Fault prediction method and system based on mutual information compared with correlation coefficient

Abstract

The application discloses a fault prediction method and system based on mutual information compared with a correlation coefficient, and relates to the technical field of electronic measurement. The method has the advantages of small overhead and superior performance, and does not require additional hardware, training data, or understanding of system structure and device parameters. The method comprises the following steps: collecting multi-dimensional parameters output by a control system; performing smoothing filtering processing on the multi-dimensional parameters to obtain filtered multi-dimensional parameters; calculating the maximum mutual information between the filtered multi-dimensional parameters; performing attribute set division on the multi-dimensional parameters according to the maximum mutual information to obtain a plurality of attribute sets; calculating the average maximum mutual information of the attribute sets; and performing fault identification according to the average maximum mutual information. The application is suitable for predicting faults that will occur in the multi-dimensional parameters output by the control system and mainly aims at attribute set fault prediction.

Description

Technical Field

[0001] This invention relates to the field of electronic measurement technology. Background Technology

[0002] Existing multi-dimensional parameter fault prediction methods for control systems can be divided into four categories:

[0003] I. Fault prediction based on traditional statistical methods: such as linear discriminant analysis and principal component analysis. These methods are computationally simple, but require a large amount of historical data and have poor fitting effect on the nonlinear and complex relationships of the system.

[0004] II. Fault prediction based on machine learning: such as support vector machines, random forests, and neural networks. These methods can handle nonlinear relationships and perform well, but require a large amount of labeled sample data for training.

[0005] III. Fault prediction based on deep learning: such as convolutional neural networks, recurrent neural networks, and deep belief networks. These methods can directly learn features from the raw data and are suitable for small samples, but they require high computational resources.

[0006] IV. Multi-source data fusion forecasting: By comprehensively utilizing multi-source monitoring data, a more comprehensive health assessment index can be constructed to improve forecast accuracy. However, it is necessary to address the heterogeneity between different data sources.

[0007] Therefore, there is an urgent need for a fault prediction method with low computational overhead and superior performance, which requires no additional hardware, no training data, and no knowledge of system structure and device parameters. Summary of the Invention

[0008] This invention provides a fault prediction method based on mutual information and correlation coefficient, which has the advantages of low overhead and superior performance, and requires no additional hardware, no training data, and no knowledge of system structure and device parameters.

[0009] To achieve the above objectives, the present invention provides the following solution:

[0010] This invention provides a fault prediction method based on mutual information and correlation coefficient, the method being:

[0011] S1. Acquire multidimensional parameters output by the control system;

[0012] S2. Perform smoothing filtering on the multidimensional parameters to obtain the filtered multidimensional parameters;

[0013] S3. Calculate the maximum mutual information between the filtered multidimensional parameters;

[0014] S4. Divide the multidimensional parameters into attribute sets according to the maximum mutual information to obtain several attribute sets;

[0015] S5. Calculate the average maximum mutual information of the attribute set;

[0016] S6. Identify faults based on the average maximum mutual information.

[0017] Furthermore, in a preferred embodiment, the smoothing filtering process in step S2 above is implemented using a moving average filter.

[0018] Furthermore, in a preferred embodiment, the maximum mutual information I in step S3 above... xy Represented as:

[0019] I xy ,{x,y}∈{1,2...18}.

[0020] Furthermore, in a preferred embodiment, the formula for calculating the maximum mutual information in step S3 above is expressed as:

[0021] I(X,Y)=H(X)-H(X|Y);

[0022]

[0023]

[0024] Where I(X|Y) is the degree of reduction in uncertainty of parameter X after obtaining the distribution information of parameter Y, H(X) is the entropy, p(x) is the probability of taking the value x, H(X|Y) is the conditional entropy, and p(x,y) is the joint probability of taking two values.

[0025] Furthermore, in a preferred embodiment, the partitioning condition for the attribute set in step S4 above is: attributes with a maximum mutual information greater than 0.9 are partitioned into the same attribute set.

[0026] Furthermore, in a preferred embodiment, the set of attributes in step S4 above is represented as follows:

[0027] G1 = {i1, i3, i7};

[0028] G2 = {i2, i5, i6};

[0029] G j ={i x i y i z ...}

[0030] Furthermore, in a preferred embodiment, step S6 above specifically comprises:

[0031] If the average maximum mutual information of an attribute set is less than 0.9, the attribute set is faulty.

