Digital oscilloscope intelligent abnormal capture implementation method

Abstract

The application discloses an intelligent abnormality capturing implementation method of a digital oscilloscope. A to-be-measured signal is input into the digital oscilloscope, the to-be-measured signal is collected by the digital oscilloscope to obtain a sampling signal containing multiple complete periods, then adaptive processing is performed on the sampling signal in amplitude and time, so that the sampling signal meets the requirements of subsequent processing; then multi-dimensional feature extraction is performed on the preprocessed signal, classification of the signal is realized based on multi-dimensional feature information, and then a suitable abnormality detection model is selected according to the classification result of the signal to perform abnormality detection.

Description

A method for intelligent anomaly capture in digital oscilloscopes Technical Field

[0001] This invention belongs to the field of digital oscilloscope technology, and more specifically, relates to a method for intelligent anomaly capture in digital oscilloscopes. Background Technology

[0002] In the rapid development of the electronic information field, the increased sampling rate of acquisition systems has led to a dramatic increase in the amount of data that needs to be processed and analyzed. This presents a significant challenge to the real-time performance requirements of acquisition systems in accurately capturing abnormal data from massive datasets. In this context, methods for checking and capturing abnormal signals become particularly important.

[0003] Different signal acquisition systems or instruments typically have limited capabilities in analyzing signal characteristics. For example, oscilloscopes primarily analyze the time-domain characteristics of signals, spectrum analyzers analyze the spectral characteristics, serial data analyzers analyze eye diagrams, and vector analyzers analyze vector diagrams. Different signal acquisition systems also employ different triggering methods.

[0004] Oscilloscopes commonly use analog and digital triggering methods. Analog triggering is simple to control, but due to analog-to-digital synchronization issues, it suffers from large jitter and low accuracy. Furthermore, because the input signal passes through signal conditioning channels and analog trigger channels with different delays, there is a phase deviation between the resulting sampled signal and the analog trigger signal. Digital triggering offers higher accuracy, but as the sampling rate of acquisition systems increases, the number of parallel data channels used for trigger positioning also increases. More parallel channels mean a smaller trigger point positioning range, thus requiring higher triggering accuracy.

[0005] Common triggering functions for spectrum analyzers include automatic triggering, single triggering, external triggering, video triggering, and level triggering. Common triggering functions for vector analyzers include manual triggering, automatic triggering, external triggering, internal triggering, event triggering, and internal triggering.

[0006] Currently, these methods and functions all have certain shortcomings in capturing abnormal signals, which can be summarized in the following two points: 1. Under the existing structure, the capture rate of triggering is low, and there is a dead time, which may lead to the omission of target signals. 2. The abnormal signal morphology is unknown, and it cannot be captured by relying on the existing multiple triggering methods.

[0007] Therefore, this paper proposes an abnormal signal capture method applicable to various testing instruments and acquisition systems. The proposed method first preprocesses the signal, adaptively processing its amplitude and time to ensure it meets the requirements of subsequent processing. Then, it extracts multi-dimensional features from the preprocessed signal, classifies the signal based on these features, and finally selects an appropriate anomaly detection model based on the classification results for anomaly detection. Summary of the Invention

[0008] The purpose of this invention is to overcome the shortcomings of the prior art and provide a method for intelligent anomaly capture in digital oscilloscopes. This method classifies signals based on multi-dimensional feature information, and then selects an appropriate anomaly detection model based on the classification results to perform anomaly detection. This method is applicable to intelligent anomaly capture in various test instruments and acquisition systems.

[0009] To achieve the above-mentioned objectives, the present invention provides a method for intelligent anomaly capture in a digital oscilloscope, characterized by comprising the following steps:

[0010] (1) Input the signal to be tested into a digital oscilloscope and acquire the sampled signal containing multiple complete cycles through the digital oscilloscope;

[0011] (2) Extract the feature values ​​of the sampled signal, including the period and peak-to-peak value of the sampled signal;

[0012] (3) Set the vertical and horizontal sensitivity of the digital oscilloscope according to the characteristic values ​​of the sampled signal;

[0013] (4) Extracting features of the sampled signal based on the period of the sampled signal;

[0014] (5) Classify the sampled signals into: time-domain characteristic signals to be detected, frequency-domain characteristic signals to be detected, and statistical domain characteristic signals to be detected.

