A fault diagnosis method and system for EMC test system

Abstract

The present invention relates to the field of EMC instrument control technology, and in particular to a fault diagnosis method and system for an EMC test system, wherein the method extracts historical feature vectors and real-time feature vectors of multiple monitoring time series data of a target EMC test system in the past period of time by means of a feature extraction window, and then calculates a fault risk value by the difference between these feature vectors, so that a fault diagnosis action can be performed. Compared with the prior art, the present invention determines the possibility of a preset abnormality by the difference between the historical feature vector and the real-time feature vector corresponding to the same preset abnormality, and calculates the fault risk value by the difference in the change of the feature vector sequence corresponding to different types of preset abnormalities, so that the EMC test system can autonomously identify faults by comparing the current data with the past data without relying on the preset rules, and solves the problem of how to diagnose the fault risk of unknown features outside the predefined rules.

Description

Technical Field

[0001] The present invention relates to the technical field of EMC instrument control, and in particular to a fault diagnosis method and system for an EMC test system. Background Art

[0002] An EMC (Electromagnetic Compatibility) test system is a specialized device used to evaluate the performance and stability of electronic equipment in an electromagnetic environment. The system can simulate real-world electromagnetic interference conditions to detect whether a device generates electromagnetic interference or is affected by external electromagnetic interference. EMC test systems typically include emission and anti-interference tests to ensure that the device does not interfere with other devices during operation, while also being able to resist electromagnetic interference from the external environment.

[0003] The fault diagnosis of EMC test system is of great significance. It can help engineers quickly locate faults in equipment and take effective measures to fix the problems. By analyzing the faults of EMC test system, engineers can identify electromagnetic compatibility problems in equipment and find out the root causes that may cause equipment to interfere with other equipment or be subject to external interference. This timely fault diagnosis helps to improve the stability and reliability of equipment and reduce production stagnation or product quality problems caused by electromagnetic compatibility problems.

[0004] However, the existing fault diagnosis methods of EMC test systems mostly rely on pre-defined rule sets to identify faults. When the system detects behavior that does not conform to the rules, it will issue an alarm or perform fault diagnosis. Although this method is simple and easy to understand, it relies too much on experience and cannot detect the risk of faults with unknown characteristics outside the pre-defined rules. Summary of the invention

[0005] Therefore, the present invention provides a fault diagnosis method and system for an EMC test system, so as to solve the problem in the prior art of how to diagnose the fault risk of unknown characteristics outside the predefined rules.

[0006] The present invention provides a fault diagnosis method for an EMC test system, comprising:

[0007] Acquire multiple monitoring time series data of the target EMC test system within a preset time period, set a feature extraction window on each monitoring time series data, and move the feature extraction window to traverse the monitoring time series data, wherein the end time of the preset time period is the current time;

[0008] When moving the feature extraction window, for each preset anomaly, feature extraction is performed based on the feature extraction window to obtain a feature vector sequence corresponding to each preset anomaly. In the feature vector sequence corresponding to each preset anomaly, the last feature vector is the real-time feature vector corresponding to the preset anomaly, and the other feature vectors are all historical feature vectors corresponding to the preset anomaly.

[0009] The fault risk value is calculated based on the difference between the historical feature vector and the real-time feature vector corresponding to the same type of preset anomaly, and the difference in the change of the feature vector sequence corresponding to different types of preset anomalies;

[0010] According to the fault risk value, a fault diagnosis action is performed.

[0011] In a further solution, the fault risk value is calculated based on the difference between the historical feature vector and the real-time feature vector corresponding to the same type of preset abnormality, and the difference between the change of the feature vector sequence corresponding to different types of preset abnormalities, including:

[0012] According to the difference between the historical feature vector and the real-time feature vector corresponding to the same preset abnormality, a first risk value corresponding to each preset abnormality is calculated, and the first risk value is proportional to the difference between the historical feature vector and the real-time feature vector corresponding to each preset abnormality;

[0013] Calculate a second risk value according to the change difference of the feature vector sequences corresponding to different types of preset anomalies, where the second risk value is inversely proportional to the change difference of the feature vector sequences corresponding to different types of preset anomalies;

[0014] A failure risk value is calculated according to the first risk value and the second risk value. The failure risk value is proportional to the first risk value and inversely proportional to the second risk value.

[0015] In a further solution, the first risk value corresponding to each preset abnormality is calculated according to the difference between the historical feature vector and the real-time feature vector corresponding to the same preset abnormality, including:

[0016] Taking Manhattan distance as the distance measurement method of feature vector, cluster the historical feature vectors and real-time feature vectors corresponding to the target preset anomaly based on DBSCAN algorithm to obtain the sample points corresponding to each feature vector after clustering, where the target preset anomaly is a preset anomaly to be analyzed.