[0032] The fault prediction method based on mutual information and correlation coefficient described in this invention can be entirely implemented using computer software. Therefore, correspondingly, this invention also provides a fault prediction system based on mutual information and correlation coefficient, the system comprising:

[0033] Storage device for acquiring multidimensional parameters output by the control system;

[0034] Storage device for performing smoothing filtering on the multidimensional parameters to obtain the filtered multidimensional parameters;

[0035] A storage device for calculating the maximum mutual information between the filtered multidimensional parameters;

[0036] Storage device for dividing the multidimensional parameters into attribute sets based on the maximum mutual information to obtain several attribute sets;

[0037] Storage device for calculating the average maximum mutual information of the attribute set;

[0038] A storage device for identifying faults based on the average maximum mutual information.

[0039] The present invention provides a computer-readable storage medium storing a computer program, which, when executed by a processor, performs the fault prediction method based on mutual information and correlation coefficient as described above.

[0040] The present invention provides a computer device, which includes a memory and a processor. The memory stores a computer program. When the processor runs the computer program stored in the memory, the processor executes the fault prediction method based on mutual information and correlation coefficient as described above.

[0041] The beneficial effects of this invention are as follows:

[0042] 1. This invention provides a fault prediction method based on mutual information and correlation coefficient. First, it collects multidimensional parameters output by the control system and performs smoothing filtering on these parameters to eliminate noise. Then, it calculates the maximum mutual information between the filtered multidimensional parameters. Calculating the maximum mutual information captures the nonlinear relationships between the multidimensional parameters, does not require assumptions about data distribution, and is suitable for complex data. Furthermore, it quantifies information gain and is robust to outliers. Based on the maximum mutual information, the multidimensional parameters are divided into attribute sets to obtain several attribute sets. Then, the average maximum mutual information of the attribute sets is calculated. Finally, fault identification is performed based on the average maximum mutual information. This method offers advantages such as low overhead and superior performance, requiring no additional hardware, no training data, and no knowledge of system structure and device parameters.

[0043] This invention is applicable to the prediction of faults that will occur in the multidimensional parameters of the output of a predictive control system and fault prediction that is mainly oriented towards a set of attributes. Attached Figure Description

[0044] Figure 1 This is a flowchart of the fault prediction method based on mutual information and correlation coefficient as described in Implementation Method 1;

[0045] Figure 2 It is the result table of the attribute set partitioning described in Implementation Method Eleven;

[0046] Figure 3 It is the fault identification set described in Implementation Method Eleven. Detailed Implementation

[0047] The specific embodiments of the present invention will be further described in detail below with reference to the accompanying drawings and examples. The following examples will help those skilled in the art to further understand the present invention, but do not limit the present invention in any way. It should be noted that those skilled in the art can make several changes and improvements without departing from the concept of the present invention, and these all fall within the protection scope of the present invention.

[0048] Implementation Method 1. See [link / reference] Figure 1 This embodiment describes a fault prediction method based on mutual information and correlation coefficient. The method is as follows:

[0049] S1. Acquire multidimensional parameters output by the control system;

[0050] S2. Perform smoothing filtering on the multidimensional parameters to obtain the filtered multidimensional parameters;

[0051] S3. Calculate the maximum mutual information between the filtered multidimensional parameters;

[0052] S4. Divide the multidimensional parameters into attribute sets according to the maximum mutual information to obtain several attribute sets;

[0053] S5. Calculate the average maximum mutual information of the attribute set;

[0054] S6. Identify faults based on the average maximum mutual information.

[0055] In practical applications, this implementation method, such as Figure 1 As shown, the multidimensional parameters i1, i2...i are acquired and controlled by the system. 18 Smoothing filters are applied to the multidimensional parameters to remove noise interference, resulting in... For the filtered multidimensional parameters Calculate the maximum mutual information I between them xy According to the maximum mutual information I xy The multidimensional parameters are divided into attribute sets to obtain several attribute sets; the average maximum mutual information of the attribute sets is calculated, and finally, fault identification is performed based on the average maximum mutual information.