[0015] (6) Select the corresponding features according to the different domain characteristics, and set a detection threshold Ts for each feature based on the normal signal. c The benchmark score and the corresponding feature anomaly detection model are used to construct the feature anomaly detection model for the corresponding domain. The time-domain feature detection model corresponding to the signal to be detected is denoted as P1, and the detection threshold corresponding to each component feature is denoted as Ts. 1,j The baseline score is denoted as SD. 1,j The frequency domain feature detection model corresponding to the signal to be detected is denoted as P2, and the detection threshold corresponding to each component feature is denoted as Ts. 2,j The baseline score is denoted as SD. 2,j The feature detection model corresponding to the signal to be detected in the statistical domain is denoted as P3, and the detection threshold corresponding to each component feature is denoted as Ts.3,j The baseline score is denoted as SD. 3,j j is the feature number;

[0016] (7) Determine the anomalies of the signal under test through feature detection model;

[0017] The domain feature to be detected signal classified in step (6) is input into the feature anomaly detection model, and then it is calculated whether each feature corresponding to the domain feature to be detected signal belongs to the judgment range (SD). i,j -Ts i,j SD i,j +Ts i,j Within the range, i = 1, 2, 3 represents the type number of the domain feature signal to be detected. If all are within the judgment range, the signal to be detected is judged as a normal signal; otherwise, it is an abnormal signal.

[0018] The objective of this invention is achieved as follows:

[0019] This invention discloses a method for intelligent anomaly capture using a digital oscilloscope. The method involves inputting the signal under test into a digital oscilloscope, acquiring a sampled signal containing multiple complete cycles, and then adaptively processing the sampled signal in terms of amplitude and time to ensure it meets the requirements of subsequent processing. Next, multi-dimensional feature extraction is performed on the pre-processed signal, and the signal is classified based on this multi-dimensional feature information. Finally, an appropriate anomaly detection model is selected based on the classification results to detect anomalies.

[0020] Meanwhile, the intelligent anomaly capture method for digital oscilloscopes of the present invention also has the following beneficial effects:

[0021] (1) By selecting features to form a feature network, it can automatically compare signal features and determine whether they are abnormal. It can quickly locate abnormal signals in complex signals, saving a lot of time and manpower costs. It has obvious advantages, especially in long-term monitoring or large-scale testing.

[0022] (2) The feature network composed of selected features does not rely on a single signal feature for judgment, but comprehensively considers multiple features, which can more comprehensively and accurately identify abnormal signals, avoid misjudgment or omission caused by the limitations of a single feature, and has better adaptability to complex and ever-changing signal environments.

[0023] (3) The feature network composed of selected features and detection thresholds is highly flexible in detecting signal anomalies. Users can select targeted features to form a feature network, and when setting the detection threshold, they can flexibly set the threshold size based on prior knowledge and adjust the importance of different features in the feature network. Attached Figure Description

[0024] Figure 1 is a flowchart of a method for intelligent anomaly capture in a digital oscilloscope according to the present invention.

[0025] Figure 2 is a flowchart of the period extraction process for the sampled signal;

[0026] Figure 3 is a flowchart of the sampling signal classification process;

[0027] Figure 4 is a schematic diagram of the feature anomaly detection model. Detailed Implementation

[0028] The specific embodiments of the present invention will now be described with reference to the accompanying drawings to enable those skilled in the art to better understand the invention. It should be particularly noted that in the following description, detailed descriptions of known functions and designs that might obscure the main content of the invention will be omitted here.

[0029] Example

[0030] In this embodiment, as shown in Figure 1, the present invention provides a method for intelligent anomaly capture in a digital oscilloscope, comprising the following steps:

[0031] (1) Signal acquisition;

[0032] The signal to be tested is input into a digital oscilloscope, and the digital oscilloscope acquires the signal to be tested to obtain a sampled signal containing multiple complete cycles;

[0033] (2) Extract the feature values ​​of the sampled signal;

[0034] (2.1) Extract the period of the sampling signal. The extraction process is shown in Figure 2.

