Anomaly detection method based on time domain joint features
By deploying a trigger detection module with multi-dimensional time-domain characteristics in a digital oscilloscope, the shortcomings of traditional oscilloscopes in detecting complex signals are solved, enabling efficient capture and accurate identification of abnormal signals.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2025-04-21
- Publication Date
- 2026-03-24
AI Technical Summary
Traditional digital oscilloscopes have a limited triggering function, making it difficult to effectively detect complex abnormal signals. They are also prone to false triggering due to noise and measurement errors, and cannot meet diverse signal detection needs.
An anomaly detection method based on joint time-domain features is adopted. By deploying detection modules triggered by extreme values, pulse width, rise time, and fall time, multi-dimensional time-domain features are extracted and template comparison is performed to determine the anomaly of the signal.
It improves the ability to capture abnormal signals, reduces the false trigger rate, enhances detection accuracy and sensitivity, and is suitable for various signal detection scenarios.
Smart Images
Figure CN120352672B_ABST
Abstract
Description
Technical Field
[0001] This invention belongs to the field of digital oscilloscope technology, and more specifically, relates to an anomaly detection method based on joint time-domain features. Background Technology
[0002] With the continuous development of electronic information technology, signal complexity and testing demands are constantly increasing, placing increasingly stringent requirements on the diverse triggering capabilities and abnormal signal detection capabilities of oscilloscopes. Triggering, as a fundamental function in digital oscilloscopes used to capture waveforms of interest to the user and display them completely and stably, is an indispensable part of digital oscilloscopes. Traditional triggering functions directly send signals to the trigger module, which suffers from disadvantages such as single triggering characteristics, cumbersome triggering processes, and a lack of abnormal signal detection capabilities. When faced with the diversity of environmental noise, measurement errors, and complex dynamic systems, the detection capability of a single-feature triggering mechanism is limited, making it difficult to meet diverse application needs and accurately and effectively detect the presence of abnormal signals from time-domain signals.
[0003] Traditional triggering methods mainly include edge triggering, level triggering, and pulse width triggering. These methods only focus on one aspect of the signal characteristics, such as the rising edge or the threshold value of the signal amplitude, ignoring information about the overall dynamic changes of the signal. They are prone to "blinding" complex abnormal signals, and some abnormal signals may exhibit multiple characteristics at the same time. For example, pulse width abnormalities and extreme value exceedances may occur simultaneously. A single triggering condition is prone to false triggering and cannot effectively capture these abnormal signals.
[0004] Meanwhile, traditional triggering methods place high demands on hardware circuit design, and signal noise and trigger edge distortion can lead to false triggering, making it difficult to capture abnormal signals. When detecting multiple types of abnormal signals, it is usually necessary to set different trigger conditions one by one, resulting in frequent adjustments to the trigger configuration, which is inefficient and complex to operate.
[0005] In conclusion, there is currently no effective method for detecting abnormal signals. Therefore, designing an anomaly detection method suitable for oscilloscopes is of paramount importance. This method can enhance the ability of digital oscilloscopes to capture abnormal signals in different scenarios, in order to meet the ever-increasing demand for signal detection. Summary of the Invention
[0006] The purpose of this invention is to overcome the shortcomings of the prior art and provide an anomaly detection method based on joint time-domain features. By constructing triggering conditions through multi-dimensional time-domain features, the ability of digital oscilloscopes to detect abnormal signals is improved.
[0007] To achieve the above-mentioned objectives, the present invention provides an anomaly detection method based on temporal joint features, characterized by comprising the following steps:
[0008] (1) Deploy a trigger detection module in a digital oscilloscope. The trigger detection module specifically includes: an extreme value trigger detection module, a pulse width trigger detection module, and a rise time and fall time trigger detection module.