[0056] This embodiment provides a fault prediction method based on mutual information and correlation coefficient. First, it collects multidimensional parameters output by the control system and performs smoothing filtering on these parameters to eliminate noise. Then, it calculates the maximum mutual information between the filtered multidimensional parameters. Calculating the maximum mutual information captures the nonlinear relationships between the multidimensional parameters, does not require assumptions about data distribution, and is suitable for complex data. Furthermore, it quantifies information gain and is robust to outliers. Based on the maximum mutual information, the multidimensional parameters are divided into attribute sets, obtaining several attribute sets. Then, the average maximum mutual information of the attribute sets is calculated. Finally, fault identification is performed based on the average maximum mutual information. This method offers advantages such as low overhead and superior performance, requiring no additional hardware, no training data, and no knowledge of system structure and device parameters.

[0057] Implementation Method 2. This implementation method is an example of the smoothing filtering process in step S2 of the fault prediction method based on mutual information and correlation coefficient described in Implementation Method 1. The smoothing filtering process is implemented using moving average filtering.

[0058] In practical applications, this implementation uses multi-dimensional parameter smoothing filtering to remove noise. Smoothing filtering is a signal processing technique used to reduce high-frequency noise or details in a signal to obtain a smoother signal. This implementation uses moving average filtering for smoothing; moving average filtering is expressed as:

[0059]

[0060] in, Here, i[nk] is the filtered output signal, i[nk] is the input signal, and N is the window size or filter length, usually a positive integer. The summation symbol indicates that the first N samples within the window are added together and the result is divided by the window size N. This formula means applying a sliding window of size N to the input signal i[nk], calculating the signal mean within the window, and using the mean as the output signal value. This helps smooth the signal and reduce the influence of high-frequency noise.

[0061] Implementation Method 3. This implementation method illustrates step S3 of the fault prediction method based on mutual information and correlation coefficient described in Implementation Method 1, specifically the maximum mutual information I. xy Represented as:

[0062] I xy ,{x,y}∈{1,2...18}.

[0063] In practical applications, this implementation method involves filtering the multidimensional parameters. Calculate the maximum mutual information I between them xy Specifically:

[0064] I xy ,{x,y}∈{1,2...18}.

[0065] Implementation Method 4. This implementation method illustrates the calculation formula for the maximum mutual information in step S3 of the fault prediction method based on mutual information and correlation coefficient described in Implementation Method 1. The formula for calculating the maximum mutual information is as follows:

[0066] I(X,Y)=H(X)-H(X|Y);

[0067]

[0068]

[0069] Where I(X|Y) is the degree of reduction in uncertainty of parameter X after obtaining the distribution information of parameter Y, H(X) is the entropy, p(x) is the probability of taking the value x, H(X|Y) is the conditional entropy, and p(x,y) is the joint probability of taking two values.

[0070] In practical applications, the advantage of mutual information over correlation coefficient lies in the fact that the correlation between parameters X and Y is not limited to linearity. The formula for calculating mutual information is as follows:

[0071] I(X,Y)=H(X)-H(X|Y);

[0072]

[0073]

[0074] Here, H(X) is the entropy, representing the uncertainty of parameter X; x is the value of parameter X, and p(x) is the probability of taking the value x. H(X|Y) is the conditional entropy, representing the uncertainty of parameter X given the distribution of parameter Y; y is the value of parameter Y, and p(x,y) is the joint probability of the two values. Therefore, I(X|Y) represents the degree to which the uncertainty of parameter X is reduced after obtaining the distribution information of parameter Y. The range of values ​​for I(X,Y) is non-negative, and the larger the value, the greater the reduction in the uncertainty of parameter X, and the stronger the correlation between parameter X and parameter Y.

[0075] Normalizing the mutual information makes comparing the mutual information values ​​between different parameters more intuitive. When comparing the mutual information values ​​of parameter X with the other parameters, the normalization operation is as follows:

[0076]

[0077] Where, I(X,Y) ratio This represents the percentage reduction in uncertainty of parameter X after utilizing information from parameter Y. For example, when H(X) = 1 and I(X,Y) = 0.3, then I(X,Y)... ratio =30%, which means that using parameter Y can reduce the uncertainty of parameter X by 30%. Therefore, the parameter Y that reduces the uncertainty of parameter X by the highest percentage is the parameter with the strongest correlation to parameter X.

