The weighting indicates the proportion of spatial correlation credibility. The formula for calculating the weighting of the time correlation reliability is as follows: in, Indicates the number of spatially correlated sensors. Represents reliability metrics over time; The formula for calculating the time-dimensional reliability index is as follows: in, Indicates the sample length. Indicates sample The measured value at time, This represents the mean of the sample data.

[0012] Furthermore, a true value estimation model based on slow time-varying statistical parameters is constructed, and the reliability of the temporal correlation of the sensor under evaluation is calculated using historical monitoring data. Specific steps include: Obtain the standard deviation of the rate of change of historical monitoring data as a slow time-varying characteristic parameter. The specific steps include: Obtain the rate of change of historical monitoring data N periods prior to the current time t; Based on the rate of change of the historical monitoring data, calculate the standard deviation of the rate of change of the historical monitoring data for the previous N periods: in, Indicates the first Standard deviation of time This represents the rate of change of historical monitoring data at time k. The average rate of change of historical monitoring data; Based on the change in the historical monitoring data at time t, determine the sign of the change direction: in, Symbols indicating the direction of change This represents the change in historical monitoring data at time t; The slow time-varying characteristic parameters from the previous time step are substituted into the truth estimation model to calculate the truth estimate for the current time t. The truth estimation model is expressed as: in, This represents the true estimate at the current moment. This represents the sensor monitoring data from the previous moment. Indicates the weighting coefficient. This represents the standard deviation of the rate of change over the previous N periods at the current moment; Calculate the deviation between the current true estimate and the sensor monitoring data, and calculate the reliability of the temporal correlation of the current measurement value according to the following mapping function: in, Indicates the credibility of time correlation. This represents the deviation between the current true estimate and the sensor monitoring data. This indicates the preset control threshold. This represents the adjustment factor, with a value range of [1,2]. This represents the maximum statistical value of the prediction deviation of sample data under normal sensor operation.

[0013] Furthermore, the weighting coefficients The specific steps involved in minimizing the estimation error of the sample data include: Based on a period of historical monitoring data, the goal is to minimize the difference between the estimated and actual values. Using the root mean square error as the objective, the optimal weighting coefficients are determined in the range of [0.5, 2.0] through a grid search method.

[0014] Furthermore, it also includes steps for correcting anomalous data: When the difference between the sensor monitoring data at the current time t and the true value estimate exceeds the control threshold, the true value estimate at the next time t+1 is calculated by substituting the true value estimate at the current time as the reference value at the previous time into the true value estimation model.

[0015] Furthermore, a spatial correlation model based on fluctuation consistency is constructed, and the spatial correlation reliability of the sensor to be evaluated is calculated using monitoring data from the correlated sensors. Specific steps include: Select m sensor types for confidence level correlation calculation, and for the i-th type, select n associated sensors. Combining the sensor type influence weight and the individual sensor correlation weight within the same type, calculate the normalized correlation influence weight of each associated sensor according to the following formula: in, This represents the normalized correlation influence weight of the j-th associated sensor in the i-th type. This indicates that the sensor type affects the weight. Represents the correlation weight of individual sensors. This represents the number of sensors of the i-th type; Establish a reliability evaluation model for spatially correlated sensors and calculate the reliability of a single spatially correlated sensor; The spatial correlation confidence scores of the associated sensors are weighted and fused to obtain the spatial correlation confidence scores of the sensors to be evaluated. in, This indicates the spatial correlation confidence level of the sensor to be evaluated. This represents the normalized correlation influence weight of the j-th associated sensor in the i-th type. This represents the confidence assessment value of the j-th associated sensor in the i-th type to the target sensor.

[0016] Furthermore, a reliability evaluation model for spatially correlated sensors is established, and the reliability of individual spatially correlated sensors is calculated. Specific steps include: For the j-th spatially correlated sensor of the i-th type With respect to the sensor S to be evaluated, calculate the confidence assessment value of the associated sensor to the target sensor at the current time t according to the following steps. : Calculate the correlation of the sensors according to the following relative volatility calculation formula. The relative volatility of the sensor S to be evaluated at time t , : in, Represents relative volatility. This represents the sensor's measurement value at time t. This represents the sensor's measurement value at time t-1. This represents the maximum statistical value of the volatility of historical sample data under normal sensor operating conditions. Using a fluctuation quantization function, the associated sensors are... The relative volatility of the sensor S to be evaluated is mapped to a quantified value of volatility within the range [0,1]. , ; Construct a volatility consistency mapping function based on the quantized value of the volatility level. , The spatial correlation confidence level of a single sensor is calculated: in, Indicates the reliability of spatial correlation of a single sensor. This represents the quantified value of the fluctuation level of the sensor to be evaluated. This represents a quantified value indicating the degree of fluctuation of the associated sensor.