[0009] (2) Signal acquisition;
[0010] Different types of input signals are sequentially input into a digital oscilloscope, which then acquires the input signals. N segments of each type of input signal are acquired, where the j-th segment of the i-th type of input signal is denoted as X. ij X ij ={x ij1 ,x ij2 ,…,x ijk ,…,x ijM}, x ijk X represents ij There are k sampling points in the M, where k is the index of the sampling point and M is the number of sampling points;
[0011] Record each sampled signal X ij steady-state high level steady-state low level And the sampling interval Δt for each sampling point;
[0012] (3) For each sampled signal X ij Perform multi-dimensional temporal feature extraction;
[0013] (3.1) Extracting extreme value features;
[0014] Traverse each sampled signal X ij Find all amplitudes that satisfy V from the sampling points in the range. ijk-1 <V ijk >V ijk+1 sampling point x ijk Store in the maximum value sequence Find all amplitudes that satisfy V ijk-1 >V ijk <V ijk+1 The sampling points are stored in the minimum value sequence. Then, the extreme value range of each sampled signal segment is calculated:
[0015] Maximum range:
[0016] Minimum range:
[0017] (3.2) Extracting the rise time and fall time features;
[0018] (3.2.1) Calculate the rise time in, When the level is low, The voltage level rises 10% from low. This is the moment when the voltage rises 90% from a low level;
[0019] (3.2.2) Calculate the descent time in, When the voltage is high, The voltage level drops by 10% at the high level. The voltage level drops by 90% at the high level.
[0020] (3.2.3) For each sampled signal X ij Repeat the above steps to obtain N sets of rise times. and descent time
[0021] (3.2.4) Extract the rise time and fall time range features of the i-th type of input signal;
[0022] Rise time range characteristics:
[0023] Descent time range characteristics:
[0024] (3.3) Extracting positive and negative pulse width features;
[0025] (3.3.1) Set the sampling signal X for each segment ij threshold
[0026]
[0027] (3.3.2) Traverse each segment of the sampled signal X ij Find the sampling points in the middle that satisfy The sampling points are then identified and marked as zero-crossing points.
[0028] (3.3.3) Calculate the pulse width between two adjacent zero-crossing points;
[0029] If the sampled signal X ij The amplitude V at the l-th zero crossing point ijl satisfy: Then calculate the positive pulse width between the l-th zero-crossing point and the (l+1)-th zero-crossing point.
[0030]
[0031] in, This indicates that the l-th zero-crossing point is in the sampled signal X. ij The sampling point number in the sample;
[0032] If the sampled signal X ij The amplitude V at the l-th zero crossing point ijl satisfy: Then calculate the negative pulse width between the l-th zero-crossing point and the (l+1)-th zero-crossing point.
[0033]
[0034] (3.3.4) Acquire the sampled signal X ij Positive pulse width range and negative pulse width range:
[0035] Normal pulse width range:
[0036] Negative pulse width range:
[0037] (4) Multi-dimensional time-domain features are stored as template comparison parameters in the trigger detection module. Specifically, the extreme value range is stored as an extreme value template comparison parameter in the extreme value trigger detection module; the rise time and fall time range features are stored as rise time and fall time template comparison parameters in the rise time and fall time trigger detection module; and the positive and negative pulse width ranges are stored as pulse width template comparison parameters in the pulse width trigger detection module.
[0038] (5) Input the signal to be tested into a digital oscilloscope, acquire a sampled signal containing M sampling points, and then extract the feature values of the sampled signal, including the maximum value V. max The minimum value is V min Rise time T rise descent time T fall Positive pulse width W pos Negative pulse width W neg The extracted feature values are sent to the trigger detection module for template comparison. If all feature values are within the template comparison parameter range, the signal under test is determined to be normal; otherwise, the signal under test is determined to be abnormal.
[0039] The objective of this invention is achieved as follows:
[0040] This invention provides an anomaly detection method based on joint time-domain features. First, different types of signals are sampled, and their time-domain features, including extreme values, rise time, fall time, and pulse width, are extracted to form a comparison set for anomaly signal detection. Then, the signal to be tested is identified to realize an anomaly signal detection method based on joint time-domain features.
[0041] Furthermore, the anomaly detection method based on temporal joint features of this invention also has the following beneficial effects:
[0042] (1) Joint triggering of multi-dimensional features: By combining multiple time-domain features such as pulse width, extreme values, rise time, and fall time, joint triggering conditions can be set, which can more comprehensively describe the dynamic characteristics of the signal and improve the ability to capture abnormal signals.
