Fault detection method for electric wire for new energy automobile

Abstract

The invention relates to the technical field of data processing, in particular to a fault detection method for a wire for a new energy automobile, and the method comprises the steps: obtaining a data sequence of detection data; determining the fluctuation time difference weight of the acquisition period based on the change of the autocorrelation strength between the data sequences when the data sequences fluctuate in the acquisition period; setting an amplitude threshold based on the amplitudes of the two kinds of detection data on all frequency components of the frequency domain, determining a fitting straight line based on the screened peak points by using a data fitting mode so as to determine a difference trend factor of the detection data, and determining a fault anomaly weight by combining a fluctuation time difference weight and an energy spectrum density difference so as to realize fault detection. Determining a connection influence coefficient of the detection data by combining the variable quantity of the elements in the data sequence; and obtaining an abnormal data point based on the connection influence coefficient by using a data clustering algorithm, and determining a fault detection result. According to the method, the clustering algorithm is optimized, so that the distinction degree of different fluctuation characteristic data is improved, and a more accurate fault detection result is obtained.

Description

Technical Field

[0001] The present application relates to the field of data processing technology, and in particular to a fault detection method for electric wires used in new energy vehicles. Background Art

[0002] The development of new energy vehicles can not only promote the use of renewable energy and reduce dependence on traditional fossil energy, but also help promote the innovation and improvement of clean technology and promote the development of the economic industrial chain. Compared with traditional vehicles, there are a large number of different types of wires in the production and processing of new energy vehicles, including high-voltage battery connection wires for connecting battery packs with motors and drive systems; charging connection wires for providing charging functions; communication data wires and motor transmission wires for assisting intelligent driving; and low-voltage control wires for transmitting low-voltage signals inside the car. These wires work together to ensure the safe driving and normal use of new energy vehicles.

[0003] In order to ensure the safety of drivers and passengers and prevent the dangerous phenomenon of spontaneous combustion of new energy vehicles due to insulation failure and other problems, fault detection of wires used in new energy vehicles is particularly important. It plays an important role in discovering hidden dangers in the lines, reducing failure rates, and increasing service life.

[0004] Intelligent detection technology uses advanced intelligent algorithms and data analysis technology to detect faults in different types of power lines for new energy vehicles, improving detection efficiency and consistency. However, during the performance detection of power lines, the instability of the access current and signal transmission network, and the use of sensors, oscilloscopes and other equipment to collect and analyze detection data often introduces unavoidable noise factors, which also makes it difficult to distinguish between normal data and abnormal data at the time of failure when fault detection is performed on power lines, making it impossible to accurately determine whether the power lines are faulty, and the detection results are unstable and the accuracy is reduced. Summary of the invention

[0005] In order to solve the above technical problems, the present application provides a fault detection method for electric wires for new energy vehicles to solve the problems existing in the existing detection methods.

[0006] The fault detection method for electric wires for new energy vehicles of the present application adopts the following technical solutions: An embodiment of the present application provides a fault detection method for electric wires for new energy vehicles, the method comprising the following steps: Obtaining a data sequence of each detection data in each acquisition cycle; Determine the fluctuation time difference weight of each collection period based on the change in the strength of the autocorrelation between the data series when the data series of the detection data between the high-voltage battery connection wire, the battery, and the motor in each collection period fluctuates; In each acquisition cycle, each detection data is converted into time-frequency domain, an amplitude threshold is set based on the amplitude of the two detection data at all frequency components, and a fitting straight line of the peak point screened by the amplitude threshold is obtained by data fitting; and a difference trend factor of each detection data is determined based on the fitting straight line; Based on the energy spectrum density diagram obtained after the data sequence of each detection data is converted into the frequency domain, the fault anomaly weight of each detection data in each acquisition cycle is determined in combination with the fluctuation time difference weight of each acquisition cycle and the difference trend factor of each detection data; Based on the fault anomaly weight of each detection data in each acquisition cycle, the connection influence coefficient of each detection data in each acquisition cycle is determined in combination with the change amount of the elements in the data sequence of each detection data; A data clustering algorithm is used to obtain abnormal data points in a data sequence of each detection data based on a connection influence coefficient of each detection data, and a fault detection result is determined based on the frequency of the abnormal data points in multiple acquisition cycles.

