Search

CN-121558086-B - Method for detecting state abnormality of nonspecific multivariable sensor based on error characteristics

CN121558086BCN 121558086 BCN121558086 BCN 121558086BCN-121558086-B

Abstract

An unspecified multivariable sensor state anomaly detection method based on error characteristics belongs to the technical field of intelligent sensing and control. The method comprises the steps of firstly collecting measurement data of a plurality of sensors in the same physical state at the same moment, establishing a multisource error boundary model, combining error distribution characteristics of the multisource error boundary model to construct a measurement error boundary containing noise and delay influences, then dynamically correcting noise deviation and delay deviation, carrying out translation and expansion on the error boundary of each sensor, and finally carrying out overlap analysis on the error boundary of all the sensors. The invention realizes the self-adaptive consistency detection and anomaly identification of the measuring result of the multi-source sensor by fusing the error characteristics of noise and delay, does not need to depend on specific physical variables, has strong model generalization, high instantaneity and good robustness, and can effectively improve the measuring reliability and operation safety of the system under the conditions of partial sensor failure, drift or attack.

Inventors

  • CHEN YANFENG
  • ZHANG JIAYI
  • WANG YAN
  • LIU HAOQIAN
  • ZHANG ZONGYE

Assignees

  • 辽宁大学

Dates

Publication Date
20260512
Application Date
20260122

Claims (6)