[0043] (2) Enhanced anomaly detection accuracy: By triggering multiple features together, the false trigger rate caused by a single feature is effectively reduced, while the sensitivity and detection accuracy of complex anomaly signals are improved.
[0044] (3) Real-time performance and flexibility: Based on the oscilloscope, joint feature triggering is implemented, which can respond to signal changes in real time and adapt to various signal detection scenarios through flexible condition settings. Attached Figure Description
[0045] Figure 1 This is a flowchart of the anomaly detection method based on temporal joint features of the present invention;
[0046] Figure 2 This is a schematic diagram of a sinusoidal sampling signal;
[0047] Figure 3 This is a schematic diagram of an amplitude modulation sampling signal;
[0048] Figure 4 This is an anomaly detection flowchart;
[0049] Figure 5 This is a diagram showing the anomaly identification results for a sinusoidal signal;
[0050] Figure 6 This is a diagram showing the results of anomaly identification for amplitude modulation signals. Detailed Implementation
[0051] 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.
[0052] Example
[0053] In this embodiment, as Figure 1 As shown, the present invention provides an anomaly detection method based on temporal joint features, comprising the following steps:
[0054] (1) Deploy a trigger detection module in a digital oscilloscope. The trigger detection module specifically includes: an extreme value trigger detection module, a pulse width trigger detection module, and a rise time and fall time trigger detection module.
[0055] Among them, the extreme value trigger detection module is used to store the extreme value template comparison parameters of different types of input signals; the pulse width trigger detection module is used to store the pulse width template comparison parameters of different types of input signals; and the rise time and fall time trigger detection module is used to store the rise time and fall time template comparison parameters of different types of input signals.
[0056] (2) Signal acquisition;
[0057] Different types of input signals are sequentially input into a digital oscilloscope, which then acquires the input signals. 100 segments of each type of input signal are acquired. The j-th segment of the i-th type of input signal is denoted as X. ij X ij ={x ij1 ,x ij2 ,…,x ijk ,…,x ijM}, x ijk X represents ij There are k sampling points in the M, where k is the index of the sampling point and M is the number of sampling points;
[0058] In this embodiment, a digital oscilloscope is connected to a signal generator, which generates different types of input signals. Taking sinusoidal and amplitude-modulated signals as detection examples, the digital oscilloscope is used to acquire the sinusoidal and amplitude-modulated signals, and the steady-state high level of each sinusoidal sampling signal is recorded. steady-state low level The sampling interval time Δt = 0.001s for each sampling point was recorded; the steady-state high level of each amplitude-modulated sampling signal was recorded. steady-state low level And the sampling interval time Δt = 0.001s for each sampling point, we obtain Figure 2 The sinusoidal sampling signal shown and Figure 3 The amplitude modulation sampling signal shown.
[0059] (3) For each sampled signal X ij Perform multi-dimensional temporal feature extraction;
[0060] (3.1) Extracting extreme value features;
[0061] Traverse each sampled signal X ij Find all amplitudes that satisfy V from the sampling points in the range. ijk-1 <V ijk >V ijk+1 sampling point x ijk Store in the maximum value sequence Find all amplitudes that satisfy V ijk-1 >V ijk <V ijk+1The sampling points are stored in the minimum value sequence. Then, the extreme value range of each sampled signal segment is calculated:
[0062] Maximum value range:
[0063] Minimum range:
[0064] (3.2) Extracting the rise time and fall time features;
[0065] (3.2.1) Calculate the rise time in,
[0066] (3.2.2) Calculate the descent time in,
[0067] (3.2.3) For each sampled signal X ij Repeat the above steps to obtain N sets of rise times. and descent time
[0068] (3.2.4) Extract the rise time and fall time range of the i-th type of input signal;
[0069] Ascent time range:
[0070] Descent time range:
[0071] (3.3) Extracting positive and negative pulse width features;
[0072] (3.3.1) Set the sampling signal X for each segment ij threshold
[0073]
[0074] (3.3.2) Traverse each segment of the sampled signal X ij Find the sampling points in the middle that satisfy The sampling points are then identified and marked as zero-crossing points.