[0007] Preferably, the fluctuation time difference weight of each acquisition cycle is determined in the following manner: For each acquisition cycle, the data sequences corresponding to the motor current, the battery current, and the current at both ends of the high-voltage battery connection line are lagged respectively. After the lag, the autocorrelation coefficient between the data sequence obtained after the lag and the data sequence before the lag is calculated; where, is the number of hysteresis bits of the data sequence corresponding to the current; Calculate the data series corresponding to each current at lag Bit and Hysteresis +1 bit corresponds to the difference between the autocorrelation coefficients; Determine the data series corresponding to all currents at lag The product of the distribution variance of the obtained autocorrelation coefficient and the difference is The cumulative result within the value range of ; The accumulated mean value of the accumulated results on the data sequences corresponding to all currents is used as the fluctuation time difference weight of each acquisition cycle.

[0008] Preferably, the setting of the amplitude threshold based on the amplitudes of the two detection data at all frequency components includes: The data sequence of each detection data is converted from the time domain to the frequency domain data using the fast Fourier transform FFT to obtain the amplitude and phase information of each frequency component, and the sequence composed of the amplitudes of all frequency components obtained after each data sequence is converted to the frequency domain is used as the amplitude sequence of each data sequence; The element values ​​of the same order in the amplitude sequences of the two detection data are subjected to difference processing to obtain an amplitude difference sequence, and the amplitude threshold is determined using the maximum peak value and the minimum peak value in the amplitude difference sequence.

[0009] Preferably, the step of determining the amplitude threshold using the maximum peak value and the minimum peak value in the amplitude difference sequence includes: The ratio of the difference between the maximum peak value and the minimum peak value in the amplitude difference sequence and 3 is used as the amplitude threshold of the amplitude difference sequence.

[0010] Preferably, the method of obtaining a fitting straight line of a peak point screened by an amplitude threshold by means of data fitting includes: The peak points whose amplitudes are greater than the amplitude threshold in the amplitude difference sequence are screened, and the amplitudes and frequencies of all the screened peak points are used as dependent variables and independent variables for linear fitting, respectively, to obtain the fitting straight line corresponding to the amplitude difference sequence.

[0011] Preferably, the determining of the difference trend factor of each detection data based on the fitting straight line includes: Each type of detection data is used as target data, and the slope of the fitting straight line corresponding to the amplitude difference sequence between the target data and any type of detection data is used as the input of the inverse tangent function, and the sum of the product of the preset smoothing coefficient and the output of the inverse tangent function and 1 is calculated; The product of the sum and the maximum peak value in the amplitude difference sequence between the target data and any one of the detection data is accumulated over all the detection data as the difference trend factor of each detection data.

[0012] Preferably, the fault anomaly weight of each detection data in each acquisition cycle is determined as follows: Perform square calculation on the amplitude of each frequency component of the data sequence of each detection data to determine the energy value of each frequency component, and draw an energy spectrum density diagram of the data sequence in the frequency domain with the frequency as the abscissa and the energy value as the ordinate; The sum of the difference between the maximum value and the minimum value of the energy spectrum density of the data sequence of each detection data in each acquisition cycle in the energy spectrum density diagram in the frequency domain and a preset constant is used as the denominator; The ratio of the product of the fluctuation time difference weight of each acquisition cycle and the difference trend factor of each detection data in each acquisition cycle to the denominator is used as the fault anomaly weight of each detection data in each acquisition cycle.

[0013] Preferably, the connection influence coefficient of each detection data in each acquisition cycle is determined as follows: The product of the accumulation result of the absolute value of the difference between adjacent elements in the data sequence of each detection data in each acquisition cycle and the fault anomaly weight of each detection data in each acquisition cycle is taken as the connection influence coefficient of each detection data in each acquisition cycle.

[0014] Preferably, the method of obtaining abnormal data points in the data sequence of each type of detection data based on the connection influence coefficient of each type of detection data by using a data clustering algorithm includes: The data sequence of each detection data in each acquisition cycle is used as input, and the connection influence coefficient of each detection data in each acquisition cycle is input into the sigmoid function to obtain the function result as the shrinkage factor for clustering; Using the CURE clustering algorithm to divide the data sequence of each detection data in each acquisition cycle into a plurality of clusters based on the shrinkage factor; For each data point in each cluster, the Euclidean distance between the data point and the representative point of the cluster to which it belongs is calculated, and the data point whose Euclidean distance is greater than the preset outlier distance is regarded as an abnormal data point.

