Intelligence board fault diagnosis method and system based on multi-sensors

Abstract

This application relates to the technical field of information board fault data processing, and specifically relates to a method and system for information board fault diagnosis based on multi-sensors. The method includes: determining the fusion factor for each type of fault, using the fusion factors of all types of faults as the input of the threshold segmentation algorithm, and marking the faults with fusion factors greater than the segmentation threshold as complementary faults; determining the first similarity by analyzing the similarity between the sensing signals of each sensor and all modal components; comprehensively determining the second similarity based on the number and average distribution of all the same frequencies of the sensing signals of any two sensors in the frequency domain, as well as the phase difference in the frequency domain, and combining the first similarity to perform time synchronization correction on all observed sensors under complementary faults in the process of information board fault diagnosis by multi-sensors. This application improves the accuracy of the information board fault diagnosis result by performing synchronization correction on different sensors in terms of time.

Description

Technical Field

[0001] This application relates to the technical field of information board fault data processing, and specifically relates to a method and system for information board fault diagnosis based on multi-sensors. Background Art

[0002] The LED variable message sign, abbreviated as the information board, is an electronic information display device used in the fields of traffic management, urban informatization, etc. It has functions of displaying text and graphics, including but not limited to traffic flow, route guidance, travel time, weather conditions, traffic control information, emergency event notifications, etc. The information board is often integrated with other intelligent transportation system (ITS) components, such as cameras, sensors, vehicle detectors, etc., to provide more comprehensive road information.

[0003] During the long-term operation of the information board, various faults will inevitably occur. The information board fault diagnosis technology based on multi-sensors mainly uses methods such as data mining and machine learning, which require the fusion analysis of multi-source sensor data. However, the information board is generally installed outdoors, and is affected by hardware aging, frequent environmental interference, electrical interference, system processing delay, etc. The quality and synchronization of multi-source monitoring data are affected, further reducing the accuracy of the information board fault diagnosis results. Summary of the Invention

[0004] In order to solve the above technical problems, the purpose of this application is to provide a method and system for information board fault diagnosis based on multi-sensors, and the specific technical solutions adopted are as follows:

[0005] In the first aspect, the embodiment of this application provides a method for information board fault diagnosis based on multi-sensors, and the method includes the following steps:

[0006] In the historical fault log of the information board, obtain the sensor data at all acquisition times of each sensor under different preset sensor types within a preset time period before and after the occurrence times of each type of fault for multiple times;

[0007] For any occurrence time of each type of fault, determine the abnormal data of each sensor according to the abnormal distribution of all sensor data of each sensor, map the abnormal data of all sensors at any occurrence time of each type of fault to the same time series, and determine the abnormal redundancy of each type of fault by analyzing the time interval between adjacent abnormal data;

[0008] Form the abnormal sequences of each sensor under each type of fault with all the abnormal data of each sensor at all occurrence times of each type of fault. Determine the fusion factor of each type of fault by analyzing the correlation between the abnormal sequences of all sensors under each type of fault and the abnormal redundancy. Use the fusion factors of all types of faults as the input of the threshold segmentation algorithm, and mark the faults with a fusion factor greater than the segmentation threshold as complementary faults;

[0009] For each occurrence time of each type of complementary fault, all sensor data of each sensor are combined into a sensing signal, the sensing signals of all sensors are fused into a feature signal, and a modal decomposition algorithm is used to decompose the feature signal into multiple modal components. By analyzing the similarity between the sensing signal of each sensor and all modal components, the first similarity of each sensor at each occurrence time of each type of complementary fault is determined.

[0010] At each occurrence time of each type of complementary fault, by comprehensively considering the number and average distribution of all the same frequencies of the sensing signals of any two sensors in the frequency domain, as well as the difference in phase in the frequency domain, the second similarity between any two sensors is determined. Combining with the first similarity, all sensors are sorted to obtain the observed sensors of each sensor at each occurrence time of each type of complementary fault, and time synchronization correction is performed on all the observed sensors under complementary faults during the multi-sensor information board fault diagnosis process.

[0011] Preferably, the method for determining the abnormal data of each sensor is as follows:

[0012] For each occurrence time of each type of fault, all sensor data of each sensor are used as the input of the anomaly detection algorithm, and all the abnormal data of each sensor are output.