[0075] (3.3.3) Calculate the pulse width between two adjacent zero-crossing points;
[0076] If the sampled signal X ij The amplitude V at the l-th zero crossing point ijl satisfy: Then calculate the positive pulse width between the l-th zero-crossing point and the (l+1)-th zero-crossing point.
[0077]
[0078] in, This indicates that the l-th zero-crossing point is in the sampled signal X. ij The sampling point number in the sample;
[0079] If the sampled signal X ij The amplitude V at the l-th zero crossing point ijl satisfy: Then calculate the negative pulse width between the l-th zero-crossing point and the (l+1)-th zero-crossing point.
[0080]
[0081] (3.3.4) Acquire the sampled signal X ij Positive pulse width range and negative pulse width range:
[0082] Normal pulse width range:
[0083] Negative pulse width range:
[0084] (4) Multi-dimensional time-domain features are stored as template comparison parameters in the trigger detection module. Specifically, the extreme value range is stored as an extreme value template comparison parameter in the extreme value trigger detection module; the rise time and fall time range features are stored as rise time and fall time template comparison parameters in the rise time and fall time trigger detection module; and the positive and negative pulse width ranges are stored as pulse width template comparison parameters in the pulse width trigger detection module.
[0085] In this embodiment, the template comparison parameters obtained from the sinusoidal sampling signal are as follows:
[0086] Rise time range: [0.045, 0.055]
[0087] The time range for the descent is [0.046, 0.056].
[0088] Maximum value range: [0.95, 1.05]
[0089] The minimum value range is: [-1.05, -0.95]
[0090] Positive pulse width range: [0.095, 0.105]
[0091] Negative pulse width range: [0.095, 0.105]
[0092] The template comparison parameters for amplitude modulation sampling signals are as follows:
[0093] Rise time range: [0.008, 0.010]
[0094] The time range for the descent is [0.012, 0.016].
[0095] Maximum value range: [0.98, 1.07]
[0096] The minimum value range is: [-1.02, -0.96]
[0097] Positive pulse width range: [0.090, 0.103]
[0098] Negative pulse width range: [0.085, 0.101]
[0099] (5) Input the signal to be tested into a digital oscilloscope, acquire a sampled signal containing M sampling points, and then extract the feature values of the sampled signal, including the maximum value V. max The minimum value is V min Rise time T rise descent time T fall Positive pulse width W pos Negative pulse width W neg The extracted feature values are sent to the trigger detection module for template comparison. The detection process is as follows: Figure 4 As shown, if all feature values are within the template comparison parameter range, the signal under test is determined to be normal; otherwise, the signal under test is determined to be abnormal. The abnormal signal under test is buffered alternately using multiple FIFOs and sent to the host computer for display, thereby realizing the detection and capture of abnormal signals and realizing an abnormal signal detection method based on time-domain joint features.
[0100] In this embodiment, the comparison results of the feature values extracted from the sinusoidal sampling signal and the template parameters are as follows:
[0101] Maximum value: 1.2 (abnormal)
[0102] Minimum value: -0.98 (normal)
[0103] Rise time: 0.032s (abnormal)
[0104] Descent time: 0.025s (abnormal)
[0105] Normal pulse width: 0.099s
[0106] Negative pulse width: -0.098s (normal)
[0107] The comparison results of the feature values extracted from the amplitude-modulated sampling signal and the template parameters are as follows:
[0108] Rise time: 0.036s (abnormal)
[0109] Descent time: 0.052s (abnormal)
[0110] Maximum value: 0.26 (abnormal)
[0111] Minimum value: -0.15 (abnormal)
[0112] Normal pulse width: 0.99
[0113] Negative pulse width: 0.98 (normal)
[0114] Based on the comparison results, an abnormal signal can be detected. This abnormal signal is then sent to a subsequent FIFO buffer and transmitted to the host computer for display. The abnormal signal extracted from the sinusoidal sampled signal is shown below. Figure 5 Abnormal signals extracted from amplitude-modulated sampling signals, such as Figure 6 As shown.