  1. 1. An unspecified multivariable sensor state anomaly detection method based on error characteristics is characterized by comprising the following steps: Step 1) multisource data acquisition and modeling, namely acquiring measurement data of a plurality of sensors on the same physical state at the same moment, establishing a multisource sensor measurement set, and establishing a multisource error boundary model based on statistical information of measurement noise and communication delay, wherein the multisource error boundary model is used for describing an uncertainty range of output of each sensor; Step 2) error boundary construction, namely on the premise that measurement noise and communication delay of each sensor are known, combining error distribution characteristics to determine measurement noise deviation and communication delay range, and forming a measurement error boundary containing the influence of the measurement noise and the communication delay, wherein the measurement error boundary is used for representing a preliminarily determined interval range of a real physical state; Step 3) error boundary correction and expansion, namely dynamically correcting the measurement noise deviation and the communication delay range according to the change rate of the sensors, translating and expanding the error boundary of each sensor by calculating the product of the change rate and the communication delay range, and obtaining a dynamic error boundary with time self-adaptability; Step 4) boundary overlap analysis, namely carrying out overlap calculation on dynamic error boundaries of all sensors to form a boundary intersection area, if the boundary of one sensor is not intersected with the overlapping range of a plurality of sensors, considering that the output data of the sensor is abnormal, and if the boundary of one sensor is intersected with the overlapping range of a plurality of sensors, considering that the output data of the sensor is normal; and 5) data fusion and state estimation, namely after eliminating the abnormal sensor, carrying out majority vote fusion on error boundaries of the residual sensor, and when a plurality of effective intersections exist in the overlapping area, taking the middle point of the overlapping area as an estimated value of the system state and recording corresponding reliability indexes.
  2. 2. The method for detecting abnormal states of the unspecified multivariable sensor based on the error characteristics according to claim 1, wherein the specific method in the step 1) is as follows: In the running process of the system, measurement data of a plurality of sensors on the same physical state are synchronously collected to form a multi-source measurement set corresponding to time Wherein, the Representing a set of all sensor measurements acquired at time t n ; j Representing the measured value of the jth sensor at a time t n , wherein M is the number of sensors; Each sensor corresponds to a unique number j (j=1, 2,., M) during acquisition for subsequent error analysis and consistency determination; According to the historical measurement sample, the experimental calibration result or the sensor technical parameter, respectively determining the measurement noise amplitude range and the communication delay range of each sensor; The measurement noise amplitude range is defined as Wherein ε + represents the maximum deviation of the sensor measurement noise and the communication-on delay range is defined as Where τ - and τ + represent minimum and maximum communication delays, respectively; and determining the error distribution characteristics of the measurement output of each sensor through the parameters, establishing an error boundary model capable of comprehensively reflecting measurement noise and communication delay uncertainty, and providing a quantitative basis for subsequent boundary correction and fusion analysis.
  3. 3. The method for detecting abnormal states of the unspecified multivariable sensor based on the error characteristics according to claim 2, wherein the specific method in the step 2) is as follows: after the measuring noise and communication delay range are determined, the dynamic change characteristics of the sensor output are comprehensively considered, and the j-th sensor is defined at the moment The measurement error boundaries of (2) are: Wherein, the And Respectively representing the lower limit and the upper limit of the error boundary of the jth sensor; For the change rate of the sensor, reflecting the change speed of the measured value of the sensor along with time, wherein τ + is the upper limit of communication delay; Rate of change The calculation mode of (a) is as follows: Wherein, the For the measurement of the sensor at the last sampling instant, Sampling period for the system; Determining upper and lower boundary limits according to the change rate direction: When (when) And (3) when the temperature is equal to or higher than 0: When (when) < 0: through the method, an error boundary which dynamically changes along with measurement is generated in real time at each sampling moment, and the measurement uncertainty range under the combined action of measurement noise disturbance and communication delay is represented; Sequentially performing boundary calculation on all sensors in the system to obtain a complete multisource sensor error boundary set Wherein, the Representing a set of all sensor error boundaries at time t n , each element in the set All represent the interval range of the corresponding sensor which possibly contains the real physical state under the conditions of measurement noise and communication delay; The error boundary set is used as an input basis for subsequent anomaly detection and boundary overlap analysis and is used for judging consistency relations among different sensor measurement results.
  4. 4. The method for detecting abnormal states of an unspecified multi-variable sensor based on error characteristics according to claim 3, wherein the specific method in the step 3) is as follows: Dynamically correcting the corresponding error boundary according to the change rate of each sensor at the current sampling moment, and increasing the contribution of a communication delay term to the boundary length if the change rate of the output signal of the sensor is larger than the ratio of the maximum error to the maximum communication delay; The corrected error boundary is expressed as: Wherein, the Indicating the margin of error after the correction, A communication delay compensation term dynamically adjusted according to the communication delay characteristic of the sensor at the current moment; Aiming at the conditions of measurement drift, sensor aging or nonlinear measurement noise in a system, part of boundaries are moderately expanded, and the expanded boundaries are expressed as: Wherein, the The parameter kappa is a boundary expansion coefficient, the value range of the parameter kappa is [0,1], and the parameter kappa is self-adaptively set according to the system measurement noise level.
  5. 5. The method for detecting abnormal states of unspecified multivariable sensor based on error characteristics according to claim 4, wherein the specific method in the step 4) is as follows: Dynamic error boundary set for all sensors at the same time t n Wherein the method comprises the steps of Representing error boundaries of the jth sensor to respectively And Representing a lower limit and an upper limit; performing interval overlapping operation on boundaries of all sensors to identify consistent regions between measurement results of different sensors, The overlap interval is defined as: When (when) When the boundary of all sensors has a common overlap interval ; When (when) When the boundary of each sensor is free from intersection, the system has obvious consistency deviation, and the abnormal sensor needs to be further judged; To quantify the degree of coincidence of each sensor boundary with the overall overlap interval, the overlap of the jth sensor is defined The method comprises the following steps: Wherein, the Representing the length of intersection of the sensor boundary with the overlap interval, Indicating the length of the sensor boundary, The value of (1) is 0,1, when The closer to 1, the more consistent the sensor measurement with most sensors; setting a consistency threshold eta epsilon (0, 1), if it meets Judging that the jth sensor is abnormal or attacked at the time t n , and marking the jth sensor as an abnormal set omega a , otherwise, judging the jth sensor as a normal sensor, and marking the jth sensor as a normal set omega n ; The abnormality detection formula is that if Then If (1) Then 。
  6. 6. The method for detecting abnormal states of the unspecified multivariable sensor based on the error characteristics according to claim 5, wherein the specific method in the step 5) is as follows: After determining the normal sensor set omega n , carrying out normalization processing on the corresponding overlapping degree to obtain a weight coefficient : Weighted fused global boundaries Expressed as: Thereby obtaining the weighted overlap interval of each sensor boundary on the basis of consistency, realizing abnormal rejection and fusion estimation, and globally weighting the overlap interval As a state estimate for the system at time t n : Wherein, the And The upper and lower limits of the fusion boundary, respectively.