[0015] Preferably, determining the fault detection result based on the frequency of the abnormal data points in multiple acquisition cycles includes: The ratio of the number of collection cycles in which abnormal data points appear in all types of detection data to the total number of collection cycles is counted. When the ratio is greater than a preset threshold, it is considered that there is a fault in the high-voltage battery connection line.

[0016] This application has at least the following beneficial effects: The present application proposes a fault detection method for electric wires used in new energy vehicles. First, based on the change in the strength of the autocorrelation between the data sequences when data fluctuations occur in the data sequences of the detection data between the high-voltage battery connection wire, the battery, and the motor in each acquisition cycle, the fluctuation time difference weight of each acquisition cycle is determined; in each acquisition cycle, each detection data is converted into the time-frequency domain, and an amplitude threshold is set based on the amplitudes of the two detection data on all frequency components. The amplitude threshold is set so that when there is a large degree of difference in each frequency component of the two detection data in the frequency domain, there will be multiple peak points in the amplitude difference sequence, and the peak points greater than the amplitude threshold are screened to eliminate the interference caused by the use of peak features to analyze the data characteristics caused by the high-voltage battery connection wire fault when the peak point value is small. Afterwards, based on the energy spectrum density diagram obtained after the data sequence of each detection data is converted to the frequency domain, the weight of the fluctuation time difference of each acquisition cycle and the difference trend factor of each detection data are combined to determine the fault anomaly weight of each detection data in each acquisition cycle; the influence of the equipment connected to the high-voltage battery connection line on the detection data is considered to facilitate the subsequent determination of the true abnormal situation of each detection data in each acquisition cycle; then, based on the fault anomaly weight of each detection data in each acquisition cycle, the connection influence coefficient of each detection data in each acquisition cycle is determined in combination with the change amount of the elements in the data sequence of each detection data; finally, the data clustering algorithm is used to obtain the abnormal data points in the data sequence of each detection data based on the connection influence coefficient of each detection data, and the shrinkage factor is adaptively adjusted to enable it to adaptively adjust the shrinkage strength of the cluster according to the fluctuation and change trend of the detection data in each acquisition cycle, thereby changing the judgment factor of the outlier and improving the accuracy of fault detection. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] In order to more clearly illustrate the technical solutions and advantages in the embodiments of the present application or the prior art, the drawings required for use in the embodiments or the prior art descriptions are briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0018] Figure 1 A schematic diagram of the implementation process of the fault detection method for electric wires for new energy vehicles provided in an embodiment of the present application. DETAILED DESCRIPTION

[0019] In order to further explain the technical means and effects adopted by this application to achieve the predetermined invention purpose, the following is a detailed description of the fault detection method for electric wires for new energy vehicles proposed in this application, its specific implementation method, structure, features and effects, in combination with the accompanying drawings and preferred embodiments. In the following description, different "one embodiment" or "another embodiment" does not necessarily refer to the same embodiment. In addition, specific features, structures or characteristics in one or more embodiments may be combined in any suitable form.

[0020] Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs.

[0021] The specific scheme of the fault detection method for electric wires for new energy vehicles provided by the present application is described in detail below with reference to the accompanying drawings.

[0022] An embodiment of the present application provides a fault detection method for electric wires used in new energy vehicles. Specifically, the following fault detection method for electric wires used in new energy vehicles is provided. Please refer to Figure 1 , the method comprises the following steps: Step S001, collecting detection data for electric wire fault detection and preprocessing the obtained data.

[0023] Due to the different types of wires, this application uses the connection of wires with relevant hardware devices in new energy vehicles for performance testing as a detection condition for fault detection. The relevant hardware devices include but are not limited to new energy vehicle motors, batteries, sensors, and vehicle air conditioners. Preferably, the embodiment of this application uses high-voltage battery connecting wires as the detection object to perform fault detection, and the high-voltage battery connecting wires are respectively connected to the battery pack and the motor, or the battery pack and the electric drive system. After the connection is completed, the detection data is collected, including the current and voltage of the motor, the current and voltage of the battery, and the current and voltage at both ends of the high-voltage battery connecting wire.