[0017] If the sample point corresponding to the real-time feature vector is a noise point, a first preset value is obtained as a first risk value corresponding to the target preset anomaly;

[0018] If the sample point corresponding to the real-time feature vector is a boundary point, a second value is calculated as a first risk value corresponding to the target preset abnormality according to the density of the cluster where the sample point corresponding to the real-time feature vector is located;

[0019] If the sample point corresponding to the real-time feature vector is a core point, obtaining a third preset value as a first risk value corresponding to the target preset anomaly;

[0020] The first preset value is greater than the second value, and the second value is greater than the third preset value.

[0021] In a further solution, the step of calculating the second value as the first risk value corresponding to the target preset abnormality according to the density of the cluster where the sample point corresponding to the real-time feature vector is located includes:

[0022] Obtain the number of sample points corresponding to the historical feature vector directly reaching the sample point density corresponding to the real-time feature vector, and obtain a first density value;

[0023] Obtain the number of sample points corresponding to the historical feature vector that are densely connected to the sample points corresponding to the real-time feature vector to obtain a second density value;

[0024] According to the first density value and the second density value, the second value is calculated as the first risk value corresponding to the target preset abnormality, wherein the first density value is inversely proportional to the second value, and the second density value is directly proportional to the second value.

[0025] In a further solution, the step of calculating the second value as the first risk value corresponding to the target preset abnormality according to the first density value and the second density value includes:

[0026] The second value is calculated by the following formula:

[0027]

[0028] Among them, V is the second value, d1 is the first density value, d2 is the second density value, e is a natural constant, and a, b and c are different preset weights respectively.

[0029] In a further solution, the second risk value is calculated according to the change difference of the feature vector sequence corresponding to different types of preset abnormalities, including:

[0030] Fit the historical feature vector corresponding to each preset anomaly to obtain the hyperplane corresponding to each preset anomaly;

[0031] Calculate the distance between the real-time feature vector corresponding to each preset abnormality and the hyperplane corresponding to the preset abnormality to obtain the change feature value corresponding to each preset abnormality;

[0032] A second risk value is obtained according to the difference in the change characteristic values ​​corresponding to the multiple preset abnormalities, wherein the second risk value is inversely proportional to the difference in the change characteristic values ​​corresponding to the multiple preset abnormalities.

[0033] In a further solution, obtaining the second risk value according to the difference of the change characteristic values ​​corresponding to the plurality of preset abnormalities includes:

[0034] The standard deviation of the change characteristic values ​​corresponding to the various preset anomalies is calculated as the second risk value.

[0035] In a further solution: the calculating the fault risk value according to the first risk value and the second risk value includes:

[0036] The failure risk value is calculated by the following formula:

[0037]

[0038] Wherein, R is the fault risk value, r1 is the first risk value, r2 is the second risk value, ω1 and ω2 are different preset adjustment parameters.

[0039] In a further solution: the preset abnormality includes electromagnetic radiation abnormality, conduction interference abnormality and signal abnormality, wherein:

[0040] The monitoring time series data corresponding to the electromagnetic radiation anomaly includes electromagnetic field intensity data, and the elements in the feature vector corresponding to the electromagnetic radiation anomaly include: the main frequency components, mean, variance, maximum value and minimum value of the electromagnetic field intensity data within the feature extraction window;

[0041] The monitoring time series data corresponding to the conduction interference anomaly includes current and voltage data, and the elements in the feature vector corresponding to the conduction interference anomaly include: the mean, variance and peak value of the current and voltage data within the feature extraction window;

[0042] The monitoring time series data corresponding to the signal anomaly includes signal waveform data, and the elements in the feature vector corresponding to the signal anomaly include: the main frequency components, period, waveform average rise time and amplitude modulation index of the signal waveform data within the feature extraction window.

[0043] The present invention also provides a fault diagnosis system for an EMC test system, comprising:

[0044] A data monitoring module is used to obtain a variety of monitoring time series data of the target EMC test system within a preset time period, set a feature extraction window on each monitoring time series data, and move the feature extraction window to traverse the monitoring time series data, wherein the end time of the preset time period is the current time;

[0045] A feature extraction module is used to extract features based on the feature extraction window for each preset anomaly when moving the feature extraction window, and obtain a feature vector sequence corresponding to each preset anomaly. In the feature vector sequence corresponding to each preset anomaly, the last feature vector is the real-time feature vector corresponding to the preset anomaly, and the other feature vectors are all historical feature vectors corresponding to the preset anomaly;

[0046] The abnormality analysis module is used to calculate the fault risk value based on the difference between the historical feature vector and the real-time feature vector corresponding to the same type of preset abnormality, and the difference in the change of the feature vector sequence corresponding to different types of preset abnormalities;

[0047] The fault diagnosis module is used to perform fault diagnosis actions according to the fault risk value.