[0013] Preferably, the method for determining the abnormal redundancy of each type of fault is as follows:

[0014] The abnormal data of all sensors at each occurrence time of each type of fault are numbered in ascending order of time in the time sequence. Among them, if there are multiple abnormal data corresponding to the same time, the same number is assigned to them;

[0015] For each occurrence time of each type of fault, calculate the variance of the time intervals between all adjacent numbered abnormal data, and take the reciprocal of the mean value of the variances of all occurrence times of each type of fault as the abnormal redundancy of each type of fault.

[0016] Preferably, the expression of the fusion factor of each type of fault is: W i = Q i ×(1 - P i ); In the formula, W i represents the fusion factor of the i-th type of fault; Q i represents the abnormal redundancy of the i-th type of fault; P i represents the normalized value of the mean of the correlation coefficients between the abnormal sequences of all sensors under the i-th type of fault.

[0017] Preferably, the fusion of the sensing signals of all sensors into a feature signal includes:

[0018] The sensing signals of all sensors at any moment when each type of complementary fault occurs are normalized, and the data at the corresponding positions of all normalized signals are added together to obtain the characteristic signal.

[0019] Preferably, the first similarity of each sensor at any one time when each type of complementary fault occurs is the maximum value of the similarities between the sensing signal of each sensor and all modal components at any one time when each type of complementary fault occurs.

[0020] Preferably, the method for determining the second similarity between any two sensors is:

[0021] The sensor corresponding to the maximum value of the main frequency of the sensor signals of any two sensors in the frequency domain is used as the characteristic sensor. The proportion of all the common frequencies between the characteristic sensor and the sensor signal of the other sensor in the total number of frequencies in the sensor signal of the characteristic sensor is calculated, and recorded as the overlap between any two sensors;

[0022] The second similarity U between the sensing signals of sensor m and sensor n m,n The expression is:

[0023] Where, ρ m,n represents the mean value of all the same frequencies between the sensing signals of sensor m and sensor n; E m,n Indicates the overlap between sensor m and sensor n; A m,n It represents the phase difference between the sensing signals of sensor m and sensor n; exp() represents an exponential function with a natural constant as the base.

[0024] Preferably, the process of obtaining the observation sensor of each sensor at any time when each type of complementary fault occurs is:

[0025] At any occurrence of each type of complementary fault, all sensors are randomly sorted. The ratio of the first similarity between each sensor and its adjacent previous sensor in any sorting result is calculated, and the product of the ratio and the second similarity between the corresponding sensors is used as the similarity product between each sensor and its adjacent previous sensor. The average of the similarity products between all sensors and their adjacent previous sensors is used as the similarity value of any sorting result.

[0026] At any moment when each type of complementary fault occurs, the arrangement order corresponding to the maximum value among the similarity values of all sorting results is used as the optimal sorting. In the optimal sorting, the previous sensor of each sensor is used as the observation sensor of each sensor under each type of complementary fault.

[0027] Preferably, during the fault diagnosis of the multi-sensor information board for complementary faults, time synchronization correction is performed on all observed sensors, including:

[0028] At any occurrence moment of each type of complementary fault, the first sensor in the optimal sorting is used as the target sensor, and the moment corresponding to its first abnormal data is used as the target moment. The deviation between the moment corresponding to the first abnormal data of each sensor and the target moment is used as the time deviation of each sensor.

[0029] In the optimal sorting, the previous sensor of each sensor is used as the observed sensor. The time deviation between each sensor and the observed sensor is used as the input of the Kalman filtering algorithm, and the estimated value of the time deviation of each sensor is output. The sensing data of each sensor is synchronously corrected in time according to the estimated value of the time deviation.

[0030] In a second aspect, an embodiment of the present application also provides a multi-sensor-based information board fault diagnosis system, including a memory, a processor, and a computer program stored in the memory and running on the processor. When the processor executes the computer program, the steps of the above-mentioned multi-sensor-based information board fault diagnosis method are implemented.