[0024] Among them, the current and voltage data of the motor are collected by an ammeter and a voltmeter respectively; the current and voltage data of the battery are collected by installing an ammeter and a voltmeter at the circuit input end of the battery respectively; the current and voltage at both ends of the high-voltage battery connection line are collected by using an ammeter and a voltmeter installed at both ends respectively.

[0025] Furthermore, taking time T as the collection cycle, each type of detection data collected above is uniformly collected 60 times in one collection cycle. In this embodiment, the value of T is 1 minute. All types of detection data collected in one collection cycle are subjected to Z-score normalization respectively, and the normalized results of each detection data are arranged according to the collection time to obtain the data sequence of each detection data in each collection cycle. Among them, Z-score normalization is a commonly used technology in the field of data preprocessing, and the specific process is not repeated here.

[0026] Step S002, determining the fluctuation time difference weight of each acquisition period based on the consistency of the change in the strength of the autocorrelation between the data sequences within the acquisition period; determining the difference trend factor of each type of detection data based on the overall difference trend between different types of detection data within the acquisition period, and determining the connection influence coefficient of each type of detection data in combination with the influence of the connection device on the data anomaly.

[0027] During the battery charging and discharging process, the battery will be connected to the power grid using charging equipment such as a charging gun to charge the battery, and the charged battery will provide power for the electric motor. In this process, due to the harmonic oscillation of the power grid or the change in the driving state of the new energy vehicle, the irregular fluctuations in the data sequence of the detection data will increase, making it difficult to distinguish between normal changes and abnormal fluctuations caused by faults in the high-voltage battery connection line, such as insulation failure and electrolyte corrosion.

[0028] In this application, the fluctuation characteristics of the detection data between the high-voltage battery connection wire, the battery, and the motor are analyzed, and the order of occurrence of the fluctuation phenomenon of the detection data between the high-voltage battery connection wire, the battery, and the motor is compared. The reason for this is that if the irregular data fluctuation is caused by factors such as the power grid and the acquisition instrument, the time when the fluctuation phenomenon occurs in the corresponding data sequence under the same parameters of the high-voltage battery connection wire, the battery, and the motor is relatively close, so the changes in the strength of the autocorrelation between the data sequences are relatively consistent.

[0029] Specifically, the data sequences corresponding to the motor current, the battery current, and the current at both ends of the high-voltage battery connection line are delayed by After the a-th data sequence is lagged, the autocorrelation coefficient between the lagged data sequence and the data sequence before the lag is calculated. The autocorrelation coefficient of the bit calculation is recorded as , The value range is Among them, the autocorrelation coefficient is a commonly used technology in the field of data processing, and the specific process will not be repeated here.

[0030] It should be noted that the setting of the value range of the hysteresis bit number is affected by the length of the data sequence. The implementer can set it according to the specific collection frequency and period of the detection data, and this application does not impose any special restrictions on this.

[0031] Based on the above analysis, the fluctuation time difference weight is constructed to reflect the influence of fluctuation phenomenon on the change of autocorrelation strength between the data sequences corresponding to multiple currents. The calculation formula of the fluctuation time difference weight of each acquisition cycle is as follows: ; In the formula, T is the fluctuation time difference weight of each acquisition cycle, n is the number of data sequences involved in the calculation, is n series with lag The distribution variance of the autocorrelation coefficient calculated after the lag is 20. m is the upper limit of the number of lag digits of the data sequence. m is set to 20. , The ath data series is lagged +1 bit, hysteresis The calculated autocorrelation coefficient.

[0032] in, It characterizes the consistency of data changes when different data series are affected and fluctuate under the same lag number. The larger the value, the greater the difference in fluctuation characteristics of the data sequences corresponding to the motor current, the battery current, and the current at both ends of the high-voltage battery connection line. It reflects the stability of a single data sequence as the lag time increases. If the high-voltage battery connection line fails, the longer the fault occurs, the greater the impact on the detection data, and the lower the autocorrelation. will be less than That is, the larger the value of T is, the worse the consistency of the fluctuation characteristics of the motor current, the battery current, and the current data at both ends of the high-voltage battery connecting wire in the time dimension is, and the more likely it is that the data fluctuation is caused by a fault in the high-voltage battery connecting wire, and the greater the impact on subsequent fault detection results.