[0017] Furthermore, the volatility quantification function is defined as follows: For any sensor's relative volatility at time t, the quantification value of the volatility is calculated using the following formula: in, This represents a quantitative value indicating the degree of volatility. This represents the fluctuation threshold of the sensor, with a value range of [1,2]. This represents the relative volatility of any sensor at time t.

[0018] Furthermore, the steps for obtaining the sensor type influence weight and the correlation weight of individual sensors of the same type include: Based on spatial distance and structural influence intensity indices, quantitative evaluation and scoring are performed on m associated sensor types to obtain a type score for each sensor type. The sensor type influence weight is then calculated using the following formula: in, This indicates that the sensor type affects the weight. This indicates the type score for each type of sensor; For the i-th type of sensor, based on spatial distance and structural influence intensity indicators, the n-th sensor of this type is quantitatively evaluated and scored to obtain an individual score for each sensor. The individual sensor correlation weight is then calculated using the following formula after normalization: in, Represents the correlation weight of individual sensors. This represents the individual score for each sensor.

[0019] In conjunction with the first feasible method, a second feasible method provides a multi-source sensor data reliability assessment system, characterized by comprising: The time assessment module is used to build a true value estimation model based on slow time-varying statistical parameters and to calculate the reliability of the time correlation of the sensor under assessment using historical monitoring data of the sensor under assessment. The spatial assessment module is used to construct a spatial correlation model based on multi-source heterogeneous sensors and to calculate the spatial correlation reliability of the sensors to be evaluated using the monitoring data of the correlated sensors. The decision fusion module is used to calculate the overall credibility of the current measurement data through linear weighted fusion based on the aforementioned temporal correlation credibility and spatial correlation credibility. in, Indicates overall credibility. Indicates the credibility of time correlation. Indicates the reliability of spatial association. The weighting indicates the proportion of credibility of time-related relationships. The weighting represents the proportion of spatial correlation credibility.

[0020] As can be seen from the above technical solution, the beneficial technical effects of the present invention are as follows: 1. In time-related assessment, this invention introduces a weighted standard deviation calculation method based on historical reliability. Unlike traditional statistical methods, this invention can automatically reduce the weight of outliers in historical data, thereby maintaining the stability of true value prediction even in noisy environments.

[0021] 2. When calculating spatial correlation, this invention does not simply rely on physical distance, but comprehensively considers spatial distance, structural influence intensity, and differences in sensor types. This hierarchical weighted method fully respects the force transmission laws and physical correlation mechanisms of engineering structures, making the collaborative evaluation between heterogeneous sensors more accurate.

[0022] 3. Compared with evaluation methods that rely on massive sample libraries and complex deep learning models, this invention does not require large-scale database support. It only needs to use historical data and related data within a sliding window for statistical calculations, which reduces data transmission and processing latency and can meet the real-time online evaluation needs of low-computing-power edge computing devices. Attached Figure Description

[0023] To more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the accompanying drawings used in the description of the specific embodiments or the prior art will be briefly introduced below. In all the drawings, similar elements or parts are generally identified by similar reference numerals. In the drawings, the elements or parts are not necessarily drawn to scale.

[0024] Figure 1 This is a flowchart of the method in Embodiment 1 of the present invention; Figure 2 This is a true value evaluation diagram of the temperature sensor in Embodiment 1 of the present invention; Figure 3 This is a correlation quantification diagram of the stress sensor in Embodiment 1 of the present invention; Figure 4 This is a reliability evaluation diagram of the strain sensor in Embodiment 1 of the present invention; Figure 5 This is a system flowchart of Embodiment 2 of the present invention; Figure label: 21-A multi-sensor data reliability assessment system, 21-Time assessment module, 22-Spatial assessment module, 22-Decision fusion module. Detailed Implementation

[0025] The embodiments of the technical solution of the present invention will now be described in detail with reference to the accompanying drawings. These embodiments are merely illustrative of the technical solution of the present invention and are therefore intended to limit the scope of protection of the present invention.