[0035] (2.1.1) Traverse each sampling point of the sampled signal and record the feature information of the maximum or minimum point, denoted as (V, L), where V is the amplitude of the extreme point and L is the number of points between the previous extreme point and the current extreme point; the feature information of the first extreme point X0 is denoted as (V0, L0), where V0 is the amplitude of X0 and L0 is denoted as 0; the information of the second extreme point X1 is denoted as (V1, L1), where V1 is the amplitude of X1 and L1 is the number of points between X0 and X1; and so on, for the Nth extreme point X N-1 The information is denoted as (V) N-1 ,L N-1 ), L N-1 For X N-2 To X N-1 The number of points between;

[0036] (2.1.2) Based on the extreme points recorded above, take m as 1, starting from X0, and take 2m+1 consecutive extreme points [X0, X1, ..., X... m-1 ,Xm ,X m+1 ,…,X 2m Then divide it into two groups [X0, X1, ..., X]. m ]、[X m ,X m+1 ,…,X 2m ]

[0037] (2.1.3) Set judgment condition A: Determine whether V0 is equal to V m Is V1 equal to V? m+1 ... V m Is it equal to V? 2m If any equation is not equal, increment the value of m by 1 and return to step (2.1.2); if all equations are equal, proceed to step (2.1.4).

[0038] (2.1.4) Set judgment condition B: Determine whether L1 is equal to L m+1 Is L2 equal to L? m+2 ... L m -L 2m If the differences are all equal, increment the value of m by 1 and return to step (2.1.2); if they are all equal, record L0, L1, ..., L... m Then calculate the period T of the sampled signal. Among them, f s This refers to the real-time sampling rate of the data acquisition system.

[0039] (2.2) Extract the peak-to-peak value of the sampled signal;

[0040] Extract the maximum value V of the sampled signal within one period max and minimum value V min Then calculate the peak-to-peak value V of the sampled signal. pp =V max -V min ;

[0041] (3) Set the vertical and horizontal sensitivity of the digital oscilloscope according to the characteristic values ​​of the sampled signal;

[0042] (3.1) Set the vertical sensitivity;

[0043] According to the formula Find S0, where S0 is the full-scale amplitude of the entire oscilloscope screen, and h is the ratio of the peak-to-peak value of the signal to the full scale, with a value of 0.8.

[0044] Round S0 up to obtain S1, and then calculate the vertical sensitivity s of the digital oscilloscope according to the following formula:

[0045]

[0046] Where ALL_DIV_V is the number of divisions in the vertical direction of the digital oscilloscope;

[0047] Finally, set the vertical sensitivity setting of the digital oscilloscope to s;

[0048] (3.2) Set the horizontal sensitivity;

[0049] According to the formula Find q0, where q0 is the number of horizontal divisions on the digital oscilloscope that one cycle of the signal occupies, ALL_DIV_H is the number of horizontal divisions on the digital oscilloscope, and R is the number of signal cycles displayed on the oscilloscope screen as set by the user.

[0050] Round q0 up to get q1, then calculate the horizontal sensitivity q of the digital oscilloscope using the following formula:

[0051]

[0052] Finally, set the horizontal sensitivity setting of the digital oscilloscope to q;

[0053] (4) Extracting features of the sampled signal based on the period of the sampled signal;

[0054] Extract the time-domain features of the sampled signal, including: period, amplitude, duty cycle, overshoot, preshoot, phase difference, and offset;

[0055] Extract the frequency domain features of the sampled signal, including: spectral distribution, spectral energy, spectral envelope, spectral distance, fundamental frequency, maximum frequency, intermediate frequency, and maximum peak value.

[0056] Extract the statistical domain features of the sampled signal, including: maximum value, minimum value, mean, median, skewness, kurtosis, histogram, mean absolute error, median absolute error, root mean square, standard deviation, and variance.

[0057] (5) Classify the sampled signals. The classification process is shown in Figure 3.

[0058] (5.1) Detect the number of extreme points of the sampled signal within one period, denoted as k; determine whether the count value k is less than the preset threshold t. If k < t, classify the sampled signal as a time-domain feature to be detected signal, and then proceed to step (6); otherwise, proceed to step (5.2).