[0033] Furthermore, converting the time series data into frequency domain data can identify the fault characteristic frequencies existing in the data sequences of different detection data, making it easier to detect and diagnose possible faults. In addition, frequency domain analysis plays an important role in filtering noise and other interference, revealing the dynamic response characteristics of the system, extracting key features, and supporting pattern recognition. By comparing the frequency domain signal differences between the data sequences of different detection data, it is possible to more comprehensively analyze the changes in the use status of the high-voltage battery connection cable when it is connected to different devices.

[0034] Specifically, for any acquisition cycle, the data sequence of each detection data is converted from the time domain to the frequency domain using the fast Fourier transform FFT to obtain the amplitude and phase information of each frequency component, and the sequence composed of the amplitudes of all frequency components obtained after converting each data sequence to the frequency domain is used as the amplitude sequence of each data sequence. Among them, the fast Fourier transform is a commonly used technology in the field of data processing, and the specific process will not be repeated.

[0035] Furthermore, for different types of detection data in each acquisition cycle, the element values ​​of the same order in the amplitude sequence of the i-th detection data and the amplitude sequence of the b-th detection data are subjected to difference processing to obtain an amplitude difference sequence , using the amplitude difference sequence The maximum peak value in , minimum peak Determining the amplitude threshold In the embodiments of this application The purpose of setting the amplitude threshold is that when there are large differences in the frequency components of the two detection data in the frequency domain, there will be multiple peak points in the amplitude difference sequence, and the peak points greater than the amplitude threshold are screened to eliminate the interference caused by the use of peak features to analyze the data characteristics caused by the high-voltage battery connection line fault when the peak point value is small.

[0036] Secondly, filter the amplitude difference sequence The medium amplitude is greater than the amplitude threshold The peak points of the selected peak points are used as the amplitude and frequency of all the peak points as the dependent variable and the independent variable for linear fitting. Corresponding fitting straight line. Among them, commonly used methods of linear fitting include but are not limited to polynomial fitting, least squares fitting, and regularized linear regression fitting. Preferably, the least squares fitting method is used in the embodiment of the present application to obtain the fitting straight line. Linear fitting is a common technology in the field of data processing, and the specific process will not be repeated.

[0037] Here, a difference trend factor is constructed to characterize the overall difference trend between different types of test data in the same acquisition cycle. The calculation formula for the difference trend factor of the i-th test data is as follows: ; In the formula, is the difference trend factor of the i-th test data, N is the number of types of test data collected, is the amplitude difference sequence The maximum peak value in is the smoothing coefficient, which controls the growth rate of the arctan() function. In this embodiment, the value is , arctan() function is the inverse tangent function, yes The slope of the corresponding fitted line.

[0038] in, It reflects the difference trend between the i-th detection data and the b-th detection data in some frequency components. If the value of is greater than 0, it means that the difference between the i-th test data and the b-th test data has an increasing trend, and if it is less than 0, it means that the difference between the i-th test data and the b-th test data has a decreasing trend; It reflects the most significant difference between the data sequences of the two detection data in all frequency components, and can further amplify the degree of difference between the data sequences of the two detection data caused by data fluctuation.

[0039] On the other hand, the fault detection of the high-voltage battery connection line is carried out under the premise that the devices are connected at both ends of the high-voltage battery connection line, so the data fluctuation in the detection data will also be affected by the connected devices. For example, when the battery power supply is unstable or the current intensity increases, the rotor and stator wires inside the motor will vibrate when subjected to electromagnetic induction force, especially when the current intensity increases, the vibration will be enhanced, and mechanical vibration will occur. This vibration can be transmitted to the sensor or measuring instrument through the high-voltage battery connection line or structure, generating noise data that cannot be removed, affecting the fault detection results of the high-voltage battery connection line.

[0040] Furthermore, the amplitude of the data sequence of each detection data on each frequency component is squared to determine the energy value on each frequency component, and the energy spectrum density diagram of the data sequence in the frequency domain is plotted with the frequency as the horizontal coordinate and the energy value as the vertical coordinate. The energy spectrum density diagram is a commonly used technology in the field of frequency domain data processing, and the specific process is not repeated here. Based on the energy distribution in the energy spectrum density diagram, combined with the fluctuation time difference weight of each acquisition cycle and the difference trend factor of each detection data, the probability that the data fluctuation in each detection data in each acquisition cycle is affected by the connected device is analyzed.