[0026] It should be noted that, unless otherwise stated, the technical or scientific terms used in this application should have the ordinary meaning as understood by one of ordinary skill in the art to which this invention pertains.

[0027] This embodiment provides a method for evaluating the reliability of multi-sensor data. The working principle of Embodiment 1 is explained in detail below: The method flowchart of this embodiment is as follows: Figure 1 As shown, the specific steps include: A true value estimation model based on slow time-varying statistical parameters is constructed, and the reliability of the temporal correlation of the sensor under evaluation is calculated using historical monitoring data of the sensor under evaluation. ; A spatial correlation model based on multi-source heterogeneous sensors is constructed, and the spatial correlation reliability of the sensors to be evaluated is calculated using the monitoring data of the correlated sensors. ; Based on the aforementioned temporal correlation reliability and spatial correlation reliability, the overall reliability of the current measurement data is calculated through linear weighted fusion: in, Indicates overall credibility. Indicates the credibility of time correlation. Indicates the reliability of spatial association. The weighting indicates the proportion of credibility of time-related relationships. The weighting represents the proportion of spatial correlation credibility.

[0028] In this embodiment, further, the weighting of the temporal correlation credibility and the weighting of the spatial correlation credibility are... satisfy: in, The weighting indicates the proportion of credibility of time-related relationships. The weighting indicates the proportion of spatial correlation credibility. The formula for calculating the weighting of the time correlation reliability is as follows: in, Indicates the number of spatially correlated sensors. Represents the reliability index over a time dimension (calculated and determined based on the autocorrelation coefficient of sample data over a period of time). The formula for calculating the time-dimensional reliability index is as follows: in, Indicates the sample length. Indicates sample The measured value at time, This represents the mean of the sample data.

[0029] In this embodiment, a true value estimation model based on slow time-varying statistical parameters is further constructed, and the reliability of the time correlation of the sensor to be evaluated is calculated using historical monitoring data of the sensor to be evaluated. The specific steps include: Obtain the standard deviation of the rate of change of historical monitoring data as a slow time-varying characteristic parameter. The specific steps include: To obtain the rate of change of historical monitoring data N periods prior to the current time t, the specific steps include: Let the sensor monitoring data at the current moment be... Based on the previous time t Periodic measurement data { Calculate the rate of change of data between adjacent time points: in, Let k represent the sensor monitoring data at time k, where k = t - N + 1, ..., t - 1, thus obtaining a sequence containing N - 1 rate of change values.

[0030] Based on the rate of change of the historical monitoring data, calculate the standard deviation of the rate of change of the historical monitoring data for the previous N periods: in, Indicates the first Standard deviation of time This represents the rate of change of historical monitoring data at time k. The average rate of change of historical monitoring data; Based on the changes in historical monitoring data at time t Determine the sign of the direction of change: in, Symbols indicating the direction of change This represents the change in historical monitoring data at time t; The slow time-varying characteristic parameters from the previous time step are substituted into the truth estimation model to calculate the truth estimate for the current time t. The truth estimation model is expressed as: in, This represents the true estimate at the current moment. This represents the sensor monitoring data from the previous moment. Indicates the weighting coefficient. This represents the standard deviation of the rate of change over the previous N periods at the current moment; Calculate the deviation between the true estimate and the sensor monitoring data at the current moment. And calculate the time correlation confidence level of the current measurement value according to the following mapping function: in, Indicates the credibility of time correlation. This represents the deviation between the current true value estimate and the sensor monitoring data. This indicates the preset control threshold. This represents the adjustment factor, with a value range of [1,2]. This represents the maximum statistical value of the prediction deviation of sample data under normal sensor operation.

[0031] The weighting coefficients The method involves minimizing the estimation error of the sample data, and the specific steps include: Based on a period of historical monitoring data, the goal is to minimize the difference between the estimated and actual values. Using the root mean square error as the objective, the optimal weighting coefficients are determined in the range of [0.5, 2.0] through a grid search method.

[0032] In this embodiment, further, the weighting coefficient The specific steps involved in minimizing the estimation error of the sample data include: Based on a period of historical monitoring data, the goal is to minimize the difference between the estimated and actual values. Using the root mean square error as the objective, the optimal weighting coefficients are determined in the range of [0.5, 2.0] through a grid search method.