[0048] The beneficial effects of adopting the above embodiment are:

[0049] The present invention provides a fault diagnosis method and system for an EMC test system, wherein the method extracts historical feature vectors and real-time feature vectors of multiple monitoring time series data of a target EMC test system in the past period of time by means of a feature extraction window, and then obtains a calculated fault risk value that can represent the probability of a current fault by means of the difference between the historical feature vector and the real-time feature vector corresponding to the same type of preset abnormality, and the difference in the change of the feature vector sequence corresponding to different types of preset abnormalities, and then the fault diagnosis action can be performed. Compared with the prior art, the present invention determines the possibility of a preset abnormality by means of the difference between the historical feature vector and the real-time feature vector corresponding to the same type of preset abnormality, and at the same time takes into account the potential association between different types of abnormalities, and further calculates the fault risk value by means of the difference in the change of the feature vector sequence corresponding to different types of preset abnormalities, so that the EMC test system can autonomously identify faults to a certain extent by comparing the current data with the past data without relying on some preset rules, and solves the problem of how to diagnose the fault risk of unknown features outside the predefined rules. BRIEF DESCRIPTION OF THE DRAWINGS

[0050] Figure 1 A method flow chart of a fault diagnosis method for an EMC test system provided by the present invention;

[0051] Figure 2 for Figure 1 Specific method flow chart of step S103;

[0052] Figure 3 for Figure 2 Specific method flow chart of step S201;

[0053] Figure 4 for Figure 3Specific method flow chart of step S303;

[0054] Figure 5 for Figure 2 Specific method flow chart of step S202;

[0055] Figure 6 This is a system structure diagram of the fault diagnosis system of the EMC test system provided by the present invention. DETAILED DESCRIPTION

[0056] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0057] Combination Figure 1 As shown, a specific embodiment of the present invention discloses a fault diagnosis method for an EMC test system, comprising:

[0058] S101, obtaining a plurality of monitoring time series data of a target EMC test system within a preset time period, setting a feature extraction window on each monitoring time series data, and moving the feature extraction window to traverse the monitoring time series data, wherein the end time of the preset time period is the current time;

[0059] S102. When moving the feature extraction window, for each preset abnormality, feature extraction is performed based on the feature extraction window to obtain a feature vector sequence corresponding to each preset abnormality. In the feature vector sequence corresponding to each preset abnormality, the last feature vector is a real-time feature vector corresponding to the preset abnormality, and the other feature vectors are all historical feature vectors corresponding to the preset abnormality.

[0060] S103, calculating the fault risk value according to the difference between the historical feature vector and the real-time feature vector corresponding to the same type of preset abnormality, and the difference between the change of the feature vector sequence corresponding to different types of preset abnormalities;

[0061] S104: Execute a fault diagnosis action according to the fault risk value.

[0062] In the prior art, faults are usually identified based on a set of predefined rules, such as detecting the value of voltage. When the voltage is higher than a certain threshold, it is considered that a fault has occurred. However, this method is particularly dependent on the formulation of rules. For example, the selection of a specific threshold in the above process cannot be used to identify the risk of faults with some unknown characteristics. Compared with the prior art, the present invention determines the possibility of a preset abnormality by the difference between the historical feature vector and the real-time feature vector corresponding to the same preset abnormality, while taking into account the potential correlation between different types of abnormalities, and further calculates the fault risk value by the difference in the change of the feature vector sequence corresponding to different types of preset abnormalities. This enables the EMC test system to autonomously identify faults to a certain extent by comparing the current data with the past data without relying on some preset rules, thereby solving the problem of how to diagnose the risk of faults with unknown characteristics outside the predefined rules.

[0063] Specifically, in the above step S101, the preset time period refers to the time period of the most recent period with the current moment as the end moment. The monitoring time series data refers to the data obtained within the preset time period that can be organized based on the time sequence, such as the change data of current and voltage, signal waveform data, etc. The feature extraction window is a sliding window set on the monitoring time series data. The data in each window is used to extract a feature vector. As the feature extraction window can slide overlappingly in the monitoring time series data, multiple feature vectors arranged in time sequence can be extracted. Obviously, in this embodiment, because the end time of the preset time period is the current moment, when the feature extraction window slides to the end, the feature vector obtained can be regarded as a feature vector that can reflect the characteristics of the data at the current moment, and it can be regarded as real-time data.

[0064] Furthermore, in a preferred embodiment, in the above step S102, the preset abnormality includes electromagnetic radiation abnormality, conduction interference abnormality and signal abnormality, wherein:

[0065] The monitoring time series data corresponding to the electromagnetic radiation anomaly includes electromagnetic field intensity data, and the elements in the feature vector corresponding to the electromagnetic radiation anomaly include: the main frequency components, mean, variance, maximum value and minimum value of the electromagnetic field intensity data within the feature extraction window;

[0066] The monitoring time series data corresponding to the conduction interference anomaly includes current and voltage data, and the elements in the feature vector corresponding to the conduction interference anomaly include: the mean, variance and peak value of the current and voltage data within the feature extraction window;

[0067] The monitoring time series data corresponding to the signal anomaly includes signal waveform data, and the elements in the feature vector corresponding to the signal anomaly include: the main frequency components, period, waveform average rise time and amplitude modulation index of the signal waveform data within the feature extraction window.