[0059] (5.2) Detect the number of spectral components of the sampled signal in one period, denoted as p, and determine whether the count value p is less than the preset threshold f. If p < f, classify the sampled signal as a frequency domain feature to be detected signal, and then proceed to step (6); otherwise, classify the sampled signal as a statistical domain feature to be detected signal, and then proceed to step (6).

[0060] (6) Construct a feature anomaly detection model;

[0061] Based on the characteristics of the signal to be detected in different domains, corresponding features are randomly selected, and a detection threshold Ts is set for each feature. c The benchmark score and the corresponding feature anomaly detection model are used to construct the feature anomaly detection model for the corresponding domain. The time-domain feature detection model corresponding to the signal to be detected is denoted as P1, and the detection threshold corresponding to each component feature is denoted as Ts. 1,j The baseline score is denoted as SD. 1,j The frequency domain feature detection model corresponding to the signal to be detected is denoted as P2, and the detection threshold corresponding to each component feature is denoted as Ts. 2,j The baseline score is denoted as SD. 2,j The feature detection model corresponding to the signal to be detected in the statistical domain is denoted as P3, and the detection threshold corresponding to each component feature is denoted as Ts. 3,j The baseline score is denoted as SD. 3,j j is the feature number; thus, the feature detection model for each domain can be obtained, as shown in Figure 4.

[0062] (7) Determine the anomalies of the signal under test through feature detection model;

[0063] The domain feature to be detected signal classified in step (6) is input into the feature anomaly detection model, and then it is calculated whether each feature corresponding to the domain feature to be detected signal belongs to the judgment range (SD). i,j -Ts i,j SD i,j +Ts i,j Within the range, i = 1, 2, 3 represents the type number of the domain feature signal to be detected. If all are within the judgment range, the signal to be detected is judged as a normal signal; otherwise, it is an abnormal signal.

[0064] Below, we will use a sinusoidal signal with glitches, u(t) = cos(2π × 1000t), as an example to illustrate the above processing flow, as follows:

[0065] (1) Acquire a sine wave signal to obtain a sampled signal;

[0066] (2) Extract the time-domain feature values ​​of the sampled signal;

[0067] (2.1) Extracting the period of the sampling signal;

[0068] After enabling the "Smart Anomaly Capture" function in the user's acquisition system or oscilloscope interface and selecting "Maximum" as the extreme value type, the acquisition system or oscilloscope will detect the maximum value point X0 of the signal and record it as (V). i ,L i ), i = 0, 1, 2…. Based on the recorded (V) i ,L i Let m be 1, starting from X0, and take 2m+1 consecutive extreme points.

[0069] [X0,X1,…,X m-1 ,X m ,X m+1 ,…,X 2m Then divide it into two groups [X0, X1, ..., X]. m ]、[X m ,X m+1 ,…,X 2m ]

[0070] Executing the judgment conditions A and B, the final value of the period T is 1ms, that is, the period of the signal to be detected is 1ms.

[0071] (2.2) Extract the peak-to-peak value of the sampled signal;

[0072] Extract the maximum value V of the signal within a 1ms period. max The minimum value is V, which is 1V. min The value is -1V. The peak-to-peak value V of the signal is calculated. pp =V max -V min It is 2V.

[0073] (3) Set the vertical and horizontal sensitivity of the digital oscilloscope according to the characteristic values ​​of the sampled signal;

[0074] According to the formula Setting h to 0.8, we get S0 as 2500mV. Rounding S0 up gives S1, which is also 2500mV. According to the formula... Setting ALL_DIV_V to 10 yields an s value of 250 mV / div. Set the oscilloscope's vertical sensitivity setting to 250 mV / div.

[0075] According to the formula With R set to 5, meaning the oscilloscope screen displays 5 complete signal cycles, and ALL_DIV_H is 10, we can obtain q0 as 2. Rounding q0 up gives q1 as 2. According to the formula... We can deduce that q is 500µs / div. Set the oscilloscope's horizontal sensitivity setting to 500µs / div.