[0033] In this embodiment, a further step is included: a step for correcting abnormal data. When the difference between the sensor monitoring data at the current time t and the true value estimate exceeds the control threshold, the true value estimate at the next time t+1 is calculated by substituting the true value estimate at the current time as the reference value at the previous time into the true value estimation model.

[0034] In response to situations where sensor data is missing at multiple sampling times, due to the long time intervals, the data collected by the sensor at time t fluctuates significantly compared to the data collected at the previous time, but the sensor measurement data may still be correct. In this case, because the changes in slowly time-varying characteristic parameters during the data gap period cannot be obtained, it is impossible to estimate the true value of the monitoring data. This situation can be handled in two ways: For cases where the missing data is less than the set threshold, interpolation estimation is performed. For cases where missing values ​​are greater than or equal to the set threshold, the statistical calculation of data features is re-performed, and the true values ​​of the monitoring parameters are re-estimated.

[0035] The core of the above method lies in using the statistical characteristics of historical data (such as standard deviation σ) to estimate the characteristic changes. The slow time-varying characteristic parameter is the carrier for realizing this estimation, and its specific form is not limited to the rate of change. For different monitored physical quantities (such as displacement, pressure, and humidity), the characteristic parameter can also be a derived quantity after physical meaning transformation or standardization, as long as it satisfies the "slow time-varying" requirement and has statistically significant fluctuation characteristics within the historical window.

[0036] In this embodiment, a spatial correlation model based on multi-source heterogeneous sensors is further constructed, and the spatial correlation reliability of the sensor to be evaluated is calculated using the monitoring data of the correlated sensors. The specific steps include: Select m sensor types for confidence level correlation calculation, and for the i-th type, select n associated sensors. Combining the sensor type influence weight and the individual sensor correlation weight within the same type, calculate the normalized correlation influence weight of each associated sensor according to the following formula: in, This represents the normalized correlation influence weight of the j-th associated sensor in the i-th type. This indicates that the sensor type affects the weight. Represents the correlation weight of individual sensors. This represents the number of sensors of the i-th type; Establish a reliability evaluation model for spatially correlated sensors and calculate the reliability of a single spatially correlated sensor; The spatial correlation confidence scores of the associated sensors are weighted and fused to obtain the spatial correlation confidence scores of the sensors to be evaluated. in, This indicates the spatial correlation confidence level of the sensor to be evaluated. This represents the normalized correlation influence weight of the j-th associated sensor in the i-th type. This represents the confidence assessment value of the j-th associated sensor in the i-th type to the target sensor.

[0037] In this embodiment, a reliability evaluation model for spatial correlation sensors is further established, and the reliability of a single spatial correlation sensor is calculated. Specific steps include: For the j-th spatially correlated sensor of the i-th type With respect to the sensor S to be evaluated, calculate the confidence assessment value of the associated sensor to the target sensor at the current time t according to the following steps. : Calculate the correlation of the sensors according to the following relative volatility calculation formula. The relative volatility of the sensor S to be evaluated at time t , : in, Represents relative volatility. This represents the sensor's measurement value at time t. This represents the sensor's measurement value at time t-1. This represents the maximum statistical value of the volatility of historical sample data under normal sensor operating conditions. Using a fluctuation quantization function, the associated sensors are... The relative volatility of the sensor S to be evaluated is mapped to a quantified value of volatility within the range [0,1]. , ; Construct a volatility consistency mapping function based on the quantized value of the volatility level. , The spatial correlation confidence level of a single sensor is calculated: in, Indicates the reliability of spatial correlation of a single sensor. This represents the quantified value of the fluctuation level of the sensor to be evaluated. This represents a quantified value indicating the degree of fluctuation of the associated sensor.

[0038] The mapping satisfies the following condition: the confidence level is 1 when the fluctuation levels of the reference sensor and the target sensor are consistent, and the confidence level is lower when the difference is greater.

[0039] In this embodiment, the fluctuation quantification function is further defined as: For any sensor's relative volatility at time t, the quantification value of the volatility is calculated using the following formula: in, This represents a quantitative value indicating the degree of volatility. This represents the fluctuation threshold of the sensor, with a value range of [1,2]. The smaller the value, the more sensitive the reliability assessment. This represents the relative volatility of any sensor at time t.

