Analog signal online anomaly detection method, electronic device and readable storage medium

Abstract

A method for online anomaly detection of analog signals adaptively generates upper and lower limits of analog signals based on historical data and real-time collected data, and discovers abnormal over-limits, trends and fluctuations in the data through signal decomposition. This enables online anomaly detection of analog signals, improves the accuracy and precision of anomaly detection, and ultimately issues corresponding alarms based on the anomaly detection results and displays the health status of the equipment in real time, making it easier for on-site staff to quickly locate the source of the fault, thereby swiftly resolving the fault and ensuring the safe operation of trains and signal equipment.

Claims

1. A method for online anomaly detection of analog signals, characterized in that: The following steps are involved: Step S1: generating a signal data set to be analyzed and a reference signal data set, wherein the signal data set to be analyzed and the reference signal data set constitute original signal data; The analog signal source is collected to obtain n data to form a signal data set to be analyzed, which is recorded as X = [x1, x2, ..., x n ]; Select normal historical data of any time period with the same collection items as the data in the data set to be analyzed to form a reference benchmark signal data set, denoted as XH = [xh1, xh2, ..., xh m ], the reference signal data set has m data; Step S2, decomposing the original signal data to obtain trend components and residual components; The trend component of the signal data set X to be analyzed is denoted as The residual component of the signal data set X to be analyzed is recorded as The trend component of the reference signal dataset XH is denoted as The residual component of the reference signal data set XH is denoted as Step S3, performing statistical calculations on the original signal data and its trend component and residual component; Calculate the mean and standard deviation of the original signal data and its trend component and residual component respectively, where the signal data to be analyzed X and the trend component X of the signal data to be analyzed are T , the residual component X of the signal data to be analyzed R The mean of The standard deviation is denoted as X std 、 Reference signal data XH, trend component XH of reference signal data T , the residual component XH of the reference signal data R The mean of The standard deviation is denoted as XH std 、 Step S4, calculating the normal upper limit and normal lower limit of the signal data to be analyzed and its trend component and residual component; Assuming that the collected data obeys the normal distribution, the normal upper limit and normal lower limit of the signal data to be analyzed and its trend component and residual component are calculated according to the z-sigma criterion, z∈[2,3]; Step S5: performing abnormal analysis on the signal data to be analyzed and its trend component and residual component according to the normal upper limit and the normal lower limit; Step S6: Send the data anomaly detection result and display an alarm.

2. The method for online anomaly detection of analog signals according to claim 1, wherein: If the time series data sampled in the signal data set to be analyzed is stationary, the larger the value of m is, the higher the accuracy of the algorithm is; if the time series data sampled in the signal data set to be analyzed is not stationary, m is set to 0.

3. The method for online anomaly detection of analog signals according to claim 1, wherein: In step S2, the method of decomposing and obtaining the trend component and the residual component includes: Taking the decomposition of the signal data X to be analyzed as an example, the trend component X is calculated by the sliding window mean filtering algorithm based on the center point. T , the calculation formula is: Where r is the sliding window radius, r is less than n, r∈[1,n / 3]; For x i-k , if the subscript index ik<1, then let x i-k =x1; For x i+k , if the subscript index i+k>n, then let x i+k =x n ; According to the signal data X to be analyzed and the trend component X T The difference between the two is the residual component X R , the calculation formula is:

4. The method for online anomaly detection of analog signals according to claim 3, wherein: In step S3, the method for performing statistical calculation includes: To calculate the mean of the signal data X to be analyzed and standard deviation X std For example:

5. The method for online anomaly detection of analog signals according to claim 4, wherein: In step S4, the method for calculating the normal upper limit and the normal lower limit of the signal data to be analyzed and its trend component and residual component includes: The upper limit of the normal value of the signal data X to be analyzed is denoted as ub z , the lower limit of normal is denoted as lb z ; The trend component X of the signal data to be analyzed T The upper limit of normal is recorded as The lower limit of normal is recorded as The residual component X of the signal data to be analyzed R The upper limit of normal is recorded as The lower limit of normal is recorded as 6. The method for online anomaly detection of analog signals according to claim 5, wherein: In step S5, the method for performing abnormality analysis includes: For the original signal data X to be analyzed: If x i < ub2 or x i > lb2, it is determined that the signal to be analyzed does not exceed the limit at time i; If ub2 < x i <ub3 or lb3 < x i <lb2, then it is determined that the signal to be analyzed is slightly over-limit at time i; If x i > ub3 or x i < lb3, then it is determined that the signal to be analyzed is severely over-limit at time i; For the trend component X of the original signal data to be analyzed T : like or It is judged that the signal to be analyzed does not show abnormal trend at time i; like or Then it is judged that the signal to be analyzed has a slight abnormal trend at time i; like or It is judged that the signal to be analyzed has a serious abnormal trend at time i; For the residual component X of the original signal data to be analyzed R : like or It is judged that the signal to be analyzed has no abnormal fluctuation at time i; like or It is judged that the signal to be analyzed has a slight abnormal fluctuation at time i; like or It is judged that the signal to be analyzed has serious abnormal fluctuations at time i.

7. An electronic device, characterized in that: The method comprises a processor and a memory, wherein a computer program is stored in the memory, and when the computer program is executed by the processor, the method for online anomaly detection of analog signals according to any one of claims 1 to 6 is implemented.

8. A readable storage medium, characterized in that: The readable storage medium stores a computer program, and when the computer program is executed by a processor, the online anomaly detection method for analog signals according to any one of claims 1 to 6 is implemented.

Images