[0076] (4) Extracting features of the sampled signal based on the period of the sampled signal;

[0077] Taking the period of glitches as an example, the time-domain features of the signal under test are extracted: the period is 1ms, the amplitude range is -1000mV to 1100mV, and a 100mV glitch anomaly occurs; the frequency-domain features of the signal under test are extracted, and the signal energy is mainly concentrated at 1000Hz and -1000Hz; the statistical domain features of the signal under test are extracted. The maximum value is 1100mV, the minimum value is -1000mV, and the average value is 50mV.

[0078] (5) Classify the sampled signals;

[0079] First, the number of extreme points within one period of the signal under test is 2, which is less than the set parameter of 10. This indicates that the signal under test is a signal with low complexity in the time domain, and it is classified as a "time-domain characteristic signal to be detected".

[0080] (6) Construct a feature anomaly detection model;

[0081] Based on the time-domain features extracted from the normal signal u(t)=cos(2π×1000t), for the "time-domain feature detection signal" classified as such above, two time-domain features, period and amplitude, are selected to construct the "time-domain feature detection model". The period and amplitude features extracted from the normal signal are used as benchmarks SD1 and SD2, respectively, and detection thresholds Ts1 and Ts2 are set for the two features, where Ts1 is 0.05ms and Ts2 is 50mV.

[0082] (7) Anomaly Judgment: Based on the above "feature anomaly detection model", the period judgment range for signal features meeting normal requirements is (0.95ms, 1.05ms), and the amplitude judgment range is (-1050mV, 1050mV). For the signal to be detected, the period is 1ms, which meets the judgment condition, while the amplitude range is -1000mV to 1100mV, which does not meet the judgment condition. Therefore, the signal is judged to be an abnormal signal.

[0083] Although the illustrative specific embodiments of the present invention have been described above to enable those skilled in the art to understand the invention, it should be understood that the invention is not limited to the scope of the specific embodiments. For those skilled in the art, various changes are obvious as long as they are within the spirit and scope of the invention as defined and determined by the appended claims, and all inventions utilizing the concept of the present invention are protected.

Claims

1. A method for intelligent anomaly capture in a digital oscilloscope, characterized in that, Includes the following steps: (1) Input the signal to be tested into a digital oscilloscope and acquire the signal to be tested through the digital oscilloscope to obtain a sampled signal containing multiple complete cycles; (2) Extract the feature values ​​of the sampled signal, including the period and peak-to-peak value of the sampled signal; (3) Set the vertical sensitivity and horizontal sensitivity of the digital oscilloscope according to the feature values ​​of the sampled signal; (4) Extract the features of the sampled signal based on the period of the sampled signal. Extract the time domain features of the sampled signal, including: period, amplitude, duty cycle, overshoot, preshoot, phase difference, and offset; extract the frequency domain features of the sampled signal, including: spectral distribution, spectral energy, spectral envelope, spectral distance, spectral fundamental frequency, spectral maximum frequency, spectral intermediate frequency, and spectral maximum peak value; extract the statistical domain features of the sampled signal, including: maximum value, minimum value, mean, median, skewness, kurtosis, histogram, mean absolute error, median absolute error, root mean square, standard deviation, and variance; (5) classify the sampled signal; (5.1) detect the number of extreme points of the sampled signal within one period, denoted as k; determine whether the count value k is less than the preset value. Threshold δ1, if k < δ1, then classify the sampled signal as a time-domain feature to be detected signal, and then proceed to step (6); otherwise, proceed to step (5.2); (5.2), detect the number of spectral components of the sampled signal in one period, denoted as h, and determine whether the count value h is less than the preset threshold δ2. If h < δ2, then classify the sampled signal as a frequency-domain feature to be detected signal, and then proceed to step (6); otherwise, classify the sampled signal as a statistical domain feature to be detected signal, and then proceed to step (6); (6), construct a feature anomaly detection model; select the corresponding features according to different domain features, and set a detection threshold Ts for each feature based on the normal signal. c The benchmark score and the corresponding feature anomaly detection model are used to construct the feature anomaly detection model for the corresponding domain. The time-domain feature detection model corresponding to the signal to be detected is denoted as P1, and the detection threshold corresponding to each component feature is denoted as Ts. 1,j The baseline score is denoted as SD. 1,j The frequency domain feature detection model corresponding to the signal to be detected is denoted as P2, and the detection threshold corresponding to each component feature is denoted as Ts. 2,j The baseline score is denoted as SD. 2,j The feature detection model corresponding to the signal to be detected in the statistical domain is denoted as P3, and the detection threshold corresponding to each component feature is denoted as Ts. 3,j The baseline score is denoted as SD. 3,j , j is the feature number; (7) Determine the anomaly of the signal to be tested through the feature detection model; Input the domain feature signal to be tested classified in step (6) into the feature anomaly detection model, and then calculate whether each feature corresponding to the domain feature signal to be tested belongs to the judgment range (SD). i,j -Ts i,j SD i,j +Ts i,j Within the range, i = 1, 2, 3 represents the type number of the domain feature signal to be detected. If all are within the judgment range, the signal to be detected is judged as a normal signal; otherwise, it is an abnormal signal.