[0115] 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 time-domain joint feature-based anomaly detection method, characterized in that, Includes the following steps: (1) Deploy a trigger detection module in a digital oscilloscope. The trigger detection module specifically includes: an extreme value trigger detection module, a pulse width trigger detection module, and a rise time and fall time trigger detection module. Among them, the extreme value trigger detection module is used to store the extreme value template comparison parameters of different types of input signals; the pulse width trigger detection module is used to store the pulse width template comparison parameters of different types of input signals; and the rise time and fall time trigger detection module is used to store the rise time and fall time template comparison parameters of different types of input signals. (2) Signal acquisition; The different types of input signals are sequentially input to the digital oscilloscope, and the input signals are collected by the digital oscilloscope The first segment of each type of input signal is collected The signal collected in the first segment of each type of input signal is recorded as , The sampling points in are represented by , the number index of the sampling points, and , the number of the sampling points recording each segment of the sampled signal a steady high level a steady low level and a sampling interval time of each sampling point ; (3) performing multi-dimension time domain feature extraction on each segment of the sampling signal (3.1) Extracting extreme value features; Traverse each sampled signal Find all amplitudes that satisfy the sampling points in the data. sampling points Store in the maximum value sequence Find all amplitudes that satisfy The sampling points are stored in the minimum value sequence. Then, the extreme value range of each sampled signal is calculated: Maximum interval: ; Minimum value interval: ; (3.2) Extract the rise time and fall time features; (3.2.1) Calculate the rise time ,in, = , = ; When the level is low, The voltage level rises 10% from low. This is the moment when the voltage rises 90% from a low level; (3.2.2), calculating the falling time wherein, = , = ; is the time at which the high level is 10% lower, is the time at which the high level is 90% lower, and is the time at which the high level is 10% lower. (3.2.3), for each segment of the sampled signal The above operation is repeated to obtain Group rise time And fall time , ; (3.2.4), extracting the rise time and fall time range characteristics of the input signal of the type Rising time range feature: ; Rising time range feature: ; (3.3) Extract positive and negative pulse width features; (3.3.1) setting a threshold value for each segment of the sampled signal ; ； (3.3.2), traversing each segment of the sampled signal to find the sample points that satisfy and then marking these sample points as zero-crossings. (3.3.3) Calculate the pulse width between two adjacent zero-crossing points; If the sampling signal The amplitude of the first zero-crossing point in the middle of the first half of the period of the sampling signal The amplitude of the first zero-crossing point in the middle of the first half of the period of the sampling signal Satisfies: The positive pulse width between the first zero-crossing point and the second zero-crossing point is calculated The positive pulse width between the first zero-crossing point and the second zero-crossing point is calculated The positive pulse width between the first zero-crossing point and the second zero-crossing point is calculated The positive pulse ； wherein represents the number of the zero-crossing point in the sampling points of the sampling signal ; If the sampling signal The Middle The amplitude at each zero point satisfy: Then calculate the first... The zero-crossing point and the first Negative pulse width between zero points ; ； (3.3.4) Acquiring the sampling signal Positive pulse width range and negative pulse width range: Normal pulse width range: ; Negative pulse width range: ; (4) Store the multi-dimensional time domain features as template comparison parameters in the trigger detection module. Specifically, store the extreme value range as extreme value template comparison parameters in the extreme value trigger detection module; store the rise time and fall time range features as rise time and fall time template comparison parameters in the rise time and fall time trigger detection module; and store the positive and negative pulse width ranges as pulse width template comparison parameters in the pulse width trigger detection module. (5) Input the signal to be tested into the digital oscilloscope and acquire the data including... The sampled signal is obtained from each sampling point, and then the feature values of the sampled signal are extracted, including the maximum value. Minimum value Rise time descent time Zhengmaikuan negative pulse width The extracted feature values are sent to the trigger detection module for template comparison. If all feature values are within the template comparison parameter range, the signal under test is determined to be normal; otherwise, the signal under test is determined to be abnormal.
Citation Information
Patent Citations
Continually responsive and anticipating automatic setup function for a digital oscilloscope
US6571185B1
Radio frequency power supply signal acquisition method and apparatus
WO2024032432A1