[0031] The present application has at least the following beneficial effects:

[0032] The present application calculates the data fusion type of each type of fault through different fault manifestation characteristics. For faults of the complementary fusion type, by calculating the first similarity of its multi-source sensor data and the second similarity between different sensors, the observed sensors of each sensor under each type of fault are obtained, and a cascaded filtering framework is adopted to estimate the time deviation of correcting the data of a certain sensor for faults of the complementary fusion type, which can avoid the problems of weak observation relationships between multi-source sensor data of this type of fault and inability to randomly select observation variables to the greatest extent, improve the time synchronization in the data fusion process, and thus improve the accuracy of the information board fault diagnosis result. Description of the Drawings

[0033] In order to more clearly illustrate the technical solutions and advantages in the embodiments of the present application or the prior art, the following will briefly introduce the drawings required for use in the description of the embodiments or the prior art. Obviously, the following drawings are only some embodiments of the present application. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0034] Figure 1 It is a flowchart of the steps of a multi-sensor-based information board fault diagnosis method provided by an embodiment of the present application;

[0035] Figure 2Schematic diagram of the first similarity extraction process provided by an embodiment of the present application. Detailed implementation manners

[0036] In order to further elaborate on the technical means and effects adopted by the present application to achieve the predetermined invention purpose, the following combines the accompanying drawings and preferred embodiments to detail the specific implementation manners, structures, features and effects of the multi-sensor-based information board fault diagnosis method and system proposed according to the present application. In the following description, different "one embodiment" or "another embodiment" do not necessarily refer to the same embodiment. In addition, the specific features, structures or characteristics in one or more embodiments can be combined in any suitable form.

[0037] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by those skilled in the technical field to which this application belongs.

[0038] The following specifically describes the specific solutions of the multi-sensor-based information board fault diagnosis method and system provided by the present application with reference to the accompanying drawings.

[0039] Please refer to Figure 1 , which shows the flowchart of the steps of the multi-sensor-based information board fault diagnosis method provided by an embodiment of the present application. The method includes the following steps:

[0040] Step S1: In the historical fault log of the information board, obtain the sensor data of each sensor at all acquisition times under different preset sensor types within a preset time period before and after the occurrence times of each type of fault multiple times.

[0041] In the historical fault log of the information board, obtain the sensor data of each sensor at all acquisition times under different preset sensor types within a preset time period before and after the occurrence times of each type of fault multiple times, where the sampling frequency is set to f.

[0042] It should be noted that for different types of faults, different preset sensor types are selected. For example, when a fault occurs in the display screen of the information board, such as LED lamp damage or control circuit problems, fault characteristics will appear in the monitoring data of brightness sensors, ambient light sensors, current sensors, etc. Therefore, select the sensors whose sensor data is greatly affected when a fault occurs in the display screen of the information board, and analyze and study them.

[0043] In addition, it should be understood that the sensor data corresponding to different sensors is also different. For current sensors, the collected sensor data is current data; the sensors collected by voltage sensors are voltage data. Therefore, the type of sensor data is related to the sensor body.

[0044] Among them, the values of the preset duration and the sampling frequency are both set artificially. In this embodiment, the sensor data before and after the occurrence of the fault for 5 minutes each are collected, and the value of the sampling frequency is 1 Hz. Implementers can also set it by themselves according to the specific situation, and this embodiment does not make special restrictions.

[0045] Step S2: For any occurrence time of each type of fault, according to the abnormal distribution of all sensor data of each sensor, determine the abnormal data of each sensor. By analyzing the time interval between adjacent abnormal data, determine the abnormal redundancy of each type of fault; by analyzing the correlation between the abnormal sequences of all sensors under each type of fault and the abnormal redundancy, determine the fusion factor of each type of fault to determine complementary faults.

[0046] Each sensor will have a time offset after running for a period of time, which has a great impact on the fault diagnosis model to fuse multi-source sensor data and extract fault features. Calibrating the synchronization of multi-source sensor data is a major challenge. However, in the process of multi-source sensor data fusion, there are generally two types of fusion: competitive fusion and complementary fusion. When the multi-source sensor data fusion type of the fault is complementary fusion, the fault information contained between any two sensors is almost non-overlapping. This means that when estimating the time deviation of one sensor, it is impossible to directly use the data of another sensor as the observation variable to obtain the observation error of the time deviation estimation, and there may be a situation where the observation error and the observation noise are confused.

[0047] Therefore, according to the abnormal distribution of the sensor data of each sensor at the occurrence time of each type of fault, and combined with the correlation between the abnormal sequences of all sensors under each type of fault, determine the complementary faults, specifically:

[0048] For any occurrence time of each type of fault, take all the sensor data of each sensor as the input of the anomaly detection algorithm, and output all the abnormal data of each sensor.

[0049] It should be noted that there are many commonly used anomaly detection algorithms. In this embodiment, the LOF anomaly detection algorithm is used. In the actual application process, as other implementation manners, implementers can also select other anomaly detection algorithms according to the specific situation, and this embodiment does not make special restrictions.