Description

Method for detecting state abnormality of nonspecific multivariable sensor based on error characteristics Technical Field The invention belongs to the technical field of intelligent sensing and control, and particularly relates to an unspecified multivariable sensor state anomaly detection method based on error characteristics, which can be widely applied to intelligent driving vehicles, unmanned aerial vehicle systems, industrial process monitoring and other complex systems requiring multi-sensor cooperative sensing and is used for improving the sensing reliability and fault tolerance of the system. Background With the wide application of intelligent equipment and automatic control systems, multi-source sensors inside the system have become key components for realizing high-precision sensing and stable control. However, since the sensor is susceptible to external environmental disturbance, electromagnetic noise, communication delay, malicious attack and other factors in actual operation, the output signal of the sensor often has random fluctuation, delay offset and even abnormal drift. These error factors may cause deviation in the state estimation of the system, and in severe cases, control instability, functional failure or safety accidents may be caused. Currently, conventional methods for sensor abnormality detection mainly include threshold judgment based on statistical analysis, residual analysis based on model prediction, abnormality recognition based on machine learning, and the like. Although the methods can realize anomaly detection to a certain extent, the method has the following defects that (1) most methods only rely on single-point data of a sensor to judge and lack comprehensive analysis on consistency of multi-source measurement, (2) dynamic change characteristics of noise and delay are ignored, and an error boundary of the sensor cannot be accurately defined, and (3) under the condition that part of sensors are attacked or drift, the traditional method is difficult to keep stability and continuity of system state estimation. Therefore, an anomaly detection method that can comprehensively consider random noise, communication delay and error distribution characteristics of a sensor is highly needed to realize consistency analysis and highly robust state estimation of multi-source data. Disclosure of Invention In order to overcome the defects of the prior art, the invention provides an unspecified multivariable sensor state abnormality detection method based on error characteristics. According to the method, a consistency model of multi-sensor measurement is established, the noise characteristics and delay distribution of the sensors are combined, the measurement error boundary of each sensor is dynamically established, and abnormal detection and state estimation are realized through boundary overlap analysis, so that the running stability of the system can be maintained under the condition that part of the sensors are attacked or performance is degraded. The technical scheme of the invention is that the method for detecting the state abnormality of the unspecified multivariable sensor based on the error characteristics comprises the following steps: Step 1) multisource data acquisition and modeling, namely acquiring measurement data of a plurality of sensors on the same physical state at the same moment, establishing a multisource sensor measurement set, and establishing a multisource error boundary model based on statistical information of measurement noise and communication delay, wherein the multisource error boundary model is used for describing an uncertainty range of output of each sensor; The specific method comprises the following steps: In the running process of the system, measurement data of a plurality of sensors on the same physical state are synchronously collected to form a multi-source measurement set corresponding to time Wherein, the Representing a set of all sensor measurements acquired at time tn;j Representing the measured value of the jth sensor at the time tn, wherein M is the number of sensors; Each sensor corresponds to a unique number j (j=1, 2,., M) during acquisition for subsequent error analysis and consistency determination; according to the historical measurement sample, the experimental calibration result or the sensor technical parameter, respectively determining the random noise amplitude range and the communication delay range of each sensor; The noise amplitude range is defined as Wherein ε + represents the maximum deviation of the sensor measurement noise and the communication delay range is defined asWhere τ - and τ + represent minimum and maximum delays, respectively; And determining the error distribution characteristics of the measurement output of each sensor through the parameters, establishing an error boundary model capable of comprehensively reflecting noise and delay uncertainty, and providing a quantitative basis for subsequent boundary correc