[0041] Here, the connection influence coefficient is constructed to characterize the probability that the collected test data is affected by the vibration of the equipment connected to the high-voltage power supply wiring. The calculation formula for the connection influence coefficient of each test data in each collection cycle is as follows: ; ; In the formula, is the fault anomaly weight of the i-th detection data in each acquisition cycle, T is the fluctuation time difference weight of each acquisition cycle, is the difference trend factor of the i-th test data, It is the difference between the maximum and minimum energy spectrum density of the data sequence of the i-th detection data in each acquisition cycle in the energy spectrum density diagram of the frequency domain, is a preset constant used to prevent the denominator from being 0. The empirical value of is 0.001; is the connection influence coefficient of the i-th detection data in each acquisition cycle, is the number of elements in the data sequence of the i-th detection data in each acquisition cycle, is the absolute value of the difference between the jth element and the j+1th element in the data sequence.

[0042] in, It reflects the degree of data fluctuation caused by abnormal phenomena such as vibration of the equipment connected to the high-voltage power supply wire in each acquisition cycle. The more serious the vibration of the connected equipment during the detection process, the more serious the energy abnormality caused. The larger the value of ; the larger the value of T, the worse the consistency of the fluctuation characteristics of the detection data between the high-voltage battery connection line and the connected motor and battery in the time dimension, and the more likely it is that the data fluctuation is caused by a fault in the high-voltage battery connection line; The larger the value of , the greater the overall difference trend between the i-th test data and the rest of the test data; that is, The smaller the value of is, the greater the probability that the data fluctuation of the i-th detection data within the collection period is affected by the connection device, and the higher the impact on the high-voltage battery connection line fault detection; It reflects the data change trend of the i-th detection data at adjacent moments in the collection cycle. The smaller the value is, the more stable the data of the i-th detection data is within a collection cycle, and data fluctuations with different change trends will not occur frequently.

[0043] Step S003, optimizing the clustering algorithm based on the connection influence coefficient of each detection data in each collection cycle, determining the abnormal data in each collection cycle, and determining the fault detection result.

[0044] Furthermore, the fault detection of the high-voltage battery connecting wire requires detecting abnormal data from a large amount of detection data, so as to determine whether there is a fault in the high-voltage battery connecting wire.

[0045] Specifically, according to the above steps, the connection influence coefficient of each detection data in each acquisition cycle is determined respectively, and the clustering algorithm is optimized based on the connection influence coefficient to distinguish data fluctuations caused by faults and data fluctuations caused by factors such as vibration of connected equipment, thereby improving the accuracy of fault detection.

[0046] Based on the above analysis, this application considers using data clustering to divide the detection data into different clusters to achieve abnormal data detection. The CURE (Clustering Using Representative) clustering algorithm uses a set of points to define a cluster, rather than using only one center point like most clustering algorithms. This method can effectively process clusters and eliminate outliers, thereby reducing the impact of outliers on clustering results. The CURE algorithm can also recognize clusters of various shapes, including non-convex shapes and ring clusters, which makes it more adaptable when processing complex data structures and can obtain more accurate abnormal data detection results when facing fault detection.

[0047] Specifically, the data sequence of each detection data in each acquisition cycle is taken as input, and the data sequence is divided into multiple clusters using the CURE clustering algorithm. The number of representative points in the algorithm is set to B. During the clustering process, the convergence factor in the algorithm is adaptively determined based on the connection influence coefficient of each detection data in each acquisition cycle. , to ensure that the contraction of representative points can be adjusted more flexibly when facing different types of detection data in different acquisition cycles, and the size of B takes an empirical value of 5. Among them, the CURE clustering algorithm is a commonly used technology in the field of data processing, and the specific process is not repeated here. The formula for determining the convergence factor when clustering the data sequence of the i-th detection data in one acquisition cycle is: ; In the formula, is the shrinkage factor for clustering the data sequence of the i-th detection data within a collection cycle, is the sigmoid function, It is the connection influence coefficient of the i-th detection data in the acquisition period.