[0050] Among them, the LOF anomaly detection algorithm is a well-known technology, and its specific principle will not be elaborated here.

[0051] Further, the abnormal data of all sensors at any occurrence time of each type of fault are numbered in ascending order of time in the time sequence. Among them, if there are multiple abnormal data corresponding to the same time, the same number is assigned to them; for any occurrence time of each type of fault, calculate the variance of the time intervals between all adjacent numbered abnormal data, and take the reciprocal of the mean value of the variances of all occurrence times of each type of fault as the abnormal redundancy of each type of fault; the smaller the variance, the denser the distribution of abnormal data, the closer they are in time, and the greater the abnormal redundancy. On the contrary, the larger the variance, the more dispersed the distribution of abnormal data, the farther apart they are in time, and the smaller the abnormal redundancy.

[0052] Further, all the abnormal data of each sensor at all occurrence times of each type of fault are arranged in chronological order to form an abnormal sequence of each sensor under each type of fault.

[0053] The fusion factor W of the i-th type of fault i has the following expression: W i = Q i ×(1 - P i ); where Q i represents the abnormal redundancy of the i-th type of fault; P i represents the normalized value of the mean of the correlation coefficients between the abnormal sequences of all sensors under the i-th type of fault.

[0054] It should be noted that there are many methods for calculating the correlation coefficient between sequences. In this embodiment, the absolute value of the Pearson correlation coefficient between abnormal sequences is used as the correlation coefficient between abnormal sequences. In actual application, as other implementation manners, the implementer can also use the Spearman correlation coefficient or the Kendall rank correlation coefficient. Regarding the selection of the correlation coefficient method, this embodiment does not make special restrictions.

[0055] Among them, the calculation method of the Pearson correlation coefficient is a well-known technology, and its specific calculation process will not be elaborated here.

[0056] From the fusion factor, it can be understood that the greater the abnormal redundancy, the smaller the normalized value of the mean of the correlation coefficients, and the greater the obtained fusion factor, indicating that the greater the possibility that the fault corresponds to a complementary fault, that is, the greater the possibility that the fault manifestations of different sensors do not overlap; on the contrary, the smaller the abnormal redundancy, the greater the normalized value of the mean of the correlation coefficients, and the smaller the obtained fusion factor, indicating that the smaller the possibility that the fault corresponds to a complementary fault, that is, the greater the possibility that the fault manifestations of different sensors overlap.

[0057] Further, take the fusion factors of all types of faults as the input of the threshold segmentation algorithm, and mark the faults with fusion factors greater than the segmentation threshold as complementary faults.

[0058] It should be noted that there are many commonly used threshold segmentation algorithms. In this embodiment, the Otsu threshold segmentation algorithm is adopted. In actual application processes, as other implementation manners, implementers can also select other threshold segmentation algorithms. This embodiment does not make special restrictions on the selection of threshold segmentation algorithms.

[0059] Step S3: For any occurrence moment of each type of complementary fault, all sensor data of each sensor are combined into a sensing signal, the sensing signals of all sensors are fused into a feature signal, and a modal decomposition algorithm is used to decompose the feature signal into multiple modal components. By analyzing the similarity between the sensing signal of each sensor and all modal components, the first similarity of each sensor at any occurrence moment of each type of complementary fault is determined.

[0060] Complementary faults are obtained from step S2. Since there is a large time delay between sensors under complementary faults, therefore, by analyzing the distribution characteristics of the sensing data of sensors under complementary faults, the first similarity is determined, specifically:

[0061] The sensing signals of all sensors at any occurrence moment of each type of complementary fault are normalized, and the data at the corresponding positions of all the normalized signals are added to obtain a feature signal.

[0062] A modal decomposition algorithm is used to decompose the feature signal into multiple modal components. In this embodiment, the intrinsic time-scale decomposition algorithm is used to decompose the feature signal. In actual application processes, as other implementation manners, implementers can also adopt the empirical mode decomposition algorithm. This embodiment does not make special restrictions on the selection of modal decomposition algorithms.

[0063] Among them, the intrinsic time-scale decomposition is a well-known technology, and its specific principle will not be elaborated here.

[0064] The first similarity of each sensor at any occurrence moment of each type of complementary fault is the maximum value among the similarities between the sensing signal of each sensor and all modal components at any occurrence moment of each type of complementary fault.