[0068] It should be noted that, since the purpose of this embodiment is to detect the risk of failure of unknown characteristics, the final fault type of the EMC test system is positional, so the above-mentioned preset abnormality is not the type of fault to be diagnosed in the end, but is only used as an evaluation indicator to enable people to analyze the change of data through a known standard. It can be understood that the above-mentioned preset abnormality and the corresponding feature vector structure are only a preferred implementation method. In practice, other types of preset abnormalities and corresponding feature vector structures can also be selected and formulated according to specific circumstances.

[0069] Further, combined with Figure 2 As shown, in a preferred embodiment, the above step S103, calculating the fault risk value according to the difference between the historical feature vector and the real-time feature vector corresponding to the same type of preset abnormality, and the difference in the change of the feature vector sequence corresponding to different types of preset abnormalities, specifically includes:

[0070] S201, calculating a first risk value corresponding to each preset abnormality according to the difference between the historical feature vector and the real-time feature vector corresponding to the same preset abnormality, wherein the first risk value is proportional to the difference between the historical feature vector and the real-time feature vector corresponding to each preset abnormality;

[0071] S202, calculating a second risk value according to the change difference of the feature vector sequences corresponding to different types of preset abnormalities, where the second risk value is inversely proportional to the change difference of the feature vector sequences corresponding to different types of preset abnormalities;

[0072] S203: Calculate a fault risk value according to the first risk value and the second risk value. The fault risk value is proportional to the first risk value and inversely proportional to the second risk value.

[0073] The above process reflects the possible failure probability of each preset abnormality through the first risk value. For example, in the above electromagnetic radiation abnormality, in the data represented by the real-time feature vector, unexpected frequency components appear in the spectrum corresponding to the electromagnetic field intensity data, which is significantly different from the historical feature vector. In this case, the higher the probability of failure of the target EMC test system, the greater the first risk value.

[0074] Because in practice, different preset exceptions may affect each other. For example, for the multiple preset exceptions in the above embodiment:

[0075] Abnormal electromagnetic radiation is associated with abnormal conducted interference: The electromagnetic radiation generated by the device may interfere with other devices or systems through the conduction path, so there may be a certain correlation between abnormal radiation and abnormal conducted interference. When the radiation field strength of the device exceeds the specified limit, it may cause an abnormal increase in voltage and current in the conducting medium, thereby causing conducted interference.

[0076] Abnormal electromagnetic radiation is associated with abnormal signals: Abnormal electromagnetic radiation of a device may cause abnormal output signals of the device. For example, interference signals in the radiation spectrum may affect the transmission of normal signals of the device, resulting in abnormal signals.

[0077] Abnormal conduction interference is associated with abnormal signals: abnormal conduction interference may affect the quality of the output signal of the equipment. For example, abnormal voltage and current in the conductive medium may cause distortion of the signal waveform, thereby causing signal abnormality.

[0078] So similarly, the above process reflects the similarity of data changes between different types of preset anomalies through the second risk value. When the real-time feature vectors corresponding to multiple preset anomalies simultaneously undergo drastic changes compared to the historical feature vectors (that is, the smaller the difference in the change of the feature vector sequences corresponding to different types of preset anomalies), it can be considered that the probability of failure of the target EMC test system is higher and the second risk value is larger. Then, by combining the considerations of the first risk value and the second risk value, and referring to the meaning and numerical values ​​of the two, a failure risk value that can comprehensively reflect the current failure rate of the target EMC test system can be obtained.

[0079] Specifically, combined Figure 3 As shown, in a preferred embodiment, the above step S201, according to the difference between the historical feature vector and the real-time feature vector corresponding to the same type of preset abnormality, calculates the first risk value corresponding to each preset abnormality, specifically including:

[0080] S301, using Manhattan distance as a distance measurement method for feature vectors, clustering historical feature vectors and real-time feature vectors corresponding to a target preset anomaly based on a DBSCAN algorithm to obtain sample points corresponding to each feature vector after clustering, wherein the target preset anomaly is a preset anomaly to be analyzed currently;

[0081] S302: If the sample point corresponding to the real-time feature vector is a noise point, obtaining a first preset value as a first risk value corresponding to the target preset abnormality;

[0082] S303: If the sample point corresponding to the real-time feature vector is a boundary point, a second value is calculated as a first risk value corresponding to the target preset abnormality according to the density of the cluster where the sample point corresponding to the real-time feature vector is located;

[0083] S304: If the sample point corresponding to the real-time feature vector is a core point, obtaining a third preset value as a first risk value corresponding to the target preset abnormality;

[0084] The first preset value is greater than the second value, and the second value is greater than the third preset value.