[0048] in, It reflects the probability of abnormal data in the i-th detection data due to interference noise or vibration of the connected equipment during a collection cycle. The larger the value of , the larger the shrinkage factor after adaptive adjustment, making the clustering process more tolerant to smaller clusters or outliers. The smaller the value of , the smaller the value of the shrinkage factor after adaptive adjustment, making the clusters obtained in the clustering process more compact and the recognition accuracy of outliers higher. Through this optimization method, the identified outliers are more likely to cause abnormal data caused by high-voltage battery connection line failure.

[0049] Furthermore, for each data point in the cluster, the Euclidean distance between it and the representative point of the cluster to which it belongs is calculated, and the data point whose Euclidean distance is greater than the preset outlier distance is regarded as an abnormal data point. Preferably, the preset outlier distance in the embodiment of the present application is 7, which can be selected by the implementer according to the actual situation, and the present application does not make any special limitation on this.

[0050] During the high-voltage battery connection line detection process, data of H collection cycles are collected. In this embodiment, the value of H is 24. The number of collection cycles in which abnormal data points appear in all types of detection data is counted as J. The threshold is set In this embodiment The value is 0.1. When the high voltage battery connection cable is faulty.

[0051] It should be noted that the above sequence of the embodiments of the present application is for description only and does not represent the advantages and disadvantages of the embodiments. The above is a description of a specific embodiment of this specification. In addition, the processes depicted in the accompanying 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.

[0052] The various embodiments in this specification are described in a progressive manner, and the same or similar parts between the various embodiments can be referenced to each other, and each embodiment focuses on the differences from other embodiments.

[0053] The embodiments described above are only used to illustrate the technical solutions of the present application, rather than to limit them. Modifications to the technical solutions recorded in the aforementioned embodiments, or equivalent replacement of some of the technical features therein, do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present application, and should all be included in the protection scope of the present application.

Claims

1. A fault detection method for electric wires for new energy vehicles, characterized in that: The method comprises the following steps: Obtaining a data sequence of each detection data in each acquisition cycle; Determine the fluctuation time difference weight of each collection period based on the change in the strength of the autocorrelation between the data series when the data series of the detection data between the high-voltage battery connection wire, the battery, and the motor in each collection period fluctuates; In each acquisition cycle, each detection data is converted into time-frequency domain, an amplitude threshold is set based on the amplitude of the two detection data at all frequency components, and a fitting straight line of the peak point screened by the amplitude threshold is obtained by data fitting; based on the fitting straight line, a difference trend factor of each detection data is determined; Based on the energy spectrum density diagram obtained after the data sequence of each detection data is converted into the frequency domain, the fault anomaly weight of each detection data in each acquisition cycle is determined in combination with the fluctuation time difference weight of each acquisition cycle and the difference trend factor of each detection data; Based on the fault anomaly weight of each detection data in each acquisition cycle, the connection influence coefficient of each detection data in each acquisition cycle is determined in combination with the change amount of the elements in the data sequence of each detection data; A data clustering algorithm is used to obtain abnormal data points in a data sequence of each detection data based on a connection influence coefficient of each detection data, and a fault detection result is determined based on the frequency of the abnormal data points in multiple acquisition cycles.

2. The fault detection method for electric wires for new energy vehicles according to claim 1, characterized in that: The fluctuation time difference weight of each acquisition cycle is determined as follows: For each acquisition cycle, the data sequences corresponding to the motor current, the battery current, and the current at both ends of the high-voltage battery connection line are lagged respectively. After the lag, the autocorrelation coefficient between the data sequence obtained after the lag and the data sequence before the lag is calculated; where, is the number of hysteresis bits of the data sequence corresponding to the current; Calculate the data series corresponding to each current at lag Bit and Hysteresis +1 bit corresponds to the difference between the autocorrelation coefficients; Determine the data series corresponding to all currents at lag The product of the distribution variance of the obtained autocorrelation coefficient and the difference is The cumulative result within the value range of ; The accumulated mean value of the accumulated results on the data sequences corresponding to all currents is used as the fluctuation time difference weight of each acquisition cycle.