[0065] According to the first similarity, it can be understood that if the similarity between the sensing signal and the modal component is greater, it means that after fusing the sensing signal and then re-decomposing it, a component signal containing fault information can still be decomposed, indicating that the sensor data is continuously stable at a certain fault occurrence moment rather than intermittent and sudden; on the contrary, if the similarity between the sensing signal and the modal component is smaller, it means that the possibility of decomposing a component signal containing fault information after fusing the sensing signal and then re-decomposing it is smaller, indicating that the sensor data is intermittent, sudden and discontinuous at a certain fault occurrence moment.

[0066] Preferably, the schematic diagram of the first similarity extraction process provided in this embodiment is asFigure 2 as shown

[0067] Step S4: At any occurrence time of each type of complementary fault, comprehensively consider the number and average distribution of all the same frequencies of the sensing signals of any two sensors in the frequency domain, as well as the phase difference between them in the frequency domain, determine the second similarity between any two sensors, and combine with the first similarity to sort all the sensors, obtain the observation sensors of each sensor at any occurrence time of each type of complementary fault, and diagnose the information board faults of multiple sensors.

[0068] For complementary fault types, their fault manifestations are continuously deteriorating, and there is a coupling effect between the multi-source sensor data during the continuous deterioration process of the fault, that is, a fault detected by one sensor may affect other sensors through physical connections or signal transmissions. Through the gradual deterioration and expansion trends between different sensor data, the strength of the observation relationship between different sensor data can be inferred; for example, for an internal fault in the power supply module of an information board and uneven power distribution, over time, unstable current may cause more display modules to fail.

[0069] Therefore, by analyzing the similarity between the sensing signals of different sensors in the frequency domain, the second similarity is determined, specifically:

[0070] Regarding the sensor corresponding to the maximum value in the main frequencies of the sensing signals of any two sensors in the frequency domain as the characteristic sensor, calculate the proportion of the number of all the same frequencies between the sensing signal of the characteristic sensor and the sensing signal of another sensor in the total number of all frequencies of the sensing signal of the characteristic sensor, and denote it as the coincidence degree between any two sensors;

[0071] The second similarity U m,n between the sensing signals of sensor m and sensor n is expressed as:

[0072] In the formula, ρ m,n represents the average value of all the same frequencies between the sensing signals of sensor m and sensor n; E m,n represents the coincidence degree between sensor m and sensor n; A m,n represents the phase difference between the sensing signals of sensor m and sensor n; exp() represents the exponential function with the natural constant as the base.

[0073] It can be understood from the second similarity that if the overlap between different sensors is higher, the average value of all the same frequencies is larger, and the phase difference between the sensing signals is smaller, then the second similarity is larger, indicating that the similarity between the sensor signals of the sensors is larger; conversely, if the overlap between different sensors is lower, the average value of all the same frequencies is smaller, and the phase difference between the sensing signals is larger, then the second similarity is smaller, indicating that the similarity between the sensor signals of the sensors is smaller.

[0074] For any occurrence time of each type of complementary fault, all sensors are randomly sorted, the ratio of the first similarity between each sensor and its adjacent previous sensor in any sorting result is calculated, and the product of the ratio and the second similarity between the corresponding sensors is used as the similarity product between each sensor and its adjacent previous sensor, and the average value of the similarity products between all sensors and their adjacent previous sensors is used as the similarity value of any sorting result;

[0075] At any occurrence time of each type of complementary fault, the arrangement order corresponding to the maximum value among the similarity values of all sorting results is used as the optimal sorting, and the previous sensor of each sensor in the optimal sorting is used as the observation sensor of each sensor under each type of complementary fault.

[0076] At any occurrence time of each type of complementary fault, the first sensor in the optimal sorting is used as the target sensor, and the time corresponding to its first abnormal data is used as the target time. The deviation between the time corresponding to the first abnormal data of each sensor and the target time is used as the time deviation of each sensor;

[0077] In the optimal sorting, the previous sensor of each sensor is used as the observation sensor, the time deviations of each sensor and the observation sensor are used as the input of the Kalman filter algorithm, the estimated values of the time deviations of each sensor are output, and the sensing data of each sensor is synchronously corrected in time according to the estimated values of the time deviations.

[0078] Among them, the principle of the Kalman filter algorithm is a well-known technology, and its specific principle process will not be elaborated here.