[0085] In the above process, DBSCAN (Density-Based Spatial Clustering of Applications with Noise) is a density-based clustering algorithm. Its uniqueness lies in that it does not need to pre-specify the number of clusters, can automatically discover clusters of any shape, and has good robustness to noise data. The algorithm constructs clusters by identifying core points and their density accessibility. First, the core points are determined, and then the clusters are gradually expanded through density accessibility. DBSCAN is more suitable for processing data sets with large density differences. It has high practicality and scalability. It is widely used because it does not need to know the number of clusters in advance. Compared with other clustering algorithms, the DBSCAN algorithm is more suitable for this application scenario. When the DBSCAN algorithm is applied to this embodiment, it has better interpretability. For example, when the sample point corresponding to the real-time feature vector is a noise point, it can be considered that the real-time feature vector is significantly different from other historical feature vectors (for example, a certain indicator such as voltage and frequency has a mutation at the current moment), and the higher the fault probability it represents. Similarly, when the sample point corresponding to the real-time feature vector is a core point, it can be considered that the real-time feature vector is similar to other historical feature vectors (for example, the voltage and frequency indicators are not much different from the normal level), and the fault probability it represents is lower. When the sample point corresponding to the real-time feature vector is a boundary point, the fault probability it reflects can be further obtained based on the clustering result, that is, the density of the cluster where the sample point corresponding to the real-time feature vector is located, so as to obtain a more scientific and reasonable result.

[0086] In addition, in this embodiment, Manhattan distance is selected as the distance measurement method between feature vectors. Compared with other conventional distance measurement methods (such as cosine similarity, etc.), this embodiment can greatly shorten the calculation amount of the distance between different feature vectors, thereby improving the operation speed of the entire method. It can be understood that the above-mentioned Manhattan distance and the specific clustering process of the DBSCAN algorithm are both prior arts that can be understood by those skilled in the art. Similarly, the meanings of noise points, core points, boundary points, and the density direct access, density connection, etc. mentioned later are also common knowledge in the art, so this article will not make too much explanation.

[0087] Further, combined with Figure 4 As shown, in a preferred embodiment, in the above step S303, according to the density of the cluster where the sample point corresponding to the real-time feature vector is located, calculating the second value as the first risk value corresponding to the target preset abnormality specifically includes:

[0088] S401, obtaining the number of sample points corresponding to the historical feature vector directly reaching the sample point density corresponding to the real-time feature vector, and obtaining a first density value;

[0089] S402, obtaining the number of sample points corresponding to the historical feature vector that are densely connected to the sample points corresponding to the real-time feature vector, and obtaining a second density value;

[0090] S403. Calculate the second value as the first risk value corresponding to the target preset abnormality according to the first density value and the second density value, wherein the first density value is inversely proportional to the second value, and the second density value is directly proportional to the second value.

[0091] In the above process, the number of sample points corresponding to the historical feature vectors directly connected to the sample point density corresponding to the real-time feature vector can be understood as the number of core points around the sample point corresponding to the real-time feature vector (it can also be understood as the number of clusters including the sample point corresponding to the real-time feature vector). Obviously, the more the number, the more clusters the real-time feature vector is accommodated by. The more conventional the real-time feature vector is, the lower the probability of failure reflected by the real-time feature vector is. The number of sample points corresponding to the historical feature vectors connected to the sample point density corresponding to the real-time feature vector can be understood as the number of non-core points in the same cluster as the sample point corresponding to the real-time feature vector. The more the number, the more feature vectors that are significantly different from other historical feature vectors but have a high degree of similarity to the real-time feature vector. At this time, the more unconventional feature vectors are reflected (for example, if the temperature of the target EMC test system suddenly rises and lasts for a period of time, the above situation will be obtained), and the higher the probability of failure reflected by the real-time feature vector is. This embodiment uses the above two density values ​​to obtain a scientific and reasonable second numerical value.

[0092] Further, in a preferred embodiment, the above step S403, calculating the second value as the first risk value corresponding to the target preset abnormality according to the first density value and the second density value, specifically includes:

[0093] The second value is calculated by the following formula:

[0094]

[0095] Wherein, V is the second value, d1 is the first density value, d2 is the second density value, e is a natural constant, and a, b and c are different preset weights. The preset weights may be preset values ​​obtained based on experience.

[0096] The significance of the above formula lies not only in that the first density value and the second density value are weightedly calculated according to the increase and decrease relationship mentioned above to obtain a second value, but also in that: on the one hand, the natural constant in the first term used in the formula ensures the order of magnitude of the second value, so that the final result easily meets the condition that the first preset value is greater than the second value, and the second value is greater than the third preset value; on the other hand, the formula reflects the relationship between the first density value and the second density value through the third term, so as to achieve a more comprehensive and accurate result.

[0097] Further, combined with Figure 5 As shown, in a preferred embodiment, the above step S202, calculating the second risk value according to the change difference of the feature vector sequence corresponding to different types of preset abnormalities, includes:

[0098] S501, fitting the historical feature vector corresponding to each preset anomaly to obtain a hyperplane corresponding to each preset anomaly;

[0099] S502, calculating the distance between the real-time feature vector corresponding to each preset abnormality and the hyperplane corresponding to the preset abnormality, and obtaining the change feature value corresponding to each preset abnormality;

[0100] S503. Obtain a second risk value according to differences in the change characteristic values ​​corresponding to the plurality of preset abnormalities, wherein the second risk value is inversely proportional to the differences in the change characteristic values ​​corresponding to the plurality of preset abnormalities.