3. The fault detection method for electric wires for new energy vehicles according to claim 1, characterized in that: The step of setting the amplitude threshold based on the amplitudes of the two detection data at all frequency components comprises: Use fast Fourier transform FFT to convert the data sequence of each detection data from the time domain to the frequency domain data, obtain the amplitude and phase information of each frequency component, and convert each data sequence into the frequency domain to obtain the sequence composed of the amplitudes of all frequency components as the amplitude sequence of each data sequence; The element values ​​of the same order in the amplitude sequences of the two detection data are subjected to difference processing to obtain an amplitude difference sequence, and the amplitude threshold is determined using the maximum peak value and the minimum peak value in the amplitude difference sequence.

4. The fault detection method for electric wires for new energy vehicles according to claim 3, characterized in that: The method of determining the amplitude threshold by using the maximum peak value and the minimum peak value in the amplitude difference sequence includes: The ratio of the difference between the maximum peak value and the minimum peak value in the amplitude difference sequence and 3 is used as the amplitude threshold of the amplitude difference sequence.

5. The fault detection method for electric wires for new energy vehicles according to claim 1, characterized in that: The method of obtaining a fitting straight line of a peak point screened by an amplitude threshold by means of data fitting includes: The peak points whose amplitudes are greater than the amplitude threshold in the amplitude difference sequence are screened, and the amplitudes and frequencies of all the screened peak points are used as dependent variables and independent variables for linear fitting, respectively, to obtain the fitting straight line corresponding to the amplitude difference sequence.

6. The fault detection method for electric wires for new energy vehicles according to claim 1, characterized in that: The determining of the difference trend factor of each detection data based on the fitting straight line includes: Each type of detection data is used as target data, and the slope of the fitting straight line corresponding to the amplitude difference sequence between the target data and any type of detection data is used as the input of the inverse tangent function, and the sum of the product of the preset smoothing coefficient and the output of the inverse tangent function and 1 is calculated; The product of the sum and the maximum peak value in the amplitude difference sequence between the target data and any one of the detection data is accumulated over all the detection data as the difference trend factor of each detection data.

7. The fault detection method for electric wires for new energy vehicles according to claim 1, characterized in that: The method for determining the fault anomaly weight of each detection data in each acquisition cycle is as follows: Perform square calculation on the amplitude of each frequency component of the data sequence of each detection data to determine the energy value of each frequency component, and draw an energy spectrum density diagram of the data sequence in the frequency domain with the frequency as the abscissa and the energy value as the ordinate; The sum of the difference between the maximum value and the minimum value of the energy spectrum density of the data sequence of each detection data in each acquisition cycle in the energy spectrum density diagram in the frequency domain and a preset constant is used as the denominator; The ratio of the product of the fluctuation time difference weight of each acquisition cycle and the difference trend factor of each detection data in each acquisition cycle to the denominator is used as the fault anomaly weight of each detection data in each acquisition cycle.

8. The fault detection method for electric wires for new energy vehicles according to claim 1, characterized in that: The connection influence coefficient of each detection data in each acquisition cycle is determined as follows: The product of the accumulation result of the absolute value of the difference between adjacent elements in the data sequence of each detection data in each acquisition cycle and the fault anomaly weight of each detection data in each acquisition cycle is taken as the connection influence coefficient of each detection data in each acquisition cycle.

9. The fault detection method for electric wires for new energy vehicles according to claim 1, characterized in that: The method of obtaining abnormal data points in the data sequence of each detection data by using a data clustering algorithm based on the connection influence coefficient of each detection data includes: The data sequence of each detection data in each acquisition cycle is used as input, and the connection influence coefficient of each detection data in each acquisition cycle is input into the sigmoid function to obtain the function result as the shrinkage factor for clustering; Using the CURE clustering algorithm to divide the data sequence of each detection data in each acquisition cycle into a plurality of clusters based on the shrinkage factor; For each data point in each cluster, the Euclidean distance between the data point and the representative point of the cluster to which it belongs is calculated, and the data point whose Euclidean distance is greater than the preset outlier distance is regarded as an abnormal data point.

10. The fault detection method for electric wires for new energy vehicles according to claim 1, characterized in that: The determining of the fault detection result based on the frequency of the abnormal data point in a plurality of acquisition cycles comprises: The ratio of the number of collection cycles in which abnormal data points appear in all types of detection data to the total number of collection cycles is counted. When the ratio is greater than a preset threshold, it is considered that there is a fault in the high-voltage battery connection line.

Images