[0079] For complementary faults, that is, faults with a fusion factor greater than the segmentation threshold, there are time deviations in the sensor data of the sensors under complementary faults. Therefore, by synchronously correcting the sensor data in time in the sensors under complementary faults and fusing the sensor data of all the corrected sensors, the faults are diagnosed according to the fused sensor data; for faults with a fusion factor less than or equal to the segmentation threshold, there are no time deviations in the sensor data of different sensors under them, so the sensor data of all sensors can be directly fused.

[0080] Obtain the sensor data of each sensor at all acquisition times within a preset duration before and after the occurrence time of the current fault. According to the calculation method of the fusion factor described in step S2, calculate the fusion factor of this fault. If the fusion factor of this fault is greater than the segmentation threshold, synchronously correct and then fuse the sensor data of all sensors under this fault according to the methods of steps S3 - S4, and diagnose the information board fault based on the fused sensor data; otherwise, if the fusion factor of this fault is less than or equal to the segmentation threshold, directly fuse the sensor data of all sensors under this fault, and diagnose the information board fault based on the fused sensor data.

[0081] Fuse the sensor data of all sensors under various faults in the historical fault log, and mark all types of faults. Use the fused data and the labels of the corresponding types of faults as the input of the neural network to output a fault diagnosis model. Use the result after fusing the sensor data of all sensors under the current fault as the input of the fault diagnosis model to output the fault type.

[0082] Based on the same inventive concept as the above method, the embodiment of the present application also provides an information board fault diagnosis system based on multi - sensors, including a memory, a processor, and a computer program stored in the memory and running on the processor. When the processor executes the computer program, it implements the steps of any one of the above - mentioned information board fault diagnosis methods based on multi - sensors.

[0083] It should be noted that: the above - mentioned sequence of the embodiments of the present application is only for description and does not represent the superiority or inferiority of the embodiments. And the specific embodiments of this specification have been described above. In addition, the processes depicted in the drawings do not necessarily require the specific order or continuous order shown to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0084] Each embodiment in this specification is described in a progressive manner. The same or similar parts between each embodiment can be referred to each other, and the key point of each embodiment is to illustrate the differences from other embodiments.

[0085] The above - mentioned are only the preferred embodiments of the present application and are not used to limit the present application. Any modifications, equivalent replacements, improvements, etc. made within the principle of the present application shall be included in the protection scope of the present application.

Claims

1. An intelligence board fault diagnosis method based on multi-sensors, characterized in that, The method includes the following steps: In the historical fault log of the information board, obtain the sensor data of all acquisition moments of each sensor under different preset sensor types within a preset duration before and after the occurrence moments of each type of fault. For any occurrence moment of each type of fault, determine the abnormal data of each sensor according to the abnormal distribution of all the sensor data of each sensor, map the abnormal data of all sensors at any occurrence moment of each type of fault to the same time series, and determine the abnormal redundancy of each type of fault by analyzing the time intervals between adjacent abnormal data. Form the abnormal sequences of each sensor under each type of fault with all the abnormal data of each sensor at all occurrence moments of each type of fault, determine the fusion factor of each type of fault by analyzing the correlation between the abnormal sequences of all sensors under each type of fault and the abnormal redundancy, use the fusion factors of all types of faults as the input of the threshold segmentation algorithm, and mark the faults with fusion factors greater than the segmentation threshold as complementary faults. For any occurrence moment of each type of complementary fault, form the sensing signals with all the sensor data of each sensor, fuse the sensing signals of all sensors into a characteristic signal, and use the modal decomposition algorithm to decompose the characteristic signal into multiple modal components, and determine the first similarity of each sensor at any occurrence moment of each type of complementary fault by analyzing the similarity between the sensing signals of each sensor and all modal components. At any occurrence moment of each type of complementary fault, comprehensively determine the second similarity between any two sensors by the number and average distribution of all the same frequencies of the sensing signals of any two sensors in the frequency domain and the difference in phase in the frequency domain, and combine the first similarity to sort all sensors to obtain the observed sensors of each sensor at any occurrence moment of each type of complementary fault, and perform time synchronization correction on all observed sensors under complementary faults in the multi-sensor information board fault diagnosis process.

2. The multi-sensor-based information board fault diagnosis method according to claim 1, wherein The method for determining the abnormal data of each sensor is as follows: For any occurrence moment of each type of fault, use all the sensor data of each sensor as the input of the anomaly detection algorithm, and output all the abnormal data of each sensor.