[0078] Implementation Method 5. This implementation method is an example of the judgment condition for the attribute set division in step S4 of the fault prediction method based on mutual information and correlation coefficient described in Implementation Method 1. The division condition for the attribute set division is: the attributes with the maximum mutual information > 0.9 are divided into the same attribute set.

[0079] In practical applications, this implementation method is based on the calculated maximum mutual information I. xy The multidimensional parameters are divided into attribute sets, and I is... xy A value greater than 0.9 is considered to be part of the same attribute set. Using 0.9 as the partitioning criterion allows multidimensional parameters to be assigned to appropriate attribute sets, which is consistent with prior knowledge.

[0080] Implementation Method Six. This implementation method illustrates step S4 of the fault prediction method based on mutual information and correlation coefficient described in Implementation Method One with an example of several attribute sets. These attribute sets are represented as follows:

[0081] G1 = {i1, i3, i7};

[0082] G2 = {i2, i5, i6};

[0083] G j ={i x i y i z ...}

[0084] In practical applications, this implementation method is based on the calculated maximum mutual information I. xy The multidimensional parameters are partitioned into attribute sets, such as G1 = {i1, i3, i7}, G2 = {i2, i5, i6}, G... j ={i x i y i z ...}

[0085] Implementation Method Seven. This implementation method illustrates step S6 in the fault prediction method based on mutual information and correlation coefficient described in Implementation Method One. Step S6 specifically involves:

[0086] If the average maximum mutual information of an attribute set is less than 0.9, the attribute set is faulty.

[0087] In practical applications, this implementation method calculates the average I of the attribute set. xy When the average value of a certain attribute set is less than 0.9, it indicates that the attribute set has a fault, thus achieving the function of fault prediction. Calculating maximum mutual information can capture nonlinear relationships between variables without requiring assumptions about data distribution, making it suitable for complex data. It can quantify information gain in feature selection in machine learning and is robust to outliers. Mutual information can also be used for independence testing and signal processing to improve the accuracy and efficiency of models.

[0088] Implementation Method 8. This implementation method provides a fault prediction system based on mutual information and correlation coefficient. The system is as follows:

[0089] Storage device for acquiring multidimensional parameters output by the control system;

[0090] Storage device for performing smoothing filtering on the multidimensional parameters to obtain the filtered multidimensional parameters;

[0091] A storage device for calculating the maximum mutual information between the filtered multidimensional parameters;

[0092] Storage device for dividing the multidimensional parameters into attribute sets based on the maximum mutual information to obtain several attribute sets;

[0093] Storage device for calculating the average maximum mutual information of the attribute set;

[0094] A storage device for identifying faults based on the average maximum mutual information.

[0095] Implementation Method Nine. This implementation method provides a computer-readable storage medium storing a computer program, which, when executed by a processor, performs the fault prediction method based on mutual information comparison correlation coefficient as described in any one of Implementation Methods One to Seven.

[0096] Implementation Method 10. This implementation method provides a computer device, which includes a memory and a processor. The memory stores a computer program. When the processor runs the computer program stored in the memory, the processor executes the fault prediction method based on mutual information comparison correlation coefficient as described in any one of Implementation Methods 1 to 7.

[0097] Implementation Method 11. See also Figure 2 and Figure 3 This embodiment describes and verifies the fault prediction method based on mutual information and correlation coefficient described in any one of embodiments one through seven. Specifically:

[0098] va, vb, vc are the grid-side voltages of VSC, L is the series inductor, the system has three phases a, b, and c, ia, ib, ic are the currents at that location, T1-T6 are IGBTs, C is the capacitor, UDC is the output voltage, s1-s6 are the PWM control signals, and Vref is the reference voltage. The closed-loop control method is SVPWM.

[0099] A fault prediction method based on mutual information and correlation coefficient is used to process the multidimensional parameters of the above system, resulting in an attribute set partitioning table, such as... Figure 2 As shown, fault identification is performed based on the result table partitioned according to the attribute set, such as... Figure 3 As shown, {1 17} and {18} are fault sets.

[0100] Therefore, the fault prediction method based on mutual information and correlation coefficient described in this embodiment can accurately predict the fault state of the converter, and requires no additional hardware, no training data, and no knowledge of the system structure and device parameters.