[0040] In this embodiment, the steps for obtaining the sensor type influence weight and the correlation weight of individual sensors of the same type further include: Based on spatial distance and structural influence intensity indices, quantitative evaluation and scoring are performed on m associated sensor types to obtain a type score for each sensor type. The sensor type influence weight is then calculated using the following formula: in, This indicates that the sensor type affects the weight. This indicates the type score for each type of sensor; For the i-th type of sensor, based on spatial distance and structural influence intensity indicators, the n-th sensor of this type is quantitatively evaluated and scored to obtain an individual score for each sensor. The individual sensor correlation weight is then calculated using the following formula after normalization: in, Represents the correlation weight of individual sensors. This represents the individual score for each sensor.

[0041] To verify the effectiveness of the time correlation reliability assessment method in this embodiment, temperature monitoring data from a sensor to be evaluated was selected for verification. Weighting coefficients were set. Monitoring data from 10 consecutive time points were selected, and outlier data (outliers) were artificially introduced at time points 3 and 9. The evaluation results are as follows: Figure 2 As shown. At time 3, the measured value was 30.00℃, a drastic change compared to the previous time. Using the true value estimation model described in this invention, the estimated true value was 24.22℃. Since the deviation (5.78) far exceeded the control threshold, the algorithm determined that the data had low reliability.

[0042] It is worth noting that when calculating the estimated true value for the next time step (indication 4), this invention employs an anomaly iteration correction mechanism. That is, instead of using the anomalous 30.00℃ as the baseline, it uses the estimated true value of 24.22℃ from indication 3. The algorithm was used in the calculation. Test results show that the estimated true value of sequence number 4 was not affected by the gross error of the previous time step, proving that the algorithm has good robustness to historical gross errors.

[0043] Furthermore, in this embodiment, a stress sensor YL at a certain cross-section of the tunnel is used. 01For the object to be evaluated, surrounding temperature sensors, displacement sensors, and other stress sensors on the same cross-section were selected as correlation references. Based on the physical correlation mechanism, displacement and stress were determined to have the strongest correlation, assigned a "different types of correlation influence weight" of 35%; temperature was second, assigned 30%; within the same type of sensor, scores were given based on spatial distance. For example, temperature sensor T1, being closer, scored 80, while T2, being slightly farther away, scored 70; finally, the normalized correlation influence weight of each sensor was calculated using the normalization formula. ,like Figure 3 As shown.

[0044] Furthermore, based on the normalized correlation influence weight, the spatial correlation reliability of the No. 4 vibrating wire sensor at a certain tunnel section is evaluated. Relative volatility is selected as the characteristic parameter, and the evaluation results are as follows: Figure 4 As shown.

[0045] At times 1, 2, 5, and 6, the relative volatility of sensor 4 was close to that of its neighboring sensors, thus the reliability was rated at 100%. Further, to verify the consistency of the volatility assessment, abnormal volatility data at time 3 was inserted. The associated data from sensors 1, 2, and 3 fluctuated, while the data from target sensor 4 remained unchanged, decreasing the spatial correlation reliability of the sensors. After weighted fusion, the overall reliability was 0.39. Simultaneously, at time 4, the associated data from sensors 1, 2, and 3 recovered, while the data from target sensor 4 remained unchanged, resulting in an overall reliability of 0.31. At time 7, the data from target sensor 4 fluctuated, while the associated sensor data remained unchanged. After weighted fusion, the overall reliability was 0.0, indicating that the method accurately identified the anomaly at that time.

[0046] In this embodiment, by integrating the dual dimensions of temporal continuity and spatial consistency, and introducing weighted statistics and structured correlation mechanisms, accurate and real-time assessment of the reliability of sensor data is achieved.

[0047] Example 2 In conjunction with the method provided in Embodiment 1, Embodiment 2 provides a multi-sensor data reliability assessment system 21, the system structure of which is shown in the figure below. Figure 5 As shown, it includes: The time assessment module 22 is used to construct a true value estimation model based on slow time-varying statistical parameters and to calculate the time correlation confidence of the sensor under assessment using historical monitoring data of the sensor under assessment. The spatial assessment module 23 is used to construct a spatial correlation model based on multi-source heterogeneous sensors and to calculate the spatial correlation confidence of the sensor to be assessed using the monitoring data of the correlated sensors. Decision fusion module 24 is used to calculate the comprehensive reliability of the current measurement data through linear weighted fusion based on the temporal correlation reliability and spatial correlation reliability. in, Indicates overall credibility. Indicates the credibility of time correlation. Indicates the reliability of spatial association. The weighting indicates the proportion of credibility of time-related relationships. The weighting represents the proportion of spatial correlation credibility.