3. The multi-sensor-based intelligence board fault diagnosis method according to claim 1, characterized in that The method for determining the abnormal redundancy of each type of fault is as follows: Number the abnormal data of all sensors at any occurrence moment of each type of fault in ascending order of time in the time series, where if there are multiple abnormal data corresponding to the same moment, they are given the same number. For any occurrence moment of each type of fault, calculate the variance of the time intervals between all adjacent numbered abnormal data, and take the reciprocal of the mean value of the variances of all occurrence moments of each type of fault as the abnormal redundancy of each type of fault.

4. The method for fault diagnosis of an information board based on multi-sensors according to claim 1, wherein, The expression for the fusion factor of each type of fault is: W i = Q i ×(1 - P i ); where, W i represents the fusion factor of the i-th type of fault; Q i represents the abnormal redundancy of the i-th type of fault; P i represents the normalized value of the mean of the correlation coefficients between the abnormal sequences of all sensors under the i-th type of fault.

5. The multi-sensor based fault diagnosis method for information boards according to claim 1, characterized in that The fusion of the sensing signals of all sensors into a characteristic signal includes: Normalize the sensing signals of all sensors at any occurrence moment of each type of complementary fault, and add the data at the corresponding positions of all the normalized signals to obtain a characteristic signal.

6. The multi-sensor-based intelligence board fault diagnosis method according to claim 1, characterized in that, The first similarity of each sensor at any occurrence time of each type of complementary fault is the maximum value among the similarities between the sensing signals of each sensor and all modal components at any occurrence time of each type of complementary fault.

7. The method for fault diagnosis of an information board based on multiple sensors according to claim 1, characterized in that The method for determining the second similarity between any two sensors is as follows: Take the sensor corresponding to the maximum value among the main frequencies of the sensing signals of any two sensors in the frequency domain as the characteristic sensor, and calculate the proportion of the number of all the same frequencies between the sensing signal of the characteristic sensor and the sensing signal of the other sensor in the number of all frequencies in the sensing signal of the characteristic sensor, which is denoted as the coincidence degree between any two sensors; The second similarity U between the sensing signals of sensor m and sensor n m,n is expressed as follows: In the formula, ρ m,n represents the mean value of all the same frequencies between the sensing signals of sensor m and sensor n; E m,n represents the coincidence degree between sensor m and sensor n; A m,n represents the phase difference between the sensing signals of sensor m and sensor n; exp() represents the exponential function with the natural constant as the base number.

8. The method for fault diagnosis of an information board based on multiple sensors according to claim 1, characterized in that, The process of obtaining the observation sensors of each sensor at any occurrence time of each type of complementary fault is as follows: For any occurrence time of each type of complementary fault, randomly sort all sensors, calculate the ratio of the first similarity between each sensor and its adjacent previous sensor in any sorting result, and take the product of the ratio and the second similarity between the corresponding sensors as the similarity product between each sensor and its adjacent previous sensor, and take the average value of the similarity products between all sensors and their adjacent previous sensors as the similarity value of any sorting result; At any occurrence time of each type of complementary fault, take the arrangement order corresponding to the maximum value among the similarity values of all sorting results as the optimal sorting, and in the optimal sorting, take the previous sensor of each sensor as the observation sensor of each sensor under each type of complementary fault.

9. The method for fault diagnosis of an information board based on multiple sensors according to claim 8, characterized in that, The time synchronization correction for all observation sensors under complementary faults in the multi-sensor information board fault diagnosis process includes: At any occurrence time of each type of complementary fault, take the first sensor in the optimal sorting as the target sensor, and take the time corresponding to its first abnormal data as the target time, and take the deviation between the time corresponding to the first abnormal data of each sensor and the target time as the time deviation of each sensor; In the optimal sorting, take the previous sensor of each sensor as the observation sensor, take the time deviations of each sensor and the observation sensor as the input of the Kalman filtering algorithm, output the estimated value of the time deviation of each sensor, and perform time synchronization correction on the sensing data of each sensor according to the estimated value of the time deviation.

10. An intelligence board fault diagnosis system based on multi-sensors, comprising a memory, a processor, and a computer program stored in the memory and running on the processor, characterized in that, When the processor executes the computer program, it implements the steps of the multi-sensor-based information board fault diagnosis method described in any one of claims 1-9.

Images