[0101] This embodiment provides a fault prediction method based on mutual information and correlation coefficient. First, it collects multidimensional parameters output by the control system and performs smoothing filtering on these parameters to eliminate noise. Then, it calculates the maximum mutual information between the filtered multidimensional parameters. Calculating the maximum mutual information captures the nonlinear relationships between the multidimensional parameters, does not require assumptions about data distribution, and is suitable for complex data. Furthermore, it quantifies information gain and is robust to outliers. Based on the maximum mutual information, the multidimensional parameters are divided into attribute sets, obtaining several attribute sets. Then, the average maximum mutual information of the attribute sets is calculated. Finally, fault identification is performed based on the average maximum mutual information. This method offers advantages such as low overhead and superior performance, requiring no additional hardware, no training data, and no knowledge of system structure and device parameters.

[0102] Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated. Thus, a feature defined as "first" or "second" may explicitly or implicitly include at least one of that feature. In the description of this invention, "a plurality of" means at least two, such as two, three, etc., unless otherwise explicitly specified.

[0103] In the description of this specification, the references to terms such as "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of the present invention. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples. Moreover, without contradiction, those skilled in the art can combine and integrate the different embodiments or examples described in this specification, as well as the features of different embodiments or examples.

[0104] The above description is merely an embodiment of the present invention and is not intended to limit the invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of the claims of the present invention.

Claims

1. A fault prediction method based on mutual information and correlation coefficient, characterized in that, The method is as follows: S1. Acquire multidimensional parameters output by the control system; S2. Perform smoothing filtering on the multidimensional parameters to obtain the filtered multidimensional parameters; S3. Calculate the maximum mutual information between the filtered multidimensional parameters; S4. Divide the multidimensional parameters into attribute sets according to the maximum mutual information to obtain several attribute sets; S5. Calculate the average maximum mutual information of the attribute set; S6. Identify faults based on the average maximum mutual information.

2. The fault prediction method based on mutual information and correlation coefficient according to claim 1, characterized in that, The smoothing filtering process in step S2 is implemented using moving average filtering.

3. The fault prediction method based on mutual information and correlation coefficient according to claim 1, characterized in that, The maximum mutual information I in step S3 xy Represented as: I xy ,{x,y}∈{1,2...18}。 4. The fault prediction method based on mutual information and correlation coefficient according to claim 1, characterized in that, The formula for calculating the maximum mutual information in step S3 is as follows: I(X,Y)=H(X)-H(X|Y); Where I(X|Y) is the degree of reduction in uncertainty of parameter X after obtaining the distribution information of parameter Y, H(X) is the entropy, p(x) is the probability of taking the value x, H(X|Y) is the conditional entropy, and p(x,y) is the joint probability of taking two values.

5. The fault prediction method based on mutual information and correlation coefficient according to claim 1, characterized in that, The partitioning condition for the attribute set in step S4 is: attributes with a maximum mutual information greater than 0.9 are partitioned into the same attribute set.

6. The fault prediction method based on mutual information and correlation coefficient according to claim 1, characterized in that, The set of attributes in step S4 is represented as follows: G1 = {i1, i3, i7}; G2 = {i2, i5, i6}; G j ={i x ,i y ,i z ...}。 7. The fault prediction method based on mutual information and correlation coefficient according to claim 1, characterized in that, Step S6 specifically involves: If the average maximum mutual information of an attribute set is less than 0.9, the attribute set is faulty.

8. A fault prediction system based on mutual information and correlation coefficient, characterized in that, The system is: Storage device for acquiring multidimensional parameters output by the control system; Storage device for performing smoothing filtering on the multidimensional parameters to obtain the filtered multidimensional parameters; A storage device for calculating the maximum mutual information between the filtered multidimensional parameters; Storage device for dividing the multidimensional parameters into attribute sets based on the maximum mutual information to obtain several attribute sets; Storage device for calculating the average maximum mutual information of the attribute set; A storage device for identifying faults based on the average maximum mutual information.

9. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program that, when executed by a processor, performs the fault prediction method based on mutual information and correlation coefficient as described in any one of claims 1-7.

10. A computer device, characterized in that, The device includes a memory and a processor, wherein the memory stores a computer program, and when the processor runs the computer program stored in the memory, the processor executes the fault prediction method based on mutual information and correlation coefficient as described in any one of claims 1-7.

Images