[0048] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention, and they should all be covered within the scope of the claims and specification of the present invention.

Claims

1. A method for evaluating the reliability of multi-sensor data, characterized in that, include: A true value estimation model based on slow time-varying statistical parameters is constructed, and the reliability of the temporal correlation of the sensor to be evaluated is calculated using historical monitoring data of the sensor to be evaluated. A spatial correlation model based on fluctuation consistency is constructed, and the spatial correlation confidence of the sensor to be evaluated is calculated using the monitoring data of the correlated sensors. Based on the aforementioned temporal correlation reliability and spatial correlation reliability, the overall reliability of the current measurement data is calculated through linear weighted fusion: in, Indicates overall credibility. Indicates the credibility of time correlation. Indicates the reliability of spatial association. The weighting indicates the proportion of credibility of time-related relationships. The weighting represents the proportion of spatial correlation credibility.

2. The method for evaluating the reliability of multi-sensor data according to claim 1, characterized in that, The weighting of the reliability of temporal correlation and the weighting of the reliability of spatial correlation. satisfy: in, The weighting indicates the proportion of credibility of time-related relationships. The weighting indicates the proportion of spatial correlation credibility. The formula for calculating the weighting of the time correlation reliability is as follows: in, Indicates the number of spatially correlated sensors. Represents reliability metrics over time; The formula for calculating the time-dimensional reliability index is as follows: in, Indicates the sample length. Indicates sample The measured value at time, This represents the mean of the sample data.

3. The method for evaluating the reliability of multi-sensor data according to claim 1, characterized in that, A true value estimation model based on slow time-varying statistical parameters is constructed, and the reliability of the time correlation of the sensor under evaluation is calculated using historical monitoring data. The specific steps include: Obtain the standard deviation of the rate of change of historical monitoring data as a slow time-varying characteristic parameter. The specific steps include: Obtain the rate of change of historical monitoring data N periods prior to the current time t; Based on the rate of change of the historical monitoring data, calculate the standard deviation of the rate of change of the historical monitoring data for the previous N periods: in, Indicates the first Standard deviation of time This represents the rate of change of historical monitoring data at time k. The average rate of change of historical monitoring data; Based on the change in the historical monitoring data at time t, determine the sign of the change direction: in, Symbols indicating the direction of change This represents the change in historical monitoring data at time t; The slow time-varying characteristic parameters of the previous time step Substitute the values ​​into the truth estimation model to calculate the truth estimate at the current time t. The truth estimation model is expressed as: in, This represents the true estimate at the current moment. This represents the sensor monitoring data from the previous moment. Indicates the weighting coefficient. This represents the standard deviation of the rate of change over the previous N periods at the current moment; Calculate the deviation between the current true estimate and the sensor monitoring data, and calculate the reliability of the temporal correlation of the current measurement value according to the following mapping function: in, Indicates the credibility of time correlation. This represents the deviation between the current true value estimate and the sensor monitoring data. This indicates the preset control threshold. This represents the adjustment factor, with a value range of [1,2]. This represents the maximum statistical value of the prediction deviation of sample data under normal sensor operation.

4. The method for evaluating the reliability of multi-sensor data according to claim 3, characterized in that, The weighting coefficients The specific steps involved in minimizing the estimation error of the sample data include: Based on a period of historical monitoring data, the goal is to minimize the difference between the estimated and actual values. Using the root mean square error as the objective, the optimal weighting coefficients are determined in the range of [0.5, 2.0] through a grid search method.

5. The method for evaluating the reliability of multi-sensor data according to claim 3, characterized in that, It also includes steps for correcting abnormal data: When the difference between the sensor monitoring data at the current time t and the true value estimate exceeds the control threshold, the true value estimate at the next time t+1 is calculated by substituting the true value estimate at the current time as the reference value at the previous time into the true value estimation model.