2. The method for intelligent anomaly capture in a digital oscilloscope according to claim 1, characterized in that, The sampling signal period extraction method is as follows: (2.1) Traverse each sampling point of the sampling signal and record the feature information of the maximum or minimum point, denoted as (V, L), where V is the amplitude of the extreme point and L is the number of points between the previous extreme point and this extreme point; the feature information of the first extreme point X0 is denoted as (V0, L0), where V0 is the amplitude of X0 and L0 is denoted as 0; the information of the second extreme point X1 is denoted as (V1, L1), where V1 is the amplitude of X1 and L1 is the number of points between X0 and X1; and so on, the Nth extreme point X N-1 The information is denoted as (V) N-1 ,L N-1 ), L N-1 For X N-2 To X N-1 (2.2) Based on the extreme points recorded above, take m as 1, starting from X0, take 2m+1 consecutive extreme points [X0,X1,…,X]. m-1 ,X m ,X m+1 ,…,X 2m Then divide it into two groups [X0, X1, ..., X]. m ]、[X m ,X m+1 ,…,X 2m (2.3) Determine whether V0 is equal to V m Is V1 equal to V? m+1 ... V m Is it equal to V? 2m If any equation is not equal, increment the value of m by 1 and return to step (2.2); if all equations are equal, proceed to step (2.4); (2.4) Determine if L1 is equal to L m+1 Is L2 equal to L? m+2 ... L m -L 2m If the differences are all equal, increment the value of m by 1 and return to step (2.2); if they are all equal, record L0, L1, ..., L m Then calculate the period T of the sampled signal. Among them, f s This represents the real-time sampling rate of the data acquisition system.

3. The method for intelligent anomaly capture in a digital oscilloscope according to claim 1, characterized in that, The method for extracting the peak-to-peak value of the sampled signal is as follows: extract the maximum value V of the sampled signal within one period. max and minimum value V min Then calculate the peak-to-peak value V of the sampled signal. pp =V max -V min .

4. The method for intelligent anomaly capture in a digital oscilloscope according to claim 1, characterized in that, The method for setting the vertical sensitivity of the digital oscilloscope is as follows: according to the formula... Find S0, where S0 is the full-scale amplitude of the oscilloscope screen, and h is the percentage of the signal peak-to-peak value to the full-scale range; round up S0 to obtain S1, and then calculate the vertical sensitivity s of the digital oscilloscope according to the following formula: Where ALL_DIV_V is the number of divisions in the vertical direction of the digital oscilloscope; finally, set the vertical sensitivity setting of the digital oscilloscope to s.

5. The method for intelligent anomaly capture in a digital oscilloscope according to claim 1, characterized in that, The method for setting the horizontal sensitivity of the digital oscilloscope is as follows: according to the formula Find q0, where q0 is the number of horizontal divisions on the digital oscilloscope that one cycle of the signal occupies, ALL_DIV_H is the number of horizontal divisions on the digital oscilloscope, and R is the number of signal cycles displayed on the oscilloscope screen as set by the user. Round q0 up to get q1, and then calculate the horizontal sensitivity q of the digital oscilloscope according to the following formula: Finally, set the horizontal sensitivity setting of the digital oscilloscope to q.

Images