[0101] Although both analyze the differences between feature vectors, it is different from calculating the first risk value. In this embodiment, a method of fitting multiple high-dimensional vectors to obtain a hyperplane (which can be fitted by existing methods such as support vector machines and multivariate regression) is used to represent the change trend of the historical feature vector, and then the distance from the real-time feature vector to the hyperplane is used to represent the degree of change of the data at the current moment. Compared with the clustering method in the previous text, this embodiment analyzes the difference characteristics of data changes from another perspective, thereby improving accuracy. After obtaining the change feature value corresponding to each preset abnormality, the difference of the change feature values ​​corresponding to multiple preset abnormalities (which can be represented by statistical features such as variance, the greater the discrete degree between multiple change feature values, the greater the difference between multiple change feature values, and the lower the probability of the preset abnormalities occurring at the same time) can be used to reflect whether multiple preset abnormalities occur at the same time. The more preset abnormalities occur at the same time, the lower the probability of data abnormalities caused by data jumps, and the higher the probability that the target EMC test system actually fails.

[0102] Furthermore, in a preferred embodiment, obtaining the second risk value according to the difference of the change characteristic values ​​corresponding to the multiple preset abnormalities includes:

[0103] The standard deviation of the change characteristic values ​​corresponding to the various preset anomalies is calculated as the second risk value.

[0104] It is understandable that in practice, other methods may be selected to evaluate the difference between the change characteristic values ​​to obtain the second risk value.

[0105] Further, in a preferred embodiment, the above step S203, calculating the fault risk value according to the first risk value and the second risk value, specifically includes:

[0106] The failure risk value is calculated by the following formula:

[0107]

[0108] Wherein, R is the fault risk value, r1 is the first risk value, r2 is the second risk value, ω1 and ω2 are different preset adjustment parameters.

[0109] The significance of the above formula is to calculate the sum of the first risk values ​​corresponding to multiple preset anomalies, and combine the second risk value to obtain a more accurate fault risk value. In practice, the fault risk value can also be calculated by comparing the first risk value with the second risk value. However, the weighted summation method used in this embodiment can highlight the influence of the two risk values ​​and obtain a more universal result.

[0110] After the above process is completed, step S104 can be performed to perform a fault diagnosis action according to the fault risk value. For example, when the fault risk value exceeds a certain threshold, a fault diagnosis action can be performed. It should be noted that the fault diagnosis action can be any action such as shutting down the machine to run a self-check program, issuing a fault alarm prompt, etc.

[0111] Combination Figure 6 As shown, the present invention also provides a fault diagnosis system for an EMC test system, comprising:

[0112] The data monitoring module 610 is used to obtain a plurality of monitoring time series data of the target EMC test system within a preset time period, set a feature extraction window on each monitoring time series data, and move the feature extraction window to traverse the monitoring time series data, wherein the end time of the preset time period is the current time;

[0113] The feature extraction module 620 is used to extract features based on the feature extraction window for each preset abnormality when the feature extraction window is moved, and obtain a feature vector sequence corresponding to each preset abnormality. In the feature vector sequence corresponding to each preset abnormality, the last feature vector is a real-time feature vector corresponding to the preset abnormality, and the other feature vectors are all historical feature vectors corresponding to the preset abnormality;

[0114] The abnormality analysis module 630 is used to calculate the fault risk value according to the difference between the historical feature vector and the real-time feature vector corresponding to the same type of preset abnormality, and the difference between the change of the feature vector sequence corresponding to different types of preset abnormalities;

[0115] The fault diagnosis module 640 is used to perform a fault diagnosis action according to the fault risk value.

[0116] It should be noted here that the fault diagnosis system of the corresponding EMC test system provided in the above embodiment can implement the technical solutions described in the above method embodiments. The specific implementation principles of the above modules or units can refer to the corresponding contents in the above method embodiments, which will not be repeated here.

[0117] The present invention provides a fault diagnosis method and system for an EMC test system, wherein the method extracts historical feature vectors and real-time feature vectors of multiple monitoring time series data of a target EMC test system in the past period of time by means of a feature extraction window, and then obtains a calculated fault risk value that can represent the probability of a current fault by means of the difference between the historical feature vector and the real-time feature vector corresponding to the same type of preset abnormality, and the difference in the change of the feature vector sequence corresponding to different types of preset abnormalities, and then the fault diagnosis action can be performed. Compared with the prior art, the present invention determines the possibility of a preset abnormality by means of the difference between the historical feature vector and the real-time feature vector corresponding to the same type of preset abnormality, and at the same time takes into account the potential association between different types of abnormalities, and further calculates the fault risk value by means of the difference in the change of the feature vector sequence corresponding to different types of preset abnormalities, so that the EMC test system can autonomously identify faults to a certain extent by comparing the current data with the past data without relying on some preset rules, and solves the problem of how to diagnose the fault risk of unknown features outside the predefined rules.