6. The method for evaluating the reliability of multi-sensor data according to claim 1, characterized in that, A spatial correlation model based on fluctuation consistency is constructed, and the spatial correlation reliability of the sensor to be evaluated is calculated using monitoring data from correlated sensors. The specific steps include: Select m sensor types for confidence level correlation calculation, and for the i-th type, select n associated sensors. Combining the sensor type influence weight and the individual sensor correlation weight within the same type, calculate the normalized correlation influence weight of each associated sensor according to the following formula: in, This represents the normalized correlation influence weight of the j-th associated sensor in the i-th type. This indicates that the sensor type affects the weight. Represents the correlation weight of individual sensors. This represents the number of sensors of the i-th type; Establish a reliability evaluation model for spatially correlated sensors and calculate the reliability of a single spatially correlated sensor; The spatial correlation confidence scores of the associated sensors are weighted and fused to obtain the spatial correlation confidence scores of the sensors to be evaluated. in, This indicates the spatial correlation confidence level of the sensor to be evaluated. This represents the normalized correlation influence weight of the j-th associated sensor in the i-th type. This represents the confidence assessment value of the j-th associated sensor in the i-th type to the target sensor.

7. The method for evaluating the reliability of multi-sensor data according to claim 6, characterized in that, Establish a reliability evaluation model for spatially correlated sensors and calculate the reliability of individual spatially correlated sensors. Specific steps include: For the j-th spatially correlated sensor of the i-th type With respect to the sensor S to be evaluated, calculate the confidence assessment value of the associated sensor to the target sensor at the current time t according to the following steps. : Calculate the correlation of the sensors according to the following relative volatility calculation formula. The relative volatility of the sensor S to be evaluated at time t , : in, Represents relative volatility. This represents the sensor's measurement value at time t. This represents the sensor's measurement value at time t-1. This represents the maximum statistical value of the volatility of historical sample data under normal sensor operating conditions. Using a fluctuation quantization function, the associated sensors are... The relative volatility of the sensor S to be evaluated is mapped to a quantified value of volatility within the range [0,1]. , ; Construct a volatility consistency mapping function based on the quantized value of the volatility level. , The spatial correlation confidence level of a single sensor is calculated: in, Indicates the reliability of spatial correlation of a single sensor. This represents the quantified value of the fluctuation level of the sensor to be evaluated. This represents a quantified value indicating the degree of fluctuation of the associated sensor.

8. The method for evaluating the reliability of multi-sensor data according to claim 7, characterized in that, The volatility quantification function is defined as follows: For any sensor's relative volatility at time t, the quantification value of the volatility is calculated using the following formula: in, This represents a quantitative value indicating the degree of volatility. This represents the fluctuation threshold of the sensor, with a value range of [1,2]. This represents the relative volatility of any sensor at time t.

9. The method for evaluating the reliability of multi-sensor data according to claim 6, characterized in that, The steps for obtaining the sensor type influence weight and the correlation weight of individual sensors of the same type include: Based on spatial distance and structural influence intensity indices, quantitative evaluation and scoring are performed on m associated sensor types to obtain a type score for each sensor type. The sensor type influence weight is then calculated using the following formula: in, This indicates that the sensor type affects the weight. This indicates the type score for each type of sensor; For the i-th type of sensor, based on spatial distance and structural influence intensity indicators, the n-th sensor of this type is quantitatively evaluated and scored to obtain an individual score for each sensor. The individual sensor correlation weight is then calculated using the following formula after normalization: in, Represents the correlation weight of individual sensors. This represents the individual score for each sensor.

10. A multi-sensor data reliability assessment system, characterized in that, include: The time assessment module is used to build a true value estimation model based on slow time-varying statistical parameters and to calculate the reliability of the time correlation of the sensor under assessment using historical monitoring data of the sensor under assessment. The spatial assessment module is used to construct a spatial correlation model based on multi-source heterogeneous sensors and to calculate the spatial correlation reliability of the sensors to be evaluated using the monitoring data of the correlated sensors. The decision fusion module is used to calculate the overall credibility of the current measurement data through linear weighted fusion based on the aforementioned temporal correlation credibility and spatial correlation credibility. in, Indicates overall credibility. Indicates the credibility of time correlation. Indicates the reliability of spatial association. The weighting indicates the proportion of credibility of time-related relationships. The weighting represents the proportion of spatial correlation credibility.

Images