[0118] It should be noted that the various embodiments in this specification are described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The same or similar parts between the various embodiments can be referenced to each other.

[0119] The above description of the disclosed embodiments enables one skilled in the art to implement or use the present invention. Various modifications to these embodiments will be apparent to one skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention will not be limited to the embodiments shown herein, but rather to the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A fault diagnosis method for an EMC test system, characterized in that: include: Acquire multiple monitoring time series data of the target EMC test system within a preset time period, set a feature extraction window on each monitoring time series data, and move the feature extraction window to traverse the monitoring time series data, wherein the end time of the preset time period is the current time; When moving the feature extraction window, for each preset anomaly, feature extraction is performed based on the feature extraction window to obtain a feature vector sequence corresponding to each preset anomaly. In the feature vector sequence corresponding to each preset anomaly, the last feature vector is the real-time feature vector corresponding to the preset anomaly, and the other feature vectors are all historical feature vectors corresponding to the preset anomaly. The fault risk value is calculated based on the difference between the historical feature vector and the real-time feature vector corresponding to the same type of preset anomaly, and the difference in the change of the feature vector sequence corresponding to different types of preset anomalies; Execute fault diagnosis actions according to the fault risk value; The calculation of the fault risk value according to the difference between the historical feature vector and the real-time feature vector corresponding to the same type of preset abnormality and the difference between the change of the feature vector sequence corresponding to different types of preset abnormalities includes: According to the difference between the historical feature vector and the real-time feature vector corresponding to the same preset abnormality, a first risk value corresponding to each preset abnormality is calculated, and the first risk value is proportional to the difference between the historical feature vector and the real-time feature vector corresponding to each preset abnormality; Calculate a second risk value according to the change difference of the feature vector sequences corresponding to different types of preset anomalies, where the second risk value is inversely proportional to the change difference of the feature vector sequences corresponding to different types of preset anomalies; A fault risk value is calculated according to the first risk value and the second risk value, wherein the fault risk value is proportional to the first risk value and inversely proportional to the second risk value; The step of calculating the first risk value corresponding to each preset abnormality according to the difference between the historical feature vector and the real-time feature vector corresponding to the same preset abnormality includes: Taking Manhattan distance as the distance measurement method of feature vector, cluster the historical feature vectors and real-time feature vectors corresponding to the target preset anomaly based on DBSCAN algorithm to obtain the sample points corresponding to each feature vector after clustering, where the target preset anomaly is a preset anomaly to be analyzed. If the sample point corresponding to the real-time feature vector is a noise point, a first preset value is obtained as a first risk value corresponding to the target preset abnormality; If the sample point corresponding to the real-time feature vector is a boundary point, a second value is calculated as a first risk value corresponding to the target preset abnormality according to the density of the cluster where the sample point corresponding to the real-time feature vector is located; If the sample point corresponding to the real-time feature vector is a core point, obtaining a third preset value as a first risk value corresponding to the target preset anomaly; Wherein, the first preset value is greater than the second value, and the second value is greater than the third preset value; The step of calculating the second risk value according to the change difference of the feature vector sequences corresponding to different types of preset anomalies includes: Fit the historical feature vector corresponding to each preset anomaly to obtain the hyperplane corresponding to each preset anomaly; Calculate the distance between the real-time feature vector corresponding to each preset abnormality and the hyperplane corresponding to the preset abnormality to obtain the change feature value corresponding to each preset abnormality; A second risk value is obtained according to the difference in the change characteristic values ​​corresponding to the multiple preset abnormalities, wherein the second risk value is inversely proportional to the difference in the change characteristic values ​​corresponding to the multiple preset abnormalities.

2. The fault diagnosis method of the EMC test system according to claim 1, characterized in that: The calculating, according to the density of the cluster where the sample point corresponding to the real-time feature vector is located, the second value as the first risk value corresponding to the target preset abnormality includes: Obtain the number of sample points corresponding to the historical feature vector directly reaching the sample point density corresponding to the real-time feature vector, and obtain a first density value; Obtain the number of sample points corresponding to the historical feature vector that are densely connected to the sample points corresponding to the real-time feature vector to obtain a second density value; According to the first density value and the second density value, the second value is calculated as the first risk value corresponding to the target preset abnormality, wherein the first density value is inversely proportional to the second value, and the second density value is directly proportional to the second value.

3. The fault diagnosis method of the EMC test system according to claim 2, characterized in that: The step of calculating the second value as the first risk value corresponding to the target preset abnormality according to the first density value and the second density value includes: The second value is calculated by the following formula: Among them, V is the second value, d1 is the first density value, d2 is the second density value, e is a natural constant, and a, b and c are different preset weights respectively.

4. The fault diagnosis method of the EMC test system according to claim 1, characterized in that: The second risk value is obtained according to the difference of the change characteristic values ​​corresponding to the multiple preset abnormalities, including: The standard deviation of the change characteristic values ​​corresponding to the various preset anomalies is calculated as the second risk value.

5. The fault diagnosis method of the EMC test system according to claim 1, characterized in that: The calculating the fault risk value according to the first risk value and the second risk value comprises: The failure risk value is calculated by the following formula: Wherein, R is the fault risk value, r1 is the first risk value, r2 is the second risk value, ω1 and ω2 are different preset adjustment parameters.

6. The fault diagnosis method of the EMC test system according to claim 1, characterized in that: The preset anomalies include electromagnetic radiation anomalies, conduction interference anomalies and signal anomalies, among which: The monitoring time series data corresponding to the electromagnetic radiation anomaly includes electromagnetic field intensity data, and the elements in the feature vector corresponding to the electromagnetic radiation anomaly include: the main frequency components, mean, variance, maximum value and minimum value of the electromagnetic field intensity data within the feature extraction window; The monitoring time series data corresponding to the conduction interference anomaly includes current and voltage data, and the elements in the feature vector corresponding to the conduction interference anomaly include: the mean, variance and peak value of the current and voltage data within the feature extraction window; The monitoring time series data corresponding to the signal anomaly includes signal waveform data, and the elements in the feature vector corresponding to the signal anomaly include: the main frequency components, period, waveform average rise time and amplitude modulation index of the signal waveform data within the feature extraction window.

7. A fault diagnosis system for an EMC test system, characterized in that: include: A data monitoring module is used to obtain a variety of monitoring time series data of the target EMC test system within a preset time period, set a feature extraction window on each monitoring time series data, and move the feature extraction window to traverse the monitoring time series data, wherein the end time of the preset time period is the current time; A feature extraction module is used to extract features based on the feature extraction window for each preset anomaly when moving the feature extraction window, and obtain a feature vector sequence corresponding to each preset anomaly. In the feature vector sequence corresponding to each preset anomaly, the last feature vector is the real-time feature vector corresponding to the preset anomaly, and the other feature vectors are all historical feature vectors corresponding to the preset anomaly; The abnormality analysis module is used to calculate the fault risk value based on the difference between the historical feature vector and the real-time feature vector corresponding to the same type of preset abnormality, and the difference in the change of the feature vector sequence corresponding to different types of preset abnormalities; A fault diagnosis module is used to perform fault diagnosis actions according to the fault risk value; The calculation of the fault risk value according to the difference between the historical feature vector and the real-time feature vector corresponding to the same type of preset abnormality and the difference between the change of the feature vector sequence corresponding to different types of preset abnormalities includes: According to the difference between the historical feature vector and the real-time feature vector corresponding to the same preset abnormality, a first risk value corresponding to each preset abnormality is calculated, and the first risk value is proportional to the difference between the historical feature vector and the real-time feature vector corresponding to each preset abnormality; Calculate a second risk value according to the change difference of the feature vector sequences corresponding to different types of preset anomalies, where the second risk value is inversely proportional to the change difference of the feature vector sequences corresponding to different types of preset anomalies; A fault risk value is calculated according to the first risk value and the second risk value, wherein the fault risk value is proportional to the first risk value and inversely proportional to the second risk value; The step of calculating the first risk value corresponding to each preset abnormality according to the difference between the historical feature vector and the real-time feature vector corresponding to the same preset abnormality includes: Taking Manhattan distance as the distance measurement method of feature vector, cluster the historical feature vectors and real-time feature vectors corresponding to the target preset anomaly based on DBSCAN algorithm to obtain the sample points corresponding to each feature vector after clustering, where the target preset anomaly is a preset anomaly to be analyzed. If the sample point corresponding to the real-time feature vector is a noise point, a first preset value is obtained as a first risk value corresponding to the target preset abnormality; If the sample point corresponding to the real-time feature vector is a boundary point, a second value is calculated as a first risk value corresponding to the target preset abnormality according to the density of the cluster where the sample point corresponding to the real-time feature vector is located; If the sample point corresponding to the real-time feature vector is a core point, obtaining a third preset value as a first risk value corresponding to the target preset anomaly; Wherein, the first preset value is greater than the second value, and the second value is greater than the third preset value; The step of calculating the second risk value according to the change difference of the feature vector sequences corresponding to different types of preset anomalies includes: Fit the historical feature vector corresponding to each preset anomaly to obtain the hyperplane corresponding to each preset anomaly; Calculate the distance between the real-time feature vector corresponding to each preset abnormality and the hyperplane corresponding to the preset abnormality to obtain the change feature value corresponding to each preset abnormality; The second risk value is obtained according to the difference of the change characteristic values ​​corresponding to the multiple preset abnormalities, wherein: The second risk value is inversely proportional to the difference in the change characteristic values ​​corresponding to the multiple preset abnormalities.

Images